FIFTH CANADIAN CONFERENCE ON NONDESTRUCTIVE ... - IAEA
FIFTH CANADIAN CONFERENCE ON NONDESTRUCTIVE ... - IAEA
FIFTH CANADIAN CONFERENCE ON NONDESTRUCTIVE ... - IAEA
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
q 11<br />
Atomic Energy<br />
of Canada Limited<br />
L'Énergie Atomique<br />
du Canada, Limitée<br />
AECL-8707<br />
<strong>FIFTH</strong> <strong>CANADIAN</strong> <strong>C<strong>ON</strong>FERENCE</strong> <strong>ON</strong><br />
N<strong>ON</strong>DESTRUCTIVE TESTING<br />
CINQUIÈME C<strong>ON</strong>FÉRENCE CANADIENNE<br />
SUR DES ESSAIS N<strong>ON</strong> DESTRUCTIFS<br />
i October / Octobre 28-31,1984<br />
TOR<strong>ON</strong>TO, <strong>ON</strong>TARIO<br />
"NDT in Canada - The Next 20 Years'<br />
20th Anniversary Celebration<br />
Les essais non destructifs au Canada<br />
Les 20 années prochaines"<br />
20ieme Anniversaire<br />
f-dito' Éditeur .<br />
C.A Kittmer<br />
Chalk River Nuclear Laboratories<br />
Laboratoires nucléaires de Chalk River<br />
Chalk River
ATOMIC ENERGY OF CANADA LIMITED AECL-8707<br />
<strong>FIFTH</strong> <strong>CANADIAN</strong> <strong>C<strong>ON</strong>FERENCE</strong> <strong>ON</strong><br />
N<strong>ON</strong>DESTRUCTIVE TESTING<br />
CINQUIÈME C<strong>ON</strong>FÉRENCE CANADIENNE<br />
SUR DES ESSAIS N<strong>ON</strong> DESTRUCTIFS<br />
October 1984 Octobre<br />
TOR<strong>ON</strong>TO, <strong>ON</strong>TARIO<br />
PROCEEDINGS /<br />
NOTES DE LA C<strong>ON</strong>FÉRENCE<br />
Editor / Editeur<br />
C.A. KITTMER<br />
Atomic Energy of Canada Limited<br />
Chalk River Nuclear Laboratories<br />
L'Énergie Atomique du Canada, Limitée<br />
Laboratoires nucléaires de Chalk River<br />
Chalk River<br />
January 1985 Janvier
ATOMIC ENERGY OF CANADA LIMITED<br />
<strong>FIFTH</strong> <strong>CANADIAN</strong> <strong>C<strong>ON</strong>FERENCE</strong> <strong>ON</strong> N<strong>ON</strong>DESTRUCTIVE TESTING<br />
EDITOR<br />
C.A. KITTMER<br />
ABSTRACT<br />
The theme for the Fifth Canadian Conference on Nondestructive Testing was "NDT in Canada - The<br />
Next 20 years". The three day conference with 42 presentations provided a short overview of NDT<br />
in Canada, a look at NDT in pipeline, materials, offshore, nuclear and training applications, and a<br />
glimpse into the next 20 years with recent advances in research and development as related to this<br />
"hi-tech" field of work.<br />
The three keynote speakers, Andrew B. Mitchell (Canada), Harold Berger (U.S.A.) and Roy S. Sharpe<br />
(England), were united in their concern for increased dialogue between research and field<br />
application. This conference was one small step in the right direction.<br />
Chalk River Nuclear Laboratories<br />
Chalk River, Ontario KOJ 1J0<br />
1985 January<br />
AECL-8707
L'ENERGIE ATOMIQUE DU CANADA LIMITÉE<br />
CINQUIÈME C<strong>ON</strong>FÉRENCE CANADIENNE DES ESSAIS N<strong>ON</strong> DESTRUCTIFS<br />
ÉDITEUR<br />
C.A. KITTMER<br />
RÉSUMÉ<br />
La cinquième conférence Canadienne des essais non destructifs (END) avait pour thème: "END au<br />
Canada - Les 20 années prochaines". La conférence, d'une durée de trois jours, a permise une vue<br />
d'ensemble rapide des END au Canada grâce à 42 présentations. L'utilisation des END dans<br />
plusieurs domaines; construction de pipelines, examens de matériaux, exploration au large des<br />
côtes, domaine nucléaire et d'entraînement de même qu'un aperçu des progrès récents en<br />
recherche et développement de cette haute-technologie furent examinés.<br />
Les trois orateurs invités; Andrew B. Mitchell (Canada), Harold Berger (U.S.A.) et Roy S. Sharpe<br />
(Royaume-Uni) furent unanimes dans leurs demandes pour un dialogue accru entre les chercheurs<br />
et les usagers. Cette conférence représente un premier pas dans cette direction.<br />
Laboratoires Nucléaires de Chalk River<br />
Chalk River, Ontario KOJ 1J0<br />
Janvier 1985<br />
AECL-8707
CINQUIEME <strong>C<strong>ON</strong>FERENCE</strong> CANADIENNE SUR DES ESSAIS N<strong>ON</strong> DESTRUCTIFS<br />
PRÉFACE<br />
II a été décidé de tenir la cinquième conférence Canadienne sur des essais non destructifs pour<br />
deux raisons principales. La première raison est qu'il s'est écoulé huit ans depuis la dernière<br />
conférence Canadiennne. De grands progrès ont été accomplis à la technologie Canadienne depuis<br />
la dernière conférence mais peu de gens sont au courant. Il y a certes des problèmes de<br />
communication dans l'industrie Canadienne puisque l'industrie Canadienne est éparpillée sur une<br />
étendue de 150 km de large par 7500 km de long. Les investissements principaux quitte le pays<br />
étant donné que les Canadiens n'ont pas connaissance que la technologie des essais non<br />
destructifs est disponible au Canada. Conséquemment, un des objectifs de la conférence était de<br />
promouvoir la communication et de rassembler les gens de l'est à l'ouest et également d'établir les<br />
progrès établis à l'intérieur du Canada d'un océan à l'autre.<br />
Une seconde raison pour tenir la conférence était qu'elle coincide avec le 20ième anniversaire de la<br />
Société Canadienne des Essais Non Destructifs et corresponds à 20 ans de progrès et de succès<br />
dans le domain des essais non destructifs. Depuis la fondation de la société Canadienne en 1964<br />
(qui était initiallement appelée le Centre Canadien pour la Technologie Non Destructif jusqu'en<br />
1970), la croissance de la société a été tout à fait remarquable atteignant 1500 membres et même<br />
plus. Durant cette époque, la société a été active en publiant son propre journal (maintenant<br />
intitulé le journal CSNDT), en présentant des cours et des séminaires et en participant à de<br />
multiples conférences et expositions. Cependant, c'était la première fois que la société décidait de<br />
promouvoir une activité majeure par elle-même.<br />
Le succès de la conférence indique le travail acharné et la persévérance du committé d'organisation.<br />
Président:<br />
John H. Zirnhelt<br />
Président technique:<br />
Charles A. Kittmer<br />
Président de l'exposition:<br />
Marvin Horenfeldt<br />
Présidents des sessions techniques:<br />
Jean F. Bussiere<br />
Norman G. Harding<br />
Mirek Macecek<br />
Lynda B. Manzer<br />
Secrétaire de la conférence:<br />
Yvonne A. Carroll<br />
Bureau Canadien de Soudre<br />
L'Énergie Atomique du Canada, Limitée<br />
Du Pont Canada Limited<br />
Conseil National de Recherches Canada<br />
la Société Canadienne des Essais Non<br />
Destructifs<br />
Techno-Scientific Inc.<br />
la Fondation de la Société Canadienne<br />
des Essais Non Destructifs<br />
Conference and Seminar Management<br />
La présence des notes de la conférence est un hommage aux personnes impliquées dans les essais<br />
non destructifs au Canada. Nous tenons encore à exprimer notre gratitude aux auteurs et<br />
également à ceux qui ont présenté les publications pour leur compréhension et leur collaboration<br />
tout en respectant les horaires. Nous espérons qu'ils maintenant reconnaissent que l'effort était<br />
valable et que la conférence a apporté une contribution à la technologie Canadienne des essais<br />
non destructifs pour maintenant et pour le futur.<br />
La technologie Canadienne, un chef de têtel Tel était un fait reconnu tout au long de la<br />
conférence. Voici maintenant les notes de la conférence... à vous d'en juger.<br />
Charles Kittmer
<strong>FIFTH</strong> <strong>CANADIAN</strong> <strong>C<strong>ON</strong>FERENCE</strong> <strong>ON</strong> N<strong>ON</strong>DESTRUCTIVE TESTING<br />
FOREWORD<br />
It was decided to hold the Fifth Canadian Conference on Nondestructive Testing (NDT) for two<br />
main seasons. First, it had been eight years since the last Canadian conference. A great deal had<br />
happened to Canadian NDT technology in that time, but few knew about it. With Canadian industry<br />
located in a strip of land 150 km wide by 7500 km long, communications is a problem. Major<br />
funding is leaving this country because Canadians are not aware of what NDT technology is<br />
available in Canada. Therefore, one of the objectives of the Conference was to provide a forum for<br />
communication.... to bring east and west together and give people an opportunity to find out what<br />
is happening in Canada from coast to coast.<br />
A second reason for holding the Conference was to mark the 20th Anniversary of the Canadian<br />
Society for Nondestructive Testing (CSNDT). What more fitting way to celebrate twenty years of<br />
achievement in NDT? Since formal establishment in 1964 (it was initially called the Canadian<br />
Council for Nondestructive Technology until 1970) the society has grown remarkably to a<br />
membership of over 1500. In that time, it has been an active society publishing its own newsletter<br />
Inow the CSNDT Journal), sponsoring education courses and seminars, and participating in<br />
numerous conferences and exhibitions. However, this was the first time for the society to sponsor<br />
a major activity on its own.<br />
The success of the Conference reflects the hard work and perseverance of the working committee:<br />
Chairman:<br />
John H. Zirnhelt Canadian Welding Bureau<br />
Technical Chairman:<br />
Charles A. Kittmer Atomic Energy of Canada Limited<br />
Exhibitor Chairman:<br />
Marvin Horenfeldt Du Pont Canada Limited<br />
Session Chairmen:<br />
Jean F. Bussiere National Research Council of Canada<br />
Norman G. Harding Executive Director, CSNDT<br />
Mirek Macecek Techno-Scientific Inc.<br />
Lynda B. Manzer CSNDT Foundation<br />
Conference Secretariat:<br />
Yvonne A. Carroll Conference and Seminar Management<br />
The presence of these proceedings reflects, and is a tribute to the hard work of the many people<br />
involved in NDT in Canada. We again express our appreciation to our keynote speakers and<br />
authors who responded with understanding and cooperation as they were exposed to our very<br />
tight time schedule. We hope in retrospect they feel the effort worthwhile, and that the Conference<br />
has made a contribution to awareness of NDT in Canada both now and for the future.<br />
Canadian technology, second to none! This was a stated theme throughout the Conference. Here<br />
are the proceedings... you be the judge.<br />
Charles Kittmer
iii<br />
ACKNOWLEDGEMENT<br />
A special note of thanks to Donna McCarthy of the Nondestructive Testing Development Branch, Chalk<br />
River Nuclear Laboratories, for her long hours spent in bringing these proceedings to print.
FOREWORD<br />
TABLE OF C<strong>ON</strong>TENTS<br />
DAY 1<br />
KEYNOTE ADDRESS: Five Years Experience with Eddy Current Testing of NBEPC 1<br />
Generating Stations<br />
- A.B. Mitchell, H.M. O'Connor, R. Pickles<br />
Canadian Forces Non-Destructive Program Origins 11<br />
- J.S. Cavers<br />
Reducing Unwanted Effects -- Recent Developments in Eddy Current Equipment 16<br />
t-eatures<br />
- A.G. Julier<br />
Nondestructive Examination Plays Key Role in Reactor Rehabilitation Program 27<br />
- CA. Wallis<br />
Evaluation of Ultrasonic Methods for the Detection and Sizing of Real Defects in the 40<br />
Base of Rails<br />
- L. Piche, J.F. Bussiere, J.P. Monchalin<br />
Optical Detection of Ultrasound at Distance 51<br />
- J.P. Monchalin<br />
Automated Ultrasonic Testing System for the Characterization of Defects in Weldments 66<br />
- M. Macecek, K. Luscott, J. Wells, D.K. Mak<br />
The Influence of Stress on the Inspection of Steel with Particular Reference to Gas 79<br />
Pipelines<br />
- D.L. Atherton, D.C. Jiles, C. Welbourn<br />
Defect Characterization and Sizing in Pipeline Weldments by Analysis of Ultrasonic 97<br />
Echo Features<br />
- J.P. Monchalin, J.F. Bussiere, R.W.Y. Chan, D. Sharp, D.R. Hay<br />
Capabilities/Limitations of Focused Ultrasonic Beams for Sizing Pipeline Girth 104<br />
Weld Defects<br />
- CA. Kittmer, J.B. Halle«, V.N. Sycko<br />
Flaw Characterization Using the Time-of-Flight Method and Ultrasonic Frequency 115<br />
Analysis<br />
- D.K. Mak<br />
The Introduction of Real-Time Radiography for the Inspection of Butt Welds in 130<br />
Offshore Pipelines<br />
- F.E. Reynolds Jr. (presented by L. Rathwell)<br />
Continuous Acoustic Emission Monitoring 145<br />
- J.S. Mitchell, M.N. Bassim
DAY 2<br />
KEYNOTE ADDRESS: Advanced Radiography for Transportation and Energy Systems 152<br />
- H. Berger<br />
NDT - Radiographie Processing Quality Control, Making the Transition to 163<br />
Automatic Processing in Radiographie NDT, Sensitometric Values for Industrial<br />
Radiography<br />
- W.E.J. McKinney<br />
Recent Microfocus X-Ray Imaging Applications (abstract only) 176<br />
- R.S. Peugeot<br />
Wet Channel Inspection Sysiems for CANDU Nuclear Reactors....CIGAR and 178<br />
CIGARette<br />
- M.D.C. Moles, M.P. Dolbey, K.S. Mahil<br />
An Advanced Heat Exchanger Eddy Current Inspection System 194<br />
- M. DeVerno, H. Ghent, H. Licht<br />
Wet Channel Measurement of Pressure Tube to Calandria Tube Spacing in CANDU 207<br />
Reactors<br />
- J.H. Sedo<br />
Checking for Cracks in Gas Turbine Rotor Discs 222<br />
- J. van den Andel, A.B. Nieberg<br />
Eddy Current Inspection of Mildly Ferromagnetic Tubing 235<br />
- W.R. Mayo, J.R. Carter<br />
On the Relation Between Ultrasonic Attenuation and Fracture Toughness in Type 250<br />
403 Stainless Steel<br />
- F. Nadeau, J.F. Bussiere, G. Van Drunen<br />
Acoustic Emission Testing of Man-Lift Devices 262<br />
- J.A. Baron<br />
Acoustic Sorting of Grinder Balls 276<br />
- F. Nadeau, J.F. Bussiere<br />
The Benefits of NDT Training for Canadians 289<br />
- L.B. Manzer<br />
Fear Detection and Removal - The Psychological Implications of the Technological 297<br />
Age (abstract only)<br />
- T. Helliwell<br />
Certification of NDT Operators in Canada - An Update 298<br />
- V. Caron
DAY 3<br />
KEYNOTE ADDRESS: Looking into the Future 303<br />
- R.S. Sharpe<br />
NDT of Structural Ceramics by High Frequency Ultrasonics 307<br />
- A. Fahr<br />
Hot Pressed Piezo-Electnc Ceramic Elements for Ultrasonic Transducers 316<br />
- N.D. Patel, J. van den Andel, P.S. Nicholson<br />
Ultrasonic Analysis of Voids in Glass: Theory and Practice 331<br />
- A..I. Stockman, J. van den Ande!, P.S. Nicholson<br />
Computer Simulation of Ultrasonic Testing 345<br />
- D.B. Duncan<br />
A Novel Approach to Eddy Current Imaging of Defects 356<br />
- D. Leemsns, M. Macecek<br />
Developments in X-Ray Stress Measurement, The CANMET Portable Stress 372<br />
Diffractometer<br />
- R.A. Holt<br />
Applications of Neutron Diffraction to Engineering Problems 387<br />
- T.M. Holden, G. Dolling, S.R. MacEwan, J. Winegar, B.M. Powell, R.A. Holt<br />
NDE in Polymers, an Example: Ultrasonics for the Determination of Density in 398<br />
Polyethylene<br />
- L Piche, A. Hamel<br />
An Industrial Application of Computer Assisted Tomography: Detection, Location 408<br />
and Sizing of Shrink Cavities in Valve Castings<br />
- P.D. Tonner, G. Tosello<br />
Computer Operated Composite Panel Testing (abstract only) 424<br />
- S. DeWalle<br />
Some Unconventional Techniques for the Inspection of Layered Materials 426<br />
- P. Cielo<br />
Materials Effects on Acoustic Emission During Deformation and Fracture 445<br />
- M.N. Bassim<br />
List of Delegates 454
DAY1<br />
KEYNOTE ADDRESS: Five Years Experience with Eddy Current Testing of NBEPC 1<br />
Generating Stations<br />
- A.B. Mitchell, H.M. O'Connor, R. Pickles<br />
Canadian Forces Non-Destructive Program Origins 11<br />
- J.S. Cavers<br />
Reducing Unwanted Effects -- Recent Developments in Eddy Current Equipment 16<br />
Features<br />
- A.G. Julier<br />
Nondestructive Examination Plays Key Role in Reactor Rehabilitation Program 27<br />
- CA. Wallis<br />
Evaluation of Ultrasonic Methods for the Detection and Sizing of Real Defects in the 40<br />
Base of Rails<br />
- L. Piche, J.F. Bussiere, J.P. Monchalin<br />
Optical Detection of Ultrasound at Distance 51<br />
- J.P. Monchalin<br />
Automated Ultrasonic Testing System for the Characterization of Defects in Weldments 66<br />
- M. Macecek, K. Luscott, J. Wells, D.K. Mak<br />
The Influence of Stress on the Inspection of Steel with Particular Reference to Gas 79<br />
Pipelines<br />
- D.L. Atherton, D.C. Jiles, C. Welbourn<br />
Defect Characterization and Sizing in Pipeline Weldments by Analysis of Ultrasonic 97<br />
Echo Features<br />
- J.P. Monchalin, J.F. Bussiere, R.W.Y. Chan, D. Sharp, D.R. Hay<br />
Capabilities/Limitations of Focused Ultrasonic Beams for Sizing Pipeline Girth 104<br />
Weld Defects<br />
- CA. Kittmer, J.B. Halle«, V.N. Sycko<br />
Flaw Characterization Using the Time-of-Flight Method and Ultrasonic Frequency 115<br />
Analysis<br />
- D.K. Mak<br />
The Introduction of Real-Time Radiography for the Inspection of Butt Welds in 130<br />
Offshore Pipelines<br />
- F.E. Reynolds Jr. (presented by L. Rathwell)<br />
Continuous Acoustic Emission Monitoring 145<br />
- J.S. Mitchell, M.N. Bassim
FIVE YEARS EXPERIENCE WITH EDDY CURRENT TESTING AT NBEPC GENERATING<br />
STATI<strong>ON</strong>S<br />
A.ß. Mi tali ell<br />
Wen 1 Bïunnvick Raea-tcli and P r o d u c t i v i t y Council, Fxede x i c t o n , N.B.<br />
H.M. 0'Co 11 not<br />
{ büxme tiij with New Brunswick Research and Productivity Council)<br />
R. Pickles<br />
" a . G \ itii M.'(' d: f (' e c t '; • c Powe.fi C o m m i s s i o n . F x e d e ï i c t o n . N . B .<br />
SUMMARY<br />
Five years ago NBEPC established an eddy current test program at its<br />
generating stations in which all condensers were to be periodically inspected to<br />
determine future maintenance requirements. This program has proved to be<br />
extremely successful.<br />
By reference to specific situations encountered, the present paper<br />
describes how the aims of the program, its logistics, the techniques used and<br />
the assimilation of the results have gradually been upgraded to provide a<br />
powerful tool to assist in preventive maintenance.<br />
Faced with the operation of several older plants having a high incidence of<br />
condenser tube failure, New Brunswick Electric Power Commission (NBEPC) took a<br />
major step forward when five years ago it established an eddy current test<br />
program at fossil fuelled generating stations. All condensers were to be<br />
periodically inspected to determine future maintenance requirements. This<br />
program proved to be extremely successful in improving condenser reliability<br />
ana since that time has expanded to include numerous auxiliary heat exchangers<br />
and also heat exchangers at hydraulic and nuclear generating stations. The<br />
New Brunswick Research and Productivity Council (RPC) has worked with NBEPC<br />
throughout the program, providing technical advice and carrying out all the<br />
inspections.<br />
The three co-authors of this paper (and the program) represent what should be<br />
an iaeal balance of interests in any preventive maintenance program. There was<br />
an operations engineer whose interests are plant efficiency, preventive<br />
maintenance ana cost effectiveness, an NDE specialist who developed many of the<br />
special inspection techniques used and a metallurgist interested in the<br />
performance of materials and the implications of the inspection results on tube<br />
life expectancy. However, even with all this combined expertise, it has still<br />
taken several years to overcome all the problems inherent with such a program<br />
and only in the last year or so have we been able to consider most inspections<br />
as "routine". Experience has taught us that each large heat exchanger must be<br />
considered on an individual basis and will probably have its own unique problem<br />
requiring a "tailor-made" inspection program.<br />
In the short time available in this presentation, we would like to concentrate<br />
on some of the lessons learned and problems encountered in establishing the<br />
program. However, before we get to that, we should say something about the<br />
program in general.
MB Power has a total generating capacity of about 2500 MW of which<br />
approximately b0% is fossil fuel, 35« hydro and 15% nuclear. The Eddy Current<br />
inspection program has been concentrated on the larger fossil and nuclear units<br />
in which aluminum brass, cupro-nickel and titanium tubes are used in a wide<br />
range of different condenser sizes and designs. Station ages range from<br />
2 years to 20 years and several condensers have been partially retubed in the<br />
last five years as a result of the program. Station details and corresponding<br />
condenser tube materials are shown in Table 1.<br />
Because of the type of corrosion detected, it has been found beneficial to<br />
inspect most Al-brass and Cu-Ni tubed condensers on an annual basis. Results<br />
obtained from such inspections have been used in:<br />
- establishing accurate corrosion rates; which in turn have been used tc<br />
monitor chemistry control effectiveness and to calculate tube life<br />
expectancy:<br />
- defining susceptible areas within the condenser; such as, air removal zone,<br />
regions affected by flow channelling, regions effected by flow induced<br />
vibration, poor performance due to variable tube quality, etc., and:<br />
- guiding preventive maintenance by identifying individual tubes for plugging<br />
and in defining the extent of zones for retubing.<br />
Titanium tubed condensers have received an initial inspection and then, since<br />
corrosion damage is minimal, have only been reinspected to investigate specific<br />
technical problems occurring during service. As described later, tube damage<br />
due to vibration required a major inspection program. At another station<br />
damage due to weld spatter which occurred during construction was investigated<br />
when discovered at a later stage.<br />
Although several techniques for mechanizing inspection have been tried, RPC<br />
have always reverted to manual inspection methods as being the most versatile<br />
for condenser inspection. An experienced team of an inspector and helper can<br />
achieve an inspection rate of 100-150 tubes per shift, the slower rate of<br />
inspection being required for tubes exhibiting numerous significant<br />
indications. An immediate diagnosis of each tube is made at the time of<br />
inspection and all signals are recorded for later review. Tubes with defects<br />
deeper than 25?; of wall thickness are noted for future reference and those with<br />
deeper than 7b% recommended for immediate plugging.<br />
The first lesson learned in the program was how to be cost effective.<br />
At the outset of the program it was thought that condenser tube leakage could<br />
be brought down to extremely low frequencies by inspecting very large numbers<br />
of tubes. This illusion was soon dispelled by the realization that a<br />
significant number of leaks developed quickly by random events (impact damage,<br />
erosion at mussel shells, etc.) for which inspection provided no forewarning.<br />
It also became clear after the first few inspections that a large sample<br />
approaching a complete inspection was uneconomical both in cost and outage<br />
time. Experience has now shown that a sample of 300 to 600 tubes (3 to 7'= of<br />
the condenser) if judiciously chosen, is both adequate for monitoring generic<br />
degredation and of a short enough duration to be acceptable to plant<br />
operations. With this sample size, the program objective became clarified as<br />
monitoriny the overall condition of the condenser rather than preventing<br />
individual tube leakage.
- 3 -<br />
The second lesson learned was the value of "credibility" and how this is only<br />
established over a period of time.<br />
The commonest form of corrosion found in copper based condenser tubes is<br />
condensate grooving (Figure 1). This type of corrosion is worst in air<br />
removal regions and is attributed to ammonia and oxygen dissolved in condensate<br />
running down the baffle plates. It forms a narrow band of wastage around the<br />
tube which steadily progresses with years of service. Experience in station<br />
operation worldwide suggests a "typical" condenser tube life is 15 years and<br />
our own experience with condensate grooving fits this projection. NBEPC has<br />
found condensate grooving to be troublesome in most aluminum brass tubed<br />
condensers.<br />
Evidence of severe condensate grooving was found during an inspection made<br />
after only one year's operation at at one oil-fired generating station<br />
(Figure 2). If remedial action were to be taken to save the tubes it had to be<br />
immediate since grooving in some tubes had already apparently exceeded 25% of<br />
wall thickness. However, this was an unexpected problem for a brand new<br />
station, and a problem which had been identified by a new team still developing<br />
the required technique. No remedial action could be taken until these findings<br />
were verified and our task over the next three years became that of carefully<br />
monitoring the rate of progress of the grooving. This proved possible by<br />
comparing the phase rotation of the defect signal with calibration grooves of<br />
similar width machined in a piece of new tubing. The accuracy of this cross<br />
correlation was verified by the removal of typical corroded tubes from the<br />
condenser. After three years, an accurate corrosion rate had been established<br />
that left no doubt that the air removal zone of the condenser would need<br />
complete retubing after five years service. It also identified the need for<br />
mechanical changes to improve venting of the condenser. This knowledge allowed<br />
ample time for retubing preparations and also the selection of a more resistsnt<br />
material (Cu-Ni) for the air removal zone. Inspection was also able to define<br />
the exact extent of the zone requiring retubing.<br />
The third lesson learned was that an inspector must be knowledgeable enough to<br />
handle the unexpected.<br />
The inspector has an ongoing responsibility to correctly diagnose all signals<br />
picked up by his °ddy current instrument. Unexpected signals which cannot be<br />
related to previous interpretations are always worrying and can significantly<br />
reduce the rate of inspection by dominating the inspector's attention. Many<br />
such signals are generated by various magnetic effects and these have often<br />
proved troublesome to interpret. Weld spatter, magnetic inclusions, material<br />
of variable permeability and the effect of high stress can all produce<br />
anomolous signals needing very careful analysis. Our ability to recognize such<br />
signals has slowly improved with familiarity but we are still sometimes misled.<br />
Another annoying magnetic effect has been encountered with cupro-nickel tubes<br />
which show variable permeability. This causes signal drift and slows down<br />
inspection because of the need for frequent balance point adjustments.<br />
Metallurgical studies have demonstrated that the permeability of a piece of<br />
cupro-nickel tube varies widely depending on its residual cold work and heat<br />
treatment.
The fourth lesson learned was that there is a practical limit to an inspector's<br />
ability to interpret indications and that there should be no "loss of face"<br />
involved in requesting a tube to be removed to diagnose an unusual indication.<br />
As a demonstration that an inspector can be tempted to extrapolate theoretical<br />
experience too fer, we give the following example:<br />
At another plant in the examination of a conaenser tube which had been in<br />
service for many years, our inspector attributed a particular type of signal to<br />
a "magnetic" corrosion film which he felt must be on the inner surface of the<br />
tube (Figure 3). He was even able to demonstrate how this signal could be<br />
generated by magnetic tape wrapped around the inside of the tube (Figure 4).<br />
the signal was of completely the wrong phase angle to be interpreted as either<br />
an internal or an external defect. Metallographie examination of a tube pulled<br />
from the condenser a year later, failed to reveal this "magnetic film". It<br />
did, however reveal numerous corrosion pits on the internal tube surface<br />
(Figure 5). It was then demonstrated that the combined effect of closely<br />
spaced pits could rotate the angle of the resultant signal to such a degree<br />
that it would be unrecognizable as a partial through-wall defect. Five years<br />
ago when this incident occurred, the effect of multiple defects was not<br />
reported in the literature. Some research is now being carried out to<br />
elucidate these effects.<br />
The fifth and final lesson learned was that certain problems were beyond the<br />
capability of conventional eddy current techniques and that research laboratory<br />
back-up was essential.<br />
As a last illustration, I would like to talk about the detection of<br />
circumferential cracks in titanium condenser tubes at Pt. Lepreau G.S. All<br />
eddy current inspectors know that conventional bobbin probes used for tube<br />
inspections are very insensitive to circumferential cracks. When it was<br />
discovered that flow induced vibration had caused fatigue failure of some outer<br />
condenser tubes at Pt. Lepreau, we were faced with the problem of identifying<br />
within a few hours, a sufficiently sensitive method to detect other cracked<br />
tubes before they became major leakers. The solution was a new technique<br />
called "3D"*.<br />
This novel technique is so called because of its ability to add another<br />
dimension to eddy current testing — in addition to defect depth and volume,<br />
this equipment can also be used to determine its anqular position in the tube<br />
Wäll.<br />
A "3D" eddy current instrument is similar to a conventional eddy current<br />
instrument except that the 3D generates a three phase sine wave source. The<br />
special 3D probe is tx three coil probe connected in a three phase star<br />
configuration.<br />
* Developed and manufactured by Eddy Current Technology.
_ c _<br />
If the 3 phase exitation is a constant voltage source of equal magnitude in<br />
each coil and the test object is symmetrical with respect to all 3 coils, then<br />
the offset voltage will be zero. However, any non-symmetrical defect, change<br />
in dimension, electrical or magnetic property will cause an impedence in<br />
balance and a defect indication.<br />
In practice this means that symmetrical features such as; support plates, tube<br />
fins, rolled joints, etc., will be invisible to "3D", whilst a non-symmetrical<br />
defect such as; an axial crack, corrosion or uneven fretting will give strong<br />
indications. In the present instance, the equipment was proved sensitive to an<br />
incomplete circumferential crack.<br />
In Figure 6 we can see an example of a circumferential crack in the titanium<br />
tubing. To the left we can see a stainless steel stake which in actual<br />
practice was installed exactly over the cracks. In Figure 7 we have compared<br />
the signals of conventional differential eddy current with "3D" for a support<br />
plate and a circumferential crack. The advantages of 3D for the latter are<br />
clearly apparent. The stainless steel "staking" proved to be an added<br />
complication since this was also unsymmetrical to the probe. Fortunately,<br />
however, the application of a very high frequency could be used to almost<br />
eliminate the staking signal whilst still retaining sufficient sensitivity in<br />
the 3D probe for the radial cracks.<br />
In this inspection program, several cracked tubes were found by "3D" in a water<br />
box where sensitive chemical detection equipment was unable to detect leakage.<br />
To conclude, we would like to summarize our experiences with regard to<br />
condenser inspection as follows:<br />
(a) Retubing a condenser costs 50 to 500 K$. We consider the cost of a<br />
typical inspection small, for the benefits accrued.<br />
(b) Each condenser inspection program must be "tailor-made" to its own<br />
particular operating problems.<br />
(c) Whereas we do not recommend annual inspections throughout the life of the<br />
condenser, we do recommend a couple of inspections during early in-service<br />
life to identify corrosion degradation at a stage where counter-measures<br />
are possible and worthwhile.<br />
(d) A first inspection should be considered both as baseline and an<br />
opportunity to identify any special techniques that need to be developed<br />
to measure a particular deterioration.<br />
(e) We are convinced that a well planned condenser inspection program will<br />
provide a great amount of information on material performance, degradation<br />
rate, susceptible zones and remaining tube life which will be of great<br />
benefit to plant operations.
STATI<strong>ON</strong> UNIT<br />
Coleson Cove<br />
Coleson Cove<br />
Coleson Cove<br />
Courtenay Bay<br />
Dalhousie<br />
Grand Lake<br />
Pt. Lepreau<br />
1<br />
2<br />
3<br />
3<br />
4<br />
1<br />
2<br />
8<br />
1<br />
TABLE 1 - DETAILS OF NEW BRUNSWICK ELECTRIC POWER COMMISSI<strong>ON</strong><br />
STATI<strong>ON</strong>S AND C<strong>ON</strong>DENSERS<br />
CAPACITY<br />
(MW) TYPE C<strong>ON</strong>DENSER MATERIALS<br />
202<br />
195<br />
202<br />
100<br />
100<br />
108<br />
180<br />
60<br />
600<br />
Oil<br />
Oil<br />
Oil<br />
Oil<br />
Oil<br />
Oil<br />
Oil/Coal<br />
Coal<br />
Nuclear<br />
Aluminum Brass/Partly<br />
Retubed Cu-Ni<br />
Aluminum Brass/Partly<br />
Retubed Cu-Ni<br />
Aluminum Brass/Partly<br />
Retubed Cu-Ni<br />
Aluminum Brass<br />
Aluminum Brass<br />
Aluminum Brass<br />
Titanium<br />
Aluminum Brass/Partly<br />
Retubed Cu-Ni<br />
Titanium<br />
NUMBER OF<br />
TUBES/C<strong>ON</strong>DENSER<br />
6,800<br />
6,800<br />
6,800<br />
6,600<br />
6,600<br />
8,200<br />
8,950<br />
6,900<br />
12,000<br />
NUMBERTOF<br />
C<strong>ON</strong>DENSERS
- 7 -<br />
Figure 1. Condensate grooving in aluminum brass condenser tubing.<br />
80<br />
70<br />
60<br />
50<br />
40<br />
30<br />
20<br />
10<br />
1<br />
1<br />
1<br />
1 y<br />
Is'<br />
*'<br />
1<br />
/I<br />
y<br />
0 2 4 6 8 10 12 14 16<br />
YEWS OF SERVICE<br />
Figure 2. Progression of condensate grooving in an oil fired<br />
generating station.<br />
y<br />
y
0*'?<br />
- 8 -<br />
Ä |<br />
'••• '~-!.if~ z '--Y-<br />
'••
- 9 -<br />
Figure 5. Corrosion pitting on inside surface of aluminum brass<br />
condenser tube.<br />
Figure 6. Cracked titanium tube. Note "staking" bar in practice<br />
covers crack.
- 10 -<br />
7/8 inch Titanium Tube<br />
3D<br />
i '<br />
t ; ,<br />
-r SuDOort Plate 1<br />
Figure 7. Comparison of signals obtained with conventional and "3D" techniques.
- 11 -<br />
<strong>CANADIAN</strong> FORCES N<strong>ON</strong>-DESTRUCTIVE PROGRAM ORIGINS<br />
J.S. Ca.ve.fLA<br />
Ae.tioipa.ee. Maintenance Development Unit<br />
Canadian ¥oicei> Base T.'iznton, k&tn.a, On.tan.io<br />
ABSTRACT<br />
This paper presents an overview of how non-destructive testing in the<br />
Canadian Forces has evolved since its inception in 1959, with review of<br />
present training programs, development projects, condition monitoring,<br />
and a few related problems.<br />
<strong>CANADIAN</strong> FORCES N<strong>ON</strong>-DESTRUCTIVE PROGRAM ORIGINS<br />
Traditional non-destructive testing in the Canadian Forces began in<br />
1959 with two Senior Non-Comissioned Officers being sent on course to the<br />
Department of Energy, Mines and Resources in Ottawa. On-job training for<br />
the next year and a half was carried out by an individual known to many of<br />
us; Mr. Bill Havercroft. In 1962 these two Senior Non-Commissioned Officers<br />
were transferred to what was then called the No. 6 Repair Depot (now the<br />
Aerospace Maintenace Development Unit) to establish an Non-Destructive<br />
Testing Shop.<br />
Between 1962 and 1964 approximately ten personnel had been absorbed<br />
into the NDT and were trained at various locations throughout North America<br />
and the United Kingdom. These ten personnel provided NDT services<br />
to all RCAF Bases in Canada, flying Bases, that is,as the CF NDT program was<br />
then, and still is today, an Air Element organization. Services to the land<br />
and sea elements were provided on a "when required" basis. These ten personnel<br />
personnel were constantly on the road carrying out inspections, in<br />
fact so much so that expansion became a must. As a result, the initial buildup<br />
began in 1965 with the first formal course being held at AMDU Trenton.<br />
Plans were now being formulated to estalish facilities at most aircraft<br />
operation locations. The facilities would not only be responsible for<br />
their own flying base, but for their geographical area as well-<br />
By 1967 there were approximately 40 trained technicians at Trenton,<br />
with support work to other Bases still dominating their workload. From then<br />
to present day, there has been a drastic expansion with facilities being<br />
established at 11 locations across Canada and Europe.
The facility at CFB Greenwood was closed in 1982, due mainly to the<br />
lark of N'DT inspectioi-.a required on the new Aurora aircraft, which replaced<br />
'he Ar.i;us. At that sant? tine, considerable support work was being carried<br />
• ut .'.t CF:i Sumraerside by CFB Greenwood personnel. With an increasing ND1<br />
requirement anticipated in the future, a division was made to establish a<br />
i.ev facility at CFB Summeraide.<br />
The total number of personnel now employed in NDT is approximately<br />
'•••ht with the largest number (40) being employed at AMDU Trenton. Table 1<br />
• : i splays total Service strength versus the number of personnel employed in<br />
."'I ". from 1959 to tha present. Our N'on-Dcstructive Testing Branch is currently<br />
'-: -ffpa with one Senior Engineering Officer with a post-graduate degree in<br />
r.,:,-destructive testing, and 3 Junior Officers, one of whom also has a postradu.i'-.e<br />
degree in nc --destructive testir.g. In addition, 27 Junior and 9<br />
Seniur NI "'• technicians with varying qualifications and experience arc- on<br />
s t n i f .<br />
1967 Edmonton - 6<br />
1968 Baden - 10<br />
Ottawa - 5<br />
1970 Moose Jaw - 5<br />
SKi.rXTl<strong>ON</strong> PROCESS<br />
TABLE 1: NDT Service Strength<br />
1971 Chatham - 4<br />
1973 Cold Lake - 12<br />
1974 Shearwater - 6<br />
1975 Bagotville - 8<br />
1982 Summerside - 6<br />
Personnel selected for NDT training nu.st have at least<br />
live ve.ir' in their b.isic aircraft trade, and tl.ey must meet specific<br />
. iitii-.it. i ••ii requirener I;;. They also must have at l'.'ast seven years remaining<br />
i :, the Service. Volunteers are subjected to a fairly extensive interview by<br />
ilie Ideal NDT facilitv Supervisor, who will then forward ):iü rocomme dations<br />
to tiic Career M'.napel at National Defence Headquarter iii Ottawa, who is<br />
re ,jM.iTi;> ihle for NDT personnel selection.<br />
It sf.i.-il.i bo pointed out thnt NDT in the Canadian i-j;-ces is a Branch<br />
•il the basic aircraft trades. As such, many individuals will be re-employed<br />
in their original trade at a future date. Normally personnel will be employed<br />
in *:IJT for 7-10 years, allowing the Canadian Forces to utilize a fully-<br />
'i lined technician for a period of 2 - 5 years.<br />
Since J9'>2 approximately 250 aircraft technicians have been<br />
M)T-t r.i Lned, with, a larp(; number of those who have retired irom the Service,<br />
, rfi'Mtlv b' 1 ing enplcr.-ed In industry in the NDT field.
AMVU N<strong>ON</strong>-DESTRUCTIVE TH.STING BRANCH<br />
- Li -<br />
Two of the responsibilities ot the Non-Destructive Testing Branch are<br />
a. to monitor development in industrial NDT, which includes<br />
evaluating equipment, and new or improved inspection<br />
b. develop and approve NDT inspection technique for Canadian<br />
Forces aircraft.<br />
As an example of how a technique may be developed, one of our NDT<br />
facilities, in co-operation with the local Aircraft Maintenance<br />
and Engineering Branch, identified the need to develop an inspection<br />
technique for the forward rotor shaft of the CH113/113A<br />
helicopter. The situation was addressed to Air Command who, in<br />
turn, recommended NDHQ that an NDT technique be developed for<br />
carrying out this inspection. NDHQ then created a project to<br />
carry out this tasking and forwarded it to AMDU for action. One<br />
of the areas which required inspection was the threaded portion<br />
of the rotor shaft at the carrier bearing retaining nut, which is<br />
approximately 27 inches from the top of the shaft opening.<br />
AMDU received the tasking on a Friday afternoon, and by the<br />
following Tuesday, two technicians had designed and manufactured<br />
a test apparatus and were carrying out successful inspections.<br />
It goes without saying that this is an<br />
exception rather than the norm for the amount of time personnel<br />
would require, or be given, to develop a viable inspection. However,<br />
it is typical of the type of work that is carried out at<br />
AMDU. There are between 30 and 40 inspections developed each<br />
year utilizing all NDT disciplines.<br />
Other responsibilites of the Non-Destructive Testing Branch include:<br />
a. carry out all NDT training in the Canadian Forces, provide<br />
training material, and screen all examination applications<br />
from Area Facilities;<br />
b. provide technial assistance to all ten NDT Area Facilities.<br />
This may be an aid in the interpretation of inspection results<br />
to guidance on equipment repair, and the operation of new<br />
equipment;<br />
c. provide supplementary personnel support to Area Facilities when<br />
they are short staffed, thus ensuring priority inspection deadlines<br />
are met;<br />
d. provide NDT support on aircraft periodic and sampling inspections<br />
at AMDU Trenton; and<br />
e. evaluate condition monitoring programs for possible inclusion in<br />
routine aircraft maintenance.
C<strong>ON</strong>DITI<strong>ON</strong> M<strong>ON</strong>ITORING<br />
- 14 -<br />
The Condition Monitoring Program in the Canadian Forces is based on<br />
the following four analysis methods:<br />
1. Spectrometric Oil Analysis was first introduced to the air<br />
Element of the Canadian Forces in 1968, with the first<br />
laboratory being located at AMDU. This was instituted in<br />
response to a requirement to monitor the wear state of the<br />
J-79 engine of the CF 10A Starfighter. As additional<br />
components were monitored, a shorter response time was<br />
considered desirable, thus additional labs were opened at<br />
CFB Edmonton in 1971 and at CfB Baden (Germany) in 1972.<br />
As maintainers of various other aircraft realized the<br />
validity of the SOA program, labs were installed at<br />
Comox, Moose Jaw, Shearwater, Bagotville, Chatham and<br />
Cold Lake. These labs all operated with atomic absorption<br />
(AA) instruments. In the late 1970's the Canadian Forces<br />
began a transition to Atomic Emission (AE) spectrometers,<br />
and are presently in the final stages of conversion. This<br />
conversion from AA to AE spectrometers was deemed necessary<br />
for the following reasons:<br />
(1) to standardize guidelines with the U.S. Forces,<br />
since a large number of our oil wetted systems<br />
are the same as our American counterparts;<br />
(ii) operation of the AE spectrometers is simpler and<br />
faster; and<br />
(iii) accessories such as nitrous oxide, acetylene and<br />
keytone are eliminated.<br />
Approximately 35,000 samples per year are analyzed with a considable<br />
number providing enough information on various components to<br />
allow for early detection of problems which could have led to<br />
catastrophic failure and has proven to be a valuable maintenance<br />
tool.<br />
2. Flash Point Testing is utilized todetect fuel contamination of<br />
engine oils. It is carried out as a regular part of maintenance<br />
on aircraft engines equipped with fuel/oil heat exchangers.<br />
3. Ferrography was found to be an excellent complementary tool to<br />
the spectrometric oil analysis method. For this reason we use<br />
it to further analyze those oil samples which are found<br />
suspicious by spectrometry. In addition, we have started an<br />
evaluation on filter debris analysis-
- 15 -<br />
4. Vibration Analysis is also used extensively in the<br />
Canadian Forces, and has been adopted as a maintenance tool for<br />
Sea King helicopters on the east coast. There have been several<br />
successful maintenance actions resulting from this program, and<br />
in time, other aircraft will be included in the Vibration Analysis<br />
program.<br />
The Condition Monitoring Section is presently preparing a course for<br />
technicians involved in this field to meet the requirements which will ensue<br />
from the expansion of this program.<br />
Several long-range NDT-related projects are being initiated, amongst<br />
them is one to define an ultrasonic transducer performance, with the<br />
objective being acceptance/rejection specifications. Another project is on<br />
reliability of NDT inspections; the aim of which is to update our present<br />
data to allow us to more accurately determine inspection frequencies, and as<br />
a double check on our training and equipment.<br />
A Real Time Microfocus Radiography System has been purchased for our<br />
Branch, and will be installed shortly. It will be used for training,<br />
technique development and for the inspection of composite materials.<br />
C<strong>ON</strong>CLUSI<strong>ON</strong><br />
In conclusion, the Canadian Forces has been supporting the technology<br />
of non-destructive testing for the past 25 years. Our expertise has grown<br />
from a nucleus of two technicians to an organization of 106 personnel, and<br />
our Area Facilities have grown to ten. Our main purpose is to provide the<br />
Air Element with non-destructive testing support for both operational and<br />
maintenance environments. To this end, we have created our own training<br />
system, developed an in-depth technique development program, and just<br />
recently, a Condition Monitoring program. We also monitor new equipment<br />
entering the market, and liaise with other Serices in order to keep abreast<br />
of the latest developments in this ever advancing field.
- 16 -<br />
REDUCING UNWANTED EFFECTS ~ RECENT DEVELOPMENTS IN EDDY CURRENT<br />
EQUIPMENT FEATURES<br />
A Cati G. Jut let<br />
Hocking Et cet ion les Inc.<br />
Haio Iiiit'iumeufi Div-Lii.cn<br />
Temp te H(££.i, MP<br />
1. ABSTRACT<br />
It has long been established that the distribution of eddy currents is<br />
affected by many variables, some of which are desirable and some are not.<br />
This paper will seek to illustrate by reference to recent applications how<br />
eddy current equipment features have been developed to overcome some of these<br />
problem effects.<br />
2. PRESENTATI<strong>ON</strong><br />
It has long been established that the distribution of eddy currents is<br />
affected by many variables, some of which are desirable and some of which are<br />
not. Within the time available today, it would be impossible to try to<br />
enumerate all of the problems and all of the solutions. So I will<br />
concentrate on four particular applications—two involving probes and two<br />
involving instrumentation. I will illustrate how their designs have been<br />
developed in order to overcome the associated problem effects.<br />
A recent task was the requirement to detect a fatigue crack at the root of<br />
small titanium helicopter engine rotor blades (Figure 1). In this example,<br />
the distance of the crack from the fir tree root was unknown, but expected to<br />
be less than 2 mm and could occur anywhere between the leading and trailing<br />
blade edges.<br />
Using conventional high frequency pencil probes the unwanted effects in this<br />
application were:<br />
1. Zero changes due to the proximity of edges and corners<br />
2. Variable surface geometry<br />
3. Difficulty in identification of indications<br />
The design, manufacture and use of shielded probes has been described at<br />
conferences and in technical publications previously, and so I do not propose<br />
to dwell on that aspect. Nevertheless, it should be remembered that a<br />
shielded probe has a smaller magnetic field than a normal unshielded pencil<br />
probe, and that the field is further reduced by incrasing the frequency.<br />
For this application, we chose to use an instrument that had three switch<br />
selectable test frequencies of 500 kHz, 2 MHz and 6 MHz. Using this unit we<br />
were able to construct some curves which allowed us to examine the effect of<br />
shielding at each of the three frequencies.
- 17 -<br />
Figures 2 and 3 show the response to edge effect of the unshielded probes of<br />
each frequency compared with the same response from the shielded type.<br />
A sinusoidal scan pattern was described within the radius of the blade<br />
between the fin and fir tree root. The output from the eddy current<br />
instrument was connected to a simple strip chart recorder.<br />
Figure 4 shows the results obtained for shielded and unshielded 2 MHz probes,<br />
followed by those for the shielded and unshielded 6 MHz probes, Figure 5.<br />
It can be clearly seen that the shielded 6 MHz probes provide a distinct<br />
advantage to this application. The final step was to design it for in situ<br />
usage.<br />
My next example involves the application of multi-frequency eddy current<br />
equipment to the examination of nonferrous heat exchanger tubing such as that<br />
found in air conditioner, condensor and nuclear steam generator systems.<br />
It is nothing new to be using eddy current techniques to examine heat<br />
exchanger tube bundles. We have been doing it for the last 10 to 15 years.<br />
Throughout all this time we have been troubled by one major problem.<br />
Tubes can vary from a few feet in length in a simple air condition unit up to<br />
nearly 100 feet in a nuclear steam generator, but in general the tube is<br />
supported along its length by what are known as support or baffle plates.<br />
These are generally made of steel and, as such, generate a large eddy current<br />
signal when the probe passes underneath one.<br />
The eddy current method can discriminate between any defects that are on the<br />
inside, the outside and also through wall of the tube. However, when a<br />
defect exists, as they often do, under the support plate area, the eddy<br />
current signal produced is a combination of that from the defect with that<br />
from the support plate. If the defect is large, then the resulting modified<br />
signal would indicate to an operator that there was a defect within that<br />
area. However what he is not able to determine is whether the defect is on<br />
the inside, outside or through wall.<br />
Even worse, if the defect is very small, such as a pin hole, then the<br />
modification of the support plate signal may be so slight as to prevent its<br />
detection. Thus the opertor fails to detect it.<br />
By driving the probe coils at two frequencies simultaneously, we can<br />
electronically mix the two frequencies and use this technique to effectively<br />
eliminate the support plate signal.<br />
In effect, we choose two different frequencies and display the support plate<br />
signal for each frequency on the screen of the eddy current instrument. By<br />
using what are known as "mixing" contrôle, we attempt to make the size, shape<br />
and phase angle of each signal identical to each other.
- 18 -<br />
When this happens, the instrument subtracts the second frequency from the<br />
first and, of course, if they are identical, the resulting signal is a zero<br />
(Figure 6). So by making the support plate signal identical in both<br />
channels, we can effectively eliminate it.<br />
For those of you who are interested in the mathematics, they are given in<br />
Figure 7.<br />
Taking the example of a titanium tube with a mild steel support, we can show<br />
how this occurs.<br />
If a defect such as a 1 mm diameter hole is introduced into the tube, we can<br />
show the resultant signals before and after mixing, with and without the<br />
support plate (Figure 8).<br />
We can extend this application into the aerospace industry for the<br />
examination of cracks extending from bolt or fastener holes which have steel<br />
bolts or fasteners inserted (Figure 9), and again show the resultant signal<br />
before and after mixing (Figure 10).<br />
My next example still, keeps us in the heat exchanger tubing industry, but<br />
this time deals with a probe design.<br />
In the development of heat exchangr tubing, the major criterion is thermal<br />
performance. Improvements in the heat transfer characteristics can<br />
frequently be gained by increasing the surface area of the tube relative to<br />
its length and also by altering the flow characteristics of the fluid inside<br />
the tube. This has led to tubes which are thinned, roped or studded on both<br />
the inside and the outside as illustrated in Figure 11.<br />
A further complication has been the presence of skip-finned areas which occur<br />
at the support plate regions (Figure 12).<br />
Even with dual frequency techniques, saturation of the instrument occurs at<br />
fairly low gain levels, therefore the sensitivity to defects in a<br />
conventional test is poor. A standard differential type probe will also not<br />
satisfactorily detect the long splitting type of defect, commonly called<br />
"zipper" cracks found in these tubes.<br />
One attempted solution to the problem of this type of defect developed here<br />
.in the USA is the cross axis probe which has absolute type detection<br />
characteristics with a differential response.<br />
The latest design is a special dual differential probe in which signals from<br />
opposing sides of the tube are cancelled. This enables the cylindrically<br />
symmetrical signals to be balanced out and gives good sensitivity to metal<br />
loss type defects.<br />
A comparison of the relative sensitivities to a hole defect and defects in<br />
the skip-fin area for the differential probes is illustrated in Figure 13.<br />
Finally, my last example involves the problem of eddy current instrument<br />
calibration. There is a requirement by both user and manufacturer to have an
- 19 -<br />
absolute method of ensuring that eddy current instruments are able to respond<br />
in the correct manner. Normally we use test blocks having machined or<br />
electrical discharge machined (EDM) slots for this purpose. However, a<br />
number of unwanted effects can occur and these are:<br />
1. Variations in slot depth from test block to test block<br />
2. Variations in slot width from test block to test block, but also<br />
possible on the same block<br />
3. Variations in test block material conductivity<br />
4. Variations in probe impedance due to age, wear and probe type<br />
There are normally three positions of the probe which are critical to the<br />
calibration procedure:<br />
1. Probe on a test block<br />
2. Probe with a small lift-off<br />
3. Probe over the calibration slot<br />
Using absolute values of inductance and resistance, we have devised an<br />
electronic reference unit which simulates these probe actions and test block<br />
variables, but without the need for either.<br />
Consider a probe on an aluminum test block. It will have an impedance Zl<br />
which can be represented by values of inductance LI and resistance Rl. There<br />
is also a dampening effect which is a function of resistance and is really an<br />
energy loss.<br />
When the probe is lifted off the material slightly, there will be an<br />
impedance change. In this case, the inductance will rise, the resistance<br />
will rise and the energy loss, or dampening, will be reduced.<br />
Finally, when the probe is over a machined slot, the inductance and dampening<br />
will increase, but the resistance will decrease.<br />
The electronic reference unit connects to the eddy current instrument through<br />
the normal probe connection and allows for simulation of the setting up<br />
procedure as well as for different materials and different test frequencies.<br />
It can be set up to simulate the response from a certain size slot.<br />
At a certain sensitivity level, we should obtain a certain meter deflection.<br />
This provides an opportunity for users to check and calibrate their<br />
Instruments and be confident the Instruments are performing and responding<br />
correctly.<br />
After doing this, then probes can be connected in the normal way and<br />
calibrated against standard test blocks in order to achieve the necessary<br />
sensitivity for the particular task in hand.
- 20 -<br />
(CJUIMTTCJW<br />
ROTOR BLADE SHOWING<br />
nrncu. pomion or ocncT<br />
TYPICAL POSITI<strong>ON</strong> OF DEFECT<br />
AND SCAN PATTERN<br />
Figure 1: Rotor blade showing typical position of defect and scan<br />
pattern.<br />
Voter<br />
Rc.iding<br />
F SO<br />
00<br />
50<br />
\<br />
TOR UH tVITH STANDARD PENCIL PROBE <strong>ON</strong> STAINLESS STEEL<br />
2MHJ<br />
\<br />
\6MHi \<br />
\<br />
\<br />
\<br />
\<br />
\<br />
\<br />
\<br />
A<br />
\\500KHz<br />
\<br />
\<br />
\<br />
\<br />
V \<br />
\ \<br />
\ \<br />
\ \<br />
\ \<br />
1 2 3 4 8 8<br />
Distance of centre of probe from edge (mm)<br />
Figure 2: Edge effect of a standard probe on stainless steel at<br />
frequencies of 500 kHz, 2 MHz and 6 MHz.
Figure 3:<br />
Meier<br />
• -.1 ! i I<br />
.0<br />
00<br />
50<br />
OMHt<br />
FLAT<br />
/ .<br />
\ \<br />
• > V M M »<br />
\ \<br />
1 \<br />
- 21 -<br />
EDGE EFFECT CHARACTERISTICS - ZERO CHANl.l<br />
1 [X.ATOR UM WITH SHIM Of 0 PHOtit S <strong>ON</strong> S1AINI | S'. silli<br />
\<br />
\<br />
\<br />
t<br />
\900KHz<br />
\ROUND<br />
\ENO<br />
\<br />
\<br />
\<br />
\<br />
\ \<br />
\ \<br />
1<br />
N<br />
I s I 1 I<br />
1 1 3 4 5 •<br />
Distance of centre of probe from edge (mm)<br />
Edge effect of a shielded probe on stainless steel at<br />
frequencies of 500 kHz, 2 MHz and 6 MHz.<br />
Unihlaldad prob« - good oiada<br />
UniW«td*d pfoba - dafaellv* blade<br />
f Maftfad proba • good blada<br />
Shlaldad proba - dafacllva blada<br />
RESULTS WITH SCAN PATTERN FROM FIN TO ROOT - 6MH/ pmh.-.<br />
Figure 4: Results of inspection of good and defective blades with 6 MHz<br />
shielded and unshielded probes.
- 22 -<br />
Un»hl«ld«d prob* - good biad«<br />
8hl*ld*d prob* - good bisd«<br />
Shlvldsd prob« - d*t*cllva blad«<br />
RESULTS WITH SCAN PATTERN FROM FIN TO ROOT - ?MH; proh.-<br />
Figure 5: Results of inspection of good and defective blades with 2 MHz<br />
shielded and unshielded probes.<br />
TITAMUM TUBE WITH MILD STEEL SUPPORT<br />
CHANNEL 1<br />
15OKH1<br />
PHASEC D6 OUTPUT:<br />
MIX<br />
CANCELLATI<strong>ON</strong> OF SUPPORT SIGNAL<br />
Figure 6: Illustration of cancellation of support plate signal by using<br />
dual frequency "mixing" of signals.
- 23 -<br />
BASIC THEORY OF DUAL<br />
FREQUENCY INSPECTI<strong>ON</strong><br />
Ai fraquancy F1 d«f«ct signal D1, IM«rf«rlnf<br />
F2 • • D2 " • a<br />
it In» rasullanla al tha alngls fraquanclaa ara RI and R2<br />
R1 = D1 • 11 .<br />
R2 = D2 • 12 ...(2)<br />
NO. ,i MI2 = 11
STEEL FASTENERS<br />
EXAMINATI<strong>ON</strong> OF FASTENER HOLES WITH IN-SITU STEEL FASTENER<br />
Figure 9: Schematic of examination of fastener holes with in situ steel<br />
fastener.<br />
INSPECTI<strong>ON</strong> FOR CRACKS NEAR<br />
STEEL FASTENERS<br />
PHASEC 06 OUTPUT FOR NORMAL FASTENER<br />
l'HASEC D6 OUTPUT FOR FASTENER AND SLOT<br />
Figure 10: Illustration of the effects of dual frequency mixing to inspect<br />
for cracks in fastener holes near steel fasteners.
Figure 11: Examples of developments in heat exchanger tubing to increase<br />
surface area.<br />
Figure 12: Examples of skip-finned areas in heat exchanger tubing at<br />
support plate locations.
- 26 -<br />
SKIP FINNED TUHt<br />
•< IG 5 INSPECTI<strong>ON</strong> OF SKIP FINNED TUBE<br />
WITH DIFFERENT PROBE TYPES<br />
Figure 13: A comparison of inspection results for a skip-finned area using<br />
differential, cross axis and dual differential probes.
N<strong>ON</strong>DESTRUCTIVE EXAMINATI<strong>ON</strong> PLAYS KEY ROLE IN REACTOR REHABILITATI<strong>ON</strong><br />
PROGRAM<br />
C liante.-!, A. Watt is<br />
Ontario Hijdio<br />
Bruce Nuclear Poivcn Vnvc ïcpmeu t, Tiverton, Ontario<br />
ABSTRACT<br />
In October 1983 a program of major proportions was initiated to<br />
reposition spacers supporting reactor pressure tubes in several Candu<br />
Reactors. Nondestructive examination played a major role in the<br />
successful rehabilitation program. Novel radiographie and eddy current<br />
techniques were developed to locate and monitor the movement of spacers<br />
during the program.<br />
REACTOR REHABILITATI<strong>ON</strong> PROGRAM<br />
On August 1, 1983, a pressure tube failed in Unit 2 at Pickering Nuclear<br />
Generating Station. During the investigation of the failed tube it was<br />
discovered that the central spacers, more commonly referred to as garter<br />
springs, were out of their design location. It was felt at the time that<br />
this was a contributing factor to the failure. Although the failure has<br />
since been attributed to metallurgical causes, the location of garter<br />
springs remains a concern because of heat economy and the uncertain<br />
effects their position may have on pressure tube life.<br />
The spacers have two main functions. One is to maintain concentricity<br />
between the calandria and pressure tubes so that when the reactor is in<br />
operation the hot pressure tubes will not come in contact with the<br />
relatively cool calandria tubes. Contact would result in poor heat<br />
economy and also a high temperature gradient in the material at the point<br />
of contact. The springs also have the ability to roll and thus act as<br />
bearings between the two tubes preventing fretting and allowing for<br />
unequal thermal expansion (see Figure 1).<br />
In the Bruce reactors there are 480 zirc-niobium pressure tubes which are<br />
approximately 6.5 meters long and have a nominal wall thickness of 4.1<br />
mm. Each of the tubes is supported by four centrally located garter<br />
springs each being spaced a meter apart. The 130 mm. diameter springs<br />
are made from square sectioned zirc-niobium wire wound 7 1/2 coils per<br />
cm. with a cross section of 4 1/2 mm. Running through the center of the<br />
spring is a girdle wire which has its ends resistance welded to each<br />
other. This maintains the circular shape of the spring assembly (see<br />
Figure 2). Pressure tubes and springs are installed at site as part of<br />
the construction program. Once springs are installed there is no<br />
possible access to them by any other means than cutting into the pressure<br />
or calandria tubes.
To gain primary statistical information/ tests were initiated soon after<br />
the failure to determine spring locations in uncommissioned units at<br />
Bruce and Pickering. It was soon evident that numerous springs had moved<br />
from their design locations in a seemingly random manner.<br />
Tests were conducted in a pressure tube mock-up to simulate actual<br />
reactor conditions. It was found that the springs/ which are compressed<br />
and fitted in the calandria tubes in the vertical position, had relieved<br />
the spring compression by assuming a slanted orientation (see Figure 2).<br />
The spring would then "walk" away from the design location in small<br />
incremental steps each time the pressure tube moved in relation to the<br />
calandria tube.<br />
By experimentation it was determined that mechanical work around the<br />
reactor accounted for a large portion of spring movement. It was also<br />
discovered that devices such as soil compactors working up to 500 meters<br />
away could also cause motion. Inspections conducted before and after<br />
commissioning indicated that circulation of water through pressure tubes<br />
and around calandria tubes would also contribute a significant amount of<br />
movement.<br />
Investigations in uncommissioned reactors indicated that by the time a<br />
reactor was ready for fuelling, the cumulative out of tolerance positions<br />
of springs varied from 200 to 500 meters and that springs would be<br />
randomly spaced. In some cases the four springs were clumped together at<br />
one end of a pressure tube which provided no effective support for that<br />
tube.<br />
There were three difficult choices that could be made: leave as is and<br />
accept lower efficiency and the risk of shortened pressure tube life; cut<br />
and replace using new or shortened tubes with redesigned spacers; or<br />
reposition existing spacers and somehow immobilize them in their design<br />
location.<br />
After an in-depth study, Ontario Hydro decided to correct spring<br />
locations in all uncommissioned units before going critical.<br />
Repositioning appeared to be the most attractive of the remaining two<br />
options as replacement would be costly in terms of time and materials.<br />
The repositioning option boiled down to; "How is it possible to<br />
selectively move springs in a 6 meter annulus without being able to<br />
either touch or see them?"<br />
A number of methods of moving springs were tabled during the numerous<br />
brainstorming sessions that ensued. Interest spread and ideas flowed<br />
in. The policy was to fully review any suggestion no matter how<br />
unconventional it sounded. Psychokinesis, exploding hydrogen and filling<br />
the annulus with urea formaldehyde were just a few that left the option<br />
list quietly and quickly. After sifting, the short list of options<br />
centered around electrical, vibratory and simple mechanical levering<br />
methods. Development programs were started in all three areas.
- 29 -<br />
In the electrical field, the discharge of large scale capacitors through<br />
a coil looked promising. This method was vigorously pursued. Mechanical<br />
levering was the simplest to perform but would only work successfully on<br />
springs slanted in the direction of desired travel. This was, in real<br />
life, a rare occurance as the springs were normally tilted away from<br />
their starting point. A system using a large sonic vibrator was also<br />
experimented with and eventually proved successful.<br />
The capacitor discharge method was furthest along in development when it<br />
came time to start repositioning on the first unit. It was to the state<br />
where it would consistently move springs in the desired direction. The<br />
system was made up of an extremely large capacitor bank, low resistance<br />
cables and an insulated copper coil that was placed in the pressure tube<br />
adjacent to the garter spring. The capacitor bank discharged through the<br />
coil causing a high intensity magnetic field that propelled the spring<br />
away from the coil. The maximum potential of the bank is 200,000 amps.<br />
Typical settings used on site were in the range of 140-160 ka. Each<br />
discharge of the bank would move the spring an average of 5 mm.<br />
Moving the springs solved half the problem. The other half was finding<br />
where the springs wWere and monitoring their movement during the<br />
repositioning program.<br />
Information on spring location was needed on a large sample very early in<br />
the program. The only workable examination method with immediate<br />
availability was radiography. Fifteen film cassettes were attached end<br />
to end on eight long aluminum rods. The rods were then inserted into<br />
reactor pressure tubes in a square pattern surrounding an unoccupied<br />
tube. An iridium source was then advanced in half meter increments down<br />
the pressure tube located in the center of the square. When the source<br />
had traversed the full length of the tube it was retracted, and the films<br />
were removed for developing. It took approximately one full week working<br />
around the clock to perform the initial inspection on the first reactor.<br />
It was obvious that radiography was a too costly and a too time consuming<br />
method of looking for springs over the long term. Eddy current, real<br />
time radiography, acoustic emission, ultrasonics and infrared detection<br />
were all methods that had the capability to find springs, but of these<br />
eddy current stood out as the method most suited to our needs.<br />
At first glance it appeared to be a simple task. Using basic theory, it<br />
seemed quite likely that a large absolute pancake type probe working at a<br />
low frequency would have sufficient penetrating power and sensitivity to<br />
locate springs. However it turned out that the geometry of the springs<br />
was such that insufficient eddy currents were generated by this method to<br />
produce a suitable signal. The eddy current signal is in fact generated<br />
by the girdle wire. Springs by themselves, or springs with broken girdle<br />
wires produce relatively no signal at all.
- 30 -<br />
The next step in the development program was the designing and building<br />
of a large inside differential coil close to the inside diameter of the<br />
pressure tube. The first probe consisted of two, two hundred turn coils<br />
separated by 15 mm. This arrangement provided a strong signal from<br />
vertically oriented spring assemblies. However, springs that had flopped<br />
over and lay slanted produced almost no signal at all (see Figure 3). A<br />
probe was quickly produced that had the coils on a 20° slant. This set<br />
up worked fine on slanted springs providing the spring was tilted in the<br />
same direction as the coils. To pick up springs inclined in the opposite<br />
direction or slued somewhere in between meant rotating the probe until<br />
the maximum signal was achieved. This was far too time consuming a<br />
procedure when it is considered that 1720 springs per reactor had to be<br />
located and mapped.<br />
AECL's Chalk River Laboratories had also tackled the problem of designing<br />
a single probe that could locate springs irrespective of orientation.<br />
Their solution was a one 25 mm. wide exitation coil with 6 mm. receive<br />
coils on either side. This configurate gave consistently strong signals<br />
from all springs with the exception of those within the 150 mm. long<br />
belled portion at each end of a calandria tube. Unfortunately, many of<br />
the springs had migrated to the bells and these had to be located as<br />
well. As an interim measure to solve this problem, a shielded isotope<br />
camera holder was fabricated that allowed a camera to be directly<br />
attached to an end fitting. It was then possible to radiograph the<br />
adjacent bell without evacuating the area. Paper radiography with<br />
instant development speeded the process to the point where results were<br />
available in less than five minutes. Later, Chalk River was able to<br />
provide a probe with segmented coils that was capable of detecting<br />
springs in the belled ends.<br />
The next hurdle was that of accuracy. In order to have any effect on the<br />
garter spring, the capacitor discharge coxl had to be placed with an<br />
accuracy of + 1 mm. and this up to eight meters down a tube.<br />
The initial scans had been performed by a probe mounted on an automatic<br />
stem unit. The heart of the stem unit is two foil elements wrapped on<br />
drums that form a strong lightweight tube as they are unrolled. They<br />
extend in much the same manner as a tape measure. A wheel encoder riding<br />
on top of the elements provides a digital readout equivalent to the axial<br />
distance (see Figure 4).<br />
To perform an inspection, the stem unit is placed on an end fitting and<br />
by means of a controller the probe is driven to the far end of the tube.<br />
The probe is retracted and as it does so, the axial distance and the eddy<br />
current signals are recorded on a strip chart. The finished chart is<br />
about a meter long and is scaled in proportion to the tube length. The<br />
accuracy of the chart is approximately + 10 mm. The encoder that<br />
provides the axial scan distance is also subject to error due to skipping<br />
on the element, deformation of the rubber rim of the wheel and slight<br />
variations in the wheel diameter. With a fine tuned system, accuracies<br />
of + 3 mm. can be achieved. However, during general use, a tolerance of
- 31 -<br />
jf 25 mm. was achieved 85% of the time. So although the automatic system<br />
was fast (5 minutes per tube) and gave a permanent record of acceptable<br />
accuracy, it was not accurate enough to place the capacitor discharge<br />
coils.<br />
To achieve the 1 mm. accuracy required for setting coils, a rigid system<br />
with an attached measuring tape was developed. A 30 mm. diameter<br />
aluminum push rod 7 meters long with a 3 meter extension was used.<br />
Cables ran to the probe through the center of the rod, which was<br />
centrally supported in the calandria tube by nylon disks. The probe was<br />
mounted on a wheeled carriage which helped reduce the lift-off effect.<br />
The carriage assembly was attached to the push rod by means of a<br />
universal joint (see Figure 5).<br />
The probe assembly is manually inserted into a pressure tube to a<br />
selected spring while watching the eddy current instrument. As the probe<br />
passes a spring, a distorted Figure 8 appears on the cathode ray tube.<br />
The probe is then slowly moved back and forth over the spring until the<br />
signal dot is at dead center of the Figure 8. At this point the center<br />
of the probe would be aligned with the center of the mass of the spring.<br />
The tape measure is then consulted and the distance recorded. To place<br />
the capacitor discharge coil, the probe is backed off an amount<br />
equivalent to the predetermined distance from the center of the coil to<br />
the bumper at the end of the probe. The capacitor discharge coil is then<br />
slowly fed in from the opposite end of the tube until it comes in contact<br />
with the bumper. It is then accurately positioned for firing.<br />
It took approximately three months of around the clock work to reposition<br />
the springs in the first reactor. Nine complete sets of eddy current<br />
equipment, each manned by a two-man crew, were available at all times.<br />
Seven units were committed to production work at the reactor face. The<br />
remaining units acted as standbys and also served the ongoing development<br />
programs. As with any program of this magnitude, there were a multitude<br />
of problems.<br />
The supplier of the eddy current equipment claimed the instrumentation<br />
was capable of operating without breakdown during normal usage. We were<br />
also advised that portions of the internal circuitry were so fine that<br />
static electricity from a person's hand could cause destruction. Our<br />
needs and expectations were for equipment that would operate on a 100%<br />
duty cycle for months at a time within meters of equipment producing<br />
currents in the neighbourhood of 150,000 amps. The external<br />
electromagnetic fields are of such magnitude that as the capacitor banks<br />
fire the cathode ray tubes are completely cleared. Because of this<br />
numerous breakdowns were experienced at the beginning. The situation has<br />
now largely been overcome by the use of isolating transformers and minor<br />
modifications to the electronic circuits. The equipment now operates<br />
almost without fault.
Each firing of the capacitor banks causes localized heating in the tube<br />
adjacent to the coil. The rise in temperature changes the material<br />
resistivity to the point where meaningful eddy current readings are<br />
unobtainable. Natural cooling through conduction and convection takes up<br />
to thirty minutes. This is an intolerable delay so cooling manifolds<br />
were fabricated that provided a cooling air blast through several jets on<br />
the effected area. This, in conjuction with working on several adjacent<br />
tubes in parallel, reduced down time to an insignificant amount.<br />
Unfortunately there are always mysterious signals that must be explained<br />
away. Reactivity mechanisms run vertically through the reactor adjacent<br />
to the calandria tubes. These produce a signal easily misinterpreted by<br />
the inexperienced as they are the mirror image of a garter spring<br />
indication. A spring adjacent to a mechanism produces signals that<br />
cancel each other out and results in undetectable springs. For this<br />
situation and where it was expected that two springs were nestled<br />
together, radiographie equipment had to be always ready.<br />
A general sequence of events established for repositioning in any tube is<br />
as follows:<br />
(1) Perform eddy current inspection and locate all four springs.<br />
(2) Feed spring locations into a computer which would suggest<br />
various options with respect to which springs to move and by how<br />
much.<br />
(3) Select a spring to move, place capacitor discharge coil and fire.<br />
(4) Cool tube and eddy current inspect to determine motion.<br />
The above sequence or slight variations on the above would be<br />
repeated until the tube was accepted by the computer.<br />
One of the alternate systems for moving a spring involved flexing the<br />
pressure tube repeatedly with a hydraulic jack located in the center of a<br />
2 1/2 meter long rigid beam. An eddy current coil was built into the<br />
beam adjacent to the jack and it was possible to monitor spring movement<br />
in real time. It turned out to be self defeating as the coil diameter<br />
interfered with the flexing of the tube and only modest movement was<br />
achieved. The probe was eventually mounted on the end of the beam which<br />
allowed progress readings to be taken by partially withdrawing the jack<br />
assembly.<br />
Another system that was developed by Ontario Hydro research utilized a<br />
large sonic vibrator that generated high intensity sound over a wide<br />
range of frequencies. The vibrator was located near the end of the<br />
selected pressure tube. An expanding head was fed into the prassure tube<br />
and clamped near the spring to be moved. An aluminum tube coupled the<br />
vibrator to the head. Various combinations of frequency and intensity<br />
were selected until the real time eddy current system indicated spring
movement in a positive direction. This system depended on having the<br />
tube free to vibrate and for that reason an air floatation system was<br />
used to support a lightweight eddy current probe. The sonic vibration<br />
method worked well but because of large sound deadening insulation<br />
cabinets it took up great quantities of valuable space in a very<br />
congested area. This coupled with vibrator breakdowns and improvements<br />
to the other available systems has led to its present limited use.<br />
As the first reactor program got underway, we were prepared for defeat.<br />
Opinion varied as to whether the repositioning was a cost effective<br />
solution as compared to retubing. It was also impossible to guarantee<br />
the equipment would perform for sustained periods. It was quite possible<br />
that coils would fail and cause arcing or that the large discharge of<br />
current through the coils would cause deformation to the tubes. Knowing<br />
of all the possible problems, it was easy to be pessimistic. The very<br />
starting of this program took a large measure of faith.<br />
It was well known the program could only function with teamwork and the<br />
constant exchange of ideas and the positive attitude that we could always<br />
do better. The first few days on the reactor were disheartening as only<br />
minimal travel was achieved and some days negative secondary motion on<br />
adjacent springs would result in a net loss for the day. Slowly daily<br />
movement totals increased as techniques, materials and teamwork<br />
improved. During the early stages four meters of travel a day was<br />
considered acceptable progress, whereas twenty meters of travel a day was<br />
not unusual by the time the second reactor had been completed. The first<br />
reactor required 72,000 firings of the capacitor banks and as many eddy<br />
current inspections to bring the springs to acceptable positions.<br />
Experimentation during commissioning has shown that when a tube is loaded<br />
with fuel the resulting sag pinches the springs between the calandria and<br />
pressure tubes and thus in most cases immobilizes them. It has therefore<br />
been our practice to load each tube with fuel as soon as the computer<br />
indicates acceptance based on maximum allowable spans between springs.<br />
Because of this pinch effect only minimal movement is expected in<br />
service. Periodic inspection during the life of the reactor will be used<br />
as confirmation.<br />
Garter springs have now been redesigned in such a way that they will stay<br />
fixed in their design location in any future units. The major change is<br />
having the spring in tension around the pressure tube rather than in<br />
compression inside the calandria tube. Bruce Unit 8 now has a complete<br />
set of the new springs installed and their positions are being monitored<br />
periodically during the construction program. So far, none have moved.<br />
Methods are now being developed to find and reposition springs on<br />
reactors that went into service before this problem was identified. The<br />
present plan calls for inspection and repositioning in tubes still filled<br />
with water.
- 34 -<br />
At the Bruce we have completed the rehabilitation of two reactors and are<br />
presently working on the third. The second unit was completed at about a<br />
quarter of the cost of the first. The garter spring repositioning<br />
program has been a high profile work assignment for quality control. In<br />
most cases, quality control inspections are the last task to take place<br />
in a long chain of production operations. By the time it gets to us,<br />
it's usually behind schedule and everyone is quite willing to point the<br />
finger at quality control for holding up production. This, however, has<br />
been one of the gratifying experiences where quality control was an<br />
integral part of the production process and as such played a positive<br />
role in achieving a rather remarkable goal.<br />
I would like to thank the following for the immeasurable help they have<br />
provided :<br />
- Ontario Hydro Central Nuclear Services for their help in getting us<br />
started.<br />
- AECL's Chalk River Nuclear Laboratories who did miracles in designing<br />
probes and automatic scanning equipment.<br />
- Hy staff of forty-four inspectors and supervisors who made the job look<br />
easy.
END FITTING BELLOWS<br />
1<br />
END FITTING (TYR) v • IOOO mm.<br />
(TYP. )mmJ—«-<br />
'•> I<br />
2750 mm. — *- 6500 mm.<br />
—TUBE SHEET<br />
o> I 2000 mm.<br />
GARTER SPRING<br />
GARTER SPRINGS (TYP.)<br />
SIMPLIFIED VIEWS OF PRESSURE TUBE<br />
( NTS. ) NOTE= All Measurements Approximate<br />
•CALANDRIA TUBE<br />
PRESSURE TUBE<br />
, CALANDRIA TUBE<br />
r .<br />
-*— M 2750mm.<br />
FIGURE 1
1<br />
GARTER SPRING<br />
IN VERTICAL POSITI<strong>ON</strong><br />
GIRDLE WIRE<br />
ENDS OF GIRDLE WIRE<br />
RESISTANCE WELDED<br />
GARTER SPRING<br />
Zirc - Niobium<br />
of<br />
Travel<br />
z-Direction<br />
TILTED SPRING<br />
GARTER SPRINGS<br />
(NTS.)<br />
CALANDRIA TUBE<br />
GARTER SPRING<br />
PRESSURE TUBE<br />
FIGURE
- 37 -<br />
VERTICAL SPRING TILTED SPRING<br />
ISJ KSI<br />
1<br />
BET PU<br />
fr<br />
<br />
1 : Differential Probe<br />
2- Send/Receive Probe<br />
Trace on Cathode<br />
Ray Tube<br />
Strip Chart<br />
VARIATI<strong>ON</strong>S IN RESP<strong>ON</strong>SE FROM<br />
VERTICAL and TILTED SPRINGS with<br />
DIFFERENTIAL and SEND/RECEIVE PROBES<br />
FIGURE 3
DIGITAL READ OUT<br />
3.595<br />
TOIL ELEMENTS<br />
•TELESCOPING TUBE<br />
STRIP CHART RECORDER<br />
PRESSURE TUBE,<br />
MKI" STEM DRIVE<br />
(N.T.S.)<br />
FIGURE 4<br />
00'<br />
I
MEASURING TAPE<br />
HAND HELD PROBE<br />
(N.T.S.)<br />
FIGURE 5<br />
I
- 40 -<br />
EVALUATI<strong>ON</strong> OF ULTRAS<strong>ON</strong>IC METHODS FOR THE DETECTI<strong>ON</strong> AND SIZING OF<br />
REAL DEFECTS IN THE BASE OF RAILS<br />
l. Picke, and J.F.<br />
National Resea-tc/i Council Canada, Bsucii^ iv iHe , Qunhcc<br />
J.-P. Monchalin<br />
Encïgij, M-ôiei and Reiou-ïcciJ,<br />
Ottawa, Canada<br />
Abstract<br />
The presence of a seam or a crack in the base of rails is not uncommon and<br />
can, in some cases, lead to catastrophic failure if their depth should exceed<br />
= 0.5 mm (0.020 in). Such flaws should be detected during the production<br />
stage and it is required that the evaluation be done automatically in a<br />
continuous way and, of course, be nondestructive. We have carried out a<br />
survey of the different ultrasonic techniques which were applicable to this<br />
specific problem. In particular, the possibility of using surface (Rayleigh)<br />
waves was investigated, and compared to the more classical bulk wave<br />
pulse-echo techniques. Both artificial (machined) flaws and real flaws were<br />
studied. In the case of real flaws, results were found to be sensitive to<br />
whether the crack was closed and/or filled, especially when Rayleigh waves are<br />
used.<br />
1 INTRODUCTI<strong>ON</strong><br />
An important task of NDE is that of detecting and sizing surface breaking<br />
cracks. Such defects act as stress concentrators and depending on their<br />
origin, nature, and shape may lead to catastrophic failure by fast fracture.<br />
An example is giv&n by railways in Canada where the danger is enhanced since<br />
the rails are submitted to particularly harsh weather conditions. For this<br />
reason Canadian railroads impose strict specifications on newly manufactured<br />
rails; the head and the web [1] are inspected but also the base, where<br />
experience shows that the presence of surface flaws can be highly hazardous.<br />
Defects in the base occur usually in the form of seams and cracks, along the<br />
principal axis of the rail and located near the center. The main criterion<br />
for acceptability is that their depth be less than 0.020 in (0.5 mm).<br />
Different techniques are used and surface flaws are detected by visual,<br />
dye-penetrant, magnetic particles and ultrasonic techniques. Ultrasonics has<br />
often proven to be a very powerful tool for NDE, however, for surface flaws, a<br />
reliable [2] inspection procedure is not available as yet. Various methods<br />
have been proposed which may give accurate results for artificial defects but<br />
their relevance [3] in characterizing real defects can be debated. Here, we<br />
describe the work that we have done to appraise different techniques which use<br />
standard NUT equipment applicable to plant conditions and purposely leave<br />
aside more sophisticated but less easily applicable techniques. In
- 41 -<br />
particular, we investigated surface (Rayleigh) wave propagation and compared<br />
the results to shear wave pulse echo measurements performed both on artificial<br />
idealized cracks and on real cracks, which were then cut open for micrographie<br />
evaluation of the depths.<br />
II METHODOLOGY<br />
The basic physical properties which are used to describe the propagation of<br />
ultrasound are rather well established [4]. Unfortunately, the application is<br />
not simple and the results are often qualitative rather than quantitative.<br />
The techniques and possible approaches are numerous and have been given full<br />
reviews [5, 6], and we shall refer to them as we try to find and evaluate a<br />
practical method of characterizing our surface defects.<br />
A Experimental arrangement<br />
We have used "off the shelf" commercially available contact longitudinal<br />
transducers. By fitting these with plastic wedges of different angles both<br />
shear waves and Rayleigh surface waves could be launched. After some<br />
experience, we settled for 0.5 inch diameter Panametrics probes of the high<br />
resolution type (well-damped). We used the smaller probes instead of the<br />
larger ones because a) the resolution was better (typically 3 or 4<br />
oscillations); b) the signal to noise ratio was greater (circa 6dB); c) the<br />
companion wedge being smaller, there was less attenuation and dead time due to<br />
the plastic; d) the coupling to the irregular surface of the rail was easier.<br />
The transducers were excited with a Metrotek (Mod. MP 203) puiser, the<br />
receiver also Metrotek (Mod. MR 101A) had a 60 dB gain and a 20 MHz<br />
bandwidth. The output was fed to a gated peak detector and to an<br />
oscilloscope. In sum, our arrangement is classical and corresponds to what is<br />
usually found where ultrasonic NDT is routine.<br />
B Description of the samples<br />
We shall report on two different kinds of defects that can be found in rails.<br />
These were detected in the production plant with magnetic particles and their<br />
depth estimated by shear wave pulse-echo measurements. The first (SI) is an<br />
obvious crack 18 cm (7 in.) in length with a depth (h) estimated to exceed<br />
0.7 mm (0.030 in.) near the center; the other (S2) is a small hairline, hardly<br />
detectable crack 8 cm (3 in.) long.<br />
Before we attempted measurements on real defects, we evaluated the different<br />
techniques on artificial flaws. These artificial flaws were narrow slits 0.25<br />
mm; 0.010 in.), 5 cm (2 in.) in length which were EDM machined near the middle<br />
of the base of the rail. Twenty such slits with depths ranging from 0.15 mm<br />
(0.006 in.) to 1.3 mm (0.050 in.) were fabricated.
III MEASUREMENTS AND RESULTS<br />
- 42 -<br />
In the midst of available techniques we can distinguish those which involve<br />
time of flight considerations and those which rely on the measurement of the<br />
scattered amplitude. Methods of the first category are based on the measurement<br />
of the delay introduced in the time of flight of a sound pulse by the<br />
presence of an obstacle. In principle these methods are very accurate and<br />
reliable because a) time measurements can be made with a high degree of<br />
precision; b) only the arrival time of the signal is considered and not its<br />
amplitude so that the influence of variations in transducer coupling and<br />
effects of attenuation in the material are minimized. The possible approaches<br />
are numerous [7] using either bulk waves or surface waves, but as a general<br />
rule the delay corresponds to the time for a wave to travel up and down the<br />
lateral faces of the flaw.<br />
We attempted several variations of the time of flight technique but to no<br />
avail. The cause can be traced to the probes, for which the damping time is<br />
still not strong enough. Immersion type transducers may allow to produce the<br />
very short pulses but even then the sensitivity to such small cracks would<br />
still be small. So for this first approach to our problem, time of flight<br />
techniques will be considered as advanced NDT, which is outside the scope of<br />
this paper. We shall thus turn to the more usual technique of scattered<br />
amplitude measurements, first with bulk waves then with surface waves.<br />
A Bulk wave techniques<br />
The amplitude of the signal, which is scattered when an acoustic wave encounters<br />
an obstacle, is used as a signature of the defect. Simple theory [8]<br />
assumes that the defect produces a mirror-like reflexion with an amplitude<br />
that increases regularity with the size of the mirror. Actually the interaction<br />
is much more complex [9], In even the simplest case, one has to consider<br />
that part of the energy will be reflected, part will be transmitted and<br />
that mode conversion will occur along with diffraction. For surface breaking<br />
cracks, there will be multiple scattering from the surface and this will further<br />
complicate [10] the picture. For real situations, the number of factors<br />
which will influence the results increases dramatically: transducer coupling,<br />
sonic frequency bandwidth and mode [11], crack shape and orientation [8], the<br />
internal roughness [12] of the crack, the state of stress [13] etc. However,<br />
some of these difficulties might be overcome if the defects are more or less<br />
of the same type and this appears to be the case for the problem at hand.<br />
We needed first to establish the ability of the technique to detect a flaw.<br />
The investigation procedure is illustrated in Fig. 1 where the inspection is<br />
shown to be performed from the base. The frequency is 1 MHz, which was found<br />
to be an adequate compromise between resolution and signal to noise ratio.<br />
An angle of 45° was found to be the most suitable. In Fig. 1, the flaw is a<br />
1.2 mm (0.048 in) EDM slit. The transducer is located near the edge of the<br />
rail and the ultrasonic beam is first reflected by the face opposite the base<br />
where it spreads before it reaches the flaw. Because of this wide angle
- 43 -<br />
search beam, the defect is more easily found then by direct insonification and<br />
results are not so sensitive to the position of the probe. However the<br />
amplitude of the signal is reduced. The calibration curve for the amplitude<br />
(A) of the echo versus flaw depth (h) was found to be linear.<br />
B Surface (Rayleigh wave)<br />
One way to circumvent the problem of geometry and sensitivity to exact probe<br />
location is to propagate surface (or Rayleigh) waves [14]. This technique has<br />
been used mainly for the detection of surface breaking fatigue cracks [15] and<br />
the published results indicate that it should be a valuable tool here also.<br />
The simplest way to launch a surface wave is to use a 90° angle probe, as<br />
illustrated in Fig. 2. We noted that in the absence of a crack, the<br />
propagation is not very sensitive to the surface finish: the attenuation does<br />
not increase noticeably when going from a polisheo. surface to the rough<br />
surface of the rail, however, the surface must be clean of water or traces of<br />
oil. In Fig. 2, the photograph shows the signal from a 0.5 mm (0.020 in) EDM<br />
defect. The frequency is 1 MHz and the gain is half that of Fig. 1: (1)<br />
corresponds to the initial pulse, (2) is the wedge/rail interface, (3) the<br />
signal from the flaw, (4) a reflexion from the far end of the rail. The flaw<br />
signal is large and very easily found, so that the technique is as a very<br />
sensitive means of detecting the presence of even the smallest flaws (0.2 mm,<br />
0.006 in). We have also experimented with leaky Rayleigh waves [16]. In this<br />
approach, where the surface is immersed, the sensitivity is even better.<br />
However, the surface wave is highly damped by the liquid and the inspected<br />
zone is reduced; the alignment of the transducers is critical and great care<br />
must be taken to eliminate spurious echos.<br />
Using the contact method, we went on to establish a correlation between the<br />
amplitude of the echo (R) and the depth of the flaw (h). The results for R<br />
versus the ratio crack depth/wavelength (h/X) are shown in Fig. 3, for 20 EDM<br />
defects at different frequencies (.5,1, 2.5 and 4 MHz). In the domain of long<br />
wavelengths or small defects, the relation between the amplitude of the echo<br />
signal and the crack depth can be reasonably approximated by a straight line<br />
(at 1 MHz for h = 1.3 mm; 0.050 in). The oscillatory behaviour which is<br />
observed in Fig. 3 is well-known [14, 17] and it is the basis for the spectral<br />
analysis [18, 19] approach. One may see the similarity with the problem of a<br />
transmission line, the crack acting as a cavity: the non-linearity corresponds<br />
to a multimode, multifrequency operation. We tried this approach where the<br />
crack is identified by the spectral density of the reflected pulse. The<br />
effects are rather small and the technique, if it holds promise, belongs to<br />
what we classified as advanced NDE.<br />
The defect-cavity analogy brings out the importance of the geometry factor:<br />
the propagation will be sensitive to the shape [19] of the defect (cavity) and<br />
more so at higher frequencies. If the EDM defects differ by factors other<br />
than their depth (h), this will manifest itself by scatter of the data points,<br />
as observed in Fig. 3 where the scatter exceeds our experimental uncertainty.
C Real defects<br />
- 44 -<br />
Here we report our results using both techniques in evaluating real defects in<br />
rails. The probe (1 MHz) is displaced along the length (x) of the rail; the<br />
depth (h) is estimated (.) using the linear calibration curve for bulk, waves<br />
and Fig. 3 for Rayleigh waves. The results obtained on rail sample SI using<br />
the bulk wave technique are shown in Fig. 4 and those using surface waves in<br />
Fig. 5. For the case of sample S2, the bulk wave signal was too weak to allow<br />
sizing of the depth; with surface waves we estimated a uniform dîpth h =0.13<br />
mm (0.005 in) along the length. Measurements were also performed at .5 and 2<br />
MHz, which were overall identical.<br />
The samples were then cut open and micrographs such as those of Fig. 7 (sample<br />
SI) and Fig. 6 (sample S2) were made. The actual depth of the defect from<br />
mouth to tip was measured. The values (V) which have been reproduced in Fig.<br />
4 and 5 show that the total depth is underestimated by both methods. This is<br />
even more so for the case of rail S2 for which the micrographs indicate that<br />
depth is of the order of 1 mm (.040 in). However, it can be seen in Fig. 6<br />
that the cracks are very often closed or filled, at least partly. Those<br />
sections of the cracks which are partly filled or closed will scatter<br />
ultrasound non coherently [12]. Since we have measured the amplitude of the<br />
reflected signal, it is to be expected that the main contribution was from<br />
that part of the flaw which is open. Even though the exercise is not<br />
foolproof, we have attempted to evaluate the open length of the crack. The<br />
values obtained are shown ([]) in Figs. 4 and 5. For the bulk wave<br />
experiments (Fig. 4), the agreement between the ultrasonically determined and<br />
the measured open depths appears to be better.<br />
In the case of surface wave reflexion, the analysis is more delicate, given<br />
that the scattering mechanism is more complex [20]. Various [18] studies have<br />
shown that reflected signals can be seen as the added contribution of a<br />
reflexion from the lip of the crack and one from the bottom. Even if the<br />
crack is closed, the lip will cause a signal to be reflected with an amplitude<br />
that is not related to the actual depth. This is what occurs in the case of<br />
sample S2. Concerning the contribution from the bottom of the crack, it is by<br />
definition correlated to the open depth. Then the correlation will also<br />
depend on the details of the geometry as was mentioned above. The calibration<br />
we have used was for 0.2 mm wide EDM slits, whereas real defects have widths<br />
which are two orders of magnitude smaller. This difference in geometry might<br />
in part explain why the agreement between the estimated depth (.) and the<br />
measured open flaw depth ([]) is not as good as for the bulk wave technique<br />
(Fig. 4). The description for the contribution of the closed or filled<br />
section of the crack becomes very involved. In a recent theoretical paper,<br />
W.M. Visscher established [21] that the response function of such a defect is<br />
multivalued and that the inverse scattering problem is undetermined.
C<strong>ON</strong>CLUSI<strong>ON</strong><br />
Classical bulk and surface wave pulse-echo techniques were evaluated in their<br />
ability to detect and size surface breaking cracks in the base of rails.<br />
Although both of these techniques give good results for EDM slots, echo<br />
amplitudes from real cracks depend on crack closure and the presence of<br />
filling material. Rayleigh waves offer considerable advantages in terms of<br />
singal to noise and ease of detection [22] but their high sensitivity to<br />
factors other than crack depth make them unusable for sizing. Classical shear<br />
wave pulse-echo techniques were found to be in good agreement with the open<br />
flaw depth when calibrated on EDM slots. For proper sizing of the full depth,<br />
the techniques used were found inadequate. Since the closed or filled part of<br />
cracks are partially transparent to ultrasound, it is likely that more<br />
advanced NDT techniques using feature extraction in the time or frequency<br />
domain would also be inadequate without adaptation. This work points out the<br />
need for theoretical modelling and experimentation on the interaction of<br />
ultrasound with closed cracks and in particular of the scattering phenomena<br />
occuring at the tip [23],<br />
REFERENCES:<br />
[1] K.G. Hall; Non-Destructive Testing, p 121, June (1976)<br />
[2] H.J. Mayer; Mater. Eval., j42, 793 (1984)<br />
[3] J. Drury; Brit. J. NDT, p 200, July (1979)<br />
[4] E.A. Kraut; IEEE Trans. Sonics Ultrasonics, Su-^3_, 162 (1976)<br />
[5] P.A. Doyle, CM. Scala; Ultrasonics, 16, 164 (1978)<br />
[6] R.B. Thompson, D.O. Thompson; J. Metals, p. 29, July (1981)<br />
[7] K. Date, H. Shimada, N. Ikenaga; NDT Intl, JJ5» 315 (1982)<br />
[8] S. Serabian; Mater. Eval., J39_, 1243 (1981)<br />
[9] K. Harumi, H. Okada, T. Saito, T. Fujimori; IEEE Ultrasonics<br />
Symposium, p. 904 (1982)<br />
[10] T. Kundu, A.K. Mai; Trans. ASME, 48, 570 (1981)<br />
[11] V.U. Schmitz, F.L. Becker; Mater. Eval., _40, 191 (1982)<br />
[12] L. Adler, K. Lewis, M. de Billy, G. Quentin; "New Procedures in<br />
Nondestructive Testing" P. Holler, Ed. (Springer-Verlag 1983) p. 163<br />
[13] R.B. Thompson, C.J. Fiedler; "Review of Progress in Quantitative<br />
Nondestructive Evaluation 3"; D.O. Thompson and D.E. Chimenti, Eds.<br />
(Plenum Press, NY, 1984)<br />
[14] I.A. Viktorov; "Rayleigh and Lamb Waves", (Plenum Press, NY 1967)<br />
[15] J.M. Carson, J.L. Rose; Mater. Eval. p. 27, April (1980)<br />
[16] L. Adler, M. de Billy, G. Quentin; J. Appl. Phys. j)3, 8756 (1982)<br />
[17] M. Hirao, H. Fukuoda, Y. Mirura; J. Acoust. Soc. Ara. 7^, 602 (1980)<br />
[18] G.P. Singh, A. Singh; Mater. Eval. J39_, 1232 (1981)<br />
[19] Y.C. Angel, J.D. Achenbach; J. Acoust. Soc. Am., 7j>, 313 (1984)<br />
[20] B.W. Reinhardt, J.W. Dally; Mater. Eval., ^8, 213 (1970)<br />
[21] W.M. Visscher; ibid [15]<br />
[22] H. Seiger; NDT Intl., p. 131, June (1982)<br />
[23] R.B. Thompson; Applied Mechanics Division of ASME, Symposium on Wave<br />
Propagation in Inhomogeneous Media, San Antonio, Texas, June 1984
Flaw<br />
Figure 1: Schematic of experimental set-up used for bulk waves and<br />
oscilloscope trace for an EDM defect 0.040" deep.<br />
Flaw t<br />
Figure 2: Schematic of Kayleigh wave set up and oscilloscope trace obtained<br />
with a 0.020" deep EDM slot using half the gain used in Figure 1.
R<br />
.8<br />
.7<br />
.6<br />
.5<br />
A<br />
.3<br />
2<br />
.1<br />
—<br />
—<br />
A /<br />
/ °<br />
D A<br />
D /A<br />
— /°<br />
/ °<br />
j<br />
f<br />
STo •<br />
i<br />
B<br />
I<br />
\ \<br />
o \<br />
AA I<br />
t<br />
IvV - 47 -<br />
/<br />
I<br />
O /<br />
/<br />
/A<br />
/ A<br />
A<br />
A • 0.5<br />
° 1.0<br />
o 2.25<br />
* 4.0<br />
i<br />
0.5 1.0 1.5<br />
(h/X)<br />
i<br />
—<br />
—<br />
—<br />
—<br />
MHz "<br />
MHz<br />
MHz "<br />
MHz —<br />
Figure 3: Reflection coefficient for Rayleigh waves, versus flaw depth<br />
normalized to ultrasonic wavelength, h/X, for EDM slots. The<br />
solid curve was obtained from the literature.<br />
i
0.5<br />
E10<br />
Q.<br />
8<br />
1.5<br />
2.0<br />
- 48 -<br />
Position along defect<br />
01 2 3 4 5 6 7 8 9 10 X(cm)<br />
\ I<br />
• Bulk wave estimate<br />
v True full depth<br />
o Open flaw depth<br />
S1<br />
10<br />
20<br />
30<br />
40<br />
D<br />
CD<br />
O<br />
50 _<br />
Figure 4: Depth of defect SI along its length obtained using shear wave<br />
pulse-echo (•) compared to the full depth (A) and open flaw depth<br />
([]) measured metallographically.<br />
60<br />
70<br />
80
CD<br />
Û<br />
0.5<br />
1.0<br />
1.5<br />
2.0<br />
- 49 -<br />
Position along defect<br />
0 "i : 2 : 3 4 ! 5 (<br />
H.<br />
—<br />
> ;<br />
—<br />
i<br />
i<br />
i<br />
i<br />
|<br />
•<br />
i i1<br />
i<br />
i<br />
i I<br />
i<br />
i<br />
I<br />
i<br />
!<br />
\ .<br />
i<br />
\<br />
\<br />
\<br />
)<br />
L<br />
i<br />
i<br />
•<br />
i<br />
ir<br />
<<br />
•j<br />
7 8 ino<br />
X(cm)<br />
I"'<br />
<<br />
i<br />
1 '<br />
î *~ i<br />
""* 1<br />
- 50 -<br />
Figure 6: Optical micrograph of seam SI used for the comparison between<br />
Rayleigh wave and shear pulse-echo mearurements. (64X)<br />
mmm<br />
Figure 7: Optical micrograph of seam S2 showing large closed regions. (40X)
- 51 -<br />
OPTICAL DETECTI<strong>ON</strong> OF ULTRASOUND AT DISTANCE<br />
Jean-Piz um Monchalin<br />
Enzngu, Mines and Re souicc s, Ottawa, Ontario<br />
+ Work performed in the laboratories of the Industrial Materials Research<br />
Institute, Boucherville.<br />
ABSTRACT<br />
Various probes which are based on optical interferometry and are used to<br />
detect ultrasound at distance (distances from less than 1 millimeter to<br />
several meters are considered) are reviewed. Their properties and sensitivity<br />
(detection limit) are critically discussed. We analyze a Michelson<br />
interferometer probe which senses the surface displacement and another one,<br />
working as an optical frequency discriminator, which senses the surface<br />
velocity. A differential arrangement is also discussed. We present also<br />
preliminary experimental results obtained by laser generation and optical<br />
detection on steel at distances ranging from 1 to 2 meters.<br />
1. INTRODUCTI<strong>ON</strong><br />
In recent years, considerable attention has been focused on the generation of<br />
ultrasound using pulsed lasers and numerous results have been published (1, 2,<br />
3, 4). This technology has the advantage to enable truly generation at<br />
distance, so it enables to produce ultrasound in hot pieces (e.g. in the steel<br />
industry) and at locations with difficult access. It is also worth noting<br />
that very short ultrasonic pulses can be produced, which enables an excellent<br />
resolution close to the surface. This technology, in order to be useful, has<br />
however to be coupled with a detection means which can also operate at<br />
distance. Optical interferometry has been recognized to be such a means.<br />
Numerous interferometric systems for the detection of ultrasound have been<br />
previously described, but if we except the one developped by a private company<br />
(5), none can really be applied to an industrial environment. Industrial<br />
applications require a system with a sufficiently high sensitivity to be able<br />
to be used on pratical rough surfaces. These surfaces have not been specially<br />
polished, as it can be done in the laboratory for more fundamental<br />
experimentation, and they scatter incident light in various directions,<br />
leaving a small fraction to be collected by the detection apparatus. This<br />
system has also to be immune from ambient vibrations. Also, except for very
- 52 -<br />
specialized applications such as microscopy, a probing area of the order of<br />
that produced by a conventional transducer is desirable. In this paper, the<br />
techniques for the detection of ultrasound based on optical interferometry are<br />
reviewed and discussed, especially their potential for application in the<br />
industry. We will also present some experimental results obtained by<br />
generating and detecting ultrasound at distances over 1 meter (both).<br />
First, it is important to understand how the ultrasonic vibration of a surface<br />
affects the light which impinges on it. It is well known that the surface<br />
motion shifts the frequency of the reflected or scattered light by the Doppler<br />
effect, but it is also useful to see the phenomenon in a different way.<br />
Let us assume that the surface moves by S(t) at a location where a light field<br />
Eo cos ut is incident,
- 53 -<br />
performing the beating right on the surface by creating a light grating on<br />
it. This technique is similar to differential Doppler anemometry (9) and is<br />
described in section III.<br />
II THE DISPLACEMENT MICHELS<strong>ON</strong> INTERFEROMETER<br />
Such a system is sketched in fig. 2. The beam from the laser is focused onto<br />
the surface which acts as one of the mirror of the interferometer. Without a<br />
frequency shifter in the reference arm (homodyne interferometer), the detected<br />
signal can be written as follows (excluding constant terms):<br />
ID = 2 IL fk Js cos [4 IT 6 (t)/X -*(t)] (1)<br />
where Ij^ is the laser power, R is the effective transmission coefficient in<br />
intensity for the beam in the reference arm and S is that for the beam<br />
reflected off the surface (S is much less than unity for practical unpolished<br />
surfaces), $(t) is a phase factor which depends upon the interferometer path<br />
difference and is affected by vibrations. In practice, the reference mirror<br />
is mounted on a piezoelectric pusher which enables to ajust the path<br />
difference in such a way that $=+_ IT/2 + 2mr (n is an integer). It then<br />
follows that the signal is proportional to the displacement. In practice,<br />
electronic stabilization circuitry should be used to keep this adjustment in<br />
spite of ambient vibrations (10, 11, 12). It is also possible to have a<br />
system insensitive to vibrations by generating inside the Michelson<br />
interferometer two optical beams in quadrature (13, 14). This last design<br />
needs critical alignment and both have the drawback to be not easily<br />
calibrated (14).<br />
If the detection is performed about a higher frequency by shifting the optical<br />
frequency in the reference arm (heterodyne interferometer), calibration can be<br />
easily performed in real time: the signal from the detector (-2Tifgt is added<br />
to the phase in eq. 1) includes a carrier at the shifting frequency fg and<br />
two sidebands (as shown in fig. 1, except the spectrum is now centered on<br />
fg) and the calibration can readily be performed by comparing the sidebands<br />
amplitude to that of the carrier. An interferometric probe based on such a<br />
design has been realized (the frequency shifting is given by an opto-acoustic<br />
cell driven at 40 MHz). This probe has already been presented (15), its<br />
performances for the detection of pulsed ultrasound will be described<br />
elsewhere (16), as well its electronic circuitry for continuous (17) and<br />
pulsed ultrasound (18). It can be shown that the signal-to-noise ratio,<br />
assuming quantum noise limited detection and detection of both sidebands, is<br />
expressed by:<br />
is/iN = (4 n 6/A) ^2 11 S n/(B h v) (2)<br />
where n is the quantum efficiency of the detector, e is the electron charge,<br />
hv is the energy of a quantum of light of frequency v and B is the detection<br />
bandwidth.
- 54 -<br />
As the probed surface is not polished to a mirror like finish, the scattered<br />
light shows the well known speckle phenomenon. It can also be shown that the<br />
best signal-to-noise ratio is obtained when one speckle is detected (18) and<br />
when the speckle size is of the order of the incident beam size, which can be<br />
obtained by focusing on the surface with diffraction limited optics (19).<br />
Then, if the beam diameter on the lens is noted a and if the focusing distance<br />
between the lens and the surface is noted D, the solid angle of one speckle is<br />
" (a/D) . Assuming isotropic scattering and neglecting the losses produced by<br />
various optics, we then find that S » (a/2D) 2 . Taking IL = 5 mW, a = 4 mm,<br />
D = 15 cm a detection limit of = 1.5 Â is found for a bandwidth of 10 MHz. A<br />
detection limit of the order mentionned above has been observed with our<br />
system on nearly isotropic scattering surfaces (18). This limit can be<br />
lowered by using a higher laser power (the detection limit scale as l/^/T^),<br />
or a closer working distance D (the detection limit scales as 1/D) (20, 21),<br />
but closer focusing produces a smaller spot size and a shorter depth of<br />
focus. These results are consistant, as shown below, with the antenna theorem<br />
for heterodyne detection (22) which states that the étendue of the receiving<br />
system (i.e. its light gathering efficiency, exactly the product of the viewed<br />
area by the solid angle of the system aperture from the surface) is X at<br />
most. The maximum received power is obtained for the minimum spot size on the<br />
surface and is then equal, to II X /TT WQ , where WQ is the focal spot<br />
radius. Since WQ - XD/a we retrieve equation 5.<br />
Since the displacement Michelson interferometer probe produces a useful<br />
detection limit only for sharp focusing and close working range, its<br />
usefulness for industrial NDT applications is rather limited. However, it<br />
should be noted that it does not require a single frequency laser and as far<br />
as the path lengths in the interferometer are compensated, it can use a rather<br />
broad band laser. It can be very useful for ultrasonic field mapping and<br />
ultrasonic transducer characterization, since, in this case the signal being<br />
repetitive, the signal-to-noise ratio can be greatly increased by simple<br />
averaging. This probe measures most easily displacements normal to the<br />
surface, but it can also be used when inclined to measure in-plane motion<br />
(17).<br />
In conclusion, this type of optical receiver which is rather simple to realize<br />
has however a small étendue (its figure of merit) of the order of X . We will<br />
see below that velocity probes can be made with a much larger étendue.<br />
Ill THE DIFFERENTIAL DISPLACEMENT PROBE<br />
(OR OPTICAL GRATING DISPLACEMENT PROBE):<br />
As seen in fig. 3, this probe is made by having two beams issued from the same<br />
laser and separated by an angle 26 focused at the same location on the surface<br />
(18, 23). A grating is then produced on the surface and a part of the speckle<br />
field scattered from the surface is viewed by a detector. The grating is<br />
fixed when the frequency of the 2 beams is the same (homodyne probe) and<br />
moving when it is heterodyne. If N speckles of mean intensity Isp are seen
- 55 -<br />
by the detector, the detected signal in the case of the heterodyne probe can<br />
be written (the speckles add up incoherently) (18):<br />
l D = sß Isp cos f 2irf ßt + (?i -t2).t(t) + * (t)] (3)<br />
where fjj is the shifting frequency kj and k2 are the wavevectors of the 2<br />
beams, (ki - k2>.ô = 4n sin 6 (sin a 6x + cos a 6z)/A, a being the angle<br />
between k^ - k2 and the normal to the surface, ôx and Sz being respectively<br />
the in-plane and out-of-plane displacements. As seen above, this expression<br />
gives a carrier signal at fß and two sidebands, which enables easy<br />
calibration. In-plane and out-of-plane displacements can be detected, but<br />
this probe is particularly useful to detect displacements parallel to the<br />
surface (in this case ki-&2 is perpendicular to the surface and a = TT/2).<br />
When both sidebands are detected, it can be shown that the signal-to-noise<br />
ratio is given by (it is indépendant of the number of speckles):<br />
is/iN = (?i -k2).t \/n Isp/(2 B h v) (4)<br />
Eq. 4 also shows that, in order to improve the signal-to-noise ratio,<br />
ISpShould be maximized, which means that the speckle size should be of the<br />
order of the detection aperture and consequently that this aperture should be<br />
large enough to collect a large part of the scattered light. This means, in<br />
turn, very sharp focusing. When typical numerical values are used, a<br />
detection limit of the same order as the one found before is obtained.<br />
Therefore, this probe has the same drawback as the one described in the<br />
previous section. Furthermore, it tends to be more bulky since 0 cannot be<br />
too small for a reasonable detection limit.<br />
IV THE MICHELS<strong>ON</strong> VELOCITY INTERFEROMETER<br />
A typical setup is sketched in fig. 4. We will note the main difference<br />
between this setup and the one of fig. 2: here the interferometer does not<br />
make use of the surface as a mirror, but views the light scattered by it.<br />
This type of system has been used in shock wave research (25) and has been<br />
also considered for the detection of ultrasound (26). The detected intensity<br />
for a given direction 6 of an incident ray is:<br />
ID = Ai + A2 cos (2TT V/AV + $) (5)<br />
where A, B and are constant, the free spectral range Av = 1/T =c/2Ad(8),<br />
(T is the delay time, Ad(8) is the difference of arms lengths for the incident<br />
ray inclined by 6. As sketched in the boxed diagram of fig. 4, which plots<br />
the spectral response given by eq. 5, frequency discrimination is obtained<br />
when the laser frequency (a single frequency laser source is necessary) is<br />
tuned to a zero crossing 2ir v/Av + = +_ IT/2 + 2mir, m being an integer). The<br />
surface being rough and irregular acts as an incoherent source, so the fringes<br />
are only observed at infinity (24) or at the focus of a lens. It can be<br />
readily shown that Ad(6) = Ad cos 6, where Ad is the path difference for the<br />
ray perpendicular to the mirrors. The fringes are then concentric rings, and
- 56 -<br />
in practice the central fringe is selected by using a circular hole at the<br />
lens focus (at the expense of a much higher complexity a mask pattern matching<br />
the fringes can be used). Taking a maximum path difference of A/4, it follows<br />
a limit aperture 654 « >J A/4Ad. In order to be able to detect ultrasound in<br />
the range 1 to 10 MHz, the free spectral range Av (the bandwidth can be<br />
defined as Av/2) should not be too large: taking Av = 25 MHz yields Ad = 6 in<br />
(even if folding mirrors are used, the system will be large and bulky) and<br />
8j4 = 2 10~ rd for A = 1.06 pm. Using an entrance aperture of 10 cm in<br />
diameter (and mirrors of this size), the system has an étendue of = 10" mm .<br />
This is much larger than the optical heterodyne systems described in section<br />
II and III, but not sufficient to use all the light scattered by an area of<br />
1/4 inch in diameter, 2 m away from a 10 cm aperture: the required étendue is<br />
- 0.06 mm . We present below a way to increase the étendue.<br />
The time response of this interferometer can be understood by considering that<br />
the path length from the laser source is changed by 6(t) for one arm and by<br />
6(t-T) for the other (27). Then, when the laser frequency is properly<br />
adjusted to the zero crossing, the detected signal is given by (6
used, a front lens too large to he feasible. Therefore, it can be concluded<br />
that the étendue will be limited by the viewed area and the front optics size<br />
(= 0.06 mm for example chosen) and that there is no need to increase further<br />
that of the interferometer (29).<br />
We evaluate below the sensitivity of such a system. We assume a beam splitter<br />
with 50% reflexion and transmission and neglect any other reflexion losses in<br />
the interferometer. We also assume isotropic scattering, a surface<br />
displacement U cos(2irfut + ) with fu ^> Av and quantum noise limited<br />
detection. Then it can be shown that the signal-to-noise ratio is given by:<br />
is / iN = 4(fu/Av) (U/X) JIT iL (1-A) Ü n/(hv a S)' (6)<br />
where II is the laser power, E is the étendue, A is the absorption<br />
coefficient Oi. the surface and S is the viewed area. Taking II = 10 W, A =<br />
0, X = 1.06 pm, fu = 2.5 MHz, Av = 25 MHz, B =10 MHz, n = 0.5, E = 0.06 mm 2 ,<br />
S = area corresponding to a 1/4 inch in diameter, we find a detection limit of<br />
0.2Â. This limit is generally appropriate for the detection of laser<br />
generated ultrasound on a metal surface. For more absorbing surfaces, such as<br />
carbon fibers composites, a higher power may be required. One should note<br />
some improvement with a narrower bandwidth Au/2 and a higher étendue E,<br />
although there are limitations on the physical size permissible for such a<br />
system. The detection limit calculated above assumed a single frequency laser<br />
source which is very stable, both in amplitude and frequency. In practice,<br />
the laser should be stabilized with respect to the interferometer and the<br />
fluctuations of amplitude and frequency should be compensated by dividing or<br />
substracting electronically the signal coming from the sample with a signal<br />
representative of these fluctuations (28).<br />
V EXPERIMENTAL RESULTS OBTAINED WITH A VELOCITY INTERFEROMETER<br />
CANMET and IMRI are engaged in the development of an ultrasound laser<br />
generating and receiving system. We present below some results obtained with<br />
a preliminary version of this system.<br />
Fig. 7a shows the present experimental setup, which uses as sample a half inch<br />
steel plate. The generating laser is a Q-switch Nd-YAG laser (made by<br />
Lasermetrics, typically half a Joule multimodcs, 10 to 20 ns pulses), which is<br />
focused with a 2 m focal length lens on thfî plate. Ultrasonic displacements<br />
are detected in the present setup from the opposite side of the plate. Fig.<br />
7b shows the echos obtained from multiple reflexions in the plate. The noise<br />
observed on the picture is quantum noise, except the one at the beginning of<br />
the trace which originates from electromagnetic interference from the laser.<br />
The ultrasonic displacement for the generating conditions used is unipolar,<br />
but since the interferometer detects (as seen above in first approximation)<br />
the optical frequency change and the surface velocity, each echo appears as a
- 53 -<br />
bipolar pulse, as it is seen, using an expanded scale, in Fig. 7c. Although<br />
the laser pulse lasts 10 to 20 ns, the ultrasonic pulse, because of the very<br />
high attenuation of high frequencies, lasts about 10 times longer, as it is<br />
also observed. It should be mentionned that the surface which was illuminated<br />
by the laser pulse was covered by a water film, which had the effect to<br />
increase the ultrasonic source strength, but also that only a low power<br />
helium-neon laser (2 mW) was used. Developments are scheduled which should<br />
boost the signal by several orders of magnitude and bring the whole setup to a<br />
practically useful stage for defect detection and also thickness gauging.<br />
VI C<strong>ON</strong>CLUSI<strong>ON</strong><br />
We have presented a review of the various interferometric means to detect<br />
ultrasound at distance. Displacement probes based on a Michelson<br />
interferometer and optical heterodyning are relatively simple, have a high<br />
spatial resolution, a wide bandwidth, but a limited sensitivity- Their field<br />
of application is the mapping of ultrasonic fields, and they can be very<br />
useful for the evaluation of various ultrasonic NDT procedures and equipment.<br />
In particular, these probes could be used to check the operation ultrasonic<br />
transducers (in time and space over the transducer surface) and also could<br />
avantageously replace such ultrasonic field evaluation procedures using<br />
artificial reflectors (e.g. drilled holes). When shear displacements parallel<br />
to the surface have to be measured, a differential arrangement should be<br />
prefered, but both types of probes ("absolute" or differential) can be used to<br />
detect shear or compression displacements and have similar detection limits,<br />
although the differential probe tends to be more bulky. Velocity probes which<br />
use an interferometer such as the Michelson interferometer as an optical<br />
filter or an optical frequency discriminator are more sensitive and can be<br />
used for ultrasonic NDT at distance in an industrial environment. When<br />
coupled with ultrasonic generation by a pulsed laser, a NDT ultrasonic system<br />
with unmatched performances can be realized, enabling in particular to probe<br />
very near a surface, to work with geometries with difficult access, to probe<br />
pieces at elevated temperature and an easy scanning.<br />
ACKNOWLEDGEMENT<br />
This work has benefited from the skillful technical assistance of Mr. R. Hgon.<br />
REFERENCES<br />
1. C.A. Calder and W.W. Wilcox, "Noncontact material testing using laser<br />
energy deposition and interferometry", Materials Evaluation, 38, 86<br />
(1980).<br />
2. C.B. Scruby, R.J. Dewhurst, D.A. Hutchins, and S.B. Palmer, "Laser<br />
generation of ultrasound in metals", in Research techniques in NDT,<br />
R.S. Sharpe Ed., Vol. 5, Academic Press, 1983, pp. 411-413.
- 59 -<br />
3. D.A. Hutchins, "Non-contact ultrasonics using lasers", in Advanced NDE<br />
Technology, Proceedings of the Conference held in Montreal May 31 -<br />
June 1, 1982, J.F. Bussiêre ed., p. 125.<br />
4. P. Cielo, F. Nadeau and M. Lamontagne, "Laser generation of<br />
convergent acoustic waves for materials inspection", to be published<br />
in Ultrasonics, 1984.<br />
5. Krautkramer-Branson Inc., Lewistown, PA, Commercial note on the Laser<br />
Ultrasound system.<br />
6. A recent review is found in M.D. Levenson "Introduction to Nonlinear<br />
Laser Spectroscopy", Academic Press, New York, 1982.<br />
7. Wolf Bickel, "Method and apparatus for receiving ultrasonic waves by<br />
optical means", US patent // 4,345,475, Aug. 1982, assigned to<br />
Krautkramer-Branson Inc.<br />
8. Wolf Bickel, "Method and apparatus for receiving ultrasonic waves by<br />
optical means", US patent it 4,129,041, Dec. 1978, assigned to<br />
Krautkramer-Branson Inc.<br />
9. L.E. Drain, "The Laser Doppler Technique", J. Wiley, New York, 1980,<br />
see chap. 5.<br />
10. C.H. Palmer and R. Green Jr., "Optical detection of acoustic emission<br />
waves", Appl. Opt. 16, 2333 (1977).<br />
11. M. Kroll and B.B. Djordjevic, "A laser stress-wave probe with<br />
sub-angstrom sensitivity and large bandwidth", IEEE Ultrasonics<br />
Symposium Proceedings, p. 864 (1982).<br />
12. J.-P. Monchalin, Progress report IGM 83-231-02, Appendix B<br />
(unpublished).<br />
13. D. Vllkomerson, "Measuring pulsed picometer-displacement vibrations b<br />
optical interferometry", Appl. Phys. Lett. ^9_, 183 (1976).<br />
14. J.-P. Monchalin and R. Hgon, Progress Report IGM 84-402-01, section I<br />
(unpublished).<br />
15. Presented at the Canadian Steel Industry Research Association Sensor<br />
Research Workshop, Hamilton, Ontario, May 8-9, 1984.<br />
16. J.-P. Monchalin, "Optical detection of ultrasound at distance", to be<br />
published in the Canadian Society for Nondestructive Testing Journal.<br />
17. J.-P. Monchalin, Progress Report IGM 84-402-01, Appendix A<br />
(unpublished).
- 60 -<br />
18. J.-P. Monchalln and R. Héon, Progress Report IGM 84-402-02, in<br />
preparation.<br />
19. A. E. Ennos, in Progress in Optics, vol XVI, edited by E. Wolf, North<br />
Holland, (1978), see section IV, Speckle Interfereomety.<br />
20. J.E. Bowers, R.L. Jungerman, B.T. Khuri-Yakub and G.S. Kino, "An all<br />
fiber-optic sensor for surface acoustic wave measurements", J.<br />
Lightwave Tech. LT-1, 429 (1983).<br />
21. R.L. Jungerman, B.T. Khuri-Yakub and G.S. Kino, "Characterization of<br />
surface defects using a pulsed acoustic laser probe", Appl. Phys.<br />
Lett. 4£, 392 (1984).<br />
22. A.E. Siegman, "The antenna properties of optical heterodyne<br />
receivers", Appl. Opt. 5, 1588 (1966).<br />
23. R. Dändliker and J.-F. Willemin, "Measuring microvibrations by<br />
neterodyne speckle interferometry", Optics Lett., 6, 165 (1981).<br />
24. M. Born and E. Wolf, "Principles of Optics", Pergamon Press, 4th ed.<br />
1970, chap. 7.5.4, p. 301.<br />
25. L.M. Barker, "Laser Interferoraetry in Shock-wave Research", Exp.<br />
Mech., May 1972, p. 209.<br />
26. W. Kaule, "Interferometrlc method and apparatus for sensing surface<br />
deformation of a workpiece subjected to acoustic energy", US patent //<br />
4,046,477, (Sept. 1977), assigned to Krautkramer-Branson Inc.<br />
27 R.J. Clifton, "Analysis of the Laser Velocity Interferometer", J.<br />
Appl. Phys., hl_, 5335 (1970).<br />
28. W. Kaule, "Method and apparatus for receiving ultrasonic waves by<br />
optical means", US patent // 4,388,832, (June 1983), assigned to<br />
Krautkramer-Branson Inc.<br />
29. C.W. Gillard, "Optical differential interferometer discriminator for<br />
FM to AM conversion", US patent it 3,503,012 (March 1970), assigned to<br />
Lockeed Aircraft Corp. Calif.
JMU V J*fu Optical<br />
frequency<br />
a. Continuous ulrasonic<br />
excitation<br />
Sample<br />
- 61 -<br />
b. Pulsed ultrasonic<br />
excitation<br />
V Optical<br />
frequency<br />
Fig. 1 Optical spectrum of the light scattered from the surface.<br />
Ultrasound<br />
Lens<br />
Mirror<br />
S Frequency shifter<br />
-.--J<br />
Detector<br />
Beam splitter<br />
Fig. 2 Schematic of the displacement Michelson interferometer: without the<br />
frequency shifter, the probe is called homodyne and with it in the<br />
reference arm, it is called heterodyne.
Sample<br />
Lens<br />
Ultrasound<br />
- 62 -<br />
Detector<br />
Frequency shifter<br />
Beam splitter<br />
Mirror<br />
Fig. 3 Schematic of the differential displacement probe (or optical grating<br />
displacement probe): the probe in homodyne without the frequency<br />
shifter and heterodyne with it.
- 63 -<br />
Sample •D<br />
Av<br />
Optical<br />
P^ frequency<br />
Fig. 4 Schematic of the Michelson velocity interferometer. The boxed diagram<br />
indicates the response of the interferometer to optical frequency.
CD<br />
CO<br />
c<br />
O<br />
Q.<br />
W<br />
0<br />
- 64 -<br />
Sin(7TfuT)<br />
Ultrasonic frequency<br />
Fig. 5 Response of the Michelson velocity interferometer to ultrasonic<br />
frequency (the laser frequency is set at a zero crossing, as shown in<br />
the boxed diagram of fig. 4).<br />
Mirror M2<br />
f u<br />
Image of M2<br />
Mirror M1<br />
Fig. 6 Schematic of the Michelson interferometer with increased étendue.
Nd-YAG<br />
generating laser<br />
Interferometer<br />
& detector<br />
Illuminating laser<br />
Steel plate (Vz inch thick) Scope<br />
Fig. 7 Generation and detection of ultrasound at distance. a) Present<br />
experimental setup for the system under development, b) Ultrasonic<br />
echos sequence observed from a i inch steel plate, c) One echo from<br />
the sequence.
- 66 -<br />
AUTOMATED ULTRAS<strong>ON</strong>IC TESTING SYSTEMS FOR THE CHARACTERIZATI<strong>ON</strong><br />
OF DEFECTS IN WELDMENTS<br />
M. Macecefe, K. Lu&cott, 3. Wells,<br />
Techno Scientific Inc., Voioniview, Ontario<br />
V.K. Mafc<br />
CAWMET, Ottama, Ontatio<br />
ABSTRACT<br />
An ultrasonic immersion inspection system, intended for research<br />
work on large, heavy section we laments, has been built. During<br />
its design special attention was given to three factors. First<br />
was the capacity and rigidity of the mechanical scanner and<br />
immersion tank. Second was the importance of flexibility in the<br />
data acquisition subsystem, to support a number of different data<br />
collection modes. Finally, there was a desire to provide the<br />
operator with a variety of display methods for test data.<br />
The resulting system provides for simultaneous collection of up<br />
to 8 channels of ultrasonic data, and allows digitization and<br />
storage of ultrasonic waveforms. Processing of the collected<br />
data is performed by a general purpose microcomputer, with test<br />
results displayed using interactive 3-d'' n ension colour graphics.<br />
It is expected that future work will involve the addition to the<br />
system of more advanced algorithms for the location and sizing of<br />
weld defects.
1. INTRODUCTI<strong>ON</strong><br />
- 67 -<br />
Because of increased activity in the offshore exploration and<br />
petro-chemical fields, a need has arisen within the research<br />
community for a means of performing ultrasonic immersion<br />
inspection on large, heavy section weldments.<br />
Techno Scientific Inc. was contacted by the Physical Metallurgy<br />
Research Laboratories (PMRL), at the Canadian Centre for Mineral<br />
and Energy Technology (CANMET) to provide an inspection system<br />
which would be suitable for this type of work. It was specified<br />
that the system must:<br />
i) include an immersion scanner having a capacity of<br />
roughly 6 feet by k feet by 3 feet deep;<br />
ii) support simultaneous collection and processing of data<br />
from up to 8 transducers;<br />
iii) allow long-term storage of test data with later<br />
retrieval for processing and display;<br />
iv) provide a level of flexibility in data collection and<br />
processing methods appropriate to a research "tool".<br />
The last point should be emphasized. While many automated<br />
ultrasonic inspection systems have been described in the<br />
literature (1-7), all have been designed for a particular type of<br />
inspection, such as that of pipelines or composite panels, with<br />
the display of test results limited to 1 or 2 standard modes.<br />
2. SYSTEM DESCRIPTI<strong>ON</strong><br />
The system can be divided into k major modules or subsystems:<br />
î) a scanning bridge and immersion tank,<br />
ii) a scanner controller,<br />
iii) an ultrasonic signal processing subsystem,<br />
iv) a data acquisition, storage, and display computer.<br />
Two separate computer systems are included. The first, referred<br />
to as the "host", provides the basic processing power and data<br />
storage for the system, as well as coordinating the operations of<br />
the the various subsystems. The second, or "slave" system,<br />
includes a programmable scanner contr . 1er (which drives the<br />
scanning bridge motors) and the ultrasonic signal processing<br />
hardware. Figure 1 is an overall block diagram of the system,<br />
showing the division between the host and slave units.
- 68 -<br />
2.1 Scanning Bridge and Immersion Tank<br />
The 3-a*es scanning bridge provides control over the movement of<br />
the transducer(s) within a volume of 72" (183 cm) by 48" (122 cm)<br />
by 40" (1O2 cm), corresponding to the size of the stainless—steel<br />
immersion tank. Each axis is controlled by a leadscrew driven by<br />
a 200-step-per-revolution stepper motor, and has a linear<br />
resolution of 0.0005" (0.01 mm). The scanning bridge and<br />
immersion tank are shown in Figure 2.<br />
Also included as part of this subsystem is a stainless—steel,<br />
stepper-motor-driven, drop-in turntable unit for the inspection<br />
of cylindrical testpieces (see Figure 3)«<br />
2.2 Scanner Controller<br />
The microprocessor-based scanner controller converts high-level<br />
commands from the host, such as "move" and "scar,", to the lowlevel<br />
control functions required by the stepper motors driving<br />
the scanning bridge and turntable. This frees the host computer<br />
to perform data acquisition and processing.<br />
Any two of the four possible axes X, Y, Z or T (for turntable)<br />
may be selected for a scan, with the step size and velocity<br />
specified independently for each. While the scan pattern used is<br />
fixed, the size of the scan and the sampling density i.e., the<br />
distance between ultrasonic samples along the scan axis, are<br />
programmable.<br />
Although normally controlled by the host computer, the scanner<br />
controller may also be used as a stand-alone unit by connecting a<br />
computer terminal to one of its RS-232 serial ports and entering<br />
commands directly to initialize parameters and begin a scan.<br />
2.3 Ultrasonic Subsystem<br />
Figure 4 shows a block diagram of the ultrasonics subsystem. It<br />
can utilize up to 8 transducers in transceive mode, or 8 pairs in<br />
transmit-receive mode. Under the control of the slave processor,<br />
each transmit transducer is pulsed in sequence while an eightchannel<br />
time-division multiplexer directs the received echo<br />
through the signal processing circuitry. Either a linear or a<br />
logarithmic amplifier is used to increase the amplitude of the<br />
ultrasonic signal, while an adjustable attenuator is available to<br />
provide control of the signal level in discrete logarithmic<br />
steps. Separate gating is supplied for each of the 8 ultrasonic<br />
channels, permitting independent control of the gate delay and<br />
width for each transducer. Additional signal processing can be
performed by a video detector stage.<br />
Peak detector and x-y display modules art also included to allow<br />
the ultrasonics subsystem to be used, along with the scan<br />
controller, in a stand-alone mode. During scanning, the x-y<br />
display interface combines the output from the peak detector with<br />
position information from the scanner controller to produce realtime<br />
modified c-scan images on the built-in crt display or on an<br />
external x-y chart recorder.<br />
Outputs are provided from each stage to allow viewing on an<br />
oscilloscope or digitization by the host computer.<br />
2.h Host Computer<br />
The host computer is a Multibus-based system utilizing an Intel<br />
8086 microprocessor and 8087 math co-processor, 768 kilobytes of<br />
program memory, and both fixed and floppy-based mass storage. A<br />
digital transient recorder (DTR) installed on the Multibus acts<br />
as a high-speed analog-to-digital (a/d) converter, permitting the<br />
host to digitize and store ultrasonic signal data. The DTR can<br />
digitize arbitrary signals at rates up to 20HHz, while repetitive<br />
waveforms may be effectively sampled at up to 160 MHz.<br />
A colour graphics module, consisting of 3 additional Multibus<br />
boards, provides a resolution of 1024 by 769 points. Up to 16<br />
colours can be displayed simultaneously, chosen from a pallette<br />
of 4096. Support for 3~dimensional graphics is provided by an<br />
ACM-SIGGRAPH Core library of subroutines (8).<br />
3. SYSTEM OPERATI<strong>ON</strong><br />
Initially, the host prompts the user for all the necessary<br />
scanning parameters: the axes to be used, the scan dimensions and<br />
velocity, the sample interval, and the number of transducers.<br />
These are then sent to the slave, followed by a signal for<br />
scanning to begin.<br />
During an inspection, the slave signals the host computer<br />
whenever the scanning bridge reaches a position at which data is<br />
to be acquired. Then, for each of the transducers in use, the<br />
slave enables the appropriate ultrasonic channel while the host<br />
triggers the puiser and the DTR. Once triggered, the DTR<br />
collects 4096 amplitude data points, each with a resolution of 1<br />
part in 256. While these may be stored directly for later<br />
analysis, additional peak detection processing is usually<br />
performed to reduce the stored data to a manageable amount.<br />
If time-of-f1ight data is being collected, the host extracts from
- 70 -<br />
each waveform digitization the time position of the two largest<br />
peaks, assumed to represent a front reference surface, and a flaw<br />
or back reference surface. If the user has requested that peak<br />
amplitude data be collected, it is assumed that he has gated out<br />
the reference surfaces and so a single value corresponding to the<br />
amplitude of the highest peak in the data is stored instead.<br />
In some situations it may be useful to combine the 2 methods of<br />
data acquistion, by collecting both amplitude and time-of-f1ight<br />
values for each peak. This is done, however, at the cost of<br />
increased data storage space requirements.<br />
Data is stored on disk, from which it can be retrieved for<br />
further processing and display. There are k types of displays<br />
that the system can currently produce:<br />
i) rf or video waveforms (A-scans),<br />
ii) 2-dimensional sections through 3~dimensional arrays of<br />
data (including B-scans and C-scans),<br />
iii) modified C-scans<br />
iv) true 3-dimensional displays created by converting timeof-flight<br />
data into linear measurements.<br />
3.1 Waveform Display<br />
Figure 5 shows a display of the 4096 data points collected from<br />
an rf signal which had been sampled at 20 Mhz. This type of<br />
display is most useful for retaining a representative waveform<br />
from a scan for the evaluation of noise levels, signal strength,<br />
etc. at some later time.<br />
3.2 2-Dimensional Sections<br />
Amplitude or time-of-f1ight data collected from a 2-dimensional<br />
rectilinear scan is generally stored, either implicitly or<br />
explicitly, in a 3~ d imensional array as a function of scanning<br />
bridge position. The common B- and C-scans can be thought of as<br />
displaying subsets of this array in the form of sections or<br />
slices normal to one of the 3 dimensions. For example, the Cscan<br />
of Figure 6 is simply a slice made normal to the "amplitude<br />
axis" in a peak amplitude data array. Similarly, a B-scan<br />
results from making a slice through a time-of-f1ight data array,<br />
normal to the scan increment axis.<br />
However the display subsystem is not limited to standard B- and<br />
C-scan modes. For any given data array, 3 different types of<br />
sections can be displayed, with the thickness and location of<br />
each under user control.
3.3 Modified C-Scans<br />
- 71 -<br />
If a series of amplitude profiles are drawn on the same page or<br />
screen, each offset from the previous one by a small horizontal<br />
and vertical displacement, the effect is much like that of a 3dimensional<br />
projection. This method of display is known as<br />
modified C-scan, and can be thought of as all the possible Cscans<br />
of the testpiece, stacked one on top of another in order of<br />
increasing amplitude threshold.<br />
Beam profiles, used to aid in the characterization of ultrasonic<br />
transducers, are a common form of modified C—scan display.<br />
Figure 7 is an example, showing the echo response of a 3.5 MHz<br />
transducer to a 2 mm diameter target as a function of their<br />
relative positions. It shows graphically the ultrasonic beam<br />
strength of a transducer at various points within it's field.<br />
3-Dimensional Time-of-F1ight Displays<br />
Since the tirne-of-f1ight from a transducer to a reflector is<br />
proportional to their separation, 3-dimensional reflector<br />
coordinates can be calculated for a scan knowing only the<br />
inspection geometry and the measured time-of-f1ight values.<br />
These can then be displayed in 3-dimensional form using standard<br />
perspective or parallel projections (9) as supplied by the Core<br />
library (10). Figure 8 shows time-of-f1ight data for the<br />
testpiece of Section 3.2 plotted in this way. By combining the<br />
information from a complete set of B-scans into one image, it<br />
shows not only the extent of defects, but also allows some<br />
estimation of their depth.<br />
Because of the general nature of the 3~dimensional projections<br />
used, it is possible to view the object from any angle or<br />
distance.<br />
h. DISCUSSI<strong>ON</strong><br />
The most important criterion for the system, as well as the most<br />
difficult one to satisfy, was that of flexibility. It has been<br />
achieved, we feel, through the use of a modular ultrasonic<br />
subsystem and by the incorporation of the DTR into the system.<br />
In particular, the ability of the DTR to digitize a received<br />
waveform allows almost any desired numerical algorithm to be<br />
applied to the data, while the storage of results as numeric data<br />
files simplifies the data processing task. The softwarecontrolled<br />
display subsystem also aids flexibility,<br />
by simplifying modification and improvement ot' the display routines).
- 72 -<br />
Further development of the system will proceed along 2 divergent<br />
paths. On one hand, there is a desire to build a production<br />
version of the current research system. In this case flexibility<br />
can be sacrificed, to some extent, for the sake of reducing<br />
costs. Two changes likely to be made will be the use of a<br />
somewhat lower resolution display, and the replacement of the DTR<br />
with a peak detector and a hardware circuit for time-of-f1ight<br />
measurements.<br />
The second development path will see the implementation of more<br />
sophisticated defect sizing and location techniques. Since the<br />
system already offers the capability to digitize rf echo signals,<br />
no specialized hardware should be necessary in order to<br />
experiment with the use of methods such as amplitude-time locus<br />
curves (ALOK) and the synthetic aperature focussing technique<br />
(SAFT), among others (11).<br />
ACKNOWLEDGEMENTS<br />
The authors wish to thank the staff of Physical Metallurgy<br />
Research Laboratories, CANMET and in particular Mr. V. Caron and<br />
Dr. J. T. McGrath for their numerous helpful suggestions,<br />
continued interest in the project, and a number of special<br />
references.
REFERENCES<br />
- 73 -<br />
1. Udagawa, T., et al. "Automatic Ultrasonic Testing and<br />
Recording System for Welds Utilizing a Microcomputer,"<br />
Materials Evaluation (March 1982), pp. 305-311.<br />
2. Woodmansee, W.E. "An Interactive Graphics System Developed<br />
for NDT," Materials Evaluation (March 1980), pp. 33-36.<br />
3. Nielsen, N. "P-Scan System for Ultrasonic Weld<br />
Inspection," Proc. 3rd Int. Conf. NDE in Nuclear Industry,<br />
(Salt Lake City, 1980), pp. 171-195.<br />
4. Coote, R.I., et al. "Ultrasonic Inspection of Girth<br />
Welds," Proc. Int. Conf. Pipeline Inspection, CANMET<br />
ERP/PMRL 83-68 (OP-J), pp. 263-282.<br />
5. de Raad, J.A., et al. "Mechanical Ultrasonic Test<br />
Systems for Pipeline Welds," ibid., pp. 283-303.<br />
6. "Improved Ultrasonic System for Nondestructive<br />
Evaluation of Composite Materials," NTIAC Newsletter,<br />
vol.10, no.2, pp. 1-2.<br />
7. A. Singh, et al. "Automated Inspection of Corroded Steel<br />
Structures," Materials Evaluation (April 1983), pp. 568-570.<br />
8. Newman, W.M., Van Dam, A. "Recent Efforts Towards Graphics<br />
Standardization," Computing Surveys, vol.10, no.A, pp. 365-<br />
380.<br />
9. Foley, J., Van Dam, A. "Viewing in Three Dimensions,"<br />
Fundamentals of Interactive Computer Graphics. Addison-<br />
Wesley, 1982, pp 267-316.<br />
10. Bergeron, R., Bono, P., Foley, J. "Graphics Programming<br />
Using the Core System," Computing Surveys, vol.10, no.k, pp.<br />
389-W3.<br />
11. Schmitz, V., et al. "Improved Methods for ultrasonic Defect<br />
Classification Reconstruction and Reliability," Proc. Int.<br />
Conf. Quantitative NDE in the Nuclear Industry (San Diego,<br />
1982), pp. 258-264.
TRANSDUCER SELECT/<br />
PULSE C<strong>ON</strong>TROL<br />
- 74 -<br />
Figure 1 — Block diagram of the inspection system, showing<br />
division between host and slave units.<br />
TRANSDUCER 1<br />
TRANSDUCER 8<br />
PULSES<br />
FROM SCANNER<br />
. C<strong>ON</strong>TROLLER
Figure 2 — Overall view of the immersion tank<br />
Figure 3 — Drop-in turntable unit, with immersion tank and slave<br />
system in background
SCAIIIIING<br />
BRIDGE<br />
COLOUR<br />
GRAPHICS<br />
DISPLAY<br />
HA S S<br />
STORAGE<br />
SCANNER<br />
C<strong>ON</strong>TROLLER<br />
HOST<br />
COMPUTER<br />
OPERATOR<br />
TERMINAL<br />
- 76 -<br />
OPTI<strong>ON</strong>AL<br />
X-Y<br />
PLOTTER<br />
X-Y CRT<br />
DISPLAY<br />
ULTRAS<strong>ON</strong>IC<br />
SUBSYSTEM<br />
DIGITIZER<br />
PRINTER<br />
Figure U — Block diagram of the ultrasonic subsystem<br />
SLAVE<br />
SYSTEM<br />
fr<br />
uHOST<br />
SYSTEM
- 77 -<br />
Figure 5 — A096 point rf waveform as displayed on colour monitor<br />
Figure 6 — C-scan image of test block
Figure J — Beam profile of 3.5 MHz transducer<br />
Figure 8 — 3-dimensional perspective image of test block using<br />
t iiTie-of-f 1 ight
- 79 -<br />
THE INFLUENCE OF STRESS <strong>ON</strong> THE INSPECTI<strong>ON</strong> OF STEEL WITH PARTICULAR<br />
REFERENCE TO GAS PIPELINES<br />
V.L. Athenian, U.C. Jil&i and C. Wzlbouin<br />
Q.ueew'4 Un.lve.mIty, Kingiton, Ontanio<br />
*Research supported in part by the Department of Energy, Mines and Resources<br />
and the National Research Council (IMRI).<br />
ABSTRACT<br />
It is commonly recognized that stress induced permeability changes are a factor<br />
capable of giving signals or spuria in magnetic or electromagnetic non destructive<br />
inspection of steel. In fact the situation is more complex because ferromagnetic<br />
materials exhibit hysteresis with the result that their magnetic<br />
behaviour is dependent not only on the ambient stress and magnetic fields but<br />
also on both the magnetic and stress histories.<br />
Although stress effects usually complicate N.D.I, they can also be exploited,<br />
for example, far side corrosion signals can be enhanced by the application of<br />
stress which is concentrated in the regions of wall thinning. Magnetometer<br />
surveys of pipelines also show stress anomalies and changes in stress can be<br />
detected. The technique appears promising for monitoring stresses in other<br />
high performance steel structures such as offshore drilling rigs.<br />
Introduction<br />
It is well known that stress alters the magnetic behaviour of ferromagnetic<br />
materials and that changes in permeability give signals or spuria in magnetic<br />
or electromagnetic non destructive testing of steel. Generally permeability<br />
variations are an unwanted complication, particularly in the case of eddy<br />
current inspection, but they can be exploited when properly understood. The<br />
effect of stress on the magnetic behaviour of ferromagnetic materials is more<br />
complex than had been generally recognized until recently (1) because ferromagnetic<br />
materials are hysteretic with respect to both magnetic field and<br />
stress. Their magnetic behaviour therefore depends not only on the ambient<br />
stress and magnetic fields but also on both the magnetic and stress histories.<br />
It is possible that magnetic monitoring may therefore t' used to indicate<br />
ambient stress and stress history.<br />
Figure 1 shows as an example a log of a magnetometer survey above a gas pipeline.<br />
There are large fluctuations in the magnetic fields and some of these<br />
are attributed to the stresses introduced in the line at cold bends made in<br />
the field in order to follow undulating topography. Principally, however<br />
the fluctuations are due to the individual joints from which the line is<br />
constructed behaving as randomly orientated individual bar magnets as shown in<br />
Figure 2. The effect of stress introduced at cold bend deformations is to add
- 80 -<br />
an additional reverse magnet to the pipe bar magnet to give the magnetic<br />
profiles shown in Figure 3.<br />
These figures show that stress in ferromagnetic structures can produce significant<br />
but complex magnetic effects. : now review the basic results of the<br />
effects of uniform stress on simple sarii£-\es of ferromagnetic material such as<br />
line pipe steel.<br />
Effects of Constant Stress<br />
Figure 4 shows hysteresis loops for simple samples of line pipe steel under<br />
constant tension or compression. The effect of stress is to give a slight<br />
tilt to the hysteresis loops. At low field tension tends to increase the<br />
magnetisation, compression to decrease it although both tend to decrease it<br />
at high field. At first this seems to be explicable by applying Le Chatelier"s<br />
principle (which states that the application of a constraint to a system in<br />
equilibrium tends to shift the position of equilibrium to reduce the applied<br />
constraint) to the magnetostrictive effect (the change in length on magnetisation)<br />
. Figure 5 shows typical magnetostrictive behaviour for iron. The<br />
magnetostrictive coefficient (strain) is dependent on both field and applied<br />
stress. A positive magnetostrictive coefficient i.e. expansion on magnetisation,<br />
would suggest that tensile strain would cause an increase in<br />
magnetisation in order to reduce the applied stress and this seems to be in<br />
good accord with Figure 4. unfortunately the results of changing the stress<br />
in a constant field are more complex.<br />
Effects of Changing Stress in a Constant Field<br />
Figure 6 shows the changes in magnetisation occur ing from the same initial<br />
conditic.is during a tension cycle and a compression cycle under constant<br />
applied field. Both of these cycles result in increased magnetisation and<br />
are therefore contrary to predictions based on Le Chateliei's principle.<br />
Figure 7 shows the changes in magnetisation occuring as a result of similar<br />
stress cycles applied at different fields around a hysteresis loop. The<br />
results depend on the initial conditions and are due to a reduction in the<br />
magnatic hysteresis.<br />
Magnetic hysteresis occurs because magnetic domains are not free to move to<br />
their equilibrium magnetisated but are pinned (2). The effect of stress is<br />
to reduce this domain pinning and to let the magnetisation approach the true<br />
equilibrium (anhysteretic) value. The dotted line in Figure 7 shows the<br />
anhysteretic magnetisation curve, explained in Figure 8. The stress induced<br />
shift towards the anhysteretic magnetisation is further exemplified in<br />
Figure 9 which shows the changes in magnetisation induced by similar stress<br />
cycles. For cycles originating at points on the initial magnetisation curve<br />
the changes appear to be related to .he differences between the initial and<br />
anhysteretic magnetisation curves and this is confirmed by the very small<br />
changes resulting from cycles originating on the anhysteretic curve itself.
Irreversible and Reversible Changes<br />
- 81 -<br />
Le Chatelier's principle cannot be directly applied to ferromagnetic mate-ials<br />
since they exhibit hysteresis and are therefore not in true reversible<br />
equilibrium. The first effect of stress applied to a ferromagnet is to cause<br />
a reduction in the hysteresis, i.e. a shift toward the anhysterstic magnetisation.<br />
This effect is irreversible and is generally the dominant effect during<br />
the initial application of stress. The second effect is that the anhysteretic<br />
curve is itself stress dependent (3) as shown in Figure 9. This effect is<br />
essentially reversible. It could therefore be correlated with the magnetostrictive<br />
effect except that magnetostrictive coefficients are generally<br />
measured along the initial magnetisation curve which is irreversible and are<br />
therefore not directly applicable.<br />
The two effects appear to be approximately addative. In steel, the reduction<br />
in hysteresis is the dominant effect during the initial stressing but during<br />
subsequent stress cycles the reversible shift of the anhysteretic becomes<br />
relatively more significant.<br />
In stress monitoring applications both these effects may need to be considered.<br />
Stress Enhancement of Far Side Corrosion Signals<br />
A simple example showing the exploitation of stress effects is illustrated by<br />
Figure 10. The signal obtained from magnetic flux leakage detectors are<br />
greater for near side defects such as corrosion pits than for far side defects<br />
because the anomalous leakage fluxes are concentrated near the defect. If<br />
however a gas pipeline with external corrosion is stressed by the gas pressure<br />
external corrosion pitting will cause anomalous highly stressed regions on<br />
the inside of the pipewall. These may be detected using active or residual<br />
leakage flux techniques and used to enhance the far side corrosion signals<br />
as shown. The optimum procedure needs rather careful consideration as<br />
discussed below.<br />
Bending Stress Monitoring<br />
We have built (4,5) a laboratory scale demonstration of pipe bending stress<br />
monitoring using two 3.4 m long 11 cm diameter pipes sprung together. The<br />
apparatus is shown in Figure 11. Changes in magnetisation as indicated by<br />
the peak to peak amplitude of the external radial field profile logged with<br />
a fluxgate magnetometer one pipe diameter from the wall, are shown during<br />
two stress cycles in Figure 12. There are large irreversible changes occuring<br />
during the initial stressing after magnetisation and smaller reversible<br />
changes occuring thereafter. Figure 13 shows similar changes in magnetisation<br />
occuring during five bending stress cycles followed by remagnetisation and<br />
further stress cycles. The processes are shown conceptually on an M-H<br />
diagram in Figure 14. The effect of remagnetisation is basically to
- 82 -<br />
reinitialise the stress dependence so that the next application of stress<br />
will again cause relatively large irreversible changes in magnetisation.<br />
It should be noted that the magnetisation changes occuring in a complex case<br />
such as a non uniformly magnetised, non uniformly stressed pipe, indicated by<br />
external leakage field measurements as shown in Figure 13 are well understood<br />
in terms of the M-H diagram of Figure 14 even though it derived from the behaviour<br />
of a small sample with uniform magnetisation and stress.<br />
The technique shows promise for monitoring the maximum stress that a pipeline<br />
has been subjected to since last magnetisation (usually by magnetic inspection<br />
pigging). It might therefore be used to detect dangerous anomalous bending<br />
stresses due to foundation subsidence in pipelines constructed over permafrost<br />
or river crossings etc. Other applications to high performance steel structures<br />
not subject to periodic remagnetisation such as offshore drilling platforms<br />
may be simpler, particularly for statically loaded structures where stresses<br />
increase steadily without cyclic variations.<br />
Conclusions<br />
Stress has long been known to have major effects on ferromagnetic properties.<br />
For many N.D.T. techniques this is an undesireable complication although it<br />
can also be exploited both to enhance defect signals and as a potential<br />
technique for monitoring stress. This requires some knowledge of the basic<br />
effects of stress on a ferromaghet and until recently the situation has been<br />
confusing. It is now clear that stress causes both reversible and irreversible<br />
changes in magnetisation and that these are not directly related to magnetostriction<br />
as had generally been supposed. The irreversible changes, which are<br />
relatively large and occur mainly during the initial application of stress,<br />
are due to overcoming magnetic domain pinning and thereby reducing the<br />
hysteresis. The reversible effects are related to the stress dependence of<br />
the reversible (anhysteretic) magnetisation. We hope that with this recent<br />
improvement in understanding it will be possible to devise and develop<br />
further techniques for exploiting the effects of stress on ferromagnetic<br />
behaviour to augment current N.D.T. methods.<br />
References<br />
(1) D.L. Atherton and D.C. Jiles, 'Effects of stress on the magnetisation of<br />
steel', IEEE Trans, on Magnetics, Vol MAG-19, pp. 2021-2023 Sept. '83.<br />
(2) D.C. Jiles and D.L. Atherton, "Theory of ferromagnetic hysteresis',<br />
J. Appl. Phys., Vol. 55 pp. 2115-2120 March '84.<br />
(3) L. Dobranski, "The effects of uniaxial isostress on the anhysteretic<br />
and initial magnetisations of various pipeline steels", M.Sc. (Eng.)<br />
Thesis, Queen's University at Kingston, April '84.<br />
(4) D.L. Atherton, L.W. Coathup, D.C. Jiles, L. Longo, C. Welbourn and<br />
A. Teitsma, 'Stress induced magnetisation changes of steel pipes -<br />
Laboratory tests', IEEE Trans, on Magnetics Vol MAG-19, pp. 1564-1568<br />
July '83.<br />
(5) D.L. Atherton, C. Welbourn, D.C. Jiles, L. Reynolds and J. Scott-Thomas,<br />
'Stress induced magnetisation changes of steel pipes - Laboratory tests.<br />
Part II', IEEE Trans, on Magnetics Vol MAF-20, November "84.
FIELD<br />
VARIATI<strong>ON</strong> IN<br />
KILOGAMMA<br />
WELD WEcD<br />
FACTORY<br />
GIRTH WELD<br />
1°<br />
SAG<br />
BEND<br />
7Y<br />
LEVEL OF EARTHS FIELD COMP<strong>ON</strong>ENT<br />
(APPROX. 60 KILOGAMMA)<br />
Figure 1: A fluxgate magnetometer traverse made 1.2 m above a 200 m long string<br />
of 1.2 m diameter 13 mm wall pipeline showing fluctuations in the<br />
vertical component of magnetic field. These are due to the constituent<br />
20 m jointe acting as randomly aligned bar magnets superimposed<br />
on which are the effects of cold bends.<br />
I<br />
oo
BAR<br />
MAGNET<br />
IDEAL<br />
FIELDS<br />
MEASURED<br />
FIELDS<br />
N<br />
Figure 2: Pipelines are constructed typically from 7 m joints welded together.<br />
Each individual joint behaves rather like a bar magnet. There are<br />
north and south magnetic poles located a little way in from the<br />
ends. The external magnetic field lines run from north to south.<br />
If a magnetometer scan is made parallel to the pipe axis and at a<br />
constant distance outside it this gives radial and axial field logs<br />
as shown ideally and as measured above a real pipe.
IDEAL<br />
BEND<br />
FIELDS<br />
MEASURED<br />
BEND<br />
FIELDS<br />
5 ky<br />
Sky<br />
- 85 -<br />
Figure 3: Magnetometer radial and axial magnetic field component profiles above<br />
a real pipeline joint before and after cold bending show almost ideal<br />
field patterns after. The more complex patterns before arise<br />
because this joint consists of two joints factory welded together<br />
and there are small relict poles near the center of the double<br />
joint. The stress induced changes are a significant percentage of<br />
the earth's field itself and much greater than the natural variations<br />
in the earth's field which are normally less than 100 gamma in<br />
a day.
- 86 -<br />
Figure 4: Hysteresis loops for a small sample of pipeline steel measured<br />
under constant applied stresses.
EXTENSI<strong>ON</strong><br />
C<strong>ON</strong>TRACTI<strong>ON</strong><br />
- 87 -<br />
COMPRESSED<br />
FIELD (H)<br />
UNSTRESSED<br />
TENSI<strong>ON</strong>ED<br />
Figure 5: Magnetostriction is the fractional change in length, the strain of<br />
a ferromagnetic material when magnetised. Initially iron expands<br />
on magnetisation but at higher fields it begins to contract, as<br />
shown. The effect is also stress dependent. Compression tends to<br />
enhance the region of expansion, tension to reduce it. The inverse<br />
effect, the change of magnetisation with stress variation, is the<br />
magnetomechanical effect which was expected to have a reciprocal<br />
behaviour.
-150<br />
H= 1.6 kA/m<br />
COMPRESSI<strong>ON</strong><br />
-100 -50<br />
0 03<br />
à B<br />
TEH.<br />
0 02 - •<br />
0 01--<br />
H= 1.6 kA/rn<br />
TENSI<strong>ON</strong><br />
50 100<br />
Figure 6: The changes in magnetisation, AB, during a compression cycle and a<br />
tension cycle both starting from the same initial conditions.<br />
a MPQ<br />
150<br />
00<br />
CO
- 89 -<br />
Figure 7: The effects of identical stress cycles at different points around<br />
a magnetic hysteresis loop. The results appear to depend on the<br />
initial conditions. The dotted line is the anhysteretic magnetisation<br />
curve described in Figure 8.
- 90 -<br />
MAGNETISATI<strong>ON</strong><br />
ANHYSTERETÎC<br />
INITIAL<br />
FIELD<br />
Figure 8: Magnetisation curves. The initial or normal magnetisation curve is<br />
obtained with an unmagnetised or demagnetised sample as the field<br />
is increased. If the field is then decreased, reversed and cycled<br />
a repeatable but irreversible path is followed. The area inside<br />
this hysteresis loop gives the irreversible energy requirement or<br />
loss. The anhysteretic curve is reversible and is the magnetisation<br />
curve which would be obtained if hysteresis effects were<br />
suppressed. Points on it are obtained, rather similarly to demagnetisation,<br />
by using an alternating field of steadily diminishing<br />
amplitude but superimposed on a steady bias field.
M<br />
Stress,<br />
MPa<br />
•*-200(cr)<br />
0(H)<br />
H, kA/m<br />
Figure 9: The anhysteretic surface as a function of applied field at constant<br />
stress for a sample of line pipe steel.
(b) 19 mm<br />
diameter<br />
o<br />
Hard<br />
spot<br />
19mm<br />
diameter<br />
o<br />
Interior<br />
- 92 -<br />
19mm 50 mm long<br />
diameter % 12 mm wide<br />
12 mm long<br />
x 50 mm wide<br />
19 mm<br />
diameter<br />
o<br />
V7A///7- LJ -(/ r/rV// / WT////77<br />
-500 mm-<br />
Exterior<br />
Figure 10: (a) Hall probe scan of the radial residual field leakage flux<br />
above a short previously magnetised piece of pipe with the simulated<br />
defects shown in (b). (c) The same, but with reduced scale, after<br />
a stress simulating the effects of line pressure has been applied.<br />
There are now clearer superimposed signals from the anomalously<br />
stressed regions due to pipewall thinning. These 'stress shadows'<br />
tend to preferentially enhance the signals from far side corrosion<br />
since their anomalous stresses are on the near side to the detector.
Figure 11: Laboratory scale monitoring of stress near pipes stressed by bending<br />
or by internal pressure. The pipes are supported by a central<br />
saddle and tied together at one end. They are bent by jacking<br />
the other ends together. Bending stresses are determined by strain<br />
gauges bonded to the peak stress point of the pipe and are verified<br />
by stress calculations from both the measured deflection and the<br />
required force. Internal pressure is provided hydraulically. A<br />
fluxgate magnetometer is being used to monitor the stress induced<br />
changes in the external field. This is a sensitive device used<br />
for geomagnetic prospecting, mine detection, etc. The pipes can<br />
be arranged at any horizontal and vertical angle to the earth's<br />
magnetic field in order to investigate its effect.
Q.<br />
I<br />
CO<br />
zL 2<br />
(Ky)<br />
TYPICAL<br />
UNCERTAINTY<br />
0 20 40 60 80<br />
MAXIMUM BENDING STRESS (MPa)<br />
Figure 12: Peak to peak values of changes in radial fields measured at one pipe<br />
diameter as a function of maximum bending stress during two bending<br />
cycles of a demagnetised pipe.<br />
100 120<br />
i
(K/)<br />
-fr— BASELINE; MAGNETISATI<strong>ON</strong> DUE TO<br />
FIRST 3A CURRENT PULSE<br />
O — BEFORE SEC<strong>ON</strong>D 3A PULSE! ^O*'lNITIAL STRESSING<br />
"^^tf^NEXT 5 STRESS CYCLES<br />
— AFTER SEC<strong>ON</strong>D 3A PULSE: *D*'lNITIAL STRESSING<br />
20 40 60 80<br />
MAXIMUM BENDING STRESS (MPa)<br />
Figure 13: Peak to peak radial field changes as a function of maximum bending<br />
stress for a pipe magnetised by a 3A current pulse in a coaxially<br />
wound solenoidal coil then subjected to 5 bending stress cycles<br />
followed by a similar magnetising current pulse and a further<br />
stress cycle.<br />
STRESS CYCLE<br />
TYPICAL<br />
UNCERTAINTY<br />
100 110
- 96 -<br />
Figure 14: Conceptual M-H behaviour during the processes of Figure 12. Initial<br />
magnetisation 0-a-b. 1st stressing: b-c, remainder of 1st 5 cycles:<br />
c *•—^ d, remagnetisation: d-e-f, 6th stress cycle: f-g-h. The<br />
application of stress tends to cause an initial large irreversible<br />
shift in magnetisation towards the anhysteretic magnetisation, zero<br />
in this case because there is no applied field. During subsequent<br />
stress cycles there are smaller nearly reversible changes.<br />
Remagnetisation produces a slightly greater magnetisation than from<br />
the initial unmagnetised state and tends to reinitialise the stress<br />
situation.
- 97 -<br />
DEFECT CHARACTERIZATI<strong>ON</strong> AND SIZING IN PIPELINE WELDMENTS BY<br />
ANALYSIS OF ULTRAS<strong>ON</strong>IC ECHO FEATURES<br />
J. P. MonckaZin<br />
Encigij, Wines and Resources Canada<br />
Ottawa, Ontario<br />
J. F. Bus sie-in<br />
Nat ional Re seaicli Council o {
- 98 -<br />
characterization program based on these lines. Classification uses the<br />
algorithms based on a linear discriminant function, an empirical Bayesian<br />
classifier and a K-nearest neighbor method previously developed for acoustic<br />
emission by Tektrend Inc. In this paper, we report preliminary work made on<br />
this project, the application being considered at the beginning being the<br />
characterization and sizing of pipeline weld defects.<br />
EXPERIMENTAL<br />
Since the application considered first in this program is pipeline welds,<br />
samples with defects were prepared from 1/2 inch (12.5 mm) thick steel<br />
plates. Planar flaws were obtained by butt welding two plates laid side by<br />
side: one weld pass was made with insufficient penetration, then the plates<br />
were turned upside down and a second pass was made, leaving in the middle an<br />
unfused area, running the whole plate length. In this way, planar defects,<br />
which perfectly simulate internal lack of side wall fusion, were obtained.<br />
Long defects (10 inches or more) having uniform through-wall dimension of = 1,<br />
2 and 3 mm were obtained. A flaw of 1 mm or less is generally considered as<br />
benign whereas larger flaws are considered critical (1).<br />
Four plates (labelled #3, #4, #5 and #6) were made in this way and then cut in<br />
half. One half of each plate was machined to remove the weld crown, whereas<br />
the other half was left unaltered. Defect size was measured visually from<br />
both ends of the plates. The following results were found:<br />
Plate #3: 2.75 mm - 3 mm<br />
Plate #4: 2 mm - size not well defined from one sick?<br />
Plate #5: 3.5 mm - 3 mm<br />
Plate #6: 1 mm - 1 mm<br />
Experimental data were taken from these samples at several locations along the<br />
defect using the pulse-echo immersion technique and broad-band commercial<br />
transducers at 5 MHz. Both planar and focused transducers were used with<br />
incidence angles of 45°, 60° and 75° in steel. Transducer height was adjusted<br />
in order that the defect was in the far field region of the planar transducer<br />
or in the focal zone of the focused transducer. The position of the<br />
transducer along the direction perpendicular to the defect was determined by<br />
maximizing the echo height for the planar transducer and somehow by trying to<br />
observe the maximum number of peaks with the focused one. The signal echo was<br />
digitized using a digital oscilloscope and the data was then stored for<br />
further computer processing. So far data has only been taken from the plates<br />
with the crown removed.<br />
DISCUSSI<strong>ON</strong> OF OBSERVED ECHO FEATURES<br />
The data taken with the planar transducer at 45° and 75° do not present any<br />
readily obvious distinguishing features, except some higher frequency<br />
oscillation at = 7 MHz for the 3 mm defects at 45°. At t>0°, trailing pulses<br />
are observed. The transducer being well damped poor signal-to-noise ratio is<br />
encountered and diffracted edge waves as predicted by elementary theory could<br />
not be clearly observed: theory predicts for a vertical defect that two pulses
- 99 -<br />
separated in time by 2d cosr/vs should be observed where d is the defect<br />
size, r is the refraction angle in steel and vs is the shear velocity. This<br />
time interval amounts to about 1.4 \is, 1 ys and 0.5 ys for a 3 mm defect at<br />
45°, 60° and 75" refraction angle, respectively.<br />
With the focused transducer properly adjusted, a much higher signal-to-noise<br />
ratio was obtained and several pulses could be observed as shown in figs. 1, 2<br />
and 3 obtained at 60° with plate //3, plate ilk and plate //6, respectively. The<br />
first two pulses can be linked to shear waves diffracted by the edges and the<br />
observed time delay corresponds approximatively to that given by elementary<br />
theory (these pulses can be hardly separated for the 1 mm flaw sample). The<br />
other peaks which follow (one or two are observed) are linked to conversion<br />
into a compressional wave, to reflections of this wave by the surfaces of the<br />
plate and reconversion into a shear wave. This phenomenon was demonstrated by<br />
using a second compression transducer (planar) oriented normally to the plate<br />
and located right above the defect (see fig. 4). Using this transducer in a<br />
pitch-catch mode, the multiple reflections of the converted compressional wave<br />
can be seea as shown in fig. 5. The first pulse seen with this transducer is<br />
produced by a parasitic wave directly scattered from the first surface.<br />
Further evidence was obtained by milling a layer of constant thickness from<br />
the bottom of the plate. The periodic spacing observed with the compression<br />
transducer was observed to diminish by a time interval corresponding to the<br />
propagation of a compressional wave through the removed thickness. In the<br />
same way, the time interval between the two edge pulses and the third echo<br />
(also the fourth when observable) was decreased by the same amount. The<br />
converted wave was observed to be very strong at the 60° incidence angle and<br />
was generally very weak at 45° and 75°. This observation is consistent with<br />
the fact that 60° corresponds to 30° incidence on the defect surface which is<br />
close to the limit angle for longitudinal wave generation. It should be noted<br />
that in the experiments conducted so far, the weldment crowns were machined<br />
off. In the presence of the crowns, the use of the second transducer is not<br />
obviously possible and the crowns may diminish or enhance the conversion into<br />
shear waves observed in pulse-echo according to their shapes. Experiments are<br />
planned to observe the effect of the weld crowns.<br />
CLASSIFICATI<strong>ON</strong> BY PATTERN RECOGNITI<strong>ON</strong><br />
Using the classifier and the features extraction programs developed by<br />
Tektrend Inc., we have analyzed about 400 ultrasonic signals taken as<br />
mentioned above from the 4 samples, at 45°, 60° and 75 C , with a planar and<br />
with a focused transducer. The collected signals for each experimental setup<br />
(angle and transducer) were grouped according to their defect size and<br />
geometry, and they were further divided into two sets; one set of signals was<br />
used to train the classifier to recognize the signals based on their defect<br />
sizes and the other set was used to verify the performance of this<br />
classifier. The tests performed are indicated below with the recognition<br />
rates (expressed as percentages):
- 100 -<br />
1. Planar transducer:<br />
l.a. Signals from plate #3 and plate<br />
representative of a 3 mm flaw.<br />
lmm vs 2mm<br />
lmm vs 3mm<br />
2mm vs 3mm<br />
45 C<br />
91.67<br />
100.00<br />
75.00<br />
#5 are taken together as signals<br />
60 c<br />
83.33<br />
91.67<br />
91.67<br />
75°<br />
91.67<br />
91.67<br />
83.33<br />
l.b. The classifier was trained on signals from plate #3 and then tested<br />
with signals from plate #5.<br />
lmm vs 3mm<br />
2mm vs 3mm<br />
45° 60 c<br />
100.00<br />
100.00<br />
2. Focused transducer: (same tests as above)<br />
2.a Signals from plate #3 and plate #5<br />
representative of a 3mm flaw.<br />
lmm vs 2mm<br />
lmm vs 3mm<br />
2mm vs 3mm<br />
45°<br />
90.00<br />
93.75<br />
81.25<br />
80.00<br />
80.00<br />
75°<br />
80.00<br />
80.00<br />
are taken together as signals<br />
60 c<br />
80.00<br />
93.75<br />
75.00<br />
75 C<br />
85.00<br />
93.75<br />
81.25<br />
2.b The classifier was trained on signals from plate #3 and then tested<br />
with signals from plate #5.<br />
lmm vs 3mm<br />
2mm vs 3mm<br />
45° 60° 75°<br />
100.00<br />
78.57<br />
83.33<br />
83.33<br />
87.50<br />
71.43<br />
These classification results show that 100% recognition rates can be<br />
obtained in some cases. Also, they show that a resolution of lram in size<br />
which is required by fracture mechanics analysis is possible. Worse<br />
recognition rates are likely to be caused by a variation of the flaw size at<br />
the location where the data is taken. This problem will be resolved later<br />
when the samples will be sectioned. Also work is required to elucidate what<br />
are the most significant features for the planar transducer data as well as<br />
the focused one.
C<strong>ON</strong>CLUSI<strong>ON</strong> AND PERSPECTIVES<br />
- 101 -<br />
We have reported preliminary work, on sizing of planar weld defects by<br />
ultrasonic echo feature analysis. With proper ultrasonic excitation<br />
diffracted edge waves were observed and found to be easily interprétable for<br />
sizing. However, in many cases, the echo pattern is complicated by other<br />
phenomena. In particular, this work has shown that the flaw cannot be<br />
considered as isolated since interactions with the sample boundaries produce<br />
extraneous features which have to be taken into account. We have also shown<br />
that classification of planar flaws according to size can be accomplished with<br />
high recognition rates and resolutions of = 1 mm. The approach presented<br />
therefore has a good potential for practical applications such as<br />
distinguishing between benign and critical flaws in pipeline weldments.<br />
ACKNOWLEDGEMENTS<br />
This work has benefited from the technical assistance of Jean-Guy Allard and<br />
Claude Michel for making the samples and of René Héon for taking the data.<br />
Software development and classification work performed at Tektrend Inc. is<br />
supported by government contract it ISQ83-00025, jointly funded by CANMET,<br />
Energy, Mines and Resources Canada and IMRI, National Research Council of<br />
Canada.<br />
REFERENCES<br />
1. J.L. Rose, "A feature based ultrasonic system for reflector<br />
classification", in New Procedures in Nondestructive Testing<br />
(Proceedings), ed. by P. Höller, Springer-Verlag, Heidelberg, 1983,<br />
p. 239-250.<br />
2. M.F. Whalen and A.N. Mucciardl, "Application of Adaptive Learning<br />
Networks for the Characterization of two-dimensional and<br />
three-dimensional flaws in solids", Review of Progress in Quantitative<br />
NDE, 5th Annual report, p. 482-508, July 1980.<br />
3. M.F. Whalen and A.N. Mucciardi, "Inversion of physically recorded<br />
ultrasonic waveforms using adaptive learning network models trained on<br />
theoretical data", Review of Progress in Quantitative NDE, 4th Annual<br />
Report (July 77 to June 78), p. 341-367, January 1979.<br />
4. D.R. Hay and R.W.Y. Chan, "Identification of deformation mechanism by<br />
classification of acoustic emission signals", 14th Symposium on<br />
Nondestructive Evaluation, San Antonio, Texas, April 1983.<br />
5. R.P. Reed, H.I. MeHenry and M.B. Kasen, A Fracture-Mechanics Evaluation<br />
of Flaws in Pipeline Girth Welds, Welding Research council Bulletin #<br />
245.
Fig. 1<br />
1/JS<br />
- 102 -<br />
Echos observed from plate #3 (vertical planar defect, = 3 mm)<br />
refraction angle in the plate = 60°, 5 MHz focused transducer.<br />
1/iS<br />
Fig. 2 Echos observed from plate #4 (vertical planar defect = 2 mm),<br />
refraction angle in the plate = 60°, 5 MHz focused transducer.<br />
1//S<br />
Fig. 3 Echos observed from plate #6 (vertical planar detect = 1<br />
refraction angle in the plate - 60°, 5 MHz focused transducer.<br />
mm} t
Transducer<br />
(Pulse-echo mode)<br />
- 103 -<br />
Transducer<br />
Edge shear waves / Converted<br />
Back converted compression wave<br />
shear wave<br />
Fig. 4 Experimental set-up to observe diffracted and converted waves.<br />
Fig. 5 Oscilloscope traces observed with the transducers shown in figure<br />
4. Top: pulse-echo signal from angled transducer - (the first pulse<br />
is parasitic). Bottom: echos from transducer directly above flaw -<br />
the first pulse is a direct reflection from the top surface,<br />
following pulses are reflections of mode-converted wave inside the<br />
plate.
- 104 -<br />
CAPABILITIES/LIMITATI<strong>ON</strong>S OF FOCUSED ULTRAS<strong>ON</strong>IC BEAMS FOR SIZING<br />
PIPELINE GIRTH WELD DEFECTS<br />
C.A. Kittrwi, J.8. Hallztt, V.N. Sycko<br />
Atomic Energy o {
- 105 -<br />
orientation) it is possible to compare it against a "tolerable" defect as<br />
determined by fracture mechanics which could safely be left in the pipeline<br />
with minimum risk of failure. This permits a realistic decision to be made<br />
as to repairs and avoids unnecessary rework of welds.<br />
Although ultrasonic inspection is capable of detecting significant defects<br />
(many of which are missed by radiography), defect sizing is still a major<br />
concern. Using standard transducers it is difficult to determine the size<br />
of girth weld defects by standard mapping techniques(3). This arises<br />
because the defect (or characteristic) being sized is comparable to or<br />
smaller than the beam used for mapping. The tendency therefore is to<br />
profile the beam rather than size the defect. Size predictions for a<br />
defect would consequently vary depending on the size of the beam used.<br />
One solution is to use a focused beam. This can be achieved simply and<br />
economically by attaching an acoustic lens to the face of a standard<br />
transducer. A test program was sponsored by the Pipeline Research<br />
Committee of the American Gas Association to investigate the capability of<br />
focused ultrasonic beams in sizing girth weld defects.<br />
APPROACH<br />
The main objective of the program was to investigate the defect sizing<br />
accuracy of an inspection procedure using focused ultrasonic beams with<br />
standard mapping techniques presently in use throughout industry. A<br />
secondary objective was to evaluate the technique as applied to field<br />
usage. Accordingly, the following approach was chosen:<br />
i. Use standard ultrasonic transducers which are readily available<br />
commercially "off-the-shelf".<br />
ii. Use contact inspection (as opposed to immersion) requiring various<br />
lens/wedge/probe combinations.<br />
iii. Use a simple scanning device (e.g. linear potentiometers to monitor<br />
probe position) so as to de-emphasize the equipment and concentrate<br />
on the ultrasonic technique itself. (See Figure 1 for the inspection<br />
equipment used).<br />
Lens Design<br />
Theory has been developed to design acoustic lenses for focused beams in<br />
defect sizing (4,5). The fundamental equations of interest are:<br />
1 - _ak<br />
N<br />
!|k _ 0.635 1-^1+0.2128 l-^l I [1]<br />
(Deviation + 9%)*<br />
* Deviation of analytical approximation from the true relationship
- 106 -<br />
V = f* - 0.9<br />
ak<br />
N [2]<br />
h. /^k)7__i<br />
N " l N / \1+ I<br />
It ^[ 1 + (1 -^'<br />
(Deviation + 5%)<br />
(Deviation + 3%)<br />
(Deviation + 5%)<br />
Ultrasonic theory also shows that the beam diameter within the acoustic<br />
focal range can be given approximately by:<br />
where fa^ = acoustic focus (note f ,/N = focusing factor, Z)<br />
fopt = optical focus<br />
Lg = the 6 dB working range<br />
AL = the part of L distributed before f ak<br />
V = gain (increase in sensitivity)<br />
N = transducer nearfield length<br />
The theory is relatively easy to apply, and focused beams can be readily<br />
reproduced (6,7) to meet the various inspection requirements.<br />
Weld Specimens<br />
The pipe specimens were in a range of sizes (610 mm diameter x 6.7 mm wall,<br />
to 1067 mm diameter x 15 mm wall thickness) typical of what is being used<br />
in the field. They were fabricated using gas-metal-arc (one sample) and<br />
shielded-metal-arc (seven samples) welding processes with representative<br />
beveled joint specifications. Defects were manually introduced by the<br />
operator during the welding process and included: inadequate penetrations,<br />
incomplete root fusions, lack of sidewall fusions, internal undercuts, slag<br />
inclusions (wagon tracks) and root cracks. They ranged in size from over<br />
50% wall thickness (root cracks) to a few millimetres (slag inclusions).<br />
Radiographic inspection carried out prior to ultrasonic inspection detected<br />
all but one of the manufactured defects (incomplete root fusion), but could<br />
give little information concerning defect sizes.<br />
[3]
ACOUSTIC LENSES<br />
- 107 -<br />
Ten acoustic lenses were designed to cover the range of wall thicknesses<br />
and weld preparations suggested for inspection. Selection of defects for<br />
inclusion in the program finalized the inspection requirements (inspection<br />
angle, defect location), and four lens/wedge/transducer combinations were<br />
then chosen to be used. Each assembly was calibrated by signal amplitude<br />
and time—of-flight using a calibration block containing 1.5 ram diameter<br />
side drilled holes spaced 2.5 and 4.0 mm apart (Figure 2a). The high<br />
resolution capability of focused beams is readily capable of<br />
differentiating between such small targets (Figure 2c) in contrast to the<br />
broad beam results for a standard transducer (Figure 2b). The calibration<br />
data was used to determine the beam inspection angle and time-of-flight of<br />
sound in the lens/wedge assembly.<br />
DEFECT SIZING<br />
Each weld specimen was examined ultrasonically from both sides of the outer<br />
surface weld bead using a simple scanning mechanism (Figure 1). The<br />
pulse-echo technique, which is commonly used in the field, was used<br />
throughout in contact mode with a light oil as couplant. The inspection<br />
data was recorded as multiple A-scans. In this technique, the probe is<br />
incrementally moved axlally away from the weld bead. The gate of the<br />
ultrasonic instrument is swept across the CRT screen, and the signal<br />
amplitude versus time display (shown on the CRT) is plotted for each probe<br />
position (see Figure 3). As the path length to the defect increases, the<br />
time-of-flight increases and the defect response on the A-scan moves to the<br />
right on the time axis. Conversely, any extraneous signal (e.g. from the<br />
lens/wedge assembly) remains at constant time and is readily<br />
distinguishable on the multiple A-scan record. The through-wall depth of<br />
the defect can then be calculated based on the extent of the defect<br />
response along the time axis.<br />
Metallurgical Examination<br />
After ultrasonic examination, each specimen was broken open by cooling it<br />
in liquid nitrogen and using three point impact loading to fracture along<br />
the centre of the weld. Where samples did not break through the defect<br />
(primarily for the smaller buried defects) the sample was sectioned and<br />
metallographically prepared to reveal the defect size.<br />
RESULTS<br />
A typical inspection record (Figure 4a) and specimen cross-section (Figure<br />
4b) are shown for a planar root crack giving a comparison of predicted<br />
versus actual defect size. These results are representative, as in all<br />
cases size predictions were in good agreement with the actual defect<br />
depths. This excellent correlation is further illustrated in Figure 5<br />
where the majority of data points fall within a band +1.5 mm of the line as<br />
determined by regression analysis. The 95% confidence limits for the<br />
predicted mean flaw depths are shown. There was little distinction in<br />
sizing accuracy between volumetric and planar defects. The technique is<br />
conservative in that defects tend to be slightly oversized (average error<br />
was 0.37 mm).
DISCUSSI<strong>ON</strong><br />
- 108 -<br />
The results of this project illustrate that focused beams can be used<br />
successfully in conjunction with a simple mapping technique for defect<br />
sizing. Other, more sophisticated, techniques have shown comparable or<br />
better accuracy (3,8). The added complexity, however, does tend to isolate<br />
them to the laboratory with several years of development before being<br />
ready for field use, and require a higher qualification of inspectors than<br />
is presently employed for pipeline inspection.<br />
The project also restated standard problems encountered during contact<br />
inspection due to weld bead geometry. Although the leading edges of the<br />
wedge assemblies were ground off, the outer weld bead was sometimes still<br />
wide enough to prevent full coverage of a known defect. Depending on the<br />
nature of the defect and the working range limits of the lens, this could<br />
result in loss of sizing accuracy. However, by lens design for longer<br />
steel paths (i.e., in the region of 1 to 2 skips of the reflected beam,<br />
instead of 0 to 1 skip as used here) this can be readily overcome.<br />
Weld bead geometry can also confuse interpretation by introducing<br />
extraneous signals, or significantly altering the sound path length. When<br />
the ID weld bead is offset from the centre of the weld, sound may reflect<br />
off the bottom of the weld bead, instead of off the pipe ID, before<br />
reaching the target defect. In such instances an ID defect may appear to<br />
be alternately located at midwall or at the weld root, depending on the<br />
relative position of the transducer. Are there two defects located here,<br />
or one? Focused beam inspection is as vulnerable to operator<br />
interpretation in this area as any other technique which relies on<br />
time-of-flight information to perform its function.<br />
Transducer selection is also an area of vulnerability for this technique.<br />
The use of transducers with inappropriate beam characteristics would<br />
significantly degrade the outcome of any sizing attempt. Similarly,<br />
variations due to instrumentation effects (instrument/probe interaction)<br />
must be considered for optimum accuracy(6). At the extreme, the unfocused<br />
beam characteristics must be determined using the actual instrument/<br />
transducer combination intended for the inspection for input to the lens<br />
design process.<br />
Although the basic ultrasonic theory is well developed, it does not readily<br />
lend itself to overcoming the specific limitations outlined above. In<br />
response to such needs, computer modelling is presently being used as an<br />
aid to understanding the physical processes involved. One such example is<br />
based on solution of the first order equations of elasticity using finite<br />
difference methods to model sound wave propagation through solids(9). The<br />
results of such modelling (e.g. see Figure 6) will be used to develop<br />
guidelines for optimization of transducer selection and inspection<br />
procedure specification.<br />
C<strong>ON</strong>CLUSI<strong>ON</strong>S<br />
- Focused ultrasonic beams can be used in conjunction with simple mapping<br />
techniques used in standard practice to give accurate sizing of<br />
pipeline girth weld defects.
- 109 -<br />
- The technique can be applied without the use of sophisticated support<br />
equipment and is readily field implementable. Operator interpretation<br />
is minimized through the use of a multiple A-scan format for recording<br />
inspection data.<br />
Contact inspection of girth welds using focused beams is subject to the<br />
same problems with weld bead geometry encountered during standard<br />
ultrasonic inspections.<br />
REFERENCES<br />
1 COOTE, R.I., FINGERHUT, M.P., 'Ultrasonic Inspection of Girth Welds',<br />
Proceedings of the International Conference on Pipeline Inspection,<br />
Edmonton, June 1983, pp. 263-279.<br />
2 GLOVER, A.G., COOTE, R.I., PICK, R.J., 'Engineering Critical Assessment<br />
of Pipeline Girth Welds', Fitness for Purpose Validation of Welded<br />
Connections, Paper 30, Welding Institute U.K., London, 1981.<br />
3 GRUBER, G.J., HENDRIX, G.J., SCHICK, W.R., 'Characterization of Flaws<br />
in Piping Welds Using Satelite Pulses', Materials Evaluation, Volume<br />
42, April 1984, pp. 426-432.<br />
4 WUSTENBERG, H., KUTZNER, J., MOHRLE, W., 'Fokussierende Prufkopfe zur<br />
Verbesserung der Fehlergro enabschatzung bei der Ultraschallprüfung von<br />
dickwandigen Reaktorkomponenten', Materialpruf, Volume 18, number 5,<br />
May 1976, pp. 152-161.<br />
5 SCHLENGERMANN, U., 'Schallfeldausbildung bei ebenen Ultraschallquellen<br />
mit fokussierenden Linsen', Acustica, Volume 30, Issue 6, 1974, pp.<br />
291-300.<br />
6 KITTMER, CA., 'Acoustic Lenses - Focusing in on Defects', Proceedings<br />
of the International Conference on Pipeline Inspection, Edmonton, June<br />
1983, 161-178.<br />
7 LARSEN, R.E., 'Zone-Focused Search Units for Bore Ultrasonic<br />
Examination Series', Proceedings of the Twelfth Symposium on<br />
Nondestructive Evaluation, San Antonio, April 1979, pp. 207-210.<br />
8. MACECEK, M., 'Precise Ultrasonic Depth Measurements of Internal<br />
Undercut in Pipeline Girth Welds', Proceedings of the Conference on<br />
Pipeline and Energy Plant Piping, Calgary Canada, November 1980, pp.<br />
291-299.<br />
9. DUNCAN, D., 'Computer Simulation of Ultrasonic Testing', To be<br />
published in Proceedings of the 5th Canadian Conference on<br />
Nondestructive Testing, October 1984.
- 110 -<br />
FIGURE 1: Equipment Used for Sizing Defects showing Ultrasonic<br />
Instrument, X-Y Recorder and Simple Scanning Mechanism.
TIME<br />
TRANSDUCER<br />
LENS<br />
WEDGE<br />
* 1.5 mm DIA.<br />
' SDH<br />
- Ill -<br />
NORMAL<br />
POSITI<strong>ON</strong><br />
FOCUSSED<br />
FIGURE 2: (a) Test Calibration Configuration and Ultrasonic A-Scans<br />
Comparing Signal Response for (b) Standard Transducer Beam and<br />
(c) Focused Beam.<br />
POSITI<strong>ON</strong> 123456789<br />
I I II I I I I I<br />
DEFECT THROUGHWALL<br />
OEPTH, h = 3.23 lU-Uj COS a<br />
FIGURE 3: Schematic of Test Setup Illustrating the Muliple A-Scan Format<br />
for Recording Inspection Data.
FIGURE 4:<br />
- 112 -<br />
5 10 IS 10<br />
Depth From Inspection Surface (mm)<br />
(a) Typical Results of a Multiple A-scan Corresponding to the<br />
Right Hand Most Position as shown in Figure 4(b).<br />
FIGURE 4: (b) Cross-Sectional View of Planar Root Crack in 15 mm Thick<br />
Specimen showing Actual Versus Predicted Defect Sizes.
3<br />
(UIUI)1 DEPTH<br />
CTED Fl<br />
PREDI<br />
IUM<br />
MAXIM<br />
12<br />
10<br />
8<br />
6<br />
2<br />
- 113 -<br />
95% C<strong>ON</strong>FIDENCE y<br />
INTERVAL ' '<br />
Y = 0.735 • 0.879X<br />
O VOLUMETRIC DEFECTS<br />
• PLANAR DEFECTS<br />
2 l* 6 8 10<br />
MEASURED FLAW DEPTH (mm)<br />
FIGURE 5: Plot of Measured Versus Maximum Predicted Flaw Depth Using<br />
Focused Ultrasonic Beams for Sizing Purposes. Also shown is<br />
the 95% Confidence Interval for the Predicted Mean Values.<br />
The Defects were Generally Slightly Oversized with an Average<br />
Error of 0.37 mm.<br />
12
FIGURE 6: Output of Computer Modelling of Longitudinal Wave Entering a<br />
Test Sample and Reflecting off the Backwall. Note the<br />
Presence of a Surface Wave Travelling Across the Top of the<br />
Test Sample at Half the Longitudinal Wave Velocity.
- 115 -<br />
FLAW CHARACTERIZATI<strong>ON</strong> USING THE TIME-OF-FLIGHT METHOD AND<br />
ULTRAS<strong>ON</strong>IC FREQUENCY ANALYSIS<br />
V.K. Mak<br />
EnzKgij, Mitiei and Re.ioun.ce.i, Ottawa, Ontanio<br />
ABSTRACT<br />
The time-of-flight method and ultrasonic frequency analysis are valuable tools<br />
for measuring the sizes and shapes of defects. Experiment work was carried<br />
out using both methods simultaneously to measure the diameter and orientation<br />
of circular rods with diameters ranging from 2 mm to 7 mm. A 5 MHz broadband<br />
transducer was used as a transmitter/receiver. Each rod was immersed in a<br />
water bath and located in the far field region of the transducer. The difference<br />
in travel time (At) from opposite edges of the circular rod was measured.<br />
Sound waves diffracted from the circular scatterer were analyzed at<br />
the same time to yield a frequency spectrum and the spacing between consecutive<br />
frequency maxima (Af) was determined. The measurements were repeated by<br />
changing the orientation of the transducer, Both methods have the capability<br />
to determine independently the diameter and orientation of the circular rods.<br />
It was also shown that the product AtAf is approximately equal to 1. Time and<br />
frequency are mutually complementary variables. The two parameters can reinforce<br />
each other in terms of the information each one produces. A merger<br />
of the time-of-flight method and ultrasonic frequency analysis will provide a<br />
powerful nondestructive testing method in the future.
- 116 -<br />
"Events on earth consist in the interplay of two opposed forces, the yin and<br />
the yang, the female and male principles respectively . The interaction<br />
of these two forces together produce most of the universal phenomena. One<br />
yin and one yang together equal the Tao (the Great Ultimate)".<br />
- James K. Feibleman [1].<br />
INTRODUCTI<strong>ON</strong><br />
The ultrasonic examination of materials has played a major role in nondestructive<br />
testing. One technique for sizing defects is called the time-of-flight<br />
method [2]. It is based on measuring the travel time of sound waves diffracted<br />
or reflected from a defect, and has usually been used to measure crack<br />
depths in one dimension [33. Recently the method has been extended to twodimensional<br />
[t,5] and three-dimensional [6,7] cases. Another technique which<br />
has flourished is the ultrasonic frequency analysis method [8], This method<br />
makes use of the analysis of ehe frequency spectrum of the signal scattered<br />
from the defect and attempts to extract information about its size and physical<br />
properties. Both techniques have provided valuable information and have<br />
been used in research and development for the last ten years. An experiment<br />
is described below in which both techniques are used simultaneously to measure<br />
the size and orientation of simulated flaws.<br />
EXPERIMENTAL PROCEDURE<br />
Figure 1 is a block diagram of the time-of-flight method and ultrasonic frequency<br />
analysis system. The immersion technique is illustrated schematically<br />
with a single transducer acting as both puiser and receiver (pulse-echo). A<br />
Metrotek high energy pulser MP215 was used to excite the 5 MHz 1.27 cm diam<br />
transducer. The highly damped transducer emits a relatively broadband multifrequency<br />
pulse with strong frequency components from 1 to 9 MHz. A brass rod<br />
placed in water acted as an idealized flaw whose size and orientation could be<br />
easily controlled. The flattened smooth ends of brass rods with various diameters<br />
(Fig. 2) were used to scatter the broadband ultrasonic pulse. The signal<br />
was detected by a Metrotek wideband receiver MR106 and fed into a Metrotek<br />
stepless RF gate MG701 where the relevant signals were gated. Both the signal<br />
and the gated signal were displayed on an oscilloscope. The time-of-flight<br />
was measured by using the delayed sweep magnification feature of the oscilloscope.<br />
The gated signal was also applied to an HP8557A Spectrum analyzer<br />
connected to an HP853A Spectrum analyzer display which produced a display of<br />
the relative amplitude as a function of frequency on a cathode ray tube (CPT).<br />
The instrument had a measurement range of 10 kHz to 350 MHz. Adjustable<br />
center-frequency and bandwidth enabled any portion of a displayed spectrum to<br />
be examined in detail.<br />
A mechanical system provided accurate control of the orientation of the transducer<br />
and allowed its orientation to be measured to *0.1°. The distance<br />
between the transducer and the specimen (the water path ) was ~18 cm. The<br />
specimen was thus located in the far field region of the transducer.<br />
Each circular brass rod was positioned vertically at the bottom of the water<br />
tank. The transducer was set at a specific orientation. It was then X- and<br />
Y-scanned to provide an optimum signal from sound waves diffracted from the
- 117 -<br />
rod. After time and frequency measurements were taken, the transducer was<br />
rotated about the x-axis to a different orientation and was Y-scanned to<br />
optimize the signal.<br />
THEORY<br />
T .e geometrical theory of diffraction assumes that in addition to reflected,<br />
refracted and incident rays, diffracted rays are produced when a ray hits an<br />
edge [91. The incident ray produces an infinite number of diffracted rays,<br />
traveling in directions determined by the law of diffraction. This law stateo<br />
that each diffracted ray which lies in the same medium as the incident ray<br />
makes the same angle with the edge as the incident ray. Furthermore, the incident<br />
and diffracted rays lie on opposite sides of the plane normal to the<br />
edge at the point of diffraction. However, the diffracted ray need not lie<br />
in the same plane as the incident ray and the edge. Therefore the diffracted<br />
rays form the surface of a cone with its vertex at the point of diffraction.<br />
When a plane wave generated by a transducer is incident on a circular scatterer<br />
at an angle, the only sound wavelets diffracted back into the transducer<br />
originate from two opposite edges of the scatterer (Fig. 3). The wavelets are<br />
coherent sources that interfere at the face of transducer. Because these<br />
wavelets contain a broad range of frequencies, the condition of constructive<br />
interference will always be satisfied by some frequencies. From geometrical<br />
consideration, the frequencies of the wavelets that interfere constructively<br />
at the center of the transducer may be expressed as [10]<br />
f =<br />
n<br />
nv<br />
dsine + [D + Dd sin 9 + ^ à r' - [D - Dd sin 6 + d ] 1 /? (1)<br />
where n is an integer<br />
v = velocity of sound in the medium<br />
d = diameter of the circular scatterer<br />
D = distance of the circular scatterer from the center of the<br />
transducer<br />
9 = orientation of the axis of the transducer with respect to the<br />
axis of the circular rod.<br />
In the far field, d«D, Eq. (1) reduces to<br />
nv<br />
f n = 2dsine<br />
Let e = 8 + y<br />
where ß is the orientation of the axis of the transducer with respect to<br />
the vertical and can be measured and y is the orientation of the axis of<br />
the circular scatterer with respect to the vertical and is an unknown.<br />
Substituting Eq. (3) into Eq. (2) and rearranging terms<br />
2 _n cos S<br />
n<br />
= d cos y tan 6 + à sin<br />
(2)<br />
(3)
- 118 -<br />
now _n = Af (5)<br />
n<br />
where Af is the spacing between consecutive frequency maxima which is the<br />
slope of the plot of f vs n.<br />
Plotting 2hfa—ß" vs tan ß wil1 y ield a strai Sht line with<br />
gradient = d COSY (6)<br />
and intercept = d sinY (7)<br />
Thus, the diameter and orientation of the circular scatterer can be found<br />
from<br />
d = \ (gradient 2 + intercept 2 ) (8)<br />
-1 /intercept \<br />
' I gradient /<br />
For d«D, it can be shown from simple trigonometry that<br />
VAt = 2dsine (10)<br />
where At is the difference in round trip travel time between the two opposite<br />
edges of the circular scatterer.<br />
Substituting Eq. (3) into Eq. (10) and rearranging terms<br />
VAt<br />
2 cos<br />
= d cos y tan s + d sin Y<br />
Plotting „ —— vs tan B will yield a straight line with gradient<br />
c. COS p<br />
and intercept again given by Eq. (6) and (7). Thus d and y can be found<br />
from Eq. (8) and Eq. (9).<br />
The methods described above thus show that ultrasonic frequency analysis and<br />
time-of-flight method can independently yield information about d and Y of<br />
a circular scatterer.<br />
Using Eq. (2), (5) and (10), it can be shown that<br />
AtAf ~ 1 (12)
Results<br />
- 119 -<br />
Figure *l shows the signals diffracted from opposite edges of a circular seatterer.<br />
Constructive interference of the sound wavelets can be seen lying<br />
between the two signals. Figures 5-10 show the frequency spectrum of sound<br />
waves diffracted from the 7 mm brass rod at different orientations. Plotting<br />
fn vs n, the slope yields if as illustrated in Figs. 11 and 12. A frequency<br />
analysis plot and a time analysis plot yielding d and y are shown in<br />
Figs. 13 and It. The actual diameters of the brass rod were measured with a<br />
pair of vernier calipers to ± .02 mm. The values are compared in Table 1.<br />
The axes of the brass rods were vertical (i.e., y = 0) in all cases because<br />
no mechanical system providing accurate control of y was available at the<br />
time of experimentation, but there was no loss of generality in the method<br />
presented here as experimentally, the transducer sees only the relative orientation<br />
of the circular rod with respect to itself.<br />
Figure 15 compares the diameters deduced from frequency analysis and time analysis<br />
with those measured from the vernier calipers. Figure 16 compares y<br />
deduced from frequency analysis and time analysis with the known y. The<br />
average of the error is ~1°. Table 2 illustrates that Ataf~l. The 5.02 mm<br />
diam brass rod was taken as an example.<br />
C<strong>ON</strong>CLUSI<strong>ON</strong>S AND DISCUSSI<strong>ON</strong>S<br />
The time-of-flight method and ultrasonic frequency analysis have been used to<br />
measure the diameters and orientation of circular rods. Commercially available<br />
equipment has been used. In the present experiment, both methods yield<br />
the same information. However it is possible that in other cases, the two<br />
methods can yield information which complement each other. It should be noted<br />
that time measurement can be deduced from the rectified rf signal (Fig. 17).<br />
When the signal is rectified, a lot of information is lost. For unrectified<br />
rf signals frequency analysis can be used to find the bulk and surface properties<br />
of the flaws.<br />
Time and frequency are mutually complementary variables. The two parameters<br />
can reinforce and supplement each other in terms of information each one produces.<br />
The properties of a physical system can be described in terms of pairs<br />
of mutually complementary variables or properties. Other examples of complementary<br />
aspects are the position and linear momentum of a particle, the angular<br />
orientation of a system and its angular momentum, and so on.<br />
The time-of-flight method and ultrasonic frequency analysis have become important<br />
research tools in characterizing flaws for the last ten years. The merger<br />
of the two methods will become a powerful tool for nondestructive testing<br />
in the next twenty years. And hopefully, Canada will play an important role<br />
in their marriage.<br />
ACKNOWLEDGEMENT<br />
The author thanks Mr. M. Hafer for help with plotting the graphs.
- 120 -<br />
REFERENCES<br />
1. J.K. Feibleman, Understanding Oriental Philosophy, p. 86 (Mentor; New<br />
York 1976).<br />
2. M.G. Silk, Research techniques in nondestructive testing; Vol. 3; ed.<br />
R.S. Sharpe; p. 51 (Academic Press; London 1977).<br />
3. K. Date, H. Shiraada, and N. Ikenaga, NDT International V15, p. 315<br />
(1982).<br />
4. D.K. Mak, Physical Metallurgy Research Laboratories Report ERP/PMRL<br />
8*1-1(J), CANMET, Energy, Mines and Resources Canada (1984).<br />
5. D.K. Mak, Physical Metallurgy Research Laboratories Report ERP/PMRL<br />
84-3(J), CANMET, Energy, Mines and Resources Canada (1984).<br />
6. D.K. Mak, Physical Metallurgy Research Laboratories Report ERP/PMRL<br />
84-2KINT), CANMET, Energy, Mines and Resources Canada (1984).<br />
7. M. Macecek, K. Luscott, J. Wells and D.K. Mak, presented at the Fifth<br />
Canadian Conference on Nondestructive Testing, (1984).<br />
8. D.W. Fitting and L. Adler, ultrasonic spectral analysis for nondestructive<br />
evaluation; (Plenum Press; New York 1981).<br />
9. J.B. Keller, J. Appl. Physics Vol. 28, p. 426 (1957).<br />
10. L. Adler and H.L. Whaley, J. of Acoustical Society of America; V. 51,<br />
p. 881 (1972).
d (mm)<br />
Y (deg)<br />
d (mm)<br />
Y (deg)<br />
d (mm)<br />
Y (deg)<br />
d (mm)<br />
Y (deg)<br />
d (mm)<br />
Y (deg)<br />
d (mm)<br />
Y (deg)<br />
- 121 -<br />
Table 1 - Comparison of diameters and orientations of the brass rods<br />
measured from frequency and time analysis with the caliper<br />
values. All brass rods were held vertically<br />
Caliper<br />
values<br />
2.02<br />
0.0<br />
2.96<br />
0.0<br />
4.02<br />
0.0<br />
5.02<br />
0.0<br />
6.01<br />
0.0<br />
7.00<br />
0.0<br />
Frequency<br />
analysis<br />
1.78<br />
2.2<br />
2.99<br />
-0.3<br />
3.81<br />
0.8<br />
4.76<br />
0.4<br />
4.93<br />
2.7<br />
5.99<br />
2.1<br />
% error<br />
-12.1<br />
1.1<br />
-5.3<br />
-5.2<br />
-18.0<br />
-14.5<br />
Time<br />
analysis<br />
2.20<br />
-1.5<br />
2.92<br />
0.3<br />
3.94<br />
0.8<br />
4.27<br />
3.2<br />
5.93<br />
0.1<br />
6.77<br />
0.9<br />
Table 2 - The table shows that AtAf~l for all orientations<br />
The 5.02 mm diam brass rod was taken as an example<br />
8 (deg)<br />
6.4<br />
12.7<br />
18.9<br />
25.1<br />
31.1<br />
at (us)<br />
0.82<br />
1.60<br />
2.20<br />
2.80<br />
3.20<br />
if (MHz)<br />
1.29<br />
0.69<br />
0.47<br />
0.36<br />
0.30<br />
AtAf<br />
1.06<br />
1.11<br />
1.04<br />
1.00<br />
0.96<br />
% error<br />
8.7<br />
-1.4<br />
-2.1<br />
-15.0<br />
-1.4<br />
-3.2
PULSER<br />
IMMERSI<strong>ON</strong><br />
TRANSDUCER<br />
WATER<br />
^CIRCULAR<br />
//ROD<br />
SYNC<br />
RECEIVER<br />
- 122 -<br />
INPUT<br />
OUTPUT<br />
TRIG/EXT<br />
, GATE<br />
GATE<br />
GATED<br />
OUTPUT<br />
DUAL TRACE<br />
EXT. TRIGGER OSCILLOSCOPE<br />
SPECTRUM<br />
ANALYZER<br />
1. Block diagram of the time of flight method and ultrasonic frequency analysis<br />
system.<br />
4 cm.<br />
2 cm.<br />
dem<br />
2 cm<br />
d~0.2.0.3,0.4,0.5,0.6.0.7<br />
2 Brass rods used in experiment.<br />
3. Sound waves diffracted from a circular soatterer. The incident plane<br />
waves are denoted by P.
- 123 -<br />
4. Signals gated for spectrum analysis.<br />
5. Ultrasonic frequency spectrum of signals reflected from the 7 mm diam<br />
brass rod, e = 0.0°.
- 124 -<br />
6. Ultrasonic frequency spectrum of signals diffracted from the 7 mm diam<br />
brass rod, e = 6.0°.<br />
7. Ultrasonic frequency spectrum of signals diffracted from the 7 mm diam<br />
brass rod, e = 12.5°.
- 125 -<br />
8. Ultrasonic frequency spectrum of signals diffracted from the 7 mm diam<br />
brass rod, e = 18.8°.<br />
9. Ultrasonic frequency spectrum of signals diffracted from the 7 mm diam<br />
brass rod, 9 = 25.0°.
i<br />
I<br />
1<br />
i l<br />
- 126 -<br />
—j<br />
10. Ultrasonic frequency spectrum of signals diffracted from the 7 mm diam<br />
brass rod, e = 31.2°.<br />
11. Plot of frequency vs n for 2.02 mm diam rod, S = 25.0°.
PLOT OF FREQ .vs. n CflSE 1- 3<br />
O , KNOHN - 7.000 MTI.<br />
ß - ia.800 DEGREES<br />
0 2 4 6 » 10 12 14 16 IS 20<br />
13. Ultrasonic frequency analysis<br />
plot for the 4.02 mm diam rod.<br />
- 127 -<br />
12. Plot of frequency vs n for 7.00 ram<br />
diam rod, B = 18.8°.<br />
FREQUENCY flNRLYSIS PLOT CflSE 1<br />
0.0 0.1 0.2 0.3 0.4 0.S 0.6 0.7
TIME RNFOSIS PLOT CRSF:<br />
- 128 -<br />
1H. Time-of-flight analysis plot for<br />
the 4.02 mm diam rod.<br />
0-0 1.0 2.0 3.0 *.O 5.0 6 0 7.0 8 0<br />
KNOWN DIFIMETER , MM.<br />
°- FREQUENCY RNHLÏSIS<br />
+ - TIME RNHLYSIS<br />
15. Experimental diameters measured from ultrasonic frequency analysis and<br />
time analysis plotted against the diameter measured from a pair of vernier<br />
calipers. Points in exact agreement would fall on the straight line with<br />
gradient = 1.
- 129 -<br />
°- FREOUENCï flNHLïSIS !<br />
•- TIME HNHLÏSIS I<br />
16. Orientation of the circular rod measured from ultrasonic frequency analysis<br />
and time analysis plotted against the diameter measured from a pair of<br />
vernier calipers. As the circular rods are all held vertically, points in<br />
good agreement would fall on the x-axis.<br />
-4-<br />
A<br />
17. rf signals and rectified rf signals.
- 130 -<br />
THE INTRODUCTI<strong>ON</strong> OF REAL-TIME RADIOGRAPHY FOR THE INSPECTI<strong>ON</strong> OF<br />
BUTT WELDS IN OFFSHORE PIPELINES<br />
T.E. Re.ynold& In.<br />
Bfiown S Root Inc.<br />
Houston, Texas, U.S.A.<br />
This paper was presented at the 16th Annual OTC in Houston, Texas, May 7-9,<br />
1984. The material is subject to correction by the author. Printed with<br />
permission of the Offshore Technology Conference.<br />
1. ABSTRACT<br />
Real time radiography may significantly reduce the costs and efforts, improve<br />
the reliability, and eliminate some subjectivity arising from the conventional<br />
film radiographie inspection of butt welds in offshore pipeline<br />
production. This paper describes this state of the art technology and discusses<br />
its introduction to the production line of traditional pipe laying<br />
barges.<br />
2. INTRODUCTI<strong>ON</strong><br />
Traditional offshore pipeline installation operations involve an assembly<br />
line process. Welding, inspecting, and coating the pipe occur simultaneously<br />
along the assembly line. New weld methods and pipe handling systems are<br />
being developed to speed this fabrication operation. In order to avoid<br />
having the inspection station become the slowest, and thereby the controlling<br />
station in the timing of this process, quicker inspection techniques must be<br />
developed. Because conventional film radiography is currently being pushed<br />
to its limits, other inspection methods were investigated. Furthermore, the<br />
current mood among pipeline contractors' clients is to remain with radiography<br />
and not stray into other non-destructive examination techniques such<br />
as ultrasonics.<br />
Real time radiography provides the required speed of inspection and helps<br />
solve the problem of subjectivity in interpretation. This paper is the<br />
result of efforts to bring a real time radiographie system to field testing<br />
on a pipe laying vessel. Potential manufacturers of these systems were<br />
contacted and encouraged to pursue development of this technology. Field<br />
testing is to be conducted once successful prototype demonstrations have been<br />
completed. This paper attempts to define the considerations to be made in<br />
real time radiography systems by describing the function of the systems and<br />
mentioning current applications of them. The current status of various codes<br />
governing pipeline inspection is reviewed with respect to their acceptance of<br />
real time radiography as a viable inspection method.<br />
Interpretation of radiographs can be a subjective art. Disputes arise among<br />
the three parties whose concern is weld quality: the client, the contractor,<br />
and the inspector. Limiting the range of interpretability of radiographs is
- 131 -<br />
a useful aspect of real time radiography. Completely automated inspection<br />
for the pipeline industry is possible in the future with real time radiography.<br />
3. BASIS FOR THE WORK<br />
The effort to product x-ray or gamma ray pictures in a real time fashion<br />
produced the fluoroscope, a device which converts radiation directly into a<br />
visible image. These are currently employed in some pipe mills around the<br />
country. Real time radiographie imaging is the instantaneous collection and<br />
conversion of x- or gamma radiation into an interprétable image. However,<br />
even with the advancement of technologies in image collectors, cameras, and<br />
displays, the fluoroscope alone has its limitations. In this paper, real<br />
time radiography will refer to computer enhancement of radiographie<br />
"television images" to perform weld inspection.<br />
When coupled with a computer, a fluoroscope type of system takes on new<br />
dimensions. Computer enhanced images expand the range of human perception<br />
and decrease the subjectivity of interpretation, while sacrificing only<br />
seconds in inspection time. Theoretically, computer enhanced radiography is<br />
just short of being real time; it takes a small amount of time for the<br />
computer to execute the functions that produce the enhanced image. The<br />
computer assigns values to each small picture element (pixels) which comprise<br />
the whole raw image. These values are assigned in terms of the location and<br />
intensity of the pixels. By performing algorithms on these values, image<br />
analyses can be performed so that scrutiny of the radiography can be<br />
facilitated. With smaller computers now having the capability to quickly<br />
process huge amounts of data, the gap between theoretical real time and<br />
actual image creation time is becoming negligible for inspection purposes.<br />
Advances in computer technology and simplification of computer programs has<br />
allowed for relatively inexpensive systems to be developed which meet the<br />
requirements of a laybarge operation.<br />
4. STANDARD FOR COMPARIS<strong>ON</strong> (Conventional Radiography)<br />
Real time radiography is applicable to the inspection of butt welds in<br />
pipelines only if it exceeds the following performance parameters of conventional<br />
radiography:<br />
1. Qualification of the image.<br />
2. Cycle time.<br />
3. Total overall cost.<br />
4. Operating reliability.<br />
5. Record keeping ability.<br />
6. Safety to personnel.<br />
A word on these parameters as they relate to conventional film radiography is<br />
necessary before continuing. Conventional film radiography refers to the<br />
inspection process in which a film media is exposed to x- or gamma radiation,<br />
chemically developed to form an image, and then interpreted to help ascertain<br />
weld acceptability.
5. QUALIFYING THE IMAGE<br />
- 132 -<br />
In comparing two methods of producing radiographie images, the basic consideration<br />
is the qualification of the image against a specified standard.<br />
Using a given exposure geometry or technique, an image must be qualified<br />
before it is interpreted. Qualified images meet sensitivity and density<br />
requirements; they are also properly identified for purposes of traceability.<br />
Sensitivity of the film is assured with the use of properly placed image<br />
quality indicators (IQI's). Placed next to the weld at specified increments,<br />
image quality indicators assure the inspector that image contrast sensitivity<br />
and resolution are adequate to visibly detect discontinuities in accordance<br />
with the governing standard. In the United States, the standard API 1104 1<br />
or ASME^ hole penetrameter is the image quality indicator. "2%-2T" sensitivity<br />
Is a common requirement of the film Images. The penetrameters are<br />
chosen with regard to pipewall thickness and weld reinforcement. "2%" refers<br />
to the thickness of the penetrameter as related to the thickness of the steel<br />
that the radiation must penetrate. When "2T" is specified, it is required<br />
that the "2T" hole in the penetrameter appears in the radiograpic image.<br />
In Europe, wire "DIN" image quality indicators are the standard. The ability<br />
to resolve wires of decreasing diameter on a given specimen relates directly<br />
to increasing image sensitivity. A lesser known image quality indicator is<br />
called the CERL IQI. It is composed of three parts; the first part is a<br />
flexible step wedge which is used to measure thickness sensitivity. The<br />
second and third parts are metal wires which are closely spaced and dimensionally<br />
graded in geometric progression. The unsharpness of the image is<br />
measured by the least discernable spacing between two wires. CERL IQI's<br />
might be best suited for measuring the resolution and contrast sensitivity of<br />
television type radiographie images, where the resolution and sensitivity are<br />
partially a function of the pixel density that comprises the image.<br />
The density or relative darkness of an image is a second consideration in<br />
qualifying a radiograph. Density is measured on a scale called Hurter-<br />
Dreffield (H&D) units. Lighter film obtains lower valued H&D units while<br />
darker film obtains higher values. For film radiography, the range of<br />
acceptable density falls approximately between 1.8 and 4.0.<br />
Most specifications require that identification of weld images appear on the<br />
radiograph so that the image'can be traced back to the weld. Improper identification<br />
can be cause for rejection of the radiograph. Currently, lead<br />
numbers and flashing are the primary means for identifying radiographs.<br />
6. INSPECTI<strong>ON</strong> TIME<br />
The inspection cycle time is the time commencing with the arrival of the pipe<br />
at the inspection station and ending with the final interpretation decision<br />
on weld quality. In conventional film radiography, the following sequence of<br />
events occur during the cycle time:
- 133 -<br />
1. Pipe stops at inspection station.<br />
2. Radiation source is positioned at the weld location.<br />
3. Film is wrapped around the joint.<br />
4. Film exposure is made.<br />
5. Film is transported to the dark room.<br />
6. Film is developed.<br />
7. Film is dried.<br />
8. Film is interpreted.<br />
The minimum cycle time for conventional radiography on pipeline operations is<br />
four minutes to read film wet and four and one-half minutes to read film dry.<br />
These numbers represent the physical minimum times using hot chemicals and<br />
high speed dryers while not sacrificing image quality.<br />
7. COSTS OF C<strong>ON</strong>VENTI<strong>ON</strong>AL RADIOGRAPHY<br />
The obvious cost items for provision of offshore film radiographie inspection<br />
services are labor, materials, and transportation. Based on estimates<br />
obtained from proposals submitted to Brown & Root during the past year, an<br />
average cost breakdown was developed for an example job. The chart is based<br />
on a single joint 30" 0 pipeline contract where 180 joints are radiographed<br />
on the firing line every twelve hour shift. This assumes a four minute cycle<br />
time. Total cost of the job, neglecting mobilization and demobilization, is<br />
figured at $53,800 when a 10 day duration of work is assumed. One major cost<br />
item in the job is the cost per joint which is 50% of the total cost of the<br />
work. This single item accounts for the expendable items involved in the<br />
work, film and processing chemicals. These figures were derived from formal<br />
telexed quotes for work from inspection companies for jobs in 1983. As pipe<br />
becomes smaller in size, the expendable costs become a smaller fraction of<br />
total operation costs. For any diameter pipe, the total day rate for labor<br />
and equipment remains constant.<br />
Some costs of film radiography are not so obvious but are substantial. In<br />
bidding work, the difference in cost between the requirement of reading wet<br />
film and reading dry film might be disparate. In bidding, the difference has<br />
translated to time and money lost. Another subtle cost factor is attributed<br />
to barge downtime caused by the inspection station. The time and cost of<br />
this factor have not yet been quantified. Also, radiographers do not know<br />
until at least four minutes after an exposure is made whether or not the<br />
radiograph is qualifiable. The costs of reshoots and crawler breakdowns<br />
comprise other unseen costs of film radiography.<br />
8. OPERATING RELIABILITY<br />
Mechanical and human operations are the main parameters which govern the<br />
operating reliability of conventional radiographie systems. Operating<br />
personnel control such functions as exposure time, chemical temperatures,<br />
placement of film in cassettes, lead shielding, record keeping, and<br />
interpretation. Exposure time is currently regulated manually and controlled<br />
with a timer. Film development has become almost fully automated and less
- 134 -<br />
C<strong>ON</strong>VENTI<strong>ON</strong>AL FILM RADIOGRAPHIC INSPECTI<strong>ON</strong><br />
(Typical Costs in 1983 Dollars)<br />
Single joint operation. 30" 0 pipeline, 180 joints per 12 hour shift, 10 day<br />
duration of job, internal radiography.<br />
6 man radiographie crew<br />
(3 men per shift)<br />
2 Level II inspectors<br />
($400 each/day)<br />
4 Level I inspectors<br />
(§250 each/day)<br />
1 crawler technician on<br />
24 hour call ($400)<br />
Cost per joint (expendables)<br />
$1.00/llnear foot of weld<br />
($8.00 per joint)<br />
Cost of equipment (crawler<br />
facility, dark room,<br />
dryer, maintenance room)<br />
10 day job cost<br />
800<br />
1,000<br />
400<br />
$ 2,200 Labor<br />
$ 2,880 Expendables<br />
300<br />
$ 5,380<br />
$ 53,800<br />
Equipment<br />
Day Rate<br />
Note: This table does not account for transportation costs.<br />
subject to human variability. Shielding must sometimes be replaced to avoid<br />
scratches appearing on the radiographie images. Film is interpreted on a<br />
high intensity light viewer screen. The viewing area must be kept clean and<br />
dry. Film interpreters must stay alert throughout their shift.<br />
A major source of mechanical downtime is due to malfunction of the Internal<br />
crawler. Crawlers provide the radiation source when internal shots are made.<br />
They have been known to run down the line, lose power, and generally break<br />
down.<br />
9. TRACEABILITY OF RECORDS<br />
Accurate record keeping is required when radiographie weld examination is<br />
performed. The ability to trace the radiographie record of a weld back to<br />
the original weld and specific location on the weld is necessary. Currently<br />
lead number belts and the flashing of identification Information onto the<br />
film are common means of keeping identification information. Discontinuities<br />
and sections deemed unacceptable are noted on paper separate from the film.<br />
For retrieval of records, the paper record must be matched to the film<br />
radiograph. Also, with film processes running at such high speeds, the
- 135 -<br />
archivability of film or its ability to retain the original film record, is<br />
very questionable.<br />
10. SAFETY OF PERS<strong>ON</strong>NEL<br />
Radiation safety must also be considered when comparing different systems.<br />
Currently, the total time of radiation exposure is calculated and safety<br />
areas are delimited so that personnel will not be overexposed. The total<br />
time of exposure at a given radiographie station on the laybarge during one<br />
day usually will not exceed two hours if 180 joints per day (4 minute cycle<br />
time) is assumed. This also assumes a 20 second exposure. Radiation safety<br />
zones are determined by marking off an area outside of which exposure will be<br />
limited to two milliroentgen per hour (2mR).<br />
11. REAL TIME RADIOGRAPHY<br />
11.1 System Description<br />
Real time radiography is the non-destructive examination procedure which<br />
collects radiation that penetrates a specimen, converts the radiation to an<br />
electronic image, and then visibly displays the image. The technology of<br />
real time radiography poses three problems:<br />
1• the physics of image formation,<br />
2. the computer management of information, and<br />
3. the application of the complete system to its environment.<br />
11.2 Physics of Image Formation<br />
The formation of real time radiographs follows the same principle as the<br />
formation of conventional film radiographs. Sources utilized for real time<br />
radiography are identical to those utilized in conventional film radiography.<br />
The quantity of radiation energies absorbed by the specimen results in the<br />
contrast density differences that form the image. Radiation sources have the<br />
same influence in the quality of real time image produced as they do in the<br />
quality of film image produced. Focal spot size and energy output directly<br />
influence geometric unsharpness and contrast sensitivity, respectively.<br />
The physics of producing a real time image is more complicated than producing<br />
a film radiographie image. X- or gamma radiation must be converted into a<br />
useable electronic image. Conversion of the radiation to an electronic image<br />
begins when x- or gamma radiation interacts with special phosphors, metals,<br />
or crystalline structures sometimes called scintillators. Scintillators<br />
convert x- or gamma rays into luminescence or electron charges that can be<br />
readily made into a visible image. Common equipment which convert x- or<br />
gamma ray energies into analog information are intensifiers, linear arrays,<br />
and charged coupling devices (CCD). New technologies are aiding in the<br />
development of miniaturized and solid state detecting components. Regardless<br />
of the converter, this analog information must be digitally addressed so that<br />
the computer can create and retain an image. The computer image is called a<br />
digital image because the computer has assigned numerical values to the<br />
densities (brightness) and locations of each picture element (pixel).
- 136 -<br />
11.3 Computer Management of Image Information-* »^<br />
Radiographie images provide information beyond the human range of perception.<br />
The digital values assigned to each pixel in terms of location and intensity<br />
are combined and stored in matrix array as one complete image.<br />
Enhancement is never a substitute for good x-ray technique. Maximizing the<br />
signal (primary radiation) and minimizing noise (scattered radiation) are<br />
necessary; a high signal to noise ratio (S/N ratio) is desirable. Therefore,<br />
an unvarying technique that keeps scatter and geometrical unsharpness to a<br />
minimum is critical to obtaining a qualified radiographie image. In butt<br />
weld radiography of pipelines, the technique is held constant throughout the<br />
job.<br />
Once the raw digitized image is stored in the computer memory, the computer<br />
can begin to enhance it. The S/N ratio of the raw image depends not only or,<br />
the technique, but also depends on the efficiency of the detector.<br />
Brightness transfer functions can be applied to images so that different<br />
results occur. For example, if the intensity variation (film density<br />
variation) across an image falls within some limited narrow range, then that<br />
range can be expanded to encompass the human range of perception. The<br />
result is an image with broader intensity (density) range. Even though real<br />
time radiography provides positive images, the field can be reversed to give<br />
negative images which the film interpreter is accustomed to seeing. By<br />
relating intensity values to quantity of penetrating energies, the computer<br />
can calculate the thickness properties of the inspected piece.<br />
Sometimes the density variation across an image is variable. In the case of<br />
radiographing a pipe section using a film set flat against the pipe rather<br />
than rounded about the pipe, the edges of the developed film appear much<br />
lighter than the center. This is referred to as blowout. And, though image<br />
detectors are analogous to flat film, the computer can assign a function to<br />
the variation of this intensity across the image and, with enhancement,<br />
provide a consistent intensity across the image. This process is called<br />
gradient removal. A shortcoming of gradient removal might be that more of<br />
the image appears to be qualified because of the consistent density.<br />
Histograms relating the thickness of the weld to the distance along a line<br />
drawn across the weld can be drawn by the computer. Any discontinuities can<br />
be measured for length and width. By changing slightly the orientation of<br />
the source on the weld, it is also possible to locate and determine<br />
discontinuities in the third dimension. With such capabilities, three<br />
dimensional topographical radiographie images of a weldment can result.<br />
Computers excel at organizing volumes of information, so records management<br />
is greatly facilitated with real time radiography. The computer can record<br />
information such as operator identification and operating functions<br />
performed. The traceability of welds is made easy by the ability of these<br />
systems to record joint identification, location, and discontinuity data<br />
directly on the imaging media. A variety of media can be used to store<br />
digitized data. These include videotape, magnetic discs, and laser discs.
11.4 Fully Automated Inspection<br />
- 137 -<br />
Three major phases have been identified in moving toward fully automated real<br />
time radiography:<br />
1. The first phase would establish the feasibility of replacing<br />
conventional radiographie film with real time radiography while<br />
using the weld inspector to make all decisions regarding<br />
acceptability.<br />
2. The second phase would use the computer to scan over acceptable<br />
regions of the weld image and stop at questionable areas. The<br />
interpreter would be shown the areas and would be asked to make<br />
a decision regarding acceptance. Image enhancement features<br />
would be available to aid in the decision.<br />
3. Finally, fully automated inspection would involve having the<br />
computer make the radiographie acceptability decisions. The<br />
computer will compare the image against a specification criteria<br />
and make the "go" or "no go" decision.<br />
Laboratory real time radiographie systems are currently entering the second<br />
phase of development toward full automation. Enhancement of questionable<br />
areas within images is possible, and software being developed will enable<br />
computers to locate those questionable areas.<br />
11.5 Application of the System to the Production Environment<br />
The transition from the laboratory to the production environment presents the<br />
single development step that must be bridged before a real time radiographie<br />
system is placed on a laybarge. Although many systems are now capable of<br />
producing qualified images of thick, steel sections, the functional reliability<br />
and efficiency of the systems on a laybarge must be proven. The<br />
following is a list of some considerations to be made in judging the functional<br />
applicability of a system on a laybarge:<br />
Front end of the system. Any pipe mounting band or ring must be<br />
easily detachable to allow for passage of the pipe and miscellaneous<br />
joints, such as valves and tees, through the system. The band or<br />
ring must also be easily attachable and must allow for the image<br />
collector and, if necessary, the source of radiation to travel over<br />
or on it. Consideration must be made for concrete coating on the<br />
pipe.<br />
The image collector should be rugged enough to function in extreme<br />
climatic conditions and should not be affected by the heat of cooling<br />
evolved from the weld. Magnetic fields should not harm its function<br />
either. If double wall exposure, single wall viewing is employed,<br />
the radiation source and detector should track each other at 180°<br />
separation and should remain within strict tolerances.
- 138 -<br />
Back end of system. The computer facility must also be housed to<br />
endure the rigors of transportation and environmental extremes. A<br />
climate controlled, insulated housing facility will protect the<br />
computer system from these extremes. The computer facility must be<br />
free from failure after power outages and should not be affected by<br />
the mechanical vibration of operating laybarges.<br />
12. REAL TIME IMAGE QUALIFICATI<strong>ON</strong><br />
Real time radiographie images should be qualified in the same ways as film<br />
images. The ultimate factor should be the interpreter's ability to discern<br />
the required portion of the image quality indicator. In contrast to film<br />
radiography, consideration should be made to the "graininess" of the digital<br />
image. Because these images are comprised of pixels, the spatial resolution<br />
is higher when there are more pixels comprising a unit area of the image. If<br />
conventional image quality indicators prove insufficient to determine image<br />
sensitivity and contrast resolution, it may be necessary to use the CERL<br />
image quality indicator or one of tailored design.<br />
13. REAL TIME RADIOGRAPHIC INSPECTI<strong>ON</strong> CYCLE TIME<br />
In real time radiography, the sequence of events involved in an inspection<br />
cycle is:<br />
1. Pipe stops at inspection station.<br />
2. Radiation source is positioned at the weld location.<br />
3. Radiation detector is positioned at the weld.<br />
4. Radiographic image is collected and interpreted simultaneously.<br />
Based on demonstrations provided by manufacturers of these real time radiography<br />
systems, it is estimated that image collection will require, at most,<br />
five seconds per diameter inch. Total cycle time will be a little more than<br />
one minute for 8" 0 pipe, and a little more than three minutes for 40" 0<br />
pipe. This allows 30 seconds for positioning of the source and mechanical<br />
equipment. There is also the possibility of mounting more than one source<br />
and detector on the pipe. In turn, the cycle time would decrease by a factor<br />
equal to the reciprocal of the number of detectors.<br />
14. REAL TIME RADIOGRAPHY COSTS<br />
Pipeline contractors and owner companies traditionally contract inspection<br />
work to third party inspection companies. If these companies will be<br />
required to invest in the equipment that comprises a real time radiographie<br />
system, then they will have to recover the cost of their investment through<br />
charges to their clients.<br />
One such scenario might be the following. Given that real time radiographie<br />
systems range in price from $60,000 to $200,000 for a laybarge type system; a<br />
price of $150,000 will be assumed. The inspection contractor would like to<br />
recover the cost of the system, plus $30,000 each for interest and profit.<br />
The inspector estimates that the useable life of the system will be three<br />
years, and it will be used 60 days per year. The total cost which must be<br />
recovered is $210,000. Based on the assumptions stated, the inspection
- 139 -<br />
contractor will have to charge a day rate of $1,170. If another twenty percent<br />
is added for equipment spares, the total day rate for equipment totals<br />
$1,400. The cost of video and computer recording media in relation to the<br />
cost of the expendables has been determined to be about $0.25 per linear foot<br />
of weld. Labor costs are assumed to remain constant. The trend will be to<br />
use less, but more technically trained personnel; therefore, the cost tradeoff<br />
will balance. The following chart is included for comparison to the 30"<br />
0 job described for conventional radiography.<br />
REAL TIME RADIOGRAPHIC INSPECTI<strong>ON</strong><br />
(Estimated Cost in 1983 Dollars)<br />
Single joint operation. 30" 0 pipeline, 180 joints per 12 hour shift, 10 day<br />
duration of job, internal radiography.<br />
Labour (same as conventional film $ 2,200 Labor<br />
radiography)<br />
Cost per joint (expendables)<br />
$0.25/linear foot of weld<br />
($2.00/joint)<br />
720<br />
Equipment and spares 1,400<br />
10 day job cost<br />
$ 4,320<br />
$43 ,200<br />
Expendables<br />
Equipment<br />
Day Rate<br />
Although the above is a hypothetical presentation of how an inspection company<br />
may charge for work undertaken with a real time radiographie system, it<br />
points out two factors:<br />
1. The cost of real time radiography might be essentially the same,<br />
or might be greater than conventional film radiography, depending<br />
on the cost of expendables and size of the pipe.<br />
2. Where the daily cost of operation of a laybarge may range from<br />
$75,000 ($3,125/hour) for a small spread in the Gulf of Mexico<br />
to $250,000 ($10,420/hour) or more for a large operation in the<br />
North Sea, total inspection cost for a whole job can be returned<br />
in a few hours of increased reliability. Pipeline contractors<br />
may want to consider the purchase of a real time radiographie<br />
system and the subcontracting of inspection work performed with<br />
that system. The return on investment could be substantial.<br />
15. OPERATING RELIABILITY<br />
The reliability of real time radiographie systems is more dependent on<br />
mechanical devices than on personnel. Film and proper screens do not have
- IAO -<br />
to be loaded properly into cassettes; no wrapping of the cassette around the<br />
weld joint is necessary. Only the front end mechanical system has to be<br />
latched to the joint. In real time radiography, the image is collected<br />
automatically, scanning rate and radiation output are preset. Once the<br />
scanning begins, the operator has only to begin interpreting radiographie<br />
images, a qualified image is assured by presetting and establishing a<br />
standard radiographie technique. Only minor adjustments may be required<br />
during a shift; any adjustments or enhancement features used are documented<br />
alongside the image. Once scanning of the weld is completed, the image<br />
collecting system is removed from the weld to await the arrival of the next<br />
joint.<br />
Reliability becomes a function of the ability of the equipment to endure the<br />
nonstrop operation on the laybarge. This reliability will be ascertained in<br />
field testing.<br />
To further increase the reliability at the inspection station, investigation<br />
into eliminating the internal crawler is being conducted. One possibility<br />
that exists with real time radiography is the double wall exposure, single<br />
wall viewing (DWE/SWV) of large diameter pipes. With a fractional sacrifice<br />
in image collecting time, internal crawlers might be eliminated from use by<br />
placing the source diametrically opposite from the image collector. Though<br />
size limits of this technique have not been established, demonstrations have<br />
shown qualified real time radiography images using the DWE/SWV technique on<br />
28" 0 x 0.68" W.T. pipe.<br />
16. TRACEABILITY<br />
Information about joint identification, discontinuities observed, enhancement<br />
features used, operator identity, and time and date can be stored alongside<br />
the radiographie image. Rather than have a dual system of paper and film<br />
records, all records for real time radiographie systems are stored on one<br />
medium. To locate repairs, the real time radiography system can refer to any<br />
location on the weld in terms of degrees or linear distances. Hard copies of<br />
images can be made on polaroid film and paper, although they lose some<br />
clarity. If desired, radiographie images can be transmitted, with minimal<br />
lag time, to land based operations via phone or satellite link. The<br />
archivability of the images is a function of only the imaging medium, not of<br />
the development technique as it is in conventional film radiography.<br />
17. SAFETY<br />
Although the time of radiation exposure is substantially increased with the<br />
use of real time radiographie systems, demonstrations have shown that the<br />
radiation intensity required to develop an image will be less than that<br />
required by conventional film radiography. Coupled with the fact that the<br />
DWE/SWV technique will use a collimated radiation beam, rather than the<br />
panoramic beam of conventional film radiography, radiation will pose no new<br />
health hazards to personnel. Safety zones within the 2roR range must still be<br />
delimited.
- 141 -<br />
18. EXISTING APPLICATI<strong>ON</strong>S OF REAL TIME RADIOGRAPHY<br />
Fluoroscopic inspection of pipelines has been ongoing for years in many pipe<br />
mills around the United States. Kaiser Steel Corporation has achieved<br />
qualifiable radiographie weld images on pipelines up to 36" 0 in their Napa<br />
Valley Plant using an unenhanced real time radiography system. They claim<br />
that the savings generated by using this sytem for one and one-half years was<br />
more than $750,000, or more than twice the cost of the system. Kaiser also<br />
sees a great reduction in record storage space with use of video tape.5<br />
Arco Alaska, Inc. desired to assess and quantify corrosion damage in some of<br />
their insulated crude oil flow lines at Prudhoe Bay. They measured corrosion<br />
using a consistent radiographie technique and relating film density (or<br />
radiation penetration) directly back to the quantity of corrosion pitting.<br />
When they applied real time radiography to their problem, they increased<br />
their inspection rate from 27 feet of line per day to 480 feet. Total<br />
inspection cost per foot of pipe inspected was reduced 90%.6<br />
The primary application of real time radiography today is in ordinance<br />
inspection. Defense related manufacturing requires high quality assurance.<br />
To meet these demands, defense contractors have manufactured sophisticated<br />
radiographie systems which can inspect such products as solid fuel in<br />
missiles for porosities or honeycomb sections of jet wings for minor<br />
deformities. With cost not being a major consideration in utilizing these<br />
systems, the resolution and sensitivities obtained are unsurpassable.<br />
19. STATUS OF GOVERNING CODES<br />
API 1104. At its annual meeting in September, 1983, the API 1104 governing<br />
committee considered amending its code to allow for imaging systems other<br />
than film systems to obtain radiographs. Subsquently, a subcommittee met and<br />
reworded Section 8.0 of the code to allow for other imaging systems. The<br />
rewording will be put to a vote before the full committee at the 1984 annual<br />
meeting. Any revised standard will probably not appear until 1985. For now,<br />
Section 5.2 of the code states, "Nondestructive testing may consist of<br />
radiographie inspection or another company specified method". This is the<br />
sentence in the code which implies that real time radiography inspection<br />
systems might be used if they are specified by the owner company.<br />
BS 4515. The BS 4515 committee is currently considering a rewritten draft of<br />
the complete specification with inspection and radiographie testing being<br />
subsections under it. With regard to draft section 25, "Inspection and<br />
Testing", and draft section 26, "Radiographie Examination", both are written<br />
with a generalized wording so that the use of real time radiography is not<br />
prevented. BS 4515 refers one to BS 2910 for radiographie<br />
techniques.''»®<br />
DNV. DNV has no current intentions to rewrite its 1981 code. In its present<br />
form, the DNV code is slanted more toward conventional film radiography and<br />
gives little leeway for real time radiography."
20. OPERATI<strong>ON</strong>AL UTILIZATI<strong>ON</strong><br />
- 142 -<br />
Film interpreters do not, at present, have tools available with which to<br />
enhance conventional film radiographs. It is envisioned that the main<br />
functions of real time radiography will be to inspect more quickly with<br />
higher quality and reliability. If real time radiographie images qualify<br />
according to specifications, no enhancement schemes need be used. The<br />
inspector will continue to interpret images the same as he did with<br />
conventional radiography. If questions arise, enhancement might help answer<br />
them. It is not foressen that special enhancement functions will be<br />
performed during normal operations• The keys to the real time radiography<br />
system will be to avoid a bottleneck in production and to provide a better<br />
system with which to perform radiography.<br />
21. C<strong>ON</strong>CLUSI<strong>ON</strong><br />
This paper has outlined the basis for comparison of conventional film<br />
radiography to real time radiography. Technical evaluations of existing real<br />
time radiographie systems have shown that it is feasible to bring a<br />
laboratory system onto a laybarge and achieve results which will outshine<br />
conventional radiography. Quantified comparative data will result when field<br />
testing is complete. Field testing will also help the offshore industry warm<br />
to the idea of real time radiography as a viable non-destructive weld<br />
examination technique for offshore pipelines.<br />
21.1 Acknowledgements ;<br />
The author wishes to thank the management of Brown & Root for continued<br />
support of this project, and Mr. James H. Walker for his contagious<br />
enthusiasm and guidance from the inception of this work.<br />
22. REFERENCES<br />
1. "Standard for Welding Pipelines and Related Facilities", American<br />
Petroleum Institute, Fifteenth Edition, 1980.<br />
2. "Standard Method for Controlling Quality of Radiographie Testing<br />
SE-142", American Society of Mechanical Engineers, Section V, Article<br />
22, "ASME Boiler and Pressure Vessel Code", 1983.<br />
3. Jacoby, M.H., "Image Data Analysis", Section 15, The Nondestructive<br />
Testing Handbook on Radiography and Radiation Testing, 1982.<br />
4. Pratt, W.K., Digital Image Processing John Wiley and Sons (1978).<br />
5. Cutting, C.C., "X-Raying Pipe on the Line", Welding Design &<br />
Fabrication, April 1983.<br />
6. Hill, D.E. and Galbraith, J.M. "Radiographie Surveys for the<br />
Evaluation of Corrosion Damage in Crude Oil Lines in the Eastern<br />
Operating Area of Prudhoe Bay", paper 52, presented at the<br />
International Corrosion Forum, Anaheim, California, April 18-22,<br />
1983.
- 143 -<br />
7. "Specification for the Process of Field Welding of Steel Pipelines",<br />
British Standards Institute, draft standard of BS 4515, July 1983.<br />
8. "Methods for Radiographie Examination of Fusion Welded<br />
Circumferential Butt Joints in Steel Pipes", British Standards<br />
Institute, BS 2910, 1973.<br />
9. "Rules for Submarine Pipelines", Section 10, Det Norske Veritas,<br />
1981.
%<br />
-•'v/ FILM AND SCREENS<br />
' \ INSIDE CASSETTE<br />
INTERNAL MECHANICAL CRAWLER<br />
PROVIDES X- OR GAMMA RAY<br />
SOURCE<br />
IMAGE COLLECTING<br />
A J<br />
f ^\ Uj I / DEVICE<br />
EXTERNAL RADIATI<strong>ON</strong> SOURCE<br />
FOR DOUBLE WALL EXPOSURE<br />
DARKROOM FACILITY<br />
. FOR RAW AND<br />
ENHANCED<br />
IMAGES<br />
MINICOMPUTER<br />
(HARDWARE AND<br />
SOFTWARE IMAGE<br />
MANIPULATI<strong>ON</strong>S)
- 145 -<br />
C<strong>ON</strong>TINUOUS ACOUSTIC EMISSI<strong>ON</strong> M<strong>ON</strong>ITORING<br />
J.S. Mitchell<br />
Viatec Ra-ioufice Sij.bte.mt> Inc., Calgaiij, Albe-ita<br />
M.SI. Baiiim<br />
Unive-ï-Sity o f 6 Manitoba.<br />
Winnipeg, Manitoba<br />
1. Abstract:<br />
A novel concept that essentially reconfigures the architecture of current<br />
acoustic emissions (A/E) monitoring instruments has been developed and is<br />
now the focus of prototype construction. This system will enable the<br />
continuous A/E monitoring of structures with enhanced reliability.<br />
Applications include integrity monitoring of large structures (pressure<br />
vessels, pipelines, etc.) and process control monitoring (valve operation,<br />
catalyst performance, etc.). Each application requires a thorough<br />
understanding if the metallurgical and acoustic regimes operating in the<br />
structures of interest. With this information the A/E monitor is<br />
programmed to observe and to some extent identify significant acoustic<br />
emissions, and present this information to personnel otherwise not skilled<br />
in interpretation of A/E data.<br />
2. Review:<br />
The present method of acoustic emission monitoring of large structures is<br />
primarily based upon equipment developed within the past 15 years. This<br />
development was enabled by the availability of rugged peizo-electric<br />
transducers, and the advent of "portable" mini computers. There have been<br />
some improvements with this equipment (primarily in the software areas) but<br />
the overall configuration has essentially remained unchanged, see Figure 1.
- 146 -<br />
By design, these A/E monitoring equipments have been targeted toward<br />
inspection type applications. These are typically short term monitoring<br />
tasks that attempt to identify (and in some cases locate) the presence of<br />
active defects in a structure undergoing some form of stress. This<br />
equipment is versatile in that they may be used to monitor several<br />
structure types (pressure vessels, piping, bridges, cranes, etc). However,<br />
with few exceptions, the operation of the equipment and interpretation of<br />
the results requires the constant attention of a skilled operator. These<br />
skills range from an understanding of electronic instrumentation and their<br />
software operating systems, to metallurgical damage mechanism that are<br />
active in the structure being monitored. Hence, the A/E monitor operator<br />
is an incredible individual. Their experience on a given structure is<br />
acquired over a relatively short term test. Also, they are asked to<br />
interpret A/E data without the benefit of learning what affect the service<br />
history of a structure may have had on its acoustic emission behavior.<br />
This is further complicated if the structure is monitored while being<br />
stressed under conditions that do not replicate the service environment<br />
(eg. hydro-static pressure testing of pulp digesters).<br />
Where this equipment has been used for continuous monitoring of large<br />
structures (eg. offshore platforms in the North Sea) the long range<br />
transmission of analog signals appeared as one problem that affects the<br />
overall A/E monitoring reliability. This hardware problem has several<br />
remedies, one of which is to improve the packaging, connectors and to<br />
transmit the A/E signal on a frequency modulated carrier. The overall<br />
system configuration is not changed by this, but it is an improvement in<br />
reliability.<br />
3. Present Concept:<br />
The present concept for the continuous A/E monitoring of large structures<br />
borrows heavily from seismological techniques. As shown in Figure 2 the<br />
concept can be described in terms similar to the operation of several<br />
seismic observatories co-ordinated by a central seismic analysis centre.<br />
In addition, the type of data processing performed by seismologists may<br />
also prove beneficial for the interpretation of A/E signals (ie. assessing<br />
the relavence of A/E signals amongst background noise). Aside from this<br />
concept, the methods of A/E signal analysis presently being investigated<br />
is not possible on conventional A/E monitoring equipment; nor was it<br />
intended to be.
- 147 -<br />
Figure 3 shows a block diagram of the prototype system currently under<br />
development. This system may be considered a redeployment of the<br />
components that make up an advanced conventional multi-channel A/E monitor;<br />
but the advantages that accrue to continuous monitoring of large<br />
structures, to advanced signal analysis, and to reduced system hardware<br />
cost are substantial.<br />
Similar to seismic laboratories located about the earth, this system uses<br />
acoustic emission surveillance units located about a subject large<br />
structure. The surveillance unit is capable of a variety of A/E signal<br />
processing tasks. The current prototype is designed to measure six A/E<br />
parameters that are dealt with during conventional A/E analysis. With the<br />
advent of digital signal processing techniques, subsequent units will have<br />
a variety of enhanced capabilities enabled by this technology.<br />
A surveillance unit is placed in close proximity to each A/E transducer.<br />
In some applications the transducer may be installed within the<br />
surveillance unit enclosure, thus simplifying installation and minimizing<br />
the risks to low level analog signal transmission. The surveillance unit<br />
consists of three electronic subsystems. These are:<br />
a) Analog section for A/E signal processing; the<br />
b) Digital section for measuring A/E transient parameters; and the<br />
c) Microprocessor section which controls the function of the surveillance<br />
unit, may perform some A/E data processing and storage, and transmits<br />
this data to a central control point.<br />
Data transmission is possible by several mechanisms, the most common using<br />
wires to conduct electronic signals. This concept is not restricted to<br />
hard wired systems, but the prototype under development will use a four<br />
wire Modem type data link.<br />
The control point for a network of surveillance units will vary between a<br />
simple status indicator display, to mini computer based data collection and<br />
analysis facilities. Generally, as the complexity of the structure<br />
increases, or as the number of variables that affect the acoustic emission<br />
regime increases, the required computing facility will increase. To<br />
illustrate this, the required central computing facility to monitor the<br />
operation of several valves would be less complex than that to monitor the<br />
acoustic emissions from an offshore platform (the number of A/E transducers<br />
being the same in both cases).<br />
This equipment deployment scheme essentially supports the two levels of<br />
software necessary when establishing the interpretive functions of a<br />
continuous A/E monitoring system. It is the interpretive function that:
- 148 -<br />
a) enables an A/E monitor system to be a part of, or stand alone as, a<br />
structural integrity monitoring system; and/or<br />
b) permits A/E monitoring for process control operations.<br />
Hence, unlike conventional inspection type A/E monitoring, the source and<br />
characteristics of the acoustic emission within a subject structure must be<br />
learned before a continuous A/E monitoring is commissioned. Further, once<br />
continuous monitoring begins, this interpretive function is further refined<br />
in recognition of unique service or operational environments. This can<br />
also be expressed as allowing experience while monitoring a subject<br />
structure (learning curve) to improve the reliability of results. These<br />
results (or output) of the system are presented to otherwise unskilled A/E<br />
operators, or as control signals to other equipment. In addition, the<br />
results of this form of continuous monitoring may be interfaced with other<br />
monitoring functions. The most promising of which has been vibration<br />
analysis; but temperature, pressure, strain, etc. can also be incorporated.<br />
4. Application:<br />
From a technical point of view, the number of applications for continuous<br />
A/E monitoring are considerable. Three applications are currently under<br />
investigation, and at least one other will be undertaken when resources<br />
permit. The structures investigated to date have been a cross country<br />
transmission pipeline, an inlet manifold to a thermal power boiler, and a<br />
semi-submersible drilling platform. In each case, a metallurgical study is<br />
conducted under laboratory conditions using the materials (steel types and<br />
damage mechanisms) that make up the subject structure. Also, acoustic<br />
emission data is collected from the structure, and this information is<br />
combined to program the microprocessor supported surveillance units and the<br />
computer based controller. The pipeline application has involved field<br />
testing of the system hardware, and preliminary results have been very<br />
positive.<br />
Other structures that are readily applicable to continuous A/E monitoring<br />
are:<br />
a) Plant wide piping systems (steam lines);<br />
b) Pressure vessels (pulp digestors, mixing vessels, packed reaction<br />
towers);<br />
c) Boilers and heat exchangers (deairators, manifolds);
d) Offshore structures (subsea pipelines, production platforms).<br />
This is not a complete list, but these applications satisfy both the<br />
criteria for large structures and continuous monitoring for either<br />
integrity or process control. Applications such as ice breaker monitoring<br />
have been considered, but will require a significant effort in acoustic<br />
emission signal processing and analysis before its feasibility can be<br />
established.<br />
5. Summary:<br />
A continuous acoustic emission monitoring concept has been described. The<br />
theme of continuous monitoring is to provide information regarding the<br />
integrity of a structure, or provide feedback for process control, or<br />
both. The acoustic emission behavior of each structure would be determined<br />
through a study of materials, damage mechanisms, and service conditions.<br />
With this information the continuous A/E monitoring system may be<br />
configured to present results to personnel otherwise unskilled in A/E data<br />
interpretation. The system hardware has been developed to prototype form,<br />
and essentially represents a redeployment of conventional, advanced multichannel<br />
A/E instrumentation. However, the seismological resemblance<br />
discussed above is a better model of the system and its operational<br />
concept.<br />
6. Acknowledgements:<br />
We wish to acknowledge the support of Petro-Canada Resou-ces with many<br />
aspects of this concept; and the encouragement and logistics support of<br />
Ontario Hydro Research Division, and Petro-Canada Exploration.
- 150 -<br />
OTHER .<br />
CHANNELS<br />
C<strong>ON</strong>VENTI<strong>ON</strong>AL Mil LTI- CHANNEL<br />
A/E M<strong>ON</strong>ITOR<br />
Figure 1: Hardware Deploynent of a<br />
Conventional A/E Monitor<br />
C<strong>ON</strong>TINUOUS SEISMIC M<strong>ON</strong>ITORING<br />
• SEISMIC OBSERVATORY<br />
(77| SEISMIC ANALYSIS CENTER<br />
Figure 2: Seismic Model of A/E Monitoring<br />
/£ SIGNAL<br />
OUTPUT<br />
MINI<br />
coup.
R*NSoucE ft<br />
SURVEILLANCE<br />
UNIT<br />
- 151 -<br />
MULTI-DROP<br />
OATA LINK TO<br />
OTHER S.U.'«<br />
C<strong>ON</strong>TINUOUS A/E M<strong>ON</strong>ITOR<br />
Figure 3 Hardware deployment of<br />
Continous A/E Monitor<br />
COMPUTER<br />
COM TROLLER
DAY 2<br />
KEYNOTE ADDRESS: Advanced Radiography for Transportation and Energy Systems 152<br />
- H. Berger<br />
NDT - Radiographie Processing Quality Control, Making the Transition to 163<br />
Automatic Processing in Radiographie NDT, Sensitometric Values for Industrial<br />
Radiography<br />
- W.E.J. McKinney<br />
Recent Microfocus X-Ray Imaging Applications (abstract only) 176<br />
- R.S. Peugeot<br />
Wet Channel Inspection Systems for CANDU Nuclear Reactors....CIGAR and 178<br />
CIGARette<br />
- M.D.C. Moles, M.P. Dolbey, K.S. Mahil<br />
An Advanced Heat Exchanger Eddy Current Inspection System 194<br />
- M. DeVerno, H. Ghent, H. Licht<br />
Wet Channel Measurement of Pressure Tube to Calandria Tube Spacing in CANDU 207<br />
Reactors<br />
- J.H. Sedo<br />
Checking for Cracks in Gas Turbine Rotor Discs 222<br />
- J. van den Andel, A.B. Nieberg<br />
Eddy Current Inspection of Mildly Ferromagnetic Tubing 235<br />
- W.R. Mayo, J.R. Carter<br />
On the Relation Between Ultrasonic Attenuation and Fracture Toughness in Type 250<br />
403 Stainless Steel<br />
- F. Nadeau, J.F. Bussiere, G. Van Drunen<br />
Acoustic Emission Testing of Man-Lift Devices 262<br />
- J.A. Baron<br />
Acoustic Sorting of Grinder Balls 276<br />
- F. Nadeau, J.F. Bussiere<br />
The Benefits of NDT Training for Canadians 289<br />
- L.B. Manzer<br />
Fear Detection and Removal - The Psychological Implications of the Technological 297<br />
Age (abstract only)<br />
- T. Helliwell<br />
Certification of NDT Operators in Canada - An Update 298<br />
- V. Caron
VAV TOO
- 152 -<br />
ADVANCED RADIOGRAPHY FOR TRANSPORTATI<strong>ON</strong> AND ENERGY SYSTEMS<br />
Haiold BzKge.1<br />
lndui,t?i.Lal Quality, Inc.<br />
Gaithe.iibu.iQ, UV, U.S.A.<br />
ABSTRACT<br />
Radiography is a code-approved NDT method for inspections involving pipelines,<br />
pressure vessels and many other products. Radiography provides several advantages<br />
for inspection including image quality indicator (IQI) sensitivity of<br />
2% or better, and an image for ease of interpretation and long-term inspection<br />
record retention. New advances in radiographie testing are expected to broaden<br />
the use of radiography even more. One example involves improved real-time<br />
radiographie methods; these are now demonstrating an IQI sensitivity of 2%<br />
or better. Real-time radiography is being applied in areas that include cast<br />
aluminum automobile wheels, aircraft maintenance and pipeline welds, as examples.<br />
Microfocus radiographie equipment is available in energies up to 160kV<br />
in both conventional and rod anode configurations. The geometric magnification<br />
one can achieve improves the radiographie results by displaying extremely<br />
small discontinuities and by the improved contrast that results from the distance<br />
between object and detector. Applications in inspection of composites<br />
and ceramics have been particularly useful. The rod anode equipment has been<br />
applied to welds in small tubing such as used in heat exchangers. Relatively<br />
low energy radioisotope sources such as ytterbium-169 have also been used in<br />
similar applications. A notable advance in the energy field is the development<br />
of a portable linear accelerator for in-service inspection. This new<br />
Minac, a 3.5MeV machine whose radiation head weighs about 100 kg, has been<br />
used in several electric utility inspections for components such as pumps and<br />
piping. A new smaller radiation head, the Shrinkac, has been particularly useful<br />
for piping inspections where access is often limited. Neutron radiography<br />
is now available for on-site industrial applications because of developments<br />
in transportable sources. Corrosion detection in aircraft and other maintenance<br />
inspections are attractive applications for this novel radiographie method.<br />
These examples illustrate that there are many new concepts available in<br />
radiographie testing and that many of the applications for this new technology<br />
are in the transportation and energy fields.<br />
INTRODUCTI<strong>ON</strong><br />
Inspection requirements for the transportation and energy industries are often<br />
critical. Lives and safe use of property depend on the performance of systems<br />
in fields such as aerospace, rail, highway, ship and pipeline transportation<br />
and the production and distribution of energy from nuclear and fossil fuel<br />
plants. It is often these industries that lead the way in the use of advanced
- 153 -<br />
nondestructive testing (NDT) equipment and procedures. This report includes<br />
discussions of some advanced radiographie methods that have been applied or<br />
show promise in these industries [1].<br />
Current methods for the use of radiographie inspection [2,3] mainly involve<br />
X-ray film. These methods yield radiographs that typically display penetrameter<br />
or image quality indicator (IQI) sensitivity [A] values of 0.5 to 2 percent.<br />
The films themselves provide an inspection record that can be retained;<br />
in some cases such as nuclear power, the inspection record retention requirement<br />
may be 40 years or more.<br />
Some new radiation methods that are coming into broader use, particularly in<br />
the transportation and energy industries are electronic real-time detection<br />
methods, improved radiation sources and the use of other radiations such as<br />
neutrons. The following sections expand on these ideas.<br />
REAL-TIME METHODS<br />
Fluoroscopy to present a prompt picture of an X-ray image has been used for<br />
many years for both medical and industrial applications. Television methods<br />
to bring these radiation images to a safe, remote location and provide electronic<br />
control of brightness and contrast prompted widespread medical use beginning<br />
in the 1950's. Industrial interest in these systems clearly existed<br />
but broad use of real-time methods is becoming a strong factor in industrial<br />
inspection only in this decade [5].<br />
Improvements in radiation sensitive screens and image intensifier tubes have<br />
been factors in this increased interest in real-time radiography [5-10], Better<br />
electronic methods for handling and retrieving data have also contributed<br />
to increased use of electronic-based radiation inspection systems [5,11].<br />
Although capital costs for real-time systems can be relatively high, the promise<br />
of efficient, relatively low-cost operation has contributed significantly<br />
to increasing industrial use of real-time methods [7].<br />
Modern real-time X-ray inspection systems are capable of presenting images that<br />
display contrast sensitivity in the 1-3% range and resolution in the 2-5 line<br />
pairs (lp)/mm range. Applications have included inspections of assemblies,<br />
castings, welds, electronic components and a variety of parts and materials<br />
[12], Some specific examples pertinent to this report include the real-time<br />
X-ray inspection of cast aluminum wheels at the rate of 3/min. [ll and the<br />
possibility of inspection of pipeline welds in a quantitative manner [11].<br />
Practical approaches for pipeline weld inspection by mechanized equipment have<br />
been described; diagrams of equipment for both double-wall and single-wall<br />
inspection are shown in Fig. 1 [11].<br />
RADIATI<strong>ON</strong> SOURCES<br />
Radiation sources obviously are a critical part of any radiation inspection<br />
system. A number of advances have been made in recent years including field<br />
emission X-ray tubes for flash radiography [13] and metal-ceramic X-ray tubes
- 154 -<br />
for light weight and resistance to shock [14]. Advances of particular interest<br />
in transportation and energy include a portable linear accelerator [15,16],<br />
a newly available, medium energy radioisotope source [17,18] and new microfocus<br />
X-ray tubes that permit inspections with high geometric magnification<br />
[19,20].<br />
The portable linear accelerator, the Minac, is a lightweight (about 100 kg)<br />
X-ray machine that provides an output of 100 R/min. at 1 meter at an X-ray<br />
energy of 3.5MeV. The unit is portable and has been used for a variety of<br />
field applications including inspections of pumps, valves and piping in nuclear<br />
power plants. Wall thicknesses up to about 20 cm of steel have been inspected.<br />
A new modification, called the Shrinkac, is extremely small (about<br />
25 cm on a side). It is coupled to the main radiation head by a flexible waveguide<br />
up to 6 m in length. This provides excellent versatility for placing<br />
the radiation head in the tight quarters often found in field situations.<br />
At the low energy end of the source scale is the radioactive gamma source<br />
ytterbium-169. This source yields gamma rays in the energy range 53 to 310kV,<br />
with much of the radiation intensity in the 180 to 240kV energy range [17,18].<br />
Therefore, this source is useful for many medium thickness inspections in<br />
which the higher energy of conventional sources such as iridium-192 or cobalt-<br />
60 makes them unsuitable. The first half-value-layer for 1 °^Yb is 4.3 mm of<br />
steel (about 180kV effective). The radiation output is 0.125 R/hour at 1<br />
meter/Ci. Sources up to 10 Ci in size are available. The short half-life,<br />
31 days, is a disadvantage but the portability and medium energy make it useful<br />
for jobs such as panoramic inspections of welds in tubes, a common inspection<br />
problem in the transportation and energy fields.<br />
X-ray units with very small focal spots, 100 jam or less, permit a geometric<br />
magnification of the X-ray image of 5, 10, 20X or more. In addition to providing<br />
magnified views so that small detail, such as porosity in cast turbine<br />
blades, can be seen more readily, these microfocus X-ray images also display<br />
improved contrast because less scattered radiation from the inspection object<br />
reaches the detector. Several commercially available units claim focal spots<br />
in the 1 to 10 um range and yield X-rays up to 160kV in energy. Rod anode<br />
units are available in rod diameters from 8 to 15 mm and lengths from 300 to<br />
1000 mm. Such units are especially useful for panoramic radiography of tube<br />
to tubesheet welds in heat exchangers and other similar applications.<br />
The relatively low energy of these microfocus systems and the excellent detail<br />
and contrast that can be obtained make them particularly useful for inspections<br />
of nonmetallic materials and components such as composites and ceramics<br />
[21], Ar. example of a microfocus X-ray inspection of a ceramic sample showing<br />
small inclusions is shown in Fig. 2. In this case the microfocus X-ray was<br />
combined with a real-time detection system so that the part could be oriented<br />
to give the best inspection view; microfocus X-ray and real-time detection<br />
can provide excellent inspection results.
NEUTR<strong>ON</strong> RADIOGRAPHY<br />
- 155 -<br />
Neutron radiography has been used for many years for inspections of nuclear<br />
fuel and control material, explosive components in aerospace, and high performance<br />
turbine blades from jet engines to cite a few examples from the transportation<br />
and energy fields [2,22,23]. It has also been recognized for some<br />
time that neutron radiography would provide particular sensitivity to hidden<br />
corrosion because the hydrogen-containing components in the corrosion products<br />
would attenuate slow neutrons while most metals that might be in the radiation<br />
path would be relatively transparent [24,25], Recent developments in<br />
transportable neutron sources, coupled with improved neutron detectors (including<br />
real-time methods [26]) have increased interest in neutron radiography.<br />
Much of this new interest involves detection of corrosion in aircraft and space<br />
structures. Neutron radiographie equipment now available includes a mobile,<br />
accelerator-based system developed for the Army [27], An example of corrosion<br />
detection in aluminum structures with this unit is shown for both film and<br />
real-time detectors in Fig. 3.<br />
Other systems soon to be available include higher neutron output acceleratorbased<br />
systems for the Air Force and the Navy [28]. In addition, a large neutron-inspection<br />
facility for aircraft maintenance is being planned at McClellan<br />
Air Force Base; aircraft will be positioned in a special inspection hangar<br />
for neutron radiography [29]. The inspection of aluminum honeycomb structure<br />
will be particularly useful for detecting water and corrosion.<br />
A novel application of real-time neutron radiography is the observation of<br />
fluid movements inside metallic structures such as engines, pumps, etc. Studies<br />
of lubrication and fuel movements in engines by a Rolls-Royce, Harwell and<br />
Burmah-Castrol team [30-32] have excited interest in this technique for better<br />
understanding of the movements of these fluids in operating systems. An example<br />
of results from these studies is shown in Fig. 4.<br />
DISCUSSI<strong>ON</strong> AND C<strong>ON</strong>CLUSI<strong>ON</strong>S<br />
Several novel radiation test methods have been described along with inspection<br />
applications in the transportation and energy industries. The techniques described<br />
represent some highlights concerning novel radiation inspection advances<br />
but the review was not meant to all-inclusive. Other developments such as<br />
semiconductor array detectors, image enhancement, scattering methods and automated<br />
inspection [1,2,5,12,33-35] have not been covered here; these also represent<br />
significant advances in radiography.<br />
Also not covered are advances in computerized axial tomography [36], There is<br />
now much industrial interest in tomography as an inspection method for missiles,<br />
engines, turbine blades, space hardware and other components and assemblies.<br />
Tomography provides a cross-sectional view of the inspection object to<br />
yield information about dimensions and discontinuities. An example of a tomographic<br />
scan of a turbine blade to show wall thinning is given in Fig. 5.<br />
This limited review has shown that many new radiation test techniques are
- 156 -<br />
available. Application examples in the transportation and energy fields have<br />
been presented.<br />
REFERENCES<br />
1. H. Berger and V.J. Orphan, "New Developments in Radiography - An Overview,"<br />
Quantitative NDE in the Nuclear Industry, R.B. Clough, ed., Metals Park,<br />
OH, American Society for Metals, 1983, p. 267.<br />
2. L.E. Bryant, ed., "Radiography and Radiation Testing," Vol. 3, Nondestructive<br />
Testing Handbook, Columbus, OH, American Society for Nondestructive<br />
Testing, 1984.<br />
3. R.C. McMaster, ed., "Nondestructive Testing Handbook," 2 volumes, New<br />
York, Ronald Press, 1959.<br />
4. Anon., "Standard Method for Controlling Quality of Radiographic Testing,"<br />
ASTM E142, Philadelphia, PA, American Society for Testing and Materials,<br />
1983.<br />
5. D.A. Garrett and D.A. Bracher, eds., "Real-Time Radiologie Imaging:<br />
Medical and Industrial Applications," ASTM STP 716, Philadelphia, PA, American<br />
Society for Testing and Materials, 1980.<br />
6. S. Nudelman, H.D. Fisher, M.M. Frost, M.P. Capp and T.W. Ovitt, "A<br />
Study of Photoelectronic-Digital Radiology - Part I: The Photoelectronic-<br />
Digital Radiology Department," Proc. IEEE, V. 70, No. 7, 700 (1982).<br />
7. S. Nudelman, J. Healy, M.P. Capp, "Part II: Cost Analysis of a Photoelectronic-Digital<br />
Versus Film-Based System for Radiology," Proc. IEEE, V. 70,<br />
No. 7, 708 (1982).<br />
8. S. Nudelman, H. Roehrig and M.P. Capp, "Part III: Image Acquisition<br />
Components and System Design," Proc. IEEE, V. 70, No. 7, 715 (1982).<br />
9. P. Schagen, "X-Ray Imaging Tubes," NDT International, V. 14, No. 1, 9<br />
(1981).<br />
10. W. Kuhl, "Detection of X-Rays for Real-Time Imaging," ref. 5, 33-44<br />
(1980).<br />
11. S.A. Wenk, R.A. Brokaw and R.G. Schonberg, "The Potential of Real-Time<br />
Radiography for API 1104," Materials Evaluation, V. 41, No. 9, 1069 (1983).<br />
12. Anon., "Real-Time Imagery Topical Meeting," Atlanta, August 14-16,<br />
1984, Materials Evaluation, V. 42, No. 8, 957 (1984).<br />
13. L.E. Bryant, ed., "Flash Radiography Symposium," Columbus, OH, American<br />
Society for NDT, 1977.
- 157 -<br />
14. H. Putzbach, "Metal-Ceramic X-Ray Tubes," Proc. 9th Uorld Conf. NBT,<br />
Parkville, Australia, Australian Inst. for NDT, Paper 4C-4, 1979.<br />
15. R. Schonberg, "Portable Linear Accelerator Developments," Report<br />
EPRI NP-2831, Palo Alto, Electric Power Research Institute, 1983.<br />
16. M.E. Lapides, "Radiographie Detection of Crack-Like Defects in Thick<br />
Sections," Materials Evaluation, V. 42, No. 6, 788 (1984).<br />
17. D. Pullen and P. Hayward, "Gamma Radiography of Welds in Small Diameter<br />
Steel Pipes Using Enriched Ytterbium-169 Sources," Brit. J. NDT, V. 21,<br />
No. 4, 179 (1979).<br />
18. J.D. Hislop, "Radiography Using Ytterbium-169 - Conference Report,"<br />
NDT International, V. 14, No. 5, 295 (1981).<br />
19. R.W. Parish and D.W.J. Cason, "High Definition Radiography of Cast<br />
Turbine Blades as a Method of Detecting and Evaluating the Incidence of Microporosity,"<br />
NDT International, V. 10, No. 4, 181 U977).<br />
20. B.E. Foster and R.W, McClung, "A Study of X-Ray and Isotopic Techniques<br />
for Boreside Radiography of Tube-to-Tubesheet Welds," Materials Evaluation,<br />
V. 35, No. 7, 43 (1977).<br />
21. D.S. Kupperman, H.B. Karplus, R.B. Poeppel, W.A. Ellingson, H.<br />
Berger, C. Robbins and E. Fuller, "Application of NDE Methods to Green Ceramics:<br />
Initial Results," Report ANL/FE-83-25, Argonne, IL, Argonne National<br />
Laboratory, March, 1984.<br />
22. H'. Berger, ed., "Practical Applications of Neutron Radiography and<br />
Gaging," ASTM STP 586, Philadelphia, PA, American Society for Testing and Materials,<br />
1.976,<br />
23. J.P. Barton and P. von der Hardt, eds., "Neutron Radiography - Proc.<br />
First World Conference," Dordrecht, Holland, D. Reidel Publishing Co. (1982).<br />
24. J. John, "Californium-Based Neutron Radiography for Corrosion Detection<br />
in Aircraft," ref. 22, 168 (1976).<br />
25. H. Berger, "Neutron Radiographie Detection of Corrosion," Proc. Symposium<br />
on NDT and Electrochemical Methods of Monitoring Corrosion in Industrial<br />
Plants, Philadelphia, PA, American Society for Testing and Materials, in press.<br />
26. H. Berger, "Real-Time Neutron Radiographie Observations of Fluid<br />
Motion," 1982 Paper Summaries, Columbus, OH, American Society for Nondestructive<br />
Testing, 487 (1982).<br />
27. W.E. Dance, LTV Aerospace and Defense Co., Dallas, TX, private communication,<br />
1983.
- 158 -<br />
28. V.J. Orphan, Science Applications, Inc., San Diego, CA, private communication,<br />
1983.<br />
29. D. Froom, McClellan Air Force Base, CA, private communication, 1983.<br />
30. P.A.E. Stewart, "Cold Neutron Imaging for Gas Turbine Inspection,"<br />
réf. 5, 180 (1980).<br />
31. D.A.W. Pullen, "Radiographic Photogrammetry Yields Valuable Data in<br />
Studies of Running Gas Turbines," Mechanical Engineering Technology, 27 (July,<br />
1981).<br />
32. P.A.E. Stewart and J. Heritage, "Cold Neutron Fluoroscopy of Operating<br />
Automotive Engines," in ref. 23, 635 (1982).<br />
33. J. McGinnis, "X-Ray Sensitive Photodiode Array," Industrial Res. &<br />
Dev., 143 (March, 1980).<br />
34. Anon., "Digital Radiography," Physics News in 1981, P.F. Schewe, ed.,<br />
New York, American Institute of Physics, 75, 1981.<br />
35. CG. Gardner, "Automated Radiography - A State-of-the-Art Review,"<br />
Report NTIAC-78-1, San Antonio, TX, Nondestructive Testing Information Analysis<br />
Center, 1978.<br />
36. H. Berger, ed., "Topical Issue on Tomography," Materials Evaluation,<br />
V. 40, No. 12, 1982.
FIGURE CAPTI<strong>ON</strong>S<br />
- 159 -<br />
Fig. 1. Illustrations of drive mechanisms for real-time radiographie inspection<br />
of pipeline welds for the source outside the pipe (upper view, double<br />
wall technique) and for the source inside the pipe (lower view, single wall<br />
technique). Photo from ref. 11; courtesy Southwest Research Institute.<br />
Fig. 2. A microfocus, real-time X-ray image of a ceramic spinel disc<br />
across the diameter showing images of dark inclusions in the size range 50<br />
to 100 um. Geometric magnification was 14X.<br />
Fig. 3. Examples of film and real-time neutron images taken with a mobile,<br />
accelerator-based neutron system. The dark areas show neutron attenuation<br />
due to corrosion in the upper and lower views; test pieces, including a modified<br />
ASTM E545 Type A IQI (middle, left) are visualized in the middle region.<br />
Reference 27; courtesy LTV Aerospace and Defense Co.<br />
Fig. 4. Two neutron radiographs of an engine, a static view at the left and<br />
a dynamic view at the right. Differences can be seen such as the pipe full<br />
of oil (in the middle of the lower part) in the static view while the dynamic<br />
view at the left shows much less neutron attenuation showing aeration of the<br />
oil in the scavenger pipe. Reference 31, courtesy NDT Centre, Harwell.<br />
Fig. 5. Tomographie image of the tip of a turbine blade showing thinning of<br />
the wall at the left lower region. Wall thinning in the order of 0.05 mm can<br />
be detected. Courtesy, Scientific Measurement Systems, Inc.
CROSS SLIDE<br />
- 160 -<br />
AXIAL DRIVE : 3"<br />
AZIMUTH DRIVE<br />
UKE CAMERA<br />
PLATFORM IMAGE<br />
INTENSIFIES<br />
it TV CAMERA<br />
ASJUITAIll<br />
TIM«<br />
COUNTERWEIGHT<br />
AXIAL DfflVf<br />
IMME '»» "««* » •<br />
INTENSIFIE*<br />
» TV CAMERA<br />
Fig. 1. Illustrations of drive mechanisms for real-time radiographie inspection<br />
of pipeline welds for the source outside the pipe (upper view, double<br />
wall technique) and for the source inside the pipe (lower view, single wall<br />
technique). Photo from ref. 11; courtesy, Southwest Research Institute.<br />
Fig. 2. A microfocus, real-time X-ray image of a ceramic spinel disc across<br />
the diameter showing images of dark inclusions in the size range 50 to 100<br />
urn. Geometric magnification was 14X.
- 161 -<br />
MOBILE ACCELERATOR N-RAY SYSTEM<br />
DETECTI<strong>ON</strong> OF CORROSI<strong>ON</strong> IN ALUMINUM STRUCTURES<br />
• 1 I«<br />
a<br />
o<br />
ni» ma Ni A« RIAL :i"i I 1 . :r«.-i<br />
VOUGHT<br />
Fig. 3. Examples of fiIra and real-time neutron images taken with a mobile,<br />
accelerator-based neutron system. The dark areas show neutron attenuation<br />
due to corrosion in the upper and lower views; test pieces, including a modified<br />
ASTM E545 Type A IQI (middle, left) are visualized in the middle<br />
reeion. Reference 27; courtesv. LTV Aerospace and Defense Co.<br />
Fig. 5. Tomographie image of the tip of a turbine blade showing thinning of<br />
the wall at the left lower region. Wall thinning in the order of 0.05 mm<br />
can be detected. Courtesy, Scientific Measurement Systems, Inc.
- 162 -<br />
Fig. 4. Two neutron radiographs of an engine, a static view at the left and<br />
a dynamic view at the right. Differences can be seen such as the pipe full<br />
of oil (in the middle of the lower part) in the static view while the dynamic<br />
view at the left shows much less neutron attenuation showing aeration of the<br />
oil in the scavenger pipe. Reference 31, courtesy, NDT Centre, Harwell.
- 163 -<br />
NDT - RADIOGRAPHIC PROCESSING QUALITY C<strong>ON</strong>TROL<br />
W.E.J. McKinncy<br />
E.I. Vu Pont Vz Nzmoun.* S Co., Inc.<br />
iflilm-ington,<br />
U.S.A.<br />
NDT - RADIOGRAPHIC<br />
PROCESSING QUALITY C<strong>ON</strong>TROL<br />
ABSTRACT<br />
In radiography much skill and equipment go into the latent image formation:<br />
source, film, holder, technique, set-up time, exposure time, and safety.<br />
After the exposure the film is processed in a chemical to form the useful<br />
visible image. Actually many films, hundreds of films, are processed by one<br />
central processor. The best technique and the best film, subjected to<br />
inferior processing will produce an inferior visible image. This paper<br />
discusses ways to control processing quality.<br />
INTRODUCTI<strong>ON</strong><br />
The fundamental tenant of Statistical Quality Control is to define the<br />
process, monitor the process, and respond when the process fails. The<br />
process in processing is the oxidation/reduction chemical reaction that<br />
converts the latent image into the visible image. It is a process that<br />
operates close to optimal. When and how it deviates from optimal directly<br />
affects film quality equally.
INTRODUCTI<strong>ON</strong> (Continued)<br />
- 164 -<br />
Processing, being a chemical process, is controlled by elements of time,<br />
temperature, and activity. "Time" is the immersion time the film is in the<br />
chemical and is controlled by a human in manual processing or transport speed<br />
in the transport system of an automatic processor. Temperature control is<br />
very specific and is related to the time factor and the activity factor. The<br />
shorter the time for a given developer activity level or quality, the higher<br />
the temperature required to produce a certain density. Thus, there is a<br />
balance between time and temperature based on a given chemical activity.<br />
This balance allows longer times at lower temperatures or shorter times and<br />
higher temperatures. However, these relationships only work for a givtn<br />
activity. This means the developer must be of a certain quality as a result<br />
of correct mixing, starter solution, and an accurate replenishment/regeneration<br />
rate. Thus, to control processing quality there are three test approaches:<br />
TESTING<br />
o Electromechanical<br />
o Chemical<br />
o Sensitometric<br />
Testing is performed periodically to determine if the process is in control.<br />
Walter A. Shewhart in the early 1940's created the "Trend Chart" which had two<br />
goals: indicate a problem requiring a decision or action, and help indicate<br />
the right decision. He stated that if we control the major variables we do<br />
not have to worry about the minor variables. In radiographie processing there<br />
is time, temperature, and activity or three major variables. Shewhart states<br />
that a trend chart has upper and lower control limits to either side of a goal<br />
or aim value. If there is a large spike or fall-off from the average track of<br />
the chart this is due solely to one of the major variables being out of<br />
mni-rol. Two other interesting facts about trend charts:<br />
o The uncontrolled variables will prevent a straight,<br />
line track. There should be some random fluctuation.<br />
o Although the trend chart demonstrates a problem, the<br />
goal of a quality control program is to never have a<br />
problem.
TESTING (Continued)<br />
- 165 -<br />
Anything can be monitored and placed on a chart. However, not everything<br />
need be tested all of the tine. For instance, unless there is some indication<br />
of a problem then line voltage might only be checked annually. Transport<br />
rates in inches or centimeter per minute might be measured monthly.<br />
Replenishment/regeneration rates should be measured weekly. Temperature is<br />
often monitored daily. Chemical tests are difficult to perform, but could<br />
include bromide, pH, hydroquinone, and archival quality testing performed by<br />
a film manufacturer. Clearing time tests, silver content tests, and specific<br />
gravity tests could be performed periodically in-house. However, the sum<br />
total of the process is the effect of the chemicals, controlled by the<br />
electromechanics, on a controlled exposure in a controlled film emulsion.<br />
This then is sensitometric quality control and is the best test and the most<br />
common test.<br />
SENSITOMETRIC QUALITY C<strong>ON</strong>TROL<br />
Ferdinand Huerter and Vero Driffield in the late 1890's were experimenting as<br />
amateur photographers with different photographic emulsions for scientific<br />
recording of events. They found different emulsion formulas produced different<br />
results so they set out to characterize how a certain film emulsion<br />
responds to a controlled exposure and controlled development. By giving the<br />
candidate emulsion a controlled exposure and controlled processing they could<br />
see how it compared to other emulsions. Thus, there are three components to<br />
this test: film, exposure, and processing (development). If any two components<br />
or variables are held constant, the third may be tested. Sensitoroetric<br />
quality control is a measure of the sum total of the image forming process.<br />
We borrow from the past to see us into the future. But, do not confuse<br />
sensitometry with densitometry which is the measurement of a density in a<br />
pentrameter to meet a work specification.<br />
In sensitometry there are four measurements:<br />
o D-Max: Maximum density for maximum exposure<br />
o Contrast: Difference between two density levels<br />
o Speed: The amount of exposure to produce a density of 2.00<br />
o D-Min: Minimum density for minimum exposure<br />
Base plus fog (B+F) is a measurement of the inherent fog, base, and tint<br />
values combined and occur prior to an exposure. B+F is not a sensitometric<br />
value.
- 166 -<br />
SENSITOMETRIC QUALITY C<strong>ON</strong>TROL (Continued)<br />
The sensitometric values may be detennined directly from a stepped wedge or<br />
from calculations derived from an H&D curve. The indicated value for speed<br />
is measured by reading the density of the step on a wedge closest to D 2.00,<br />
but above D 2.00. This is adeguate for quality control work because it<br />
states that for a given film emulsion and a given exposure we wish to achieve<br />
a certain density. The actual density level is unimportant as long as it is<br />
above D 2.00. The step chosen is called the speed point.<br />
Indicated contrast is measured by selecting two steps, one to either side of<br />
the speed point, on the stepped wedge. The resulting densities are substracted<br />
to produce the difference or contrast value. This is also called the<br />
delta D (AD) meaning difference of density or contrast.<br />
Calculated speed is derived frcm the H&D curve. The coordinate is D 2.00<br />
plus B+F. At this point on the H&D curve a vertical line is dropped to the<br />
abscissa where the relative mAs is read. An alternative is to use Du Pont<br />
Graph Paper which has a built-in Bit Speed Scale which relates to relative or<br />
arithmetic speed. Step 11 or Bit Speed of 16 is assigned a relative value of<br />
100.<br />
Calculated contrast may be calculated one of two ways:<br />
o Gross or Work which has coordinates of 1.00 and 2.00<br />
o ANSI, with coordinates of 1.50 plus B+F and 3.50 plus B+F<br />
In both cases a line is drawn between the coordinate points to form the<br />
average gradient slope line. There are three common ways to measure the<br />
slope or rate of inclination:<br />
o Tangent of the angle<br />
o Parallel rule method<br />
o Rise divided by run (Y/X)<br />
Obviously, indicated values are faster and they can be just as accurate for<br />
quality control work. The values derived should be tabulated for a month and<br />
then an average found. This forms the aim point or index or standard which<br />
is placed on a chart. Upper and lower control limits are established at +/-<br />
10%.<br />
WHEN TO TEST<br />
Test before a work film shows a problem because then it is often too late and<br />
the work film, plus all the other films in processing, must be repeated which<br />
is very expensive. Test prior to the processing of work film such as at the<br />
beginning of each shift. Test during a shift to insure the process is<br />
holding consistent; that the process is not changing. Record the data on the<br />
trend chart. Testing is cheaper than repeats and downtime.
HOW TO TEST (Continued)<br />
- 167 -<br />
Set aside a box of film, sufficient in quantity to last a month or two, in<br />
a good environment such as 20 C (70 F). Label the box for test purposes<br />
only and record the film type, brand, class, emulsion number and expiration<br />
date. Choose a film that is the same as the film most often used. If both<br />
Class I and II films are regularly used choose a higher speed Class I film.<br />
When the first batch of film is about to be used up choose another batch and<br />
check for similarity by proocessing both emulsions together.<br />
Choose an easy, fast exposure technique. Expose a stepped wedge and, if<br />
possible and practical, a defective part that has known flaws. Label this<br />
part "for test purposes only". Attach appropriate penetrameters to the wedge<br />
and part. The film holder or cassette, screens, FFD, etc. must be constant.<br />
The essence of sensitcmetrie quality control is to give a controlled piece<br />
of film a controlled exposure and then process it to see if the processing<br />
is in control. Hold everything reasonably constant in order to monitor one<br />
of the variables.<br />
To help you get started, and everyone must have a sensitometric quality control<br />
program, the Du Pont Company offers a Sensitometric Quality Control Kit consisting<br />
of 50 pre-exposed NDT 55 Class I film, clearing time strips, hypo retention<br />
solution and reference strips for sensitometry, and archival quality. There is<br />
an instruction booklet and a copy of the Du Pont Radiographer's Reference with<br />
technique and processing specifications. Du Pont produces this kit as an<br />
industry service to help get everyone on QC, it is a no-profit item, and it is<br />
intended only to get you started. You should be able to start-up your own<br />
program within the first month.<br />
C<strong>ON</strong>CLUSI<strong>ON</strong><br />
Sensitometric quality control measures the sum total effect of processing<br />
chemicals, time, temperature on a controlled exposure to a controlled piece<br />
of filai. An image is measured to see if processing is in control. This is<br />
the best method to effect quality control on Quality Control. Radiography<br />
is a quality control tool of production, but, is it, itself, under control?<br />
Has the quality been defined and is it consistent? Quality Control is an<br />
economic as well as a technical tool. It is absolutely necessary today...<br />
not twenty years frcm now.<br />
REFERENCE READING<br />
"Radiographic Latent Image Processing" by William E. J. McKinney, Section<br />
Seven, ANSI Nondustructive Testing Handbook, Columbus, Ohio, 1982.<br />
WEJM/daw
- 168 -<br />
MAKING THE TRANSITI<strong>ON</strong> TO AUTOMATIC PROCESSING IN RADIOGRAPHIC NDT<br />
W.E.J. UcKinnzij<br />
E.I. Va Vont Ve. Ue.moan.is ê Co., Inc.<br />
Wilmington, Ve.la.waie.<br />
U.S.A.<br />
MAKING THE TRANSITI<strong>ON</strong> TO AUTOMATIC PROCESSING IN RADIOGRAPHIC NDT<br />
ABSTRACT<br />
Automatic processing offers the potential for high volume, consistent quality<br />
radiographie visible image formation. But, it is only partially automatic<br />
and careful planning and execution are required to achieve the inherent<br />
value. This paper discusses the advantages, and problems of automatic<br />
processing and the procedures necessary in making the transition fran manual<br />
to automatic processing.<br />
INTRODUCTI<strong>ON</strong><br />
Radiographie latent image processing, whether manual or automatic, by hand or<br />
by a machine, is a chemical reaction wherein the latent image is amplified<br />
into the useful visible image. This is an oxidation/reduction reaction. The<br />
developer is oxidized, releasing electrons which selectively attack exposed<br />
silver halide crystals and reduces them to black metallic silver. This<br />
reaction must be controlled by elements of time, temperature, and activity.<br />
A radiographer chooses appropriate time, temperature, and activity levels.<br />
In manual processing a person controls these values. In an automatic processor<br />
the processor controls these functions. To many, this appears to be a<br />
loss of control. Actually, by eliminating the human variable there is<br />
increased consisfency and productivity. But, what if the technique is<br />
incorrect? In manual it can be salvagedi This observation is a classic<br />
dicotomy because yes the film can be saved, but should it be since obviously<br />
the wrong technique was used. Allowing that processing, whether manual or<br />
automatic, must neither overdevelop or underdevelop, why would it be manipulated<br />
to save a poor technique? The answer is, in order to save an exposure<br />
and this is referred to as "sight development". Its major failing is that,<br />
it provides a way to mask the need for technique reform or control.
INTRODOCTIOM (Continued)<br />
- 169 -<br />
Quality control means consistent, measurable quality. An automatic processor<br />
provides this to a degree superior to controlled manual processing.<br />
It is important to understand the prejudices and history before trying a new<br />
method, if the new method is to succeed.<br />
SELECTING A PROCESSOR<br />
Many people gravitate toward the smallest, cheapest unit available. Unfortunately<br />
this means a medical unit slowed down. The result is low capacity<br />
and perhaps increased mechanical load. The ideal processor for NOT applications<br />
is one with 5 to 10 gallon tanks, preferably plastic for greatest<br />
temperature stability, bellows replenishment pimps which are the most accurate<br />
and a transport rate sufficient to produce 50 films per hour on an eight<br />
minute cycle. Of course price must be considered, but it should be considered<br />
in relationship to film capacity and darkroom personnel costs. This<br />
translates into a cost per film. If the processor has built-in variable<br />
drive, cold water washing, water saver, and automatic standby then other<br />
savings are applied to reduce the cost per film.<br />
INSTALLING A PROCESSOR<br />
Processors may be located through darkroom walls, or be placed totally in<br />
the dark. Or it may be totally in the light as when a Du Pont Daylight<br />
Module is used. Some processors can also be placed in trucks, trailers,<br />
airplanes, and on barges and ships. The only criteria is to provide service<br />
room to all sides and to consider the ambient conditions. In the first<br />
case, about two feet is needed on all sides of a processor. In the second<br />
case, ambient conditions include the room air, moisture, and quantity of<br />
metal filings and/or general dust present. The processor has a powerful<br />
blower that draws in 100-200 cubic feet per minute of air to dry the film.<br />
The air should be relatively clean, cool, and dry. Metal filings affect all<br />
developers, but would tend to drawn toward an automatic processor more than<br />
a manual one. Physical installation is described in the installation manual<br />
of the processor.<br />
ADVANTAGES/DISADVANTAGES<br />
The automatic processor offers high volume, consistent quality in a small<br />
package. An untrained darkroom person can produce good films immediately,<br />
or in eight minutes. If the processor is monitored periodically, which is<br />
referred to as sensitometric quality control testing, to insure that it is<br />
indeed consistent, then it is a simple device important to production.
(Continued)<br />
- 170 -<br />
The primary mechanical disadvantage is the presence of 50-100 rollers which<br />
touch the film. But, in manual we learned how to avoid or control air bells,<br />
reticulation, streaks, and hanger scratches. In an automatic processor we can<br />
learn how to keep the rollers clean and because it is automatic it does not have<br />
any of the problems listed above.<br />
The primary advantage is consistency. This is achieved with automatic replenishment<br />
that is only +_ 3% inaccurate for Gorman Rupp Bellows Pumps. It is<br />
achieved with solid state electronic temperature control with inaccuracies of<br />
only +_ 1/4°F in a Du Pont NOT 100 processor or similar unit. It is achieved<br />
with motors, gears, pumps, switches, and thermostats that should last 5 to 10<br />
years.<br />
C<strong>ON</strong>CLUSI<strong>ON</strong><br />
More and more radiographers are buying automatic processors and more important<br />
is the fact that more radiographers are expressing confidence in automatic<br />
processing. Automatic processing is more consistent, it handles higher<br />
film volumes with less mess in a smaller space, it is cost effective. But most<br />
of all, it is more consistent than manual processing. Once this is realized<br />
the next step is check out the costs. But what about the occasinal jam or<br />
roller nark or guide shoe nark? Artifacts such as these will occur occasionally<br />
but are not generic to automatic processing. They are controllable. In<br />
fact, more films will be produced artifact-free than are now possible with the<br />
human manually processing. And improved quality, consistent quality with<br />
improved capacity and reduced costs is attainable only through the modern NOT<br />
radiographie processor.<br />
And, finally, how does one make the transition to automatic processing? By<br />
matching the automatic processing quality to the manual, or if automatic is<br />
superior then adjust techniques. In either case density and contrast are<br />
maintained.<br />
READING<br />
"Radiographic Latent Image Processing" by William E. J. McKinney Section<br />
Seven, ASNT Nondustructive Testing Handbook, Columbus, Ohio, 1982.<br />
WEJM/daw
- 171 -<br />
SENSITOMETRIC VALUES FOR INDUSTRIAL RADIOGRAPHY<br />
W.E.3. UcK.Lnne.ij<br />
E.I. Vu Font Vz Ne.mou.ii, ä Co., Inc.<br />
Wilmington, Vzta.wa.ne.<br />
U.S.A.<br />
SENSITOMETRIC VALUES FOR INDUSTRIAL RADIOGRAPHY<br />
ABSTRACT<br />
In measuring sensitometric values of different film products, or in the use<br />
of sensitometry as a quality control tool, several methods may be used. This<br />
paper discusses indicated versus calculated values. Examples are given of<br />
the various calculated values.<br />
INTRODUCTI<strong>ON</strong><br />
In NOT Quality Control, it is often necessary to produce film that meets<br />
specific sensitometric values. Perhaps more important is to log the values<br />
of test films and to monitor the trends so as to insure consistency. Trend<br />
charts may be constructed frcm indicated or calculated values as outlined<br />
below.<br />
SPEED<br />
o Indicated: Read the density of the step closest to Density 2.00 If<br />
one step reads D 1.85 and the next higher step reads D 2.20,<br />
choose the higher step as standard. Remember this "value"<br />
is influenced by base plus fog. Indicated speed may also be<br />
derived by "eye-balling" (visual comparative analysis).<br />
o Calculated: Locate the coordinate of D 2.00+B+F on the H&D curve. Draw<br />
a line down from this point to intercept the Bit speed calculation<br />
bar*. If your graph paper lacks the calculator set<br />
one up beginning at Step 11 = 16.0 Bits, Step 9 = 17.0 Bits,<br />
etc. A Bit Speed of 16.0 = Arithmetic Speed of 100; 17 = 200;<br />
15 = 50. Alternately, if Step 11 represents a 100 speed, then<br />
Step 12 is a 41% increase in energy or about 20% slower speed.<br />
Step 13 represents half the speed of Step 11 or 50.<br />
*Du Pont H&D Graph Paper.
UJNXKAST<br />
o Indicated:<br />
o Calculated:<br />
- 172 -<br />
Read the density of the steps one above and one below the<br />
speed point. Subtract these densities. The answer is the<br />
"difference" or "delta D" (4 D).<br />
Gamma (
C<strong>ON</strong>TRAST (Continued)<br />
NO. 1 PARALLEL LINES:<br />
- 173 -<br />
Once the slope or line is drawn, any one of the following<br />
three methods may be used to produce the same answer<br />
{*_ 5%). The following graphs demonstrate each method.<br />
In the lower right hand corner of the graph paper, there is a slash line or<br />
a full line at a 45 angle that intercepts the density scale at D 1.00.<br />
Using parallel rulers, opposing triangles or by measuring with a ruler, move<br />
the H&D average gradient slope line over so that it intersects with the slash<br />
reference line at the bottom of the graph paper. Project the line upward; where<br />
it meets the vertical rise of numbers, read the number as the contrast value.<br />
As you can easily see the rise in numbers stops too soon. By measurement, the<br />
vertical scale can easily be extended. Another method would be to cut out a<br />
portion of graph paper and make a scale extender. Another approach is to change<br />
the density scale readings. To do this move the parallel slope line to the<br />
right, just past Step 19. Draw a 45 angle and line so that it exactly meets<br />
D 0.50 on the vertical rise. Relabel this as 1.00; 1.00 becomes 2.00, etc. The<br />
scale has been doubled.<br />
NO. 2 TANGENT OP THE ANGLE:<br />
Using a protractor read the angle of the incline or slope (H&D average gradient<br />
line) to the half or quarter of a degree, if possible. Look up the<br />
tangent of that angle in the trignometric functions of a mathematics book or<br />
use a scientific calculator. The answer is the contrast value.<br />
NO. 3 Y OVER X:<br />
This is also called X into Y or "Rise (Y) over Run (X)". From the upper<br />
coordinate (Y) of the average gradient slope line, draw a vertical line down,<br />
very straight and parallel. From the lower coordinate (X) draw a line horizontal<br />
to the right very straight and parallel. Where they intersect should<br />
be a 90 angle. Measure the vertical Y axis in centimeters. Measure the<br />
horizontal X axis in centimeters and divide into the Y value. The answer is<br />
the contrast value.<br />
It is only necessary to perform one of these methods; the one method you<br />
prefer. There are, also, other methods such as the difference in relative<br />
log exposure divided into the density difference range of 2.00.<br />
WEJM/daw
- 174 -<br />
F»LM_i5=l£B_ DATE22=22=ß3_<br />
EXPOSURE.<br />
PROCESSING<br />
" " " • '<br />
C mv /U '<br />
BIT SPEED/*-S""* M*S/ «2.<br />
CRADIENT<br />
AVERACE CRADIENT *•*?<br />
STEP/5 DENSITY 3<br />
BASE + FOC<br />
r 3 *<br />
VJMe
0 0<br />
- 175 -<br />
FILM_£ciffiî_ DATEB2=22=83_<br />
EXPOSURE 1S7 KVP St—1 ¥>dg»<br />
PROCESSING<br />
t Wo:<br />
BIT SPEED/»"'<br />
CRADIENT<br />
STEP 13 DENSITY<br />
BASE 1- FOC<br />
VJMo<br />
CR0Ö.3K mm<br />
• g oc<br />
-f ~J.OO<br />
t.to<br />
2.SC
- 176 -<br />
RECENT MICROFUCUS X~RAY IMAGING APPLICATI<strong>ON</strong>S<br />
R-icha/id S. Pzagtot<br />
Rldgz Incoipoiatzd<br />
Tu.ck.zK, Go.on.Qia, U.S.A.<br />
The advent of reliable, commercially-available high output microfocus<br />
x-ray generating systems has enabled real time x-ray imaging to successfully<br />
compete with film radiography. The advantages of a near point-source<br />
of radiation in reducing x-ray image penumbral unsharpness are well known.<br />
While tnicrofocus x-ray equipment has existed for more than four decades,<br />
available microfocus x-ray equipment has been characterized by low x-ray<br />
output, rendering the technique unsuitable for real time x-ray imaging.<br />
Recently, advances in microfocus x-ray tubehead design have made available<br />
microfocus x-ray equipment operating at energies up to 160kV with focal<br />
spots on the order of 10 microns while at the same time producing x-ray<br />
output intensities comparable to conventional x-ray generating systems.<br />
The result is that microfocus x-ray imaging can be successfully applied<br />
to a wide range of test projects ranging from thin, lightweight materials<br />
to materials of substantial cross section and density up to the equivalent<br />
of approximately 5/8" of steel.<br />
The advantages of using microfocus imaging techniques are twofold. First,<br />
image resolution is improved by orders of magnitude according to the<br />
amount of geometric enlargement used. Resolutions on the order of 25 lint:<br />
pair per millimeter are routinely achieved at 10X geometric enlargement<br />
using a commercially-available imaging system having a resolving power of<br />
only 2.5 line pair per millimeter. Second, image contrast is improved.<br />
Thii» comes about through a reduction in the amount of test object scatter<br />
radiation reaching the imaging system due to two factors: 1) the x-ray<br />
illuminated area of the test object is reduced, thereby reducing the<br />
amount of scatter generated, and 2) the test object is positioned away from<br />
the image receptor reducing the scatter intensity and intercepted solid<br />
angle.<br />
The result is that high output microfocus x-ray imaging systems can<br />
routinely produce equivalent penetrameter sensitivities on the order of<br />
2-1T or even 1 — IT in some instances. The resulting high quality real time<br />
video x-ray image may be processed using analog and/or digital image
- 177 -<br />
processing techniques. Images may be archived by electronic means, thereby<br />
eliminating costly x-ray film.<br />
This paper will highlight several practical applications of microfocus<br />
x-ray imaging in the area of micro-electronics, investment castings and<br />
composite aerospace materials. Test part sizes range from small microcircuits<br />
to large sections of aerospace structures. Geometric magnifications<br />
ranging from 10X to more than 200X are illustrated.<br />
One interesting application which will be discussed involves the automated<br />
in-place inspection of bonded components on a complete airframe. Robotic<br />
positioning of the microfocus x-ray imaging system over an 85' x 90'<br />
working envelope is provided. All inspection functions except the interpretation<br />
of the actual microfocus x-ray image are under computer control.<br />
Planned future system enhancements include a digital optical disk archival<br />
image storage and retrieval system allowing past inspection images to be<br />
compared with present inspection images for tost part performance tracking.<br />
It is also anticipated that much of the operator-dependent real time x-ray<br />
image interpretation can be placed under computer control providing a truly<br />
automated x-ray inspection system m-ide possible by the high quality real<br />
time x-ray image resulting from the use of high output microfocus x-ray<br />
imaging techniques.
ABSTRACT<br />
- 178 -<br />
WET CHANNEL INSPECTI<strong>ON</strong> SYSTEMS FOR CANDU NUCLEAR REACTORS,..<br />
CIGAR AND CIGARette<br />
M.V.C. l\olzi>, M.P. Volbzy and K.S. Uakil<br />
Ontaftlo Hydto, To Konto, Ontario, Canada<br />
This paper describes and compares the operation of the CIGAR and CIGARette wet<br />
channel inspection systems that have been developed by Ontario Hydro for CANDU<br />
nuclear reactor pressure tubes. Initially, the highly sophisticated,<br />
automated CIGAR (£hannel inspection and Gauging Apparatus for Reactors) is<br />
described. When the hydride blistering problem was discovered in Pickering<br />
NGS Units 1 and 2 in late 1983, CIGAR was not operational, so a less<br />
automated, slower system called CIGARette was rapidly designed and built. The<br />
operation and some problems experienced with CIGARette are also described.<br />
1. INTRODDCTIOH<br />
Pressure tubes in CANDU nuclear reactors contain the bundles of natural<br />
uranium fuel and act as a pressure boundary (see figure 1). There are several<br />
hundred pressure tubes per reactor, mostly made of Zr-2.5%Nb, and these are<br />
surrounded by a calandria tube containing dry gas between the two tubes.<br />
Pressure tubes are 103 mm inside diameter, 4.0-4.5 mm wall thickness and<br />
about 6m long. The pressure tube and calandria tube are separated by garter<br />
spring spacers, from two to four per pressure tube, depending on the unit.<br />
There are two general reasons for inspecting pressure tubes - regulatory and<br />
economic. So far the mandatory regulatory requirements have been less<br />
demanding than special requirements, only requiring gauging of a few pressure<br />
tubes per year/1/. The economic reasons vary from wanting information on<br />
pressure tube behaviour, to predicting and correcting pressure tube problems,<br />
to major, well-publicized outage problems. The latter include the rolled joint<br />
delayed hydride cracking in Pickering units 3 and 4 in 1973-74/2/, and the<br />
more recent hydride blistering problem in Pickering units 1 and 2, which was<br />
the impetus for this paper.<br />
There are two general approaches to inspecting pressure tubes, namely 'dry'<br />
and 'wet' channel systems, and Table 1 gives a sample list of both types. Dry<br />
channel systems require pressure tubes to be defuelled, isolated, drained and<br />
opened, so they require more time and man-rem than wet systems. Also<br />
ultrasonic volumetric inspection is difficult to perform in a dry channel. Wet<br />
channel inspections are performed in a defuelled channel still connected to<br />
the primary coolant circuit. Channels can he defuelled on-line beforehand to<br />
save time and costs with either approach.
- 179 -<br />
This paper describes the concept and evolution of the sophisticated CIGAR<br />
pressure tube inspection system, and the very rapid development and<br />
operational experience with the less automated CIGARette system,both 'wet'<br />
systems.<br />
2. CIGAR<br />
2.1 Concept<br />
Following the problems experienced with the rolled joint cracking in Pickering<br />
units 3 and 4 in 1973-74, it was decided that a sophisticated system for<br />
inspecting and gauging pressure tubes (PT) was required/1/. The system was<br />
called CIGAR (Channel Inspection and Gauging Apparatus for Reactors), and the<br />
specifications have evolved somewhat with time/2-4/. The inspection objectives<br />
are listed in Table 2 and the operational objectives in Table 3.<br />
The CIGAR system can be sub-divided into three areas 1) the delivery system<br />
(drive mechanism and drive rods); 2) the nondestructive evaluation systems<br />
(eddy current,sag and ultrasonics); and 3) control and data handling (storage<br />
and display). These are described briefly below.<br />
2.2 The Delivery System<br />
The drive mechanism, shown on the Ontario Hydro Research mock-up (Figure 2),<br />
was developed by CGE in Peterborough. It is designed to be moved into the<br />
reactor vault and mounted on the fuelling machine bridge, which moves it to<br />
the appropriate channel for inspection. The CIGAR drive is then connected to<br />
the remote control console which is outside containment (Figure 3) by a cable<br />
approximately 100 m long.<br />
The CIGAR inspection head (described in more detail later) has been loaded<br />
beforehand into a defuelled channel by the fuelling machine. The drive<br />
mechanism pushes a drive rod containing the instrumentation cables through a<br />
special closure plug/carrier after connecting to the enclosed CIGAR inspection<br />
head. This connects the nondestructive evaluation (NDE) systems transducers<br />
directly to their instruments in the remote control location. Later the CIGAR<br />
drive automatically inserts a second section into the drive rod to make it<br />
long enough to push the inspection head to the end of a Darlington channel<br />
with 150 mm of elongation.<br />
2.3 The NDE Systems<br />
The CIGAR NDE system utilizes the following transducers:<br />
1) an eddy current probe to measure the gap between the PT and calandria<br />
tube (CT);<br />
2) an inclinometer to measure the sag deflection of the pressure tube;<br />
3) four ultrasonic probes mounted to produce 50° shear waves for volumetric<br />
inspection;<br />
4) a three transducer, four channel ultrasonic gauging system for measuring<br />
PT diameter and wall thickness.
- 180 -<br />
5) an eddy current probe to locate the position of garter spring spacers.<br />
These transducers are operated from instruments in the remote control location<br />
using up to 100 metres of cable.<br />
The NDE transducers are mounted in a CIGAR inspection head (Figure 4), which<br />
consists of two centering modules, a universal joint connector, plus eddy<br />
current, sag measurement and ultrasonic modules. The heads are modular so<br />
inspection systems can he readily interchanged. The limiting factor at the<br />
present time is the number of miniature coaxial cables and connectors that<br />
can be inserted down the drive rods, namely fifteen. A radiation resistant<br />
in-channel multiplexer is under development that will allow more transducers<br />
to share the 15 available cables.<br />
2.4 Control and Data Handling<br />
CIGAR inspections are controlled by a DEC PDP11 minicomputer at the remote<br />
console, using a series of preprogrammed scanning patterns. For example, the<br />
operator might perform 1) a high speed helical ultrasonic inspection using a<br />
3mm axial step (100 minutes for the whole channel); 2) sag measurements,<br />
performed axially (15 minutes), 3) eddy current and ultrasonic wall thickness<br />
measurements for gap (15 minutes), 4) garter spring location if required (15<br />
minutes), 5) detailed ultrasonic inspection of a selected area (say 20-50<br />
minutes), for a total inspection time of about 3 hours per channel. The order<br />
is flexible, depending on the operator.<br />
CIGAR generates a vast amount of data, more than an operator can absorb while<br />
watching the various screens and displays, so some method of recording and<br />
displaying the data is necessary. Much of the data is recorded by a 14 channel<br />
magnetic tape recorder in analog form (Figure 5). This device records the<br />
axial and circumferential positions from the console, 2 channels (X and Y)<br />
from the eddy current instrument, eight channels of data from the flaw<br />
detection ultrasonic instrument (two channels each from four probes), plus a<br />
voice channel. These can he played back through a second computer to give, for<br />
example, isometric displays of the ultrasonic flaw detection data. The<br />
ultrasonic gauging data and sag data are recorded on digital cassette. Also,<br />
the ultrasonic flaw detection system triggers alarms to notify the operator<br />
when signals above threshold are observed, and these are printed out on the<br />
alarm print module, along with the axial position.<br />
3. CIGARette<br />
When Pickering unit 2 pressure tube G16 burst on August 1, 1983, due to<br />
'hydride blister' related cracking, CIGAR was not ready for operation. A mini-<br />
CIGAR, quickly nicknamed CIGARette, was improvised using many CIGAR components<br />
and basic inspection system concepts, and was under test in about a month. A<br />
small friction drive mechanism was designed and built by AECL Sheridan Park<br />
using standard components for speedy delivery/5/.
- 181 -<br />
Though CIGARette was based on CIGAR, it differed in a number of ways (see<br />
Table A). First, if CIGAR is automated (ie) computer controlled, CIGARette is<br />
mechanized (ie) operator controlled. The operators run the CIGARette console<br />
directly and watch and control the NDE instrumentation. Consequently CIGARette<br />
scans a lot slower than CIGAR. The CIGAR drive is much larger, more powerful<br />
and automatically aligns and connects to the channel being inspected whereas<br />
the CIGARette drive is smaller and is clamped to the end-fitting by two<br />
operators (Figure 6). The two drive rods are the same for both systems, but<br />
are loaded by hand for CIGARette. Both CIGAR and CIGARette use CIGAR heads,<br />
closure plugs and drive rods, with only minor modifications.<br />
CIGAR records ultrasonic data on 1A channel tape and plays back directly<br />
through a computer for quasi-on line print outs. In contrast, CIGARette<br />
records data on a similar recorder for later playback, and operators are<br />
required to make assessments of the data in situ. CIGARette play back is<br />
supplementary. Also, the CRT screen of the ultrasonic flaw detector was<br />
videotaped during CIGARette inspections to assist in post test assessment.<br />
CIGAR has a slip-ring unit that allows the inspection head to rotate<br />
continuously. CIGARette does not, so it has to scan back and forth with a<br />
maximum rotation of A00°. CIGARette is operated from an AECL-designed<br />
console/5/,(Figure 7) which can be operated manually or be made to perform a<br />
continuous scan pattern with rotational and axial steps of preset size and<br />
velocity.<br />
3.1 CIGARette NDE Inspections<br />
When CIGARette was developed, the CIGAR inspection head did not contain a<br />
garter spring spacer location system. The head was modified by installing an<br />
eddy current coil for this purpose in place of the CIGAR Sag Module.<br />
Subsequently, the CIGAR head was changed to contain both these systems. The<br />
CIGARette head includes all but the Sag Module. Once the head was in channel<br />
with both drive rods and end connector in place, the CIGARette inspection<br />
typically took the following sequence:<br />
1) An axial pass (with ultrasonic probes at 6 o'clock or bottom dead centre)<br />
to locate the garter spring spacers using the eddy current probe;<br />
2) If one or both of the two garter springs was found to be out of position,<br />
ultrasonic inspections were performed on the longest span. If the two<br />
garter springs were in their design locations (or could not be found),<br />
ultrasonic inspections were performed on the middle of the outlet span<br />
(Figure 8). Typically a number of locations along the PT were inspected<br />
if reflectors were found.<br />
3) Eddy current gap measurements were performed along the length of the PT.<br />
A) Ultrasonic gauging of the wall thickness was performed, along nominally<br />
the same path as the gap measurements, to provide correction information<br />
for the gap measurement.<br />
This sequence is illustrated in Table 5. Ultrasonic volumetric inspections<br />
were performed over an arc of only 45 degrees around bottom dead centre, using<br />
a 2mm axial step. The scans were performed over a limited length of pressure
- 182 -<br />
tube, say 0.5-1.0 metre, so in practice only a few percent of the total PT<br />
volume was inspected for hydride blisters. In contrast, CIGAR inspects<br />
nominally 100% of the PT in a similar period of time. The eddy current gap<br />
operation will be described elsewhere in this conference/6/.<br />
3.3 Problems with CIGARette<br />
After the expected teething problems, CIGARette worked pretty well. A total of<br />
66 PTs were inspected in Pickering 1 and 2, of which about one third showed<br />
reflectors. A further six PTs were inspected in Pickering unit 3. Some of the<br />
problem areas that had to be resolved are discussed below.<br />
Drive Rods/Connectors<br />
The drive rods were loaded by operators working in plastics on the FM<br />
bridge. First, they had to be clean so no slippage occurred in CIGARette's<br />
friction drive. Second, they had to be connected slowly so that the connecting<br />
pins would engage without damage. The connecting pins between the drive rods<br />
and end/head connectors were occasionally damaged during loading. Since about<br />
100 pins are used in each drive rod/head assembly, a critical shortage of<br />
these components arose. Occasionally an inspection had to be performed with a<br />
probe not functioning.<br />
End Connectors<br />
This was a major problem in the early stages with individual wires in the<br />
catinery cable becoming tangled in the pulleys and being cut. Also, the<br />
oscillating rotational scan pattern of the ultrasonic inspection fatigued the<br />
cables at a sharp bend where they came out of the end connector. These<br />
problems were solved by 1) binding the loose wires with tape, and 2) by using<br />
a strain relief spring on the end of the connector.<br />
Rotational Slippage<br />
After only a few inspections it was recognized that progressive rotational<br />
slippage between the drive rod and friction drive (containing the position<br />
readout) could occur. This had serious implications for the ultrasonic<br />
inspection. Blisters were expected only about the 6 o'clock position and a<br />
rotation scan of only 45° was performed about this position. Unrecognized<br />
slippage of only 30° would put the inspection out of the range of the<br />
potential defects. The problem was solved by putting mercury switches in the<br />
end connectors to trigger at the 6 o'clock position to indicate if slippage<br />
occurred.<br />
Operator Comfort<br />
Since CIGARette inspections centre round the operator, fatigue was a potential<br />
problem. The first inspection was done inside the vault in plastic suits, at<br />
great effort. Later, containment penetrations were made, and cabins<br />
constructed, which greatly improved operator conditions.
Problems with Eddy Current Systems<br />
- 183 -<br />
These are described in more detail in/6/. The garter spring location system<br />
coil went through several development stages before a satisfactory reliability<br />
and signal-to-noise ratio was achieved. The gap measurement system, which had<br />
been under development before the hydride blistering problem/4/, saw its first<br />
active service during these inspections and required considerable procedural<br />
modifications before satisfactory results were obtained.<br />
Problems with Ultrasonics<br />
The CIGAR heads and flaw detection (volumetric inspection) worked well,<br />
largely because they had been tested extensively in the laboratory. The probes<br />
lasted longer than expected in the reactor radiation fields. Most of the<br />
problems involved failed connectors and radiation decay of cables. Problems<br />
with standing signals were solved by modifying the ultrasonic modules.<br />
We had some problems selecting a recording threshold since hydride blisters<br />
were a new problem. This was solved in practice by using a 'grass-level plus 1<br />
approach. Probe power checks were performed in each PT using a transmit—<br />
receive method between opposite probes, then the gain increased till grass was<br />
observed on the inside surface gate. All signals a certain level above grass<br />
were noted.<br />
Initially there were problems playing hack the data from the 14 channel analog<br />
tape. Videotaping the flaw detector CRT was useful, but a composite print-out<br />
was also required. Eventually the CIGAR computer play back system was modified<br />
to provide an off line isometric playback facility. Figure 9 shows a typical<br />
computer generated plot of a blistered area, produced from the composite data<br />
of four shearwave probes.<br />
The ultrasonic gauging system in CIGARette was not a great success. The<br />
digital thickness gauges selected for this joh, could not be made to operate<br />
in a stable manner with the long cables used. Consequently discrete<br />
measurements were made on the CRT of the flaw detectors. This was very time<br />
consuming and did not provide as high a precision as desired for gap<br />
measurement correction.<br />
4. CLOSING COMMENTS<br />
The CIGAR system will be used for active channel pressure tube inspection in<br />
1985. It will provide the capability to perform a wide range of inspections<br />
on a number of tubes quickly, with minimal manrem expenditure and with<br />
comprehensive data analysis and presentation either on-line or immediately<br />
after the inspection. The CIGARette system that was improvised for the<br />
Pickering 1 and 2 inspections filled the gap adequately and is now operating<br />
well. CIGARette may he used for quick inspections and for special purposes<br />
such as carrying video cameras. Many of the problems experienced with<br />
CIGARette, such as those due to instrumentation cables and connectors, have<br />
not been experienced with CIGAR due to the different delivery methods. From an<br />
NDE viewpoint, the modular CIGAR head worked well, and in the future, one can<br />
predict that new inspection modules will be developed and used for a variety<br />
of different applications.
ACKNOWLEDGEMENTS<br />
- 184 -<br />
Many people in Ontario Hydro's Research, Design and Development, Central<br />
Nuclear Services and Pickering, plus AECL-Sheridan Park and Canadian General<br />
Electric at Peterborough have been involved in the design, development,<br />
manufacture and operation of CIGAR and CIGARette. Unfortunately there are too<br />
many to list by name, as we would like.<br />
REFERENCES<br />
1. M.P. Dolhey and O.A. Kupcis, I. Hech. E. Conf. publication no C32/79.<br />
2. O.A. Kupcis, Proceedings of the Third Conf on Periodic Inspection of<br />
Pressurized Components, I. Hech. E. Conf. Publication 1976-10.<br />
3. J.A. Baron, M.P. Dolbey, J.H. Erven, D. Booth and D.W. Murray, I. Mech.<br />
E. Conf. Publication no C139/82.<br />
4. J.A. Baron, M.P. Dolbey, J.H. Erven, D. Booth and D.W. Murray, Nuclear<br />
Engineering International, December 1981, p.45.<br />
5. P. Braun, Paper presented at Conf on Robotics and Remote Handling in the<br />
Nuclear Industry, Toronto, Sept 23-27, 1984.<br />
6. J. Sedo - Paper presented at this conference.
- 185 -<br />
TABLE 1<br />
Sample List Of Dry And Wet Channel Inspection Systems<br />
Dry Wet<br />
Closed Circuit TV Rolled Joint Inspection System<br />
Orenda Sag CIGAR<br />
Diameter and Profilometry CIGARette<br />
Inspection<br />
STEM Eddy Current Surface Flaw<br />
Inspection<br />
Laser Sag Measurement<br />
TABLE 2<br />
Inspection Objective Of CIGAR/3/<br />
Conduct mandatory in-service inspections, as required.<br />
Conduct in-service inspections to monitor pressure tube conditions<br />
to predict pressure tube life.<br />
Obtain information that could indicate required action to extend the<br />
inservice life of pressure tubes.<br />
Obtain design-related information for future reactors.
- 186 -<br />
TABLE 3<br />
Operational Objectives Of CIGAR/3/<br />
Minimize outage time required for inspections.<br />
Minimize radiation exposure to personnel.<br />
Operate on a shutdown reactor with the heat transport system cold<br />
and depressurized.<br />
Inspect defuelled channel without draining or isolating,<br />
To be operated remotely from outside containment.<br />
Acquire inspection information at the remote console and process the<br />
data while the equipment is in channel.<br />
Provide archive data for comparison with future inspections.<br />
To operate on Bruce, Pickering and Darlington reactors.<br />
TABLE 4<br />
Differences Between CIGAR and CIGARette<br />
CIGAR CIGARette<br />
Control Automated, computer Mechanized,<br />
controlled operator controlled<br />
Inspection Speed Fast Fairly Slow<br />
Drive Mechanism Large, mounted on F.M. Small, mounted by<br />
Bridge operator<br />
Drive Rod Loading Automatic Manual<br />
Data Playback Essentially on-line Off-Line, later<br />
Scan Pattern Helical or back and Back and Forth<br />
Forth only
Step 1<br />
Step<br />
2<br />
Step 3<br />
Step 4<br />
No<br />
- 187 -<br />
TABLE 5<br />
Sequence for CIGARette Inspection<br />
Action Objective<br />
Axial Scan Using Eddy<br />
Current Garter Spring<br />
Probe<br />
Garter Springs Located?<br />
\ f Yes<br />
Garter Springs in design location?<br />
No<br />
Ultrasonic Inspection<br />
of longest span<br />
Record any indications<br />
Yes<br />
Ultrasonic Inspection<br />
of span between outlet<br />
rolled joint& nearest<br />
garter springs<br />
Axial/rotational scans to measure<br />
PT/CT gap using eddy current<br />
Ultrasonic Gauging of pressure<br />
tube wall thickness<br />
Report Results<br />
Locate Garter<br />
Springs<br />
Detect<br />
Hydride<br />
Blisters,<br />
if present<br />
Eddy Current<br />
and ultrasonics<br />
to measure PT/<br />
CT gap
Pressure<br />
Tube<br />
Fuel<br />
-Carter Spring Spacer<br />
- 188 -<br />
FIGURE 1<br />
— End Fitting<br />
CANDU REACTOR FUEL CHANNEL SHOWING<br />
PRESSURE TUBE AND CALANDRIA TUBE<br />
End Fitting<br />
FIGURE 2<br />
Instrument Contaminant Enclosure<br />
CIGAR DRIVE MECHANISM IN MOCK-UP<br />
Drive Shafts
Equipment<br />
air loch<br />
FIGURE 3<br />
- 189 -<br />
RamoU control<br />
SCHEMATIC SHOWING OPERATI<strong>ON</strong> OF CIGAR<br />
<strong>ON</strong> REACTOR FACE FROM REMOTE C<strong>ON</strong>TROL C<strong>ON</strong>SOLE<br />
Centering Modules<br />
FIGURE H<br />
CIGAR/CIGARETTE INSPECTI<strong>ON</strong> HEAD SHOWING<br />
CENTERING AND INSPECTI<strong>ON</strong> MODULES<br />
to<br />
• Ultrasonic Moiule<br />
•Eddy Current Carter<br />
Spring Coils<br />
Eddy Current Cap Coil
«AC UUSU«O«HII<br />
INCUMKTM<br />
- 190 -<br />
sur««««— FIGURE 5<br />
CIGAR DATA ACQUISITI<strong>ON</strong> AND PLAYBACK<br />
• End Fitting Sleeve<br />
Clamps onto Pressure<br />
Tube<br />
FIGURE 6<br />
Rotational<br />
Drive Motor<br />
CIGARETTE DRIVE MECHANISM IN MOCK-UP<br />
Axial Drive Motor<br />
Hydraulic<br />
Levelling<br />
System
- 191 -<br />
FIGURE 7<br />
: (Axial Drive Index<br />
( Rotational Drive Index<br />
\ Digital Position Readouts<br />
Function Controls<br />
CIGARETTE C<strong>ON</strong>TROL C<strong>ON</strong>SOLE IN C<strong>ON</strong>TROL ROOM
Inlet or<br />
Outlet £nd<br />
Outlet<br />
2m<br />
- 192 -<br />
6 m<br />
2m<br />
Design Locations<br />
of Carter Springs<br />
~Area Scanned by<br />
Volumetric Ultrasonics<br />
(a) Out of Location Carter Spring<br />
7<br />
Area Scanned by<br />
Ultrasonics<br />
Design Locations<br />
Actual Locations<br />
(b) Correctly Located Carter Sprints<br />
FIGURE 8<br />
2m<br />
Actual Locations<br />
of Carter<br />
Pressure Tube<br />
J<br />
Calandria Tube<br />
•• Rolled Joint<br />
SCHEMATIC ILLUSTRATING INSPECTI<strong>ON</strong> AREAS FOR PRESSURE TUBES WITH<br />
(a) CARTER SPRINGS OUT OF LOCATI<strong>ON</strong> AND (b) SPRINGS IN THE DESIGN LOCATI<strong>ON</strong>S
TYP|CAL CIGARETTE<br />
.TAOUTPUT.<br />
ULTRAS<strong>ON</strong>IC DATA OUTPÜ.<br />
. • * • • - . *<br />
One Blister<br />
^C: \One Blister<br />
[Detected by<br />
)Four Probes<br />
OUTERSIDE<br />
9<br />
OUTER<br />
SURFACE<br />
SHOW.NC BUSTERS <strong>ON</strong>
- 194 -<br />
AN ADVANCED HEAT EXCHANGER EDDY CURRENT INSPECTI<strong>ON</strong> SYSTEM<br />
M. Ve.Ve.Mno, H. Ghznt, H. Licht<br />
Atomic Energy oß Canada Limi.te.cl<br />
Chalk Rivz/i, Ontario<br />
ABSTRACT<br />
Heat exchangers are important components in many process industries,<br />
especially nuclear power plants. Nondestructive testing techniques and<br />
equipment are required for fast, reliable testing of heat exchanger tubes<br />
during both unanticipated and scheduled inspections. An advanced eddy<br />
current inspection system has been developed at Chalk River Nuclear<br />
Laboratories to meet this need. This system was designed to meet the needs<br />
of the operator, analyst, station manager and regulatory agencies. The<br />
inspection system provides real-time data acquisition, display, recording and<br />
hardcopying to allow on-line analysis by a qualified eddy current analyst.<br />
In real time, a one page printout containing impedance images and strip chart<br />
traces of eddy current signals is produced for each tube. Tube coordinates,<br />
inspection data, print number and data acquisition validity messages are also<br />
included for record keeping purposes. This record enables the signal analyst<br />
to readily identify, analyze and comment on anomalous signals. Three pairs<br />
of eddy current test data (X and Y components) can be fed into the system.<br />
All three data sets plus probe position and tube information are stored on FM<br />
tape. The operator can select one data set for real-time display. A<br />
sophisticated tape playback feature permits a convenient means of further<br />
analyzing data. The real-time hardcopying allows the inspection team to<br />
provide the utility manager with a copy of the data and inspection results<br />
immediately following the inspection.<br />
This inspection system, known as R'Eddy Record, requires two people to<br />
operate, can realistically inspect 400 tubes/8 h shift, and is compatible<br />
with all commercially available single and multifrequency eddy current<br />
instruments using either absolute or differential type probes.<br />
A particularly impressive feature of the format used to present the data is<br />
the inherent ability of comparing previous inspection data with current<br />
inspection data. This provides a convenient means of monitoring a heat<br />
exchanger's tube performance, and permits long range planning for tube<br />
plugging and/or heat exchanger replacement.<br />
This paper describes the complete inspection system, how it is operated and<br />
field experience to date.
1. INTRODUCTI<strong>ON</strong><br />
- 195 -<br />
The development of an advanced eddy current testing (ET) nondestructive<br />
Inspection system for heat exchanger tubing in nuclear facilities and other<br />
industries presents a challenging set of problems. For the operator the<br />
inspection system must be reliable, easy to operate and maintain. For the<br />
analyst the system must present the data to facilitate on—line analysis.<br />
Stored data required for signal confirmation must be comprehensive and easily<br />
retrievable. The system must enable the complete inspection to be done<br />
quickly and reliably to reduce the overall downtime of the heat exchanger,<br />
and yet provide the utility manager and control agencies a convenient means<br />
by which the conclusions of the inspection can be verified. To provide a<br />
method of monitoring a heat exchanger's tube performance and permit long<br />
range planning of maintenance and replacement, a convenient method of<br />
comparing collected data must be available. The R'Eddy Record system,<br />
designed to meet these needs of the operator, analyst, utility manager and<br />
control agencies, represents a significant advance in eddy current inspection<br />
technology.<br />
Eddy current inspection is focused on the collection and analysis of two<br />
voltage signals, Vx and Vv, which together represent the impedance of the<br />
eddy current coil and make up one channel of eddy current data. The Vx and<br />
Vy signals respectively represent the XY cartesian mapping of coil<br />
impedance and the Lissajous figure, Vx versus V„, represents the eddy<br />
current impedance image or 'feature'. The measured coil impedance is a<br />
function of frequency and the material environment surrounding the coil.<br />
Conventional single and multifrequency eddy current instrumentation can<br />
provide X and Y outputs at both single and multiple test frequencies,<br />
frequency mixes, and provide a storage oscilloscope impedance image display.<br />
Signal analysis of the eddy current data involves interpretation of the 2—D<br />
image with the separate X and Y versus distance traces providing information<br />
on signal direction within the image.<br />
A rudimentary inspection would consist of manual probe insertion down a heat<br />
exchanger tube. An operator would analyze the tube condition by visually<br />
monitoring the oscilloscope impedance image. A single storage scope display<br />
superimposes images, thus signals can be missed and/or misinterpreted. If<br />
long terra data collection or hardcopy results were needed the data could be<br />
put on FM analog tape and/or strip chart. On-line signal analysis is very<br />
difficult and prone to errors in this type of inspection. Post inspection<br />
analysis must then be accomplished. On large inspections, literally miles of<br />
strip chart can be generated making data retrieval and record identification<br />
cumbersome and difficult. Data playback from analog tape also presents the<br />
same problems of retrieval, identification and feature display<br />
superpositioning.<br />
A logical extension to manual inspection and the CRNL forerunner to the<br />
R'Eddy Record system was a motorized probe pusher/puller (probe drive) to<br />
speed up data collection, and a separated feature playback display that<br />
allowed an operator to display non-superimposed ET features on an analog XY
- 196 -<br />
storage terminal. However, the major problems associated with off-line<br />
analysis, data retrieval and record keeping still remained. The R'Eddy<br />
Record system solves these problems.<br />
2. SYSTEM DESIGN OVERVIEW<br />
2.1 General<br />
The R'Eddy Record heat exchanger inspection system provides computer<br />
controlled data acquisition, probe positioning, data storage and real-time<br />
graphical data display with hardcopy printout. The data display and hardcopy<br />
were specifically designed to facilitate on-line analysis and provide a long<br />
term quality assurance document. This one page printout provides a permanent<br />
record of the X and Y signal traces, separated features, tube Identification<br />
and pertinent scan information of a single channel of eddy current test data<br />
for immediate analysis and future review. Three channels of eddy current<br />
data can be collected and stored along with record identifiers and position<br />
information on magnetic tape and strip chart. These data records can quickly<br />
be retrieved and displayed through the playback feature. The system will<br />
collect and display any twin set of analog signals making it a general data<br />
acquisition system independent of the type of probe or probe configuration<br />
and the particular signal generating instrumentation.<br />
The overall system configuration can be divided into a probe drive control<br />
subsystem (Figure 1) and a data processing subsystem (Figure 2). The modular<br />
design allows for flexibility in configuration, and ease of maintenance and<br />
updating. The overall weight and physical dimension of any system module<br />
(weight
- 197 -<br />
This subsystem was designed for repetitive scanning of heat exchanger tubes<br />
with automatic stop and retract upon reaching a preprogrammed position or<br />
under control from the data processing subsystem. The probe drive control<br />
subsystem communicates status, position information, tube coordinates and<br />
three twin analog channels to the data processing subsystem through a single<br />
multi-channel link.<br />
2.2.2 Probe Drive<br />
The probe drive comprises four segments designed for easy access to congested<br />
areas and assembly in minutes. The probe drive provides 400N (90 lbf) drive<br />
force using four drive wheels for both plastic and steel wound cables. A<br />
compact motor driven take-up reel with mercury-wetted slip rings is utilized.<br />
An incremental shaft encoder allows for high resolution data collection and<br />
accurate positioning. A flexible probe guide tube directs the probe cable<br />
and probe to the heat exchanger tube. The probe guide normally connects to a<br />
calibration tube and zero limit switch. This switch allows the system to<br />
correct for positioning hysteresis effects. A second mechanical switch<br />
placed within the probe guide tube just ahead of the drive wheels prevents<br />
the probe from being withdrawn through the drive wheels.<br />
2.2.3 Probe Drive Interface<br />
The probe drive interface module contains the probe drive power supply, speed<br />
regulation controller and manual probe drive controls. Manual probe drive<br />
operation without position readout is possible with this module alone.<br />
2.2.4 Probe Drive Controller<br />
The probe drive controller is a microprocessor controlled module responsible<br />
for the execution of preprogrammed scan patterns and manual scanning with<br />
full distance display plus slip and jam detection. The probe drive<br />
controller accepts input via two remote keyboards from either the operator or<br />
the analyst. Shaft encoder and limit switch inputs provide positioning<br />
information and three twin differential channels transmit eddy current data<br />
to the data acquisition subsystem.<br />
2.3 Stand-Alone Probe Controller<br />
The stand-alone probe controller module is a scaled down version of the probe<br />
drive interface and controller. This single module is fully programmable by<br />
the operator via a hand held keyboard. The module contains the probe drive<br />
power supply, speed regulator, and microprocessor control system. This<br />
module allows repetitive scanning with automatic retraction, time delayed<br />
repeat and jam detection. Programmed speeds and distance regions allow scan<br />
speed variation in the U-bend region in both the retraction and insertion<br />
phases of a scan. No eddy current signal transmitting capabilities are<br />
included in this module.
2.4 Data Processing Subsystem<br />
2.4.1 General Design<br />
- 198 -<br />
The data processing subsystem is the heart of the R'Eddy Record system. It<br />
provides user friendly operation of the overall system by means of an<br />
interactive menu which leads the operator through the entire inspection.<br />
The major functions of the data processing subsystem are:<br />
i) Scanning and display parameter set generation through the interactive<br />
nenu. Up to 16 distinct parameter sets can be created and stored in<br />
memory.<br />
ii) Storage and retrieval of scanning and display parameters on digital<br />
cassette for quick recall.<br />
iii) Transfer of scan pattern to probe drive controller,<br />
iv) Data acquisition,<br />
v) On-line graphic display of data,<br />
vi) Automatic hardcopy at the end of each scan,<br />
vii) Tubesheet signal pattern recognition to allow the scan to end at the<br />
far tubesheet. This allows tubes of varying length to be scanned<br />
without the probe exiting the far end of the tube and getting caught<br />
on retraction,<br />
viii) The generation of record identifiers and distance encoding for both<br />
strip chart and FM analog recordings,<br />
ix) The automatic control of FM analog recorders,<br />
x) Voice synthesis of the tube number (as keyed in by the operator) and<br />
scan status on scan completion. This voice synthesis can be broadcast<br />
over a speaker and on the voice channel of the analog recorder,<br />
xi) Automatic playback of tape records incorporating high speed search for<br />
tube or print numbers,<br />
xii) Overall system coordination.<br />
The data processing subsystem is modular and structured; operator interaction<br />
is centered around the menu. A main menu page gives the operator a number of<br />
tasks which the system will execute and which the operator invokes by<br />
number. Tasks that do not lead directly to scanning automatically return to<br />
the main menu page with the task listing. Tasks that lead directly to tube<br />
scanning allow the analyst the option to return to the menu before scanning<br />
commences. During scanning, allowed operator inputs are displayed on a two<br />
line display on the operator terminal. The operator can exit the scanning<br />
sequence back to the menu by single key operation at selected points.<br />
Data acquisition is based on an interrupt driven data acquisition format that<br />
ensures a minimum of one point (X,Y) of data is collected for every 0.4 mm of<br />
probe travel. As data is collected, it is displayed as separated features on<br />
a monochrome monitor.
2.4.2 Repetitive Scanning<br />
- 199 -<br />
The design of a repetitive scan function revolves around maintaining the<br />
shortest overall inspection turnover time. The system reacts to slip and jam<br />
conditions of the probe drive in order to prevent probe and probe cable<br />
damage. Maximum insertion speed of 50 cm/sec and retraction speed of 75<br />
cm/sec and the use of selected speed/location changes allow the fastest<br />
possible speed without signal distortion and cable jamming. The need to<br />
re-scan due to false tube identification is minimized since the system<br />
recognizes when the tube number has not been updated, alerts the operator and<br />
suspends any subsequent scan until updating has taken place.<br />
The operator is alerted to signal deterioration by the calibration signals<br />
that are presented on the first feature of every display. Due to the nature<br />
of the display, signal anomalies can be readily picked out by an analyst,<br />
thus allowing concurrent analysis. An average scan time of roughly 1<br />
minute/tube, provides enough time for such analysis. Also, to reduce<br />
inspection downtime, the system can be temporarily controlled by a single<br />
operator.<br />
The analyst selects a scan parameter set for use, then enters the main<br />
inspection sequence through the menu. The operator then positions the probe<br />
guide at the tube to be inspected and either the operator or analyst<br />
initiates the scan by depressing the scan 'Ready 1 key on the remote keyboard.<br />
As the probe enters the tube, data is collected, displayed and stored on<br />
tape. The insertion sequence terminates with recognition of the tubesheet or<br />
upon reaching a preprogrammed position. During probe insertion the tube<br />
coordinates are entered via the remote keyboard. At the end of the probe<br />
insertion phase the probe automatically retracts, the display hardcopy is<br />
produced and the voice synthesis announces the scan status and tube<br />
coordinates. If the tube number is not entered by the end of the insertion<br />
phase then a 'label' light is lit on the keyboard; the probe still retracts<br />
but the hardcopy and voice synthesis are delayed until the tube coordinates<br />
are entered. The end of printing, voice synthesis and probe retraction to<br />
zero, mark the end of the retraction phase. The operator then repositions<br />
the probe guide and the sequence repeats. During a scan the operator can<br />
stop and start probe motion by using the keyboard. The scan can be<br />
terminated (scan stopped and probe retracted) or aborted (probe drive motion<br />
stopped pending reinitialization) by the analyst through the terminal. The<br />
operator and remote keyboard could be replaced by an automatic tubesheet<br />
walker for scanning tubes in hostile environments such as nuclear steam<br />
generators.<br />
3. SYSTEM OPERATI<strong>ON</strong> AND DATA DISPLAYS<br />
3.1 General Operation<br />
System operation begins with equipment setup. A series of calibration scans<br />
would be done, which provides a quick mechanism to calibrate the ET<br />
instrumentation and check overall system operation. The inspection and
- 200 -<br />
display parameter set could then be loaded from cassette or entered via the<br />
menu. Feature stepping locations are normally based on the positions of tube<br />
supports and baffle plates. If these positions are known from drawings of<br />
the heat exchanger, the parameter sets could be prepared prior to the<br />
inspection. Otherwise, the information can be acquired by scanning to the<br />
end of the tube with evenly spaced feature stepping and the locations of<br />
supports and baffle plates determined from the printout. After the parameter<br />
set is generated, it would be stored on cassette and repetitive inspection<br />
scanning would then take place.<br />
3.2 Generation of a Parameter Set<br />
The generation of a parameter set is accomplished through the menu. The<br />
operator responds to a series of questions to determine the following scan<br />
parameters:<br />
a) Scan length and speed/location profile.<br />
b) The location of blind regions, which are regions where the data is not<br />
displayed on the impedance images and can be used to blank out the<br />
signals generated as the probe enters the calibration tube.<br />
c) The number of features displayed and stepping positions. Stepping can<br />
occur at operator entered positions, evenly spaced positions, on speed<br />
changes, after blind regions, or automatic stepping upon threshold<br />
detection.<br />
d) Options for system control (what to do on slip or jam conditions, and<br />
whether the operator or analyst or both control the start of scan).<br />
e) Tubesheet information for tubesheet recognition if required.<br />
After each parameter set is generated, it is automatically printed for<br />
permanent reference.<br />
3.3 Calibration Scanning<br />
Calibration of the eddy current instrument for the probe in use is aided by<br />
the R'Eddy Record 'Calibrate and Scan' program. The system requires minimum<br />
operator input and produces a hardcopy printout, as seen in Figure 3. In<br />
this case, the operator requested a 9 cm scan with 3 separated features,<br />
whlci reflects the calibration tube used. This calibration tube contained a<br />
through-wall hole, an 0D groove and an ID groove. The upper row of Impedance<br />
features was obtained by plotting X vs Y with stepping occurring as the probe<br />
passed the 3 and 6 cm positions. Below the impedance features are the<br />
individual X and Y component traces. A row of 'tic' marks below the Y<br />
component trace indicates the stepping location of each feature with the last<br />
'tic' mark denoting end of scan. Below the 'tic' marks lies bookkeeping<br />
information. The scan parameters used in the calibration run are<br />
automatically printed after being entered; thus both permanent parameter set<br />
information and the calibration scan hardcopy are obtained. No tube number<br />
Is associated with a calibration scan.
3.4 Inspection Scanning<br />
3.4.1 General Tube Scanning<br />
- 201 -<br />
Scanning can take place as soon as a parameter set has been generated and<br />
calibration has taken place. The repetitive scan task is invoked via the<br />
menu. Entry back to the menu is available to the analyst at the end of each<br />
scan.<br />
A model of a heat exchanger tube is seen in Figure 4. This tube contained<br />
two defects, a dent located under a broach plate and fretting wear (20%<br />
eccentric) under a baffle plate. The calibration tube and retract limit<br />
switch used to zero the distance count are shown.<br />
The printout collected from the mockup heat exchanger tube scan is shown in<br />
Figure 5. The first feature contains the calibration signals. The two<br />
defect signals are easily picked out of the printout. Twelve features were<br />
requested along with tubesheet recognition. The label 'Full Length Scan 1<br />
indicates an inspection scan. The label L:555 denotes that the maximum<br />
length of scan was 555 cm and the label T:555 indicates the tubesheet<br />
recognition stopped the probe at 555 cm. If desired, up to 50 impedance<br />
images could be displayed. If data collection and display could not meet the<br />
0.4 mm data resolution, the message 'Invalid Scan' would be displayed. The<br />
printout can be written upon (analyst notes) and photocopied. A complete,<br />
easy to access inspection record is thus available to reassure the station<br />
manager and regulatory agencies about inspection credibility.<br />
3.4.2 Scan of Selected Tube Regions<br />
The system gives the operator a method of scanning up to 10 tube regions with<br />
expanded signal traces on the X and Y traces of the display. This option,<br />
called 'Expanded Scan', suppresses the signals from tube regions not<br />
requested for display. As in the full length scan, the calibration tube<br />
signals appear in the first feature. An expanded scan display of the two<br />
regions of the mockup tube containing defects appears in Figure 6.<br />
4. VERIFICATI<strong>ON</strong> OF ANALYSIS AND INSPECTI<strong>ON</strong> PROCEDURE<br />
Verification of both inspection procedure and data analysis requires<br />
recording the original signals and their interpretation as permanent records.<br />
Analog FM recording of data including encoding of tube number, print number<br />
and distance markers, acts as a primary record and retrieval of selected<br />
records is automated. By using the voice synthesized tube number on the<br />
voice channel of the FM recorder, a taped record can be retrieved without the<br />
R'Eddy Record system. The hardcopy printouts of inspection parameters and<br />
data display records with analyst remarks can be bound and act as secondary<br />
records. A third record set is available, comprised of standard strip chart<br />
recordings of data with an ana.log bar code of tube number, print number and<br />
distance pulses on a separate channel. All these records are automatically<br />
generated and make verification a straight forward task.
5. PLAYBACK<br />
- 202 -<br />
A tape playback program allows easy access to R'Eddy Record data on magnetic<br />
tape. High speed search for up to 10 records identified by tube or print<br />
number, with record replay and sequential record display after search<br />
completion allows great flexibility In playback. Three sets of single or<br />
multi-frequency data can be retrieved and regular R'Eddy Record displays<br />
produced. This allows detailed post inspection analysis if required. The<br />
tape playback does not require the use of the probe drive control subsystem.<br />
6. SYSTEM APPLICATI<strong>ON</strong>S AND FIELD EXPERIENCE<br />
The system can be used for a wide range of tube inspections. System set up<br />
time and calibration time are minimal. The set-up of the Inspection system<br />
components typically takes 2 experienced personnel approximately 30 minutes.<br />
Additional setup time is required for instrumentation calibration. An<br />
experienced inspector can enter a particular scanning parameter set in<br />
approximately 2 minutes, provided feature stepping distances are known<br />
beforehand. The highly regular scans and displays are a valuable Inspection<br />
asset. The capability of providing on-line analysis coupled to a high tube<br />
scan repetition rate has demonstrated an inspection time saving of up to 40%.<br />
This is based on three field Inspections, the smallest of which was approximately<br />
450 tubes. Experience has shown that 400 tubes per 8 hour shift Is a<br />
reasonable expectation. A scan repetition rate of 1 tube/minute suggests<br />
that a higher number of inspected tubes per shift can be achieved under<br />
optimum conditions. Because the system was designed and constructed as a<br />
prototype, minor problems have occurred, but the overall system performance<br />
has been very good.<br />
The probe drive modular design has proven to be very important. Transportation<br />
and setup are now done more quickly, and with less physical effort. The<br />
new technology used in the friction drive components has enabled slippery,<br />
wet push tubes to be pushed and pulled through even partially clogged heat<br />
exchanger tubes. The jam detect feature on the probe drive controller has<br />
prevented the drive from damaging push tubes. No mechanical malfunction of<br />
this probe drive has yet occured.<br />
An example of inspection data from a condenser is shown in Figure 7. This is<br />
a direct photocopy of the silver-oxide coated paper used by the system<br />
printer. An analysis of the eddy current signals indicates there Is substantial<br />
steam erosion between the second and third tube supports (between 25%<br />
and 50% of wall loss), and lesser amounts before the first tube support, and<br />
after the third tube support. A through-wall defect is suspected just prior<br />
to the second tube support.<br />
7. CURRENT SYSTEM STATUS<br />
The R'Eddy Record inspection system is currently in the process of being<br />
licensed to a manufacturer of eddy current instrumentation. It is anticipated<br />
this sophisticated inspection system will be commercially available in<br />
the near future.
s<br />
KEYBOARD<br />
HEAT<br />
EXCHANGER<br />
„ET PROBE<br />
PROBE DRIVE PROBE DRIVE COMPUTER OR<br />
INTERFACE COHIHQLLER C<strong>ON</strong>SOLE ~<br />
Figure 1: Configuration of Probe Drive Control Subsystem<br />
with Probe Manipulation by Operator<br />
PROBE DRIVE<br />
C<strong>ON</strong>TROLLER<br />
|- [g] CASSETTE DECK<br />
•JÏ I TV M<strong>ON</strong>ITOR<br />
US VIOEO<br />
PRINTER<br />
TERMINAL<br />
KEYBOARD<br />
C<strong>ON</strong>SOLE<br />
ZJ (ZJ\ TAPE RECOROER<br />
STRIP CHART RECORDER<br />
))) AUDIO<br />
Figure 2: Configuration of Data Processing<br />
Subsystem During Inspection<br />
o
" *<br />
0 0<br />
MOkE*<br />
i<br />
SUN! SOW<br />
A<br />
1<br />
3*<br />
~V<br />
i<br />
1<br />
CALIBRATE<br />
1<br />
p»oeî<br />
00 GROOVE *<br />
J<br />
A1UH flUN<br />
*<br />
10 W00VE<br />
s«r<br />
Figure 3: Typical Printout when a Calibration Tube with<br />
Three Defects is Scanned (Note: *Denotes Comments<br />
added for Clarity, not Generated by the System)<br />
3«5 P77Jp77i BAFFLE PLATE<br />
Si I<br />
5E0<br />
S35£<br />
555<br />
Zl BAFFLE PLATE<br />
1 FRETTING WEAR.<br />
30% ECCENTRIC<br />
445 V/>\r7-7\ BAFFLE PLATE<br />
223 BAFFLE PLATE<br />
ENVIR<strong>ON</strong>METAL<br />
SHICLO<br />
ID<br />
0?<br />
HOLE<br />
«TRACT -"—<br />
LIUIT SWITCH<br />
DISTANCE (cm)<br />
BROACH PLATE<br />
|4j 3R0ACH PLATE<br />
BROACH PLATE<br />
TUBE<br />
SHEET<br />
CALIBRATI<strong>ON</strong><br />
TUBE<br />
Figure 4: Mockup of Heat Exchanger Tube<br />
O<br />
-P-
x.y FEATURES<br />
IIMPtOANCE PLANE)<br />
X-CGMP<strong>ON</strong>ENT<br />
VERSUS OISTANCE<br />
IT-COMP<strong>ON</strong>ENT<br />
VERSUS OISTANCE<br />
STEPPING MARKERS<br />
CALIBRATI<strong>ON</strong>'<br />
SIGNALS<br />
r<br />
lr<br />
SLANT HOW<br />
123/45<br />
- 205 -<br />
IUBESHEH* OENT A! BROACH* • NORMAL BROACH*<br />
r<br />
PLAIE PLATE<br />
U-8EN0 SUPPORT* NORMAL BAFFLE* 20V. FRETTINO WEAR<br />
PLAH UNOER 8AFFLE PLATE<br />
r<br />
C<br />
PRINT IS<br />
21-JUH-l)<br />
LA8tL HOTl.tG-1<br />
FREU 210 KHZ<br />
L VA T VA<br />
ENVIR<strong>ON</strong>MENTAL<br />
SHIELD<br />
FULL iCNGln SCAH<br />
TUBESHEET<br />
c<br />
afOO»<br />
RECORO<br />
(BNt<br />
Figure 5: Printout Generated when Testing the Mockup Tube.<br />
(Note': *Denotes Comments Added for Clarity, not Generated by the<br />
System).<br />
123/45<br />
\i—•<br />
Figure 6: Printout Generated by an Expanded Scan of Two Selected Regions<br />
of the Mockup Tube Containing Defects. Only the Calibration<br />
Signals (0-9 cm region) and Signals in Regions 70 to 90 cm and<br />
385 to 405 cm are Displayed.<br />
(Note: ^Denotes Comments Added for Clarity, not Generated by the<br />
System).<br />
V
SLHHT POU<br />
422/14<br />
- 206 -<br />
x •*<br />
F-RIHT 33 PRO8t:yC015 )<br />
15 FIJLL Lt:!GTH :t<br />
•d>)<br />
R-EDDY<br />
RECORD<br />
CRML<br />
Figure 7: Photocopy of Printout Generated During Actual Field Inspection<br />
using the R'Eddy Record System. Comments were Written Directly<br />
onto the Printout by the Analyst.
- 207 -<br />
WET CHANNEL MEASUREMENT OF PRESSURE TUBE TO CALANDRIA TUBE<br />
SPACING IN CANDU REACTORS<br />
J.tf. Sado<br />
Ontario Hydn.0<br />
ToA.on.to, Ontario<br />
ABSTRACT<br />
The pressure tube (PT) to calandria tube (CT) spacing in CANDU reactors is an<br />
important parameter that relates to the general condition of the fuel<br />
channels. The measurement system that was developed to measure this parameter<br />
during the wet channel inspections of Pickering Units 1 and 2 is described in<br />
this paper. A send-receive eddy current probe was designed which is<br />
primarily sensitive to variations in PT/CT spacing but is also affected by<br />
pressure tube wall thickness. A computer simulation showed that the phase<br />
angles of the response to these variables are similar for all usable<br />
frequencies, thus eliminating the possibility of mul tifrequency compensation.<br />
A marriage of technologies was proposed involving the ultrasonic measurement<br />
of wall thickness values which are then used to extract the spacing<br />
information from the eddy current signal. The accuracy of the system is<br />
aproximately ±(30% +.lmm) which has been sufficient to determine if and where<br />
any of the pressure tubes have come in contact with their calandria tube.<br />
Field experience with the new system is discussed and areas for development<br />
are also outlined.<br />
1. INTRODUCTI<strong>ON</strong><br />
Each fuel channel in commercial size CANDU Reactors is installed horizontally<br />
and contains an unsupported center span approximately 6 meters long<br />
consisting of a pressure tube surrounded by a dry gas and enclosed by<br />
calandria tube. These tubes are firmly centered at the rolled joint positions,<br />
that is at the ends, but spacers (garter springs) are required to ensure<br />
proper separation in the center span. After the reactor has been in operation<br />
for some time, the pressure tube will exhibit sagging between the garter<br />
springs as shown in Figure 1. However, if a garter spring is out of position<br />
then accelerated sagging may occur resulting in premature pressure tube (PT)<br />
to calandria tube (CT) contact.<br />
Two techniques of measuring the PT/CT spacing (gap) have been investigated<br />
during the development of a system for the remote inspection of pressure tubes<br />
(CIGAR) /I/. One method processes the output of an eddy current probe during a<br />
full 360° rotational scan in order to arrive at an estimate of the PT/CT<br />
eccentricity. The second method utilizes both an eddy current probe and an<br />
ultrasonic probe to produce a direct measurement of the change in gap from a<br />
reference point.
- 208 -<br />
The recent failure of a Zircaloy-II pressure tube in Pickering Unit 2<br />
resulted in a decision to construct a simplified version of CIGAR<br />
(CIGARette)/2/ in order to inspect the condition of the remaining fuel<br />
channels in both Units 1 and 2 . As part of the program it was decided to<br />
develop the direct gap method since it showed the potential to be more<br />
accurate over a wider range of values and because the measured data<br />
corresponds in a more straightforward manner to the unknown gap. This is an<br />
important advantage during the interpretation of data and also when trouble<br />
shooting in the field.<br />
2. THE EDDY CURRENT PROBE<br />
The eddy current gap probe contains two differentially connected receive coils<br />
encircled by a send coil as shown in Figure 2. All of the materials used in<br />
the construction of the probe are radiation resistant in order to withstand<br />
the 10 rad/hour fields in the shutdown reactors. Although the probe was<br />
originally designed for use in fuel channels with Zr-2.5% Nb pressure tubes,<br />
the offsetting wall thickness and resistivity changes in Zircaloy-II tubing<br />
result in similar eddy current behaviour in fuel channels of Pickering 1 and 2<br />
and allow the existing design to be used.<br />
2.1 Computer Predicted Response<br />
The complex plane response of the probe (Figure 3) was simulated by a<br />
computer program based on the equations and coding developed by Dodd for a<br />
similar probe configuration over multilayered planar conductors/3/. The<br />
diameter of the gap probe is small in comparison with that of the pressure<br />
tube and so it was assumed that the planar case would closely approximate the<br />
actual situation. Another departure from the configuration used by Dodd is<br />
that instead of placing the receive coils symetrically within the send coil,<br />
their position is adjusted during manufacture so as to produce a minimum<br />
signal when the probe is in contact with a piece of Zr-2.5% Nb pressure tube.<br />
This situation more closely approximates the probe's normal operating<br />
environment thus minimizing the offset signal and making available a larger<br />
overall gain from the eddy current instrument. These approximations plus the<br />
neglect of the computer program to take into account the effect of the<br />
stainless steel wear cap that surrounds the probe may result in some departure<br />
of the actual response from the predicted response.<br />
The frequency and phase settings of the eddy current instrument are chosen so<br />
as to place the signals due to lift off changes on the X axis and those due to<br />
gap changes predominantly on the Y axis. Therefore, monitoring only the Y<br />
component of the eddy current signal practically eliminates any lift off<br />
influence and substantially reduces that of pressure tube resistivity. Signals<br />
due to pressure tube wall thickness changes are also primarily in the Y<br />
direction, but the exact phase angle is influenced somewhat by the magnitude<br />
of the PT/CT spacing beneath the probe. This condition persists over a wide<br />
range of operating frequencies, illustrating why multi frequency techniques<br />
were originally rejected for wall thickness compensation. Instead, an<br />
ultrasonic system will be used to measure the pressure tube wall thickness<br />
allowing the gap information to be recovered from the eddy current signal.
2.2 Primary Factors<br />
- 209 -<br />
Computer simulation results verified that vectorial addition was valid for<br />
signals due to gap and wall thickness changes, thus allowing these effects to<br />
be studied independently. A number of sets of data were collected, each at a<br />
different fixed wall thickness and containing the response of the eddy current<br />
probe to a range of gap values. One such set is shown in Figure 4 where the<br />
non-linear response of the probe is evident. In order to relate these<br />
independent data sets and determine the effect of wall thickness changes, it<br />
is only necessary to determine the response of the probe to wall thickness at<br />
one fixed gap. This was done by performing a rotational scan on a pressure<br />
tube sample with a large wall thickness variation around the circumference and<br />
no calandria tube (infinite gap). The result, shown in Figure 5, was a good<br />
linear fit with a regression coefficient of 0.9887. The slope of this line<br />
was used to shift the sets of previously collected data so that they all<br />
referenced a common balance point at infinite gap and 5.0 mm wall thickness.<br />
Straight lines were fitted to the points of common gap with a great deal of<br />
success, the smallest regression coefficient being 0.9960. This overview,<br />
shown in Figure 6, formed the data base of the probe's response which was to<br />
be used during field inspections.<br />
2.3 Pressure Tube Resistivity<br />
The calculated complex plane response of the probe indicated that pressure<br />
tube resistivity could significantly influence the eddy current signal if<br />
large enough variations were encountered during the inspections. This<br />
possibility was explored by measuring the resistivity and wall thickness<br />
around the circumference of a pressure tube sample and then scanning the same<br />
section with the gap probe in order to compare both influences.<br />
The resistivity scan in Figure 7c shows a well defined change of about 1<br />
just after the 180° mark. Using Figure 3, it was calculated that a<br />
resistivity change of this magnitude should cause approximately the same<br />
effect in the Y component of the probe response as a 0.17 mm wall thickness<br />
change. Figure 7a shows that this is about half of the total wall thickness<br />
variation measured in the pressure tube sample. However, the response of the<br />
gap probe shown in Figure 7b does not agree with this prediction as the wall<br />
thickness influence was followed very closely with almost no sign of the<br />
resistivity influence.<br />
The apparent insensitivity of the probe to resistivity variations was<br />
encouraging for the immediate goal of the Pickering 1 and 2 inspections,<br />
however, further investigations are required for the long term development of<br />
the gap measurement system.<br />
2.4 Temperature<br />
The temperature sensitivity of the probe in its operating environment and in<br />
isolation are shown in Figures 8 and 9 respectively. The response in Figure 9<br />
is somewhat unexpected since send-receive probes are generally expected to be<br />
insensitive to temperature variations due to the small effect that variations<br />
in coil resistance have on the output signal. In this case the response is<br />
thought to be caused by a physical deformation of the probe and is currently<br />
under investigation. The linear response of -0.036V/°C obtained from the
- 210 -<br />
probe in its operating environment is the combined result of this effect plus<br />
the temperature related resistivity changes in the pressure tube and in the<br />
stainless steel wear cap.<br />
The temperature variations in each fuel channel are expected to be minimal due<br />
to the flow of coolant that exists during shutdown. As a result the<br />
temperature sensitivity of the probe should not cause finy major problems<br />
during inspections and may be covered by the stated accuracy of the results.<br />
2.5 Probe Variations<br />
The data presented thus far were obtained with one specific probe.<br />
Experiments comparing subsequent probes to the original showed a constant<br />
difference in signal outputs due to identical gap variations. The wall<br />
thickness and temperature sensitivities of the alternate probe were also<br />
verified to differ from the reference probe by the same factor. This allowed<br />
the assignment of a compensation factor to each probe generally ranging in<br />
value from 0.7 to 1.1, so that for an identical set of circumstances the<br />
change in output of the alternate probe could be referred to a standard.<br />
3. GAP ESTIMATI<strong>ON</strong><br />
The procedure for estimating gap assumes that lift off, resistivity,<br />
temperature or any other factors do not influence the eddy current signal<br />
obtained in channel. Before the inspection data can be used, these signal<br />
values must be related to those recorded in the data base, and so it is<br />
necessary to select a reference point in each channel where gap, wall<br />
thickness and the eddy current signal value are known. This generalized<br />
requirement does not necessarily involve the point at which the instrument is<br />
balanced, thus reducing the function of balancing to one of locating the<br />
output trace somewhere on the strip chart record. The wall thickness and eddy<br />
current values may be measured at the reference point but the gap must be<br />
inferred from the structure of the fuel channel. Two good reference points<br />
are located near the ends of the pressure tube, as close as possible to the<br />
rolled joints. Here the pressure tube is centered firmly in the calandria<br />
tube and the gap may be assumed to be 8 mm at all rotational positions.<br />
Once a reference point has been chosen in the channel, the corresponding point<br />
in the data base may be located using the wall thickness and gap information.<br />
From this point, the changes in eddy current signal and wall thickness data,<br />
as measured when the probes travel from the reference point to the measurement<br />
point in-channel, can be followed individually to arrive at the mesurement<br />
point in the data base. It is quite likely that the measurement point in the<br />
data base will end up between two lines of constant gap in which case simple<br />
linear interpolation is used to determine the estimated gap value. The number<br />
of lines of constant gap is large enough that this approximation to the true<br />
non-linear nature of the gap response will not introduce any significant<br />
error. This procedure may easily be implemented on a computer using only the<br />
stored slopes and intercepts of the lines in the data base.
4. FIELD EXPERIENCE<br />
- 211 -<br />
The gap measurement subsystem was incorporated into the CIGARette inspection<br />
equipment along with garter spring and flaw detection capabilities for use in<br />
Pickering Units 1 and 2/2/. Unfortunately the gap measurement function was<br />
not one that worked right away.<br />
4.1 Wall Thickness Measurements<br />
Wall thickness measurements were originally to be obtained using a modified<br />
digital thickness gauge and a highly focussed radiation resistant ultrasound<br />
probe. This method is similar to that planned for the CIGAR inspection system<br />
with the exception that a technician would be responsible for logging the<br />
data from the digital display instead of a computer. Unfortunately, stable<br />
readings could not be obtained and eventually the instrument used for the flaw<br />
detectors, a Branson KB-6000 was utilized. Thickness measurements using this<br />
modified system required that the probe be stationary so that the time beween<br />
the reflection peaks of the inner and outer pressure tube walls could be<br />
measured on a CRT display. The result was then converted to a thickness value<br />
by multiplying by 2.389 mm/p s (the velocity of sound in Zircaloy-II divided by<br />
2). This slow procedure limited the number of points that could be measured<br />
in each channel and introduced a random observation error that was estimated<br />
to be about±l% or a little more than the thickness of the screen trace.<br />
4.2 Eddy Current-Signal Quality<br />
Although the procedures for calculating gap outlined in section 3 indicate<br />
that data need only be collected at the reference and measurement points, it<br />
was soon realized that the eddy current signal would have to be recorded<br />
during a continuous scan in order to better ascertain the signal quality.<br />
Once this procedure was adopted, it became evident that in many cases the eddy<br />
current signal contained excessive noise and drift which was traced to<br />
various problems with the 80 meters of signal cabling and 8 connection points<br />
in the CIGARette system. Finally, diligent trouble shooting and the<br />
availability of tested spare components reduced this problem significantly,<br />
resulting in smooth and stable signal traces for many of the channel<br />
inspections.<br />
4.3 Types of Scans<br />
Two types of scans were used during the inspections, an axial scan along the<br />
bottom of the channel to establish a gap profile and a rotational scan,<br />
usually used at points of interest in an attempt to confirm the axial scan<br />
results. The rotational scans were evaluated assuming an infinite gap at the<br />
12 o'clock position but were generally not as successful as *he axial scans.<br />
The reasons for this difficulty were never firmly established, however the<br />
quite rapid wall thickness changes around the circumference of the pressure<br />
tube and the relatively few data points collected required exacting<br />
performance from the system for proper results. Axial scans were much more<br />
successful with the added benefit that the eddy current signal trace was<br />
usually indicative of the gap profile since very few significant axial wall<br />
thickness variations were encountered.
- 212 -<br />
The eddy current and wall thickness probes are located at the same rotational<br />
position on the inspection head, however they could not be used simultaneously<br />
during an axial scan due to the conflicting requirements of the two systems<br />
(the eddy current scan must be continuous whereas the ultrasonic probe must be<br />
stopped for each measurement). This separation of the data collection scans<br />
caused some problems on numerous occasions as the inspection drive mechanism<br />
would tend to slip rotationally and corkscrew without affecting the rotational<br />
display counter. This had serious consequences for the gap system since a<br />
small rotation in a typical pressure tube during an axial scan would result<br />
in a sharp rotational wall thickness change being sensed as a gradual axial<br />
change when in fact there was none. Such a situation, especially if only<br />
experienced during the eddy current or wall thickness scan, would result in<br />
misleading gap results.<br />
4.4 Sample Results<br />
The axial gap profile of a fuel channel exhibiting accelerated pressure tube<br />
sag due to the shift of a garter spring from its design location is shown in<br />
Figure 10. The gap system was originally conceived to discover situations<br />
such as this where the PT/CT spacing is very small in some areas.<br />
4.5 Sources of Errors<br />
Errors can generally be grouped into two categories, those that affect the<br />
accuracy of the measurements and those that affect the reliability of the<br />
system. The latter group tend to severely distort the data and would include<br />
such conditions as; cabling problems, misalignment of the wall thickness and<br />
eddy current probes, operator or analyst error and excessive temperature or<br />
resistivity changes. Many times these problems will cause the data or results<br />
to be unusual or even unreasonable when compared to historical data and known<br />
characteristics. Some checks that can be made to give credibility to the data<br />
are listed here:<br />
1. The eddy current signal trace should be smooth with no<br />
abrupt discontinuities or excessive noise or wandering.<br />
2. The wall thickness data is usually fairly constant in an<br />
axial direction.<br />
3. Some sign of both gap and wall thickness changes should<br />
be seen in the eddy current signal trace since it is<br />
highly unlikely that gap and wall thickness would vary in<br />
exactly opposing manners so as to cause the eddy current<br />
signal to remain constant.<br />
4. The gap profile should be fairly smooth.<br />
5. The gap near the rolled joints at both ends of an axial<br />
scan should be 8 mm.
- 213 -<br />
6. The axial gap profile should be similar when computed<br />
using either end of the pressure tube as a reference.<br />
Note that if the gap at both ends of the pressure tube<br />
are not equal to 8 mm, repeating the calculations using<br />
the other end as a reference will not just shift the gap<br />
profile by the difference. The shape of the profile will<br />
also change due to the non-linearity in the data base.<br />
The accuracy of the gap measurements is primarily determined by the effect of<br />
the ±0.05 mm uncertainty in the wall thickness measurements at both the<br />
reference and measurement points. Referring to Figure 3, the uncertainty in<br />
the location of the reference point, which must be located along the line of<br />
constant 8 mm gap for axial scans, can also be expressed in terms of the eddy<br />
current signal voltage as approximately ± 0.08V. Doing so allows the<br />
uncertainty to be transferred to the measurement point and to consider the<br />
reference point fixed. The additional uncertainty in the measurement point<br />
due to its associated wall thickness measurement can also be expressed in<br />
terms of the eddy current signal, but the exact conversion factor will depend<br />
on the size of the gap. The wall thickness measurements can be assumed to be<br />
indépendant and normally distributed and so both of the uncertainties in the<br />
measurement point may be combined as the square root of the sum of the<br />
squares. The result will vary from a high of +0.11V for a gap of 8 mm to a<br />
low of +0.083V for a gap of 0 mm. An error bound of ±(30% + 0.1 mm) on<br />
computed gap values approximates this overall uncertainty and is the magnitude<br />
of the error bars that have been placed around the sample results in Figure<br />
10.<br />
S. AKEAS ft» DEVELOPMENT<br />
The field experience in Pickering has served to solidify the gap measurement<br />
procedure and also to uncover various weak points in the equipment which<br />
effect the results. Further development is being done using the CIGAR<br />
inspection system with the emphasis on Zr-2.5%Nb pressure tubes now that the<br />
Zircaloy-II tubes in Pickering Units 1 and 2 are to be replaced.<br />
The use of the CIGAR inspection system addresses a number of equipment<br />
reliability problems encountered in Pickering which made it difficult to<br />
isolate and solve problems originating with the measurement technique. The<br />
most important advancement is that both the eddy current and wall thickness<br />
data are collected automatically during the same axial scan. This ensures the<br />
proper alignment of both probes and enables a large amount of data to be<br />
collected which can then be smoothed, reducing any unbiased error in the<br />
individual measurements. Another improvement is the gentler and largely<br />
automatic handling of the signal cables and their connections which has<br />
resulted thus far in stable and reliable eddy current signals. Very promising<br />
and repeatable results have been obtained to date but more testing is still<br />
required. Some development work still remains to be done in the area of<br />
pressure tube resistivity variations and in the quality of the eddy current<br />
probe, both in terms of its temperature sensitivity and its consistency in<br />
manufacture.
6. SUMMARY<br />
- 214 -<br />
The failure of a pressure tube in Pickering Unit 2 created a need to measure<br />
the pressure tube to calandria tube spacing in the remaining Zircaloy-II fuel<br />
channels in order to help ascertain their condition. A technique, previously<br />
investigated for CIGAR and which utilizes ultrasonic wall thickness<br />
measurements to compensate the response of an eddy current probe was<br />
integrated into the CIGARette inspection system. Some procedural and<br />
reliability problems were initially experienced but eventually the system was<br />
made to work with an accuracy of about ±(30% + .1 mm). This was sufficient to<br />
be able to discern regions where the pressure tube had come in contact with<br />
the calandria tube. Work is continuing to integrate this capability into the<br />
CIGAR inspection system for use with the rest of the commercial size CANDU<br />
reactors. It is hoped that the use of CIGAR will increase both the<br />
reliability and the accuracy of the measurements although final results are<br />
still pending.<br />
ACKHOWLEDGEMEHTS<br />
Appreciation is extended to Dr. D. Leemans of Cross Currents Research & Policy<br />
Consulting Ltd for discussions on the eccentricity method and the design of<br />
the gap probe and to H. Ghent of CRNL for his assistance with the computer<br />
simulation of the eddy current response.<br />
REFERENCES<br />
1. J.A. Baron, M.P. Dolbey, J.H. Erven, D. Booth, and D. Murray, I. Mech. E.<br />
Conf. Publication No. C139/82.<br />
2. M.D.C. Moles, M.P. Dolbey and K.S. Mahil. Paper presented at this<br />
conference.<br />
3. C.V. Dodd et al. 0RNL Report No. TM-4107, Oak Ridge, Tennessee, 1973.
ROLLED JOINT<br />
CALANDRIA TUBE<br />
rs.5<br />
- 215 -<br />
PRESSURE TUBE<br />
6.03 H<br />
.5<br />
MM,<br />
ROLLED JOINT<br />
FIGURE 1 FUEL CHANNEL EXHIBITINC PRESSURE TUBE SAC<br />
Pressure Tube<br />
Recei ve<br />
Coils Send Coil<br />
FIGURE 2<br />
Send<br />
Coil<br />
THE EDDY CURRENT GAPPROBE<br />
ä<br />
3<br />
Receive<br />
Coils<br />
(Wound in<br />
Opposition )<br />
> =J Pressure<br />
Calandria Tube Tube
+2.5%p<br />
Lift Off •*<br />
- 216 -<br />
+5% Wall Thickness<br />
-5% Wall Thickness<br />
-5% Wall Thickness<br />
FIGURE 3<br />
Cap = 8 mm<br />
p = 74 \iti-cm (Zirc II)<br />
Wall Thickness = 5.0 mm<br />
COMPUTER PREDICTED COMPLEX PLANE<br />
RESP<strong>ON</strong>SE OF THE CAP PROBE<br />
-2.5%p
in<br />
in Vol<br />
lent i<br />
Compor<br />
10<br />
c<br />
o<br />
o<br />
0<br />
-0 .5<br />
-1 .0<br />
-1 .5<br />
-2.0<br />
0.3<br />
-0.3 -<br />
- 217 -<br />
Gap in mm<br />
FIGURE 4<br />
Zirc II<br />
Wall Thickness =5.Omm<br />
RELATI<strong>ON</strong>SHIP BETWEEN EDDY CURRENT SIGNAL<br />
AND GAP (NOT ALL DATA SHOWN)<br />
Zirc II<br />
Cap = S mm<br />
Slope = -1.71 V/mm<br />
R = 0. 9887<br />
5.1 5.2 5.3 5.4<br />
Wall Thickness in mm<br />
FIGURE 5<br />
RELATI<strong>ON</strong>SHIP BETWEEN EDDY CURRENT<br />
SIGNAL AND WALL THICKNESS<br />
5.5
0<br />
-0.2<br />
-0.4<br />
-0.6<br />
-0.8<br />
-1.0<br />
c .1 2<br />
o<br />
o -1.4<br />
-1.6<br />
-1.8<br />
-2.0<br />
-2.2 -<br />
-2.4<br />
5.0<br />
5.1<br />
- 218 -<br />
Wall<br />
_L<br />
5.2<br />
Thickness<br />
5.3<br />
in mm<br />
Cap in mm<br />
5.4 5.5<br />
FIGURE 6<br />
EXPERIMENTALLY DERIVED DATA BASE USED FOR THE<br />
COMPUTATI<strong>ON</strong> OF CAP IN ZIRC II CHANNELS<br />
(NOT ALL LINES SHOWN)
u<br />
- 219 -<br />
180°<br />
Rotational Position<br />
FIGURE 7a WALL THICKNESS OF PRESSURE TUBE SAMPLE 841-207(ZIRC II)<br />
-375<br />
180°<br />
Rotational Position<br />
360'<br />
FIGURE 7b SCAN OF SAMPLE 841-207 USING THE EDDY CURRENT CAP PROBE<br />
76.8<br />
0° 180° 360 e<br />
Rotational Position<br />
FIGURE 7c RESISTIVITY OF PRESSURE TUBE SAMPLE 841-207<br />
360--
0<br />
-0.25<br />
-0.50<br />
-0.75<br />
-1.00<br />
-1.25<br />
i<br />
1/1<br />
•<br />
1<br />
c<br />
iponeni<br />
Con<br />
> ~<br />
- 220 -<br />
Rate of Temperature Rise<br />
0.5°C/min<br />
Slope =<br />
-0.036V/°C<br />
10 30 50<br />
Temperature in °C<br />
FIGURE 8<br />
TEMPERATURE SENSITIVITY OF THE<br />
CAP PROBE MOUNTED IN ITS HOLDER<br />
AND AGAINST ZIRC II TUBING<br />
.00<br />
0 .75<br />
0 .50<br />
0 .25<br />
0<br />
Slope = 0. 035 V/°C A<br />
/<br />
\f i I i<br />
10 30 50<br />
Temperature in °C<br />
FIGURE 9<br />
TEMPERATURE SENSITIVITY OF THE<br />
GAP PROBE AL<strong>ON</strong>E
»6<br />
5<br />
^ 4<br />
13<br />
2<br />
IL<br />
0<br />
- 221 -<br />
•Rolled Joints<br />
£t 11Uiiiiitii<br />
Detected Positions<br />
Of Carter Springs<br />
X<br />
1,X<br />
Rotational Position = 180°<br />
Error Bars are t (30% + 0. I mm)<br />
t K<br />
3 4 5 6 7<br />
Axial Position As Read By Cigarette Drive (m)<br />
FIGURE 10<br />
GAP RESULTS FROM CHANNEL P1-G14<br />
I<br />
f A •. A<br />
X X X<br />
X
- 222 -<br />
CHECKING FOR CRACKS IN GAS TURBINE ROTOR DISCS<br />
J. ran den Aude I and 4.8. Niebciq<br />
d'eifiufflicuic Canada Inc., Hamilton, Cnta'iic<br />
1. INTRODUCTI<strong>ON</strong><br />
A gas turbine consists of a compressor and a turbine. The turbine drives the<br />
compressor and a mechanical output shaft. The compressor becomes warm (from<br />
compressing air), the turbine becomes hot from the flame which burns in the<br />
compressed air in the combustion chamber. The higher the flame temperature,<br />
the higher the turbine's efficiency, but the material is also used closer to<br />
its limitations. The turbine discs (Figure 1) appear to have a finite life<br />
and although most may last seemingly forever, there is still a fair percentage<br />
that does not. This means that discs are considered to fail eventually<br />
and that a close watch is to be kept on the disc quality at overhauls, usually<br />
in the field. The type of failure is most often a crack in the tooth-serrations<br />
(Figure 2) and so far, only two nondestructive tests appear useful<br />
for the detection of root cracks, i.e., Pénétrants and Eddy currents. The<br />
former is tedious, messy and difficult to evaluate in tight locations, the<br />
latter is quicker, more definite and very sensitive, but is affected by material<br />
edges and can therefore not test the entire length of the serration.<br />
2. PENETRANT TESTING<br />
The third party is now faced with a possible residue of two previous penetrant<br />
applications in a crack and has to go through an extensive cleaning process<br />
which might or might not remove all the old penetrant residue.<br />
The cleaning process involves glass beading to remove scale, vapour degreasing<br />
to dissolve trapped penetrant and perhaps a wash with acetone to enhance<br />
a difficult cleaning process. At this point, it should be realized that we<br />
are dealing with an entire turbine rotor from which only the turbine blades<br />
were removed. Turbines come in different sizes, but a length of 7 m and a<br />
mass of 12 tonnes, is not uncommon.<br />
The pénétrants currently used are biodegradable and water washable. It was<br />
found that the post emulsifiable pénétrants although superior when properly<br />
used, lose all of their advantages when the emulsif1er is left on too long.<br />
Apart from having no control over an operator in the field, the operator has<br />
a huge component on his hands that has to be treated in its entirety, left to<br />
soak a precise time and then wash. Such a procedure is impossible in the time<br />
slots available. Water washable pénétrants are more tolerant and appeared to<br />
be much better suited for the inspection of large components.
- 223 -<br />
After washing, drying and developing, the examination is another problem.<br />
Often a number of rows of teeth have to be examined and accessibility is limited;<br />
an operator has to work with tiny pieces of mirror attached to popsickle<br />
sticks which he tries to position in the bottom groove, shine the black light<br />
onto them and look at the same time for defects.<br />
Even if defects were found, the familiar relationship between the crack depth<br />
and the amount of penetrant-bleeding is often not there because of the very<br />
tight nature of this type of defect. The cracks may be intergranular and<br />
tightly filled with combustion or corrosion products which did not wash away<br />
in the cleaning operation. The small indication of Figure 3 proved to be<br />
from a substantial flaw (Figure 4).<br />
Operators with less sophisticated cleaning facilities would find even less<br />
cracks. This becomes a problem where no cracks are found in a disc which<br />
does contain them. Fortunately, small cracks are not instantly causing a catastrophic<br />
failure; most cracks grow slowly and are found during the next overhaul<br />
at which time the disc can be replaced. Nevertheless, a flawed disc<br />
which was given a clean bill of health because the testing method was inadequate,<br />
may fail before the next overhaul and cause expensive damage. The test<br />
therefore is inadequate and better testing means ought to be found.<br />
3. EDDY CURRENTS<br />
An obvious solution to the testing of disc teeth seems to be an Eddy Current<br />
test. However, these tests are not so obvious when one actually tries them.<br />
Rotors become slightly scaled and the rough surface so far has caused problems<br />
with the Eddy Current test. Descaling by means of glass-beading has been<br />
practised before an Eddy Current test was attempted and leaves a sort of rough<br />
surface.<br />
Figure 5 shows an Eddy Current signal of a defect-free tooth. There is some<br />
horizontal wandering and an exit signal veering off to the upper left. Figure<br />
6 shows the signal when the probe passes over a defect close to the exit. The<br />
display shows loops which betray defects. The sensitivity setting of the<br />
instrument was relatively low and when increased to find small defects, the<br />
displays would grow much larger than the screen and could no longer be interpreted<br />
(Figure 7).<br />
The Eddy Current tester was adjusted such that the lift-off signal as well as<br />
the wandering of the dot while the probe travels were horizontal on the screen.<br />
The vertical component is then always related with defects and end-of-groove<br />
effects. The vertical component has an output available on the instrument<br />
which was connected to the Y direction of an ordinary storage osciloscope. A<br />
relatively simple fixture was made to sense the position of the probe and fed<br />
to the x-direction of the scope. Figure 8 shows how a differential probe senses<br />
a crack halfway along the tooth. The tooth is 91 mm long, the crack 13 mm<br />
but it shows 20 mm on the base line which means that the probe sees 4 mm more
- 224 -<br />
on the fringes. (The sensitivity setting on the instrument was very low, hence<br />
the defect would show even larger with the sensitivity increased to normal<br />
testing levels.)<br />
4. VERIFICATI<strong>ON</strong><br />
ïhe position of the cracks depends on the design of the turbine. Depending on<br />
the model, the experienced operator could tell ahead where the cracks will be<br />
found. That means that a probe can be designed with the highest sensitivity<br />
at the location of the anticipated cracks. Nevertheless, one needs a probe<br />
which covers a reasonably wide swath, in case defects with another orientation<br />
are present which ought to be detected.<br />
Test blocks were manufactured from homogeneous teeth by Electro Discharge<br />
Machining small artificial defects in the grooves (Figure 9)• One groove contained<br />
notches of different depths for calibration purposes, another groove<br />
contained identical notches which were offset to determine field-of-view of<br />
the ET probes.<br />
It was found before, that absolute probes have advantages when only one phenomenon<br />
is sought. 1 An absolute probe was made with both a wide and a limited-length<br />
field of view. A once-through test with this probe on the reference<br />
notches with respect to their position is shown in Figure 10.<br />
By changing the frequency, the optimum response was found, and it was not surprising<br />
that it was close to the design frequency. It was expected that a<br />
change in ratio of signals from small to large reference notches would also be<br />
found with a change of frequency but it was not (Figure 11).<br />
Since notch 2 is supposed to be exactly 0.2 mm deep and notch 3, 0.4 mm deep,<br />
the ideal signal ratio would be two. From 800 to 900 kHz the ratio was 2.3,<br />
considered the best obtainable. It was also decided that it was probably better<br />
to use a firm signal as a set-up rather than a smaller but noisier one and<br />
the 0.4 mm deep notch was a tentative choice.<br />
At 800 kHz a series of measurements with different instrument sensitivities<br />
was made (Figure 12). This showed that without increasing the noise level,<br />
signals could be boosted to show the 0.1 mm notch (No. 4, Figure 9) clearly.<br />
With this level of sensitivity, any dangerous crack would cause the indication<br />
at the lower sensitivity. The latter appears better for testing since too<br />
many indications would go off-screen if the sensitivity were too high. With a<br />
reduced sensitivity test may be more realistic and could be used to as close<br />
as 5 mm from the end. (Notches are at 20, 40, 60 and 78 mm from the left<br />
edge, Figure 13.) With this setting, the real crack in Figure 13 showed a<br />
start in steps but is fully established at 50 mm and carries on to the end.<br />
(The microscope showed it to fade at 4 mm from the end or 87 mm from the left<br />
which is revealed in the picture although not clearly.)<br />
1. G. Van Drunen and V.S. Cecco, CSNDT Journal, Vol. 5 No. 1, 1983 October
5. C<strong>ON</strong>CLUSI<strong>ON</strong>S<br />
- 225 -<br />
A practical Eddy Current test for disc root cracks is possible with a sensitivity<br />
high enough to detect 0.1 mm deep imperfections. The position of these<br />
defects is shown on the screen of the scope, using relatively simple position<br />
fixtures which have to be made for every type of turbine disc. The defect<br />
size is displayed as distinct measurable spikes, while testing is quick and<br />
precise. Pictures made of questionable teeth are self explanatory and can be<br />
compared with those made of the reference.<br />
Limitations of the test are the lack of definition within 5 mm from the end<br />
and a field-of-view, limited to only approximately A mm wide.<br />
Overall, a greatly simplified more precise and much quicker test is now available<br />
for testing blade roots in discs for cracks.<br />
6. ACKNOWLEDGEMENTS<br />
The encouragement given by Messrs. M. Metala and E. Fuselier of Westinghouse<br />
Electric Corporation have been greatly appreciated.
Figure 1 18 MW gas turbine rotor, partially de-bladed.<br />
S3
- 227 -<br />
Figure 2 Multiple power stage
- 228 -<br />
Figure 3 Weak penetrant indication on carefully prepared sample.<br />
II<br />
Figure 4 Previous small<br />
indication was a large defect.
Figure 5<br />
- 229 -<br />
Eddv Current sienal of defect-free groove.<br />
Figure 6 Large multiturn loop runs into exit signal.
- 230 -<br />
Figure 7 All signals superimposed.<br />
Figure 8 Vertical Eddy Current signal displayed on probe position.
REF NOTCHES <strong>ON</strong><br />
CRACK LINE<br />
CENTRIFUGAL<br />
FOHCE<br />
DISC<br />
Figure 9<br />
- 211 -<br />
Location of reference notches.<br />
LINE WHERE CRACKS OCCUR<br />
NOTCHES OFF CRACK LINE TO<br />
DETERMINE FIELDOF VIEW<br />
• STANDARD REF. NOTCH<br />
2 mm long<br />
0.1 mm wide<br />
0.2 mm deep<br />
Figure 10 Test on groove with reference notches (absolute probe).
- 232 -<br />
Figure 11 Frequency response on reference notches.
- 233 -<br />
Sens.<br />
35<br />
Figure 12 Influence of sensitivity setting on output.<br />
45<br />
55
- 234 -<br />
row 3 sent to 0 4 n<br />
1Vor?d.w<br />
v 3 wi«. to 0 à mm notch<br />
ObVorl div.<br />
(not»; testing prwible ui 5mm from<br />
ind. blade width 91 mini<br />
rtrf notch sensitivity ai -iljon-<br />
MMI cr.jck from 50 mm to tn>l<br />
w:n4 20<br />
Figure 13 Two possible sensitivity settings and a real crack.
- 235 -<br />
EDDY CURRENT INSPECTI<strong>ON</strong> OF MILDLY FERROMAGNETIC TUBING<br />
(U.R. Mayo, J.R. Cantan.<br />
Atomic Enzigy o {, Canada<br />
Chalk Rivzn, Ontatii.o<br />
ABSTRACT<br />
The past decade has seen the development of eddy current probes for<br />
inspection of the mildly ferromagnetic alloy Monel 400. Due to the rapid<br />
advances in permanent magnet technology similar probes have been upgraded to<br />
magnetically saturate, and hence inspect, the duplex stainless steel Sandvik<br />
3RE6O, which has saturation induction more than twice that of Monel 400.<br />
Prototypes of these probes have been tested in three ways: saturation<br />
capability, quality of typical eddy current data, and ability to eliminate<br />
permeability induced signals. Successful laboratory testing, potential<br />
applications, and limitations of these type probes are discussed.
1. INTRODUCTI<strong>ON</strong><br />
- 236 -<br />
In-service eddy current testing (ET) of heat exchanger tubing is often<br />
complicated by the use of magnetic materials such as nickel-copper alloys,<br />
austenitlc-ferritic stainless steels, and mild carbon steel. Since many<br />
factors must be considered when a choice of tubing material is made for a<br />
particular unit, including initial cost and resistance to deterioration,<br />
inspectability is often of secondary importance. For this reason the alloys<br />
Monel 400* and Sandvik 3RE60** are encountered as examples in the nuclear<br />
industry, and mild carbon steel is used extensively throughout industry in<br />
general.<br />
It is commonly known that magnetic materials can only be inspected by ET once<br />
they have been magnetically saturated[l]. This condition can easily be met<br />
during manufacturing testing, but places a severe restriction upon in-service<br />
inspection, where usually only the tube ID is accessible.<br />
We may non-rigorously define materials as being mildly ferromagnetic if they<br />
have saturation induction less than 1 tesla. This puts the alloys Monel 400<br />
and 3RE60 in the mild class. Carbon steel, on the other hand, is highly<br />
ferromagnetic, having saturation induction over 2 tesla. A considerable<br />
amount of ET development work in the Canadian nuclear industry has been<br />
directed toward saturation probes for inspection of mild magnetic materials.<br />
This was largely due to the use of Monel 400 in the Pickering Generating<br />
Station heat exchangers. Probes to inspect Monel 400 tubing have seen<br />
successful use in the field, and recently we have succeeded in developing<br />
probes for 19 mm 3RE60 tubing, which is more difficult to saturate. The<br />
ability to inspect 3RE60 tubing is very likely the highest level to which<br />
probes of this type, using high energy-product permanent magnets, can be<br />
elevated. Highly ferromagnetic mild carbon steel will require some new<br />
approach and is currently the subject of laboratory investigation.<br />
This paper presents data obtained from a prototype probe suitable for<br />
inspection of 19 mm 3RE60 heat exchanger tubing. Impedance monitoring shows<br />
the magnetic tubing is fully saturated by the probe. Comparison of 3RE60<br />
signals (eddy current impedance plane display) to those from a nonmagnetic<br />
316 stainless steel tube of the same dimensions also demonstrates complete<br />
saturation. Finally, stress induced permeability variation signals<br />
effectively "dissappear", as would be expected at full saturation. The<br />
results indicate 3RE60 tubing of this size is now inspectable (with<br />
limitations), representing a further advancement in the in-service inspection<br />
of ferromagnetic materials.<br />
* Reg. trademark of International Nickel Co.<br />
** Reg. trademark of Sandvik Tubular Prod.
2. BACKGROUND<br />
2.1 Theory<br />
- 237 -<br />
The thickness of conducting material inspectable by ET Is limited by the<br />
exponential decay of eddy current density with depth. This is characterized<br />
by a decay constant known as the "standard depth of penetration" (6), given<br />
by [2]:<br />
where:<br />
6 = J— (1)<br />
u ~ electromagnetic field angular frequency<br />
a » electrical conductivity of medium<br />
y - magnetic permeability of conducting medium<br />
As expressed in equation (1), 6 is the depth at which the alternating field<br />
intensity, and hence eddy current density, falls to 1/e (e = 2.71828...) of<br />
its surface value. (This is satisfactory as an approximation, but is only<br />
exact for a semi-infinite conducting medium with uniform external magnetic<br />
field intensity). Since 6 is inversely proportional to the square root of<br />
permeability, it must be much less in magnetic materials than in nonmagnetic<br />
cases. * The former have permeabilities that are typically orders of<br />
magnitude higher than the latter. This is indicative of one of two problems<br />
associated with ET in magnetic materials; eddy currents cannot penetrate the<br />
specimen sufficiently. Since the eddy current technique relies upon the<br />
presence of defects to perturb the flow of currents in the metal, there can<br />
be little sensitivity to sub-surface defects if eddy current density is<br />
vanishingly small. The application of a biasing field will increase the<br />
depth of eddy current penetration, by effectively reducing the permeability<br />
of the sample, but here the second problem arises. Large amplitude signals<br />
due to permeability variations will be detected by the ET probe unless<br />
saturation is complete. Apart from these signals overwhelming the<br />
conventional ones, we must cope with the fact that permeability variations<br />
are not restricted to defect locations. Internal stresses, non-uniform heat<br />
*Actually the situation is far more complicated for magnetic media, which<br />
have nonlinear induction vs field intensity relationships. The value of<br />
permeability that we use in equation (1) must also be carefully chosen in the<br />
magnetic case, but detailed consideration of electromagnetic fields in<br />
conductors is beyond the scope of this discussion.
- 238 -<br />
treatment, and bending are examples of non-defective sites where signals will<br />
occur in unsaturated samples. These are often indistinguishable from real<br />
defects, resulting in unnecessary tube removal or plugging. It is therefore<br />
essential, for reliable inspection, that the medium be fully saturated.<br />
To understand the phenomenon of magnetic saturation we must consider the<br />
magnetic dipole moments of the constituent atoms in a magnetic material.<br />
These are arranged in a random pattern of domains when the sample is in the<br />
demagnetized state. Application of an external field causes large scale<br />
alignment of dipoles, which contributes to the total magnetic induction.<br />
Eventually, as applied field intensity is increased, the material<br />
contribution reaches a saturating value; all dipoles are aligned parallel to<br />
each other. Any further increase in induction is then equal only to the<br />
increase in applied field. The slope of the B vs H curve becomes equal to a<br />
constant (no), where B and H are the moduli of induction and field<br />
intensity respectively, and conventional eddy current testing is possible.<br />
2.2 Saturation Probe<br />
Because of the geometrical configuration of tubing in a heat exchanger, the<br />
only source of applied saturating field must be the eddy current probe<br />
itself. For mild magnetic materials it has been found that permanent magnets<br />
housed in the probe are the most convenient method of field application.<br />
This eliminates the need to supply any high magnetizing currents, and the<br />
subsequent cooling requirements that these imply. In general, for ET<br />
saturation probe development, it is more desirable to choose permanent<br />
magnets, whenever possible, over electromagnetic coils.<br />
We have been fortunate in developing probes to saturate mild materials, due<br />
largely to the advances over the past two decades in permanent magnet<br />
manufacturing technology. Of particular use has been the fabrication of Rare<br />
Earth - Cobalt alloys (e.g., S111C05) with maximum energy-products of order<br />
200 kJ/m 3 (25 MG-Oe). Only with hard magnetic materials of this quality<br />
can alloys such as 3RE60, with saturation induction approximately 0.6T<br />
(6000G), be saturated. This is achieved by placing three magnets in the<br />
configuration shown in Figure 1, where like poles are arranged facing each<br />
other across a space filled with soft magnetic material (mild steel or<br />
Permendur "keepers"). This configuration was initially developed by Cecco<br />
and co-workers[3]; its effect is to focus magnetic flux lines into the tube<br />
wall. Higher flux levels, and hence inductions, can be obtained in the tube<br />
using this method than from a single magnet. The eddy current sensing coil<br />
normally rides with the assembly, surrounding the central magnet. As<br />
expected, the magnets experience considerable demagnetizing fields from each<br />
other, and must therefore have a very "long and flat" magnetization vs<br />
coercive field curve. This is a property of Rare Earth-Cobalt alloys, making<br />
them Ideal for this application.
- 239 -<br />
The choice of optimum magnet sizes depends largely upon the material type and<br />
tube dimensions. Assembling a probe for a given material and tube<br />
configuration can be aided by computer analysis, such as finite element<br />
solutions to Maxwell's equations for magnetostatics, but the magnetic<br />
properties for the materials involved (tube, magnets, keepers) must be known<br />
accurately. Laboratory verification is still required and constitutes the<br />
major portion of the work involved. Once a magnet configuration is<br />
"optimized", usually by impedance monitoring of magnetization levels,<br />
conventional eddy current probe design considerations, such as impedance<br />
matching and coil winding, can be determined. This is described in a paper<br />
by Cecco [3]. A cut-away view of a probe is shown in Figure 2.<br />
The probe may be described as three magnets and two keepers, arranged in the<br />
manner of Figure 1, with the sensing coil surrounding the central magnet and<br />
adjacent to the region of tube wall saturation. Some shielding is necessary<br />
to prevent coupling into the keepers by the alternating field. A reference<br />
coil is located at the rear and is shielded to prevent it from sensing the<br />
tube material. Spacer material is usually some nonconducting medium such as<br />
Dupont Delrin. All components are housed in an Inconel* (nonmagnetic) casing<br />
with electrical connections through a four pin connector at the rear.<br />
A final comment, in relation to probe dimensions, should be inserted here.<br />
Dimensions shown in Figure 1 are typical of the magnet assembly in a 19 mm OD<br />
tube, and they effectively determine the minimum probe length and diameter.<br />
In many instances it may not be possible for the probe to traverse U-bend<br />
tubes completely, restricting usage to straight tubing, or the straight<br />
sections of U-bend tubing.<br />
3. PROBE TESTING<br />
The first step in developing a saturation probe is to optimize the magnet<br />
configuration to the tube size. This cannot be done without some trial and<br />
error measurement, and a fast method of verifying saturation in a tube has<br />
been developed to accommodate it. If an AC encircling coil is placed around<br />
the outside of a nonmagnetic tube, it will have an electrical impedance that<br />
is dependent upon the tube material resistivity and geometry. Replacing the<br />
tube with a magnetic one of identical dimensions and resistivity will change<br />
the coil Impedance, but it can be brought back to its original value by<br />
saturating the magnetic tube. Eddy current impedance monitoring of<br />
saturation is based upon this phenomenon.<br />
international Nickel Co.
- 240 -<br />
With an AC coil surrounding a magnetic tube, magnets are slid down the inside<br />
and saturation is verified if the coll impedance equals the value it would<br />
have for an identical, but nonmagnetic, tube. In practice neither electrical<br />
resistivity nor dimensions can be matched identically between magnetic and<br />
nonmagnetic tubes. The former problem, that of electrical resistivity, can<br />
be tolerated if the match is within 15 to 20 percent. The latter problem,<br />
different tube dimensions, is overcome by closely matching only the outer<br />
diameter and sampling a shallow surface layer using high frequency. Figure 3<br />
shows the change in impedance of a coil encircling a 19.1 mm 00 (1.84 mm wall<br />
thickness) 3RE60 tube as a saturating magnet assembly is moved through the<br />
tube.<br />
After saturation is attained and the complete prototype probe is assembled, a<br />
comparison of ET signals from artificial defects in a nonmagnetic tube and a<br />
3RE60 tube can be made. This is shown in Figure 4, where impedance plane<br />
signals from machined calibration defects in 316 stainless steel and 3RE6O<br />
are presented. These signals were recorded at a test frequency of f9o, the<br />
frequency at which shallow ID and 0D defect indications have 90° phase<br />
separation[4].<br />
As a final check on the ability of the probe to prevent unwanted<br />
"permeability signals", a strip chart recording is taken in a 3RE6O tube with<br />
machined ID, 0D and through hole indications, and a shot-peened<br />
circumferential band. Shot-peening induces a localized permeability<br />
variation due to internal stresses, but is not a real defect site. On the<br />
strip chart recording, which is shown in Figure 5, we would expect the real<br />
defects to appear, but not the shot-peened area.<br />
The key procedures and components for probe development are given in the<br />
following summary.<br />
SUMMARY OF PROCEDURES AND COMP<strong>ON</strong>ENTS<br />
Test 1 Impedance Monitoring to Verify Saturation<br />
Detector: AC Encircling Coil<br />
Frequency 1 MHz<br />
Nonmagnetic Reference Tube: 304 stainless steel<br />
resistivity p = 72 x 10~ B fi-m<br />
dimensions 19.13 mm 0D x 1.38 mm wall<br />
Method: Monitor encircling coil impedance as magnets pass through<br />
tube.<br />
Results: Figure 3 - Impedance plane display of encircling coil.
- 241 -<br />
Test 2 Comparison of ET Signals in Nonmagnetic and Magnetic Tubing<br />
Detector: ET Saturation Probe<br />
Frequencies: £,„ (316 SS) = 70 kHz<br />
£,0 (3RE60) = 75 kHz<br />
Nonmagnetic Tube: 316 stainless steel<br />
p = 75 x 10~ 8 J2-m<br />
dimensions 19.05 mm OD x 1.85 mm Wall<br />
Method: Record ET signals from OD, ID, through hole machined<br />
calibration defects in nonmagnetic tube and 3RE60 tube<br />
Results: Figure 4 - Impedance plane display of ET probe signals<br />
Test 3 Verify Saturation Through Absence of Permeability Induced Signal<br />
Detector: ET Saturation Probe<br />
Frequency f90 (3RE60) = 75 kHz<br />
Method: Record ET signals from 3RE60 tube having OD, ID, through<br />
hole machined calibration defects and shot-peened OD band.<br />
Results: Figure 5 - Strip chart recording to show absence of signal<br />
from shot-peened area.<br />
For all tests: 3RE60 tube - resistivity p = 85 x 10~ e Q-m<br />
dimensions 19.1 mm OD x 1.84 mm wall<br />
4. DISCUSSI<strong>ON</strong> OF RESULTS<br />
Eddy Current Instrument - Automation Industries EM33OO<br />
Because there is a great deal of procedural background involved in<br />
understanding Figure 3, some discussion is warranted. The results shown may<br />
be briefly explained as follows.<br />
An AC encircling coil operating at 1 MHz In air has an impedance vector locus<br />
at point 0. Phase is set up such that an inductive increase gives a<br />
vertically upward signal. Insertion of a 304 stainless steel tube into the<br />
coil moves the Impedance locus to point 0', whereas if a 3RE60 tube is used,<br />
impedance is shifted to point A. From point A three traces can be generated.<br />
These are labelled as trace A, trace B and trace C. Trace A is generated by<br />
magnetizing the tube to saturation by means of an external solenoidal coil,<br />
bringing the impedance locus to point A 1 . If the 304 stainless steel and<br />
3RE60 tubes had identical resistivity and geometry points 0' and A' would<br />
coincide. If the vertical component of trace A Is plotted against solenoid<br />
magnetizing current, trace B is generated, where B 1 represents the point of<br />
saturation. If now, instead of a solenoid, magnets are used inside the tube<br />
and the vertical component of impedance change (from point A) is plotted
- 242 -<br />
against magnet position, trace C is generated. By definition the vertical<br />
distances from A to A' and B to B 1 , which are equal, represent the vertical<br />
impedance change necessary to saturate the tube. For the magnets to also<br />
saturate the tube, the vertical distance from C to C must equal the former<br />
two.<br />
It is evident from Figure 3 that the magnet configuration tested has<br />
saturated the 3RE60 tube at point C. This occurs when the magnets are<br />
centred under the AC sampling coil. We see three lobes in trace C, each one<br />
due to one of the three magnets in the configuration. With saturation<br />
verified, a probe can be constructed by winding an AC coil around the central<br />
magnet.<br />
The eddy current impedance plane signals in Figure 4 compare signals from 316<br />
stainless steel and 3RE60. The saturation probe is used for both, since in<br />
the case of 316 stainless steel the magnets do not influence the eddy current<br />
signals. Signals from 3RE6O and 316 stainless are essentially identical.<br />
For this type of signal analysis, eddy current instrument phase is set to<br />
give a horizontal deflection to the left for a concentric ID groove. Working<br />
at fgo, OD groove signals will be rotated 90° clockwise from the ID signals.<br />
This 90° phase separation is in itself often used as a verification of<br />
saturation, and it is amply demonstrated in the 3RE6O signals. The OD signal<br />
from the 3RE60 tube is slightly distorted near the bottom. This is<br />
considered to be due to demagnetizing effects near the corners of the groove,<br />
preventing full saturation in these areas. Some samples of 3RE60 tubing are<br />
more difficult to saturate than others. Those that are more difficult<br />
generally show this distortion; it does not occur in all tube samples. The<br />
central portion of the groove is fully saturated, as can be seen by the<br />
disappearance of the distortion as the signal grows vertically. Overall the<br />
eddy current signals from the 3RE6O tube closely natch those from the 316<br />
stainless, indicating the probe is capable of providing data that is readily<br />
interprétable in the light of present eddy current knowledge.<br />
The strip chart recording of Figure 5 shows signals of the type in Figure 4<br />
resolved into X and Y components. Each one is displayed on a separate<br />
channel. Although they are labelled for easy identification, there is no<br />
difficulty determining that signals are present from the machined calibration<br />
defects. We note however that between the two event markers there Is no<br />
discernable signal, although this is the region of an external shot-peened<br />
band. Without full saturation, a large signal would be seen in this region.<br />
The absence of such a signal further proves the probe can eliminate<br />
"permeability noise", and saturates the tube under examination.
5. C<strong>ON</strong>CLUSI<strong>ON</strong>S<br />
- 243 -<br />
It would be premature to conclude that full field inspections could now be<br />
performed in all mildly ferromagnetic materials up to those of saturation<br />
induction approximately 0.6 tesla. The laboratory testing of saturation<br />
probes for 3RE60 is to a great degree complete, and the probes must now prove<br />
themselves in the field. Laboratory results on calibration defects are<br />
encouraging. We know of at least one prototype probe that has seen field<br />
use, but at this time its performance has not yet been assessed in depth.<br />
It is believed that 3RE60 tubing of "average" magnetic characteristics (for<br />
that material) can be inspected by eddy current, using probes of the type<br />
discussed here. Although the results presented in this paper dealt<br />
exclusively with 19 mm tubing, previous work using 16 mm tubing has also been<br />
successfully carried out. With a minimal amount of laboratory work the<br />
application of these probes could likely be extended to other<br />
austenitic-ferritic stainless steels.<br />
To avoid presenting an overly optimistic picture of the situation, it is<br />
necessary to point out several important considerations. These are as<br />
follows.<br />
i) Because of the sizes of the magnets, U-bend tube sections are largely<br />
uninspectable.<br />
ii) Tube regions under magnetic support plates are uninspectable since<br />
these areas cannot be fully saturated.<br />
iii) The variation of magnetic characteristics between tube samples is not<br />
yet well known. Only a field test can determine whether all tubes in<br />
a given heat exchanger are saturable. (This is a problem also<br />
encountered in Monel 400 Inspection).<br />
iv) It is possible that the existence of heavy magnetic deposits might<br />
sufficiently perturb the saturating field to render tubing<br />
uninspectable.<br />
v) The probe cannot be used effectively in the hands of an operator<br />
unaccustomed to ferromagnetic inspection and its associated problems.<br />
The above points serve to underline that although the probe makes 3RE60<br />
tubing Inspectable, it cannot be expected to perform satisfactorily outside<br />
of its limitations.
6. SUMMARY<br />
- 244 -<br />
Conventional eddy current inspection of ferromagnetic materials requires full<br />
magnetic saturation of the test item. This has been realized In 19 mm 3RE60<br />
tubing by using a focussed arrangement of permanent magnets. The results<br />
shown here are proof that the laboratory problems associated with 3RE60<br />
inspection have been overcome, and field confirmation of probe ability is the<br />
logical next step. There is no reason to believe that the use of these<br />
probes cannot be extended to similar mildly ferromagnetic alloys. Properly<br />
utilized, this generation of saturation probes is of great potential value to<br />
ferromagnetic tube inspection, so long as they are used within the bounds of<br />
their applicability.<br />
7. REFERENCES<br />
[1] Van Drunen, G. and Cecco, V.S., Recognizing Limitations in Eddy<br />
Current Testing, Atomic Energy of Canada Report AECL-7508, (1981).<br />
[2] Lorrain, P. and Corson, D.R., Electromagnetic Fields and Waves - 2 ed<br />
(Freeman and Co., 1970), 476.<br />
[3] Cecco, V.S., Materials Evaluation, _3_7, (1979), 51.<br />
[4] Cecco, V.S., Van Drunen, G., Sharp, F.L., Eddy Current Testing Manual<br />
on Eddy Current Method Vol. 1, Atomic Energy of Canada Report<br />
AECL-7523, (1981), 124.
End Magnet<br />
- 245 -<br />
Soft Magnetic<br />
Keepers<br />
30 mm<br />
FIGURE 1: Magnetic Focusing Configuration in<br />
Ferromagnetic Tube.<br />
End Magnet
10<br />
1 Centre Magnet<br />
2 Keepers<br />
3 End Magnets<br />
4 Brass Shield Rings<br />
5 Delrin Spacers<br />
- 246 -<br />
6 Delrin Reference Coil Mount<br />
7 Brass Reference Coil Shield<br />
8 Eddy Current Coil<br />
9 Reference Coil<br />
10 Inconel Casing<br />
11 Casing Closure Plug (Inconel)<br />
FIGURE 2: Cruss Sectional View of Probe.<br />
11
Induct ive<br />
Increase<br />
Inserti on<br />
of 304<br />
Stainless Steel<br />
Tube<br />
0'<br />
Trace A<br />
B r<br />
Solenoid Currrnt<br />
(amperes)<br />
External Magnetization<br />
50 0<br />
Trace C<br />
Magnet 1'usition (cm)<br />
FIGURE 3: Impedance Monitoring of Saturation.<br />
50 100 150 200<br />
Permanent Magnet Saturation<br />
•p-
A - II' Concentric Groove<br />
0.1 mm deep x 2.5 mm<br />
long<br />
- 248 -<br />
316 Stainless Steel Signals: fy0 = 70 kHz<br />
B - OD Concentric Groove<br />
O.-i ram deep x 2.5 mm long<br />
3RE60 Signals: f 0 = 75 kHz<br />
A - ID Concentric Groove<br />
0.5 mm deep x 2.5 mm long<br />
C - OD Hole 1.5 ma Dia.<br />
0.5 mm deep<br />
D - Through Hole<br />
1.5 mm Dia.<br />
B - Through Hole<br />
1.5 mm Dia.<br />
E - OD Eccentric Groove<br />
0.8 mm deep x<br />
3.5 mm long<br />
C - 01) Fxi'i-ntric Gro<br />
0.7 S mm deep x<br />
3.5 mm long<br />
I'lGURE M : Kddv Current Impedance Plane Signals.
Vertical Signal Component (Y-Channel)<br />
A_A<br />
ID Through OD<br />
Groove Hole Groove<br />
Horizontal Signal Component (X-Channel)<br />
- 249 -<br />
Both Channels - 10 volts full scale deflection<br />
ET Frequency - 78 kHz<br />
ID<br />
Groove<br />
FIGURE 5: Strip Chart Recording.<br />
Shot-Peen<br />
(Permeability)<br />
event<br />
marker<br />
event<br />
marker
- 250<br />
<strong>ON</strong> THE RELATI<strong>ON</strong> BETWEEN ULTRAS<strong>ON</strong>IC ATTENUATI<strong>ON</strong> AND FRACTURE<br />
TOUGHNESS IN TYPE 403 STAINLESS STEEL<br />
F. Nadeau and J.-F. Bui.iie.-xe.<br />
National Reieaicli Council o & Canada, Bouckexv itie, Quebec<br />
G. Van Dianen<br />
Atomic Eneigij oj Canada Limited<br />
Chalk Rive-%] ' Ontaiio<br />
ABSTRACT<br />
Previous studies have indicated the possibility of using ultrasonic<br />
attenuation as a nondestructive tool for predicting the fracture toughness of<br />
metals. In all these studies, however, the fracture toughness was varied by<br />
changing microstructure or composition. In the present paper, ultrasonic<br />
attenuation is measured in a single sample of constant microstructure (type<br />
403 stainless steel) in which fracture toughness is changed by nearly a factor<br />
of three by varying the temperature. Results show that the frequency and<br />
temperature dependence of attenuation are in agreement with conventional grain<br />
scattering theory and do not correlate with the large changes in Kic. It is<br />
concluded that one of the main difficulties of nondestructively characterizing<br />
fracture toughness is that a small variation in one of the material's<br />
properties, such as yield stress, can alter the fracture mechanism, leading to<br />
very rapid changes in Kjc. In most cases, the modest change in yield stress<br />
will be accompanied by similarly modest changes in attenuation, which do not<br />
reflect the rapid changes in Kjc.<br />
INTRODUCTI<strong>ON</strong><br />
As the ability to detect and characterize discrete defects in structures is<br />
improved through the use of better nondestructive methods, more attention is<br />
being focused on novel techniques for the determination of microstructure<br />
'1' and mechanical properties (2). of particular interest is the<br />
possibility of using ultrasonic measurements to determine engineering<br />
mechanical properties such as tensile strength, yield strength, fracture<br />
toughness and the ductile to brittle transition temperature (DBTT). Earlier<br />
attempts in this area have been based on empirical relations between<br />
ultrasonic velocity and yield (or tensile) strength in various steels and<br />
iron-carbon alloys (3-5)t More recently, the bahavior of third-order<br />
elastic constants was shown to correlate with the heat treatment history of<br />
aluminum alloys ("'.<br />
However, most attempts at characterizing microstructure and mechanical<br />
properties ultrasonically have been based on measurements of attenuation.<br />
Attenuation results from the combined effects of scattering, the reorientation<br />
and mode conversion of acoustic energy at microstructural interfaces, and
- 251 -<br />
absorption, the conversion of acoustic energy into heat. Scattering and<br />
absorption are both caused by the presence of inhoraogeneities such as<br />
precipitates, inclusions, voids, grain and interphase boundaries,<br />
dislocations, interstitial impurities, etc. Since these same inhomogeneities<br />
also play a primary role in determining mechanical properties, it is expected<br />
that measurements of attenuation will reflect mechanical properties. The<br />
sensitivity of attenuation to small changes in microstructure which affect<br />
mechanical properties has been demonstrated for aluminum alloys »'' and<br />
other materials (8)# in cases where grain boundary scattering is dominant,<br />
various models (l>°,10) offering good agreement with experiment can be used<br />
to calculate the grain size from the frequency dependence of attenuation. The<br />
ultrasonically measured grain size, 6, can then be correlated with mechanical<br />
properties via the use of the Hall-Petch relationship.<br />
U = uo + kS-i (1)<br />
where M is the mechanical property of interest, n, is a constant dependent<br />
on composition, and k is a proportionality constant. This latter approach was<br />
used by Klinman et al (11) to successfully predict yield strength, tensile<br />
strength and DBTT's in carbon steels of known compositions. Another approach<br />
was taken by A. Vary ("), who suggested an empirical relation relating<br />
fracture toughness, K.ic, yield strength oy, and ultrasonic attenuation:<br />
Klc 2 /Oy = 8.12 x 10 6 (Va&j) 0 ' 31 " 1 (2)<br />
where V£ is the velocity of longitudinal waves in the material and ßg is<br />
the slope of the attenuation versus frequency curve at a frequency, f =<br />
Vg/6, where the wavelength is equal to the grain size. Since f is<br />
generally outside the range of most normal attenuation measurements (eg. f =<br />
0.5 GHz in steel with 6 = 10 Mm), the value of fJ$ is extrapolated from<br />
measurements in the 10 to 50 MHz range. Two ma rag ing steels and a titanium<br />
alloy, aged at various temperatures, were found to obey relation (2).<br />
However, studies on Fe-C alloys (13) have shown that increasing the grain<br />
size or increasing the cabon content, both cause an increase on the DBTT but<br />
have opposite effects on ultrasonic attenuation. This is attributed either to<br />
diminished absorption due to an increase in dislocation damping with<br />
increasing C content, or to diminished scattering due to a decrease of elastic<br />
anisotropy of the grains related to the presence of pearlite or cementite<br />
around the ferritic grain boundaries. Thus it is difficult to isolate one<br />
component of the ultrasonic attenuation, and in particular, to differentiate<br />
between absorption and scattering. It is also difficult to vary one material<br />
property such as fracture toughness while others remain constant.<br />
In the present study, this is accomplished by correlating the ultrasonic<br />
behavior and mechanical properties of a 403 stainless steel between -60 and<br />
+40°C. In this temperature range, the plain strain fracture toughness, Kjc,<br />
changes by =250% because of a smooth brittle to ductile transition, while<br />
composition and microstructure remain unchanged.
- 252 -<br />
Following a brief description of the sample, its mechanical properties, and<br />
the experimental techniques, attenuation measurements are presented as a<br />
function of temperature and frequency. They are interpreted in terms of<br />
scattering mechanisms and discussed in terms of possible correlations with<br />
mechanical properties.<br />
EXPERIMENTAL<br />
Sample<br />
The material chosen is an AISI type 403 stainless steel which has well<br />
characterized properties because of its use as end fittings to hold zirconium<br />
pressure tubes in Canada-Deuterium Uranium (CANDU) nuclear reactors. The<br />
sample was part of a compact tension (CT) specimen used in a previous study on<br />
the effects of impurity concentration and temperature on the fracture<br />
toughness of type 403 stainless steel (14). The particular sample used<br />
here, identified as LR in reference 14, has chromium as main alloying element<br />
(11.8 wt %) and a total impurity concentration of 1620 ppm. (See reference 14<br />
for exact composition and heat treatment). The microstructure consists of<br />
tempered martenslte. Carbides outline the original austenite grains (6= 25um)<br />
and also the martensite lath boundaries. Fig. 1 is a scanning electron<br />
micrograph of the sample.<br />
The Kic values were measured using compact tension (CT) specimens according<br />
to ASTM Test for Plane-Strain Fracture Toughness of Metallic Materials<br />
E 399-74. The yield stress, ov, and the plane strain fracture toughness,<br />
Kic, taken from reference 14 are given in Fig. 2 for the temperature range<br />
-60°C to +40°C. Note that the yield stress decreases by = 7% and Klc<br />
increases by approximately a factor three over this temperature range.<br />
Ultrasonic attenuation<br />
Ultrasonic attenuation was measured using the pulse-echo technique with two<br />
undamped 12 mm diameter X cut quartz piezoelectric transducers having center<br />
frequencies f0 of 5 to 10 MHz used successively. Using overtones,<br />
measurements were made at frequencies of 10 to 50 MHz in 5 MHz intervals. The<br />
ultrasonic pulses were generated and detected with a Matec heterodyne<br />
pulser-receiver and the attenuation was measured by electronically fitting an<br />
exponential curve over the decaying echos. The velocity of longitudinal waves<br />
was also measured (6075 m/s at 20°C varying by less than 1% in the temperature<br />
range studied) and used to convert the decay time into neper/m attenuation<br />
units. Temperature was varied by mounting the 5 mm thick sample on a Peltier<br />
element which allowed both cooling and heating of the sample depending on the<br />
polarity of the current supplied to the element. The temperature was measured<br />
with a thermocouple mounted directly on the sample. The experimental<br />
apparatus, along with a typical oscillogram showing a series of echos fitted<br />
with an exponential decay curve, is illustrated in Fig. 3.
Results and Discussion<br />
- 253 -<br />
Ultrasonic attenuation results are given in Fig. 4 as a function of temperature<br />
for frequencies ranging from 10 to 50 MHz. The attenuation exhibits a<br />
strong dependence on frequency but the variations with temperature are small<br />
(within experimental error) and somewhat random. The frequency dependence of<br />
the attenuation is illustrated in Fig. 5 using values averaged over the<br />
temperature range studied and corrected for diffraction losses by the method<br />
described in (15). The boundary that distinguishes Rayleigh scattering from<br />
stochastic scattering, X = 2it6, corresponds for the case of a grain size 6 =<br />
25 ym to a value of f = 38.7 MHz. We see, on Fig. 5, the initial f 1 * dependence<br />
of a typical of Rayleigh scattering, gradually transforming into an f<br />
dependence in the X = 2IT6 region where stochastic scattering is expected.<br />
Thus, the frequency dependence of a is consistent with grain boundary<br />
scattering theory.<br />
In order to examine the temperature dependence of a in the perspective of the<br />
temperature related changes in plastic properties and fracture toughness, the<br />
values of 0£ were computed, as described in ref. 2, and compared with the<br />
empirical relation given in the introduction, equation (2). Fig. 6 is a plot<br />
of V£0£ values against Kic /cv. The data points lie more or less<br />
along a horizontal line indicating no net effect on ultrasonic attenuation of<br />
the almost 3 fold change in fracture toughness.<br />
It is well-known O0) that ultrasonic attenuation in steel at typical<br />
N.D.E. frequencies (1-50 MHz) is mainly due to scattering by grain boundaries<br />
and the present results appear to corroborate this fact. Although scattering<br />
in a multiphase material such as type 403 stainless steel is not easily<br />
calculated theoretically because of the complexity of the structure and<br />
insufficient knowledge of elastic constants, the results of Fig. 5 are<br />
consistent with measurements by Papadakis and others ( 10 ), which indicate<br />
that the main scattering features are the original austenitic grains, and that<br />
the effect of the intragranular structure is to modify the effective anisotropy<br />
of the grains. In general, the presence of multiple phases is to reduce<br />
the anisotropy compared to single phase materials. This explains the<br />
relatively low attenuation of the present sample. Based on the anisotropy, A,<br />
of pure iron (**•) where A = Cilit-(Cn-C12)/2 and the Cy 's are the usual<br />
elastic constants for a cubic system, any increase of =6% in attenuation with<br />
temperature is expected between -60°C and +40°C. This is within experimental<br />
error of the results of Fig. 4.<br />
The temperature dependence of yield stress (and therefore indirectly that of<br />
Kic) may be expected to correlate better with absorption, which can be<br />
related to dislocation density and mobility ( 17 \ rather than scattering<br />
which is associated with a constant grain size and varying anisotropy. An<br />
effort was therefore made to separate any absorption contribution to a by<br />
analyzing the data of Fig. 5 in the form of plots of ct/f versus f 2 as in<br />
reference (18). Such an analysis, however, did not give a clear separation of<br />
scattering and absorption, largely because of the lack of data lying clearly<br />
^ ideated that the absorption contribution is
C<strong>ON</strong>CLUSI<strong>ON</strong><br />
- 254 -<br />
In order to verify tô what degree ultrasonic attenuation can be used to nondestructively<br />
predict the fracture toughness of a material, the ultrasonic<br />
attenuation behavior of an AISI type 403 stainless steel exhibiting a smooth<br />
ductile to brittle transition was studied as a function of temperature. Over<br />
the temperature range studied (-60°C to +40°C) the yield strength and fracture<br />
toughness changed respectively by 7 and 250%. The ultrasonic attenuation,<br />
however, varied by less than the 6% expected from grain boundary scattering<br />
theory, giving no indication of the large change in Kic. An analysis of the<br />
results in terms of Vary's empirical model (equation 2) which appears to give<br />
good results for cases where Kic is varied by changes in microstructure<br />
(grain size, heat treatment) gave no correlation between attenuation and<br />
K ic / a y<br />
One of the main difficulties of nondestructively characterizing a material's<br />
fracture toughness is that a small variation in one of the material's<br />
properties, such as yield stress, can cause the fracture mechanism to change<br />
(ductile to brittle transition) leading to very large variations in Klc<br />
values. If that property is indirectly measured by a method such as<br />
ultrasonic attenuation which is mostly sensitive to other properties of the<br />
material, the chances of obtaining an accurate K\c value are reduced even<br />
more.<br />
Nondestructive evaluation of Kic can be achieved when the method directly<br />
measures the material properties controlling its fracture toughness, for<br />
example ultrasonic attenuation for samples where variations in the grain size<br />
changes their toughness (**). Similarily, a method allowing for the direct<br />
measurement of ultrasonic absorption may provide better correlation with K^c<br />
variations caused by differences in the plastic properties of a material for a<br />
given microstructure. Such a method is currently under development (19) at<br />
the Industrial Materials Research Institute and it is hoped that it will prove<br />
to be a powerful tool for the nondestructive characterization of materials.<br />
Acknowledgement<br />
Supply of sample material for this study by R.R. Hosbons of the Metallurgical<br />
Engineering Branch at Chalk River Nuclear Laboratories is gratefully<br />
recognized.
REFERENCES<br />
- 255 -<br />
1. K. Goebbels, "Structure Analysis by Scattered Ultrasonic Radiation", in<br />
Research Techniques in Nondestructive Testing, Vol. IV, ch. 4, edited<br />
by R.S. Sharpe, Academic Press New York (1980).<br />
2. A. Vary, "Ultrasonic Measurement of Material Properties", ibid, ch. 5.<br />
3. J. and H. Krautkrämer, "Ultrasonic Testing of Materials", second<br />
edition, Springer-Verlag, New York, N.Y. (1977).<br />
4. L.Y. Levitan, A.N. Fedorchenko and A.V. Sharko, "Ultrasonic Testing of<br />
the Strength of Steel 45", Defektoskopiya, Vol 3, pp. 129-130, May-June<br />
1976.<br />
5. E.R. Leheup and J.R. Moon, "Yield and Fracture Phenomena in<br />
Powder-Forged Fe -0.2C and their Prediction by NDT Methods", Powder<br />
Metallurgy, April 1980, p. 177.<br />
6. J.S. Heyman and E.J. Chern, "Characterization of Heat Treatment from<br />
Ultrasonic Determinarion of the Second and Third Order Elastic<br />
Constant", 1981 IEEE Ultrasonics Symposium p. 936, IEEE New York<br />
(1981).<br />
7. M. Rosen, "NDE Characterization of Precipitation Hardening Phenomena in<br />
Aluminum Alloys" in Advanced NDE Technology 1982, National Research<br />
Council of Canada, Boucherville, Québec, Research Report IGM82-0-59<br />
J.F. Bussière ed., p. 54.<br />
8. R.E. Green, "Effect of Metallic Microstructure on Ultrasonic<br />
Attenuation", p. 117 in Nondestructive Evaluation: Microstructural<br />
Characterization and Reliability Strategies, edited by 0. Buck and<br />
S.M. Wolf, Conference Proceeding, The Metallurgical Society of AIME,<br />
New York, N.Y. (1981).<br />
9. Mason, W.P., Physical Acoustics and the Properties of Solids, New York,<br />
Van Nostrand, New York, (1958).<br />
10. E. Papadakis, "Scattering in Polycrystalline Media", ch. 5, pp. 2237 in<br />
Methods of Experimental Physics, Vol. 19 - Ultrasonics, P.D. Edmonds<br />
ed., Academic Press, New York, N.Y. (1981).<br />
11. R. Klinman and E.T. Stephenson, "Ultrasonic Prediction of Grain Size<br />
and Mechanical Properties in Plain Carbon Steel", Mat. Eval. Vol. 39<br />
(1981), No. 12 pp. 1116 - 1120 Nov. 1981.<br />
12. A. Vary, "Correlations among Ultrasonic Propagation Factor and Fracture<br />
Toughness Properties of Metallic Materials", Mat. Eval. Vol. 36 N.7<br />
p. 55-64, June 1978.
- 256 -<br />
13. R.L. Smith, F.L. Rusbridge, W.N. Reynolds and B. Hudson, "Ultrasonic<br />
Attenuation, Microstructure and Ductile to Brittle Transition<br />
Temperature in Fe-C Alloys", Mat. Eval. Vol. 41, pp. 219-222, (Feb.)<br />
1983.<br />
14. R..R. Hosbons, A.J. Pacey, B.L. Wotton, "Effect of Impurity<br />
Concentration on the Change in Fracture Toughness of AISI Type 403<br />
Stainless Steel with Fast Neutron Radiation", Properties of Reactor<br />
Structural Alloys After Neutron or Particle Irradiation, ASTM STP 570,<br />
pp. 103-116, (1975).<br />
15. H. Sepi, A. Granato, R. Truell, "Diffraction Effects in the Ultrasonic<br />
Field of a Piston Source and Their Effects upon Accurate Ultrasonic<br />
Attenuation Measurements", Journal of the Acoustical Society of<br />
America, Vol 28, p. 230, (1958).<br />
16. G. Simmons and H. Wang, "Single Crystal Elastic Constants and<br />
Calculated Aggregate Properties: A Handbook", The M.I.T. Press,<br />
Cambridge, Mass., 1971, p.34.<br />
17. A.B. Bathia, "Ultrasonic Absorption", Clarendon Press, Oxford, England<br />
(1967), pp. 338-370.<br />
18. R.L. Smith, W.N. Reynolds and S. Perring, presented at the Fourth<br />
European Conference on Internal Friction and Ultrasonic Attenuation,<br />
Lyon, France 1983 (to be published).<br />
19. J.P. Monchalin, J.F. Bussiêre, "Measurement of Near-Surface Ultrasonic<br />
Absorption by Thermo-Emissivity" presented at the Symposium on<br />
Nondestructive Methods for Material Property Determination, Hershey,<br />
Pennsylvania, April (1983), (Plenum Press, New York, to be published).
- 257 -<br />
Fig« I' Scanning electron micrograph of the type 403 stainless steel<br />
sample used in the present study.<br />
i<br />
820<br />
810 -<br />
800<br />
790<br />
780<br />
770<br />
760<br />
1<br />
I |<br />
750<br />
I<br />
i i<br />
-60 -40 -20 0 20 40<br />
T(°C)<br />
i<br />
-<br />
100<br />
80<br />
o<br />
i<br />
60 I<br />
- 40<br />
Fig. 2: Yield strength, ay, and plane strain fracture toughness,<br />
Kic, of the 403 stainless steel sample between -60°C and<br />
+40°C.
Ultrasonic<br />
Instrumentation<br />
- 258 -<br />
«ÏOTÏYB<br />
- Styrofoam<br />
Transducer<br />
Sample<br />
Pelletier Element<br />
— Heat Sink<br />
Fig. 3: Schematic of the experimental setup used to measure ultrasonic<br />
attenuation as a function of temperature.
-Ê<br />
70<br />
60<br />
50<br />
40<br />
6 30<br />
20<br />
—<br />
-<br />
"I<br />
I<br />
•<br />
a<br />
1<br />
*<br />
a<br />
O<br />
A<br />
X<br />
50 MH^.<br />
30 MH^..<br />
15 MH^<br />
10 MH^.<br />
- 259 -<br />
10 I<br />
I<br />
1<br />
r<br />
-60<br />
0<br />
-40 -20<br />
T(°C)<br />
0<br />
*<br />
I<br />
•<br />
O<br />
A<br />
*<br />
D<br />
0<br />
*<br />
•<br />
O<br />
*<br />
•<br />
A<br />
*<br />
a<br />
A<br />
I<br />
* *<br />
*<br />
D<br />
1<br />
a<br />
1<br />
* *<br />
D<br />
D<br />
-<br />
I.<br />
-<br />
20 40<br />
Fig. A: Ultrasonic attenuation data as a function of temperature for<br />
frequencies ranging from 10 to 50 MHz.
100<br />
a (m" 1 )<br />
10<br />
f Slope = 4<br />
10<br />
- 260 -<br />
f (MHZ)<br />
Slope = 2<br />
100<br />
Fig. 5: Frequency dependence of ultrasonic attenuation.
- 261 -<br />
V ßß =8.07x10" 2l (k 2 icÖy)<br />
K 2 1c/ay(MJ/m 2 )<br />
Flg. 6: Relation between ultrasonic parameters V^ßg and mechanical<br />
properties Kic/Oy. The solid line is equation (2), which<br />
Vary used to fit his data on Titanium alloy and maraging<br />
steels. In this case Kic and oy were changed by heat<br />
treatment. The crosses correspond to the experimental data<br />
obtained for 403 stainless steel in the temperature range -60°C<br />
to +40°C.
- 262 -<br />
ACOUSTIC EMISSI<strong>ON</strong> TESTING OF MAN-LIFT DEVICES<br />
J.A. Baton<br />
Ontario Hydto Reie.an.ah<br />
Toronto, Ontario<br />
INTRODUCTI<strong>ON</strong><br />
Safety is a very controversial issue. Whereas we may find it an acceptable risk<br />
to cross a busy street or to ski down a steep slope, we may have some reservation<br />
about flying in a commerical airliner or to be raised 20 m above the ground<br />
in a bucket of an aerial lift device unless we have some positive perception<br />
regarding the structural Integrity of the equipment. Much of this assurance is<br />
developed through nondestructive test procedures but, until recently there has<br />
been no method which could provide reliable information as to the mechanical<br />
worthiness of the fibre reinforced plastic (frp) boom of man-lift devices<br />
("bucket trucks" or "cherry pickers").<br />
Dielectric testing of booms is mandatory but this test, usually performed semiannually,<br />
is a poor indicator of mechanical integrity. Under dry conditions a<br />
boom on the point of mechanical failure could pass this electrical test.<br />
Conversely, seepage of moisture Into surface layers would result in dielectric<br />
failure but may not have implications in terms of mechanical fitness.<br />
Tradition NDT methods do not work well on composite materials such as frp.<br />
Radiography will show fibre direction and gross voids but little else. Ultrasound<br />
suffers high attenuation plus the fibres give rise to a myriad of reflections.<br />
Liquid pénétrants will show surface imperfections but may also cause<br />
chemical degradation of the plastic resin. However, acoustic emission monitoring<br />
can potentially detect and locate "active" defects in both frp and metallic<br />
components and can be used in a straightforward, all encompassing test.<br />
FAILURE MODES<br />
As a generality, nondestructive testing is performed to detect potential future<br />
failure of a component or structure. If such a test is to be worthwhile, it<br />
must be sensitive to the precusors of possible failure modes. Further, the test<br />
must be performed at a frequency such that the progression from the earliest<br />
reliable detection of incipient failure to actual failure cannot occur. This<br />
point will not be dealt with here but does raise some questions as to the<br />
efficacy of periodic testing of frp booms.
- 263 -<br />
The two components of the frp composite are very different in mechanical<br />
behaviour. The glass fibres are brittle, exhibiting little strain prior to<br />
tensile fracture, whilst the matrix is a plastic which exhibits visco-elastic<br />
flow [1] rather than sudden failure. To some degree, the composite tends<br />
towards the characteristics of the matrix when in compression and the fibres<br />
when under tension. Under normal load regimes, a boom will experience tenoion<br />
on its upper surface and compression on its lower surface. The tensile properties<br />
are superior to those under compression thus, if the load is such as to<br />
cause failure, the normal mode is compressional collapse.<br />
Corapressional failure is, from an acoustic emission standpoint, very active. It<br />
is initiated by failure of the matrix and results in the large scale expulsion<br />
of fibres, as in Figure 1. All the processes involved are good generators of<br />
acoustic emission and thus defects that would lead to this failure are readily<br />
detectable.<br />
In rare but no unknown circumstances [2], tensile failure can occur. An hypothesis<br />
describing this behaviour has been presented [3] but has not been<br />
confirmed. Briefly, the scenario is as follows. Below a certain threshold, the<br />
fibres will maintain a tensile load Indefinitely. At loads above this threshold,<br />
small scale fibre breakage will occur. The load previously supported by<br />
the broken fibres will be instantaneously transferred to the surrounding<br />
matrix. The matrix then undergoes visco-elastic flow and, in turn, transfers<br />
its load to the remaining fibres, thus adding to their overload. Again, this<br />
results in more fibre breakage and more load shedding through the matrix to the<br />
remaining fibres. It is reasonable to suggest that the time taken for the<br />
matrix to transfer its load due to fibre breakage, ie, through the flow process,<br />
decreases as the load to be transferred increases. Thus, once the process<br />
starts it accelerates to an avalanche type failure, assuming the load is<br />
maintained. This can result in a very sudden failure in contrast to the<br />
relatively slow compressional failure mode.<br />
Figure 2 illustrates the relationship between fibre breakage and time and also<br />
indicates the effect of changing load. Figure 3 illustrates how the time to<br />
failure is dependent on load and also the effect of increasing the fibre count.<br />
Both of these figures are generated based on the above hypothesis. However,<br />
McElroy [2] 1 s demonstrated the relationship in the laboratory, his results<br />
being illustrated in Figure 4. Figure 4, reproduced from reference 2, also<br />
illustrates the effect of removing or changing the load during the process.<br />
Initially, acoustic emission monitoring was applied to bucket trucks solely to<br />
enhance the inspectability of the boom. It soon became apparent that the metal<br />
components also generated acoustic emission under the proof test conditions. At<br />
first this was considered a nuisance and "noise" sources were eliminated eg, a<br />
poorly lubricated pivot pin, but advantage was taken of this and the test was<br />
soon expanded to cover all major structural items in a single, comprehensive<br />
test. In fact, current indications are that defects in metallic components are<br />
discovered far more frequently than defects In the frp components [4],
- 264 -<br />
Metal components with classical acoustic emission sources (eg, propagating<br />
crack) exhibit the Kaiser effect. This effect describes the phenomenon whereby<br />
acoustic emission will only be generated when stress exceeds the previous highest<br />
value. A new component will tend to yield copious acoustic emission when<br />
first loaded. If the load were removed and reapplied, no acoustic emission<br />
would result until the load was further increased above its previous value.<br />
This can be beneficial in that if a structure is loaded to the onset of acoustic<br />
emission, an indication of load history is gained. If a periodic proof load is<br />
imposed, such as 150% working load, acoustic emission will only be generated if<br />
active degradation has occurred since the last proof load.<br />
The Kaiser effect is less well exhibited by frp composites. This is because the<br />
resin material flows and thus relaxation occurs on load removal. This was<br />
demonstrated in the laboratory, Figure 5. Here an frp specimen was subjected to<br />
a three point bend test such that acoustic emission occurred. The load was then<br />
removed. When it was reapplied with little delay, almost no acoustic emission<br />
was generated, in accordance with the Kaiser effect. However, if a delay of<br />
24 h ensued prior to reloading, much of the acoustic emission was recovered.<br />
This phenomenon has implications where a test is performed following an accident<br />
to a boom, such as a known overload.<br />
Some reference has been made in the literature to the ultimate strength of frp<br />
components with respect to acoustic emission monitoring. The ultimate tensile<br />
strength is truly the load that just fails to initiate the avalanche failure<br />
mechanism. Whilst loads much higher than this can be sustained for finite<br />
periods of time, they do impart damage and do reduce the load level required to<br />
initiate avalanche failure. Also, as failure is a time dependent phenomenon,<br />
above the threshold load acoustic emission will continue to be generated up to<br />
ultimate failure. Initially this may be very sparse but becoming more prolific<br />
as fracture approaches.<br />
For loads below the avalanche threshold, the Kaiser effect (short-term) tends to<br />
hold. For loads above the threshold, acoustic emission will always occur. If a<br />
boom is loaded, unloaded and then reloaded and acoustic emissions recur, then<br />
the threshold load has been reached. In this case, damage will have been done<br />
and acoustic emission will be re-established at a slighly lower load dependent<br />
on the degree of degradation. The ratio of these two loads is known as the<br />
"Felicity Ratio" and is used as a measure of damage.<br />
PERFORMANCE OF THE ACOUSTIC EMISSI<strong>ON</strong> TEST<br />
With the boom in the lowered state, but with outriggers operating, sensors are<br />
attached. It is important that the sensors are well coupled to the component.<br />
This is as simple as having a clean surface, using a small amount of viscous<br />
fluid such as vacuum grease between sensor and the surface and holding the<br />
sensor firmly in position. Sensor hold-downs are extremely varied and may range<br />
;rora a magnetic clamp for steel components to a simple band of adhesive tape<br />
for the boom Figure 6. Some care must be taken to ensure that the hold-down is<br />
not going to move, and hence produce its own acoustic emissions, during the<br />
test.
- 265 -<br />
Sensors used on the major metallic components will have a maximum sensitivity in<br />
the range 100-400 kHz. However, attenuation in frp at these frequencies would<br />
demand an unacceptable number of sensors necessary for coverage of the complete<br />
boom. For this reason, it is more usual to employ sensors for the boom with<br />
peak response in the 20-50 kHz range. This is strictly a compromise as the<br />
lower frequency increases the susceptibility to extraneous mechanical noise.<br />
It is rare to use "time of flight" source location on either the boom or metal<br />
components. Sources of acoustic emission that may occur are simply identified<br />
with the particular component. In this way, two sensors may be fed into a<br />
single amplifier chain. This reduces the capital cost of equipment at the<br />
expense of discrimination.<br />
Sensor spacing may be quite important. An artificial acoustic emission source,<br />
such as a puiser or pencil lead break [5], must be used to ensure that the<br />
required degree of coverage is achieved. However, this rigorous process need<br />
only be performed once for a given model of aerial lift truck. Sensitivity,<br />
which includes sensor performance, couplant and amplifier gain, must be checked<br />
and equalized for each test. This is a simple process of stimulating each channel<br />
with the artificial source and adjusting amplifier gains for equal<br />
response. Figure 7 illustrates commonly used sensor locations.<br />
The boom is then deployed in a manner such that it is representative of working<br />
configurations and such that a controlled load may be applied to the bucket<br />
end. in Figure 8 the load is applied through a cable winch and load cell.<br />
A predetermined load regime is followed consisting of the gradual imposition of<br />
load which is then held for a few minutes and then repeated [6], The applied<br />
load must be consistent with the total vehicle design [7], but this may be as<br />
high as three times, but more typically twice, rated load [8].<br />
TEST RESULTS<br />
For each stage of the load profile acoustic emission activity is recorded for<br />
each channel. The data are plotted in a number of ways, but would normally<br />
include count versus load, Figure 9, amplitude distributions, Figure 10 and<br />
perhaps an indication of individual channel activity through the test.<br />
Experience plays a large part in both conducting the test itself and in interpreting<br />
the results. At the present, there are no solid, universal accept/<br />
reject criteria. However, if acoustic emission continues into the hold period<br />
or recurs on the second load, there is strong indication that a problem exists.<br />
If the general activity, that is the overall count, is unusually high, concern<br />
would depend on which channels were active. For example, this type of indication<br />
could be due to a worn or poorly lubricated pin. The shape of the count<br />
versus load plot is also a significant factor used in interpretation.<br />
Fibre breakage represents high energy acoustic emission generation. By observation<br />
of the amplitude distribution, Figure 10, some assessment of fibre breakage<br />
during the test is available.
- 266 -<br />
Any recognizable fibre damage may be interpreted as cause for rejection, or at<br />
least further investigation.<br />
THE VALUE OF THE TEST<br />
Many utitlities in both Canada and USA either perform the test for themselves,<br />
contract out to testing agencies or are on the point of implementating a testing<br />
program. Frequency of testing varies from semi-annual, consistent with dielectric<br />
testing, to only at rebuild time, every ~8 years. There is no doubt<br />
that defects are discovered, but primarily in metal components, and there is<br />
some debate as to whether improvements to maintenance and inspection procedures<br />
would be of greater benefit than an acoustic emission test. However, acoustic<br />
emission seems to be the only reliable monitor of the structural integrity of<br />
the boom.<br />
Industry experience is that structural problems with the boom are rara, and<br />
personal injury related to such problems is rarer still. However, it may be<br />
argued that safety is not an issue that can be weighted against the cost of<br />
downtime, but the question must be asked as to whether the acoustic emission<br />
test enhances safety. Is the frequency of testing such that the test will<br />
identify incipient failure before it can progress to actual failure? The answer<br />
is probably yes if the vehicles are used within the design limits specified by<br />
the manufacturer and enforced through the codes [7], but no if the vehicle is<br />
liable to be subjected to abuse«<br />
One USA utility has attempted to address this problem by installing on-board<br />
monitors. These are single channel, battery powered acoustic emission systems<br />
which are attached to the critical area of the upper boom. In an overload<br />
situation which results in acoustic emission the monitor produces an audible and<br />
visual alarm. Unfortunately, this is not a panacea as false alarms do occur<br />
which may require the vehicle being taken out of service for verification, only<br />
a very small portion of the critical areas are covered and it is possible that<br />
the device can generate a false sense of security in the work crew. On the<br />
other hand, the most common failure site is at the lower insert, Figure 11,<br />
which is well covered by the onboard monitor, and there is no question that<br />
enormous amounts of acoustic emission will be generated even prior to a sudden<br />
avalanche failure in tension.<br />
A recent exhaustive inspection of forty-three manlifts and forty-nine digger<br />
derricks [4] discovered no mechanical defects classed as critical or needing<br />
immediate attention associated with frp boom components. However, very many<br />
other defects were discovered but, again, whether acoustic emission testing<br />
would have uncovered all of these (eg, missing bolts holding pedestal to truck)<br />
or is even the appropriate procedure is debatable.<br />
C<strong>ON</strong>CLUSI<strong>ON</strong>S<br />
There is no question that the frp material used for manlift booms generates<br />
acoustic emission at the early stages of structural degradation. Both compressive<br />
and tensile modes of failure are acoustically "noisy".
- 267 -<br />
The utilities using an acoustic emission proof test discover defects in metallic<br />
and frp components. Most of these are trivial and could be addressed in<br />
alternate approaches* but the bottom line is that critical defects are being<br />
discovered by the test, thus increasing the level of safety in the workplace.<br />
The proof test cannot protect against critical failure brought about through<br />
abuse. The on-board monitor can potentially provide some assurance during<br />
operation, exactly where it is needed, but may generate unwanted downtime due to<br />
false alarms.<br />
Catastrophic failure of bucket trucks is rare. This makes evaluation of an<br />
acoustic emission program based on statistical evidence alone very difficult if<br />
not impossible.<br />
REFERENCES<br />
1. J.R. Mitchell and D.G. Taggart, Acoustic Emission Testing Electric Utility<br />
Fleet Management, V. 2, No. 5, October/November 1983.<br />
2. J.W. McElroy, On-Board Acoustic Emission Monitoring of Fiberglass Boom<br />
Aerial Lift Trucks Society of Automotive Engineers, Congress on Exhibition,<br />
Detroit, February 1980.<br />
3. J.A. Baron, Status Report on Assuming the Structural Integrity of Aerial<br />
Personnel Devices Canadian Electrical Association, Study SD-124.<br />
4. K. Moore, Aerial Equipment Requires Thorough, Regular Inspection<br />
Transmission and Distribution, January 1984.<br />
5. A. Nielsen, Acoustic Emission Source Based on Pencil Lead Breaking,<br />
Svejsecentralen, Copenhagen, Publication 80.14.<br />
6. Acoustic Emission Test Method for Insulated Aerial Personnel Devices<br />
American Society for Testing and Materials proposed standard, Committee F18.<br />
7. CSA Standard C225, Canadian Standard for Vehicle-Mounted Aerial Devices,<br />
Canadian Standards Association.<br />
8. Procedure for Acoustic Emission Testing of Aerial Lift Devices, PSE&G<br />
Research Corporation, 1980.
- 268 -<br />
FIGURE 1<br />
EXPULSI<strong>ON</strong> OF FIBRES<br />
IN COMPRESSI<strong>ON</strong>AL FAILURE
a<br />
<<br />
o<br />
TIME<br />
- 269 -<br />
LOAD 2L<br />
FIGURE 2<br />
LOAD L<br />
THEORETICAL RELATI<strong>ON</strong>SHIP BETWEEN NUMBER OF FIBRES<br />
TO BREAK AND TIME (C<strong>ON</strong>STANT LOAD)<br />
TÎME TO FAILURE<br />
^ — — - _<br />
FIGURE 3<br />
THEORETICAL RELATI<strong>ON</strong>SHIP BETWEEN LOAD AND TIME<br />
TO FAILURE FOR GIVEN NUMBER OF FIBRES IN BOOM<br />
(c.f. FIGURE 4)
100<br />
95<br />
LU LU<br />
5 S 90<br />
^"-85<br />
K<br />
H H 80<br />
Utu 75<br />
70<br />
- 270 -<br />
VARIABLE<br />
TIME C<strong>ON</strong>STANT<br />
PRE-LOAD STRENGT<br />
DEGRADATI<strong>ON</strong><br />
TIME TO<br />
SPECIMEN FAILURE<br />
65<br />
100 200 400 1000 2000 4000 10000 20000 40000<br />
TIME (SEC<strong>ON</strong>DS)<br />
FIGURE 4<br />
THE RELATI<strong>ON</strong>SHIP BETWEEN LOAD, TIME. AND<br />
RESIDUAL STRENGTH FOR FIBERGLASS SAMPLES<br />
REPRODUCED FROM REFERENCE<br />
50%<br />
100000
1000,-<br />
- 271 -<br />
(C) RELOAD AFTER 21 HOURS<br />
(B) RELOAD AFTER 15 MINUTES<br />
TIME<br />
FIGURE 5<br />
ACOUSTIC EMISSI<strong>ON</strong> AT FIXED LOAD<br />
SHOWING TIME DEPENDENT RELAXATI<strong>ON</strong><br />
10 mins
- 272 -<br />
FIGURE 6<br />
SENSORS AND PREAMPLIFIERS ATTACHED<br />
TO BOOM USING ADHESIVE TAPE
OUTRIGGER<br />
UPPER BOOM<br />
- 273 -<br />
FIGURE 7<br />
TYPICAL SENSOR DISTRIBUTI<strong>ON</strong><br />
FIBREGLASS INSERT
900<br />
800<br />
700<br />
i/> 600<br />
H<br />
Z<br />
O 500<br />
U<br />
400<br />
300<br />
200 -<br />
100 -<br />
0<br />
- 274 -<br />
FIGURE 8<br />
BOOM C<strong>ON</strong>FIGURATI<strong>ON</strong> DURING TEST<br />
LOAD APPLIED BY CABLE WINCH<br />
ALL FIBERGLASS CHANNELS<br />
r<br />
.J"<br />
ALL METAL CHANNELS<br />
250 500<br />
LOAD(kg)<br />
FIGURE 9<br />
SUM OF COUNT VERSUS LOAD BY<br />
BOTH FIBRECLASS AND METAL CHANNELS
ENTS<br />
LU<br />
LL<br />
o<br />
/IBER<br />
ID<br />
z<br />
400<br />
300<br />
200<br />
100<br />
1<br />
- 275 -<br />
_ 1 4<br />
MATRIX FIBRE<br />
t<br />
r •)<br />
I<br />
/<br />
BEHAVIOR<br />
yr i<br />
BREAKAGE •-<br />
20 tO 60 80<br />
100<br />
AMPLITUDE (dB)<br />
FIGURE 10<br />
N<strong>ON</strong>-ACCUMULATED AMPLITUDE DISTRIBUTI<strong>ON</strong><br />
(LIN-LOG SCALE)<br />
FIGURE 11<br />
BOOM FAILURE AT LOWER INSERT<br />
IN A DESTRUCTIVE TEST
- 276 -<br />
ACOUSTIC SORTING OF GRINDER BALLS<br />
F. Na.de.au. and J.F. Zixalnfin<br />
national Re.itax.ch Council o i Canada, Boucke.n.v±lle., Qntan-io<br />
An automated non destructive testing device for ball mill grinder balls<br />
was developed using acoustic flaw detection. An initial study of the sound<br />
generated in air by the impact of grinder balls has shown that the changes<br />
in ball velocity upon impact produce an initial low frequency pressure<br />
pulse and that various resonances exited in the impacted spheres produce the<br />
higher frequency oscillations that follow. Resonances of uncracked balls<br />
were related to a particular spheroidal mode of vibration while resonances<br />
of cracked balls are attributed to "tuning fork" modes of vibrations.<br />
Both a laboratory and an industrial prototype that perform the sorting<br />
of cracked and uncracked grinder balls using the differences in their<br />
acoustic properties were built and tested. A 97% success rate was achieved<br />
for 3" diameter balls. The device is being fitted to the production line.<br />
INTRODUCTI<strong>ON</strong><br />
Many of the more sophisticated Non Destructive Testing (N.D.T.)<br />
techniques are expensive. This often restricts their use either to the<br />
aerospace or the nuclear industries. When an N.D.T. problem involves a low<br />
mark-up non critical product, emphasis is placed on the per unit cost of the<br />
test. Automation becomes important while 100% reliability is no longer<br />
required. This paper presents such an example where a cheap, fast,<br />
automated N.D.T. method was developed using audio range acoustics to detect<br />
flaws in grinder balls. Grinder balls are hardened steel balls used in<br />
stone grinding "ball mills". Those used in this study are forged from high<br />
carbon steel rods to roughly spherical shapes and waterquenched to a<br />
hardness of about 60 Re. A non négligeable proportion of the production<br />
exhibits deep running cracks and therefore tend to rupture under operating<br />
conditions inside the ball mill. This problem is familiar to ball mill<br />
operators who often have to manually remove the broken parts from inside the<br />
mill to ensure smooth operating conditions and prevent exercise wear of the<br />
mill's inner sleeve. Testing of the production is done on a statistical<br />
basis (samples of each batch) by manual inspection. Cracks can be detected<br />
visually and also acoustically: two balls are knocked together and a<br />
characteristic high pitch sound is heard if one of them is cracked.<br />
Acoustic, or "sonic" inspection is probably one of the oldest N.D.T.<br />
method. The sound of a body's natural resonances has been used to evaluate<br />
material properties (1, 2, 3), size and shape (4), and also for flaw<br />
detection (5). In the case of grinder balls, a study of the acoustical<br />
properties of good and cracked balls confirmed that repeatabledifferences in<br />
their natural sounds could be electronically detected. These results are<br />
presented in the first part of this article.<br />
An automatic grinder ball tester was designed and both a laboratory and<br />
an industrial prototype were built and tested. The second part of this<br />
article contains description of the laboratory machine along with some test<br />
results.
- 277 -<br />
1. THE ACOUSTIC HAVE GENERATED IN AIR BY THE IMPACT OF GRINDER BALLS<br />
When two grinder balls are knocked together in ambient air, an impact<br />
sound is generated which has several distinct features. Fig. 1 shows two<br />
typical waveforms picked up by a Bruel & Kjaer 1/4" microphone (# 4135) flat<br />
to 70 kHz and captured on a digital waveform analyzer. The waveform on the<br />
left was produced by the impact of two good balls while the one on the<br />
right, by the impact of a good and a cracked ball. The balls, 3" in<br />
diameter, are hand held with the target ball near the microphone and the<br />
"anvil", always a good ball, on the opposite side. A close examination of<br />
these waveforms reveals several distinct features. First of all, an initial<br />
pulse is always present at the beginning of each waveform, whether the balls<br />
are cracked or not. Also common to both waveforms is the low frequency<br />
rumble that follows. This rumble is due to reflections on surrounding<br />
structures and low frequency room resonances. Tests conducted in an<br />
echo-free room did not show such low frequencies. Following the initial<br />
pulse, high frequency resonances are present in both waveforms. In the case<br />
of the "good" impact waveform, the amplitude of the resonance is small, its<br />
frequency is around 40 kHz and its damping is low. In the case of the<br />
"cracked" impact, the resonance's amplitude is much greater, while its<br />
frequency is lower (around 5-15 kHz) and it is more damped. These<br />
features were individually studied and the results are presented in the<br />
following discussion.<br />
The initial pulse<br />
An initial pulse of width around 200 ys such as the one of Fig. 1<br />
appears at the beginning of every waveform recorded, whether good or cracked<br />
balls are used for the impact. This relatively low frequency pulse<br />
( 3 kHz) can be explained by considering the overall motion of the balls as<br />
a source of acoustic waves in air.<br />
Let us first recall the relation between particle displacement and<br />
acoustic pressure p for plane waves in air (6):<br />
p = - poc 2<br />
where: po is the density of air<br />
c is the velocity of sound in air<br />
x is the equilibrium position.<br />
We also have for plane waves propagating in one direction in air:<br />
Replacing in (1), we obtain:<br />
Se<br />
ôx<br />
e = f(ct - x)<br />
6e 1 6e<br />
i.e. — =<br />
6x c 5t
- 278 -<br />
6e<br />
P = Po c - St<br />
or p = p0 eu where u is the particle velocity<br />
Thus an infinite flat surface moving in air at a velocity u(t) will<br />
generate a pressure wave p(t) as given by Eq. (4), considering that the<br />
particles in contact with the surface have the same displacement. For<br />
spherical surfaces, one can no longer assume plane waves which means that a<br />
directivity factor along with a radial dependence on amplitude would<br />
appear. Still the time dependence of p(t) and p(t) would essentially be the<br />
same for points very near the surface. Hence, when an object is, for<br />
example, accelerated, its front face generates a compressive wave while its<br />
back face generates a rarefaction wave. This is easily verified<br />
experimentally as illustrated in Fig. 2: a grinder ball is accelerated as<br />
it is struck by an other one. When the microphone is placed near the front<br />
face of the target ball, it registers a positive pressure pulse. When<br />
placed at the rear of the target ball, it registers a negative pressure<br />
pulse. The ball thus acts as a dipolar source of length approximately equal<br />
to d, the diameter of the ball.<br />
This was also verified by measuring the width At of pulses obtained<br />
using grinder balls of various diameter. The microphone was located on the<br />
collision axis as illustrated in the upper part of Fig. 2. Therefore we<br />
expect the fall in pressure to be delayed with regards to the rise in<br />
pressure by the propagation time over a distance equal to the diameter "d"<br />
of the ball, thus producing a positive pulse of width At equal To d/c, where<br />
c is the velocity of sound in air. Fig. 3 is a plot of t vs. d. We see a<br />
good agreement with this model as the experimental data points cluster<br />
around the straight line of slope equal to 1/c.<br />
We can also look at the shape of the pulse and relate it to p(t), the<br />
time evolution of the velocity of the ball. In particular, the rise time of<br />
the pulse is expected to be equal to the rise time of the velocity of the<br />
ball i.e. the frequency content of the pulse would be limited by the<br />
interaction time T of the collision. We therefore calculate this<br />
interaction time T for the simple case of two spheres of diameter d and mass<br />
m colliding at a velocity v0 the relative velocity of one sphere with<br />
regards to the second one is 2 . v0.<br />
The symmetry of the problem allows the second sphere to be replaced<br />
with a fixed and infinitely rigid wall. With position x = 0 and time t = 0<br />
chosen when the sphere touches the wall, we have (7):<br />
F = - 5.182 fi (d.x 3 H<br />
where F is the reaction force of the wall on the sphere. When t = T/2,<br />
the work of that force on the center of mass of the sphere has brought it to<br />
a halt. Therefore we have:<br />
jxm F(x)dx = 1/2 m vo 2
- 279 -<br />
where x8 is the maximum value of x. Solving for xm yields:<br />
/ -VA<br />
xm=l.A22l _ I<br />
\ E 2 d /<br />
We can approximate the equation of motion of the sphere by a sine:<br />
x xm sin uit<br />
and knowing that when t = 0, dx/dt = vo, we have: (o = vo/xo<br />
The interaction time, T, corresponds to a half cycle of the sine wave:<br />
Thus, finally:<br />
T = TT/ÜJ<br />
= IT Xm/Vo<br />
In the case of 3" diameter balls, the numerical parameters are:<br />
m = 1.8 kg<br />
d = 76 ram<br />
E = 2 x 10 5 MPa<br />
vo= 1 m/s<br />
width yields:<br />
T = 280 us<br />
This value is in good agreement with the experimental data and offers<br />
and explanation to the absence of frequency components higher than 2-3 kHz<br />
in the initial pulse.<br />
Resonances of uncracked grinder balls<br />
Sound measurements of uncracked grinder ball impacts were made using a<br />
high pass filter to attenuate the low frequency noise associated with the<br />
initial pulse (up to 3 kHz as mentionned above). The filtered signals show<br />
the enhanced presence of a high frequency resonance in the acoustic waveform<br />
ranging from 35 kHz for 3" dia balls to 65 kHz for H" dia balls. This<br />
oscillation can be related to one of the lower modes of resonance of a solid<br />
steel sphere. Fig. 4 is a plot of the inverse of the resonance frequency as<br />
a function of ball diameter. The straight lines represent theoretical<br />
curves for the first 3 modes of vibration of solid steel spheres as<br />
calculated from relations given in (8), and with elastic constants given in<br />
1/5
- 280 -<br />
(9) for hardened steel. All other modes fall below the 2S1 mode and so were<br />
not considered. The data points indicate that either the 1S2 mode or the<br />
1T2 mode is the one that generates the high frequency acoustic waves.<br />
However, for torsional or "T" modes in spheres there is no radial<br />
displacement and therefore it is assumed that these modes do not radiate any<br />
sound in air. Only spheroidal or "S" modes can therefore couple with air<br />
and generate an acoustic wave so that the 1S2 mode is believed to be the one<br />
responsible for the high frequency tone. A representation of the<br />
displacement associated with the 1S2 mode is included in Fig. 4. It shows<br />
that this mode is very likely to be strongly exited by an impact because<br />
such a uniaxal force will ovally deform the sphere. Other modes are<br />
probably exited as well, including higher S modes. Some inharmonic<br />
distortion was indeed observed on several high frequency waveforms. This is<br />
expected if other modes are exited since their frequencies are not exact<br />
multiples of each other.<br />
Resonances of cracked grinder balls<br />
When similar acoustic measurements are made using cracked grinder<br />
balls, the appearance of medium range frequencies (5 - 15 khz) in the<br />
waveforms is observed. These oscillations look like damped resonances.<br />
Disappearance of the high frequency 1S2 vibration is also observed,<br />
especially for severely cracked balls for which the amplitude of the<br />
mid-range resonances is often the highest. This amplitude also depends on<br />
the point of impact i.e. the orientation of the cracked ball relative to the<br />
anvil. This effect however varies from one cracked ball to another and no<br />
repeatable pattern emerged.<br />
The mid-range resonances, characteristic of cracked grinder balls, are<br />
attributed to some sort of "tuning fork" mode of vibration as the ball is<br />
partially divided by deep running cracks. Acoustic measurements made on<br />
artificially flawed grinder balls (saw-cut) reveal the presence of at least<br />
two modes of vibration at frequencies lower than the 1S2 mode of the solid<br />
sphere. The data is presented in Fig. 5 where the frequencies of the<br />
various resonances observed are plotted against the depth, a, of the<br />
saw-cut. We see that mid-range frequencies similar to those produced by<br />
naturally cracked balls appear for values of a in the range of 30% to 70% of<br />
the diameter d of the ball. The 1S2 amplitude becomes unmeasurable when a<br />
reaches about half the diameter of the ball. The assumed displacements<br />
corresponding to the two "tuning fork" modes are also illustrated in<br />
Fig. 5. This assumption is in agreement with the observed variations in the<br />
relative amplitude of the two frequencies with the orientation of the<br />
impact: when directed perpendicular to the plane of the saw-cut, the lower<br />
frequency mode was preferably exited whereas when directed parallel to the<br />
bottom of the saw-cut, the higher frequency "shear" mode was preferably<br />
exited. The frequency of both modes understandably tends towards zero when<br />
the value of a approaches the value d.
281 -<br />
Finally, resonances of saw-cut balls were practically undamped compared<br />
to those of naturally cracked balls. Evidently the latter are damped by<br />
friction occuring along the crack interface. Consequently, one can also<br />
assume that the shear motion depicted in Fig.5 for the "shear" mode is<br />
unlikely to occur in naturally cracked balls.<br />
2. DEVELOPMENT OF AN ACODSTIC CRACK DETECTOR FOR SORTING GRINDER BALLS<br />
Principle of operation<br />
We have seen that noticeable differences exist between the sound<br />
produced by the impact of uncracked grinder balls and the sound produced by<br />
the impact of cracked ones. From this, it was concluded that a device that<br />
would use these differences to acoustically detect and sort out the cracked<br />
grinder balls from the good ones was technically feasible and a laboratory<br />
scale prototype was developed. Figure 6 is a photograph of the device. The<br />
mechanical assembly allows the balls to drop from a constant height on a<br />
fixed anvil. It is composed mainly of 4" PVC tubing supported by a bolted<br />
"Dexion" frame. The anvil is a 2" steel cube mounted in a foam filled box.<br />
A 6" long aluminium rod mounted on a hinge inside the horizontal sorting<br />
tube and activated by a solenoid deflects the cracked balls toward the first<br />
opening, while the others exit by the end opening.<br />
The impact sound is picked up by a nearby microphone and electronically<br />
analyzed. The ball finally exits by one of two ports depending on whether<br />
or not a crack has been detected. The electronic circuit "listens" for<br />
those mid-range resonances typical of cracked grinder ball impacts that<br />
immediately follow the initial pulse. It measures the "amount" of such<br />
resonances and uses it as a criterion to decide whether the ball is good or<br />
bad. Figure 7 is a simplified diagram of the circuit. The microphone is a<br />
Bruel & Kjaer i" microphone and power supply flat to 30 kHz. It is located<br />
in the target area, as near as possible to the trajectory of the grinder<br />
balls, positioned downwards to prevent any buildup of dust or other<br />
pollutants on its membrane. The detector consists of a band pass filter<br />
(10-20 kHz), a rectifier and a gated integrator. The output voltage thus<br />
represents the amount of acoustic energy between 10 and 20 kHz integrated<br />
over the time window defined by a gate signal. It is compared to a<br />
reference voltage to determine whether the ball is cracked or not. The<br />
logic block carries out the measurement sequence. It uses the "initial<br />
pulse" as a trigger event, resets the integrator and holds it on reset for<br />
about 1 ms. This adjustable delay determines the start of the time window<br />
and allows the blanking out of the initial pulse. It then frees the<br />
integrator for about 10 ms and then interrupts the integration. This<br />
defines the width of the time window, the end of which is positionned just<br />
before the occurence of a second collision due to the recoil of the anvil<br />
bouncing back on the grinder ball. While the time window is open, the<br />
output of the integrator is compared to an adjustable reference and the<br />
output of the comparator commands the operation of the sorting device. As<br />
the time window closes, the trigger of the logic block is disabled for about<br />
100 ms to prevent retriggering on the second collision mentionned above or<br />
any other spurious noises. After that delay, it is ready to be retriggered<br />
when an other ball hits the anvil. Until then, the integrator is on hold,
- 282 -<br />
but not reseted, so the comparator, and therefore the sorting device, stays<br />
in the same state.<br />
One major constraint is the signal-to-noise ratio. Sources of acoustic<br />
noise, especially those occurring at the moment of impact and in the<br />
frequency range of cracked ball resonances, have to minimized. Of<br />
particular importance in this matter is the anvil. It has to be designed so<br />
that the frequency of its lowest mode of resonance is much higher than<br />
cracked ball resonant frequencies. The resonant frequencies of the<br />
prototype's anvil, a 2" steel cube, were measured and found to be all higher<br />
than 25 kHz. The anvil has to be mounted on some sort of damped suspension<br />
(foam rubber in our case) so that when it is struck by a grinder ball, the<br />
high frequency part of the impact force is not transmitted to the supporting<br />
structure and therefore does not ring high frequency resonances in it. For<br />
this reason and also to maximize the portion of the impact energy<br />
transmitted to the grinder ball, one would tend to make it as massive as<br />
possible. The limiting factor is that the resonant frequencies diminish<br />
with increasing size. It is possible to increase the mass of the anvil<br />
without decreasing its resonant frequency by optimizing its shape. In this<br />
respect, the sphere has the highest mass . frequency product followed by the<br />
square cylinder (length equal to diameter), followed by the cube etc... The<br />
cylinder is probably the most convenient compromise for automated systems<br />
but, in theory, the best anvil to test a particular size of grinder ball is<br />
another uncracked one of the same size.<br />
Another design constraint is that the rate at which grinder balls can<br />
be tested is limited by the travel time of a ball from the moment of impact<br />
to the moment it clears the mechanical sorting device; the electronics are<br />
not the limiting factor, as the good or cracked diagnosis is done before the<br />
ball actually leaves the anvil. This travel time was more than half a<br />
second for the rudimentary sorting tube of the laboratory prototype but<br />
could easily be improved for an industrial machine.<br />
Field Test Results<br />
Sorting tests were conducted mainly on 3" diameter balls- A sample of<br />
approximately 100 3" diameter balls was carefully hand sorted so as to form<br />
a "good" pile and a "cracked" pile. Each time a ball was put through the<br />
machine, the integrator output level was recorded. The data is presented in<br />
the form of histograms (Figure 8). A normal curve has been drawn through<br />
the good ball distribution and the 2.5 V. threshold is indicated.<br />
Ideally, the distributions of good balls and of cracked balls would be<br />
two delta functions and no matter how close to one another they would be,<br />
one would always be able to position the threshold between them and achieve<br />
100% success rate. However, in reality, a number of factors widen the<br />
distributions causing overlapping and, thus, diminishing the success rate.<br />
In the case of cracked balls, a number of those factors are uncontrollable.<br />
They are: severity of the crack, geometry of the crack, direction of impact<br />
with regards to crack and microphone orientation with regards to crack. The
- 283 -<br />
effect of the last two could be averaged out by having each ball run through<br />
the test several times but this is not necessary to achieve satisfactory<br />
success rates. Other factors that widen all the distributions are<br />
controllable to a degree. These are: velocity of the incoming ball,<br />
direction of impact and location of impact on the anvil. All these<br />
parameters are rerated to the trajectory of the incoming ball. The more<br />
repeatable it is, the finer the distributions will be. This allows the gain<br />
to be increased, moving the good ball distribution closer to the threshold<br />
and thus pushing the cracked ball distribution further away from it. The<br />
distributions obtained with this prototype were narrow enough to obtain very<br />
good success rates (97X of cracked balls recognized with no good balls<br />
misdiagnosed). However, one parameter in particular, ball velocity, varied<br />
significantly and is believed to cause most of the spreading of the<br />
distributions. A significant improvement is expected when the feeding<br />
system is redesigned to fit a tester on the production line.<br />
ACKNOWLEDGEMENTS<br />
The author wishes to thank Mr. Ghislain Vaudreuil of the Industrial<br />
Materials Research Institute and also Mr. George Chapman of the Steel<br />
Company of Canada for their precious contributions to this work.<br />
REFERENCES<br />
1. Grime, G., "Determination of Young's Modulus for Building Material by a<br />
Vibration Method", Philosophical Magazine, Vol. 7, No. 20, 1935,<br />
pp. 304-310.<br />
2. Prigge, R.E., "Correlation of Modulus of Rupture and Modulus of<br />
Elasticity", ß.S. Thesis, New York State College of Ceramics,<br />
(May 1951).<br />
3. Hornibrook, F.B., "Discussion on Sonic Method for Modulus of<br />
Elasticity", A.S.T.M. Proceedings, Vol. 39, 1939, pp. 996-998.<br />
4. Rubin, G.A., Leach, M.F., "Particle Size and Shape Characterization from<br />
Acoustic Emissions", Proc. 82nd Annual Conference of the ACS/CEC/NCE,<br />
Chicago, April 1980.<br />
5. Rowe, K..G., "Vibration Apparatus for Testing Articles", U.S. Patent<br />
#2 486 984 (Nov. 1949).<br />
6. Kinsley, L.E., Frey, A.R., Fundamentals of Acoustics, 2nd edition, 1962,<br />
p. 112. John Wiley & Sour, Inc., New York.<br />
7. Roark, K.J. & Young, W.C.. Formulas for Stress and Strain, 5th edition,<br />
1975, p. 516. Me Graw-Hill, Inc., New York.<br />
8. Schreiber, E., Anderson, O.L. & Soga, N. Elastic Constants and their<br />
Measurement, 1st edition, 1973, pp. 126-142. Me Graw-Hill, Inc.,<br />
New York.<br />
9. Krautkramer, J. & H., Ultrasonic Testing of Materials, 2nd edition, 1977,<br />
p. 620. Springer-Verlog, Berlin.
- 284 -<br />
GOOD CRACKED<br />
0 2 4 6 8 0 2 4 6 8<br />
t(ms)<br />
Figure 1: Acoustic waveforms produced by the impact of good balls (left)<br />
and cracked balls (right).<br />
0 2 4 6 8 0 2 4 6 8<br />
t(ms)<br />
Figure 2: Influence of the nicrophone location on the polarity of the<br />
initial pulse.
At (/xs)<br />
300 r<br />
200<br />
100<br />
- 285 -<br />
y<br />
/i-.\ Slope = c<br />
_ i i i i i<br />
0 20 40 60 80 100<br />
d (mm)<br />
Figure 3: Width At of the initial pulse, as a function of d, the ball<br />
diaaeter.<br />
(X10 5 S) 1T2<br />
j i I i i i<br />
20 40<br />
d(mm)<br />
60 80<br />
Figure 4: Besonance period as a function of ball diaaeter.
0<br />
- 286 -<br />
Figure 5: Resonance frequencies of saw-cut balls as a function of the depth<br />
"a" of the cut.
- 287 -<br />
Figure 6: Grinder ball crack detector and sorter.<br />
DETECTOR<br />
Figure 7: Schematic diagran of the crack detector.
- 288 -<br />
GOOD<br />
0 .8 1.6 2.4 3.2<br />
Level (V)<br />
4 5.6<br />
Figure 8: Histograms of integrator output level for good and cracked balls.
- 289 -<br />
THE BENEFITS OF NDT TRAINING FOR <strong>CANADIAN</strong>S<br />
Lynda B. Uanzzn.<br />
Canadian Soc.ie.ty ^on. Monde.* tractive. Tz&ting Foundation<br />
HAMILT<strong>ON</strong>, Ontario<br />
Although industry is becoming increasingly aware of the<br />
benefits to be derived from nondestructive testing, many of these<br />
same firms have yet to realize the importance and impact of effectively<br />
educating NDT personnel.<br />
The author will outline the history of NDT training in<br />
Canada, the types of programs presently available, the typical<br />
student profile, and who the target market could (and should) be.<br />
The author will then discuss the effects of NDT training and education<br />
in Canadian industry.
- 290 -<br />
THE BENEFITS OF NDT TRAINING FOR <strong>CANADIAN</strong>S<br />
It is generally ac'cepted that the role of nondestructive testing is<br />
to prevent human injuries and to save lives, to increase productivity,<br />
and to make a profit for the user. For those of us who have heard,<br />
or used, this phrase over and over through the years, let us not allow<br />
repetition to desensitize us to the enormous responsibility that this<br />
places on those involved in nondestructive testing (NDT).<br />
This inherent factor of responsibility demands that personnel responsible<br />
for, or involved with, a company's NPT section receive some<br />
form of post-secondary school education in this technology. A technology<br />
which is essential in an industrial nation such as Canada.<br />
Yet the history of NDT training in Canada, and the numbers and types<br />
of students presently in training, would seem to indicate that the<br />
educational component has not kept pace with the phenomenal growth<br />
in the use and applications of nondestructive testing. That may<br />
seem a shocking statement from someone such as myself, a representative<br />
of the Canadian Society for Nondestructive Testing Foundation<br />
which is the leading educator in NDT in Canada. But who better to<br />
relay to you our concern that not enough people in business and industry<br />
realize the importance of training their personnel, or hiring<br />
well-trained, competent individuals to carry out NDT functions, or<br />
to supervise these functions.<br />
In the early days of NDT, much of the training was conducted on-thejob;<br />
personnel became proficient (or, did not become proficient)<br />
through trial and error. In 1940, the Society for Nondestructive<br />
Testing (SNT) was established in the United States, and in 1953, the<br />
first Canadian Section of SNT was formed. In 1954, a second section,<br />
the Eastern Canada Section of SNT, was formed in Montreal. A small,<br />
select group of Canadians realized the importance of nondestructive<br />
testing and the importance of disseminating information on NDT technology<br />
.<br />
These Sections conducted a few continuing education courses and, in<br />
effect, the pattern was set for NDT training in Canada to be the<br />
short, intensive-type of course. In the 1950's, and even today,<br />
equipment suppliers offered training courses which followed the same<br />
pattern - courses of three to five days duration.<br />
In 1964, the Canadian Council for Nondestructive Technology (C.C.N.D.T.)<br />
was formed with the following objectives:<br />
-to advance scientific, engineering, and technology knowledge<br />
in the field of nondestructive testing.<br />
-to gather and disseminate information relating to nondestructive<br />
testing useful to the individual and beneficial to the<br />
general public.<br />
-to promote nondestructive testing through courses of instruction,<br />
lecture, meetings, publications or other means.
- 291 -<br />
The Canadian Society for Nondestructive Testing (C.S.N.D.T.) was<br />
formed from C.C.N.D.T., and with the similar objectives, in 1967<br />
and the NDT community had a truly Canadian Society. The increased<br />
interest in NDT an,d support of a National Society in the late 50's<br />
and early 60's paralleled a new emphasis, in Canadian education, on<br />
mathematics and sciences. It may be that the Vocational Training<br />
Act, passed in 1961, emphasized the need for training in new technologies,<br />
although the federal dollars did not filter through to a<br />
Canadian institution specifically for NDT training.<br />
It was also in 1960 that the Canadian Government Specifications<br />
Board (C.G.S.B. and now the "Canadian General Standards Board") introduced<br />
C.G.S.B. Standard 48-GP-4 "The Certification of Industrial<br />
Radiographie Personnel". This Standard established a Junior and<br />
Senior level of certification, outlining the requisite amount of<br />
work experience and the responsibilities of each level. Standards<br />
for certification in Ultrasonics, Liquid Penetrant, Magnetic Particle<br />
and Eddy Current followed.<br />
C.S.N.D.T. Chapters across Canada continued to offer continuing education<br />
programs and increased the educational input of the Society by<br />
conducting seminars. By the mid 70's, it was evident that assistance<br />
was needed to provide industry with trained NDT personnel from the<br />
Canadian labour force. It should be noted that some companies were<br />
still finding it necessary to hire Europeans who had good NDT backgrounds<br />
and to send personnel to the United States for training<br />
courses. Also at this time, the Department of Energy, Mines and<br />
Resources (the Examining Authority for C.G.S.B.) indicated their need<br />
for assistance in conducting the certification program.<br />
In 1976, the educational arm of C.S.N.D.T. was established as the<br />
Canadian Society for Nondestructive Testing Foundation. This nonprofit<br />
organization, headquartered in Hamilton, Ontario, has three<br />
main objectives:<br />
- to improve the quality of education in nondestructive<br />
testing throughout Canada.<br />
- to assist Canadian industry in its use of nondestructive<br />
testing<br />
- to assist the Department of Energy, Mines and Resources<br />
in conducting the written and practical examinations for<br />
certification of nondestructive testing personnel, in<br />
accordance with the Standards of the Canadian General<br />
Standards Board.<br />
These objectives outlined a major undertaking; previously, there had<br />
not been an organization charged with the responsibility of providing<br />
intensive, daytime NDT programs. The Nondestructive Testing Section<br />
of the Department of National Defense, Aircraft Maintenance Development<br />
Unit, based in Trenton, Ontario had been (and still is) a major<br />
contributor to education and certification in Canada. The Foundation's<br />
responsibility, however, is to the entire spectrum of NDT in Canadian<br />
i ndus try.
- 292 -<br />
C.S.N.D.T. Chapters continue to conduct evening educational programs,<br />
some of which are in co-operation with local community colleges.<br />
This automatically limits the geographical area served. The Foundation,<br />
in order to meet its objectives, must provide programs wherever<br />
and whenever they are needed.<br />
In its first full year of operation, 1977, the C.S.N.D.T. Foundation<br />
conducted eight (8) courses with a total student enrolment of one hundred<br />
and fifteen (115). These courses were the short, intensive-type<br />
and classed as "Regular Courses"; in other words, open to industry in<br />
general.<br />
In 1978, the Foundation added Practical Workshops and In-Plant Trainin]<br />
to its list of available programs. The following chart gives a statistical<br />
account of student enrolment from 1977 to 1983:<br />
FIGURE 1<br />
C.SMJDX FOUNDATI<strong>ON</strong><br />
STUOGMT ENROLMENT<br />
I97Y ~1963<br />
1977 1976 1979 1980 )9Q\ (982 1983<br />
Regular courses -Practical Workshops In-Plant Training
- 293 -<br />
These figures are much more than an exciting and encouraging indication<br />
of the Foundation's growth and its established reputation. Student<br />
enrolment in these types of programs, conducted by the Foundation,<br />
grew from a total of 115 in 1977 to 758 in 1983 - an increase of 559%!<br />
Furthermore, the C.S.N.D.T. Foundation is not the only NDT educator in<br />
Canada. Companies run their own in-house educational programs, equipment<br />
suppliers offer NDT courses, and C.S.N.D.T. Chapters have continued<br />
with their evening courses. There is, very definitely, an increase<br />
in the recognition of the importance of educating NDT personnel.<br />
In Figure 1, note the changes in the enrolment figures for Regular<br />
Courses and In-Plant Training Programs between 1981 and 1983; a decrease<br />
in attendance at Regular Courses and a dramatic increase in<br />
In-Plant Training Programs.<br />
There has also been a change in the students' training objectives<br />
since 1981. Prior to this, I estimate that 80% of the students in any<br />
of these programs were preparing for C.G.S.B. certification. After<br />
1981, there has been little change in the training objectives of students<br />
in Regular Courses, but a decrease to approximately 50% of the<br />
students in In-Plant Programs who are preparing for certification.<br />
Many of the companies now contracting with the Foundation to teach "onsite"<br />
are providing their employees with a general knowledge of nondestructive<br />
testing. These are not necessarily NDT technicians; they<br />
may be engineers, marine surveyors, procurement inspectors, quality<br />
assurance managers, quality control managers, production superintendents<br />
and foremen - all of whom are representative of the disciplines which<br />
must inter-relate with the actual NDT personnel in order to ensure an<br />
economical, quality product. The increase of programs of the "need to<br />
know" type, as opposed to the "how to do", underlines the growing<br />
recognition that knowledge of nondestructive testing is essential to<br />
the decision makers.<br />
I am not advocating, nor am I suggesting, that there has been a swing<br />
away from the certification of NDT personnel. To the contrary, Canada<br />
has one of the best certification systems in the world, and if more<br />
companies would require C.G.S.B. certification in their inspection<br />
procedures, at least, we would be assured of the minimum level of<br />
competence of their NDT technicians. We are well aware of the ramifications<br />
of a nondestructive test improperly performed or improperly<br />
interpreted.<br />
Then why aren't more of us, who know nondestructive testing and understand<br />
our responsibility for the welfare and safety of the Canadian<br />
public, demanding that all NDT technicians be educated and trained in<br />
specific applications, and their knowledge verified? In order to ensure<br />
top quality, professional and ethical technicians, we need:<br />
- a responsible employer<br />
- a receptive trainee<br />
- a responsible educator<br />
- a responsible regulatory body<br />
AND<br />
- a responsible, professional National Society
- 294 -<br />
The groundwork is in place. The increased emphasis on education,<br />
which has already been pointed out, has to reflect on the employers.<br />
In my own contact with hundreds of students, I have noticed a change<br />
in attitude; more ßtudents have a real desire to learn, and to learn<br />
more than just the basics. We certainly have responsible educators,<br />
the Foundation being one of the leaders in this field.<br />
Through the Canadian General Standards Board, and its Examining<br />
Authority (E.M.R.), we have a responsible regulatory body. The certification<br />
standards have been changed to a three level system (rather<br />
than Junior and Senior) and classroom training is now mandatory, in<br />
addition to the required work experience. There are approximately<br />
3,000 certified NDT technicians in Canada and, until the recent economic<br />
recession, there had been a steady increase in examination candidates.<br />
The C.S.N.D.T. Foundation, fulfilling its objectives, conducts<br />
about 60% of the required practical and written examinations; Figure<br />
2 shows the numbers of candidates who have made use of the Test Centre<br />
facilities at the Foundation from 1977 to 1983:<br />
FIGURE 2<br />
300<br />
CERTIFICATI<strong>ON</strong> EXAMINATI<strong>ON</strong>S CC.G-S.ß.)<br />
C<strong>ON</strong>OUCTBO BY THE C.S.N.P.T. FoUMDATIOIV<br />
\977 ~ \9S3<br />
1977 1978 1979 1980<br />
- V«nrrer* EXAMINATI<strong>ON</strong>*<br />
EXAMINATI<strong>ON</strong>S<br />
1981 1982 1983
- 295 -<br />
In the Canadian Society for Nondestructive Testing, we have a responsible<br />
National Society. The.Society provides its members with an excellent<br />
technical Journal, communicates at the national and international<br />
level with other technical Societies, provides representation<br />
on the C.G.S.B. subcommittees for certification of NDT personnel,<br />
and conducts seminars and conferences. These services are vital in<br />
keeping the NDT community in Canada up-to-date on recent developments<br />
and providing a forum for discussion. C.S.N.D.T. has taken a major<br />
step in ensuring high quality, uniform education for nondestructive<br />
testing personnel in Canada. I refer to "C.S.N.D.T. Procedure - ED-1"<br />
which states :<br />
This document identifies the conditions and procedures<br />
under which the Canadian Society for Nondestructive<br />
Testing Inc., and the Canadian Society for Nondestructive<br />
Testing Foundation (hereafter referred to as the<br />
Society and the Foundation) will issue certificates of<br />
education to students who have attended in any of the<br />
NDT methods identified in this document.<br />
The use of this procedure is mandatory and is to be<br />
adopted by the Foundation and all Chapters of the<br />
Society. Other organizations may adopt this procedure<br />
in order to conform with standardization of NDT<br />
training in Canada; such organizations must comply<br />
with ED-1 in its entirety for approval by the National<br />
Commi 11ee.<br />
ED-1 is meant to supplement the C.G.S.B. certification standards, not<br />
to replace them. The Society has simply gone a step further and provided<br />
recommended guidelines regarding the educational material,<br />
instructors' credentials, and examination format for the educational<br />
programs. Again, the objective is to ensure high quality, uniform<br />
education for NDT personnel.<br />
N.D.T. training in Canada has come a long way in a relatively short<br />
period of time. At this conference, we are celebrating the major<br />
strides that the Society has taken in just twenty years and I congratulate<br />
the Society for its emphasis on education.<br />
What strides will be taken in the next twenty years? The strides that<br />
are taken are going to have to be in step with a rapidly changing technology.<br />
Many of the registrants at this conference will listen to presented<br />
papers with NDT applications which they had not thought of, or<br />
perhaps, thought possible. But how many people, and how many of the<br />
disciplines 'hat inter-relate with NDT, are we reaching?<br />
In any technology, there is a frustrating (and potentially dangerous)<br />
gap between the engineer and the technician. Within the next five years<br />
I believe those of us in the NDT sector of Canadian industry are going<br />
to be the first to realize the impact of a fairly recent addition to<br />
some community college semester programs. In a course entitled "Materials<br />
Testing", students are taking Mathematics, Chemistry, Physics, Foundry<br />
Practice and Metallurgy, Electrical Systems and Instrumentation, Welding<br />
Computer Concepts, Statistics, Mechanics of Solids, Nondestructive and<br />
Destructive testing, and Language and Report Writing.
- 296 -<br />
The graduates of these programs, along with the growing numbers of<br />
Level III NDT technicians, will help to fill the gap. However,<br />
engineers must also be educated in NDT technology. I recently had<br />
the pleasure of hearing Dr. David W. Hoeppner, of the University<br />
of Toronto speak on "Failure Analysis and NDT Interface". Dr.<br />
Hoeppner, an educator at the university level, is appalled at the<br />
lack of a solid component of nondestructive testing techniques in<br />
the education of our engineers.<br />
What will be the outcome of an even greater emphasis on nondestructive<br />
testing education in Canada? The saving of human lives, an<br />
increased prevention of human injury, higher productivity and the<br />
saving of billions of dollars.<br />
In the United States, the National Bureau of Standards and Battelle<br />
Columbus Laboratories conducted a study to assess the costs of<br />
material fracture to the United States for the year 1978. In their<br />
report'-, it states<br />
"The costs were large. In 1982 dollars the total cost was<br />
estimated to be $119 billion per year, about 4 percent of<br />
the gross national product. An estimated $35 billion per<br />
year could be saved through the use of currently available<br />
technology. Costs could be further reduced by as much as<br />
$28 billion per year through fracture-related research...<br />
The study concluded that substantial material, transportation,<br />
and capital investment costs could be saved if technology<br />
transfer, combined with research and development,<br />
succeeded in reducing the factors of uncertainty related to<br />
structural design. Equally safe or safer structures could<br />
be produced with substantial cost savings in material,<br />
transportation, and capital investment. This could be done<br />
by reducing the uncertainty currently related to structural<br />
design through better predictions of structural performance<br />
from materials properties, better process and quality control<br />
and in-service flaw monitoring, and less materials<br />
variability."<br />
The same solutions and cost savings, albeit with a lower dollar<br />
figure, apply to Canadian industry and, what dollar figure do we<br />
apply to the value of human life.<br />
The Economic Effects of Fracture in the United States<br />
Special Publication 647-1<br />
U.S. Department of Commerce - National Bureau of Standards<br />
R.P. Reed, J.H. Smith, B.W. Christ<br />
Issued March 1983
- 297 -<br />
FEAR DETECTI<strong>ON</strong> AND REMOVAL<br />
THE PSYCHOLOGICAL IMPLICATI<strong>ON</strong>S OF THE TECHNOLOGICAL APE<br />
Tanii HdlliiMe.il<br />
With the advent of the technological age many age old fears of<br />
mankind have resurfaced. What is Fear? How does fear impact on all of our<br />
lives today? How can we learn to detect and remove fear from our lives and'<br />
help others to do the same?<br />
Fear for most of us indicates a loss of control in a situation and it<br />
leads to feelings of helplessness and powerlessness in our life. With the<br />
ever increasing speed and frequency of change surrounding us in our world<br />
many of us are thrown off balance, out of control and into the ocean of<br />
fear.<br />
Fear of failure, fear of rejection, fear of change and fear of the<br />
unknown are particularly relevant for people in industries which are becoming<br />
increasingly high tech such as the petrochemical, nuclear, plastic,<br />
steelraaking, and refineries. Older workers worry "Have I got what it takes<br />
to keep up?" Younger workers question "Will I be the first to go if<br />
employees are laid off and computers take over?" Managers wonder "How can I<br />
motivate my employees and allay their fears when I'm not sure what is going<br />
to happen myself?"<br />
These fears can lead us into depression, burnout, passivity,<br />
aggression, and work aholism. Through recognizing our fears and our reaction<br />
to them we gain insight into our roles and behaviours, our relationships and<br />
our communication with others. We must learn to free ourselves from our<br />
fears if we are to motivate ourselves and others to adapt to the changing<br />
world around us so that we can be active participators and not victims of<br />
these changes.
- 298 -<br />
CERTIFICATI<strong>ON</strong> OF N<strong>ON</strong>DESTRUCTIVE TESTING PERS<strong>ON</strong>NEL IN CANADA<br />
AN UPDATE<br />
I'. Care n<br />
CAWUET, Ottawa, Ontario<br />
ABSTRACT<br />
This paper surveys both the national and international scenes regarding the certification<br />
of nondestructive testing (NDT) personnel. In Canada, the increasing<br />
demand for NDT personnel and the implementation of the M-Standards have led to the<br />
need for automating the administration of the program. Concurrently, the increased<br />
interest towards harmomizing national certification schemes has steered activities<br />
in both the ISO and the ICNDT forums.<br />
A) THE NATI<strong>ON</strong>AL SCENE<br />
As is well-known, certification of nondestructive testing (ndt) personnel in<br />
Canada is a fully bilingual operation which is carried out through a central agency<br />
which is the federal Department of Energy, Mines and Resources. The certification<br />
standards are issued by the Canadian Government Standards Board (CGSB)<br />
and they contain all the requirements and procedure details underlying the certification<br />
process. Some 35 to 40 persons representing a broad range of interests<br />
are members of the national CGSB Committee for the Certification of NDT Personnel<br />
which meets every six months to review overall activities and to resolve any major<br />
issue arising from the certification programme. A Steering Committee supports the<br />
main committee.<br />
Certification initially started with Industrial Radiography in 1960, followed by<br />
ultrasonics in 1970, magnetic particle and liquid penetrant in 1971. Certification<br />
in eddy-currents did not start until 1982. The graph of Fig 1 illustrates<br />
the growth of the certification programme since 1970 by showing the number of<br />
certificates issued annually for the first four methods. Obviously, level I<br />
radiographers had the strongest growth during the past thirteen ( L3) years but<br />
also dropped considerably in 1983.<br />
The total number of certificates issued as of January 1, 1984, in each method,<br />
since their respective implementation date, is as follows:<br />
LEVELS RT (I960) UT (1970) MT (1971) FT (1971) ET (1982)<br />
I & II 3708 1878 1385 1490 175<br />
for a grand total of about 8636 excluding Level III. Well over 60% of the certified<br />
personnel holds certification in at least two (2) methods or more.
- 299 -<br />
A major event took place in 1979 when the new CGSB M- standards started being<br />
implemented. The advent of a new approach to ndt personnel certification followed<br />
a long period of monitoring the world trend towards a three—level system, particularly<br />
under the infLuence of SNT-TC-1A which dates from 1965. The resulting Mstandards<br />
retained a general approach and a central examination system but were<br />
extended to include a third level of competence. Provisions for employer's specific<br />
certification were added to the standards. In the process, training courses<br />
became compulsory and industrial requirements became less stringent.<br />
A major difference between the two A- and M- standards involved the amount of<br />
industrial experience needed for eligibility to examination and the new requirement<br />
of formal training courses. The A- standards required considerably more<br />
experience but training was not mandatory. As an example, to appreciate the<br />
trade-off between training and experience for purpose of eligibility, one calculates<br />
an equivalency or a equivalent trade-off of one week of training for one year<br />
of industrial experience for the particular case of radiography Level II.<br />
Impact of m-standards on administration<br />
This heading describes the status of a few problems associated with the introduction<br />
of the M-Standards.<br />
(a) The change from A- to M- Standards meant that over 30 papers totalling over<br />
60 hours of examination time would have to be prepared instead of the wellestablished<br />
11 papers lasting 19 hours. Accordingly, it was necessary to replace<br />
descriptive by multichoice questions. Considerable time savings can be achieved<br />
mainly from marking papers.<br />
(b) In order to mimimize personnel needs preparing written examinations, a bank<br />
of coded examination questions stored on diskettes is approaching completion and<br />
will contain both French and English versions. Print-outs and answer sheets allow<br />
to save much time in respectively preparing and marking examination papers.<br />
(c) The mandatory annual renewal of certification has become over the years a<br />
cumbersome operation; for instance, more than 6000 pocket certificates were issued<br />
to some 2800 persons in 1984. Software is being developed to automate those applications<br />
for renewal. Progress was impeded in 1984 when the Division undertook<br />
a global analysis of in-house computer requirements for both research and administrative<br />
operations. However it is hoped to achieve progress and run dry tests in<br />
the coming months once the computer mainframe become available.<br />
(d) The problem of space has become critical as it must compete with R&D laboratory<br />
needs. It was necessary to radically change filing customs from previously<br />
holding all examination material to nearly discarding all such material, except of<br />
course for resuLts. Even at that, space needs are increasing but resources are<br />
not presently available to put in a computer all the desirable information material,<br />
statistics, background information etc. However, the basic information which<br />
is presently recorded in the Cardex systems is being transferred onto diskettes:<br />
name, address, age, examination results, ndt method, date of certification.
- 300 -<br />
(e) A differing but persisting problem since the M- Standard on Radiography was<br />
introduced in 1979 concerns the Aircraft Structures category. The General category<br />
includes Welding, Casting and Forgings. The Aircraft Structures is a<br />
category which now pre-requires the General certification whereas the absolete Astandards<br />
did not have such a requirement. That extra requirement has always encountered<br />
strong opposition from both the civilian and military users who are only<br />
interested in maintenance work. That issue has re-surfaced in recent times and<br />
will be given further study during the coming months by the national CGSB<br />
Committee.<br />
( f ) The certification network now includes nine (9) practical test centres but<br />
it would need to be enlarged in Eastern Canada but progress is slow, one reason<br />
being the difficulty to obtain defective metal parts from users and producers and<br />
the considerable time needed to analyse them and prepare documentation.<br />
A new certifying agency (?)<br />
Following an offer made by the Canadian Welding Bureau (CSA) to assume the responsibility<br />
of Certifying Agency, the possible withdrawal of DEMR is currently being<br />
given consideration within the interested forum. Considerable resources are being<br />
devoted to both building and operating the new certification programme and those<br />
resources are indeed conflicting with the basic mission of PMRL/CANMET which is<br />
R&D. A feasibility study will be conducted for submission Lo the next meeting of<br />
the CGSB Committee in April 1985.<br />
B) THE INTERNATI<strong>ON</strong>AL SCENE<br />
This section briefly surveys the international developments within the optics of<br />
possible harmonization of national standards or practices. There is no doubt that<br />
the SNT-TC-1A practice for certification of ndt personnel which was introduced in<br />
1966 has exerted an important influence on the development and contents of such<br />
national documents. Their growing number reflects the necessity for measuring the<br />
qualification of ndt personnel in order to cope with the increasing demand and<br />
complexity of ndt requirements during manufacturing or service.<br />
Efforts at harmoning certification have been carried out for several years by Lhe<br />
ICNDT whereas ISO has become active only during the past year although its<br />
official commitment dates back to 1975. The secretariat for ISO-TC 135/SC7 on<br />
certification is now held by Canada which gives DEMR the responsibility for<br />
drafting an international document.<br />
(a) Perhaps the most striking feature which differentiates the various national<br />
schemes is the "general" and "sector" components of certification as being conducted<br />
by an independent central body, leaving the employer responsible for any<br />
additional features. From a survey of several national documents, it would appear<br />
that SNT-TC-1A is the only procedure giving the employers the sole responsibility<br />
for the whole certification.<br />
(b) The number of levels of certification may vary but are usually conceived to<br />
roughly correspond. Thus, a country with two levels will correspond to levels I<br />
and II. The range of qualifications ascribed to each level shows extensive
- 301 -<br />
sirailiarities but there may exist differences. For instance, a few standards<br />
allow Level I personnel to interpret/evaluate some results, whereas others do not<br />
allow such responsibility.<br />
(c) The categories or sectors of certification are often identical with<br />
castings, forgings, welds and aircraft structures/aerospace being commonly found<br />
but others may also be found. Specific areas may also be covered such as wrought<br />
steel plates and bars, types of welds (plate, pipe, nozzle), etc.<br />
(d) Examination schemes show some common features. Written and practical tests<br />
are applied by all schemes with oral tests being encountered in but a few schemes,<br />
either as a requirement or as a complement. Most schemes include a general<br />
written test, a sector written and practical tests.<br />
(e) Training requirements are common to most schemes but not all. Minimum<br />
hours may be specified and even related to a given degree of schooling. In order<br />
to avoid the implications of assessing and comparing national education systems,<br />
the ICNDT has endeavoured to determine a syllabus without a corresponding number<br />
of hours and schooling level. Industrial experience is usually mandatory.<br />
Several other headings are not examined herein but must also be considered, such<br />
as validity period, renewal conditions, health requirements, certificate ownership,<br />
etc.<br />
The above description points to some of the difficulties that might be expected<br />
towards harmonizing present and future national schemes for the certification of<br />
nondestructive testing personnel but there also exist definite commond grounds.<br />
The next ISO TC135/SC7 meeting will be held in February 1985 and the Board of<br />
Directors of the American Society for Nondestructive Testing recently expressed<br />
their objective to have an international standard finalized by 1988. Time will<br />
tell whether an international document or standard will be acceptable to the concerned<br />
countries and if so, what the time frame will be to achieve that objective.
250<br />
200<br />
150<br />
100<br />
50<br />
Magnetic Particle<br />
Level II O<br />
Liquid Penetrant<br />
Level II<br />
1970 72 74 76 78 80 82 1970 72 7A 76 78 80 82 1970 72 74 76 78 80 82<br />
FIGURE 1: Number of Certificates Issued Annually Between 1970 and 1983 Inclusive.<br />
I<br />
o<br />
I
DAY 3<br />
KEYNOTE ADDRESS: Looking into the Future 303<br />
- R.S. Sharpe<br />
NDT of Structural Ceramics by High Frequency Ultrasonics 307<br />
- A. Fahr<br />
Hot Pressed Piezo-Electric Ceramic Elements for Ultrasonic Transducers 316<br />
- N.D. Patel, J. van den Andel, P.S. Nicholson<br />
Ultrasonic Analysis of Voids in Glass: Theory and Practice 331<br />
- A.J. Stockman, J. van den Andel, P.S. Nicholson<br />
Computer Simulation of Ultrasonic Testing 345<br />
- D.B. Duncan<br />
A Novel Approach to Eddy Current Imaging of Defects 356<br />
- D. Leemans, M. Macecek<br />
Developments in X-Ray Stress Measurement, The CANMET Portable Stress 372<br />
Diffractometer<br />
- B.A. Holt<br />
Applications of Neutron Diffraction to Engineering Problems 387<br />
- T.M. Holden, G. Dolling, S.R. MacEwan, J. Winegar, B.M. Powell, R.A. Holt<br />
NDE in Polymers, an Example: Ultrasonics for the Determination of Density in 398<br />
Polyethylene<br />
- L. Piche, A. Hamel<br />
An Industrial Application of Computer Assisted Tomography: Detection, Location 408<br />
and Sizing of Shrink Cavities in Valve Castings<br />
- P.D. Tonner, G. Tosello<br />
Computer Operated Composite Panel Testing (abstract only) 424<br />
- S. DeWalle<br />
Some Unconventional Techniques for the Inspection of Layered Materials 426<br />
- P. Cielo<br />
Materials Effects on Acoustic Emission During Deformation and Fracture 445<br />
- M.N. Bassim<br />
List of Delegates 454
LOOKING INTO THE FUTURE<br />
R.S. Sh.an.pe.<br />
Mat-tonal NVT Ce.ntn.0.<br />
AERE Hawo.ll, England<br />
ABSTRACT<br />
- 303 -<br />
The recent rapid growth in scientific maturity of NDT is resulting in greater<br />
confidence in finding defects, higher inspection reliability, better<br />
understanding of the principles and limitations of the techniques<br />
conventionally used and a more quantitative approach to the assessment of the<br />
nature and sizes of defects through improved signal and data processing<br />
routines.<br />
In addition, the boundaries of nondestructive testing are reaching out to<br />
encompass new objectives and also provide a wider horizon than just<br />
inspection after manufacture, for which the majority of the conventionally<br />
applied techniques and application philosophies were primarily designed.<br />
The paper will show how the National NDT Centre at Harwell is operated to<br />
provide a focus of activity in the UK for spearheading this progress towards<br />
greater technological maturity. Examples taken from the current programmes<br />
of work will be used to illustrate moves towards a quantitative science,<br />
moves towards greater sensitivity and higher reliability and accuracy, moves<br />
towards linking the technology more overtly to a better understanding of<br />
fracture processes and moves towards better process control which,<br />
paradoxically, should make NDT, as an inspection process at 'the end of the<br />
line', edge towards redundancy!
- 304 -<br />
Nondestructive testing Is a subject that one might say was born of necessity,<br />
nurtured and kept alive by dedicated and experienced practitioners and has<br />
developed empirically, often through panic and expediency, to satisfy<br />
specific needs as they arise in industry.<br />
In recent years the subject has developed in a much more structured way, by<br />
encouraging formal training and operator certification, by the setting up of<br />
societies and conferences both nationally and internationally and, perhaps<br />
most significantly, by attracting the interest and participation of<br />
scientific research groups in many countries.<br />
One great advantage of this input of 'fresh blood' into the subject has been<br />
that NDT has developed much firmer foundations; the conventional techniques<br />
are, themselves, much more clearly understood and the basis for forward<br />
development is rapidly reverting to one of 'fact' and 'scientific reasoning'<br />
rather than 'hunch' and 'lets have a go'. The disadvantage, which should<br />
certainly not be underestimated, is that we now have a very dichotomous<br />
situation with, on the one hand, those practised and experienced in<br />
conventional practices and in the needs of industry, who have not amassed<br />
detailed knowledge or understanding of recent scientific developments or of<br />
research practices associated with signal and data processing; and, on the<br />
other hand, the 'scientific recruits' to the subject who have virtually no<br />
feel for the applications side of practical field NDT, scant allegiance to<br />
the technological society framework and little concern for the empirical<br />
'footslogging' of those who built the 'string and sealing wax' launch pad<br />
which they, the scientists, have used to get themselves into orbit.<br />
So if we look first into the near future one of the major needs is to bridge<br />
this "gulf before it becomes an unbrldgable chasm. One solution is to build<br />
up combined teams in which practical experience and scientific erudition can<br />
be brought together, hopefully to the mutual benefit of both disciplines. We<br />
have tried to tackle it this way in the UK with the national NDT Centre,<br />
which was set up at Harwell as long ago as 1967 with this as one of its main<br />
objectives. Both NDT applications work and research programmes are combined<br />
in the Centre's work programme and there is the added stimulus that need (as<br />
measured fairly quantitatively by contract income) can be used as the arbiter<br />
to decide the balance and range of activity.<br />
Our experience is that the type of organisation we have set up in the UK can<br />
be made to cater for both the short-term 'fire—bridgade ' type of NDT problem<br />
that tends to arise unexpectedly but, seemingly, with amazing regularity in<br />
manufacturing industry; and for the longer term R&D needs of larger<br />
organisations (such as the power generation, nuclear construction, oil and<br />
gas distribution and aerospace industries) who are perhaps more likely to<br />
anticipate and plan for the NDT needs associated with new manufacturing<br />
designs or new constructional materials and hence support underlying<br />
research.
- 305 -<br />
Another 'near future 1 need is undoubtedly going to be one of technology<br />
transfer. The rapid increase in research is resulting in a rapid fall-out of<br />
innovative Ideas and laboratory demonstration systems. The scientist often<br />
tends to lose interest when the 'physics has been worked out 1 and is then<br />
keen to transfer his innovative skills to the solution of the next<br />
challenging problem. This translation from a laboratory demonstration<br />
'happening' to an engineered, marketable and user-friendly package that can<br />
be taken over (and championed !) by the NDT practitioner is indeed a very<br />
significant problem- Too often laboratory solutions relate to a very<br />
specific problem, and the further development costs to produce an engineered<br />
system are then difficult to justify in terms of potential marketable units.<br />
This situation can sometimes lead to 'prototype systems' perpetuating the<br />
laboratory concepts, and hence requiring a 'PhD driver', getting out into<br />
practical shop-floor application and quickly leading to disillusionment and<br />
engendering loss of credibility - for the technique, the research laboratory<br />
that produced it and ultimately for NDT itself.<br />
Another 'near-future' need relates to the importance of getting the newer NDT<br />
techniques and systems through the barriers of standards, codes of practice,<br />
training and certification syllabi that collectively tend to set the norm of<br />
currently accepted NDT practices. Changes in this area are generally slow<br />
and effected by those often unfamiliar with what is happening at the 'sharp<br />
end' of research.<br />
So much for the near future, what of the longer term future for NDT and of<br />
the 'fresh fields' into which NDT practices might be expected to develop.<br />
Firstly, I detect a significant 'fresh fields' outlook away from simple<br />
defect detection of manufactured products back into the monitoring of the<br />
materials processing techniques and into the closer control of the properties<br />
of the materials themselves from which products are manufactured.<br />
The overall objective in this new approach is to provide better monitoring of<br />
the materials and the materials processing treatments so that the need for<br />
conventional NDT inspection 'at the end of the line' hopefully may become<br />
both less necessary and less critical.<br />
Then there is definitely a 'fresh fields' approach in studying and monitoring<br />
the early stages of crack formation. The objectives here are to develop a<br />
better understanding of the mechanisms of the process of crack development<br />
and to monitor the factors that contribute to failure occurring. Work on<br />
monitoring crack—tip stress and fracture toughness are very much in their<br />
infancy but could soon become a major area of study.<br />
Coming back to the more conventional defect detection role of NDT, there is a<br />
'fresh fields' approach now building up in which maximum use will be made of<br />
all test data so as to obtain more informative detail about the defect being
- 306 -<br />
interrogated, from which an interpretation of 'significance' then becomes<br />
easier to make, in support of fracture mechanics predictions.<br />
A whole range of advanced data assimilation and signal processing systems<br />
have recently been developed. We have recently developed and evaluated one<br />
such system which rapidly digitises and stores every A-scan and allows recall<br />
at will in a variety of presentational modes.<br />
We are also helping to contribute to a new generation of non-amplitude<br />
dependent ultrasonic sizing techniques with what we call the time-of-flight<br />
diffraction (TOFD) technique. We and others are also working on the more<br />
difficult 'classification' problem of identifying and characterising defects<br />
from the fine-scale information contained in the defect signal.<br />
Another fresh-fields approach in NDT that I see developing is to move away<br />
from 'eye-ball' defects in order to detect significant imperfections that are<br />
of much smaller dimension. We are contributing to this in the further<br />
development and evaluation of high definition radiography and fluoroscopy and<br />
in studying the problems associated with inspecting engineering ceramics,<br />
where the critical defects for initiating brittle failure are measured in<br />
microns rather that millimetres.<br />
Yet another trend, that should appeal to many, is aimed at hastening the<br />
obsolescence of NDT on the product line by modelling the effects of variables<br />
in the manufacturing process so as to try to reduce the incidence of<br />
defective products, by providing greater control of the process itself. As<br />
an example of this approach we are at the present time theoretically<br />
modelling the paint spray process in order to understand the effects of<br />
process variables on the final product, so that it should hopefully be<br />
possible to produce a more consistent 'quality', in this case in a robotised<br />
paint spraying process.<br />
Perhaps a final fresh fields approach as far as NDT is concerned should be<br />
that of being less introverted; by spending more time looking over one's<br />
shoulder to see which way other technologies are developing and making<br />
maximum use of such parallel experience wherever it is relevant to the<br />
furtherance of NDT practices.<br />
Bearing in mind that NDT only a few years ago was a 'bottom of the heap'<br />
technology kept alive only by a dedicated band of enthusiasts, it is<br />
particularly encouraging that it has recently attracted so much academic and<br />
management interest and as a consequence is opening up its frontiers in so<br />
many new and challenging directions. Looking into the future of NDT is, in<br />
fact, quite a stimulating and heartening experience at the present time and<br />
the possiblities certainly open up a whole new horizon of opportunity whilst,<br />
at the same time, exposing many problems and pitfalls associated with communication<br />
(up, down and sideways) that will need to be tackled and overcome.
- 307 -<br />
NDE OF STRUCTURAL CERAMICS BY HIGH FREQUENCY ULTRAS<strong>ON</strong>ICS<br />
A. fakn.<br />
National Re.izan.ck. Council o I Canada<br />
ABSTRACT<br />
A high frequency ultrasonic system with computer controlled data acquisition and<br />
analysis has been described. The system has been used to detect surface flaws in<br />
the size range of 20-300 um in hot pressed silicon nitride. The detection<br />
technique is based on Rayleigh waves which are generated as a result of the mode<br />
conversion of compressional waves incident from water onto the surface of the<br />
test material at near the Rayleigh angle. The scattering of the Rayleigh waves<br />
by surface cracks has been detected, analysed in the time and frequency domains,<br />
and correlated with the size of the cracks.<br />
This work was supported by the Defence Research Establishment, Pacific,<br />
Victoria, B.C. and was carried out at the Ontario Research Foundation,<br />
Mississauga, Ontario.
1. INTRODUCTI<strong>ON</strong><br />
- 308 -<br />
Ceramics such as silicon nitride, silicon carbide, and zirconia have excellent<br />
corrosion resistance, low thermal conductivity, high wear resistance, and good<br />
strength at elevated temperatures. These properties have generated considerable<br />
interest in the use of these ceramics in energy conversion systems, e.g., diesel<br />
engines, turbines, and heat exchangers.<br />
The hard and brittle nature of ceramics, however, inhibits the release of<br />
stresses at flaws under load by plastic deformation. This implies that a very<br />
small flaw can cause a catastrophic failure in ceramics. Therefore, flaws in the<br />
size range of 20-100 vim are considered as "critical" in ceramics for such<br />
applications as turbine blades.<br />
The reliable use of ceramics as structural parts requires nondestructive<br />
evaluation techniques which are capable of detecting and characterizing such<br />
small defects. Microfocus x-radiography and high frequency ultrasonics have been<br />
found^ ' to be the most suitable NDE methods for ceramics.<br />
This paper describes a high frequency ultrasonic system with computer controlled<br />
data acquisition and analysis capability and its application in the detection and<br />
characterization of microscopic surface cracks in silicon nitride.<br />
2. EXPERIMENTAL STUDIES<br />
2.1 Material and standard cracks<br />
Hot pressed silicon nitride is one of the most promising materials for high<br />
temperature structural applications. The majority of failures in this material<br />
originate from surface cracks introduced during grinding and machining operations.<br />
These produce arrays of semi-elliptical cracks with random inclination to the<br />
surface, but a preferred alignment parallel to the direction of motion of the<br />
abrading particles ,. A Knoop indentation technique was used to simulate<br />
surface machining cracks. In silicon nitride the Knoop indentation produces<br />
sharp semi-circular cracks with a mouth opening of the indentation less than one<br />
fifth of its depth, a, and its surface length about 2a.<br />
Knoop indentations in the depth range of 10 to 300 ym were introduced on a<br />
polished surface of a hot pressed silicon nitride (NCI32) plate. These cracks<br />
were carefully measured under an optical microscope prior to ultrasonic<br />
examination.<br />
2.2 High frequency ultrasonic system<br />
Since the critical flaw size in ceramics is very small and the high stiffness to<br />
density ratio generally gives comparatively high values of acoustic velocity and<br />
hence long wavelength at conventional ultrasonic frequencies, it is necessary to<br />
use high frequency ultrasonic waves.
- 309 -<br />
The system described in this paper (see Fig. 1) can be operated at frequencies up<br />
to 100 MHz. The puiser of the system consists of a 1 Hz to 100 MHz pulse<br />
generator (Philips PM5771) and a broadband (0.15 to 300 MHz) power amplifier<br />
(EIN403LA). The pulse generator is capable of providing pulses as narrow as 5 ns<br />
which are required for 100 MHz transducers. The high frequency transducers*<br />
employ a quartz buffer rod with a piezoelectric lithium niobate crystal mounted<br />
on one end and a lens machined on the opposite end. Focused transducers (12 mm<br />
focal length in water) with a narrow beam diameter (500 ym at focal point) were<br />
used.<br />
The detecting system uses a broadband (0.1-1300 MHz) amplifier with 48 dB gain<br />
(HP8447F) and a digital scope (HP1980). The latter has a bandwidth of 5 Hz to<br />
100 MHz and consists of two 100 MHz analogue measurement channels and an analogue<br />
to digital converter. When used together with a digital waveform storage unit<br />
(HP19860), repetitive ultrasonic signals of up to 100 MHz can be averaged (up to<br />
64 times), digitized (up to 501 points per waveform in main or delay modes), and<br />
stored in the memory. The stored waveforms are transferred to a microcomputer<br />
(HP9845) for analysis in the time and/or frequency domains.<br />
2.3 Detection Technique<br />
Surface or Rayleigh waves were used to detect surface cracks in silicon nitride.<br />
These waves are generated by the mode conversion of compressional waves incident<br />
at the Rayleigh angle, 9R, onto the surface of the test material. The<br />
conventional method of generating surface waves is to use a wedge transducer with<br />
a wedge angle of 9 g in contact with the test specimen. Preliminary work*- ' '<br />
however, indicated that the contact method is not suitable for testing ceramics<br />
at high frequencies due to coupling problems. Consequently, the technique of<br />
non-contact mode conversion in an immersion tank was used in the present work.<br />
Figure 2 shows the basic configuration for the generation and detection of the<br />
Rayleigh waves. In this configuration the Rayleigh angle is given by:<br />
6R = arc sin -y— (1 )<br />
R<br />
where V is the acoustic velocity in water and VR is the Rayleigh wave velocity in<br />
the test material. For silicon nitride 9R=15.7 . The Rayleigh waves reflected<br />
by sharp surface discontinuities radiate energy back into water at an angle of<br />
9Rand this energy is received by the transducer. These are referred to as "leaky<br />
Rayleigh waves". Thus, as the transducer is scanned parallel to the surface of<br />
the test material, strong signals from surface flaws are detected. Figure 3<br />
shows a typical signal from a 100 pm Knoop indentation detected by leaky Rayleigh<br />
waves at a frequency of 50 MHz.<br />
The detectability of surface flaws by leaky Raleigh waves is dependent on the<br />
transducer frequency. This is demonstrated in the experimental results of Fig. 4<br />
which show the smallest Knoop indentation detectable at different frequencies.<br />
This figure indicates that for the experimental conditions of the present work, a<br />
* Precision Acoustic Devices, Inc.
- 310 -<br />
frequency of about 50 MHz is optimum for the detection of surface cracks as small<br />
as 20 um in silicon nitride. This frequency is equivalent to an acoustic<br />
wavelength of 112 um in silicon nitride. Beyond 50 MHz, the detectability tends<br />
to level off due to the increase of attenuation losses. Thus, the crack<br />
measurements in this investigation were carried out at 50 MHz.<br />
The leaky Rayleigh waves propogating along the water-solid interface, lose their<br />
energy rapidly and particularily at high frequencies. Thus, only those flaws<br />
which are located near the point of incidence of the impinging beam are detected.<br />
As a result, the resolution of this method is very good since reflections from<br />
neighbouring flaws or specimen edges do not overlap with the signal from the flaw<br />
of interest. The resolution is particularily good when a narrow Rayleigh beam is<br />
employed by focusing the incident beam on the surface of the test piece using a<br />
focused transducer.<br />
Overall, the scanning simplicity, the high resolution and high sensitivity of<br />
leaky Rayleigh waves make this technique a practical and reliable method of<br />
detecting small surface flaws in ceramic materials.<br />
2.4 Measurement techniques<br />
The normal procedure for flaw size measurement by any ultrasonic technique<br />
utilizes the acoustic reflection coefficient, S, of the flaw for size estimation.<br />
S is defined as the ratio of the peak amplitudes of the flaw signal to input<br />
signal at the transducer terminal. In the case of cracks, however, the acoustic<br />
reflection is very much dependent on orientation of the crack with respect to the<br />
incident beam. This is demonstrated in the experimental results of Fig. 5<br />
obtained for a 100 ym Knoop indentation. Analytical solutions are given by<br />
Auld^ ' for calculation of the reflection coefficient at oblique angles. Both<br />
the analytical and experimental results indicate that as the direction of the<br />
crack relative to the Rayleigh beam deviates from normal, a smaller portion of<br />
the crack face reflections are received by the transducer. Thus, for complete<br />
characterization of cracks from the reflection coefficient, it is necessary to<br />
scan the test material in more than one direction.<br />
Another important factor affecting the acoustic signal amplitude from a flaw is<br />
the wavelength compared to the flaw size. Since at 50 MHz the Rayleigh<br />
wavelength in silicon nitride is about 112 pm and this is in the flaw size range<br />
of our interest (i.e. 10-200 um), therefore both the long wavelength (X^>a) and<br />
the short wavelength (A^
and<br />
- 311 -<br />
S = a 2j for XR>a (2)<br />
S = a'.a for XR
References<br />
- 312 -<br />
1. W.N. Reynolds, and R.L. Smith, British Journal of NDT, 145, 1982.<br />
2. J.J. Mecholsky, S.W. Freiman, and R.W. Rice, J. Amer. Ceram. Soc, 6^, 116,<br />
1977.<br />
3. S. Johar, A. Fahr, and M.K. Murthy, Proc. Conf. Advanced NDE Tech. sponsored<br />
by IMRI, NRC, 1982, Montreal.<br />
4. A. Fahr, S. Johar, M.K. Murthy, and W.R. Sturroek, Proc. Review Prog.<br />
Quantitation NDE, Vol. 3A, Plenum Press, 1984.<br />
5. B.A. Auld, Wavemotion, J_, 3, 1979.<br />
6. G.S. Kino, J. Appl. Physics, h% (6), 1390, 1978.<br />
7. B.T. Khuri-Yakub, G.S. Kino, and A.G. Evans, J. Amer. Ceram. Soc, 63,<br />
(1-2), 65, 1980.<br />
8. S. Ayter, and B.A. Auld, Proc. DARPA/AFWAL Rev. Prog. Quant. NDE, AFWAL-TR-<br />
80-4078, 344, 1974.<br />
9. V. Domarkas, B.T. Khuri-Yakub, and G.S. Kino, Appl. Phys. Lett., 3_3, (7),<br />
557, 1978.<br />
10. N. Saffari, and L.J. Bond, IEEE Ultrasonics Symposium, 1983.
TRANSCEIVER<br />
digital<br />
signal<br />
display<br />
1<br />
A to D<br />
converter<br />
waveform<br />
storage<br />
- 313 -<br />
focused<br />
transducer<br />
Specimen<br />
A A<br />
Plotter<br />
7-<br />
3<br />
Mag netiC<br />
tap<br />
Pf_J<br />
Fig. 1. A schematic of the set-up.<br />
LEAKY RAYLEIGH<br />
WAVES<br />
Fig. 2. The basic configuration for<br />
the generation and detection of leaky<br />
Rayleigh waves.<br />
5,us<br />
„'User's<br />
"i Graphics<br />
terminât<br />
CRACK<br />
REFLECTI<strong>ON</strong><br />
EDGE<br />
REFLECTI<strong>ON</strong><br />
Fig. 3. An oscilloscope view of a<br />
signal from a Knoop indentation (100 pm)<br />
detected by 50 MHz leaky Rayleigh waves.<br />
The indentation was located 2 mm from<br />
the specimen edge.
PLANE OF<br />
CRACK<br />
100<br />
20 40 60 60<br />
SAW Frequency (MHz)<br />
1OO<br />
Fig. 4. The detectability of surface cracks in silicon<br />
nitride by leaky Rayleigh waves as the function of frequency.<br />
EXPERIMENTAL<br />
(KNOOPCRACK)<br />
40<br />
60<br />
80<br />
I00<br />
I40<br />
THEORETICAL •<br />
20<br />
160 ISO' ANGLE OF<br />
INCIDENCE<br />
CNORMAL<br />
/ ÄCOUSTI)<br />
REFLATI<strong>ON</strong><br />
.COEFFICIENT<br />
Fig. 5. The experimental data obtained for the acoustic reflection of a Knoop<br />
indentation at different orientations compared to the theoretical curve obtained
m<br />
"O -30<br />
z<br />
o<br />
Ul -40<br />
u.<br />
Ul<br />
ce<br />
o<br />
o<br />
Ü<br />
-50<br />
-60<br />
- 315 -<br />
CRACK DEPTH (Aim)<br />
50MHz Ä * 112 um<br />
100 200 300<br />
10 20 30<br />
Ko<br />
Fig. 6. The acoustic reflection of various size<br />
Knoop indentations obtained by 50 MHz leaky Rayleigh waves.<br />
O<br />
0.<br />
<<br />
Ul<br />
Ul<br />
50 100<br />
FREQUENCY<br />
CRACK SIZE<br />
(/im)<br />
60 ^<br />
150<br />
Fig. 7. Spectra of different size Knoop<br />
indentations deconvolved by a reference signal.
- 316 -<br />
HOT PRESSED PIEZOELECTRIC CERAMIC ELEMENTS FOR ULTRAS<strong>ON</strong>IC<br />
TRANSDUCERS<br />
U.V. Patdt, J. van den Ande.1 and P.S. Nicholson<br />
HamiCtcr., Ontanic, Canada<br />
ABSTRACT<br />
Piezoelectric ceramics (PZT4, PZT5, PZT7 and SPN) were investigated for<br />
development of high frequency (30-100 MHz) normal and focussed transducers. Hot<br />
pressed piezoelectric materials with a controlled grain size of ~2jum and<br />
approaching theoretical density can be lapped down to 30 jum thickness without<br />
structural damage whereas conventional sintered piezoelectric material with grain<br />
sizes ranging frcm 2-15jum and at least 3% porosity can only be lapped down to<br />
100>um. Hot pressed piezoelectric ceramics when compared with conventional<br />
piezoelectric ceramics, show higher values of coupling coefficient (10-20%),<br />
elastic compliance ^12% and moderate increases in mechanical quality factor<br />
(Q ) and dielectric constant (€ ). The ccmpressional sound velocity measured<br />
along the poling direction is very sensitive to the switching of dipoles other<br />
than 180 ones. The velocity increases as the polarization increases and this<br />
constitutes a useful method for quality control of the piezoelectric ceramic as<br />
it gives the correct thickness resonant frequency. By accurately measuring the<br />
change by a comparison method, the dipole behaviour can be better understood and<br />
the degree of polarization and depolarization can be established. Temperature<br />
dependent polarization and depolarization is also discussed in the light of<br />
dipole switching.<br />
Hot pressed SPN and PZT5 are sufficiently transparent to be considered for<br />
use in opto-ultrasonic applications such as in medicine.<br />
Because of the high value of the thickness-frequency constant and low O ,<br />
SPN seems a better choice as a high frequency transducer material. The signal<br />
spectra of a high frequency transducer constructed frcm these materials is<br />
presented.<br />
I INTRODUCTI<strong>ON</strong><br />
High frequency transducers in the range of 30 to 100 MHz are of great<br />
importance for detection of minute flaws (~-'5/*m) in high performance ceramics.<br />
The detection of these flaws require powerful high-frequency transducers which<br />
are now being developed. As the frequency increases, the absorption in the<br />
medium being investigated also increases. Therefore, to compensate this loss,<br />
the transducer piezoelectric element should have a high transmitting coefficient<br />
d,. and receiving constant g„. Also, the sound beam should be capable of
- 317 -<br />
deep penetration in same focussed fashion.<br />
Most ccmmercially available piezoelectric discs are sintered. These<br />
sintered materials have grain sizes between 2 and 10 /um with a few percent<br />
porosity and occasionally have large flaws. In order to achieve high frequencies<br />
in the range 30-100 MHz, these elements have to be lapped down to as thin as 20<br />
jm without any structural danage, however, sintered piezoelectric ceramics, when<br />
lapped below 100 /un thickness, receive structural damage, eg. grain pull-out and<br />
have insufficient strength. Therefore, piezoelectric ceramics used for high<br />
frequency transducers should have at least 7-12 grain layers devoid of porosity.<br />
Large flaws also contribute to the electrical failure of piezoelectric ceramics<br />
on polarisation.<br />
It is veil known that grain size and bulk density markedly influence the<br />
electrical properties of sintered piezoelectric ceramics. By choosing another<br />
fabrication method, i.e., hot pressing, one can control the grain growth and<br />
densification process. This method can yield 100% dense ceramics with<br />
controlled grain sizes suitable for the extreme thinning required. The ccmpressional<br />
wave sound velocity in the direction of polarisation for the hot pressed<br />
material is higher than their sintered counterparts. This gives an extra layer<br />
of thickness when compared with the sintered material for the same thickness<br />
resonant frequency. The piezoelectric properties of hot pressed elements are<br />
also superior to sintered elements.<br />
Another important factor governing these ceramics is polarisation. One has<br />
to optimise the poling condition for each composition when such ceramics are to<br />
be lapped down to «OO/fm after polarisation. If one exceeds these cptimim<br />
poling conditions (Ref.l) (e.g. temperature, poling field or poling time),<br />
micrccracks will develop causing element failure when excited by the voltages<br />
necessary to generate sufficiently powerful bursts of high frequency ultrasound.<br />
Whereas the switching of normal dipoles (180 dipoles) does not involve<br />
any structural changes, the alignment of "mechanical dipoles" (other than 180°)<br />
during polarisation requires such changes. Since sound waves are associated with<br />
interatomic forces, one can detect these mechanical dipoles by measuring sound<br />
velocities in the material. Such measurements facilitate an understanding of<br />
dipole behaviour during polarisation and depolarisation. It is the mechanical<br />
dipole alignment associated with the thickness resonant frequency of the<br />
piezoelectric ceramic which causes the change in sound velocity. Therefore, the<br />
thickness resonant frequency will depend upon the number of mechanical dipoles<br />
aligned during polarisation.<br />
II SAMPLE PREPARATI<strong>ON</strong><br />
PZT 4, PZT 5 and PZT 7 calcined powders and SPN were pressed at 20 MPa<br />
(3000 psi) into slugs ^15 nm long, isopressed at 340 MPa (50,000 psi) and fired<br />
in a closed alunina crucible at 1270 -1350 c in a controlled atmosphere for<br />
an hour. Heating and cooling rates of 450 c/h were maintained.<br />
Samples of the same powders were pre-pressed to obtain slugs and these were<br />
placed in a molybdenum mold and packed with zirconia/alunina (200 mesh) powder.<br />
This composite was pressed in a vacuum induction furnace for ~10-20 minutes at<br />
900-1200 C at pressures between 3 and 30 MPa (500-5000 psi) employing the same
- 318 -<br />
heating and cooling rates of 450 C/h. Bulk densities were measured in butanol<br />
and distilled water.<br />
Both hot presses and sintered ceramics were sliced into « 6 mm thick discs.<br />
Tfi^ hot-gressed slices were oxidized in a controlled atmosphere for 4-24 hours at<br />
350-1200 C, depending on composition. Both sides of these discs were lapped to<br />
a thickness of 400/urn with optimum flatness and parallelism. Fired silver electrodes<br />
and sputtered gold electrodes were deposited on them. The fired silver<br />
electrodes were polished to M 2/tm on both sides of the piezoelectric discs.<br />
III ELECTRICAL AND ACOUSTIC CHARACTERIZATI<strong>ON</strong><br />
All materials were poled in oil at 80-180°C. The poling temperature, time<br />
and voltage were varied to identify the optimum poling conditions for each<br />
cemposition. All poled specimens were aged for 24 hours, prior to measurement.<br />
The resonant (f ) and antiresonant frequencies (f ) were measured according<br />
to IEEE standards. Following measurement of the ccmpressional and shear<br />
wave velocities, the Poisson's ratio for each cemposition was obtained by calculation.<br />
The piezoelectric constants, the planar coupling factor, k ; the piezoelectric<br />
strain constant, d.,, and the piezoelectric 'voltage' constan? g-., ware thus<br />
obtained and are tabulated in Table I. Capacitance and loss factors were<br />
measured at 1 kHz using an impedance bridge.<br />
The ccmpressional sound wave velocities at 50 MHz ware measured in both<br />
polarized and unpolarized piezoelectric discs. In the polarized materials,<br />
velocity was measured in the polarized direction. The pulse-echo overlap nethod<br />
was employed and velocities were measured to one part in 10 . A low-viscosity<br />
oil was used as a couplant between the 50 MHz transducer and the piezoelectric<br />
discs. The thickness variation of the discs was monitored before and after<br />
polarization and also during the velocity measurements.<br />
The microstructural grain size was determined from photomicrographs (optical<br />
and electron optical) of the test specimens. These were mechanically'polished to<br />
a surface roughness of 0.3A
- 319 -<br />
the density by »w3 to 7%.<br />
These increases are due to the increases in density, the lack of texturing<br />
and the uniform grain size of the hot pressed materials. Such materials are<br />
clearly advantageous for use in high frequency transducers. The higher sound<br />
wave velocities allow extra thickness for the same frequency, e.g. for 100 MHz<br />
longitudinal sound wave velocity, the required thickness for hot pressed PZT 7 is<br />
•* 25 /«m whereas for the sintered material it is 23 yum. The hot pressed<br />
piezoelectric ceramics drew less current during poling than the sintered ones<br />
(Figure 1). For all materials the anount of current drawn increases with the<br />
applied voltage. Current fluctuation or considerable increases in<br />
voltage during poling, indicate the development of microcracks in the ceramic.<br />
This causes a decrease in the coupling coefficient (9) as indicated in Figure 2<br />
beyond 3,600 V/mm. If either time, temperature or field are increased, the<br />
specimens fail. The dipole alignment on application of an electric field, leads<br />
to the development of internal stresses which eventually leads to microcracking.<br />
In the absence of microcracking, the dipole alignment increases k ; with microcracking,<br />
k drops. p<br />
In assembling high frequency transducers, either indium bonding or a<br />
conducting epoxy is used to bond the piezoelectric discs to the backing (damper)<br />
medium. For both cases, a temperature of 80-150°C is required to construct the<br />
transducer. Therefore, the elements should be able to withstand these temperatures<br />
without losing their piezoelectric properties. Figure 3 illustrates the effects<br />
of overheating PZT 5 discs. The curves show the value of k as a function<br />
of increasing temperature up to 175 C on the same test specBnen. It is clear<br />
that k is reduced by «'Si when the piezoelectric element is heated to ~200 o c<br />
No sucn property change was observed when specimens were heated to 140°C;<br />
k peaks at «/HO C and this can be attributed to the transition frcm one<br />
ferroelectric phase to another.<br />
To construct a high frequency transducer it is necessary to know the correct<br />
thickness resonant frequency of the element and this is difficult to measure on<br />
thin discs. By measuring the sound velocity in the polarization direction,<br />
a high degree of accuracy can be realized. Fran such data, the thickness<br />
resonant frequency; (f = c/2t, c = velocity, f = frequency and t = thickness)<br />
can be obtained and dipole behaviour during poling, elucidated.<br />
During poling the 180 -demain reversal is complete and other discrete<br />
angle dipoles are incompletely switched (the percentage of such switching depends<br />
upon the phases involved and the impurities added to the ceramic). Stress free<br />
piezoelectric ceramics would be those whose domains switch only by 180°, i.e.,<br />
experience no structural changes. Therefore, the sound velocity measured along<br />
the polarization direction for such a material, should be unchanged when compared<br />
with the virgin ceramic. This was true for the initial polarization of PZT 4,<br />
PZT 5 and PZT 7 in a low field, i.e., where only the 180 dipoles are involved<br />
(Figures 4,5,6 and 7). The k values obtained frcm alignment of these normal<br />
dipoles for PZT 4 (rhcmbohedral/tetragonal phase), sintered and hot pressed are<br />
~23% and 29%, respectively; for PZT 5 (rhanbohedral), sintered and hot<br />
pressed are 27% and 38%; for PZT 7 (tetragonal phase), sintered and hot<br />
pressed are 18% and 23%, and for SPN, sintered and hot pressed are '"19% and<br />
«/26%, respectively.
- 320 -<br />
The alignment of the mechanical dipoles involves structural changes and<br />
this causes internal stress; this will change the sound velocity in the polarization<br />
direction. The velocity change is directly proportional to the extent of<br />
mechanical dipole alignment during polarization. For hot pressed, PZT 4 the<br />
change was ~9%, for PZT 5 «'10%, for PZT 7 "7% and for SPN ~8%. Their sintered<br />
counterparts showed a few percent less sound velocity. The initial temperaturedependent<br />
depolarization occurs via the mechanical dipoles as stresses are<br />
relieved. For the PZT compositions and the SPN, the mechanical dipoles alone<br />
switch back up to 250°C and 325°C respectively. As temperatures are increased<br />
beyond 250 C for PZT canpositions and 325°C for SPN, the normal<br />
dipoles also switch back. The time and frequency spectra of hot pressed elements<br />
and transducers constructed fran them are shown in Figures 8 and 9.<br />
SPN, with a high sound velocity, low dielectric constant and low O is<br />
potentially useful for the construction of high frequency transducers. Also the<br />
acoustic impedance for SPN (31.3 x 10 _ g/an sec) is low compared with that of<br />
the PZT canpositions ( ~39 x 10 g/an sec). This property facilitates<br />
matching with transducer backing materials.<br />
Hot pressed PZT 5 and SPN can be sufficiently transparent for consideration<br />
in opto-ultrasonic applications.<br />
V C<strong>ON</strong>CIDSI<strong>ON</strong>S<br />
As a result of the reported investigation, the following conclusions can be<br />
made;<br />
(1) hot pressed piezoelectric ceramics ( /v 100% dense) can be used to construct<br />
high frequency transducers in the range 50-100 MHz.<br />
(2) the piezoelectric properties of the hot pressed materials are considerably<br />
better than their sintered counterparts.<br />
(3) the high values of sound velocity along the polarisation direction (i.e.,<br />
the high frequency constant) for hot pressed material allows an extra layer<br />
of grain thickness for the same frequency as compared to similar sintered<br />
materials.<br />
(4) the grain size of the final ceramic can be controlled by hot pressing<br />
techniques.<br />
(5) the ccmpressional wave velocity measurements in the polarisation direction<br />
facilitate an understanding of the behaviour of dipole switching during<br />
polarisation and depolarisation.<br />
REFERENCES<br />
(1) R.W. Rice and P.C. Pohanka, J. An. Ceram. Soc., 56 (8) 420-23 (1973).<br />
ACKNOWLEDGEMENT<br />
SPN calcined powder was supplied by BM Hi Tech (Collingwood, Ontario).
density 10 3 kg/m 3<br />
dissipation factor %<br />
relative dielectric<br />
kp %<br />
k 31 %<br />
d31<br />
pC/N<br />
g31 10- 3 Vm/N<br />
Vp m/s<br />
Vu m/s<br />
om<br />
required thickness In<br />
Vim for 100 MHz<br />
TABLE 1: Physical Properties of Four Piezo-Electric Ceramics<br />
sintered<br />
7-05<br />
0-4<br />
1100<br />
54<br />
32<br />
.115<br />
-11-8<br />
4420<br />
4100<br />
4SO<br />
22<br />
PZT 4<br />
hot<br />
pressed<br />
7*09<br />
0-3<br />
1250<br />
02<br />
37<br />
.138<br />
-12-3<br />
4820<br />
4420<br />
490<br />
24<br />
sintered<br />
7-73<br />
1-8<br />
15SO<br />
«3<br />
38<br />
-170<br />
-12-4<br />
4510<br />
4180<br />
70<br />
23<br />
PZT 5<br />
hot<br />
pressed<br />
8-01<br />
1-8<br />
2100<br />
72<br />
43<br />
• 230<br />
-12-4<br />
4850<br />
4420<br />
80<br />
24<br />
sintered<br />
7-45.<br />
1-5<br />
400<br />
32<br />
19<br />
-30<br />
-8*5<br />
4820<br />
4400<br />
1025<br />
23<br />
PZT 7<br />
hot<br />
pressed<br />
7 »97<br />
0.6<br />
450<br />
38<br />
23<br />
-40<br />
-10-0<br />
4940<br />
4640<br />
1180<br />
25<br />
sintered<br />
4-20<br />
2.2<br />
260<br />
37<br />
22<br />
-31<br />
-12.5<br />
6520<br />
8170<br />
140<br />
28<br />
SPN<br />
hot<br />
pressed<br />
4>49<br />
1-e<br />
380<br />
45..<br />
27<br />
-46<br />
-13-1<br />
6980<br />
6430<br />
210<br />
35
o PZT 7<br />
o PZT 4<br />
• HP 7<br />
• HP4<br />
- 322 -<br />
1-5 3-0<br />
Poling field (kV/mm)<br />
development of<br />
microcracks,<br />
piezo-eiectric<br />
properties<br />
drop<br />
Fig. 1: Dissipation Current as a Function of Poling Field at 120°C<br />
for sintered and Hot-Pressed PZT 4 and PZT 7 Ceramics.
(0<br />
60<br />
g40<br />
«MM<br />
"5.<br />
3<br />
O ü<br />
Cö<br />
§20<br />
A-HP4<br />
A-PZT4<br />
- 323 -<br />
120 °C<br />
DEVELOPMENT<br />
OF<br />
MICROCRACKS<br />
O-PZT7<br />
• -HP7<br />
i<br />
00 1500 3000<br />
Poling field (V/mm)<br />
i<br />
4500<br />
Fig. 2: The Dependence of Planar Coupling Factor kp on Poling<br />
Field at 120°C for Sintered and Hot-Pressed PZT 4 and PZT 7<br />
Ceramics.
70<br />
-65<br />
b.<br />
o<br />
•4-»<br />
O CO<br />
= 60<br />
a<br />
3 O<br />
£55<br />
PZT5<br />
- 324 -<br />
I I<br />
50 150<br />
Temperature (°C)<br />
•DECREASE IN PIEZO-ELECTRIC<br />
PROPERTIES<br />
Fig. 3: Loss of Planar Coupling Factor by Repeated Overheating of<br />
a Piezo-Electric Element.<br />
250
60<br />
PZT7<br />
Vp "<br />
- 325 -<br />
o A —polarization<br />
• * —depolarization<br />
/Vu<br />
8 10<br />
Fig. 4: The Planar Coupling Coefficient as a Function of Change<br />
in the Compressional Wave Velocity, with Depolarization<br />
caused by Heating (PZT 7.).
o<br />
CO<br />
O)<br />
60<br />
- 326 -<br />
2 4 6<br />
V P » v u /Vu<br />
A o—polarization<br />
*• —depolarization<br />
Fig. 5: As Fig. 4, but for PZT 4.<br />
8 10
- 327 -
- 328 -<br />
2 4 6<br />
o —polarization<br />
• —depolarization<br />
vP - vu /Vu )<br />
Fig. 7: As Fig. 4, but for SPN.<br />
8 10
- 329 -<br />
Fig. 8: Typical Outputs of Three Hot-Pressed, Unmounted SPN Elements<br />
of Different Thickness and Determination of Their Time Period.
- 330 -<br />
60 MHz<br />
Fig. 9: Output of Two Transducers Made with Hot-Pressed PZT 4 Elements:<br />
1. Highly Damped<br />
2. Lightly Damped<br />
3. Frequency Spectrum of 2
- 331 -<br />
ULTRAS<strong>ON</strong>IC ANALYSIS OF VOIDS IN GLASS; THEORY AND PRACTICE<br />
Aitkux J. Stockman, Jan van den Ande.1 and Patrick S. Hlcholion<br />
McMaiie-t linivdtisitij<br />
Hamilton, Ontaiio, Canada<br />
ABSTRACT<br />
The characterization of voids in glass enables the study of reflected ultrasonic signals off . known but<br />
undisturbed defects which can also be characterized by optical means. The reflected ultrasonic<br />
signals can be compared with theoretical signals transformed to a frequency spectrum by a computer<br />
and a frequency spectrum can be transformed to a time signal (A-scanl.<br />
Flaw information can now be read from such signals, and such will eventually help to characterize<br />
flaws in opaque ceramics.<br />
I INTRODUCTI<strong>ON</strong><br />
The characterization of defects in high performance ceramics is important when the components are<br />
to be used in situations of high stress. Size, shape and composition of the target defects will determine<br />
whether or not the ceramic will fail. Interest in ultrasonic testing to determine these factors has led<br />
to many advances in the field of defect characterization by scattered signal analysis.' 1 '<br />
On the link between theory and practice, it was found by the authors that a beam entry of exactly 90°<br />
with the entrance surface of the component was important.' 2 ' The essence ofthat article was that the<br />
frequency components of the ultrasonic beam pass the interface undistorted if a perfect 90° angle-is<br />
maintained. A slight misalignment will cause enough distortion to make the frequency spectrum of<br />
the reflected beam lop-sided and random misalignments from sample to sample means that the<br />
analysis will no longer yield accurate results.<br />
The importance of proper alignment is brought out even more in a study of reflections of previouslyfound<br />
defects. 13 ' Each defect responds differently to a frequency component so their reflected signals<br />
contain information about their type and shape. The change in frequency component amplitude can<br />
also be calculated, measured and the resulting spectrum compared to the perfectly aligned spectrum,<br />
i.e. the only distortion is caused by the defect rather than by off-normal set-up conditions<br />
The technique was applied to ceramics at frequencies above 10 MHz in order to test for flaws<br />
< 100 urn. A precision immersion scanner with a resolution of 20 urn was built especially for locating<br />
these small defects. Beams were focussed to obtain higher power densities and even though these<br />
beams do not provide a true plane wave, the wave theory as reported by Ying and Truell' 4 ' was<br />
applied along with the "high frequency decomposition" part, reported by Tittman et al.' 5 '
- 332 -<br />
As the scattering theory requires spherical defects, this work examines the scattering of a focussed<br />
sound beam from spherical and near spherical bubbles in glass. The use of glass rather than opaque<br />
ceramics facilitates optical characterization, the only way a small defect can ever be diagnosed.<br />
II. THEORY<br />
The pressure profile for the sound beam from a focussed transducer operating at a single frequency is<br />
a complicated function of the density p and speed Î sound cL of the medium, the diameter D and<br />
radius of curvature A of the transducer, and the frequency f. At the focal distance Zf from the<br />
transducer the pressure is:<br />
where<br />
i2nflt - (ß + z.)]<br />
2c, '<br />
c, S fe<br />
L zf<br />
8nS<br />
S = »<br />
V 1 - A/Zf<br />
(Zf - h)" + (D/2)"<br />
sin {nf (p 1 - Zf )}<br />
1/2<br />
, h = A - A" - (D/2)-<br />
and So is the displacement of the front face of the transducer. N'ear the focus the pressure function<br />
can be approximated in the form of a plane wave as:<br />
i2nf(t - AZ/Cjj<br />
where AZ is the distance along the z-axis from the focal point. L'sing the theory of Ying and Truell<br />
(4), the relative pressure amplitude for the longitudinal wave backscattered to the transducer from a<br />
spherical void of radius, a, is:<br />
p (f,a) -2n-e<br />
f L<br />
(2m<br />
m = 0<br />
where the scattering coefficients Am are determined from the stress and strain relationships at the<br />
surface of the void. Now if the sound pressure at the site of the void were known it would be possible<br />
to determine the pressure of a backscattered monochromatic longitudinal wave.<br />
Reflections from the sample surfaces make single frequency studies impractical, fnstead, short<br />
pulses of ultrasound are generated by shock excitation of the transducer so that the defect can be<br />
located. A pulse of ultrasound is composed of many waves of different frequencies The distribution of<br />
wave amplitudes and phases as a function of frequency is called the response function of the<br />
transmitter-receiver system, R(f). Since the sound pressure is maximum at the focus the waves will<br />
be in phase so that only the amplitude component of the response function needs to be considered<br />
Experimentally the function is determined by routing to a spectrum analyzer the signal reflected<br />
from a flat surface in the focal plane of the transducer. For the calculations the system response<br />
1/2
- 333 -<br />
function is approximated by a Gaussian distribution then it is multiplied by the relative pressure<br />
amplitude function to produce the expected frequency spectrum of a signal reflected from a void.<br />
The expected frequency spectra so calculated have both amplitude and phase components making it<br />
possible to obtain the real time signals and the magnitude frequency spectra expected. However, only<br />
relative amplitudes of the time signals and magnitude spectra can be determined since the exact<br />
pressure at the focus is not known. Shape comparisons between measured and calculated signals and<br />
spectra are possible and are shown below.<br />
III. EXPERIMENTS<br />
Figure 1 shows the experimental set-up used. The commercial focussed immersion transducer has a<br />
piezo-electric ceramic element and is labelled as 25 MHz resonant frequency. It has a 6.2 mm element<br />
and a focal length of 25 mm in water making it a F/4 class. Signals reflected from the sample are<br />
picked-up by the same transducer and routed to the amplifier and gating electronics; they are then<br />
passed to an oscilloscope and a spectrum analyzer. First, the samples were scanned ultrasonically to<br />
pinpoint the defects. Then the transducer was positioned and refocussed to yield the maximum signal<br />
from the void on the oscilloscope and the trace was recorded photographically. A frequency spectrum<br />
was made of the returned signals by means of a spectrum analyzer and this was also photographed;<br />
however, noise from the gating electronics obscured the spectrum of signals from small bubbles<br />
pointing to the need of improved gating circuitry, to improve the signal-to-noise ratio. Oscilloscope<br />
trace photographs for select voids were enlarged and digitized on a graphics tablet so that they could<br />
be Fourier transformed by computer to obtain frequency spectra without interfering noise. From<br />
these spectra, frequency measurements could be made accurately. Figure 2 shows the typical signal<br />
reflected from the flat surface of the samples and its corresponding frequency spectrum. The shape of<br />
this spectrum was theoretically approximated by a Gaussian distribution with a centre frequency of<br />
30 MHz and a full width at half maximum of 16 MHz (Figure 3) for use in the calculations.<br />
IV. RESULTS<br />
Figure 4 a-d is representative of the voids examined. Shown in (a) is a photograph of the void giving<br />
the major pnd minor cross-sectional dimensions; in (b) the signal trace of the void from the<br />
oscilloscope; in (c) a photograph of the spectrum analyzer trace, and in (d) a plot of the frequency<br />
spectrum as calculated from the Fourier transform of the digitization of the oscilloscope trace. (Note<br />
the absence of noise on this trace). The shape of the signals and spectra, changes with void<br />
dimensions. In order to quantify these changes, the frequency at maximum amplitude and frequency<br />
full width at half maximum (FWHM) were measured.<br />
Figures 5 and 6, a and b are measured and calculated frequency spectra for void diameters of 49 urn<br />
and 132 urn. The calculated spectra have been determined for a shock pulse and glass matrix<br />
parameters ct = 5660 ms 1 and C|/cr = 1.67, which are average values for crown glass. Rather than<br />
a one to one comparison of measured to calculated frequency spectra, repeated calculations have been<br />
performed to generate plots of frequency at maximum amplitude versus void diameter (Figure 7a)<br />
and frequency FWHM versus void diameter (Figure 7bl
- 334 -<br />
It is also possible to do the reverse and transform the complex frequency spectra into time spectra;<br />
Figure 8 is an example o[ this for a 90 pm void. Transformations for void sizes from 28 urn to 132 urn<br />
diameter show that the fargest deflection from zero was negative-going. In Figures 9a-d, peak<br />
deflections versus time have been plotted with the largest negative-going deflection normalized to -<br />
100%. From this figure, it is evident that as the void diameter increases, the largest positive-going<br />
deflection changes from the one following the largest negative-going deflection to the positive-going<br />
deflection preceding it. This behaviour has been noted previously in the experimental results.<br />
Plotting ratios of these two positive going peaks against the void dimensions (fig. 10) shows that there<br />
exists a reasonable correlation with the minor dimension of the void. Although a similar relation<br />
exists with the major void dimensions it is less apparent, especially for the larger dimensions which<br />
approach or exceed the wavelength, \c, of the main frequency component..<br />
V. C<strong>ON</strong>CLUSI<strong>ON</strong>S<br />
Calculations have shown that trends apparent in-signals reflected from voids in the focal zone of<br />
pulsed focused transducer can be reproduced by a model based on scattering of monochromatic plane<br />
waves convoluted with the frequency distribution of the incoming wavepacket. Although direct<br />
signal amplitude comparison of theory and experiment is not possible due to lack of information about<br />
the focal zone pressure, comparison of normalized signals show strong correspondences in signal<br />
shape. The calculated changes of the frequency distribution as a function of void diameter as<br />
measured by the frequency at maximum amplitude and frequency full width at half maximum have<br />
been followed closely by the measured values.<br />
The ratio of the positive going signal peaks in the time spectra have proven to be related to the size of<br />
the reflector and the wavelengths used. It can be used as a practical tool for absolute sizing without<br />
the use of artificial reference.<br />
REFERENCES<br />
1. Proceedings of the DARPA/AFWAL Review of Progress in Quantitative NDE (1981).<br />
2. Van den Andel, J Stockman, A., and Nicholson, P.S., "The Importance of Correct Alignment<br />
in Ultrasonic Testing", Advanced NDE Technology, Bussiere, J.F. ed., NRC Canada, 1982, p.<br />
21.<br />
3. Nicholson, Stockman and Van den Andel, "Ultrasonic Detection of Defects in High<br />
Performance Ceramics", CSNDT Journal, Vol. 4, No. 2, Nov. 1982.<br />
4. Ying, CF. and Truell, R., "Scattering of a Plane Longitudinal Wave by a Spherical Obstacle<br />
in an Isotropically Elastic Solid", J. Appl Phys.36, 1086(1956).<br />
5 Tittmann, B.R., Cohen, R.E. and Richardson. J.M., "Scattering of Longitudinal Waves<br />
Incident on a Spherical Cavity in a Solid", J. Acoust. Soc. Am, 63, 68(1978).
PULSER/RECEIVER<br />
OSCILLOSCOPE<br />
SPECTRUM<br />
ANALYZER<br />
- 335 -<br />
GRAPHICS<br />
TABLET<br />
FOCUSED<br />
TRANS-<br />
DUCER<br />
Figure 1: Experimental Set-Up<br />
MINI-<br />
COMPUTER<br />
SCANNER<br />
C<strong>ON</strong>TROL<br />
FOURIER<br />
TRANSFORMS<br />
MODEL<br />
CALCULATI<strong>ON</strong>S •
- 336 -<br />
-2.0 -1.0 0.0 1.0 2.0<br />
TIME (MS)<br />
10 20 30 40 50<br />
FREQUENCY (MHz)<br />
Figure 2: Signal reflected from flat surface of a sample and its corresponding<br />
frequency spectrum
to<br />
•a<br />
I a. a<br />
Ê £<br />
< IS<br />
i-o<br />
00<br />
-<br />
- 337 -<br />
Gaussian Frequency Spectrum<br />
Centre Frequency = 30 MHz<br />
%<br />
4»<br />
4»<br />
/<br />
/<br />
•*—F W V<br />
HM—-<br />
(5 30 45<br />
Fre quency ( MHz)<br />
Full Width at Half Maximum (FWHM) = 16 MHz<br />
Figure 3: Gaussian frequency distribution showing measured parameters of frequency<br />
at maximum amplitude and full width at half maximum<br />
\<br />
60
- 338 -<br />
130ymx 163 y m<br />
12 22 31 42 52<br />
FREOUENCY\MHz)<br />
130 x 163 pm Void<br />
Figure 4: Representative of characterized voids:<br />
(a) optical character<br />
(b) ultrasonic signal<br />
(c) frequency spectrum obtained from spectrum analyzer<br />
(d) frequency spectrum from Fourier transform of digitized signal<br />
* .
Figure 5: Frequency spectra from:<br />
(a) measured signal off a 50 ym x 53 ym void and<br />
(b) theory for 49 ym diameter void
Figure 6: Frequency spectra from:<br />
(a) measured signal off a 130 yra x 163 pm void and<br />
(b) theory for 132 ym diameter void
- 341 -<br />
100 150<br />
Void Diameter (jjm)<br />
50 100 150<br />
Void Diameter (pm)<br />
LEGEND<br />
+ Minor Diameter<br />
x Major Diameter<br />
Model Prediction<br />
• Input Parameter<br />
( fc =30 MHz, FWHM =1SMHz )<br />
Figure 7: Variation of frequency spectrum parameters with void diameter:<br />
(a) frequency at maximum amplitude and<br />
(b) full width at half maximum (FWHM)
TD<br />
QJ<br />
o<br />
o<br />
c/)<br />
CL -£<br />
E 5<br />
-1<br />
-1<br />
- 342 -<br />
'0 15 30 45 60<br />
Frequency (MHz)<br />
A *<br />
/ l\<br />
l\ l\<br />
^ /M<br />
V !' ^<br />
\| u<br />
0-10 0-20<br />
Tj me (JÜS)<br />
Figure 8: Theory prediction for a 90 ym void:<br />
(a) magnitude frequency spectrum<br />
(b) time si signal obtained from inverse Fourier transform of frequency<br />
spectrum<br />
75<br />
0-30
- 343 -<br />
TIME (ns)<br />
Figure 9: Peak time signal deflections predicted from theory. Signal amplitudes<br />
are normalized to the largest negative-going signal. Shown are signals<br />
for void sizes of (a) 28 pm (b)66 ym (c) 94 ym and (d) 132 pm<br />
100
if)<br />
a .1.0<br />
> 0-9<br />
o ",T;<br />
to<br />
O 08<br />
ÛL<br />
07<br />
Model Trend<br />
Data Trend<br />
- 344 -<br />
Minor Diameter +<br />
Major Diameter x<br />
Model Prediction •<br />
fc = 3 0 MHz, Ac =189jjm<br />
' h-1<br />
50 100 150 ^200<br />
Void Dimension (juin)<br />
Figure 10: Ratio for positive-going peak amplitudes for the peaks preceding and<br />
succeeding the largest negative-going peak. Lines are used to indicate<br />
trends rather than represent a fit to the data.
- 345 -<br />
COMPUTER SIMULATI<strong>ON</strong> OF ULTRAS<strong>ON</strong>IC TESTING<br />
V.B. Duncan<br />
KtomLc. Enz/igy o
1. INTRODUCTI<strong>ON</strong><br />
- 346 -<br />
We are concerned here with the computer simulation of ultrasonic testing techniques<br />
using finite difference approximations of the equations of infinitesimal<br />
elasticity. These equations are used to describe ultrasonic waves in metals and<br />
in models of seismic waves in the Earth. Note that we are interested in the<br />
dynamic (time dependent) form of the equations and that solution of these equations<br />
is only superficially related to the solution of problems in elastostatics.<br />
Another spurious connection is with the simpler Acoustic Wave<br />
Equation; it does not describe the interplay between elastic waves of different<br />
types.<br />
Computer simulation is used in nondestructive testing (NDT) in a variety of<br />
ways. One application is as an aid to the understanding of the physical<br />
processes involved in, for example, wave interactions with reflecting surfaces,<br />
cracks, inhomogeneities and other features. Computer models can be much more<br />
informative than experiments in this work because the displacements and<br />
stresses are predicted for all points in space and time. A more commercial<br />
application is the modelling of specific transducers and testing configurations<br />
where the model is already used to add detar'l to experimental work and will be<br />
used to select transducers before experimental work is started.<br />
We avoid the use of detailed mathematics in what follows and instead give a<br />
general description of the techniques used with references for more detailed<br />
study. The equations of elasticity are stated in two-dimensional form, but<br />
approximation techniques are the same in any number of space dimensions. The<br />
construction of finite difference schemes and the modelling of features such as<br />
cracks and transducers are discussed. The use of the first order form of the<br />
Elastic Wave Equation is highlighted, and its various advantages are explained.<br />
The numerical approximation of the first order form that is used in the<br />
computer model developed by the author is briefly described.<br />
2. EQUATI<strong>ON</strong>S OF INFINITESIMAL ELASTICITY<br />
The equations of infinitesimal elasticity with plane strain can be written in a<br />
number of different ways. They are most often expressed in Cartesian coordinates<br />
in the second order form relating displacements and stresses as follows
pu = T XX + T Xy<br />
tt x y<br />
pV = T Xy + T yy<br />
tt x y<br />
- 347 -<br />
where U and V are the displacements in the x and y directions respectively, the<br />
T's are the stresses, and p is the density of the material. The stresses are<br />
defined in the following way<br />
T XX = \(U + V ) + 2uU<br />
x y x<br />
T yy = \(U + V ) + 2MV<br />
x y J *" y<br />
T Xy « u(U + V )<br />
y x'<br />
where X and p. are the lame elasticities. These two sets of equations can be<br />
combined to give a pair of linear, second order, hyperbolic equations involving<br />
only the displacements U and V.<br />
Reformulating the equations in first order form was proposed by Clifton [1] and<br />
later used by Smith [2] in the form given below. The equations are written in<br />
terms of the 5-vector<br />
We have<br />
where<br />
W T = (pUt,pVt,T xx ,T yy ,T xy ).<br />
-t -x ° -y '<br />
A =/0 0 1 0 0\ , B =/0 0 0 0<br />
0 0 0 0 l \ /OOOIO<br />
cOOOO) lOaOOO<br />
aOOOO/ \ 0 c 0 0 0<br />
b 0 0 0/ \b 0 0 O 0,<br />
and a » A./p, b » u/p, c • (\+2(x)/p.<br />
Again we have a linear, hyperbolic system of equations. The displacement<br />
velocities and the stresses are calculated simultaneously in this first order<br />
form of the equations. The displacement at any point in the space domain can<br />
be found by integrating the velocity of displacement with respect to time.
- 3A8 -<br />
3. PROPERTIES OF THE ELASTIC WAVE EQUATI<strong>ON</strong><br />
We have already noted that the Elastic Wave Equation is hyperbolic (in both<br />
second and first order form) and therefore has travelling wave solutions. In<br />
the interior of the space domain disturbances are propagated with speeds /p/p<br />
and /(\+2|i)/p which are often denoted by c2 and el, respectively.<br />
Compressional waves are disturbances with displacements parallel to the<br />
propagation direction and travel with speed cl. Shear waves are disturbances<br />
perpendicular to the direction of travel and have speed c2. In ultrasonic<br />
testing, we are particularly interested in waves which are localized in space,<br />
have short wavelengths and are fast-moving relative to the length and time<br />
scales of the problem.<br />
The energy in a perfectly elastic material is the sum of its Kinetic and<br />
Elastic Strain Energies. The Kinetic Energy Density Is given by<br />
and the Elastic Strain Energy Density by<br />
1/2 (\+2(i)(U + V ) 2 + |i/2 /(U +V ) 2 - 4U V )<br />
x y V y x x y/<br />
The total energy is found by integrating these densities over the space region<br />
of interest. We note that the energy density can be expressed as an algebraic<br />
combination of the components of the solution vector In the first order form,<br />
but that it involves partial derivatives of the second order form solution.<br />
Energy is conserved in a perfectly elastic material, that Is, the rate of<br />
change of energy in a space region with respect to time is equal to the total<br />
of the energy flux through the boundary of the region. This conservation<br />
property is fundamental and should be modelled accurately.<br />
4. FINITE DIFFERENCE APPROXIMATI<strong>ON</strong><br />
Finite difference approximations of the Elastic Wave Equation were first<br />
developed In the 1960's when their main application was seismology. A major<br />
contributor to this field was Alterman (see [3]). Most of the work with finite<br />
difference approximations in NDT simulation has concentrated on approximating<br />
the second order form of the equations, but two exceptions are mentioned<br />
above. General surveys of these methods in NDT simulation are given by Aboudi<br />
[4] and Bond [5]. A general description of finite difference approximations of<br />
hyperbolic PDE's is given in chapter 4 of Mitchell and Griffiths [6J.<br />
Schemes of implicit type, that is, schemes in which the approximation formula<br />
involves coupling between the points at the latest time level, can be rejected<br />
almost immediately. The reason for this is the prohibitive cost of solving the
- 349 -<br />
linear system of equations coupling the latest time level of the solution.<br />
Marfurt agrees with this view in a recent review paper [7]. We, therefore,<br />
restrict our attention to schemes which calculate the solution at the latest<br />
time level explicitly.<br />
The usual finite difference approximation to the second order form of the<br />
equations is constructed by replacing the partial derivatives with second order<br />
accurate central difference approximations. (The order of accuracy of a finite<br />
difference approximation is the order of the leading term in its truncation<br />
error. See [6], p 21). For example, the second partial time derivative of U is<br />
approximated by<br />
U(x,t) = -^ (U(x,t+At) - 2U(x,t) + U(x,t-At))<br />
At<br />
and the second partial space derivative of U by<br />
U(x,t) = -i- (U(x+Ax,t) - 2U(x,t) + U(x-Ax,t))<br />
xx Ax 2<br />
Bond [5] gives a detailed description of this scheme.<br />
There are a large number of approximation schemes for first order hyperbolic<br />
systems and one must consider their accuracy, efficiency and reliability. The<br />
numerical simulation developed by the author uses a "leap-frog" scheme proposed<br />
by Kreiss and Öliger [8] which is accurate to fourth order in space and to<br />
second order in time. These schemes are particularly useful because they can<br />
be implemented efficiently, give good resolution of the solution and do not<br />
dissipate energy. In addition, they have been used successfully in weather<br />
prediction codes and have been the subject of extensive theoretical<br />
investigation.<br />
Detailed descriptions of the Kreiss and Öliger leap-frog scheme can be found in<br />
[8] and [9], and so we give only a brief description here. The scheme is<br />
constructed by approximating the time derivative of U by<br />
U(x,t)t = Y^r (U(x,t+At) - U(x,t-At))<br />
and the space derivatives by<br />
U(x,t) = yi-^- |U(x-2Ax,t) - 8U(x-Ax,t) + 8U(x+Ax,t) - U(x+2Ax,t)J<br />
The difference approximation of the space derivatives must be modified at<br />
points adjacent to and on the boundaries of the space region to avoid the use<br />
of points outside the region.
- 350 -<br />
5. ADVANTAGES OF USING THE FIRST ORDER FORM<br />
An important advantage of the first order form is that the most common boundary<br />
conditions can be expressed without involving space derivatives of the solution<br />
components. This makes numerical approximation at the boundaries much easier.<br />
For example, the free surface (stress free) boundary condition involves only<br />
two linear combinations of the three stresses. More general conditions on the<br />
normal and tangential stress and displacement are equally easy to apply.<br />
In general, the space domains of interest include a wide variety of shapes that<br />
do not lend themselves to modelling on a Cartesian grid. Many cases are suited<br />
to a polar coordinate system or require the use of a non-uniform grid. It is<br />
generally much more difficult to modify the second order form of the equations<br />
to use non-rectangular grids than it is to modify the first order form. The<br />
modelling of inhomogeneous materials is also easier when the first order form<br />
is used, and the modified set of equations which results is similar to the<br />
modified equation set for a non-uniform grid. In both cases, the advantage of<br />
using the first order form is due simply to the ease of manipulation of lower<br />
order space derivatives.<br />
Little is known about the properties of numerical approximations to the second<br />
order form of the equations, but approximations of first order hyperbolic PDE's<br />
have been investigated in great detail. The stability of difference<br />
approximations to initial-boundary value problems, such as we are discussing,<br />
is not easy to guarantee; results are given in [9] and [10] for the first order<br />
equations. The stability of approximations to the second order form of the<br />
Elastic Wave Equation has only been investigated experimentally and should not<br />
be assumed to hold in all cases.<br />
6. LIMITING THE COMPUTATI<strong>ON</strong>AL AREA<br />
In most cases, it is not necessary to model the whole object being tested. One<br />
can safely ignore the parts that cannot contribute to the response signal at<br />
the transducer during the time period of interest. When modelling the testing<br />
of weld joints in pipelines, for instance, there is no need to consider more<br />
than the small region that includes the transducer and weld. Any signal<br />
spilling out of this region is effectively lost down the pipeline.<br />
To limit the extent of the computational region, an artificial boundary is used<br />
to truncate the object and absorb the energy of any disturbance reaching it.<br />
Ideally, there should be no reflected disturbance at this boundary, and waves<br />
should appear to pass through it undistorted. It is possible to impose<br />
perfectly absorbing boundary conditions for hyperbolic systems in one<br />
dimension, but it is usually impossible to do so for multi-dimensional<br />
problems. A study of absorbing boundary conditions for wave equations can be<br />
found in Clayton and Enquist [11].<br />
The simplest absorbing boundary conditions have been used in our numerical<br />
model. They are designed to absorb waves travelling perpendicularly to the<br />
boundary perfectly, but have proved to be most satisfactory for waves of all<br />
orientations tried so far. In fact, the energy reflected by this artificial<br />
boundary usually amounts to only a few percent of the energy reaching it.
- 351 -<br />
These boundary conditions are constructed by assuming that the solution is<br />
uniform in the direction parallel to the boundary, then applying the perfectly<br />
absorbing conditions for the resulting one space dimension equation<br />
perpendicular to the* boundary. For example, the equation is approximated at<br />
the boundary x=0 (i.e. a vertical surface) by the one-dimensional equation<br />
W_ = AW .<br />
—t —x<br />
The perfectly absorbing boundary conditions for this equation are<br />
) - T XX = 0<br />
/ET(pV ) - T xy = 0 .<br />
7. CRACKS AND CAVITIES<br />
We are primarily concerned with the detection of cracks and cavities in<br />
metals. In an unbounded elastic material it is possible to model a wide<br />
variety of cracks and cavities by fitting the coordinate system of the<br />
simulation to the geometry of the flaw. If boundaries and flaws must all be<br />
modelled, then it can be much more difficult to fit a coordinate system to the<br />
geometry of the problem.<br />
Planar cracks with stress-free surfaces have been successfully modelled.<br />
Cracks with various orientations can be included in the approximation by using<br />
a coordinate system which is skewed so that one axis is parallel to the crack.<br />
The crack tips have so far been assumed to be stress free, but further investigation<br />
of this is required. Some cavities with a simple parallelogram shape<br />
have also been modelled, but more cojplicated geometries have not yet been<br />
attempted.<br />
8. TRANSDUCERS<br />
The simulation of transducer generated ultrasonic pulses has two closelyrelated<br />
purposes. It can simply be regarded as a way to send a pulse with<br />
reasonable physical properties into the computational region to gain insight<br />
into the physics of interactions with cracks and other features. Alternatively,<br />
accurate modelling of the input can be used to study the properties of<br />
specific transducers. We note that accurate models of transducers can be used<br />
to predict both transmission and reception of signals.
- 352 -<br />
At Chalk River, we have concentrated on modelling the physics of an ultrasonic<br />
pulse inside the object being tested and not on the detailed modelling of<br />
specific transducers. Pulses of both shear and compressional types can be<br />
input at arbitrary angles to free surfaces and absorbing boundaries. This is<br />
achieved at free surfaces by applying stresses and at absorbing boundaries by<br />
specifying a combination of stress and displacement.<br />
9. COMPUTATI<strong>ON</strong>AL COSTS<br />
Computer simulation of the Elastic Wave Equation is limited on any computing<br />
device by two main factors: the availability of fast access storage; and the<br />
cost of executing the program. The execution cost is simply the cost of<br />
performing the arithmetic required to calculate successive time levels of the<br />
solution. Past storage and retrieval of the values of the approximate solution<br />
at the two preceding time levels is important because of the large amount of<br />
data _nvolved and the fact that it must be done every timestep. Information<br />
storage and retrieval becomes very expensive if it cannot be done in the<br />
machine's central memory (or equivalent) and alternative storage, such as disk,<br />
is used.<br />
Execution and storage costs are related to the number of grid points used in<br />
the construction of the finite difference approximation. To illustrate this we<br />
consider an approximation in which the number of grid points for each of the<br />
space dimensions is directly proportional to N (a positive integer). The total<br />
number of space grid points is then proportional to N to the power D, where D<br />
is the number of space dimensions. The amount of storage required for the<br />
approximation and the number of arithmetic operations per timestep are<br />
multiples of the number of space points. The number of timesteps is proportional<br />
to N and so the total number of arithmetic operations is a multiple of N<br />
to the power (D+l).<br />
The number of space points required to approximate the solution to a given<br />
accuracy depends on the physics of the problem. Probably the most significant<br />
factor is the ratio of the wavelength of the ultrasonic pulse to the length<br />
scale of the object it is traversing. The number of subdivisions in each space<br />
dimension is inversely proportional to this ratio, making it more expensive to<br />
model shorter wavelength pulses in the same object.<br />
To achieve reasonable accuracy in typical problems encountered at Chalk River<br />
it was necessary to use values of N of at least 100. To model all the problems<br />
we have considered so far, values of N of 400 or more would be required. The<br />
storage and execution costs of two-dimensional modelling of the easier problems<br />
severely stretched the capabilities of our CYBER 175, and difficult<br />
two-dimensional problems would stretch most Supercomputers. General threedimensional<br />
modelling of most problems is certainly not possible with current<br />
technology; the execution and storage costs are prohibitive on any existing<br />
machine.
10. EXAMPLE<br />
- 353 -<br />
The following example demonstrates the application of our numerical model to a<br />
simple test problem. It represents the generation, propagation and reflection<br />
by a crack of a shear wave in steel. The computational region is a 10 mm by 10<br />
mm square and all its edges are energy-absorbing except for a 1 mm crack along<br />
the centre part of the right edge. The figures show the geometry of the region<br />
and plot the energy density by contour levels which are graded to pick out the<br />
low amplitude features.<br />
The first two figures show the generation of the wave at the left edge uf the<br />
region. It is generated by specifying a combination of stress and<br />
displacement, and it has the form of an exponentially damped wave packet wir.h<br />
frequency 3.5 MHz. Small amplitude compressional waves are generated at tho<br />
edges of the input region due to the rapid variation in stress and displacement<br />
there. There are no signs of spurious reflection by the absorbing boundaries.<br />
Figures 3 and 4 show the shear wave after reflection by the crack. A large<br />
amount of energy is lost through the absorbing boundary, but a significant<br />
portion is reflected by the crack. Another important feature is the scattering<br />
by the crack tips which shows itself mainly in the two compressional waves<br />
emanating from the tips.<br />
11. SUMMARY AND C<strong>ON</strong>CLUSI<strong>ON</strong>S<br />
We have described the construction of numerical approximations of the Elastic<br />
Wave Equation and outlined the particular advantages of using the first order<br />
form of the equations. We have discussed the current and potential uses of<br />
computer simulation in ultrasonic testing and also suggested that powerful<br />
computers are required if codes are to be run reasonably quickly and<br />
efficiently. The example given shows that the physical phenomena of reflection<br />
and crack-tip scattering can be successfully modelled and we have been equally<br />
successful in modelling other physical phenomena.<br />
12. REFERENCES<br />
1) R.J. CLIFT<strong>ON</strong>, Quart. Appl. Math., V. 25, p.97 (1967).<br />
2) P.D. SMITH, Int. J. Num. Meth. Eng. , V. 8, p.9x11 (1974).<br />
3) Z. ALTERMAN and D. LOEWENTHAL, Methods in Computational Physics Vol. 12,<br />
Ed. B.A. Bolt, pp.35-164, Academic Press (1972).<br />
4) J. ABOUDI, Modern Problems in Elastic Wave Propagation, Eds. J. Miklowitz<br />
and J.D. Achenbach, pp. 45-65, Wiley (1978).<br />
5) L.J. B<strong>ON</strong>D, Research Techniques in NDT Vol. 6, Ed. R.S. Sharpe,<br />
pp. 106-150, Academic Press (1983).
- 354 -<br />
6) A.R. MITCHELL and D.F. GRIFFITHS, The Finite Difference Method in Partial<br />
Differential Equations, Wiley (1980).<br />
7) K.J. MARFURT, Geophysics, V. 49, pp.533-549 (1984).<br />
8) H. KREISS and J. ÖLIGER, Methods for the Approximate Solution of Time<br />
Dependent Problems, Global Atmospheric Research Programme Publications<br />
Series No. 10 (1973).<br />
9) J. ÖLIGER, Math. Comp., V. 28, pp.15-25 (1974).<br />
10) D.M. SLOAN, Math. Comp., V. 41, pp.1-11 (1983).<br />
11) R. CLAYT<strong>ON</strong> and B. ENQUIST, Bull. Seis. Soc. Amer., V. 67, pp. 1529-1540<br />
(1977).
3<br />
"/A !<br />
- 355 -<br />
1 I 2 I<br />
Figs. 1-4. Energy density at times 0.43, 0.86, 3.85 and 4.27 ys, respectively,<br />
in a 10 mm by 10 mm region. The dashed lines show the energy<br />
absorbing boundaries and the thick line on the right edge<br />
represents a 1 mm crack.
- 356 -<br />
A NOVAL APPROACH TO EDDY CURRENT IMAGING OF DEFECTS IN<br />
ALUMINUM SHEETS<br />
V. Le.zma.ni, M. Macecefe<br />
Tzchno Sc-cewi-cjj.tc Inc.<br />
Ontcifi-lo<br />
ABSTRACT<br />
Eddy current testing relies on the disturbance of eddy currents<br />
induced in a conducting material by an alternating<br />
electromagnetic field. While the data acquisition is very<br />
simple, its interpretation is complex and a lifetime of<br />
experience of the operator is sometimes required.<br />
This paper describes a new approach where the flaws in aluminum<br />
sheets will be imaged in a fashion similar to an X-ray or<br />
ultrasonic C-scan image. An array of coils is used to generate<br />
the image of the flaw. Both wire wound and printed circuit coil<br />
arrays can be used. The inversion of the data is discussed and<br />
future work outlined.
INTRODUCTI<strong>ON</strong><br />
- 357 -<br />
The inspection for defects by Eddy Currents is based on an indepth<br />
analysis of the impedance changes experienced by a single<br />
coil (1,2). This can, after analysis, yield the location of the<br />
defect and an estimate of its depth. However, the evaluation of<br />
the defect type and depth from the impedance plane image or chart<br />
recording is tedious (Figure 1). Special calibration is required<br />
to obtain semi-quantitative data on flaw size. Our work proposes<br />
to take a different approach: developing an eddy current system<br />
capable of imaging flaws directly, with the goal of designing a<br />
real time eddy current imaging system to detect and image flaws.<br />
EDDY CURRENT IMAGING<br />
A breakdown of eddy current methods for imaging is based on two<br />
characteristics:<br />
1) the movement of the testing system in relation with<br />
the test object.<br />
2) the number of coils used.<br />
One method of eddy current imaging is eddy current holography,<br />
described in detail by B.P. Hildebrand (3). This technique is<br />
based on the observation that phase is dependent on the distance<br />
from the receiving coil and uses the backward wave reconstruction<br />
algorithm of Van Rooy (A). The results reported are encouraging<br />
but there are some distinct problems with focus and resolution.<br />
The experimental set-up is of the SD (single coil dynamic) type,<br />
scanning a single coil, preprocessing the signal in a multifrequency<br />
eddy current instrument to minimize the noise due to<br />
variations in lift-off, and using a digital reconstruction<br />
algorithm developed originally for ultrasonics.<br />
More sophisticated inversion procedures are discussed in the<br />
1 itérât - e (5) based on the convolution of some parameters of the<br />
test object, and a response function. C-scan type pictures have<br />
been made for eddy currents (6). This practice was also used<br />
extensively to distinguish between types of defects (pits,<br />
cracks).<br />
The AS (array-static) type of test is quite different in nature<br />
from the previous one. It consists of N coils arranged in an<br />
array which are then placed on top of the test object to produce<br />
an image.<br />
There are several approaches to array imaging. One can separate<br />
the driver and receiver function of the coils; the array of<br />
receiving coils is placed within a large driver coil. This has<br />
the advantage that phase lags can be measured with respect to one<br />
driver coil. The signal information can be processed in a binary<br />
fashion. This method is being used by DREP (7)»
- 358 -<br />
The impedance (or the phase lag) of each coil is measured and if<br />
there is a sufficiently large deviation from the measurement<br />
above a good test object a 1 is assigned to the location of the<br />
coil in question. Another approach has the coils acting<br />
simultaneously- as driver and receiver. The induced eddy currents<br />
would be similar around each probe. If the eddy current signal<br />
could be analysed at the coil location for depth (by measuring<br />
amplitude and phase at frequencies), an immediate profile of<br />
the defect could be shown. A holographic method of imaging is<br />
also possible in the AS mode. It is assumed that the response of<br />
a coil is not altered by the presence of neighbours. If no<br />
reasonable pictures can be developed from the static case (AS),<br />
it will be difficult to develop these from the dynamic case (AD).<br />
There are similarities with early radiographie techniques<br />
indicating the presence of something but no indication of the<br />
depth. Therefore a binary picture will not give the necessary<br />
information to fully evaluate the importance of a defect.<br />
However, it suffices for the detection of anomalies.<br />
In our approach, the response of each individual coil is<br />
processed, resulting in a depth profile. The experiments<br />
addressed the questions below.<br />
- Are all the coils identical or can their response be made<br />
identical?<br />
- Are neighbours altering the coil response?<br />
- Can a calibration curve be drawn for all probes?<br />
EXPERIMENTAL<br />
Design of Coils and Building of the Array<br />
We used the Hocking Phasec D6A Eddy Current instrument, which is<br />
a multi-frequency device (ranging from 1 kHz to 1 MHz), highly<br />
flexible and easy to operate. The Hocking Phasec D6A prefers<br />
driver coils with impedances larger than 50 ohms. This puts<br />
certain limitations on the choice of coils and the working<br />
frequency.<br />
Two tools assisted in coil design. The first is a computer<br />
program AIRCO, developed by Dodd et al.,(9) which allows for the<br />
calculation of the impedance of coils with air cores, taking<br />
into account geometry and wire gage. This program was adapted<br />
for the microcomputer in C BASIC. Ferrite cores were also<br />
introduced to increase the inductance of the coils.<br />
Probes were manufactured in-house or were commissioned from a<br />
reputable manufacturer for the final array. An HP 4192A<br />
impedance meter was used to measure the impedance and inductance<br />
of the coil in air. The resonance frequency of the coils was over<br />
600 kHz.
- 359 -<br />
A 3 by k array was constructed with twelve nominally identical<br />
coils. It was made by potting the ferrite ends in holes which<br />
were drilled carefully in a fiberglass plate. This assembly was<br />
then potted and the leads connected to a mechanical switch. See<br />
F igure 2.<br />
Mechanical Scanner ^R j g<br />
In order to assess the effect of coil location on the defect<br />
response, a number of scans were carried out. A computer<br />
controlled ultrasonic immersion X-Y scanner provided accurate<br />
encoding of the X,Y position of the coil, as well as an isometric<br />
display. A number of repeatable scans with variable scanning<br />
patterns could then be made. A block diagram of the experimental<br />
set-up is shown in Figure 3.<br />
INFLUENCE OF NEIGHBOURS <strong>ON</strong> THE PERFORMANCE jDF CO ILS<br />
A series of experiments was developed to assess the influence of<br />
neighbours. The first test consisted of moving the coils with<br />
respect to each other while recording the differences in signal<br />
output of the eddy current instrument. Impedance of a coil<br />
was also measured before and after it was placed into an array.<br />
The principal reason for the change in impedance when neighbours<br />
are present < Lnought to be the presence of ferrite cores. Due<br />
to this higl- permeability material there is an increase in the<br />
impedance uf the coil. Figure U shows the change in impedance<br />
as a function of distance between two coils. It shows the<br />
exponential drop off with distance, with the effect disappearing<br />
at about *i mm. In our array,coils are placed within a distance of<br />
less than two millimeters and are therefore strongly affected<br />
by neighbours. The impedances and inductances were also measured<br />
on twelve coils before and after they were arranged in an array<br />
with distance between nearest neighbours equal to 1.7 mm. The<br />
change in inductance increases with increasing Neighbouring<br />
Factor. While the standard deviation was 3-6 JJH for individual<br />
coils, it became 8.9 fiH when the coils were in an array. The<br />
changes between individual coils entail that:<br />
1. rebalancing of each coil in the array is<br />
important for traditional eddy current<br />
system approach;<br />
2. the lift-off direction will be different<br />
for each coi1 ;<br />
3. calibration curves will be coil dependent.<br />
These coils were wound within one turn of each other on<br />
Identical ferrite slugs. It is unlikely that the<br />
reproducibility of the coils can be improved with existing<br />
techniques. Photo lithography was an alternative which was<br />
exploited.
Printed Circuit Coils<br />
- 360 -<br />
To assure repeatability of the coil parameters, a printed circuit<br />
coil method was applied. Printed circuit coils (PCC) were made<br />
with planar technology. A master was first drawn on a high<br />
contrast medium, then photographically reduced (100 X) to actual<br />
size. The spiral coils were drawn by a microcomputer scanner.<br />
The masters were reduced in a photolithographic laboratory on<br />
Kodak high resolution plates.<br />
Figure 5 shows the layout of a multi turn planar inductor in a<br />
spiral configuration. A simplified design equation giving the<br />
inductance L was reported by Brown (8) as<br />
L = 0.8(r - nd/2) n<br />
(6r + 7nd)<br />
L t Inductance (nH)<br />
r : Outer radius of coil (mil)<br />
d : turn width and spacing between turns (mil)<br />
n : number of turns<br />
The agreement with the equation was within 10%; wound coils 1<br />
inductances were higher. The coils were first made on regular<br />
printed circuit material but a serious underetching problem<br />
prevented the use of the thick (0.002") Cu plate. Pre-thinning of<br />
the copper plate by etching did not produce an even surface.<br />
Therefore, vacuum deposition of aluminum was attempted giving an<br />
Al layer in excess of 1000 A. Shipley positive photoresist was<br />
applied and coil patterns were exposed to ultraviolet light.<br />
After developing, the aluminum was etched in an etchant based on<br />
phosphoric acid, HNO , and acetic acid. The strip width is about<br />
12 /urn. A number of substrates were experimented with including<br />
glass, plexiglass, machinable ceramics and nylon. Best results<br />
were obtained with machinable ceramics, but the plexiglass and<br />
other flexible materials were also mastered. The coils produced<br />
in this way had a desired geometry and repeatability but due to<br />
tne thin layer of aluminum, very high resistance. To increase<br />
the cross section of the conductor, a copper or gold layer was<br />
grown on the aluminum pattern by electrochemical means. Finally,<br />
contact holes 0.006" (0.15 "*") were drilled and contacts secured.<br />
Using this technology, repeatable and identical coils could be<br />
produced in any array configuration. The particular coil was<br />
2.18 mm in diameter and had 29 turns<br />
We have also experimented with larger coils (21.8 mm in<br />
diameter.) Kodak negative photoresist and standard copper clad<br />
plates were utilized. This type of coil has 29 turns, a<br />
resistance of A ohms and an inductance of 10.3 pH at 1 kHz.<br />
Resonance did not occur below 13 MHz. The requirement of the<br />
instrument for 50 ohms impedance indicated that higher<br />
frequencies (above 700 kHz in this case) should be used. The<br />
coils picked up all the notches including the .1 mm deep by 2 mm<br />
long flaw.
INVERSI<strong>ON</strong> PROBLEM<br />
- 361 -<br />
The inversion is a mathematical process which yields the true<br />
dimensions of the flaw based on the signals obtained with sensing<br />
coils. A number of works address purely theoretical inversion<br />
protocols and very little work is done on experimental data.<br />
The modelling of lift-off as well as the response of practical<br />
probes is well understood. A flaw response in both twodimensional<br />
and three-dimensional geometries has also been<br />
developed. Although great progress has been made in improving<br />
mathematical modelling of eddy current responses, the theory has<br />
not been adequately tested against experimental practice. A<br />
flowchart of the inversion is shown in Figure 6.<br />
We developed a novel image producing method based specifically on<br />
array concepts. A hexagonal array (AB type) is suggested as<br />
shown in Figure 7« Calculations by W.G. Simpson et.al. (9) have<br />
shown that the change in impedance due to a point defect can be<br />
approximated by a complicated relationship represented by the<br />
equation below, where F (x)=signal response, x=distance from<br />
centre of coil and r=effective coil radius. Simply put, the<br />
coil is not sensitive under its centre.<br />
F (x) = A M - cos<br />
F (x) = 2A xlexp - [x - l\ | L< x < r<br />
2A xlexp - /x - i\ 1<br />
F (x) = A exp |-2fx-r<br />
x ]<br />
r)<br />
~ I V 2 JJ<br />
The parameter of interest (eg. crack depth) in matrix form is:<br />
T =<br />
t(i,l) t(i,m)<br />
t(n,i) t(n,m)<br />
where (i,k) is the location of the coil with (i,k) coordinates.<br />
The response of the n X m coils is also described by the matrix<br />
S =<br />
s(n,i) s(n,m)<br />
The measured S matrix should be transposed into a T matrix.<br />
Let us assume that the response of the coil at location n,m is a<br />
weighted sum of the crack depth values (t) at the nearest<br />
neighbour positions.<br />
s(m,m) = t(n-1,m-1) + t(n, m-1) + t(n-1,m) + t(n, m+1)<br />
+ t(n+1, m-1) + t(n+1,m)
- 362 -<br />
This is a reasonable approximation if t(n,m) is the average value<br />
of the depth over a circular area. The sum of the responses over<br />
the six nearest neighbours of the location that is to be<br />
investigated (n,,m) is calculated overt<br />
F(n,m) = s(n-1,m) + s(n, m-1) + s(n, m+1) + s(n-1, m-1)<br />
This can be expressed as<br />
+ s(n+1, m-1) + s(n+1,m)<br />
F(n,m) = 6t(n,m) + 2 t(n-1,m) + 2t(n,m-l)<br />
+ 2t(n+l,m) + 2t(n,m+l) + 2t(n-1,m-1)<br />
+ 2t(n+T,m-1) + 2t(n+2,m) + 2t(n-2,m)<br />
+ 2t(n+1,m-2) + 2t{n+1,m+i) + 2t(n-1,m-2)<br />
+ 2t(n-1,m+1) + t(n-2,m-1) + t(n-2,m+1)<br />
+ t(n+2,m-1) + t(n+2,m+1) + t(n,m-2) + t(n,m+2)<br />
The first sufficient approximation is t(n,m) = F(n,m)<br />
6<br />
This solution can be used for an iterative procedure for the<br />
solution of n X m linear equations with n X m unknowns, thus<br />
completing the transformation of the S matrix into a T matrix.<br />
The values of the T matrix can then be color coded to produce an<br />
image with depth indications. The above methodology applies both<br />
to wire wound and printed circuit coils. This procedure, which<br />
plots depth estimates at the center of each individual coil<br />
core, can also be used in the dynamic case.<br />
TESTS PERFORMED WITH THE TSI ARRAY<br />
This array was prepared in house. In short, twelve coils were<br />
arranged in a rectangular pattern of three by four coils, with<br />
the distance between the nearest neighbours being 1.75 nm - The<br />
individual coils serve both as driver and receiver.<br />
The performance of individual coils was assessed, namely the<br />
sensitivity to detect the shallow (.1mm deep by 2mm long)<br />
slot (Figure 8). The coils picked up the defect with no<br />
difficulties.<br />
This array was also mechanically scanned over an aluminum plate<br />
containing 3 3" hole. Figure 9 shows two isometric plots<br />
obtained by scanning the array over 33 m hole.<br />
We experienced some asymmetries: coils 1 to 6 had a larger left<br />
peak than right peak with the opposite being the case for ^ to<br />
12. This was due to the higher concentration of flux lines to
- 363 -<br />
the right of coils 1 to 6, with the opposite being the case for<br />
coils 7 to 12. These flux lines were constrained by the presence<br />
of the neighbours.<br />
Tests were performed with single coils over the source plate and<br />
symmetrical curves were observed as shown in Figure 10. The<br />
asymmetries observed in the array studied were due solely to the<br />
effect of neighbours and / or asymmetries in the induced field.<br />
The coils had to be individually balanced, and lift-off<br />
directions and multipliers had to be adjusted for each coil.<br />
The eddy current signal was influenced by individual coil<br />
parameters and to a small degree by the location of the coil. It<br />
was acknowledged that coil parameters such as impedance were<br />
affected by the location of the coil in the array. Also, for our<br />
array, a simple signal processing method was not directly<br />
apparent. However, a novel inversion protocol was devised which<br />
would lead to advanced true flaw imaging.<br />
C<strong>ON</strong>CLUSI<strong>ON</strong>S<br />
Eddy Current imaging is feasible with arrays of coils. Coils,<br />
when arranged in an array, influence each other. Interactions<br />
between neighbours must be considered. The changes in impedance<br />
due to the presence of defects are very small (Az = 10" 3 - 10" 2 ).<br />
With present technology there is a need for individual treatment<br />
of each coil in the areas of balancing, lift-off orientation and<br />
calibration. We believe the balancing problem can be solved by<br />
means of software.<br />
The test notch (,1mm X 2 mm) is detected with no difficulties in<br />
the coil design used in the array. The coils should be made<br />
small, identical and powerful. Printed circuit coils can result<br />
in identical elements in the array. However these coils operate<br />
at higher frequencies (250 KHz and above) and would be useful for<br />
the inspection of open shallow cracks.<br />
We have developed a novel inversion procedure to size flaws. A<br />
hexagonal stacking of identical coils is advocated and the<br />
summing of coil responses in groups of six is shown to have<br />
significant signal to noise improvement. Utilizing arrays to<br />
produce images of defects yields improved S/N and prompt imaging,<br />
speeds the inspection and provides superior flaw<br />
characterization.<br />
The signal processing and three dimensional colour graphics<br />
display techniques developed elsewhere can be easily applied to<br />
this problem (10).
ACKNOWLEDGEMENT<br />
- 36A -<br />
The work was supported by the Defence Research Establishment<br />
Pacific. We acknowledge the input of Mr. W. Sturrock, his<br />
critical comments as well as permission to publish the work.<br />
References<br />
1. H.L. Libby, Introduction to Electromagnetic Nondestructive<br />
Test Methods, (New York; Wiley- Interseience, 1971).<br />
2. R. L. Stoll, The Analysis of Eddy Currents, (Oxford :<br />
Clarendon Press,<br />
3. B. P. Hildebrand,Eddy Current Imaging, Spectrum Development<br />
Laboratories, Inc., 1982.<br />
k. D. L. Van Rooy, Digital Ultrasonic Wavefront Reconstruction<br />
in the Near Field, IBM Report # 320.2402.<br />
5« B.A. Auld, et.ai., "Quantitative Modelling of Flaw<br />
Responses in Eddy Current Testing", Nondestructive<br />
Testing, Vol. 7, (London: Academic 'Press, 198A),<br />
PP 37-76-<br />
6. D.C. Copley, "Eddy Current Imaging for Defect Characterization<br />
", Review of Progress in Quantitative NDE, (1982.)<br />
7. W. R. Sturrock, Defence Research Establishment Pacific,<br />
private communication to M. Macecek, 1984.<br />
8. R. B. Brown, "Experimental Study of Q Dependence on Thin<br />
Film Inductor Geometry", International Microelectronics<br />
Symposium, 1973., paper 5A, pp 1-13.<br />
9. W. A. Simpson, C. V. Dodd, J. W. Luquire and W. G. Spoeri,<br />
Computer Programs for _sqnie Eddy - Current Problems -1970,<br />
ORNL-TM-3295<br />
10. M. Macecek, K.L. Luscott, J.D. Wells and D.K. Mak,<br />
"Automated Ultrasonic Testing System for the<br />
Characterization of Defects in Weldments", Paper to the 5th<br />
Canadian Conference on NDT, Toronto, Oct. 29-31, 1984.
—-<br />
••;••••<br />
'IM<br />
i lüiiaiiüiiii! au.<br />
BR JSH ACCUCHART<br />
1<br />
t—<br />
If* »fff<br />
—I—\<br />
M 2<br />
éi<br />
|<br />
- 365 -<br />
•"1<br />
H 1 1-<br />
I?»<br />
rt:<br />
4 5<br />
M,<br />
1 - .005" notch<br />
2 - .015" notch<br />
3 - .030" notch<br />
U - lift-off pad<br />
5 - ferromagnetic<br />
inclusion<br />
Figure 1. Eddy Current impedance plane images of notches<br />
and corresponding chart recording
\<br />
X-Y SCANNER<br />
OBJECT<br />
- 366 -<br />
Figure 2. Potted coil array with switch arrangement<br />
IE.C . PROBE<br />
TWO-FREQUENCY<br />
EDDY CURRENT<br />
TESTER<br />
COMPUTER<br />
C<strong>ON</strong>TROLLER<br />
~"<br />
--<br />
Figure 1. Block diagram of experimental set-up<br />
MIXER
- 3hl<br />
2 3 4<br />
OISTANCE (mm)<br />
Figure h. Influence of neighbours on coil impedance<br />
and the physical arrangement of the coils<br />
Figure 5. Layout of the printed circuit coil and<br />
finished probe (X 50)
- 368 -<br />
Multiposiiion/mullifrequency measuremenl<br />
Gross crack size<br />
Position Length Width Opening<br />
Figure 6. Flow diagram illustrating an inversion protocol<br />
(after Auld [5])<br />
Figure 7. Hexagonal array (AB type)
- 369 -<br />
a) 0.080 X 0.010 X 0.004"<br />
55 dB gain<br />
f, = 900 KHz<br />
f2= 700 KHz<br />
b) 1/4 X 0.010 X 0.004"<br />
55 dB gain<br />
f, = 900 KHz<br />
f.,= 700 KHz<br />
Figure 8. Impedance plane images of EDM notches
DISTANCE (in cm)<br />
Figure 9a. Isometric display of the response of individual<br />
TSI array coils to a 3 mm hole; probe TSI #11
- 371 -<br />
DISTANCE ;,n(m)<br />
Figure 9b• Isometric display of the response of individual<br />
TSI array coils to a 3mm hole; probe TSI #9<br />
Figure 10. Isometric plot of the response of a single coil<br />
to a 3mm hole
- 372 -<br />
DEVELOPMENTS IN X"RAY STRESS MEASUREMENT<br />
THE CANMET PORTABLE STRESS DIFFRACTOMETER<br />
R.A. Holt<br />
Enzfigy, H-Lne.i> and Rz&ou.fic.Q.i, Canada<br />
Ottawa, Ontaftlo<br />
ABSTRACT<br />
The measurement of stress in a surface with X-rays makes use of<br />
the Bragg law for the diffraction of monochromatic X-rays from a crystal<br />
to measure precisely the small changes in lattice parameter corresponding<br />
to elastic distortion, until recently, accurate stress measurements<br />
using this technique have been confined to the laboratory.<br />
The recent development of compact position sensitive proportional<br />
counters, solid state high voltage supplies and compact X-ray<br />
tubes and the availability of powerful, inexpensive computers make it<br />
possible to design a diffractometer to measure stress in engineering<br />
structures and components in the field under clement conditions.<br />
The CANMET Portable Stress Diffractometer has been designed to<br />
make accurate stress measurements in the field while retaining adequate<br />
versatility for dedicated laboratory use. It has several unique features<br />
including the precisely controlled angular alignment of the incident<br />
X-ray beam relative to the specimen surface which allows the measurement<br />
of stress in coarse grained materials and those with pronounced gradients<br />
in crystallographic texture. A commercial prototype of the instrument<br />
is currently being built.
- 373 -<br />
INTRODUCTI<strong>ON</strong><br />
X-ray diffraction is one of a few nondestructive techniques for<br />
measuring stress in engineering components and structures. The basic<br />
principle is simple and techniques using X-ray sensitive film for measuring<br />
stress in crystalline materials were developed more than 40 years<br />
ago (1). Recent developments in electronics and X-ray detector technology<br />
have allowed a reduction in the size of the equipment required and<br />
an increase in the speed with which accurate measurements can be made.<br />
Furthermore, an improved understanding of the relationship between the<br />
stress state in a polycrystal and the distortions measured by X-ray diffraction<br />
allows a more reliable interpretation of data than was previously<br />
possible (2,3).<br />
THE PHYSICS OF X-RAY DIFFRACTI<strong>ON</strong><br />
The measurements of stress by X-rays is based on the Bragg law<br />
for the diffraction of monochromatic X-rays by the planes of a crystal,<br />
i.e.:<br />
nX = 2d sin 6 Eq. 1<br />
where: n is the order of diffraction<br />
X is the wavelength<br />
d is the interplanar spacing of the diffracting planes<br />
e is the diffraction angle<br />
Unique values of X are obtained by making the anode of the<br />
X-ray tube from a pure metal ; the X-ray wavelengths normally used correspond<br />
to the transition of electrons from the L to the K orbitals of the<br />
atoms in the anode. The nature of diffraction from a three dimensional<br />
crystal lattice is such that the X-rays appear to reflect at a unique<br />
angle (S) corresponding to a unique interplanar spacing (d). As ö<br />
approaches 90° - the back reflection range - changes in d corresponding<br />
to elastic strains of less than 10"^ can be measured with acceptable<br />
accuracy.<br />
The X-ray wavelengths useful for diffraction in metals are in<br />
the range 0.15-0.25 nm and these have a penetration depth in most metals<br />
of a few hundredths of a mm. Hence X-ray diffraction is suitable for<br />
measuring elastic distortions at the surface.<br />
PRINCIPLES OF MEASUREMENT<br />
We define a set of axes x, y and z in which z is normal to the<br />
specimen and x and y are aligned with some axes of symmetry in the specimen<br />
(Fig. 1). In practical terms we can measure the lattice spacing in<br />
any direction, m, within a cone 5O~60° from the specimen normal. We
- 374 -<br />
define the axes u, v and w where w is the projection of m into the specimen<br />
surface. The angles * and 0 define the orientation of m, u, v<br />
relative to x, y and z.<br />
The elastic strain along m is<br />
d d-d d-d<br />
n m _ in o „, m o<br />
G s j_n<br />
m d d d<br />
o o<br />
where: dm is the interplanar spacing along m,<br />
d0 is the interplanar spacing with no stress and<br />
the change from the exact value d0 to the approximate value<br />
d, causes only a small error in era.<br />
Then:<br />
em = ew(cos 2 ip) YUW cosipsini|/<br />
and, for material exhibiting isotroplc elasticity:<br />
uw<br />
uw<br />
G<br />
where: ew, eu, yuw are normal and shear strains in the u,<br />
v, w coordinate system,<br />
a u> a v> a w anc ' T uw are n 01 "" 13 ! an( l shear stresses<br />
in the u, v, w coordinate system<br />
E, v, G are Young's modulus, Poisson's ratio and the shear<br />
modulus.<br />
The X-rays normally sample only a thin surface layer in which<br />
stress and structure gradients are small. With no surface tractions<br />
= Tuw = 0<br />
Eq.<br />
Eq. 3<br />
Eq. 3a<br />
Eq. 3b<br />
Eq. 3c
and<br />
- 375 -<br />
e a -^(a + a ) + —-— .a .sin ty Eq.<br />
m E u v E u<br />
Referring to equation 2, au may be found from the slope of<br />
a plot of the measured lattice spacing dm versus sin"^. If the assumptions<br />
of isotropic elasticity and absence of large stress gradients in<br />
the measured volume are valid, such plots are linear (Fig. 2) and only<br />
two measurements are necessary to obtain ou. If two detectors are<br />
used, the two measurements can be made simultaneously, intercepting diffracted<br />
beams on each side of the incident beam within the u-w plane<br />
(Fig. 3(a)). If only one detector is used, the X-ray source and detector<br />
must be rotated by several 10 f s of degrees relative to the specimen<br />
surface to obtain lattice parameter measurements at two values of t|i<br />
(Fig. 3(b)). The two methods are called single exposure and double exposure<br />
techniques respectively. With either technique, measurements in<br />
three planes normal to the specimen surface (usually 0 = 0,^5 and 90°,<br />
Fig. 1) are sufficient to define fully the surface stress tensor.<br />
If the assumption of isotropic elasticity is invalid, or if<br />
stress gradients are sufficiently steep that the stress varies significantly<br />
within the layer penetrated by the X-ray beam, or if a triaxial<br />
stress distribution can be sustained within the surface layer, then plots<br />
of d vs. sin 2 i|i may no longer be linear. In this event, both the single<br />
and double exposure techniques may give incorrect results (3), and measure<br />
ments must be made for a number of values of ty at each value of 0 (multiple<br />
exposure technique). From such measurements, a triaxial stress<br />
tensor can be extracted, or corrections made for effects of anisotropic<br />
elasticity or depth gradients in stress (see references 2 and 3 for detailed<br />
treatments).<br />
For instruments in which both the X-ray source and specimen remain<br />
stationary during a single lattice parameter measurement and the<br />
X-ray line profile is sampled by a moving slit or fixed position sensitive<br />
detector, errors occur if the grain size is coarse or if steep gradients<br />
in crystallographic texture exist. Because of the reflecting<br />
geometry, the diffracted profile arises from a range of orientations of<br />
diffracting planes corresponding to half the range of 2e over which the<br />
profile is sampled (Fig. 4). If the number density of diffracting planes<br />
varies with orientation, the diffraction profile is affected. The effects<br />
of coarse grain size can be partially compensated by averaging the<br />
diffracted profile for a small range of incident beam orientations (3)<br />
but precise control of the incident beam direction combined with data<br />
manipulation are required to make a rigorous correction (4).<br />
The single exposure technique has an important advantage over<br />
double or multiple exposure techniques. Because the effective radius of
- 376 -<br />
the instrument is the same for both detectors, the error arising from<br />
misalignment of the specimen surface relative to the center of the<br />
instrument when the instrument is rotated is eliminated (5,6). This is<br />
particularly important for a portable instrument designed to examine<br />
large specimens, because the instrument rotates about the specimen and<br />
providing an accurate center of rotation can be difficult.<br />
ADVANCES IN INSTRUMENTATI<strong>ON</strong> FOR X-RAY STRESS MEASUREMENT<br />
The explosion in the development of solid state electronics in<br />
the past 15 years has resulted in many advances which make it possible<br />
for accurate measurements of stress to be made with a portable instrument.<br />
The new developments include inexpensive, compact but very powerful,<br />
computers, compact solid state high voltage supplies capable of<br />
delivering 100-200 W and stable, accurate electronic circuits for signal<br />
analysis. Very small X-ray tubes have been developed to complement the<br />
solid state power supplies (Fig. 5(a)).<br />
However, the most important advance has been the development of<br />
small position sensitive proportional counters (7). These measure the<br />
distribution of X-ray intensity in one dimension with a resolution of<br />
50-60 um. Detectors as small as 105 x 50 x 50 mm, with an active length<br />
of 50 mm (Fig. 5(b)) and weighing only = 0.6 kg are now commercially<br />
available. A position sensitive proportional counter consists of a fine<br />
wire centered in a chamber filled with Xe-lOJCHij or Ar-10?CHij (Fig.<br />
6(a)). For the counters shown in Fig. 5 the wire is a 25 ym diameter<br />
quartz filament, coated with carbon, which has a resistance of 8 kjj/mm.<br />
The wire is at a high positive voltage relative to the walls of the chamber.<br />
X-rays passing through the chamber ionize the gas causing electron<br />
cascades onto the wire. The voltage pulses observed at each end of the<br />
wire, arising from each cascade, have rise times proportional to RC,<br />
where R is the resistance from the end of the wire to the point of the<br />
electron cascade, and C is the capacitance between the end of the wire<br />
and ground (Fig. 6(b)). The two voltage pulses are shaped and amplified<br />
to produce bipolar pulses whose cross-over times are proportional to the<br />
rise times (Fig. 6(c)). Each cross-over time is marked by a very sharp<br />
voltage pulse, one of which is delayed a constant amount 6 (Fig. 6(d))<br />
so that it always follows the other. Thus, their time separation T is<br />
proportional to the position of the X-ray photon, x, in the counter.<br />
Electronic circuitry to perform this analysis is available commercially.<br />
DESIGN OF THE CANMET PORTABLE STRESS DIFFRACTOMETER<br />
These new technologies have been applied to design a versatile<br />
diffractometer capable of accurate stress measurement by either single<br />
or multiple exposure techniques. The CANMET diffractometer will be suitable<br />
for field use under clement conditions but with sufficient versatility<br />
and accuracy for dedicated laboratory use.
- 377 -<br />
The commercial prototype is expected to be ready for testing in<br />
December 1984. It consists of the head (Fig. 7), a stand, a 10 m umbilical<br />
cord and a services package containing analysis electronics, computer,<br />
X-ray power supply, closed circuit cooler, detector gas supply and<br />
motor controllers. The head will weigh ~15 kg, the simplest stand<br />
~10 kg, the umbilical cord ~12 kg and the services package ~90 kg.<br />
The instrument may be used with either one or two position sensitive<br />
proportional counters of 50 mm active length. A unique feature<br />
(H) is the precisely controlled angular rotation of the X-ray tube about<br />
the focal spot (a-rotation, Fig. 8(a)) combined with data manipulation<br />
to allow accurate determination of the Bragg angles for specimens with<br />
large grain size or pronounced gradients in crystallographic texture.<br />
For single exposure measurements, only the a-rotation is used. With<br />
a-rotation the diffraction peaks shift in the detectors, and hence the<br />
a-rotation can be used to calibrate the position sensitivity of the<br />
detectors.<br />
For multiple exposure measurements, the X-ray tube and detector<br />
(s) can be rotated simultaneously about the focal point (w-rotation,<br />
Fig. 8(b)). The uniformity and concentricity of this rotation must be<br />
precise, but a high degree of angular accuracy is not required. The a<br />
and ai rotations are coplanar and the stress is determined in the direction<br />
of the line of intersection between the plane of rotation and the<br />
specimen surface.<br />
Because the diffractometer is designed to measure stresses in<br />
structures as well as small specimens, the design incorporates stages to<br />
position the diffractometer relative to the specimen. The adjustment<br />
along the direction of the specimen normal is most critical and is incorporated<br />
into the diffractometer head (Fig. 7). A position sensor is<br />
provided to sense the position of the diffractometer in this direction,<br />
relative to the specimen surface, and allow accurate automatic positioning.<br />
The diffractometer head also rotates about the specimen normal<br />
(^-rotation) to allow stress measurement in any direction in the specimen<br />
surface.<br />
Translations in the plane of the specimen surface are provided<br />
to allow one or two dimensional mapping of stress. The stages providing<br />
these motions will be part of a laboratory stand.<br />
Table 1 summarizes the important features of the CANMET portable<br />
stress diffractometer.<br />
PERFORMANCE OF THE CANMET DIFFRACTOMETER<br />
To date, some aspects of the design have been evaluated with an<br />
engineering version of the diffractometer.
- 378 -<br />
It incorporates many of the features of the commercial prototype<br />
although it lacks the w-rotation. Relative accuracies of ±18-20 MPa<br />
(standard deviation) have been obtained on annealed ferritic steel and<br />
cold worked Inconel 600 calibration beams, for example Fig. 9. The<br />
counting times used to obtain this accuracy are ~150s with another 90s<br />
in executing the o-motion of the X-ray tube. With streamlining of the<br />
software and improvements in data reduction measurement, times for fine<br />
grained ferritic steels will be further reduced (possibly by a factor of<br />
2).<br />
In actual applications, the engineering version of the diffractometer<br />
gives results comparable with those obtained using a standard<br />
Phillips X-ray diffractometer modified for stress measurement (8), for<br />
example Fig. 10.<br />
SUMMARY<br />
The CANMET Protable Stress Diffractometer is designed for field<br />
and laboratory use. The design takes advantage of the error compensating<br />
feature of the single exposure technique, but still allows the investigation<br />
of stress distributions requiring multiple exposures. New detector<br />
and solid state electronics technologies have been applied to make<br />
it transportable for field use under clement conditions. It will be as<br />
accurate as standard laboratory equipment and have sufficient versatility<br />
for dedicated laboratory applications. A unique feature (the a-drive)<br />
allows the elimination of texture gradient and coarse grain size effects.<br />
A commercial prototype incorporating these features is currently under<br />
construction.<br />
ACKNOWLEDGEMENTS<br />
The CANMET Portable X-Ray Stress Diffractometer was conceived,<br />
and until his retirement January 198 1 !, developed, by CM. Mitchell of the<br />
Physical Metallurgy Research Laboratories (PMRL), CANMET, Department of<br />
Energy, Mines and Resources Canada. The engineering model of the instrument<br />
was designed and made by Proto Manufacturing Ltd., Oldcastle,<br />
Ontario, in 1983 with funding from CANMET. The commercial prototype of<br />
the diffractometer will be made by the same company within the framework<br />
of the Program for Industry Laboratory Projects (PILP). We wish to thank<br />
J.A. Hunter of NRC for his enthusiastic support in the role of PILP<br />
Project Manager and D.W.G. White of PMRL for useful discussions and help<br />
in other ways.
- 379 -<br />
REFERENCES<br />
1) Klug, H.P. and Alexander, L.E. X-ray Diffraction Procedures, John<br />
Wiley and Sons, New York (1954) p. 539<br />
2) Dolle, H. Journal of Applied Crystallography, V.12, 489 (1979).<br />
3) Maeder, G., Lebrun, J.L. and Spravel, J.M. N.D.T. International,<br />
235 (October 1978).<br />
4) Mitchell, CM. Proc. International Conference on Pipeline Inspection,<br />
Canadian Government Publishing Centre, Ottawa (1984).<br />
5) Mitchell, CM. Advances in X-ray Analysis, V. 20, 379 (1977).<br />
6) Ruud, CO. and Snoha, D.J. Journal of Metals, 33 (February 1984).<br />
7) Schelten, J. and Hendricks, R.W. Journal of Applied Crystallography,<br />
V. 11, 297 (1978).<br />
8) Winegar, J.E. Report No. AECL 6961, Atomic Energy of Canada Ltd.,<br />
Chalk River, Ontario, June 1980.
- 380 -<br />
Table 1 - Important Features of CANMET Portable Stress Diffractometer<br />
HARDWARE - uses 2 position sensitive proportional counters with analog<br />
timing electronics<br />
- small 200 W X-ray tube and compact solid state power supply<br />
- IBM-PC compatible microcomputer<br />
- for large components or structures and field use under clement<br />
conditions<br />
- transportable but with adequate versatility and accuracy for<br />
dedicated laboratory use*<br />
- capable of single or multiple exposure methods»<br />
- a-drive controls angle of incident X-ray beam*<br />
- (o-drive controls tilt for multiple exposure technique<br />
- positioning stages for manipulation of diffractometer head*<br />
SOFTWARE - complete software package for calibration, operation of diffractometer<br />
and data manipulation to eliminate texture gradient<br />
and grain size effects*<br />
•unique features of CANMET Portable Stress Diffractometer
- 381 -<br />
incident x-ray beam<br />
diffracted x-ray beam<br />
Fig. 1. Coordinate system for X-ray stress analysis. z is<br />
normal to the specimen surface and m is the direction<br />
of measurement of the lattice spacing. An angle of<br />
180°-29 separates the incident and diffracted beam.<br />
The measurement direction, m, falls on the bisector of<br />
this angle.<br />
0.804<br />
z 0.803<br />
a!<br />
0.802 -<br />
O.I 0.2 0.3<br />
SIN 2
- 382 -<br />
incident x-ray<br />
beam<br />
detector I incident<br />
'x-ray beam<br />
letector 2<br />
Fig. 3. Representations of the u-w plane of Fig. 1 showing the<br />
geometry of: A - the single exposure technique and B -<br />
the double exposure technique for the measurement of<br />
stress along the "u" direction using lattice parameter<br />
measurements m, and m 2'<br />
incident x-ray<br />
beam<br />
Fig. 4. Representation of the u~w plane of Fig. 1 showing three<br />
diffracted beams, d.. , d and d~, for a given incident<br />
beam direction. That corresponding to the centre of the<br />
diffraction profile, d„, arises from lattice planes<br />
normal to m„ while those corresponding to the high and<br />
low angle tails of the profile, d. and d,, arise from<br />
lattice planes with normal m. and m., respectively.<br />
Because of the reflecting geometry, the angular separation<br />
of the diffracted beams is twice that of the plane<br />
normals.<br />
m,
- 383 -<br />
Fig. 5(a). 200 W X-ray tube manufactured by KEVEX<br />
Fig. 5(b). Small, commercially available position sensitive<br />
proportional counters manufactured by Technology for<br />
Energy Corporation (TEC) and M. Braun GmbH.
- 38A -<br />
VOLTAGE L S T S<br />
Y Y Y<br />
s<br />
TIME<br />
Fig. 6(a). Schematic diagram of a position sensitive proportional<br />
counter.<br />
Fig. 6(b). Voltage pulses produced at the ends of the counter<br />
wires by an X-ray absorbed by the counter gas at X.<br />
Fig. 6(c). Shaped voltage pulses with cross-over times proportional<br />
to rise times of the initial pulses.<br />
Fig. 6(d). Sharp, pulses marking the cross-over times. The pulse<br />
at S is delayed by an amount S to S' so that the<br />
separation T is proportional to X.
- 385 -<br />
Fig. 7. Head of the commercial prototype of the CANMET Portable<br />
Stress diffractometer. It includes tube, T, collimator,<br />
C, two detectors, D.. , and D~, a and u) rotations about<br />
the axis F-F^ in the specimen surface, translation<br />
along the specimen normal, z, and rotation about the<br />
specimen normal, .<br />
Fig. 8 a-rotation. The X-ray tube, T, rotates about the focal<br />
point F, while the angular positions of the detectors<br />
Dl and D2 remain fixed. w-rotation. The X-ray tube and<br />
detectors rotate simultaneously about the focal point.
- 386 -<br />
APPLIED STRAIN<br />
Fig. 9. Elastic strain (measured using X-ray diffraction) as a<br />
function of applied strain (measured using strain<br />
gauges) in a flat beam of Inconel 600 stressed in four<br />
point bending. Note the large compressive residual<br />
stress in the beam (indicated by the negative intercept<br />
with an applied strain of zero).<br />
MEASUREMENT POSITI<strong>ON</strong>S<br />
Fig. 10. Longitudinal residual stress as a function of circumference<br />
(relative to the tube axis) on the outside<br />
surface of "u" bend in an Inconel 600 nuclear steam<br />
generator tube. Measurement position 1 is at the<br />
inside diameter of the bend, measurement, 15 at the<br />
outside diameter.
- 387 -<br />
APPLICATI<strong>ON</strong>S OF NEUTR<strong>ON</strong> DIFFRACTI<strong>ON</strong> TO ENGINEERING PROBLEMS<br />
T.M. Holdan, G. Veiling, S.U. MacEwan, J. Wlne.gai and B.M. Pouie.ZZ<br />
Atomic. Ena-igy o () Canada Limited<br />
Chalk Rl\mK, Ontaulo<br />
R.A. Holt<br />
Physical Me.taZlu.igy Rziaafich Laboiatonle.!,<br />
Ottauia, Ontailo<br />
ABSTRACT<br />
Neutron diffraction may be applied with advantage to the study of engineering<br />
problems in areas currently served by X-ray diffraction. At Chalk River, neutron<br />
diffraction has been applied to the measurement of lattice strains in an<br />
over rolled Zr 2.5 wt% Nb pressure tube and in a bent Incoloy-800 steam generator<br />
tube. Basically, the interplanar lattice spacing, deduced from the diffraction<br />
peak, acts as an internal strain gauge. Our experiments on the pressure<br />
tube showed that a major contribution to the lattice strain comes from<br />
grain to grain interactions and allowed us to measure the strain tensor associated<br />
with the overrolling procedure. The experiments on the generator tubing<br />
showed that, even in a cubic material, the strain in the longitudinal direction<br />
is higher by a factor of two for grains aligned with an [002] crystallographic<br />
axis longitudinal than for grains aligned with a [111] axis longitudinal. The<br />
differences originate in the elastic anisotropy of Incoloy-800. A number of<br />
other potential applications of neutron diffraction are discussed.<br />
1. INTRODUCTI<strong>ON</strong><br />
Neutron diffraction can be applied to the study of engineering problems in many<br />
areas currently served by X-ray techniques. The advantages of neutron<br />
diffraction over X-ray diffraction originate in the low neutron absorption and<br />
scattering cross sections of most elements. For example a 1 Â thermal neutron<br />
beam is only attenuated by a factor of 3, 1.3 and 1.1 by a 1 cm thickness<br />
respectively of steel, zirconium and aluminum. Hence one is usually able to<br />
measure peaks diffracted by atomic planes whose normals lie along any chosen<br />
direction in the component. With this flexibility one can examine, for<br />
example, the hoop strain in a tube by passing the beam through one wall and<br />
scattering off the second wall oriented so that the tangent to the tube bisects<br />
the incident and scattered beams. With a thin-walled tube, the property (be it<br />
texture, or strain) is averaged through the wall with neutrons whereas X-rays<br />
measure the property in the first few microns from the surface. With a carefully<br />
collimated beam of neutrons one can, by moving the component through the<br />
point of intersection of the incident and scattered beam, map out the properties<br />
through the bulk [1].
- 388 -<br />
The disadvantages of neutron diffraction are few but significant. Sufficiently<br />
intense neutron beams are available only from nuclear fission reactors or from<br />
spallation reactions in the target of a charged particle accelerator; neutron<br />
beams and neutron spectrometers are both costly and few in number throughout<br />
the world. With the NRU reactor at Chalk River and its associated neutron<br />
spectrometers, we have equipment in Canada that is as good as any in the world<br />
for these kinds of experiments.<br />
The paper is arranged as follows. The basis of the neutron technique is<br />
described in section II, while section III contains two examples of recent<br />
neutron experiments at Chalk River, viz. the measurement of residual strain in<br />
overrolled Zr 2.5 wt% Nb pressure tubes and in bent Incoloy-800 steam generator<br />
tube. A number of other applications are mentioned.<br />
2. NEUTR<strong>ON</strong> DIFFRACTI<strong>ON</strong> TECHNIQUES<br />
A typical neutron spectrometer mounted at a fission reactor can select a beam<br />
of neutrons with any desired wavelength in the range » 1-5 Â, that can be<br />
directed onto the samples. Figure 1 shows such a spectrometer mounted at the<br />
L3 beam facility of the NRU reactor, Chalk River. Typically about 10 neutrons<br />
cm" s~ impinge on the sample, and the intensity of powder diffraction is of<br />
the order 10-100 counts/sec when the diffracting volume is of order one cm .<br />
The spectrometer, because of stringent shielding requirements, weighs several<br />
tons and can accommodate samples of up to about a ton on the sample table. The<br />
maximum cross-sectional area of the neutron beam is 50 mm * 50 mm; smaller<br />
sizes are obtained by suitable collimators that cut down the beam with absorbing<br />
masks. The neutron beam size in any experiment is chosen to match the<br />
spatial scale of the feature under investigation. For example in the experiment<br />
on Incoloy-800 steam generator tubing there were marked changes in strain<br />
in one dimension over a distance of order 1 mm, so the masks were chosen to be<br />
1 mm wide and 10 mm long. The angular resolution is easily altered to suit the<br />
demands of the measurement, by appropriate choice of wavelength and collimation<br />
of the incident and diffracted beams. For crystallographic texture measurements,<br />
low resolution is required to integrate over the appropriate Bragg<br />
peaks; on the other hand, the highest available resolution, of order 0.01% is<br />
often desirable for residual strain measurements. Typically the time taken to<br />
characterize the texture for a particular set of crystallographic planes is<br />
about a day, whereas the determination of the angle of diffraction 29j,kji in a -<br />
strain measurement takes a few hours. The interplanar spacings of the crystal<br />
lattice constitute miniature internal strain gauges. The angle of diffraction<br />
is related to the corresponding spacing, dhkJi, by Bragg's law.<br />
X = 2
peaks which are easily measurable.<br />
3. APPLICATI<strong>ON</strong>S<br />
- 389 -<br />
3.1 Lattice strain measurements in overrolled cold-worked Zr 2.5 wt% Nb<br />
pressure tubes<br />
Several measurements of lattice strains in model engineering components by neutron<br />
diffraction have recently been reported. [1,2]. Our experiments were<br />
initiated in response to a pressing engineering problem: the understanding of<br />
the strains that developed in overrolled présure tubes in CANDU reactors in<br />
Ontario. In a CANDU reactor, the seal between the zirconium alloy tube and its<br />
steel end-fitting is a rolled joint, made by inserting a rolling tool and<br />
expanding the tube into the end-fitting. The mouth of the fitting was slightly<br />
tapered for easy location and insertion of the tube in the fitting. In many<br />
cases the rolling tool was inserted too far, so that the tube was expanded into<br />
the taper, giving rise to a large circumferential residual tensile stress [3].<br />
Under operating conditions, hydrogen diffused to the regions of high tensile<br />
stress. Platelets of zirconium hydride were formed perpendicular to the<br />
residual tension, and cracking eventually occurred. We set out to map the<br />
lattice strain in the overrolled region with the neutron diffraction technique.<br />
Figure 2 shows measurements of lattice strain in the hoop direction in a complete<br />
overrolled tube as a function of distance from a point in the tube within<br />
the end fitting (0 mm) through the point of maximum Insertion of the rolling<br />
tool (17.5 mm) to a region characteristic of the production tube. The lattice<br />
strain is determined with respect to the interplanar spacing at the free<br />
"unstrained" end and corresponds to the incremental elastic strain induced by<br />
rolling. Well within the end-fitting, where the tool presses the tube against<br />
the fitting the lattice is compressed, as one would expect; however, in the<br />
unsupported region there is a sharp increase In the (0002) interplanar spacing<br />
corresponding to the hoop direction. This tensile strain extends beyond the<br />
insertion point of the tool. A coupon was subsequently cut from the complete<br />
overrolled tube for X-ray line broadening measurements and Fig. 2 also shows<br />
the lattice strain In the hoop direction of this coupon. Although the macroscopic<br />
hoop stresses no longer exist, as much as 50% of the original lattice<br />
strain Is still present In the coupon. We believe that this phenomenon arises<br />
from grain-to-grain interactions. One conventional method of measuring<br />
residual stress involves observing the change in strain gauge readings when<br />
such a coupon Is cut from a tube, and assuming that no stresses remain in the<br />
coupon. This approach would clearly fail to reveal the effects we have<br />
observed with neutron diffraction. Also shown In Fig. 2 is the lattice strain<br />
profile for a complete tube that was given a stress-relief heat treatment after<br />
the rolled joint was made.<br />
Measurements of the anlsotropy of the residual strain were made with this<br />
coupon and some of these are shown in Fig. 3 for the plane in the coupon
- 390 -<br />
containing the tube radius (X^) and the hoop direction (X3). At the place in<br />
the coupon where the (0002) plane spacing in the hoop direction shows residual<br />
tensile strain, the *(2ffO) plane spacing in the radial direction shows a<br />
compressive strain. The third major direction X2 (axial, corresponding to<br />
(1010)) shows a small residual tensile strain- Analysis of 19 interplanar<br />
spacings in three orthogonal planes X1X2, X2X3 and X3X1 gives the complete<br />
strain tensor including sizeable shear terms.<br />
To make this approximate one-to-one correspondence between plane normals and<br />
direction in the sample, we have made use of the marked texture of a production<br />
pressure tube. The vast majority of grains have their [01Î02 axes along the<br />
tube axis, the [0002] axes along a hoop direction and the [2110] axes along a<br />
radial direction. This is the first time such a detailed mapping of the strain<br />
has been carried out and it should lead to a realistic model of the distortion<br />
caused by overrolling.<br />
3.2 Residual stresses in Incoloy-800 steam generator tubes<br />
The steam generator tubes in CANDU nuclear generating stations are formed into<br />
hairpins. Residual stresses in a given tube after it has been bent to its<br />
design shape are undesirable because they make the tube prone to stresscorrosion<br />
cracking when a concentration of salts in areas of localized dryout<br />
coincides with high tensile stress.<br />
The spacings of the {ill} and {200} planes in the longitudinal direction of the<br />
tube were measured as a function of position around the circumference. The<br />
"relaxed" value of plane spacing was measured on a piece of straight tubing.<br />
The strains derived from {220}, {200} and {ill} planes in the radial and hoop<br />
directions were also measured as a function of circumferential position.<br />
Figure 4 shows a schematic view of the experimental arrangement in which the<br />
neutron beam beam passes first through the near wall of the tube but the<br />
incident and scattered beams only intersect in the far wall which in this case<br />
is oriented so that its axial direction lies along the scattering vector. The<br />
neutron counter is shielded so that it cannot see the scattering from the near<br />
wall. This kind of arrangement is very useful for studying tubes nondestructively.<br />
The results are shown in Fig. 5. The longitudinal (or axial)<br />
lattice strain shows a maximum just above the neutral plane of the bent tube<br />
and a minimum just below it. The maximum longitudinal strain derived from<br />
{200} planes is almost twice that derived from {ill} planes. The magnitude of<br />
the strains within the wall came as a surprise, since conventional X-ray<br />
methods [4] had indicated lower strains. The fact that the strain in the<br />
longitudinal direction could be different in differently aligned grains of a<br />
cubic material was also a surprise. Model calculations for a thin walled tube<br />
with elastic/plastic bending [5], following by elastic unloading, confirm the<br />
shape of the residual strain curve. They suggest, moreover, that the anisotropy<br />
of Young's modulus in Incoloy accounts for the difference between the {ill}<br />
and {200} strains. The Incoloy tubing is textured such that there are 5 times<br />
more {ill} planes oriented along the tube axis than {002} planes. The {ill}<br />
behaviour, showing a balanced bending moment, is more representative of the<br />
tube. However, the texture is not so strong that we can determine the complete<br />
strain tensor. The variation of lattice strain in both the hoop and radial
- 391 -<br />
directions around the circumference is the reverse of the longitudinal strain<br />
and about 25% of its value. It appears that the relationship between the<br />
strains,<br />
given by<br />
EL, ER, £H in the longitudinal, radial and hoop directions is<br />
£ R<br />
V V<br />
where v is Poisson's ratio, which for Incoloy-800 lies between 0.28 and 0.34<br />
[6] . If we assume that the strains are uniaxial the residual stress may be<br />
obtained by making use of Young's moduli derived from X--ray measurements. The<br />
residual stresses measured with neutrons are roughly 1.5 times those derived<br />
from X-ray diffraction on the same tube. This may indicate major gradients<br />
through the wall or surface relaxation effects. It is clear from these<br />
measurements that neutrons are an excellent diagnostic tool for measuring the<br />
residual elastic strain in engineering components, capable of giving Important<br />
insights Into the mechanical behaviour of deformed metals.<br />
3.3 Texture measurements<br />
Measurements of crystallographic texture by neutron diffraction are not particularly<br />
new. For a recent review, see Bunge [7] . However, they have major<br />
advantages since the number of grains sampled by a neutron beam Is much larger<br />
than with typical X-ray beams. There are few geometrical constraints for most<br />
materials and it is often possible to place an entire component in the neutron<br />
beam.<br />
3.4 Measurement of second phases in materials<br />
The existence of second phases often controls the strength and fracture<br />
behaviour of materials. The sensitivity of neutrons to second phase material<br />
is often higher than X-rays partly because the signal is an average over many<br />
grains, and partly because the light element constituents of many precipitates,<br />
such as carbon, nitrogen and oxygen, have scattering cross sections that are<br />
just as large as most metals or other materials of interest to the engineer.<br />
Detection of volume fractions less than 1% is fairly routine. As an example of<br />
a recent application of neutron diffraction, Windsor et al. [8] have studied<br />
the development of fee second phase material in bec maraging steels.<br />
3.5 Small angle scattering of neutrons<br />
Small angle scattering relies on an effective contrast in scattering between<br />
the matrix and the embedded "small particles" that are the objects of interest.<br />
The sizes of small particles in the range 10-1000 Â may be readily determined;<br />
they may be lattice distortions around impurity atoms, nucleating phases barely<br />
detectable by other means, small precipitates etc. There is already a large<br />
literature on the technological applications of small angle scattering (see for<br />
example ref. [9]).
4. THE FUTURE<br />
- 392 -<br />
The neutron techniques are straightforward and well understood. The<br />
penetrating nature of neutron beams allows examination of real components from<br />
all angles, and the scale of neutron equipment is commensurate with the size<br />
and weight of large components such as pipeline and rail steel. Our first<br />
experiments have been on topics of interest to the nuclear industry, but the<br />
extremely detailed information that we have obtained suggest a much wider<br />
domain of usefulness in the future.<br />
5. REFERENCES<br />
[1] L. Pintschovius, V. Jung, E. Macherauch and 0. Vöhringer, Materials<br />
Science and Engineering, jjl_, 43 (1983).<br />
[2] M.J. Schmank and A.B. Kravitz, Met. Trans. A., 134, 1069 (1982).<br />
[3] E.C.W. Perryman, Nucl. Energy, J7, 95 (1978).<br />
[4] J. Winegar, AECL Report 6961 (1980).<br />
[5] S.P. Timoshenko and J.M. Gere, Mechanics of Materials, New York, Van<br />
Nostrand-Reinhold, p. 113 (1970).<br />
[6] R.A. Holt, G. Dolling, B.M. Powell, T.M. Holden and J.E. Winegar. Proceedings<br />
of the 5th International Symposium on Metallurgy and Materials<br />
Science, Riso, Denmark, 3-7 Sept. 1984 (in press).<br />
[7] H.-J. Bunge, Texture Analysis in Materials Science, London,<br />
Butterworths, 1982.<br />
[8] CG. Windsor, R.N. Sinclair, S. Faulkner, V.S. Rainey and G.F. Slattery.<br />
Proceedings of the 5th International Symposium on Metallargy and<br />
Materials Science, Riso, Denmark, 3-7 Sept. 1984 (in press).<br />
[9] W. Schmatz, in The Neutron and Its Applications, ed. P. Schofield, The<br />
Institute of Physics (London, 1982) , p. 301; these conference proceedings<br />
contain several other papers dealing with special applications of<br />
small angle neutron scattering.
I.-3 TRIPLE-AXIS NEUTR<strong>ON</strong> SPECTROMETER<br />
LIQUID NITROGEN COOLED<br />
• CALANDRIA<br />
VNEUTR<strong>ON</strong> FILTER .-AUXILIARY GATE<br />
RREACTOR HOLE GATE R QUARTZ<br />
1<br />
r LIQUID NITROGEN<br />
••<br />
-COLLIMATOR K<br />
RM<strong>ON</strong>OCHR0MAT0R<br />
\<br />
CRYSTAL<br />
J COLLIMATOR ;<br />
r-SHIELDING DRUM<br />
!! r M<strong>ON</strong>ITOR<br />
SHIELDING YOKE<br />
NEUTR<strong>ON</strong> BEAM<br />
Fig. 1 Neutron spectrometer at the NRU reactor, Chalk River.<br />
NEUTR<strong>ON</strong> DFTECTOR<br />
' SPECIMEN<br />
I
I 2<br />
- 394 -<br />
AXIAL DISTANCE (CM)<br />
OVER-ROLLED PRESSURE TUBE<br />
STRESS RELIEVED TUBE<br />
COUP<strong>ON</strong> FROM PRESSURE TUBE<br />
' "**..<br />
O •-<br />
Fig. 2 Measured lattice strain in the hoop direction derived from the<br />
interplanar spacing of (0002) planes in Zr 2.5 wt% Nb pressure tubes.<br />
The dashed line indicates the result for the over-rolled pressure<br />
tube, the solid line indicates the result for the coupon cut from the<br />
tube, and the dotted curve indicates the result for the stress<br />
relieved tube.
Ad hki£<br />
d hkil<br />
xlO 3<br />
- 395 -<br />
INTERNAL STRAIN IN Zr"2.5wt%Nb PRESSURE TUBE<br />
X( X3 plane<br />
3 4<br />
AXIAL DISTANCE (CM)<br />
0002<br />
SEC<strong>ON</strong>D GRAIN<br />
Fig. 3 Measured anisotropy of lattice strain in the coupon cut from the<br />
overrolled Zr 2.5 wt% Nb pressure tube. A tensile strain is observed<br />
in the hoop direction X3, (0002) planes, and a compressive strain in<br />
the radial direction Xj, (2ÏÏ0) planes. A very small compressive<br />
strain is observed for intermediate directions corresponding to (2ÏÏ2)<br />
and other related planes.
IRRADIATED<br />
VOLUME<br />
- 396 -<br />
BEAM APERTURE<br />
ANGLE OF<br />
SCATTERING<br />
20<br />
Fig. 4 Arrangement of steam generator tube in the neutron beam for diffraction<br />
from (200) planes in one wall of the tube, directed along the<br />
bisector of the incident and scattered beams.<br />
Cd
- 397 -<br />
Fig. 5 Longitudinal, tangential (hoop) and radial strains measured around the<br />
circumference (to - 0 corresponds to the neutral plane) of an<br />
Incoloy-800 steam generator tube at the apex of the bend. The strains<br />
in differently oriented grains but directed along the same axis in the<br />
tube are markedly different, showing the effects of the elastic<br />
anisotropy.<br />
120
- 398 -<br />
NDE IN POLYMERS, AN EXAMPLE:<br />
ULTRAS<strong>ON</strong>ICS FOR THE DETERMINATI<strong>ON</strong> OF DENSITY IN POLYETHYLENE<br />
L. Vichu and A. Hame.1<br />
NatZonai. Re.i zafich* Co an.c-lt o & Canada<br />
Bouche.ivA.Zle., Quebec<br />
Abstract<br />
An ultrasonic immersion method is presented for the rapid and accurate<br />
measurement of longitudinal sound velocity (v) in polyethylene (PE). The<br />
method is based on the sole measurement of time delays and can be automated.<br />
The samples are the same press molded plaques that are prepared for the<br />
standard density (p) determination using a density gradient column. Our<br />
measurements on a representative group of various types of PE with densities<br />
in the range .92 to .96 g/cm show without ambiguity that velocity is<br />
proportional to density. The degree of correlation is such that we come to<br />
suggest that the acoustic method be used for the characterization of density.<br />
We show that the accuracy is good and compares with that of the density<br />
gradient column and then we conclude on the practical advantages of the<br />
ultrasonic method.
I - INTRODUCTI<strong>ON</strong><br />
- 399 -<br />
Polyethylene, be it low density (LDPE) or high density (HDPE) is one of the<br />
most important, if not the most important product of the plastics industry.<br />
As it is the case for all industrial materials, it is of primary importance<br />
for the manufacturer and the processor to- have means of characterizing the<br />
material.<br />
From a somewhat simplistic point of view, polyethylene can be considered as a<br />
binary mixture of an amorphous and a crystal phase and the degree of<br />
crystallinity will affect the elastic properties (elastic modulus, impact<br />
strength and melting point). On the other hand, it is known (1) that density<br />
is directly related to the degree of crystallinity: so that in the<br />
relationship between the basic parameters of PE and its end properties, the<br />
value of density is an important link. This explains why the value of density<br />
has always been the major criterion for the characterization of polyethylene.<br />
The most commonly used and accepted method for the determination of density is<br />
that of the "Density Gradient Column" defined by ASTM D1505 standard (2). The<br />
sample is placed in a column of liquid which exhibits a known density gradient<br />
and after it has sunk to its equilibrium, the density is obtained from the<br />
reading of its position in the column.<br />
As such, the procedure appears to be simple but it is in fact quite involved.<br />
First, the column must be prepared from a mixture of different solutions of<br />
known densities. This is time consuming, requires much care at all stages and<br />
is much an art. The column which is placed in a constant temperature<br />
environment is then calibrated, using standard glass floats (the calibration<br />
has to be verified periodically). To perform the measurement, the samples are<br />
cut to size, cleaned, wetted down and gently introduced in the column where<br />
they settle in roughly 20 minutes. When, after a number of measurements, the<br />
number of samples in the column is too important, these are retrieved<br />
carefully with a basket, without upsetting the gradient.<br />
Here we propose a method and technique based on the propagation of ultrasound<br />
which allows for an accurate and reliable determination of density. The<br />
technique is simple and can be used in an uncontrolled environment; the<br />
measurement is quick and can be made automatic.<br />
II - EXPERIMENTAL METHOD<br />
For a number of polymers, the ultrasonic data are available from the<br />
literature (3); sparse results are found for the sound velocity and density<br />
(4,5). The special case of polyethylene has, however, been given closer<br />
attention (6,7,8,9) and the various results tend to show that there exists a<br />
correlation between sound velocity (v) and density (p).<br />
The most complete and conclusive study is that of Davidse et al (7). They<br />
have measured, using a resonance method, the propagation velocity of low<br />
frequency oscillations (5 to 10 kHz) in bars of PE and their results on a
- 400 -<br />
great number of samples show that the velocity is proportional to the<br />
density. The technique they have used is involved: it requires preparation of<br />
samples and the determination of velocity takes into account geometrical<br />
effects and a priori knowledge of the Poisson ratio for the material.<br />
It is our objective here to verify the existence of a possible correlation<br />
between the velocity and the density, to describe the correlation and to<br />
propose a technique by which the velocity can be obtained in a rapid yet<br />
precise way so as to allow for the determination of density.<br />
In our study, we use an immersion technique (4,10) because of the ease with<br />
which the measurements can be made and repeated. The principle of the method<br />
where a single transducer is used as both the emitter and the receiver is<br />
illustrated in Fig. 1. A high frequency (MHz) sound pulse travels with a<br />
normal incidence onto the sample (a plate of thickness d). There it sees part<br />
of its energy reflected back (E^) and reaching the transducer at time t\, and<br />
part of it going through the sample where it later meets the second interface:<br />
again part of the energy is reflected, part is transmitted and so on. This<br />
gives rise to the formation of échos (Ej, E2 ...) which are seen at times tj +<br />
T, ti + 2T...(T is the time for the sound to travel back and forth in the<br />
sample). If v is the velocity of sound in the sample, then<br />
T = 2d/v<br />
In order to obtain v, one needs to know the value of d with sufficient<br />
precision so as to obtain the required accuracy. This measurement could be<br />
made using a micrometer but this would be both annoying and not very<br />
accurate. The samples being rough from press show surface and thickness<br />
irregularities so that the value of d must be some average taken at the<br />
location of insonification. We have chosen to take the thickness into<br />
account through acoustics. Part of the energy after having traversed the<br />
sample is transmitted forward in the liquid. This signal can be reflected<br />
back to the transducer by use of an acoustic reflector M. If t'i is the<br />
time it takes the sound pulse to go from the sample to the mirror and back,<br />
a signal will be detectecd at time<br />
t'=ti+T+t'i=ti+ 2d/v + t'i<br />
The sample being removed, the signal coming from the mirror appears at time<br />
t" = t! + 2d/c + t 1 !<br />
where c is the velocity of sound in the liquid.<br />
Considering the difference in the times of flight yields<br />
t' - t" = At = 2d (1/c - 1/v)
- 401 -<br />
This together with the expression of T gives<br />
v = c (At/T + 1)<br />
Thus, from the sole measurement of time delays, and the knowledge of c, one<br />
obtains the value of the velocity of sound in the sample.<br />
Ill - EXPERIMENTAL PROCEDURE AND RESULTS<br />
The samples (35) corresponding to a broad range of densities were furnished<br />
by DuPont Canada and Union Carbide Canada Ltd. They were in the form of press<br />
molded plaques (80 X 50 X 1.9 mm) prepared according to ASTM standards from a<br />
variety of commercial grades of P.E. In the measurements, they were used as<br />
is with no preparation. A number of pieces (2 to 4) were cut from the plaques<br />
and their density measured in a density gradient column following the standard<br />
ASTM procedure. The values are given with an accuracy of +_ .0007 g/cm at a<br />
constant temperature t = 23°C.<br />
The ultrasonic set up was designed to accomodate such samples and placed in<br />
a bath of demineralized water at 23°C. The transducer is of the<br />
commercially available immersion type. In order for the echos to be well<br />
separated in time, the pulses must be short: this calls for a highly<br />
damped transducer operating at high frequency. However, the attenuation of<br />
sound in P.E. rapidly increases with frequency and this sets an upper limit.<br />
We found that 3 MHz constitutes a good compromise between time resolution and<br />
signal amplitude. The reflector, located roughly 15 cm from the transducer,<br />
is a flat piece of glass with an absorbant backing so that no signals (echos)<br />
are issued other that the reflexion from the front face. The set up is rather<br />
simple and the only requirement is that the samples and reflector be<br />
perpendicular to the sound beam.<br />
Concerning the electronics, we used a Metrotec puiser (Mod. MP 203) capable of<br />
delivering short (100 n sec) high voltage (300 volts) pulses with a repetition<br />
rate or 50 to 100 Hz. The receiving amplifier was also Metrotec (Mod. MR 101)<br />
wiht a 60 dB gain factor and a 20 MHz bandwidth. The RF signal is monitored<br />
on an oscilloscope (H.P. Mod. 1743A) which is equipped with a dual time base<br />
that allows time delay measurements with an adequate resolution (3 n sec or<br />
better).<br />
The sound velocity in the water has to measured: this is accomplished by<br />
displacing the reflector by a known distance and noting the arrival times. We<br />
obtain a value of c = (1.483 +_ 0.001) X 10 5 cm/sec at 23°C, in good agreement<br />
with what is found in the literature from which we quote the temperature<br />
dependence: 200 cm/sec°C.<br />
The results for our measurements of sound velocity (at 3 MHz) in<br />
polyethylene samples of different densities are illustrated in Fig. 2. To<br />
a good approximation, the velocity is proportional to density.
- 402 -<br />
Since the velocity measured is actually a value taken over the whole region of<br />
insonification, we looked at the effect of changing the acoustic beam<br />
diameter. Our observations using transducers of increasingly larger diameters<br />
(from 0.2 cm for à focused beam up to 4 cm) show that the scatter diminishes.<br />
Increasing the diameter amounts to measuring a value that is averaged over a<br />
wider area. In order to appraise the scatter, we have measured the velocity<br />
at different locations on the same plaques. The results are illustrated in<br />
Fig. 3 for a 1.25 cm (i in) diameter transducer. The standard deviation is<br />
found to be (Sv/v)expe. = _+ 0.0018. However, if we repeat the measurement<br />
several times at the same location on the plaque, the standard deviation is<br />
(6v/v)repeat = _+ 0.0010. This tends to show that the scatter partly has its<br />
origin in non homogeneities found across the plaque.<br />
In turn, we can estimate the theoretical experimental uncertainty by<br />
considering the relation from which the velocity is calculated and the various<br />
instrumental errors. We obtain a value of (ôv/v)theo. = i 0.0012, which is<br />
to be compared with (
- 403 -<br />
(5v/v)fit must be considered as more realistic since it integrates errors<br />
from all sources.<br />
From a statistical point of view, the fit can be considered as very good and<br />
this leads us to the conclusion that for PE having a density in the range of<br />
0.92 to 0.97 g/cm , the room temperature (23°C), sound velocity is<br />
proportional to density. We may now examine the possibility of using the<br />
measurement of SDund velocity for the determination of density in PE with a<br />
model such as p = Av + B. Working the regression analysis with p as the<br />
dependent variable we obtain A = (0.0965 +_ 0.0012)10~ ? g sec/cm 4 and B =<br />
(0.07218 +_ 0.0026) g/cm . The correlation factor is of course the same as<br />
above (r = 0.998) and the standard error of estimate becomes 0.00098 g/cm<br />
yielding the relative average value (ôp/p) = _+ 0.0010. As mentioned above<br />
this integrates errors from all sources, and yet it compares quite well with<br />
what can be expected when using a density gradient column for which<br />
(6p/p)co^# = _+ 0.0007. Given this result and bearing in mind that the small<br />
difference most probably originates in the additive content, we believe that<br />
the ultrasonic approach will allow the determination of density in<br />
polyethylene with the required accuracy for quality control purposes.<br />
IV - C<strong>ON</strong>CLUSI<strong>ON</strong><br />
We have demonstrated that in commercial grades of polyethylene, there exists a<br />
linear relationship between the sound velocity at ultrasonic frequencies and<br />
the density of the material. We have shown experimentally that the velocity<br />
measurements can be performed with sufficient accuracy to allow the<br />
computation of density. The method we describe is simple and since it is<br />
based on the sole measurement of time delays, it can be made automatic through<br />
digitalization. The advantages of the method are numerous when compared to<br />
the classical density gradient column: a) no preliminary preparation, the<br />
setup is completely portable, suffers little influence from the environment<br />
and requires no special care to upkeep, b) the sample is not destructed and<br />
can be kept for files, c) the measurement is fast, reliable and repeatable,<br />
the accuracy is good (ôp = +^ 0.001 g/cm ) and the operation being simple<br />
requires no special attention, care nor skill.<br />
Further work is needed to check whether such a technique could be applicable<br />
to other polymers and also to examine the influence of fillers (silica, carbon<br />
black...). Further work is also needed in order to establish the physical<br />
foundation for the observed behaviour.<br />
ACKNOWLEDGMENT<br />
The author wishes to acknowledge the gracious participation of Dr. P. Kelly<br />
(DuPont Canada Inc, Kingston, Ontario) and Mr. I.R. Lambie (Union Carbide,<br />
Sarnia, Ontario), who provided both the incentives and the samples for this<br />
work.
REFERENCES<br />
- 404 -<br />
1. S. Kavesh, J.M. Shultz; Polymer eng. Sei. _9> 331 (1969)<br />
2. Annual book of ASTM Standards<br />
3. For a revue, see: Hartmann B., Methods of Experimental Physics,<br />
vol. 16c; R.A. Fava Ed.; Academic Press Inc., 1980.<br />
4. B. Hartmann, J. Jarzianski, J. Acous. Soc. Am _56_, 1469 (1974).<br />
5. Biing-Nan Hung, Goldstein A., IEEE Trans-sonics, Ultrasonics Su3£, 4<br />
(1983).<br />
6. J. Schuyer; Jour. Poly. Seien., J36_, 475 (1959).<br />
7. P.D. Davidse, H.I. Watermann, J.B. Westerdijk; Jour. Poiy. Seien. _59,<br />
389 (1962).<br />
8 A. Levene, W.J. Pullen, J. Roberts; Journ. Poly. Seien., A3, 687 (1965).<br />
9. Adachi K., Harrison G., Lamb J., North A.M., Pethrick R.A,: Polymer _22,<br />
1032 (1981).<br />
10 G.W. Paddison, Ultrasonics Symposium, IEEE, p. 502, (1979).<br />
Patent Pending
Transducer<br />
Signal<br />
- 405 -<br />
P.E.(p.v)<br />
M<br />
Immersion Liquid: Water, (c)<br />
T<br />
Reflector<br />
t" time<br />
Figure 1: a) Schematic description of the path followed by the sound beam as<br />
it travels through the water and interacts with the sample; b)<br />
representation of the different signals which are to be observed on<br />
an oscilloscope.
sec.)<br />
2.5<br />
F 2.4<br />
b<br />
n O<br />
Î2-3<br />
S CO<br />
f 2.2<br />
Longi<br />
—<br />
—<br />
I I<br />
2.1 - •/<br />
- 406 -<br />
I I I<br />
2<br />
.91<br />
/<br />
i<br />
.92<br />
i<br />
.93 .94 .95<br />
Density (g/cm 3 )<br />
Vv<br />
/ ;<br />
I I I<br />
—<br />
.96 .97<br />
Figure 2: Results obtained for the measurement of ultrasonic (3 MHz) sound<br />
velocity in plaque samples of commercial grade polyethylene, the<br />
density of which was measured in a density gradient column at 23°C.
ü<br />
.CO<br />
ü<br />
m O<br />
o<br />
CÖ<br />
g<br />
•4—><br />
Q<br />
2.1<br />
2.09<br />
2.08<br />
2.07<br />
±.18% ±.12%<br />
- t<br />
1<br />
0 5 10<br />
Measurement N°.<br />
Figure 3: Standard deviation as estimated (6v/v)theo. = .12% and observed<br />
(6v/v)expe = .18% from sampling at 16 different locations on the<br />
same plaque.<br />
1<br />
o<br />
5 o<br />
o<br />
o<br />
0<br />
0<br />
0<br />
0<br />
15<br />
o<br />
o<br />
o<br />
o<br />
4 o<br />
0<br />
o"<br />
16 o<br />
o<br />
I
- 408 -<br />
AN INDUSTRIAL APPLICATI<strong>ON</strong> OF COMPUTER ASSISTED TOMOGRAPHY:<br />
DETECTI<strong>ON</strong>, LOCATI<strong>ON</strong> AND SIZING OF SHRINK CAVITIES IN VALVE CASTINGS<br />
P.V. Tonne* and G. Toitllo<br />
Atomic Enz-igy o & Canada<br />
Chalk R-ti/e*, OntaKlo<br />
ABSTRACT<br />
Computer assisted tomography (CAT) scanning is a nondestructive testing technique<br />
used to obtain quantitatively accurate mappings of the distribution of<br />
linear attenuation coefficients inside an object. To demonstrate the potential<br />
of the technique for accurately locating defects in three dimensions a sectioned<br />
5 cm gate valve, with a shrink cavity made visible by the sectioning, was<br />
tomographically imaged using a Co-60 source. The tomographic images revealed a<br />
larger cavity below the sectioned surface. The position of this cavity was<br />
located with an in-plane and axial precision of approximately ± 1 mm. The<br />
volume of the cavity was estimated to be approximately 40 mm 3 .<br />
1. INTRODUCTI<strong>ON</strong><br />
Computerized Tomography (CT) is a nondestructive testing technique used to<br />
obtain quantitatively accurate mappings of the distribution of linear attenuation<br />
coefficients inside an object. One possible application of CT is to reinspect<br />
items that have failed to conform to standard QA acceptance tests in<br />
order to characterize the defect in greater detail. Since CT images contain a<br />
wealth of distortion free spatial information, CT re-inspection may be especially<br />
useful if the non-conforming item is to be repaired or upgraded. The<br />
present study demonstrates the use of CT re-inspection to detect, locate and<br />
size shrink cavities in valve castings earmarked for upgrading.<br />
Radiographs of valve castings are routinely used for QA, with the choice of<br />
acceptance standards depending on the intended end use of the valve. If an<br />
unacceptable shrink cavity is indicated on the film the casting may be upgraded<br />
by gouging out the cavity and backfilling the entire area with weld material.<br />
Because of the complex shapes involved, it is sometimes difficult to accurately<br />
locate the cavity in three dimensions by referring to radiographs alone. In<br />
some cases this had led to the removal and backfilling of considerably more<br />
material than necessary. In other cases the cavity has not been adequately<br />
repaired on the first attempt and, hence, it has been necessary to upgrade the<br />
casting a second time.
- 409 -<br />
The present study was carried out to demonstrate how CT images can be used to<br />
eliminate such costly upgrading errors. Because tomography is not presently in<br />
widespread use as an industrial NDT technique, the theory of tomography is<br />
briefly reviewed with special emphasis on the differences between conventional<br />
radiography and CT imaging. This is followed by a description of the general<br />
purpose laboratory scanner used to perform the study. Finally, results obtained<br />
with a test valve casting are presented and analyzed.<br />
2. THEORY OF TOMOGRAPHIC IMAGING<br />
As indicated in Figure l(a), the attenuation of a monoenergetic collimated beam<br />
of gamma rays as it passes through a uniform slab of material is given by 1<br />
where<br />
-uL<br />
I e<br />
o<br />
Io is the initial intensity of the gamma ray beam,<br />
I is the intensity after passing through the material,<br />
u is the linear attenuation coefficient of the material (cm" 1 ), and<br />
L is the path length in the material (cm).<br />
Equation (1) can be solved for the product (J.L:<br />
(1)<br />
In Io/I = nL (2)<br />
If the ray passes through a non-homogeneous material the path can be considered<br />
to consist of a number of elements of width w with attenuation coefficients (j-i »<br />
H2» *• *|J-n» as indicated in Figure l(b).<br />
Equation (2) then becomes<br />
In Io/I = + (3)<br />
That is, the logarithm of the measured transmission along a particular ray<br />
represents the sum of the attenuation coefficients of all the elements that the<br />
ray traverses. In tomography, the quantity In Io/I is normally called a ray<br />
sum.<br />
A single ray sum cannot, by itself, give any indication of the distribution of<br />
attenuation coefficients in the material. For example, the homogeneous sample<br />
illustrated in Figure l(c) has precisely the same ray sum as the nonhomogeneous<br />
sample shown below it.<br />
While it is not possible to use a single ray sum to learn the distribution of<br />
attenuation coefficients in an object it is intuitively obvious that ray sums<br />
obtained from other directions will provide useful information about that distribution.<br />
Multiple radiographs of a single object obtained from many viewing<br />
angles around the object are sometimes obtained for that very reason: i.e. to
- 410 -<br />
allow the radiographer to visualize the relative location of different materials<br />
inside the object. In tomography then, one obtains sets of ray sums at<br />
many different angles about the object being tested. A mathematical algorithm<br />
is then used to reconstruct the unique distribution of attenuation coefficients<br />
within the object that gave rise to the experimentally measured ray sums.<br />
In its simplest form, a tomographic scanner is designed to obtain sets of<br />
parallel ray sums in a single plane from many directions around the object. A<br />
translate-rotate tomographic scanner is schematically illustrated in Figure 2.<br />
It consists of a collimated source of radiation, either an isotopic source or,<br />
as shown here, an x-ray tube, and a collimated detector. Each parallel line in<br />
Figure 2(a) represents one ray sum. One set of parallel ray sums, obtained by<br />
translating the source and detector relative to the object (or vice versa) is<br />
called a projection. Projections from many directions around the object are<br />
obtained by rotating the source and detector relative to the object (or vice<br />
versa) and repeating the translation procedure at each new orientation. Figure<br />
2(b) illustrates a typical tomographic scan pattern consisting of translations<br />
at 1 degree increments over 180 degrees. For clarity only the 1st, 60th and<br />
120th translations are shown. More sophisticated scanners employ arrays of<br />
detectors to speed up data acquisition and make better use of the available<br />
photon flux.<br />
The rules of thumb normally applied in collecting tomographic data on a translate-rotate<br />
scanner (the type used in this study) are:<br />
1. The total number of projections, m, should be greater than or equal<br />
to nTc/4, where n is the number of ray sums per projection.<br />
2. When formed with identical source and detector collimator apertures,<br />
the effective ray width (i.e. the full-width at half-maximum) of the<br />
beam midway between source and detector is approximately half the<br />
width of the collimator aperture.<br />
3. The ray sura sample spacing should be less than or equal to half the<br />
effective ray width (Nyquist criterion).<br />
Because of the large amounts of data involved, all practical tomographic scanners<br />
are computer controlled. Hence, the acronyms CT for computed tomography<br />
or CAT for computer assisted tomography, are generally used to describe the<br />
technology. A computer is also required to reconstruct the tomographic image<br />
from the measured ray sums. The reconstruction technique most often used, and<br />
that used in this study, is termed filtered back projection. The details of<br />
this technique are described in reference 1.<br />
Tomographic images are, by their nature, digital images. They are constructed<br />
on a grid of picture elements or pixels typicaily 64 x 64 to 1024 x 1024 in<br />
dimension. The width of each pixel is normally chosen equal to the ray spacing.<br />
Each pixel is assigned a value which is proportional to the attenuation<br />
coefficient, JJ., at the corresponding location in the object.
- 411 -<br />
Since tomographlc Images are reconstructed from gamma-ray intensity measurements<br />
which are subject to statistical variation, the reconstructed |j.'s will<br />
also have an associated statistical variance a^ 2 . The standard deviation,<br />
ff„, is the precision with which a given |a can be determined from a single<br />
pixel in a tomographic image.<br />
It can be shown [2,3,4] that an average a^ 2 can be estimated using<br />
where<br />
o 2 = î_ï<br />
11 3<br />
12At N<br />
tot (4)<br />
Ntot is the total number of photons passing through the object,<br />
D is the average diameter of the object (cm), and<br />
At is the width of a pixel (cm).<br />
Equation (4) is only exactly true if the number of counts in all ray sums passing<br />
through the object are equal. The equation is, however, approximately<br />
correct even when this situation does not exist. As can be seen, UM, At and<br />
Ntot are related such that, with Ntot constant, an increase in At by<br />
a factor of 2 results in an eight-fold improvement (decrease) in o^ 2 .<br />
3. TOMOGRAPHIC SCANNING APPARATUS<br />
The tomographic scanner used in this study is illustrated schematically in<br />
Figure 3. It comprises a control and data acquisition computer, scintillation<br />
detector with associated electronics, a radioactive source, source and detector<br />
collimators and shielding and a computer controlled motorized assembly for<br />
translating and rotating the test object.<br />
The computer system consists of a DEC LSI-11/23 microcomputer with 256 k bytes<br />
of on board memory, a 20.8 M byte (Winchester) disk and a ^ M byte floppy<br />
disk.<br />
In the present study an Nal scintillator, hermetically sealed to a photo multiplier<br />
tube, was used as the detector. The detector preamp and single channel<br />
analyzer have been customized to enable measurement of pulse rates of up to<br />
50 kHz without software correction for pile up. With software correction this<br />
limit can be extended by more than an order of magnitude. Detector output<br />
pulses are counted in a standard counter/timer and interfaced to the DEC Q-bus<br />
through the Canberra 1788 and DEC DLV11-J serial interfaces.<br />
The test object is placed on a turntable mounted on a translatable carriage.<br />
The turntable and carriage are driven by individual stepper motors and their<br />
absolute positions are continuously monitored by shaft angle encoders. The<br />
stepper motors and encoders are interfaced to the bus through ADAC 1600 series<br />
interface cards.
- 412 -<br />
The photon source may be Cs-137, Ir-192 or Co-60, depending on the requirements.<br />
A Co-60 source with a useful penetration of up to 24 cm of steel was<br />
used in the present study.<br />
The source and detector collimators are Identical tungsten-alloy cylinders,<br />
165 mm in length, mounted in V blocks. The apertures are continuously variable<br />
in width from 0 to 6.4 mm and are available in three discrete heights; 1.59 mm,<br />
3.175 mm and 6.35 mm.<br />
Tomographie reconstruction is executed on the LSI-11/23 with the assistance of<br />
the Computer.Design and Applications (CDA) MSP-3 array processor. Typical<br />
reconstruction time is about 90 seconds for a 128 x 128 reconstruction grid.<br />
After reconstruction, the images are displayed on a colour graphics screen<br />
using the CDA MDP-3B graphics processor. Images can be photographed off the<br />
screen for further study and can be archived in digital form on disk or magnetic<br />
tape.<br />
4. CT DEM<strong>ON</strong>STRATI<strong>ON</strong><br />
4.1 Description of the Sample Valve Casting and Defect<br />
A class 600 flanged-end gate valve casting with a 5.08 cm (2") bore was examined<br />
in the present study. The casting had been bisected along a plane passing<br />
through the bore and the stem port axes (see Figure 4). As indicated, a shrink<br />
cavity in one of the crotch areas is exposed at the sectioned face. Shrink<br />
cavities generally represent a tear rather than a smooth internal void and<br />
therefore they may propagate, like a crack, under vibration. For this reason<br />
they are considered a serious defect. They are caused by the uneven solidification<br />
of liquid metal in the mold and often occur in regions where external<br />
and internal surface contours change rapidly. The crotch area of a gate valve<br />
is a typical example of this.<br />
The fact that shrink cavities are likely to occur near contoured surfaces presents<br />
special difficulties in the interpretation of radiographs. In particular,<br />
determining the coordinates of the defect in three dimensions from radiographs,<br />
as required for upgrading, is subject to error because of the changing<br />
film exposure due to variation in the path lengths through the irregularly<br />
shaped object. This difficulty is compounded by the ever present magnification<br />
distortion due to the divergence of the x-ray beam on its path from source to<br />
film.<br />
4.2 CT Scanning Procedure<br />
To demonstrate the effectiveness of CT in circumventing these problems inherent<br />
in conventional radiography, the valve casting was toraographed in the region of<br />
the shrink cavity. The casting was mounted in the scanner as shown in Figure<br />
4. In order to delineate the defect in three dimensions, six adjacent slices,<br />
each perpendicular to the axis of the bore and separated by 1.59 mm along the
- 413 -<br />
bore axis, were scanned. The general location of these slices is indicated by<br />
the scan plane illustrated in Figure 4. For ease of reference in the discussion<br />
that follows, the height of the scan plane above the bottommost surface of<br />
t_>e casting will be referred to as distance Z in the axial direction. The<br />
distance Z to the midpoint of the topmost slice was 109.5 mm and this was reduced<br />
for each of the six slices by slipping a succession of 1.59 mm (1/16")<br />
shims underneath the casting.<br />
The details of the scanner set-up are summarized in Table 1. The 4.4 x<br />
(12 curie) Co-60 source delivered an unattenuated gamma-ray exposure of approximately<br />
29.7 R/hr at the face of the Nal detector. The corresponding count<br />
rate, as measured using the test assembly counting electronics and corrected<br />
for pile-up, was approximately 7 x 10 counts/second.<br />
A modulation transfer function (MTF) test pattern was also included in each<br />
scan. As described in Appendix A, it was used to determine the system MTF.<br />
The 50% point on the system MTF corresponds to 2.1 line pairs/cm (see Figure<br />
A-2).<br />
4.3 Results<br />
Tomographie images of the six parallel slices are shown in Figure 5.<br />
A useful method of understanding these images is to imagine that the casting<br />
and MTF test piece have been cut through the scanning plane indicated in Figure<br />
4, and that the material above the cut has been removed. The tomographic<br />
images represent what an observer, standing over the source and facing in the<br />
direction of the detector, would see when looking down on the surface from<br />
above, except that the tomographic image is a map of linear attenuation<br />
coefficients, not a true photographic image. The series of images (a) through<br />
(f) in Figure 5 represents a succession of these surfaces revealed by a<br />
succession of cuts from top (Z = 109.5 mm) to bottom (Z = 101.6 mm).<br />
The images shown in Figure 5 are photographs taken from a CRT display. The CRT<br />
display consists of a 256 x 256 array of square picture elements or pixels too<br />
small for the human eye to discern. Each pixel represents a volume (voxel),<br />
0.85 mm on a side, and 1.59 mm high, in the object itself. Each pixel is<br />
assigned an integer number between 0 and 255 that is related to the linear<br />
attenuation coefficient in the corresponding voxel. Rather than attempt to<br />
display these integer numbers themselves, each number is assigned a unique<br />
shade of grey. In this particular assignment the largest integers (i.e. largest<br />
linear attenuation coefficients) are assigned the lightest shades of grey<br />
and the smallest integers are assigned the darkest. The operator is free to<br />
use any assignment of grey scale, or colour, he chooses. Certain choices may<br />
tend to enhance some features and suppress others. However, the data, i.e.<br />
the actual reconstructed attenuation coefficients, remain unaltered.<br />
It is immediately apparent from Figure 5 that the shrink cavity is not confined<br />
to the surface of the sectioned face of the valve but that a larger cavity is<br />
located somewhat below the surface. This subsurface cavity is particularly
- 414 -<br />
evident in Figures 5 (c) and (d) but is detectable in all the images. Also<br />
noted is a seccnd smaller subsurface cavity in Figure 5(a) that is not evident<br />
in the other images.<br />
To demonstrate the dimensional accuracy of the tomographic images, known dimensions,<br />
the lengths a,b,c and d defined in Figure 5(b), were compared with those<br />
determined from the tomographic reconstruction. The reconstructed dimensions<br />
were measured by counting pixels and converting this to mm using the conversion<br />
factor 1 pixel = 0.86 mm. Reconstructed and measured dimensions agreed to<br />
within 1 mm (see Table 2).<br />
The in plane coordinates of the centre of the main body of the subsurface<br />
shrink cavity can thus be determined in relation to a reference coordinate<br />
system. The location and orientation of one possible reference coordinate<br />
system is shown in Figure 5(c). The axes are oriented parallel and perpendicular<br />
to the straight edges of the flange, thereby providing useful directional<br />
references both before and during upgrading. With reference to this coordinate<br />
system, the centre of the main body of the subsurface shrink cavity lies 28<br />
±1 mm below the surface in the x direction and 72 ±1 mm from the edge of the<br />
flange in the y direction.<br />
The axial (Z) coordinate of the centre of the main subsurface shrink cavity is<br />
apparently somewhere between Z = 106.4 mm and Z = 104.8 mm, corresponding to<br />
Figure 5(c) and (d), respectively. The intermediate value Z = 106 ±1 mm is a<br />
reasonable compromise.<br />
The volume of the main body of the subsurface cavity may be estimated by assuming<br />
the cavity is approximately spherical in shape with a diameter of approximately<br />
6 mm, as estimated from Figure 5(c). On this basis the volume is estimated<br />
to be approximately 40 mm 3 .<br />
Tomography not only provides quantitatively accurate dimensional information<br />
but it can also be used to determine, again quantitatively, the linear attenuation<br />
coefficients of the materials in the image. The theoretical linear attenuation<br />
coefficents, (i^h' °^ brass, iron and air for 1.25 MeV gammas are compared,<br />
in Table 3, to the corresponding average coefficients determined from<br />
the reconstructed images, {Trec. The excellent agreement between nth, which<br />
has an associated uncertainty of about 5%, and |Trec is typical of tomographic<br />
imaging.<br />
The ]Irec of each material was obtained by averaging the [i's assigned to a<br />
number of pixels representing that material. Thus the uncertainty in the<br />
ÏÏreo a iï> is considerably smaller than the pixel-to-pixel standard deviation<br />
in M-reo represented by o^. In fact, op is a„//n where n is the number<br />
of f pixels il used d to determine d i Mrec* ^ i i Fi 5 i<br />
or tne i raa S es i- n Figure 5 cy is approxi-<br />
mately 0.011 cm" 1 . Since 120 pixels were use to determine free» a \î *- s<br />
0.011//Î2Ô or 0.001 cm" 1 , as quoted in Table 3.<br />
Expressed as a percentage of the n of iron, a^ is approximately 3%. It is<br />
important to note that this 3% is not equivalent to radiographie contrast
- 415 -<br />
normally measured with a shim type penetrameter. Radiographie penetrameter<br />
contrast refers to an integrated contrast along the line between source and<br />
film, taking into account all the material along this line, whereas o^ is the<br />
precision with which the n of a specific pixel anywhere within the interior of<br />
the object can be determined. If the difference between the reconstructed ^'s<br />
of two pixels is greater than 2a„ then there is a 95% chance that these two<br />
pixels represent different M-'s in the object. Since in tomographic images any<br />
two pixels or groups of pixels can be compared in this way, a^/y, is equivalent<br />
to a much smaller radiographie contrast, typically two orders of magnitude<br />
smaller.<br />
5. C<strong>ON</strong>CLUSI<strong>ON</strong>S AND DISCUSSI<strong>ON</strong><br />
The tomographic scanning and reconstruction method has been used to detect and<br />
locate a subsurface shrink cavity in the crotch area of a sectioned 5.08 cm<br />
(2") valve casting.<br />
The main body of the cavity has been located in three dimensions with a precision<br />
in each dimension of approximately ±1 mm. The volume of the main body of<br />
the cavity has been estimated to be about 40 mm 3 . The precision with which the<br />
cavity can be located would be of considerable assistance in upgrading the<br />
valve.<br />
The ability to accurately reconstruct linear attenuation coefficients has been<br />
demonstrated. The standard deviation in a given reconstructed linear attenuation<br />
coefficient was determined to be approximately 0.011 cm"* or 3% of iron.<br />
This is typical of CT images and represents an improvement of one to two orders<br />
of magnitude over the contrast obtained with film radiography.<br />
Although the demonstration scans in this study each required approximately 14<br />
hours for data acquisition, the scan time can be greatly reduced by employing<br />
arrays of detectors rather than a single detector. Scan times of 4 seconds or<br />
less are typical of current generation medical CAT scanners. There is no fundamental<br />
reason why CAT scanners intended for industrial purposes cannot achieve<br />
similar scan times.<br />
6. REFERENCES<br />
1. L.M. Zaty, "Basic Principles of Computed Tomography Scanning",<br />
Radiology of the Skull and Brain, Vol. V: Technical Aspects of<br />
Computed Tomography, 1981.<br />
2. D.A. Chesler, R.J. Stephen and N.J. Pele, "Noise Due to Photon Counting<br />
Statistics in computed X-Ray Tomography", Journal of Computer<br />
Assisted Tomography, 1(1), 1977.<br />
3. R.A. Brooks and G.D. Di Chiro, "Statistical Limitation in X-ray<br />
Reconstructive Tomography", Medical Physics, Vol. 3, No. 4, 1976.<br />
4. P. Reimers and J. Goebbels, "New Possibilities of Nondestructive<br />
Evaluation by X-Ray Computed Tomography", Materials Evaluation, 41,<br />
1983.
Source<br />
- 416 -<br />
Table 1: Details of Scanner Set-up<br />
Source type Co-60<br />
Source strength 4.4 x 10 11 Bq (12 curies)<br />
Detector<br />
Detector type Nal (Tl)<br />
Detector size 5 cm x 5 cm<br />
Geometry<br />
Collimator aperture opening (w x h x 1) 3.175 mm x 3.175 mm x 16.5 mm<br />
Source-to-detector spacing 711 mm<br />
Centre of rotation to source 394 mm<br />
Scanning Pattern<br />
Total number of images 6<br />
Vertical spacing between images 1.59 mm<br />
Projections per image 201<br />
Angular spacing between projections 0.9°<br />
Rays per projection 256<br />
Spacing bewteen rays (At) 0.86 mm<br />
Count time per ray 1 s<br />
Table 2: Comparison of Reconstructed and Actual Dimensions<br />
Dimension (see Figure 5(b)) Reconstructed (mm) Actual (mm)<br />
a<br />
b<br />
c<br />
d<br />
174.5<br />
34.4<br />
85.1<br />
49.0<br />
Table 3: Comparison of Reconstructed and Theoretical<br />
Linear Attenuation Coefficients<br />
175.0<br />
35.0<br />
85.5<br />
50.0<br />
Material "£,.»„ (cm" 1 ) art(cm~ i ) li^.cm" 1<br />
^th c<br />
brass 0.450 ± .001 0.45<br />
iron 0.422 ± .001 0.42<br />
air -0.001 ± .001 0.00
(a) L<br />
(b)<br />
ic)<br />
1<br />
- 417 -<br />
•HwK-<br />
|3AJ|3JU|3/JL^J<br />
^ S >^ s^\<br />
1JU|6JU|2AJLJ<br />
J<br />
« J<br />
,e iM-<br />
e-9/jw<br />
•=>l=lo<br />
Figure 1 : Attenuation of gamma-rays of intensity Io incident on<br />
(a) a homogeneous slab,<br />
(b) a non-homogeneous slab, and<br />
(c) homogeneous and non-homogeneous slabs having identical ray<br />
sums.<br />
X-RAY /<br />
TUBE \f<br />
1st<br />
TRANSLATI<strong>ON</strong><br />
(a)<br />
DETECTOR<br />
1st<br />
TRANSLATE<br />
1° INCREMENTS<br />
60th ""^<br />
TRANSLATI<strong>ON</strong> \<br />
(b)<br />
P 120th<br />
\ TRANSLATI<strong>ON</strong><br />
Figure 2 : A schematic illustration of a translate-rotate CT scanner.<br />
(a) The source and detector are first translated relative to the<br />
object. Each parallel line represents one ray sura.<br />
(b) Multiple projections are obtained by rotating the source and<br />
detector around the object.
DEC<br />
LSI 11/23<br />
CPU<br />
SOURCE<br />
SOURCE'<br />
FLASK<br />
ROM/FLOPPY/<br />
WINCHESTER<br />
MEMORY<br />
ADAC 1604<br />
PULSE OUTPUT C<strong>ON</strong>TROLLER<br />
STEPPER<br />
MOTOR (SM)<br />
DRIVERS<br />
O<br />
SOURCE<br />
COLLIMATOR<br />
PHOT<strong>ON</strong><br />
BEAM<br />
Figure 3<br />
PRINTER/CRT<br />
li-TRANSLATI<strong>ON</strong><br />
IjSAE<br />
^TRANSLATI<strong>ON</strong> SM<br />
LEAD<br />
SCREW<br />
CDA MSP-3<br />
ARRAY<br />
PROCESSOR<br />
CDA MDP-3B<br />
GRAPHICS<br />
PROCESSOR<br />
ADAC 1664<br />
TTL I/O INTERFACE<br />
SHAFT ANGLE<br />
ENCODER (SAE)<br />
DECODERS<br />
H.V.<br />
Nal DETECTOR<br />
DETECTOR<br />
COLLIMATOR<br />
CARRIAGE<br />
TURNTABLE<br />
PREAMP<br />
POWER<br />
PREAMP<br />
Q.-BUS<br />
GRAPHICS<br />
DISPLAY<br />
A block diagram of the CRNL toraographic scanner apparatus.<br />
DLV11-J<br />
SERIAL<br />
INTERFACE<br />
CANBERRA<br />
1788<br />
INTERFACE<br />
•Pi-<br />
1<br />
oo
Figure A : Orientation of the casting and MTF test piece on the scanner turntable.<br />
The scanning plane was changed by slipping 1.59 mm (1/16")<br />
shims underneath the casting.<br />
I
(e)<br />
- 420 -<br />
Figure 5 : Tomographie images of the six parallel slices. The height of the<br />
scanning plane, Z, is 109.5 mm for image (a) and decreases in<br />
steps of 1.59 mm in each of the subsequent images (b) through (f).<br />
The large subsurface cavity is particularly evident in images (c)<br />
and (d) but is detectable in all images. A second smaller subsurface<br />
cavity is evident in (a) only. The modulation transfer function<br />
(MTF) test pattern is analyzed in Appendix A.
- 421 -<br />
APPENDIX A<br />
Determination of the System Modulation Transfer Function (MTF)<br />
To determine the modulation transfer function (MTF) of the imaging system an<br />
MTF test pattern was included in each scan. It consisted of a cylindrical<br />
block of brass containing 21 holes 4, 3, 2, 1.5, 1.0, 0.75 and 0.5 mm in diameter<br />
arranged in seven parallel rows of three. The diameters of holes in a<br />
given row were equal and the centre to centre spacing between holes in each row<br />
was twice the hole diameter.<br />
Each of the seven sections in Figure A-l (a) through (g) show, at the left, the<br />
portion of Figure 5(a) containing the image of the MTF test pattern. The linear<br />
attenuation coefficients of the pixels along the horizontal line superimposed<br />
on each reproduction of the test pattern are plotted at the right of the<br />
image. Plots through all seven sets of holes, from largest to smallest, are<br />
shown in Figure A-l (a) through (g), respectively.<br />
The three smallest sets of holes (Figures A-l (e), (f) and (g)) show no evidence<br />
of separation. In all the others (Figures A-l (a), (b), (c) and (d))<br />
separation is evident. However, the n's associated with the holes do not extend<br />
all the way down to the air level, nor do the n's representing the brass<br />
between the holes fully recover to the n of brass. The peak to peak variation<br />
between the reconstructed \i of the holes and that of the brass between the<br />
holes, expressed as a percentage of the full range from air to brass and<br />
plotted as a function of inverse hole diameter, is a measure of the system<br />
MTF.<br />
The system MTF, based on the data shown in Figure A-l, is plotted in Figure<br />
A-2. The conversion between line pairs/mm (lp/mm) and hole diameter is<br />
1 lp/mm = l/2d. The 50% point on the MTF corresponds to 2.1 lp/cm.
- LOO -<br />
Figure A-l: Plots of the linear attenuation coefficients along lines passing<br />
through the seven rows of equal size holes in the modulation<br />
transfer function (MTF) test pattern of Figure 5(a). Hole diameters<br />
are: (a) 4 mm, (b) 3 mm, (c) 2 mm, (d) 1.5 mm, (e) 1 mm,<br />
(f) 0.75 mm and (g) 0.5 mm.
ÜJ<br />
I/)<br />
1<br />
ce<br />
TO PE/<br />
a.<br />
100<br />
90 -<br />
80 -<br />
70<br />
60<br />
50<br />
40<br />
30<br />
20<br />
10<br />
- 423 -<br />
2 3<br />
Up/cm (= 1/2d)<br />
Figure A-2: The system modulation transfer function (MTF) determined, as<br />
explained in the text, from the data in Figure A-l•
- 424 -<br />
COMPUTER C<strong>ON</strong>TROLLED CARB<strong>ON</strong> COMPOSITE PANEL TESTERS<br />
5. VaWalle.<br />
Canadian HVL Technology Limited, Rzxdate, Ontario<br />
When we speak of composites these days we generally refer to panels composed<br />
of a number of layers of cloth made of carbon, graphite or kevlar fibers held<br />
together and impregnated with a binder or adhesive which after the<br />
application of heat and pressure within a vacuum becomes a strong light<br />
panel.<br />
Since the price of the materials from which these panels are made and the<br />
construction as well as the testing has been very costly, the use of these<br />
panels has been restricted primarily to military airplanes, frames and skins.<br />
However as the cost of the material and construction is coming down and more<br />
applications are being found more aluminum and other metal components are<br />
being replaced with this superior material.<br />
The superiority of these panels for their function in airplanes or other<br />
structures are in their weight-to-strength ratio as well as in the rigidity<br />
in one direction versus flexibility in other directions.<br />
However, the graphite or carbon composites can exhibit these properties only<br />
if the layers are bonded together perfectly. Even small areas of unbond<br />
between adjacent layers can significantly weaken the panels. Particularly if<br />
these multilayer composites are glued against both sides of a metal or fibre<br />
honeycomb, this so called composite sandwich can only be inspected by means<br />
of ultrasonic through-transmission.<br />
The heart of all present ultrasonic panel testing systems consists of two<br />
transducers, one transmitter and one receiver, connected to an appropriate<br />
ultrasonic instrument.<br />
Since the honeycomb centered panel is extremely light and often very large,<br />
up to 50" x 8" wide or larger, it is difficult to submerge it and perform an<br />
immersion test on it.<br />
Therefore, the two transducers are housed in so called water jets or<br />
squirters. These water coupling devices produce a small diameter water<br />
column in front of the transducers which allows the sound beam to be conveyed<br />
from the transmitter transducer to the receiver.<br />
When the panel is placed between the transducers, the sound beam passes<br />
through the panel except where it is interrupted by an area of unbond. This<br />
loss of signal amplitude is related to an amount or area of unbond.
- 425 -<br />
This paper will concentrate on testing systems which incorporate a number of<br />
novel and some unique patented devices and features. Probably the most<br />
important of these are the recently patented so-called "Laminar Link" water<br />
column coupling devices, which produce truly laminar water columns at any<br />
orientation, horizontal or vertical at virtually any water pressure. Since<br />
this laminar flow is achieved without any object such as vanes etc. in the<br />
water stream these units can be used in pulse echo or through-transmission to<br />
allow not only the detection but also the location of unbond areas.<br />
In order to be capable of testing large, thin and fragile panels, besides<br />
thick rigid ones, the discussed systems were designed to hold these panels<br />
vertically so the water will not settle on the one side as is the case with<br />
systems holding the panels horizontally.<br />
Three types of systems will be presented:<br />
1) With opposed water coupling devised mounted on a U-shaped arm to test<br />
flat and slightly curved panels of reasonable width.<br />
2) Wif.h opposed coupling devices locked together to test very long flat<br />
panels up to 10 ft. wide<br />
3) Independently mounted coupling devices to test very long flat panels as<br />
well as complexly curved panels having widths up to at least 10 ft.<br />
Also discussed will be a newly developed T)ot-Matrix C-scan recorder which<br />
prints scan results in GO-NO-GO, lined grey scale shading or numbered defect<br />
severity indications. This unit requires no special paper but uses plain<br />
high quality bond paper.<br />
All systems are under complete computer control and are operating by means of<br />
simple menu prompted direction which will be discussed in some detail.
- 426 -<br />
SOME UNC<strong>ON</strong>VENTI<strong>ON</strong>AL TECHNIQUES FOR THE INSPECTI<strong>ON</strong> OF LAYERED<br />
MATERIALS<br />
P. Cietc<br />
National Reiea-tc/i Council o & Canada<br />
Beuche\vMi, Quebec<br />
ABSTRACT<br />
Optothermal methods using a pulsed laser have been developed for the detection<br />
of delaminated areas in layered materials such as Al-epoxy or graphite-epoxy<br />
laminates. The pulsed laser be?m heats the laminate surface to some degrees C<br />
inducing a slight thermoelastic bend of the unbonded layer.<br />
Two techniques have been tested for the detection of delaminatlons. In the<br />
first technique, an infrared detector is used to monitor the thermal decay<br />
rate of the surface after the absorption of the pulse. The slope of the<br />
signal shows a strong variation when the thermal wavefront reaches an<br />
interface such as a delamination or a layer of different thermal properties.<br />
This technique allows to detect subsurface defects and to locate their depth,<br />
or alternatively to evaluate the thermal properties of the layer.<br />
Graphite-epoxy composites were evaluated by this method. The second technique<br />
uses a laser interferometer to measure the thermoelastic bending of the layer,<br />
which is much larger If the layer is unbonded. The displacement of optically<br />
rough surfaces can be measured with great accuracy by using a focused<br />
interferometer. Delaminations on an Al-epoxy honeycomb laminate were detected<br />
by this technique.<br />
These techniques are attractive for the inspection of large and complex<br />
structures because they require no contact with the inspected surface, are<br />
easy to scan and require access from one side only. Moreover, both techniques<br />
exert a lifting thermoelastic moment on the surface which tends to widen the<br />
delamination gap. This facilitates the detection of lack-of-adhesion defects<br />
when the two layers are initially in physical contact.<br />
INTRODUCTI<strong>ON</strong><br />
Laminate structures such as metal or composite lap joints and honeycomb<br />
sandwich structures are increasingly used in the transportation and<br />
construction industries. Their combination of high stiffness and strength<br />
with low weight is particularly appreciated in the aerospace industry. Such<br />
structures are formed by adhesively bonding several layers of metallic or<br />
fiber-reinforced-plastics materials in order to obtain a more uniform stress<br />
transfer and an increased fatigue life. However, these materials require a<br />
careful inspection after being assembled to detect any extended bonding<br />
defect, such as an internal delamination or an unbonded area, which would<br />
severely affect the structural resistance to transverse or shear loading.
- 427 -<br />
Ultrasonic and radiographie methods are extensively used for the<br />
nondestructive inspection of laminate structures » . Commercially available<br />
instruments based on such methods have reached an advanced stage of<br />
development, and are now capable to detect most types of bonding defects.<br />
Research is nevertheless quite active for the development of alternate<br />
techniques which would lead to faster and less cumbersome devices, requiring<br />
no coupling media or physical contact with the inspected structure, and which<br />
can be applied where access is difficult and possible from one side only.<br />
Moreover, none of the presently used techniques is satisfactory for the<br />
detection of lack-of-adhesion defects between the layer and the adhesive,<br />
i.e. areas where the adherents are in physical contact with each other, but<br />
with a zero bond strength > .<br />
A research program has thus been established at the Industrial Materials<br />
Research Institute to explore unconventional inspection techniques for<br />
stratified materials. Laser-holographic and therr.ographic techniques are<br />
attractive because they are non-contact and provide in a relatively short time<br />
a full image of the inspected surface. However, after nearly 20 years of<br />
development efforts such techniques have found few applications out of the<br />
research laboratory. Part of the reasons for such an inertia are technical:<br />
in particular, thermography is surface-emissivity dependent and provides<br />
little quantitative Information, while holography is very sensitive to ambient<br />
vibrations and to background light, usually requiring that the inspected<br />
object be mounted on a cumbersome vibration-isolated table in a ' dark<br />
environment. Such requirements make it difficult to apply such techniques to<br />
bulky structures on-the-field. A research effort is thus under way at the<br />
Industrial Materials Research Institute to develop alternative optical<br />
techniques which, while maintaining the non-contact and scanning rapidity<br />
advantages, may provide more quantitative information and show a potential for<br />
on-the-field applicability.<br />
Two optothermal approaches under evaluation at our Institute are described in<br />
this paper. The first is a pulsed thermographie technique which has been<br />
applied to a quantitative thermal analysis of graphite-epoxy laminates. The<br />
second Is a novel thermoeleatic technique using a localized heating source and<br />
focused Interferometer to inspect the transient thermomechanical behaviour of<br />
the unbonded layer. A description of such methods and of the experimental<br />
results as well as a discussion of the relative merits and fields of<br />
application of each of these approaches Is presented.<br />
PULSED PHOTOTHERMAL INSPECTI<strong>ON</strong> OF LAYERED MATERIALS<br />
The effectiveness and convenience of thermographie techniques for the<br />
detection of delamlnattons and other subsurface defects has been demonstrated<br />
by a number of authors " . Usually, the surface of the layered structure is<br />
heated by a heat lamp or a hot-air gun and the temperature distribution during<br />
or after heating Is observed with a thermographie camera. The presence of a<br />
delamination is revealed by the appearance of a hot spot because of the<br />
thermal-barrier effect of the unbonded interface. More detailed analyses have<br />
been performed recently » » in order to better quantify the retrieved<br />
information. Part of the research effort at our Institute is directed toward<br />
the assessment of the thermographie method and the expansion of the amount of<br />
Information that can he obtained from the detected signal. In particular, the<br />
possibility of a thermal analysis of the Inspected materials by properly
- 428 -<br />
processing the time resolved signal is pointed out in this section.<br />
One of the problems encountered in thermographie NDT is related to the spatial<br />
fluctuations of the recorded thermal image caused by changes in the surface<br />
emlssivity or by non-uniformities of the heating-source distribution«. A<br />
differential approach using a short heating period and monitoring the<br />
temperature decay rate of the heated area is an effective way to avoid such<br />
problems . The temperature history within a homogeneous sample after the<br />
surface absorption of a short heat pulse can be represented in a first<br />
approximation by the one-dimensional model for an instantaneous pulse > :<br />
T(z,t) - — exp (1)<br />
(ïïKpct) 1 / 2<br />
where z and t are respectively the depth and time variables, Q is the injected<br />
energy density, K is the thermal conductivity, p is the density, c the<br />
specific heat, and a = K/pc is the thermal diffusivity. We can see from ea.<br />
(1) that the surface temperature of a homogeneous sample decays as (t) '<br />
after the absorption of the pulse.<br />
The effect of the finite pulse duration and non-uniform heat distribution are<br />
best analyzed by a numerical model. A two-dimensional (axially symmetric)<br />
finite-difference model was thus used to study the heat propagation en the<br />
stratified material. under transient surface heating . The surface<br />
temperature history obtained from such a model in a configuration typical of<br />
our experimental tests is shown in fig. 1. In this figure, curve (a)<br />
represents the thermal history after the absorption of a 50 ps pulse by a<br />
homogeneous steel sample (K = 0.7 W/cm °C, p = 7.8 g/cm and c = 0.5 j/g °C).<br />
We can see that, apart from an initial perturbation caused by the finite pulse<br />
duration, the thermal decay slope is equal to - 1/2 on the log-log scale, in<br />
agreement with eq. (1). Curve (b) corresponds to a lower-conductivity layer<br />
(K * 0.14 W/cm °C), 100 pm thick, perfectly bonded to a steel (K. = 0.7 W/cm<br />
°C) substrate. The curve slope presents a discontinuity when the thermal<br />
front reaches the interface after a period which is of the order of the<br />
thermal propagation time t = I /4a through the thickness I of the coating.<br />
The position of this discontinuity can thus be used to obtain the coating<br />
thickness I if its thermal diffusivity is known. Moreover, the vertical<br />
displacement Tjj between the coated-sample curve (curve b) and its asymptote<br />
for large values of t (curve a) can be used to obtain the ratio between the<br />
effusivity e = Kpc of the coating and the substrate:<br />
T,i = log (2)<br />
2<br />
as can be inferred from eq. (1). Curve (c) corresponds to an unbonded layer<br />
where the interface defect has been represented by a I um-thick air layer.<br />
The presence of the defect is apparent in this curve.<br />
\ graphite-epoxy laminate comprising four layers of Hercules AS 3501-6 prepreg<br />
bonded together with equal orientation was inspected by such method. Each<br />
layer has a thickness of 125 \m and is composed of a large number of<br />
equally-oriented, 8 )im-diameter graphite fibers embedded in an epoxy matrix.
- 429 -<br />
A micrograph of the laminate is shown in fig. 2. As can be seen from fig. 2a,<br />
Che upper layer was partially unbonded because of interface stress during the<br />
mechanical manipulation. The laminate was inspected by the experimental<br />
apparatus shown in ftg. 3. A CO2-TEA laser with volumetric ratio equal to<br />
0.05 provided 60mJ, 50 ps pulses with pulse shape optimized to' obtain<br />
efficient surface heating without overheating .<br />
The single-transversal-mode beam had a gaussian distribution with 1/e 2 radius<br />
of 4 mm. The surface temperature was monitored by an InSb infrared detector<br />
pointed in the center of the irradiated area. Such a detector has a spectral<br />
sensitivity range from 2 to 5.5 pm, so that it is insensitive to the 10,6 pm<br />
reflected intensity of the CO2 laser.<br />
The experimental results obtained with such a system are shown in fig. 4.<br />
Curve (a) corresponds to a well-bonded region, while curves (b) and (c) were<br />
obtained on regions where the upper layer was partially unbonded. The<br />
discontinuity in the slope of curves (b) and (c) after approximately 10 ms<br />
corresponds to the arrival of the thermal front on the thermally resistive<br />
air—filled unbonded interface. The thermal propagation time of 10 ms over a<br />
125 vim thickness corresponds to the expected thermal properties of the<br />
graphite-epoxy material.<br />
While the slope of curve (a) is nearly equal to -1/2, as for a uniform sample,<br />
curve (c) has a significantly different slope, even before 10 ms. This may be<br />
interpreted by taking into account the heterogeneous nature of this material.<br />
As can be seen in fig. 2b, the fibers are randomly distributed in the matrix,<br />
some fibers being in contact with each other while other fibers are separated<br />
by a significant thickness of the resin. As the thermal conductivity of<br />
epoxy, 4 . 10~ W/cm °C, is much smaller than the conductivity of graphite<br />
fibers 17 ' 18 , 0.5 w/cm °C, the fiber distribution may significantly affect<br />
the transverse thermal properties of the composite.<br />
Figure 5 shows the results of the fiber distribution in the composite matrix.<br />
In the three-dimensional model used to obtain such curves, 25 ym x 25 urn x 1<br />
cm graphite elements are embedded in an epoxy matrix with volume ratio of<br />
0.5. A contact thermal resistance equivalent to 2.5 urn of epoxy is assumed<br />
between adjacent graphite elements. Curves (a) and (b) in Figure 5 correspond<br />
to a well-bonded and delaminated layer, respectively, with a uniformly<br />
distributed graphite-epoxy mesh, as shown in Figure 6a. Curve (c) in Figure 5<br />
corresponds to a delaminated layer with a random matrix obtained by a<br />
random-number generator, Figure 6b. We can see that curve (c) has a more<br />
gradual slope, similar to the experimental curve (c) of Figure 4. These<br />
results suggest that photothermal techniques may provide information not only<br />
on the presence of sub-surface delaminations, but also on the fiber content<br />
and distribution in fiber-resin composite materials.<br />
As a conclusion to this first section, it can be said that pulsed photothermal<br />
techniques are an attractive unconventional approach for the inspection of<br />
coated or layered materials. Their rapidity and non-contact nature makes such<br />
techniques convenient to use for scanning the surface of large structures.<br />
Moreover, the possibility to use such techniques for an on-the-field thermal<br />
evaluation of layered materials may open new opportunities in the NDE field.
- 430 -<br />
PULSED PILATOMETRIC INSPECTI<strong>ON</strong> OF UNB<strong>ON</strong>DED LAYERS<br />
Thermographie techniques are particularly convenient for the inspection of<br />
high emissivity surfaces such as graphite-epoxy laminates or anodized<br />
coatings. However, the thermographie signal is normally lower by an order of<br />
magnitude when inspecting metallic surfaces such as Al-epoxy honeycomb panels<br />
such as the one shown in fig. 7. This is a consequence of the low emissivity<br />
of bare aluminium surfaces, which is typically of the order of 0.1. On the<br />
other hand, thermography measures an indirect property of the material i.e.<br />
thermal resistance of the layer-to-layer interface, rather than the property<br />
which is of real interest i.e. the adhesive bond at the interface. A lack of<br />
adhesion between the aluminium skin and the honeycomb core in a laminate such<br />
as the one shown in fig. 7 would be quite difficult to detect thermographically,<br />
because the thermal conductivity of the air-filled honeycomb core<br />
is lower by several orders of magnitude than the thermal conductivity of the<br />
aluminium skin.<br />
i 2 u<br />
The holographic interferometry approach • » , on the other hand, provides<br />
a direct bond evaluation by mechanically lifting the bonded layer between the<br />
two hologram exposures. Although either vibration, pressure or heat can be<br />
applied to lift the unbonded layer, thermal stressing is the most convenient<br />
technique for a fast and non-contact inspection of laminates ~ . With this<br />
technique, the structure is holographically inspected before and after surface<br />
heating by a lamp or a hot-air gun. unbonded areas are revealed by the<br />
appearance of localized fringe patterns if the therraoelastic bending forces<br />
produce a- normal displacement of the unbonded layer of more than half a<br />
wavelength.<br />
In order to take advantage of the mechanical inspection capability of<br />
holography while avoiding its sensitivity to ambient vibration and stray<br />
light, a thermoelastic interferometric technique has been developed at our<br />
Institute 21 .<br />
The basic principle is shown in fig. 8. The layer to be inspected is heated<br />
by a pulsed and focused laser beam. The pulse duration, typically 1 ms or<br />
less, is comparable to the thermal propagation time across the layer<br />
thickness. Consequently, a transient therraoelastic moment causes the layer to<br />
lift, if unbonded, while its position is monitored by a sensitive focused<br />
interferometer. Because of the short duration of each pulse, large surfaces<br />
can be scanned rapidly. Apart from eliminating fringe-counting and heat<br />
uniformity problems, this approach has the following advantages with respect<br />
to holography:<br />
1. Every point being inspected during a very short period of time,<br />
low-frequency ambient vibration noise can be avoided. Operation on the<br />
industrial floor Is thus possible with our approach. High-power pulsed<br />
lasers have been used in the past to avoid hologram blurring during the<br />
exposure time, but ambient vibrations during the heating period between<br />
the two exposures cannot be avoided with such a method. Moreover, the<br />
colllmated interferometer beam can be spatially filtered to avoid stray<br />
light, so that our system can operate under conditions of normal ambient<br />
illumination.
- 431 -<br />
2. Holography requires displacements of some wavelengths, while laser<br />
interferometers are sensitive to much smaller displacements > .<br />
Heterodyne holographic techniques can be used to detect small<br />
deformations , but this requires the use of complex double-wavelength<br />
exposure techniques demanding an accurate control of the alignment'and of<br />
the ambient conditions while still requiring point-by-point scanning.<br />
3. In the holographic approach, a broad heat source is used to heat a large<br />
area of the bonded structure. The generated thermoelastic stress<br />
distribution inside the unbonded layer is of the type shown in fig. 9a.<br />
A large longitudinal stress is produced in a direction parallel to the<br />
surface, but the bending component which makes the.-unbonded area to lift<br />
outward is quite small, unless the layer is initially slightly convex.<br />
Bending of the unbonded layer appears to be more related to differential<br />
thermal expansion between the layer and the substrate and to longitudinal<br />
thermal gradients produced by the thermal barrier effect of the unbonded<br />
interface » . Such lifting mechanisms are weak and unreliable. On<br />
the other hand, the transient strain configuration under localized pulsed<br />
heating shown in fig. 9b is much more effective in producing a bending<br />
moment, as was verified using both analytical and finite-element<br />
thermoelastic modelling.<br />
The advantages of conventional holographic techniques over interferometry are<br />
that a full picture is obtained without scanning and that scanning surfaces<br />
can be inspected. However, focused interferometers are now commercially<br />
available which can probe perfectly rough surfaces at any angle of incidence<br />
and at operating distances of several meters . As to the scanning speed, a<br />
rate of the order of 100 cm /s could be obtained with 1 ms pulses and a 3 mm<br />
resolution; such a speed compares favorably to the exposure, heating and<br />
development time required with the conventional holographic approach.<br />
The interferometric technique was experimentally tested on different<br />
adhesively-bonded samples. Some preliminary results were obtained using a<br />
Cu-Be layer, 125 um thick, epoxy-bonded to a massive plexiglass plate. Some<br />
unbonded areas at the metal-plexiglass interface could be easily localized<br />
visually through the plexiglass. A nearly 1 J, 1 ms pulse from a Nd:YAG laser<br />
was used to heat the layer surface to some degrees C above ambient. Fig. 1Ü<br />
shows two curves obtained by scanning the two concentric laser beams over two<br />
delaminated areas of different dimensions. Curve (a) was obtained over a 1<br />
cm-diameter unhond with a 6 mm-diameter heating YAG beam, while curve (b) was<br />
obtained over a 3 mm-diameter delamination with a 0.7 mm-diameter heating<br />
beam. The unbonds are clearly visible, and the signal level of some \m is<br />
much higher than the ultimate sensitivity of the interferometer.<br />
The same interferometric system was also used to inspect an Al-epoxy laminate<br />
of the kind shown in fig. 7. Some artificial bonding defects were produced by<br />
inserting some Teflon foils between the aliminium skin and the epoxy or<br />
between the epoxy and ths honeycomb core. The skin thickness was of 300 ym,<br />
and the honeycomb cells had a size of 3 mm. Fig. 11 shows some curves<br />
obtained by scanning the skin surface over a 2 cm-diameter unbonded area. The<br />
heating-beam diameter was of 1 mm, 3 mm and 6 mm, respectively for curves (a),<br />
(b) and (c). All of the three curves show a clear evidence of the unbonded<br />
area, which is situated between the 5 mm and the 25 mm positions on the<br />
abscissa. A nearly constant background signal is also obtained on the
- 432 -<br />
well-bonded region, because of the linear thermal expansion as well as the<br />
elastic deformation of the relatively compliant epoxy interface. It is<br />
interesting to note that the 3 mm underlying honeycomb structure can be seen<br />
only when the heating laser beam diameter is smaller than the cell size, in<br />
agreement with the thermoelastic considerations illustrated in Fig. 9 . '<br />
C<strong>ON</strong>CLaSI<strong>ON</strong><br />
The optothermal approach to the nondestructive inspection of layered materials<br />
appears to be an attractive alternative to conventional materials inspection<br />
techniques. Its main advantages are that it is fast, non-contact, does not<br />
require sample preparation nor access from both sides of the laminate, and<br />
provides information about the mechanical bonding properties of the layer.<br />
Two inspection techniques are presented in this paper, and the results of our<br />
experimental investigation are discussed. The infrared technique appears to<br />
be a relatively inexpensive method for the inspection of high emissivity<br />
materials such as graphite-epoxy laminates, while the interferometric approach<br />
is more appropriate for the inspection of metal-honeycomb structures and for<br />
quantitative analyses of bonded laminates.<br />
REFERENCES<br />
1. Scott, I.G. and Scala, CM.- A review of NDT of Composite Materials, NDT<br />
Int. Vol. 15, p. 75, 1982.<br />
2. Segal, E. and Rose, J.L., NDT Techniques for Adhesive Bond Joints, in<br />
"Research Techniques in NDT", R.S. Sharpe ed., Academic Press, London,<br />
1980.<br />
3. Chaskelis, H.H., and Clark, A.V., - Ultrasonic Nondestructive Bond<br />
Evaluation: an analysis of the problem, Mater. Eval. Vol. 38, p. 20,<br />
1980.<br />
A. Birnbaum, G. and Vest, CM. - Holographic NDE: status and future, in<br />
International Advances in NDT, vol. 9, p. 257, 1983.<br />
5. Vavilov, V.P. and Taylor, R. - Theoretical and practical aspects of the<br />
thermal NOT of bonded structures, in "Research Techniques in NDT", R.S.<br />
Sharpe ed., Academic Press, London, 1982.<br />
6. Vogel, P.E.J., - Evaluation of bonds in armorplate and other materials<br />
using infrared NDT techniques, Appl. Opt. Vol. 7, p. 1739, 1968.<br />
7. Lawson, W.D. and Sabey, J.W. - Infrared Techniques, in "Research<br />
Techniques in NOT", R.S. Sharpe ed., Academic Press, London, 1970.<br />
8. Trezek, G.J. and Balk, S. - Provocative Techniques in Thermal NDT<br />
Imaging, Mater. Eval., Vol. 34, p. 176, 1976.<br />
9. Engelhardt, R.E. and Hewgley, W.A. - Thermal and Infrared Testing, in<br />
"Nondestructive Testing, a Survey", NASA-SP-5113, 1973.
- 433 -<br />
10. Green, D.R., Schneller, M.D. and Sulit, R.A. - "Thermal NDE Method for<br />
Thermal Spray Coatings", 10th Int. Thermal Spraying Conf., Essen, West<br />
Germany, May 1983.<br />
11. Williams, J.H., Mansouri, S.H. and Lee, S.S., - One-dimensional Analysis<br />
of Thermal Nondestructive Detection of Delamination and Inclusion Flows,<br />
British Journal of NDT, Vol. 22, p. 113, 1980.<br />
12. Griffiths, D.H. - Toward Quantitative Industrial Thermography, SPIE 246,<br />
152, 1980.<br />
13. Carslaw, H.W. and Jaeger, T.C. - Conduction of Heat in Solids, Chapt. 10,<br />
Oxford University Press, Oxford, 1959,<br />
14. Ready, J.F. - Effects of High-Power Laser Radiation, Academic Press, New<br />
York, 1971.<br />
15. Cielo, P. - Analysis of Pulsed Thermal Inspection, 14tli Symposium on<br />
Nondestructive Evaluation, San Antonio, Apr. 19-21, 1983.<br />
16. Dallaire, S. and Cielo, P. - Pulsed Laser Treatment of Plasma-sprayed<br />
Coatings, Metall. Trans. Vol. 13B, p. 479, 1982.<br />
17. Fritz, W. Huttner, W., Maler, K. and Brandt R. - Thermophysical<br />
Properties of Carbon-fibre Carbon-matrix Composites at High Temperatures,<br />
Rev. Int. Hautes Temp. Refract., Vol. 16, p. 350, 1979.<br />
18. Balageas, O.L. and Luc, A.M. - Non Stationary Thermal Behaviour of<br />
Directional Reinforced Composites. Limit of Application of Thermal<br />
Property Horaogenization, 18th AIAA Thennophysics Conf., Montreal, June<br />
1-3, 1983,<br />
19. Kersch, L.A. - Laminate Structure Inspection, in "Holographic NDT,<br />
R.K.ERF ed., Academic Press, N.Y., 1974.<br />
20. Harris, W.J. and Woods, D.C. - Thermal Stress Studies Using Optical<br />
Holographic Interferometry, Mater. Eval. Vol. 32, p. 50, 1974.<br />
21. Cielo, P., Rousset, G. and Bertrand, L. - Nondestructive Interferometric<br />
Detection of Unbonded Layers, Optics and Lasers in Engineering, 1984 (to<br />
be published).<br />
22. Moss, G.E., Miller, L.R. and Forward, R.L. - Photon-noise Limited Laser<br />
Transducer for Gravitational Antenna, Appl. Opt. Vol. 10, p. 2495, 1971.<br />
23. Wickramasinghe, H.K., Martin, Y., Ball, S. and Ash, E.A. - Thermodisplaceraent<br />
Imaging of Current in Thin-Film Circuits, Electron Lett.<br />
Vol. 18, p. 700, 1982.<br />
24. Damm, V., DISA, Copenhagen, and Wedgwood, F.A., AERE Harwell,' England,<br />
Private communications.
0.1<br />
- 434 -<br />
0.01 0.1 1<br />
TIME (m s)<br />
Fig. 1 Calculated thermal decay curves for a<br />
plasma-sprayed coating heated by a 50 |xs pulse.<br />
Curve (a): bare steel sample;<br />
Curve (b): well-bonded Ni-Cr coating on steel substrate;<br />
Curve (c): delaminated coating
- 4 35 -<br />
Fig. 2 (a): cross-section of the 4-ply graphite-epoxy laminate<br />
used in the experiments. The upper layer was partially<br />
disbonded under stress;<br />
(b): close-up view of the fiber-epoxy matrix<br />
(a)<br />
1OO|o.m<br />
(b)
IR Detectors-<br />
Fig. 3 Experimental apparatus used to thermally inspect<br />
the four-ply laminate shown in Fig. 2.
10 _<br />
0.1<br />
1_<br />
0.01 0.1 1<br />
- 437 -<br />
TIME (m s)<br />
(c)<br />
10 100<br />
Fig. 4 Experimental curves obtained with a graphite-epoxy<br />
laminated sheet.<br />
(a): well-bonded area;<br />
(b) and (c): delaminated regions.
o ü<br />
LU<br />
LU<br />
OC<br />
Ü<br />
LU<br />
CC<br />
ICC<br />
LU<br />
Q.<br />
LU<br />
LU<br />
Ü<br />
Lf<br />
CC<br />
œ<br />
1O_<br />
1_<br />
0.1.<br />
0.01<br />
I<br />
0.1<br />
- 438 -<br />
1<br />
TIME (m s)<br />
(a)<br />
10 100<br />
Fig. 5 Theoretical curves obtained for a graphite-epoxy<br />
laminated sample.<br />
Curve (a): well-bonded sample;<br />
Curve (b): delaminated layer, uniform fiber distribution;<br />
Curve (c): delaminated layer, random fiber distribution.
• •' '
i M H<br />
I ! ! : •<br />
niy<br />
ill<br />
Fig. 9 Thermoelastic siress distributiow on a partially<br />
unbonded layer under:<br />
(a): continuous broad healing ami<br />
(b): pulsed focused heatin. j<br />
(a)
HEATING<br />
PULSES<br />
DETECTOR<br />
" -—FILTER<br />
LASER<br />
B<strong>ON</strong>DED<br />
LAYER<br />
SUBSTRATE<br />
Fig. 8 Basic principle of the thermoelastic interferometric<br />
method for detecting an unbond. A surface-heating<br />
pulse causes the layer to thermoelastically bend,<br />
while its vertical displacement is detected by a<br />
focused interferometer.<br />
.c-
Fig. 7 Al-honeycomb sandwich structure to be Inspected<br />
for unbond defects.
- 443 -<br />
5 10<br />
POSITI<strong>ON</strong> (mm)<br />
Fig. 10 Interferometrically measured vertical<br />
displacement of the heated surface of an<br />
epoxy-bonded Cu-Be layer over:<br />
(a): a 1 cm-diameter unbond and<br />
(b): a 3 mm-diameter unbond.
10 15 20 25<br />
POSITI<strong>ON</strong> (mm)<br />
30 35<br />
Fig. 11 Interferometric scan of a 2 cm-diameter<br />
deiamination on an Al-epoxy honeycomb laminate,<br />
with a heating beam diameter of:<br />
(a): 1 mm, (b): 3 mm and (c): 6 mm.<br />
I
- 445 -<br />
MATERIALS EFFECTS <strong>ON</strong> ACOUSTIC EMISSI<strong>ON</strong> DURING DEFORMATI<strong>ON</strong> AND<br />
FRACTURE<br />
M. Nabit Ba&6im<br />
Univzuity o (j Wa.ni.toba.<br />
Winnipeg, Manitoba, Canada<br />
ABSTRACT<br />
The technique of acoustic emission is increasingly used for nondestructive<br />
testing of large structures. Because of its passive nature in detection of<br />
the stress wave emissions which occur during plastic deformation and/or crack<br />
propagation, it is possible to monitor large areas by using a relatively low<br />
number of transducers. While the application of the technique is being<br />
perfected, understanding of the nature and origin of acoustic emission is<br />
less clear due to the complexity of quantifying the nature of the acoustic<br />
sources and the effect of the medium where the stress waves are propagating.<br />
In this investigation, models describing the origin of acoustic emission in<br />
terms of dislocations are reviewed. The effect of attenuation on the stress<br />
waves is examined and the extent of acoustic emission activity from a<br />
specific material is related to its microstructure. A comprehensive analysis<br />
of the nature of acoustic emission as observed in a typical application is<br />
thus presented.<br />
1. INTRODUCTI<strong>ON</strong><br />
Considerable effort has taken place in recent years to use acoustic emission<br />
as a nondestructive testing method for evaluation of the integrity of<br />
structures and components. The bulk of this effort has been directed to<br />
applications involving large structures, namely pressure vessels, nuclear<br />
reactors...etc where flaw localization is the main concern and evaluation of<br />
the severity of these flaws is the primary goal of the application of the<br />
technique. Commercially available systems, which are equipped with very<br />
sophisticated digital processing techniques including micro-processors are<br />
available for performance of such tests. The application of acoustic<br />
emission is usually conducted during proof tests of these structures and<br />
record.-, of performance of the acoustic emission activity, in terms of<br />
acoustic emission parameters such as total ringdown count, count rate, peak<br />
amplitude and energy of the signals are obtained. As well, the position of<br />
the flaw is determined from the differences in time of arrival to several<br />
transducers positioned at various locations on the structures. Bassim(l) and<br />
and Houssny-Eman(2) have summarized the different techniques used for
- 446 -<br />
analysis of the acoustic signals in the time and frequency domains and have<br />
provided a theoretical basis for the flaw localization techniques most used<br />
in industrial applications.<br />
While examples abound in the literature on successful application of acoustic<br />
emission to large structures by use of general purpose large systems which<br />
are computer-controlled to give, in real time, the location and severity of<br />
flows, the operation of such systems depends, to a large extent on the<br />
presence of highly skilled operators which can provide almost immediate<br />
interpretation of the raw data obtained during the tests. Furthermore, these<br />
sophisticated pieces of equipment have only a limited capability to perform<br />
over a long period of time beside the fact that they are costly, bulky and<br />
difficult to transport from one place to another. These drawbacks limit the<br />
use of existing first generation acoustic emission equipment to applications<br />
which have a finite length of time, such as proof tests, where they can be<br />
assembled to perform tha test, proceed with the actual test, analyse the data<br />
and disassemble to move on to another application.<br />
An alternate process to provide continuous monitoring of large structures<br />
with acoustic emission is to use surveillance units which continuously<br />
analyse acoustic emission signals, determine the relevant signals vhlch are<br />
indicative of impending failure and use these signals to trigger a warning<br />
system. These surveillance units would be compact, inexpensive, and a number<br />
of these units can be applied to a very large structure such as a pipeline,<br />
an offshore platform or a nuclear reactor for continuous monitoring of these<br />
structures. The units will each include a micro-processor which is<br />
software-controlled to perform the different functions assigned to each of<br />
these units. This new approach to use of acoustic emission for continuous<br />
monitoring of large structures has been patented by Bassim and Tangri(3).<br />
The description of the system is given by Mitchell and Bassim(4).<br />
An important feature of these surveillance units is that they are<br />
preprogrammed to identify signals corresponding to impending failure. This<br />
requires an intensive research program to identify the sources of acoustic<br />
emission in the structure under consideration and to determine the effect of<br />
attenuation on the waveform of the signal detected by the transducer. In<br />
this paper, this research program is outlined with regard to an industrial<br />
application, involving the use of the acoustic emission system to monitor a<br />
long pipeline.<br />
2. OUTLINE OF MATERIAL TESTING PROGRAM<br />
The material testing program for a continuous monitoring system used in<br />
conjunction with pipelines can be summarized in the following steps:<br />
2.1 Laboratory tests of pipeline materials<br />
The tensile and fracture properties of a pipeline material were determined<br />
using tensile and three-point bending specimens respectively. The effect of
- 447 -<br />
orientation of these specimens with respect to the rolling plate was<br />
investigated. In the fracture tests, parameters such as the stress intensity<br />
factor K-£C and the J-integral, Jjc which is used to characterize ductile<br />
fracture, were determined.<br />
Together with the mechanical testing, acoustic emission was obtained using a<br />
system similar to that in Fig. 1. Acoustic emission parameters, namely the<br />
total count, count rate and root mean square of the amplitude were determined<br />
as a function of the applied load and elongation. In the fracture tests,<br />
acoustic emission was obtained by positioning a transducer on the side of the<br />
specimen and following the emissions as the plastic zone ahead of the crack<br />
is developed followed by stable crack growth.<br />
2.2 Model tests on pipeline materials<br />
In these tests, two aspects were considered: Firstly, the continuous<br />
acoustic signals due to leaks were detected with transducers of different<br />
frequencies located at different distances from leaks induced by drilling<br />
holes in the pipe as described by Bassim and Tangri(5). Air flow at<br />
different pressures was maintained through the pipe. The experimental set-up<br />
is shown in Fig. 2. Secondly, the attenuation coefficient a was determined<br />
using ultrasonic methods described in (6). The attenuation characteristics<br />
of the pipe were thus established and calculations of the optimum spacing<br />
corresponding to a transducer with a given resonant frequency was<br />
calculated.<br />
2.3 Theoretical consideration of acoustic emission sources<br />
Theoretical modelling of the sources of emissions in terms of dislocation<br />
motion and interaction was carried out to arrive at the optimal frequencies<br />
corresponding to an acoustic emission signal. Also, once an emission has<br />
occurred, modelling of the type of waves and its propagation characteristics<br />
was carried out to determine the proper position of transducers on the pipe<br />
and their optimum spacing. The effect of pipe (or specimen geometry) on the<br />
destructive and constructive interference of acoustic signals, following the<br />
analysis of Hamel, Ballon and Bassim(7) was also investigated.<br />
3. RESULTS AND DISCUSSI<strong>ON</strong><br />
Following are representative data obtained in the course of the research<br />
program as well as an interpretation of their physical meaning and the extent<br />
that they can be used in the design of a dedicated acoustic emission system.<br />
3.1 Material characterization<br />
Flat tensile samples, from two orientations with respect to the rolling<br />
direction, were obtained from A53B pipeline steel. Fig. 3 shows a typical<br />
stress-strain and RMS data from a longitudinal specimen. It is observed that<br />
the A53B deforms initially by a complex Lüders band propagation. Several
- 448 -<br />
bands are initiated right after the first yield drop. High acoustic emission<br />
activity occurs in the Lüders strain region of the stress-strain curve. Each<br />
peak on the RMS curve corresponds to a stress drop on the stress-strain<br />
curve. This indicates that the major source of acoustic emission in this<br />
steel is due to inhomogeneous deformation. No systematic difference in<br />
bahaviour was observed between longitudinal and transverse samples. Thus<br />
acoustic emission activity is detected mostly within ehe first 3% plastic<br />
strain during tensile testing.<br />
In fracture testing, three-point bend specimens, 12.7 mm thick with a notch<br />
in the TL and LT directions were obtained from the as-received A53B pipeline<br />
steel. Fig. 4 shows a typical RMS voltage of the acoustic signal during a<br />
test as a function of load point displacement. The acoustic emission<br />
activity remains high throughout the displacement range investigated which<br />
represents the development of the plastic zone ahead of ehe crack tip without<br />
extending the crack length. Most of the emission were found to occur during<br />
the plastification of the ligament ahead crack tip and no significant<br />
emission accompany the stable crack growth process itseLî. This is in<br />
agreement with Blanchette, Bassim and Dickson(8) for A516 grade 70 steel.<br />
3.2 Leak and attenuation characterization<br />
The leak tests were performed on a segment of pipe, 400 cm long with inside<br />
and outside diameters of 160 and 170 mm respectively. The pressure and flow<br />
rate of the air through the pipe were measured accurately. Two series of<br />
tests, one using an accelometer with a range of up to 25 KHz and the other<br />
with an acoustic transducer with a resonant frequency of 1140 KHz were<br />
conducted. An artificial leak was made by drilling a hole directly in the<br />
middle of the pipe and the signal was analysed for its frequency content.<br />
Fig. 5 represents a plot of the RMS from different leak sizes as a function<br />
of pressure while Fig. 6 shows the effect of hole diameter on the leak<br />
frequency obtained at different air pressures through the pipe. It is<br />
noticed that for a given pressure, as the hole diameter decreases, the leak<br />
frequency also decreases. On the other hand, the value of the R.M.S. which<br />
represents the relative amplitude of the signal increases with the increase<br />
in hole diameter as well as increase of the pressure. Finally, measurement<br />
of the signal amplitude at different points on the pipe shows no significant<br />
attenuation for this relatively short segment of pipe. Turbulence due to end<br />
effects appears to increase the signal amplitude to some extent.<br />
In the attenuation tests, the aim is to determine the attenuation coefficient<br />
aj which is defined for a plane wave as<br />
P = Po e~ ad (1)<br />
where PQ is the initial pressure and P is the sound pressure at distance d.<br />
In this study the attenuation coefficient a, expressed as dB/m, was
- 449 -<br />
determined experimentally for pipeline steels through the thickness of plate<br />
as well as in the pipe segment described earlier. The procedure for<br />
estimation of a through plate thickness using the pulse reflection method was<br />
described by Krautkramer(9), and will not be repeated here. For X70 pipeline<br />
steel, the attenuation coefficient was found to be equal to 1.09 dB/m while<br />
for X42 steel, the attenuation was much higher with a = 2.125 dB/m.<br />
In the tests on the pipe segment, a wave generator was used to produce<br />
frequencies swept between 01 to 1 MHz. An exciter transducer was connected<br />
with the wave generator. A receiving transducer was placed at specific<br />
distances from the exciter and the signal, after amplification, pass-as<br />
through an R.M.S. voltmeter and is plotted as a function of frequency. The<br />
value of a determined in this test corresponded to 0.89dB/m. This value is<br />
affected by the reflection of the elastic waves in the free ends of the<br />
pipe.<br />
3.3 Theoretical modelling<br />
Significant effort in the theoretical modelling of the acoustic emission<br />
system and its functioning was made. This ranges from modelling the sources<br />
of burst emission in terms of lattice perturbation and dislocation<br />
interaction (Bassim and Wassef(lO)), to the prediction of the stresses and<br />
displacement in a plate due to a propagating crack (Bassim, Wassef and<br />
Tangri(ll)) and finally the prediction of the type and frequency spectre! of<br />
the noise due to the flow of a gas (or fluid) in a circular hole or<br />
rectangular slit (Wassef et al(12)).<br />
The above studies have for general conclusions that acoustic emissiori is<br />
caused by the perturbation of an otherwise perfect lattice to cause phonon<br />
emission. Involved quantum mechanics approach shows that the interaction of<br />
two or more dislocations is capable to producing a phonon emission<br />
corresponding to about 200 KHz which is the range where acoustic emission is<br />
most observed.<br />
Once acoustic emission is produced, its propagation will depend on the<br />
position of the source with respect to the structure and on the relative<br />
dimension of the emitting crack and the rest of the structure. Modelling,<br />
involving spectral representation, Green's functions as well as solutions to<br />
Fredholm integral equations are used to predict the stresses and displacement<br />
fields in a body containing a crack and subjected to a loading. These<br />
predictions give an accurate indication of the type of transducer to be used<br />
in conjunction with a given geometry as well as the most appropriate<br />
transducer configuration to use in a structure.<br />
4. C<strong>ON</strong>CLUSI<strong>ON</strong>S<br />
A summary of a research program which was carried out in conjunction with<br />
development of a new acoustic emission system is presented. The specific<br />
application is the continuous monitoring of pipelines. The program involves<br />
materials studies on pipeline steels, model tests to measure signals
- 450 -<br />
characteristics of leaks and theoretical modelling of the sources of acoustic<br />
emission and wave propagation in a solid medium.<br />
5. ACKNOWLEDGEMENTS<br />
The author acknowledges the efforts of his associates, Drs. W. Wassef, M.<br />
Houssny-Emam and D. Tseng and Mssrs. Gauthier and Hartle, in producing some<br />
of the data presented. The financial support of Petro-Canada and the Natural<br />
Sciences and Engineering Research Council is acknowledged and appreciated.<br />
6. REFERENCES<br />
(1) M.N. Bassim, in Microstructural Characterization of Materials by<br />
Non-Microscopical Techniques, 5th Riso International Symposium on Metallurgy<br />
and Materials Science, 1984, pp. 193-198.<br />
(2) M.N. Bassim and M. Houssny-Emam, in Acoustic Emission, J.R. Matthews<br />
ed., Gordon and Breach, 1983, pp. 139-164.<br />
(3) M.N. Bassim and K. Tangri, U.S. Patent 570866 (pending).<br />
(4) J. Mitchell and M.N. Bassim, Fifth Canadian Conference on Nondestructive<br />
Testing, 1984.<br />
(5) M.N. Bassim and K. Tangri, Proceedings of the International Conference<br />
on Pipeline Inspection, Published by CANMET, 1984, pp. 529-544.<br />
(6) J. Krautkramer and H. Krautkramer, ultrasonic Testing of Materials,<br />
Spring-Gerlag, 1969, pp. 89-95.<br />
(7) F. Hamel, J.P. Bailon and M.N. Bassim, Ultrasonics, 1977, pp. 125-127.<br />
(8) Y. Blanchette, M.N. Bassim and J.I. Dickson, in Proceedings of the 5th<br />
Canadian Fracture Conference, Pergamon Press, 1981, pp. 191-199.<br />
(9) .T. Krautkratner, Ultrasonic Nondestructive Testing of Materials, 1961.<br />
(10) M.N. Bassim and W. Wassef, Physical Review, Submitted for publication,<br />
1984.<br />
(11) M.N. Bassim, W. Wassef and K. Tangri, Int. J. Fracture, in press.<br />
(12) W.A. Wassef, M.N. Bassim, M. Houssny-Emam and K. Tangri, submitted for<br />
publication in J. Acoustical Soc. of America. 1984.
STRUCTURE<br />
•^ TRANSDUCER (S)<br />
- 451 -<br />
TIME DOMAIN<br />
ANALYSIS<br />
FREQUENCY<br />
DOMAIN ANALYSIS<br />
RINGOOWN COUNT<br />
ENERGY<br />
ROOT-MEAN SQUARE OF AMPLITUDE<br />
AMPLITUDE DISTRIBUTI<strong>ON</strong><br />
COMPUTER<br />
FLOW LOCATI<strong>ON</strong><br />
Figure 1: Typical Experimental Set-up<br />
-C<strong>ON</strong>STANT PRESSURE ^PIEZOELECTRIC TRANSDUCER<br />
VALVE-2 / OR<br />
fi ^ 7 /—PRESSURE GAUGE-2 /— ACCELEROMETER<br />
-FLOWMETER TRANSDUCER SfTES PRESSURE REGULATOR-<br />
AMPLFIER FILTER<br />
---<br />
t i_<br />
TERMINAL<br />
PRESSURE GAUGE- I<br />
C<strong>ON</strong>STANT PRESSURE VALVE-1 -<br />
-<br />
R. M. S.<br />
METER<br />
SPECTRUM<br />
ANALYSER<br />
BE!»<br />
Figure 2: Leak Testing on Pipeline
72Or<br />
- 452 -<br />
STRESS<br />
2.30 SX» 7.50 10.0 12 5 ISO 17.5 20.0<br />
TRUE STRAIN (%)<br />
Figure 3: Acoustic emission from tensile test of<br />
pipeline steel.<br />
7S0 -<br />
*• 500 -<br />
250<br />
0.5 1.0 1.5 2.0<br />
LOAD - POINT DISPLACEMENT ( mm )<br />
Figure 4: Load and RMS vs displacement during<br />
fracture test.
- 453 -<br />
500<br />
PRESSURE (KPO)<br />
LEAK DIAMETERS • 0.5 mm<br />
* 1.0 mm<br />
A 1.5 mm<br />
• 2.0 mm<br />
X 2.5 mm<br />
Figure 5: R.M.S. vs pressure in leak tests<br />
335 900<br />
PRESSURE (Kpo)<br />
Figure 6: Frequency vs pressure in leak tests<br />
670<br />
670
Ahmad, A.<br />
Babcock & Wilcox Canada<br />
Coronation Blvd.<br />
CAMBRIDGE, Ontario<br />
N1R 5V3<br />
Allan, Carolyn<br />
CSNDT Foundation<br />
Mohawk College<br />
135 Fennell Avenue W.<br />
HAMILT<strong>ON</strong>, Ontario<br />
L8N 3T2<br />
Allinson, J.O.<br />
Westinghouse Canada Inc.<br />
Nuclear Products Dept.<br />
PORT HOPE, Ontario<br />
L1A 3V4<br />
Andersen, H.<br />
Gulf Canada Products Co.<br />
TOR<strong>ON</strong>TO, Ontario<br />
Andress, B.L.<br />
Stelco Inc.<br />
Box 490<br />
WELLAND, Ontario<br />
L3B 5R2<br />
Atherton, D.<br />
Queen's University<br />
KINGST<strong>ON</strong>, Ontario<br />
K7L 3N6<br />
Barker, P.<br />
Atomic Energy of Canada Ltd.<br />
Sheridan Park Research Community<br />
MISSISSAUGA, Ontario<br />
L5K 1B2<br />
Baron, J.A.<br />
Ontario Hydro Research<br />
800 Kipling Avenue<br />
TOR<strong>ON</strong>TO, Ontario<br />
M8Z 5S4<br />
Bassim, Nabil<br />
University of Manitoba<br />
Dept. of Mechanical Engineering<br />
WINNIPEG, Manitoba<br />
R3T 2N2<br />
- 454 -<br />
LIST OF DELEGATES<br />
Beards, David<br />
Ontario Hydro<br />
Toronto, Ontario<br />
Behal, V.G.<br />
Dofasco Inc.<br />
HAMILT<strong>ON</strong>, Ontario<br />
Berger, H.<br />
Industrial Quality Inc.<br />
9832 Canal Road<br />
P.O. Box 2397<br />
GAITHERSBURG, MD 20879<br />
Bradbury, T.<br />
Sonco Steel Tube Ltd.<br />
14 Holtby Avenue<br />
BRAMPT<strong>ON</strong>, Ontario<br />
L6X 2M1<br />
Buby, J.<br />
ASCO Ltd.<br />
PORT HOPE, Ontario<br />
Burnay S.<br />
Viatec Resource Systems Inc.<br />
Suite 425, 20 Richmond St. East<br />
TOR<strong>ON</strong>TO, Ontario<br />
M5C 2R9<br />
Burton, J.<br />
Babcock & Wilcox<br />
Coronation Blvd.<br />
CAMBRIDGE, Ontario<br />
N1R 5V3<br />
Bussiere, J.F.<br />
National Research Council<br />
IMRI<br />
75 DeMortagne Blvd.<br />
BOUCHERVILLE, Quebec<br />
J4B 6Y4<br />
Butler, M.<br />
McMaster University<br />
1280 Main St. West<br />
HAMILT<strong>ON</strong>, Ontario<br />
L8S 4K1
Caron, V.<br />
Energy Mines & Resources<br />
CANMET<br />
568 Booth Street<br />
OTTAWA, Ontario<br />
K1A0G1<br />
Carroll, Y.A.<br />
TOR<strong>ON</strong>TO, Ontario<br />
Cavers, J.<br />
Canadian Armed Forces<br />
Aircraft Maintenance Dev. Unit<br />
TRENT<strong>ON</strong>, Ontario<br />
Chittim, K.<br />
CWB<br />
254 Merton Street<br />
TOR<strong>ON</strong>TO, Ontario<br />
Churchill, Paul<br />
Suncor Oil Sands Group<br />
P.O. Box 4001<br />
FORT MCMURRAY, Alberta<br />
T9H 3E3<br />
Cielo, P.<br />
National Research Council<br />
IMRI<br />
75 DeMortagne Blvd.<br />
BOUCHERVILLE, Quebec J4B 6Y4<br />
Cobill, P.<br />
Atomic Energy of Canada Limited<br />
Radiochemical Company<br />
413 March Road<br />
P.O. Box 13500<br />
KANATA, Ontario<br />
K2K 1X8<br />
Dalrymple, D.<br />
Atomic Energy of Canada Ltd.<br />
Chalk River Nuclear Laboratories<br />
CHALK RIVER, Ontario<br />
Derkx, S.J.<br />
ASCO Ltd.<br />
PORT HOPE, Ontario<br />
Deverno, M.<br />
Atomic Energy of Canada Ltd.<br />
Chalk River Nuclear Laboratories<br />
CHALK RIVER, Ontario<br />
- 455 -<br />
DeWalle, S.<br />
CANDET<br />
18 Canso Road<br />
REXDALE, Ontario<br />
Dumoulin, G.<br />
Sonco Steel Tube Ltd.<br />
14 Holtby Avenue<br />
BRAMPT<strong>ON</strong>, Ontario<br />
L6X 2M1<br />
Duncan, D.B.<br />
Atomic Energy of Canada Ltd.<br />
Chalk River Nuclear Laboratories<br />
CHALK RIVER, Ontario<br />
Dunn, A.L.<br />
Intertechnology Limited<br />
TOR<strong>ON</strong>TO, Ontario<br />
Fahr, A.<br />
Ontario Research Foundation<br />
Sheridan Research Community<br />
MISSISSAUGA, Ontario<br />
Foster, M.L.<br />
Atomic Energy of Canada Ltd.<br />
Radiochemical Company<br />
413 March Road<br />
P.O. Box 13500<br />
KANATA, Ontario<br />
K2K 1X8<br />
Gerrior, David<br />
Eastern Technical Services<br />
686 Rothesay Avenue<br />
SAINT JOHN, N.B.<br />
E2L 4E6<br />
Gilbert, R.E.<br />
Atomic Energy of Canada Ltd.<br />
Whiteshell Nuclear Research Est.<br />
PINAWA, Manitoba<br />
ROE 1L0<br />
Gooch, K.<br />
CSNDT<br />
HAMILT<strong>ON</strong>, Ontario<br />
Grabarczyk, J.A.<br />
Atomic Energy of Canada Ltd.<br />
Sheridan Park Research Community<br />
MISSISSAUGA, Ontario<br />
L5K 1B2
Hanrath, D.<br />
Ministry of Consumer & Commercial Rel.<br />
TOR<strong>ON</strong>TO, Ontario<br />
Harding, N.<br />
CSNDT<br />
HAMILT<strong>ON</strong>, Ontario<br />
Hartling, M.<br />
McMaster University<br />
HAMILT<strong>ON</strong>, Ontario<br />
Havercroft, W.E.<br />
Consultant NDT<br />
2373 Ridgecrest Place<br />
OTTAWA, Ontario<br />
K1H 7V4<br />
Hearn, M.<br />
Stelco Inc.<br />
General Delivery<br />
NANTICOKE, Ontario<br />
NOA 1LO<br />
Helliwell, T.<br />
191 Waverley Road<br />
TOR<strong>ON</strong>TO, Ontario<br />
M4L 3T4<br />
Hirning, C. Ross<br />
Ministry of Labour<br />
Special Studies & Services Branch<br />
400 University Avenue<br />
8th Floor<br />
TOR<strong>ON</strong>TO, Ontario<br />
M7A 1T7<br />
Holden, T.<br />
Atomic Energy of Canada Ltd.<br />
Chalk River Nuclear Laboratories<br />
CHALK RIVER, Ontario<br />
Holt, R.<br />
CANMET Dept. E.M.&R.<br />
568 Booth Street<br />
OTTAWA, Ontario<br />
K1A 0G1<br />
Horenfeldt, M.<br />
Du Pont Canada Ltd.<br />
TOR<strong>ON</strong>TO, Ontario<br />
- 456 -<br />
Huether, Doug<br />
Ontario Hydro<br />
Darlington G.S.<br />
Box 1000<br />
BOWMANVILLE, Ontario<br />
L1C 3W2<br />
Julier, Alan<br />
Hocking Electronics Inc.<br />
6631 Wakefield Drive<br />
ALEXANDRIA, VA 22307<br />
Kennedy, W.<br />
Canadian Welding Bureau<br />
254 Merton Street<br />
TOR<strong>ON</strong>TO, Ontario<br />
M4S 1A9<br />
Kittmer, C.<br />
Atomic Energy of Canada Ltd.<br />
Chalk River Nuclear Laboratories<br />
CHALK RIVER, Ontario<br />
Levesque, Yvon Robert<br />
Alcan International Kingston Labs.<br />
12 Michael Grass Cres.<br />
KINGST<strong>ON</strong>, Ontario<br />
K7M 2W3<br />
Macecek, Mirek<br />
Techno Scientific Inc.<br />
205 Champ jne Drive<br />
DOWNSVIEW, Ontario<br />
Mak, D.<br />
Dept. Energy Mines & Resources<br />
CANMET<br />
568 Booth Street<br />
OTTAWA, Ontario<br />
K1A 0G1<br />
Manzer, L.<br />
CSNDT Foundation<br />
135 Fennell Avenue West<br />
HAMILT<strong>ON</strong>, Ontario<br />
Marr, David<br />
Saint John Shipbuilding & Drydock Ltd.<br />
SAINT JOHN, N.B.<br />
Mayo, W.<br />
Atomic Energy of Canada Ltd.<br />
Chalk River Nuclear Laboratories<br />
CHALK RIVER, Ontario
Mclntyre, Dent<br />
Babcock & Wilcox Canada<br />
Coronaton Blvd.<br />
CAMBRIDGE, Ontario<br />
N1R 5V3<br />
McKinney, W.<br />
E.I. duPont de Nemours & Co. Inc.<br />
Photo Products Building No. 1<br />
Chestnut Run<br />
WILMINGT<strong>ON</strong>, DE 19898<br />
McNeill, Rob<br />
Ontario Hydro<br />
Darlington G.S.<br />
Box 1000<br />
BOWMANVILLE, Ontario<br />
L1C 3W2<br />
Mitchell, A.<br />
NB Research & Productivity Council<br />
FREDERICT<strong>ON</strong>, N.B.<br />
Mitchell, J.<br />
Viatec Resource Systems Inc.<br />
Digital Building<br />
6815-8th Street N.E.<br />
CALGARY, Alberta<br />
T2E 7H7<br />
Moles, M.D.C.<br />
Ontario Hydro Research<br />
800 Kipling Avenue<br />
TOR<strong>ON</strong>TO, Ontario<br />
M8Z 5S4<br />
Monchalin, J.P.<br />
National Research Council<br />
75 DelVortagne Blvd.<br />
BOUCHERVILLE, Quebec<br />
J4B 6Y4<br />
Mullins, K.<br />
Stelco Inc.<br />
P.O. Box 2030<br />
HAMILT<strong>ON</strong>, Ontario<br />
L8N 3T1<br />
Nadeau, F.<br />
National Research Council<br />
IMRI<br />
75 DeMortagne Blvd.<br />
BOUCHERVILLE, Quebec<br />
J4B 6Y4<br />
- 457 -<br />
Nieberg, A.B.<br />
Westinghouse Canada<br />
Box 510<br />
HAMILT<strong>ON</strong>, Ontario<br />
L8N 3K2<br />
Peugeot, R.<br />
Ridge Inc.<br />
4432 Bibb Blvd.<br />
TUCKER, GA 30084<br />
Piche, L.<br />
National Research Council<br />
75 DeMortagne Blvd<br />
BOUCHERVILLE, Quebec<br />
J4B 6Y4<br />
Rathwell, L.<br />
Monenco Offshore Limited<br />
1200 Monenco Place<br />
801-6th Avenue Southwest<br />
CALGARY, Alberta<br />
T2P JW6<br />
Reykdal, L.<br />
Canadian Welding Bureau<br />
50 Paramount Road<br />
WINNIPEG, Manitoba<br />
R2X 2W3<br />
Rotter, R.<br />
McMaster Engineering Society<br />
1, Leith CT.<br />
ANCASTER, Ontario<br />
L9G 3V8<br />
Roy, M.<br />
Guardian inspection Atlantic Ltd.<br />
552A Windmill Road, Bay A<br />
DARTMOUTH, Nova Scotia<br />
B3B 1B3<br />
SEDO, J.<br />
Ontario Hydro Research<br />
800 Kipling Avenue<br />
TOR<strong>ON</strong>TO, Ontario<br />
M8Z 5S4<br />
Sharpe, R.S.<br />
Atomic Energy Research Est.<br />
Harwell, United Kingdom
Sinclair, A. (Tony)<br />
University of Toronto<br />
5 King's College Road<br />
TOR<strong>ON</strong>TO, Ontario<br />
M5S 1A4<br />
Staples, R.<br />
Stelco Inc.<br />
Box 2030<br />
HAMILT<strong>ON</strong>, Ontario<br />
L8N 3T1<br />
Tim leck, D.<br />
Canadian Welding Bureau<br />
254 Merton Street<br />
TOR<strong>ON</strong>TO, Ontario<br />
M4S 1A9<br />
Norscan NDE Ltd.<br />
94 Ambleside Drive<br />
BRAMPT<strong>ON</strong>, Ontario<br />
Tonner, Paul<br />
Atomic Energy of Canada Ltd.<br />
Chalk River Nuclear Laboratories<br />
CHALK RIVER, Ontario<br />
Tremblay, S.<br />
Laboratoire Ferex Inc.<br />
2323 Versant-Nord, Suite 110<br />
SAINTE-FOY, Quebec<br />
G1N 4P4<br />
Truss, K.J.<br />
Atomic Energy of Canada Ltd.<br />
Whiteshell Nuclear Research Est.<br />
PINAWA, Manitoba<br />
ROE 1L0<br />
van den Andel, J.<br />
Westinghouse Canada Inc.<br />
P.O. Box 510<br />
HAMILT<strong>ON</strong>, Ontario<br />
L8N 3K2<br />
Veray, Mark<br />
Pratt & Whitney<br />
TOR<strong>ON</strong>TO, Ontario<br />
- 458 -<br />
Wallar, Jon<br />
CSNDT Foundation<br />
Mohawk College<br />
135 Fennell Avenue West<br />
HAMILT<strong>ON</strong>, Ontario<br />
L8N 3T2<br />
Wallis, C.<br />
Ontario Hydro<br />
R.R. #4<br />
KINCARDINE, Ontario<br />
NOG 2G0<br />
Wallis/Assistant<br />
Ontario Hydro<br />
R.R. #4<br />
KINCARDINE, Ontario<br />
NOG 2G0<br />
Wells, John<br />
Techno Scientific Inc.<br />
205 Champagne Drive<br />
DOWNSVIEW, Ontario<br />
Yih, G.<br />
Atomic Energy Control Board<br />
Supplies and Services<br />
OTTAWA, Ontario<br />
Zirnhelt, J.H.<br />
Canadian Welding Bureau<br />
254 Merton Street<br />
TOR<strong>ON</strong>TO, Ontario<br />
M4S 1A9
ISSN 0067-0367<br />
To identify individual documents in the series<br />
we have assigned an AECL- number to each.<br />
Please refer to the AECL- numberwhen requesting<br />
additional copies of this document<br />
from<br />
Scientific Document Distribution Office<br />
Atomic Energy of Canada Limited<br />
Chalk River, Ontario, Canada<br />
K0J1J0<br />
ISSN 0067-0367<br />
Pour identifier les rapports individuels faisant<br />
partie de cette série nous avons assigné<br />
un numéro AECL- à chacun.<br />
Veuillezfaire mention du numéroAECL- si<br />
vous demandez d'autres exemplaires de ce<br />
rapport<br />
Service de Distribution des Documents Office<br />
L'Energie Atomique du Canada Limitée<br />
Chalk River, Ontario, Canada<br />
K0J1J0<br />
Price $19.00 per copy Prix $19.00 par exemplaire<br />
s ATOMIC ENERGY OF CANADA LIMITED, 1985<br />
0445-85