Electromagnetic Nondestructive Evaluation (XIV) - Volume 35 Studies in Applied Electromagnetics and Mechanics

ELECTROMAGNETIC NONDESTRUCTIVE EVALUATION (XIV) Studies in Applied Electromagnetics and Mechanics Series Editors: K. ...

Author: T. Chady | S. Gratkowski | T. Takagi | S.S. Udpa

38 downloads 998 Views 19MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

ELECTROMAGNETIC NONDESTRUCTIVE EVALUATION (XIV)

Studies in Applied Electromagnetics and Mechanics Series Editors: K. Miya, A.J. Moses, Y. Uchikawa, A. Bossavit, R. Collins, T. Honma, G.A. Maugin, F.C. Moon, G. Rubinacci, H. Troger and S. A. Zhou

Volume 35 Previously published in this series: Vol. 34. Vol. 33. Vol. 32. Vol. 31. Vol. 30. Vol. 29. Vol. 28. Vol. 27. Vol. 26. Vol. 25. Vol. 24. Vol. 23. Vol. 22. Vol. 21. Vol. 20. Vol. 19.

S. Wiak and E. Napieralska Juszczak (Eds.), Computer Field Models of Electromagnetic Devices J. Knopp, M. Blodgett, B. Wincheski and N. Bowler (Eds.), Electromagnetic Nondestructive Evaluation (XIII) Y. K. Shin, H. B. Lee and S. J. Song (Eds.), Electromagnetic Nondestructive Evaluation (XII) A. Tamburrino, Y. Melikhov, Z. Chen and L. Udpa (Eds.), Electromagnetic Nondestructive Evaluation (XI) S. Wiak, A. Krawczyk and I. Dolezel (Eds.), Advanced Computer Techniques in Applied Electromagnetics A. Krawczyk, R. Kubacki, S. Wiak and C. Lemos Antunes (Eds.), Electromagnetic Field, Health and Environment Proceedings of EHE’07 S. Takahashi and H. Kikuchi (Eds.), Electromagnetic Nondestructive Evaluation (X) A. Krawczyk, S. Wiak and X.M. Lopez Fernandez (Eds.), Electromagnetic Fields in Mechatronics, Electrical and Electronic Engineering G. Dobmann (Ed.), Electromagnetic Nondestructive Evaluation (VII) L. Udpa and N. Bowler (Eds.), Electromagnetic Nondestructive Evaluation (IX) T. Sollier, D. Prémel and D. Lesselier (Eds.), Electromagnetic Nondestructive Evaluation (VIII) F. Kojima, T. Takagi, S.S. Udpa and J. Pávó (Eds.), Electromagnetic Nondestructive Evaluation (VI) A. Krawczyk and S. Wiak (Eds.), Electromagnetic Fields in Electrical Engineering J. Pávó, G. Vértesy, T. Takagi and S.S. Udpa (Eds.), Electromagnetic Nondestructive Evaluation (V) Z. Haznadar and Ž. Štih, Electromagnetic Fields, Waves and Numerical Methods J.S. Yang and G.A. Maugin (Eds.), Mechanics of Electromagnetic Materials and Structures

Volumes 1 6 were published by Elsevier Science under the series title “Elsevier Studies in Applied Electromagnetics in Materials”.

ISSN 1383 7281 (print) ISSN 1879 8322 (online)

Electromagnetic Nondestructive Evaluation (XIV)

Edited by

Tomasz Chady West Pomeranian University of Technology, Szczecin, Poland

Stanisław Gratkowski West Pomeranian University of Technology, Szczecin, Poland

Toshiyuki Takagi Tohoku University, Japan

and

Satish S. Udpa Michigan State University, USA

Amsterdam • Berlin • Tokyo • Washington, DC

© 2011 The authors and IOS Press. All rights reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without prior written permission from the publisher. ISBN 978 1 60750 749 9 (print) ISBN 978 1 60750 750 5 (online) Library of Congress Control Number: 2011928887 Publisher IOS Press BV Nieuwe Hemweg 6B 1013 BG Amsterdam Netherlands fax: +31 20 687 0019 e mail: [email protected] Distributor in the USA and Canada IOS Press, Inc. 4502 Rachael Manor Drive Fairfax, VA 22032 USA fax: +1 703 323 3668 e mail: [email protected]

LEGAL NOTICE The publisher is not responsible for the use which might be made of the following information. PRINTED IN THE NETHERLANDS

Electromagnetic Nondestructive Evaluation (XIV) T. Chady et al. (Eds.) IOS Press, 2011 © 2011 The authors and IOS Press. All rights reserved.

v

Preface This volume contains selected papers from the fifteenth International Workshop on Electromagnetic Nondestructive Evaluation, which was held in Szczecin, Poland, from June 13 to 16, 2010. Previous ENDE Workshops have been held in: London, United Kingdom (1995); Tokyo, Japan (1996); Reggio Calabria, Italy (1997); Chatou, France (1998); Des Moines, United States (1999); Budapest, Hungary (2000); Kobe, Japan (2001); Saarbrücken, Germany (2002); Paris, France (2003); East Lansing, United States (2004); Iwate, Japan (2006); Cardiff, United Kingdom (2007); Seoul, Korea (2008); Dayton, United States (2009). The aim of the workshop, organized by the West Pomeranian University of Technology, Szczecin, Poland and the Japanese Society of Maintenology, was to bring together scientists from universities and research institutions conducting in-depth research into the basics of electromagnetic non-destructive evaluation (ENDE) on the one hand, and engineers presenting practical problems and industrial applications on the other. Ninety nine participants from eleven European countries and from Algeria, Australia, Brazil, China, India, Japan, Korea, and the United States, were officially registered. Eighty papers were presented in all, among them five invited papers, namely: 1. 2. 3.

4. 5.

D.C. Jiles, Ł.P. Mierczak, Y. Melikhov, Detection of Surface Condition in Ground Steel Components Using Magnetic Barkhausen Measurements, S. Honda, Pipe Wall Thickness Inspection with Current Driven Thermal Method, J.S. Knopp, M. Blodgett, J. Calzada, E. Lindgren, C. Buynak, J. Aldrin, Computational Methods in ENDE: Revolutionary Capability for the Sustainment of Aerospace Systems in the 21st Century, Z. Chen, T. Hwang, L. Wang, S. Tian, N. Yusa, Investigation on the Features of the Electric Conductivity Around a Stress Corrosion Crack, S.C. Chan, R. Grimberg, J.A. Hejase, Z. Zeng, P. Lekeakatakunju, L. Udpa, S.S. Udpa, Development of a Nonlinear Eddy Current Technique for Estimating Case Hardening Depths.

Short versions of all the contributions have been published in the Book of Abstracts, and reviewed and accordingly revised full papers have been accepted and are now included in this volume: Electromagnetic Non-Destructive Evaluation (XIV) published by IOS Press in the series Studies in Applied Electromagnetics and Mechanics. In closing, we would like to thank the authors, session chairs, and reviewers for conscientiously executing their duties to maintain the high scientific quality of the papers published in this volume. We believe that the readership of this book will find the included papers interesting and inspiring. T. Chady, S. Gratkowski, T. Takagi, S.S. Udpa Co-Editors

This page intentionally left blank

vii

ENDE 2010 – Szczecin Poland Conference Organization The 15th International Workshop on Electromagnetic Nondestructive Evaluation, Szczecin, Poland, June 13–16, 2010. Organized by West Pomeranian University of Technology, Szczecin, Poland in Cooperation with Japan Society of Maintenology, Tokyo, Japan Co-Sponsors Japan Society of Maintenology, Tokyo, Japan Oddział Szczeciński Stowarzyszenia Elektryków Polskich, Szczecin, Poland Federacja Stowarzyszeń Naukowo-Technicznych NOT w Szczecinie, Poland ZAPOL grupa reklamowa, Szczecin, Poland Technika Obliczeniowa, Kraków, Poland Standing Committee F. Kojima I. Altpeter J. Bowler N. Bowler T. Chady Z. Chen D. Jiles J. Knopp D. Lesselier K. Miya G.Z. Ni J. Pavo G. Pichenot G. Rubinacci S.J. Song T. Takagi A. Tamburrino L. Udpa S.S. Udpa

Japan Germany USA USA Poland China UK USA France Japan China Hungary France Italy Korea Japan Italy USA USA

viii

List of Referees and Scientific Committee P. Baniukiewicz L. Janousek C. Bardel T. Jayakumar H. Brauer H. Kikuchi J.R. Bowler F. Kojima K. Capova D. Lesselier T. Chady P. Lekeaka-Takunju S.C. Chan N. Mahapatra Z. Chen A. Nishimizu W. Chlewicki D. Premel Y. Goto J. Pavo S. Gratkowski G. Psuj R. Grimberg J.M.A. Rebello D. He M. Shiwa H. Huang M. Smetana Organizing Committee Honorary Member: Rector ZUT Włodzimierz Kiernożycki Chairman: Tomasz Chady Co-Chairman: Stanisław Gratkowski Ryszard Sikora Stefan Domek Andrzej Brykalski Members: Piotr Baniukiewicz Wojciech Chlewicki Paweł Frankowski Justyna Jończyk Jacek Kowalczyk Krzysztof Kujawski Paweł Lesiecki Przemysław Łopato Lech Napierała Marzena Olszewska Grzegorz Psuj Tomasz Pietrusewicz Krzysztof Stawicki Barbara Szymanik Marcin Ziółkowski

K. Stawicki G. Steffes O. Stupakov T. Takagi A. Tamburrino T. Uchimoto L. Udpa S.S. Udpa G. Vertesy T. Yamamoto N. Yusa M. Ziółkowski

ix

ENDE 2010 – Szczecin Poland List of Participants Yuji Akiyama Kanagawa Institute of Technology, Yokohama 3-19-25 Shimosueyoshi, Tsurumu-ku 230-0012 Yokohama Japan [email protected]

Iris Altpeter Fraunhofer-Institut IZFP, Saarbrücken Campus E3.1 D-66123 Saarbrücken Germany [email protected]

Piotr Baniukiewicz Department of Electrical Engineering West Pomeranian University of Technology, Szczecin ul. Sikorskiego 37 70-313 Szczecin Poland [email protected]

B.P.C. Rao NDE Division, Metallurgy and Materials Group, Indira Ghandi Centre Atomic Research 603102 Kalpakkam India [email protected]

Bilicz Sándor Budapest University of Technology and Economics Egry József út 18 1521 Budapest Hungary [email protected]

Hartmut Brauer Ilmenau Univeristy of Technology Helmholtzplatz 2, P.O.Box 100565 98684 Ilmenau Germany [email protected]

Andrzej Brykalski Department of Electrical Engineering West Pomeranian University of Technology, Szczecin ul. Sikorskiego 37 70-313 Szczecin Poland [email protected]

Stephen Burke DSTO 506 Lorimer St VIC 3207 Fishermans Bend Australia [email protected]

Flavio Calvano Università degli Studi di Napoli Federico II Via Claudio 21 80125 Naples Italy [email protected]

Juan Calzada Air Force Research Laboratory 2230 Tenth Street 45433 WPAFB USA [email protected]

x

Klára Čápová DEBE, Faculty of Electrical Engineering, University of Žilina Univerzitná 1 SK-01026 Žilina Slovak Republik [email protected]

Tomasz Chady Department of Electrical Engineering West Pomeranian University of Technology, Szczecin ul. Sikorskiego 37 70-313 Szczecin Poland [email protected]

Zhenmao Chen Key Laboratory for Strength and Vibration, School of Aerospace, Xi’an Jiaotong University 28 West Xianning Road 710049 Xi-an China [email protected]

Wojciech Chlewicki Department of Electrical Engineering West Pomeranian University of Technology, Szczecin ul. Sikorskiego 37 70-313 Szczecin Poland [email protected]

Georgios Christofi Department of Physics, University of Cyprus Building:ΘΕΕ 02,20537 1676 Nicosia Cyprus [email protected]

Thanh Long Cung SATIE – ENS Cachan 61 avenue du président Wilson 94230 Cachan France [email protected]

Rémi Douvenot L2S-DRE (CNRS-SUPELEC-UNIV.PARIS SUD) Joliot Curie 3 91192 Gif-sur-Yvette France [email protected]

Dagmar Faktorová University of Žilina Univerzitná 1 SK-01026 Žilina Slovak Republik [email protected]

Mouloud Feliachi IREENA CRTT, Bd de l’Université 44600 Saint Nazaire France [email protected]

Emna Amira Fnaiech CEA List Centre Saclay; Bât 611 PC120 91191 Gif-sur-Yvette France [email protected]

Bryan Foos First Principles Inc. 800 Harman Ave 45419-3432 Oakwood USA [email protected]

Paweł Frankowski Department of Electrical Engineering West Pomeranian University of Technology, Szczecin ul. Sikorskiego 37 70-313 Szczecin Poland [email protected]

xi

Konstanty Marek Gawrylczyk Department of Electrical Engineering West Pomeranian University of Technology, Szczecin ul. Sikorskiego 37 70-313 Szczecin Poland [email protected]

Stanisław Gratkowski Department of Electrical Engineering West Pomeranian University of Technology, Szczecin ul. Sikorskiego 37 70-313 Szczecin Poland [email protected]

Raimond Grimberg National Institute of Research and Development for Technical Physics 47 D.Mangeron 700050 Iasi Romania [email protected]

Gyimóthy Szabolcs Budapest University of Technology and Economics Goldman Gy. tér 3 1111 Budapest Hungary [email protected]

Ammar Hamel A/Mira University, Bejaïa Route de Targa Ouzemour 06000 Bejaïa Algeria [email protected]

He Dongfeng National Institute for Materials Science Sengen 1-2-1 305-0047 Tsukuba Japan [email protected]

Tommy Henrisson L2S-DRE (CNRS-SUPELEC-UNIV.PARIS SUD) Joliot Curie 3 91192 Gif-sur-Yvette France [email protected]

Susanne Hillmann Fraunhofer IZFP, Dresden Maria-Reiche-Str. 2 01109 Dresden Germany [email protected]

Satoshi Honda Faculty of Science & Technology, Keio University Hiyoshi 3-14-1, Kohoku 223-8522 Yokohama Japan [email protected]

Huang Haiying University of Texas at Arlington 500 W. First Street, WH211 76019 Arlington USA [email protected]

Moayyed Hussain Benet Laboratoris Watervliet Arsel 12189 Watervliet USA [email protected]

Ladislav Janoušek DEBE, Faculty of Electrical Engineering, University of Žilina Univerzitná 1 SK-01026 Žilina Slovak Republik [email protected]

xii

David Jiles Cardiff University CF24 3AA Cardiff United Kingdom [email protected]

Hiroaki Kikuchi Iwate University 4-3-5 Ueda 020-8551 Morioka Japan [email protected]

Kim Jungmin Chosun University Seosukdong 375 501-759 Gwangju Korea [email protected]

Kiss Imre Budapest University of Technology and Economics Egry József út 18 1111 Budapest Hungary [email protected]

Jeremy Knopp AFRL/RXLP WPAFB, OH, 45433 [email protected]

Satoru Kobayashi Iwate University 4-3-5 Ueda 020-8551 Morioka Japan [email protected]

Fumio Kojima Kobe University Rokkodai 1-1, Nada 657-8501 Kobe Japan [email protected]

Eugeniusz Kornatowski Department of Electrical Engineering West Pomeranian University of Technology, Szczecin ul. Sikorskiego 37 70-313 Szczecin Poland [email protected]

Jacek Kowalczyk Department of Electrical Engineering West Pomeranian University of Technology, Szczecin ul. Sikorskiego 37 70-313 Szczecin Poland [email protected]

Marc Kreutzbruck BAM Federal Institute for Materials Research and Testing Unter den Eichen 87 12205 Berlin Germany [email protected]

Marc Lambert L2S-DRE (CNRS-SUPELEC-UNIV.PARIS SUD) Joliot Curie 3 91192 Gif-sur-Yvette France [email protected]

Brian Larson Iowa State University 175 ASC II 50011 Ames USA [email protected]

xiii

Lee Jinyi Chosun University Seosukdong 375 501-759 Gwangju Korea [email protected]

Dominique Lesselier L2S-DRE (CNRS-SUPELEC-UNIV.PARIS SUD) Joliot Curie 3 91192 Gif-sur-Yvette France [email protected]

Przemysław Łopato Department of Electrical Engineering West Pomeranian University of Technology, Szczecin ul. Sikorskiego 37 70-313 Szczecin Poland [email protected]

Lim Tekoing CEA Saclay Centre Saclay; Bât 611 PC120 91191 Gif-sur-Yvette France [email protected]

Steve Mahaut CEA Saclay Centre Saclay; Bât 611 PC12D 91191 Gif-sur-Yvette France [email protected]

Kenzo Miya Professor-emeritus of the University of Tokyo President of Japan Society of Maintenology (JSM) http://www.jsm.or.jp/jsm/en/index.html Editor in Chief of E-Journal of Advanced Maintenance (EJAM) http://www.jsm.or.jp/ejam [email protected]

Hassane Mohellebi Electrical Engineering Laboratory Tizi-Ouzou University BP17RP 15000 Tizi-Ouzou Algeria [email protected]

Lech Napierała Department of Electrical Engineering West Pomeranian University of Technology, Szczecin ul. Sikorskiego 37 70-313 Szczecin Poland [email protected]

Yoshihiro Nishimura National Institute of Advanced Industrial Science and Technology 1-2-1 Namiki 305-8564 Tsukuba Japan [email protected]

Youngmin Park Chosun University Seosukdong 375 501-759 Gwangju Korea [email protected]

Jozsef Pavo Budapest University of Technology and Economics Goldman Gy. tér 3 1111 Budapest Hungary [email protected]

Cuixiang Pei Uesaka-Demachi Laboratary, The University of Tokyo 2-11-16, Yayoi, Bunkyoku 113-0032 Tokyo Japan [email protected]

xiv

Matthias Pelkner BAM Federal Institute for Materials Research and Testing Unter den Eichen 87 12205 Berlin Germany [email protected]

Tomasz Pietrusewicz Department of Electrical Engineering West Pomeranian University of Technology, Szczecin ul. Sikorskiego 37 70-313 Szczecin Poland [email protected]

Denis Prémel CEA List Centre Saclay; Bât 611 PC120 91191 Gif-sur-Yvette France [email protected]

Grzegorz Psuj Department of Electrical Engineering West Pomeranian University of Technology, Szczecin ul. Sikorskiego 37 70-313 Szczecin Poland [email protected]

Joao Rebello Federal University of Rio de Janeiro Ilha do Fundao CT Room F-210 21941-972 Rio de Janeiro Brasil [email protected]

Verena Reimund BAM Federal Institute for Materials Research and Testing Unter den Eichen 87 12205 Berlin Germany [email protected]

Maciej Roskosz Silesian University of Technology ul. Akademicka 2A 44-100 Gliwice Poland [email protected]

Guglielmo Rubinacci Università degli Studi di Napoli Federico II Via Claudio 21 80125 Naples Italy [email protected]

Yasutomo Sakai Tohoku University 6-6-01-2, aramaki aza aoba, aoba ward 980-8579 Sendai Japan [email protected]

Adriana Savin National Institute of Research and Development for Technical Physics 47 D. Mangeron 700050 Iasi Romania [email protected]

Hans-Peter Schmidt UAS Amberg Weiden (HAW) Kaiser Wilhelm Ring 23 D-92224 Amberg Germany [email protected]

Karl Schmidt Evisive, Inc., 8867 Highland Road, #378, Baton Rouge, Louisiana, 70808 USA [email protected]

xv

Jan Sikora Lublin University of Technology Electrical Engineering and Computer Science Faculty 20-618 Lublin 38A Nadbystrzycka Str. Poland [email protected]

Ryszard Sikora Department of Electrical Engineering West Pomeranian University of Technology, Szczecin ul. Sikorskiego 37 70-313 Szczecin Poland [email protected]

Christian Seibold UAS Amberg Weiden (HAW) Kaiser Wilhelm Ring 23 D-92224 Amberg Germany [email protected]

Kisu Shin Korea Nationa Defense University 205 Susaek-dong 122-875 Seoul Korea [email protected]

Marek Smaga University of Kaiserlautern Gottlieb-Daimler-Straße 67663 Kaiserlautern Germany [email protected]

Milan Smetana DEBE, Faculty of Electrical Engineering, University of Žilina Univerzitná 1 SK-01026 Žilina Slovak Republik [email protected]

Thierry Sollier IRSN 31, av. de la Division Leclerc BP 17 92262 Fontenay-aux-Roses France [email protected]

Krzysztof Stawicki Department of Electrical Engineering West Pomeranian University of Technology, Szczecin ul. Sikorskiego 37 70-313 Szczecin Poland [email protected]

Tatiana Strapáčová DEBE, Faculty of Electrical Engineering, University of Žilina Univerzitná 1 SK-01026 Žilina Slovak Republik [email protected]

Barbara Szymanik Department of Electrical Engineering West Pomeranian University of Technology, Szczecin ul. Sikorskiego 37 70-313 Szczecin Poland [email protected]

Toshiyuki Takagi Tohoku University Katahira 2-1-1, Aoba-ku 980-8577 Sendai Japan [email protected]

Antonello Tamburrino University of Cassino V.G. Di Biasio, 43 03043 Cassino Italy [email protected]

xvi

Theodoros Theodoulidis University of Western Macedonia Bakola & Sialvera 50100 Kozani Greece [email protected]

Tetsuya Uchimoto Tohoku University 2-1-1 Katahira, Aoba-ku, Sendai, Miyagi 980-8577 Sendai Japan [email protected]

Lalita Udpa Michigan State University 2100C Engineering Building 48824-1226 East Lansing USA [email protected]

Satish S. Udpa Michigan State University 3410 Engineering Building 48824-1226 East Lansing USA [email protected]

Elmar van den Elzen Lismar Engineering BV Ambachtenstraat 55 1191JM Ouderkerk a.d. Amstel Netherlands [email protected]

Vertesy Gabor Research institute for Technical Physics and Materials Science Konkoly Thege Miklós ut 29-33 H-1121 Budapest Hungary [email protected]

Xie Shejuan Tohoku University 2-1-1 Katahira, Aoba-ku, Sendai, Miyagi 980-8577 Sendai Japan [email protected]

Katsuhiko Yamaguchi Fukushima University 1 Kanayagawa 960-1296 Fukushima Japan [email protected]

Toshihiro Yamamoto JAPEIC 14-1 Benten-cho, Tsurumi-ku 230-0044 Yokohama Japan [email protected]

Noritaka Yusa Tohoku University 6-6 Aramaki, Aza Aoba Aoba, Sendai 980-8579 Miyagi Japan [email protected]

Marcin Ziółkowski Department of Electrical Engineering West Pomeranian University of Technology, Szczecin ul. Sikorskiego 37 70-313 Szczecin Poland [email protected]

Chiara Zorni CEA Saclay Centre Saclay; Bât 611 PC120 91191 Gif-sur-Yvette France [email protected]

Marek Ziółkowski Ilmenau Univeristy of Technology Helmholtzplatz 2, P.O.Box 100565 98684 Ilmenau Germany [email protected]

xvii

Contents Preface T. Chady, S. Gratkowski, T. Takagi and S.S. Udpa Conference Organization List of Participants

v vii ix

Keynote Lecture Pipe Wall Thickness Inspection with Current Driven Thermal Method Satoshi Honda

3

Modeling and Inverse Problems Modeling and Signal Processing Sensor Tilt in Eddy Current-GMR Inspection G. Yang, Z. Zeng, L. Udpa and S.S. Udpa Non-Iterative MUSIC-Type Algorithm for Eddy-Current Nondestructive Evaluation of Metal Plates Tommy Henriksson, Marc Lambert and Dominique Lesselier Identification of Defects in 3D Space Using Computer Radiography System Wojciech Chlewicki, Piotr Baniukiewicz, Tomasz Chady and Andrzej Brykalski A Thin-Skin Model for Eddy-Current NDE of Cracks in a Borehole S.K. Burke Magnetic Response Field of Spherical Defects Within Conductive Components M. Kreutzbruck, H.-M. Thomas, R. Casperson, V. Reimund and M. Pelkner Decreasing Uncertainty in Size Estimation of Stress Corrosion Cracking from Eddy-Current Signals Ladislav Janousek, Milan Smetana and Marcel Alman Semi-Discrete Time-Domain Sensitivity Analysis for Cracks Recognition Konstanty M. Gawrylczyk Metamodel as Input of an Optimization Algorithm for Solving an Inverse Eddy-Current Testing Problem Rémi Douvenot, Marc Lambert and Dominique Lesselier Fast Multipole Method for 3D Electromagnetic Boundary Integral Equations. Application to Non Destructive Testing on Complex 3D Geometries Tekoing Lim, Gregoire Pichenot and Marc Bonnet Computation of the Magnetostatic Field in Nonlinear Media via the Integral Equation Formalism: Application to the Characterization of Magnetic Flux Leakage NDT System Emna Amira Fnaiech, Denis Prémel, Claude Marchand and Bernard Bisiaux Efficient Computation of Eddy Current Crack Signals Theodoros Theodoulidis

13

22 30

36 44

53 61

71

79

87

96

xviii

Experimental Validation of a Fast Non-Iterative Imaging Algorithm for Eddy Current Tomography Salvatore Ventre, Flavio Calvano, Guglielmo Rubinacci and Antonello Tamburrino Numerical Evaluation of Microwave Testing for Pipe Thinning Yasutomo Sakai, Noritaka Yusa and Hidetoshi Hasizume Parallelization of Crack Signal Calculation Using CUDA Imre Kiss, József Pávó and Szabolcs Gyimóthy Electromagnetic-Acoustic Transducers Modeling in Time Domain Using the Finite Element Method Marcin Ziolkowski and Stanislaw Gratkowski Eddy Current Testing of Ferromagnetic Materials: Modelling of Flaws in a Planar Medium Chiara Zorni, Christophe Reboud, Marc Lambert and Jean-Marc Decitre Efficient Propagation of Uncertainty in Simulations via the Probabilistic Collocation Method Jeremy S. Knopp, John C. Aldrin and Mark P. Blodgett Computation of the Magnetic Field Due to a Defect Embedded in a Planar Stratified Media: Application to AC Field Measurement Techniques Denis Prémel and Grégoire Pichenot Automatic Detection and Identification of Image Quality Indicators in Radiograms Piotr Baniukiewicz and Ryszard Sikora Numerical Simulation of Acoustoelastic Effect in Pre-Stressed Media Cuixiang Pei and Kazuyuki Demachi “BEMLAB” – Universal, Open Source, Boundary Element Method Library Applied in Micro-Electro-Mechanical Systems Paweł Wieleba and Jan Sikora Use of Half-Analytical Method for the Detection of Defects in Diet Pulses Hassane Mohellebi, Ferroudja Bouali and Mouloud Feliachi Imperialist Competitive Algorithm Applied to Eddy Current Nondestructive Evaluation Ammar Hamel, Hassane Mohellebi and Mouloud Feliachi

103

111 117

125

134

141

149

157 164

173 183

192

Transducers and Techniques Pulsed ECT Method for Evaluation of Pipe Wall-Thinning of Nuclear Power Plants Using Magnetic Sensor Shejuan Xie, Toshihiro Yamamoto, Toshiyuki Takagi and Tetsuya Uchimoto Developing ECT System with AMR Sensor for Combustion Chamber Dongfeng He, Mitsuhara Shiwa, Jianping Jia, Hisashi Yamawaki, Ichizo Uetake, Junji Takatsubo, Shinichi Moriya and Koichi Okita Flux Leakage Measurements for Defect Characterization Using NDT Adapted GMR Sensors Matthias Pelkner, Andreas Neubauer, Mark Blome, Verena Reimund, Hans-Martin Thomas and Marc Kreutzbruck

203

211

217

xix

Evaluation of EMAT Signals Using Magnetostrictives for Imaging Yoshihiro Nishimura, Akira Sasamoto and Takayuki Suzuki Pulsed Eddy Current Testing – Application to Defect Evaluation Milan Smetana, Ladislav Janousek, Klara Capova and Maria Michniakova An ECT Probe with Widely Spaced Coils for Local Wall Thinning in Nuclear Power Plants Toshihiro Yamamoto, Tetsuya Uchimoto and Toshiyuki Takagi PLL Based Eddy Current Measuring System for Inspection of Outer Flaw in Titanium Alloy Plate Tomasz Chady, Jacek Kowalczyk, Leon Nawos-Wysocki, Grzegorz Psuj and Ireneusz Spychalski Giant Magneto-Resistive Sensor Based Magnetic Flux Leakage Technique for Inspection of Track Ropes W. Sharatchandra Singh, B.P.C. Rao, S. Mahadevan, T. Jayakumar and Baldev Raj

225 233

241

249

256

Application of Electromagnetic Nondestructive Techniques Pipe Wall Thickness Measurement by Electro-Magnetic Acoustic Transducer Using Band Exciting Method Daigo Kosaka, Fumio Kojima and Kosuke Umetani Applicability of Magnetic Flux Leakage Method for Wall Thinning Monitoring in Nuclear Power Plants Hiroaki Kikuchi, Isamu Shimizu, Katsuyuki Ara, Yasuhiro Kamada and Satoru Kobayashi Stress Corrosion Cracks Evaluation in 316 Austenitic Stainless Steel Plate Tomasz Chady, Jacek Kowalczyk and Grzegorz Psuj Multi-Frequency Eddy Current NDE of the Distance Between Parts in Aeronautical Assemblies Thanh Long Cung, Pierre Yves Joubert and Eric Vourc’h Artificial Heart Valve Testing Using Electromagnetic Method Tatiana Strapacova, Klara Capova, Ladislav Janousek and Milan Smetana Conceptual Design of an Industrial System for Automatic Radiogram Analysis Ryszard Sikora, Tomasz Chady, Piotr Baniukiewicz, Marcin Caryk, Przemysław Łopato, Lech Napierała, Tomasz Pietrusewicz and Grzegorz Psuj

267

275

282

288 296 302

Material Characterization Quantification of Sigma Phase Precipitation by Magnetic Non Destructive Testing João M.A. Rebello, Rodrigo Sacramento, Maria C.L. Areiza and K. Santos de Assis Nondestructive Characterization of Neutron Induced Embrittlement in Nuclear Pressure Vessel Steel Microstructure by Using Electromagnetic Testing I. Altpeter, G. Dobmann, G. Hübschen, M. Kopp and R. Tschuncky

311

322

xx

In-Line Thin Film Characterization Using Eddy Current Techniques Susanne Hillmann, Marcus Klein and Henning Heuer Feasibility of Stress State Assessment on the Grounds of Measurements of the Strength of the Residual Magnetic Field of Ferromagnetics Maciej Roskosz Magnetic Characterization of Material Degradation Using Dynamical Minor Loops Satoru Kobayashi, Shinichi Tsukidate, Hiroyuki Okazaki, Yasuhiro Kamada, Hiroaki Kikuchi and Toshihiro Ohtani Local Magnetic Properties and Magnetic Particle Distribution Due to Cr Depletion in Sensitized Ni Based Alloy K. Yamaguchi, K. Suzuki, T. Takase, T. Takara, O. Nittono, T. Uchimoto and T. Takagi Metamaterials in Electromagnetic Nondestructive Evaluation Adriana Savin, Rozina Steigmann, Alina Bruma, Nicoleta Iftimie and Raimond Grimberg Evaluation of Plastic Deformation in Steels by Magnetic Hysteresis Measurements Gábor Vértesy, Shuzo Ueda, Tetsuya Uchimoto, Toshiyuki Takagi, Ivan Tomáš and Zofia Vértesy

330

339

347

355

362

371

High Frequency Techniques and Others GPR for UXO Recognition Raimond Grimberg, Adriana Savin, Nicoleta Iftimie, Sorin Leitoiu and Aurel Andreescu Application of Microwave Interferometry in Complex Engineered Dielectric Materials Karl Schmidt and Jack Little Detection of Concealed, Graphite Containing Frescos, Using Microwave Enhanced Infrared Thermography Paweł Lesiecki and Barbara Szymanik Solid Materials Complex Permittivity and Inhomogenities Determination in High Frequency Range Dagmar Faktorová Terahertz Time Domain Inspection of Composite Coatings for Corrosion Protection Przemyslaw Lopato, Tomasz Chady and Joao M.A. Rebello Simulation Assisted Diagnostic of Switching Arcs Hans-Peter Schmidt, Michael Anheuser and Sylvio Kosse Studies on Analysis Technique of Commutation Phenomena for Cleaner’s Universal Motor Using the State Variable Method Yuji Akiyama and Yuta Niwa Propositions for the Analysis of Commutation Phenomena of Universal Motors Using the State Function Method Yuji Akiyama and Yuta Niwa

381

389

402

410

417 425

432

439

xxi

Detection of CFRP’s Degradation Through Electromagnetic Procedures Bogdan Serghiac, Daniel Petrica Salavastru, Paul Doru Barsanescu, Adriana Savin, Rozina Steigmann and Raimond Grimberg Microwave Antenna Sensors for Fatigue Crack Monitoring Under Lap-Joints Justin Erdmann and Haiying Huang Subject Index Author Index

447

456 467 471


Keynote Lecture


Electromagnetic Nondestructive Evaluation (XIV) T. Chady et al. (Eds.) IOS Press, 2011 © 2011 The authors and IOS Press. All rights reserved. doi:10.3233/978-1-60750-750-5-3

3

Pipe Wall Thickness Inspection with Current Driven Thermal Method Satoshi HONDA a,1 a

Faculty of Science & Engineering, Keio University, Japan

Abstract. Electric potential difference(EPD) method is widely used for diagnosing and monitoring flow conduits. Theoretical analysis and the exact solution of the electrical potential field in pipe wall have been given, and 2D approximation is effective for small pipe wall thickness there. This study proposes to use temperature distribution driven by Joule heat with the electric current other than electrical potential difference. Under thin wall approximation, 2D electric and thermal problems are analyzed and exact solutions are given. Keywords. Nondestructive test, pipe wall thinning, forward analysis, electric potential difference

Introduction Electric potential difference(EPD) method is widely used for diagnosing and monitoring flow conduits [1]. It applies large electric currents between a pair of electrodes on pipe wall surface, and the electrical potentials are measured with multiple potential electrodes. Since the localized pipe wall thinning or a wall crack deforms the potential field, we can monitor and/or diagnose pipe wall condition under test. The method is also required to evaluate pipe wall thinning in larger span. We have studied the prospect of the method to evaluate pipe wall thinning. Based on the work on non-intrusive resistance thermometer for fast breeder reactor[2,3], theoretical analysis and the exact solution of the electrical potential field in pipe wall have been proposed[4], and 2D approximation is effective for small pipe wall thickness. We propose in the paper to use temperature distribution driven by Joule heat with the electric current other than electrical potential difference. Under thin wall approximation, 2D electric and thermal problems are analyzed and exact solutions are given. 1 Corresponding Author: Satoshi Honda, Faculty of Science & Technology, Keio University, Hiyoshi 3-14-1, Kohoku, Yokohama 223-8522, Japan; Email: [email protected]

4

S. Honda / Pipe Wall Thickness Inspection with Current Driven Thermal Method

1. Analysis of Wall Potential Problem for EPD Method 1.1. 3D Analysis Let the electric potential inside the fluid be φ1 (ρ, ϕ, z), and the one inside the pipe wall, φ2 (ρ, ϕ, z), where cylindrical coordinates are adopted with pipe axis as z coordinate. Electrical current is fed between a pair of rectangular electrodes of bw×bw on the outer surface (b, 0, 0). If we let σ1 and σ2 be electrical conductivities of the fluid and the pipe wall, the following potential problem is formulated. Δφ2 = 0,

a ! ;);7 , 5 9, 7

- 5 7 & 8 . - 5 -7 ';; ?@7 @A ?(? '2 2 7 8 7 32 ;);B2 33,

126

M. Ziolkowski and S. Gratkowski / EMATs Modeling in Time Domain

. 4 5 5 7 1 . 5 - < 5 - , - - , 5 1C 3 ,2

4 4 , /

- 4 0 1 )

, = 5

- , 15 , , = 5 1 3 D- ,1 1 2 4 , . E(7 :F3

. 7 ) . . 7 , - , 5 4 1 5 5 - ) , , 3 - 5

).

5 5 1 ,, 1 5 , 3 ) .

, 7 33 $ 2 5 7 2 5 7 . 5 / ) ; , - 5 5 0 -

/ 0 E?GHF3 , ). 5 7 - 7

. -3 " ,, . ,2 55 7

- - ) , 1 , 3 . . -, 5 1 ;

).3 " ,, ) )

55 , 5 5 5 , 7 , 7 33 8 8 $ 9

3

) 2

- /'I0 , 2 ).7 ) ,. . ) . - 5 5 E(G@F3 ' , 7 - $ 2 5 )

1 3 1

. 1 , - , -

-3 ' 5 )

) 5 5- , ). 7 ) ,, ,

3 " 5 7 ;) ,, , 3 1XPHULFDOPRGHOGHVFULSWLRQ

- 5 ) < , 5 1 5 3 - 7 , ) 5 5 ) 5 ,, 3 $ 2 5 ) .

5 )

;)

)

- u%

/(0

)> - J - - . , 7 % J 5 < . 5 , 3

127


330(0$7 5 ) 3( 5 ,

. - 8 8 /8803 ;

] , 5 - - . 4 27 -] K A7 , 5 5 ) ; 5> )[ K -\%]7

)\ K G -[%]7

) ] K - [% \ G - \% [

/:0

)LJXUH ) 8 8 ,

, 3

< 1- 5 1 - E:F>

,

:SIW · § 5 W d 1 I ° , A :SIW ¨( ¸7 ® 1 ¹ © °¯ A7 5 W ! 1 I

/?0

" 3: , 5 < 54 - , 1 )7 5 7 , . -3

)LJXUH # 5 < 7 54 - , L I - 3

M(HAA I27 ,A

:A 7 1

H

128


, -7 5 < 1 3

7 , . 1 2 1 (7 : ?7 , . -3 7DEOH , 5HPDQHQWIOX[GHQVLW\RIPDJQHW%U &XUUHQW, 6SHFLPHQPDJQHWLFUHODWLYHSHUPHDELOLW\ȝU 6SHFLPHQFRQGXFWLYLW\ı

( :A ( H3@N(A@

EF EF EF E'OF

7DEOH , (3 . 5 5 - 5 5 , 7 :3 . . 5 5 < - 1 , ) 5 5 , 3

)LJXUH 5 5 7 5 3 S3

5 5 , / 0 -3

236

M. Smetana et al. / Pulsed Eddy Current Testing – Application to Defect Evaluation

1XPHULFDOVLPXODWLRQUHVXOWV D , 5 . . 1 . 5 , 3

) -, 5 . . 1 > 4 /'V07 /+*07 (OS ). /(OS'"60 8 % ) . 3 , 5 55 , . , 5 . .- 3 " , 5 . 5 - 5 55 , 3 T 5 - ,5 7 -

5 8 % . ) ,N 5 W K ( 7 3 ?7 , 1 )3 " , ) 1. . ,7 8 %

. , 7 N 5

\ W DWH

W : E

7

/(0

) D E 3 . . /"07 "''6 (?M? @:M(7 "#' ,7 D 3 ?(7 :AAM3 ESF Q3 &3 & 7 *3U3 > 3XOVHG HOHFWURPDJQHWLF PDWKRGV IRU GHWHFW GHWHFWLRQ DQG FKDUDFWHULVDWLRQ7 6~ 7 . 3 SA7 . $ 37 :AA@7 E# F3


241

An ECT Probe with Widely Spaced Coils for Local Wall Thinning in Nuclear Power Plants Toshihiro YAMAMOTOa, Tetsuya UCHIMOTOb and Toshiyuki TAKAGIb,1 a Japan Power Engineering and Inspection Corporation b Institute of Fluid Science, Tohoku University

Abstract. Currently, local wall thinning on the inner surface of a pipe is evaluated mostly by ultrasonic thickness measurement from the outer surface of the pipe in nuclear power plants in Japan. However, it has been pointed out that there are two major issues of this evaluation method. Firstly, this method is not able to evaluate local wall thinning from the top of a reinforcing plate that covers the outer surface of a pipe because this two layer structure hinders the propagation of ultrasound. Secondly, because ultrasonic thickness measurement of a pipe wall is conducted at certain intervals on the outer surface of the pipe, it may overlook a hole like defect if the defect size is small compared to this interval. We propose an eddy current testing (ECT) method using widely spaced excitation and pick up coils to alleviate these problems. Our experimental results show this method can be applied to in spection of a double plate that consists of two 8 mm thick austenitic stainless steel plates, and a defect indication is observed in a much larger area than the defect. Keywords. Eddy current testing, Local wall thinning, Pancake coil

1. Introduction In nuclear power plants, local wall thinning on the inner surface of a pipe is evaluated mostly by ultrasonic thickness measurement from the outer surface of the pipe. However, some improvements are still required for thorough and efficient inspection of local wall thinning to maintain aging nuclear power plants. In Japan, it is considered a problem that ultrasonic thickness measurement is not able to evaluate local wall thinning from the top of a reinforcing plate that covers the outer surface of a main pipe around a branch pipe perpendicularly connected to the main pipe because this two-layer structure hinders the propagation of ultrasound (Figure 1). Also, it is of special concern that ultrasonic thickness measurement may overlook a small hole caused by liquid droplet impingement (LDI) because ultrasonic thickness measurement of a pipe wall is conducted at certain intervals on the outer surface of the pipe and the defect size is sometimes smaller than this interval. Decreasing the interval of course reduces the incidence of this problem, but piping maintenance in nuclear power plants already takes an enormous amount of time, and inspection time is not easily allowed to be extended any longer. The aim of this study is to propose an inspection method for two1

Corresponding Author: Professor, Institute of Fluid Science, Tohoku University, 2 1 1 Katahira, Aoba ku, Sendai, Miyagi, Japan; E mail: [email protected]

242

T. Yamamoto et al. / An ECT Probe with Widely Spaced Coils for Local Wall Thinning

Figure 1. Reinforcing plate

Figure 2. Absolute probe

Figure 3. Inspection procedure

layer part of piping and to prevent overlooking a hole-like defect on a pipe. The authors have recently reported an eddy current testing (ECT) technique for evaluating local wall thinning [1]. In this technique, an excitation coil and a pick-up coil are aligned at a distance to detect a defect in a deep region after the model of remote field ECT usually used for inspection of small tubes with bobbin coils aligned at a distance [2,3]. Prior to our study, remote field ECT using pancake coils to be applied to flat geometries has been introduced in [4,5], later specialized in inspection of aircrafts. In this study, we assess the applicability of our ECT technique to detection of local wall thinning of a thick-walled double plate and a hole-like defect on a thick-walled single plate.

2. Experiments 2.1. Experimental setup The ECT probe we used in this study has one excitation coil and one pick-up coil shown in Figure 2. The excitation and pick-up coils are 10 mm in outer diameter, 5 mm in inner diameter and 5 mm in height. The number of turns of the excitation and pickup coils is 301 and 3,465 respectively. The wire diameter of the excitation and pick-up coils is 0.2 mm and 0.05 mm respectively. The center-to-center distance between these coils was set to be 50 mm. The equipment used in our experiments consists of a function synthesizer (WF1945B, NF Corporation), a bipolar amplifier (HSA4011, NF Corporation) and a lock-in amplifier (LI5640, NF Corporation). In the experiments, the detection signals were measured while the pair of excitation and pick-up coils was moved on the surface of a plate as shown in Figure 3. The excitation frequency was set to be 1 kHz. 2.2. Specimens Tables 1 and 2 show the information of the specimens used for evaluation of the per-


243

Table 1. Rectangular slit specimens Material

316 austenitic stainless steel

Size

500 mm×300 mm×8 mm

Defect

Rectangular slit (Width 10 mm, Depth 1,3,5 mm) Table 2. Square hole specimens

Material

316 austenitic stainless steel

Size

500 mm×300 mm×12 mm

Defect

Square hole (Bottom 10 mm×10 mm, Depth 1,3,5 mm)

Figure 4. Double plate

(a) Plate with a rectangular slit

(b) Plate with a square hole Figure 5. Specimens

formance of our ECT technique. The material of these plates is 316 austenitic stainless steel. A reinforcing plate part of piping was simulated by two 8 mm-thick plates. An intact plate was placed on the top of the other plate that has a rectangular slit on the bottom side as shown in Figure 4. The rectangular slit was cut along a half of the center line on the bottom side in the shorter direction (Figure 5 (a)). A defect caused by LDI was simulated by a square hole. The square hole was made at the center of a 12 mmthick plate (Figure 5 (b)). For each of the rectangular slit and the square hole, three plates were prepared so as to vary the depth of these simulated defects (1 mm, 3 mm and 5 mm). 2.3. Evaluation procedure The specimens described above were inspected from the opposite side of the defect to be detected. Figure 6 shows how we scanned the surface of the specimens. For a double plate with a rectangular slit, five equally-spaced lines along the longer direction of the plate were selected as the scanning lines. Each of the scanning lines has a length of 150

244


(a) Rectangular slit

(b) Square hole Figure 6. Scanning areas

(a) Original signals

(b) Processed signals Figure 7. Signal processing

(a) Original distribution

(b) Processed distribution

Figure 8. Example of signal processing

mm at 1 mm intervals. The left three lines run above the rectangular slit while the other two do not. For a square hole, a 150 mm×60 mm area at the center of the surface of the plate was scanned. The intervals of scanning were 1 mm in the longer direction and 2 mm in the shorter direction of the plate. 2.4. Signal processing to reduce lift off noise Because the specimens used in this evaluation have a wide-area surface (500 mm×300 mm) and also have a slight warp, the lift-off distance between the probe and the surface


245

of the specimens varies with the position of the probe, which causes considerable liftoff noise. To cope with this problem, we processed raw signals obtained from the lockin amplifier. Figure 7 (a) shows the Vx and Vy components of ECT signals obtained from one of the scanning lines going over a rectangular slit in Figure 6 (a). It can be observed that the rectangular slit induces a big change in the signals around the center position. Although the end point of the scanning line is far enough away from the slit, the signal values do not return to zero at the end point. To make the signal values zero at both ends of the scanning line, the voltage value on the line that connects both ends of a signal curve in Figure 7 (a) is subtracted from the voltage value on that signal curve for each point. Figure 7 (b) is obtained in consequence of this processing. For a signal distribution obtained as described in Figure 6 (b), the above processing is applied in the X direction as well as in the Y direction so as to make the values on the boundary of the scanning area zero. Figure 8 shows an example of this signal processing. The removal of gradual changes in the signal distribution highlights a defect indication due to a square hole. 2.5. Results Figure 9 shows the signal distributions obtained from the specimens with a rectangular slit. To simulate a reinforcing plate part, an intact plate was placed on the top of a plate with a rectangular slit as shown in Figure 4. The results indicate that our ECT tech-

(a) 1 mm deep slit

(b) 3 mm deep slit

(c) 5 mm deep slit

Figure 9. Signal amplitude distributions (rectangular slit)

(a) 1 mm deep hole

(b) 3 mm deep hole

(c) 5 mm deep hole

Figure 10. Signal amplitude distributions (square hole)

246


nique can detect a 1 mm-deep defect on the bottom side of two 8 mm-thick plates. Figure 10 shows the signal distributions obtained from the specimens with a square hole. A 1 mm-deep square hole on the bottom side of a 12 mm-thick plate is clearly recognized in the signal distributions. Defect signals are observed in a much larger area than the defect (The defect area of a square hole is 10 mm×10 mm. The area where a defect indication appears due to a square hole is about 20 mm×40 mm). This property helps prevent overlooking a small defect. Figure 11 shows the relationship between the maximum signal amplitude and the defect depth obtained from Figures 9 and 10. There is an almost linear relationship between them. It implies that the defect depth can be estimated from the maximum amplitude of a defect indication. In above experiments, an LDI defect is simulated by a square hole, but a real LDI defect tends to have a cone shape. Then, a specimen with a cone-shaped hole shown in Figure 12 was also used for evaluation of the detectability of an LDI defect with our ECT technique. The specimen is 300 mm in length, 100 mm in width and 7.1 mm in thickness. The base diameter and the height of the cone-shaped hole are 10.2 mm and 3.6 mm respectively. Because the surface of this specimen is not large enough, the scanning area was reduced from the area described in Figure 6 (b) by decreasing the width in the X direction from 60 mm to 40 mm. Figure 13 (a) shows the signal amplitude distribution obtained by this scanning. The appearance of the defect indication is almost the same as Figure 10, which was obtained with square holes. Figure 13 (b) provides the profiles on the lines X=0 and Y=0 in the amplitude distribution. Whereas the defect is located from -5 mm to 5 mm in both the X and Y directions, a big variation in signal amplitude is still observed at ±10 mm in the X direction and at ±20 mm in the Y direction. This indicates that a defect can be found even if the scanning line is 5 mm away from the edge of the defect with the coils aligned parallel to the scanning line. When the coils are aligned perpendicular to the scanning line, the defect can be found even if the scanning line is 15 mm away from the edge of the defect (Figure 14).

(a) Rectangular slit

(b) Square hole

Figure 11. Relationship between the maximum signal amplitude and the defect depth

Figur 12. Specimen with a cone shaped hole


(a) Amplitude distribution

247

(b) Amplitude profiles in the X and Y directions

Figure 13. Signal amplitude distribution (cone shaped hole)

(a) Parallel to the scanning line

(b) Perpendicular to the scanning line

Figure 14. Scanning directions

3. Summary The aim of this study is to propose an inspection method to evaluate local wall thinning on a reinforcing plate part of piping and detect a defect located out of the scanning line. We proposed an ECT technique using widely spaced excitation and pick-up coils to detect a defect on the bottom side of a thick-walled plate. The experimental results show this probe can detect a 1 mm-deep rectangular slit on the bottom side of two 8 mm-thick austenitic stainless steel plates and a 1 mm-deep square hole on the bottom side of a 12 mm-thick austenitic stainless steel plate. There is an almost linear relationship between the maximum signal amplitude and the defect depth. This implies that the defect depth can be estimated from the maximum signal amplitude due to a defect. These defect signals are observed in a much larger area than the defect size. When the coils are aligned perpendicular to the scanning line, this method can detect a defect located within 15 mm from the scanning line. The proposed method can be applied to two-layer structure and detects a defect even if the defect is not on the scanning line. These properties are expected to alleviate concern over ultrasonic thickness measurement conducted for evaluation of local wall thinning of piping in nuclear power plants.

248


Acknowledgment This study was conducted as part of Nuclear and Industrial Safety Agency (NISA) project on Enhancement of Ageing Management and Maintenance of Nuclear Power Plants in Japan.

References [1] T. Yamamoto, T. Takagi and T. Uchimoto, Remote field ECT for evaluation of local wall thinning using pancake coils, The 14th International Workshop on Electromagnetic Nondestructive Evaluation (ENDE 2009), 130 132. [2] T. R. Schmidt, The remote field eddy current inspection technique, Materials Evaluation, 42 (2) (1984), 225 230. [3] T. R. Schmidt, History of the remote field eddy current inspection technique, Materials Evaluation, 47 (1) (1989), 14 22. [4] Y.S. Sun, S. Udpa, W. Lord and D. Cooley, A remote field eddy current NDT probe for the inspection of metallic plates, Materials Evaluation, 54 (4) (1996), 510 512. [5] Y.S. Sun, T. Ouang and S. Udpa, Remote field eddy current testing: one of the potential solutions for detecting deeply embedded discontinuities in thick and multiplayer metallic structures, Materials Evaluation, 59 (5) (2000), 632 637.


249

PLL BASED EDDY CURRENT MEASURING SYSTEM FOR INSPECTION OF OUTER FLAW IN TITANIUM ALLOY PLATE Tomasz CHADY1, Jacek KOWALCZYK, Leon NAWOS-WYSOCKI, Grzegorz PSUJ and Ireneusz SPYCHALSKI Department of Electrical and Computer Engineering, West Pomeranian University of Technology, al. Piastow 17, 70-310 Szczecin, Poland

Abstract. In this paper a new nondestructive eddy current system will be presented. The system was design for detection of shallow outer flaws in thick titanium alloy specimens (e.g. plates, discs). A Lock in amplifier was used in a measuring subsystem in order to increase the efficiency of inspections of minor defects. The verification of a performance of the system was done using 6.9 mm titanium plate with artificial EDM (Electrical Discharged Machinated) notches. Keywords. Eddy current testing, measuring system, titanium alloys inspections

Introduction Titanium because of its mechanical properties (high corrosion resistance, high strengthto-weight ratio) is an important material in many leading industries e.g. aerospace. It is commonly used to manufacture new components such as turbine disc. The disc is machined from a titanium billet which is a cylindrical solid bar of up to 400mm of a diameter. In-service failures of such components can often lead to catastrophic consequences with significant economic impact. NDT techniques that are commonly utilized to examine traditional aerospace materials are not always sufficient to detect sub-surface defects created during the manufacturing process in titanium alloys [1]. In the case of components having large diameter a reasonable technique to apply is Ultrasound UT. However its effectiveness is limited to shallow regions close to the component’s surface. This results in a strong need for new advanced NDT technologies for the inspection of titanium components during the manufacturing process. We introduce a novel automated quality control system for the inspection of titanium billets in response to this need in this paper. The system combines two subsystems: eddy current testing (ECT) for evaluation of the material not deeper than 5mm beneath the surface and phased array (PA) ultrasonic for deeper inspection. In this paper the ECT

1

Corresponding Author: West Pomeranian University of Technology in Szczecin, Department of Electrical and Computer Engineering, ul. Sikorskiego 37, 70 313 Szczecin, Poland; E mail: [email protected].

250

T. Chady et al. / PLL Based Eddy Current Measuring System for Inspection of Outer Flaw

a)

b)

Figure 1. Photo of the measuring system: a) front panel, b) top view.

measurement system is introduced. We verify the performance of the system by measuring the depth of a defect in a flat titanium plate using a dedicated transducer.

1. ECT measuring system NDT systems in industrial applications have to be robust in that interference coming from various sources should not affect measurement results. The presence of noise in EC signals may mislead the flaw identification algorithms. This is important especially in the case of low testing frequency used for deeper material evaluation (greater penetration depth) which however cases week EC response signal (low signal-to-noise ratio). The most widely used noise suppression technique is digital filtering. It gives the best results if the signal can be separated from noises in a frequency domain. In many practical cases this condition is not fulfilled, and therefore filters may cause distortion of the original signals. It also affects the frequency characteristic. Therefore, a single frequency system with lock-in amplifier is proposed (Figure 1). Such system allows to detect signals caused by minor defects and having the amplitude lower than the noise level by extracting from whole spectrum only that signal’s component that has the same frequency as the excitation signal (reference signal). In consequence lockin technique allows to increase the sensitivity to minor flaws located deeper in the material. The proposed system allows to monitor amplitude and phase responses of a transducer using single testing frequency ranging from 1 kHz up to 100 kHz. The main part of the system is a lock-in amplifier based pick-up signal block. 1.1. Portable device of the system The block diagram of the measuring system is shown in Figure 2. A high speed 16-bit multifunction data acquisition converter (DAQ) with D/A and A/D modules (NI USB6251) are used in the system for both generating and acquiring signals. The DAQ device in connected via USB interface to the PC class computer which accumulates, computes data and controls the system. A generated in the D/A DAQ module sinusoidal signal is gained utilizing a high speed power amplifier (TDA 7294) before driving transducer's excitation coils. TDA 7294 is very low distortion and noise power amplifier with DMOS power stage. The driving current was controlled through LEM Components electronic transducer with no galvanic connection to the main circuit and was cut off if its value exceeds the maximum calculated for a measuring transducer.


LIA BVD 150 H

NI 6251 REAL

LOCK IN

IMAG

SIGNAL IN

C/A TDA 7294

SWICH

TRANSDUCER

LOCK IN REF IN

DIGITAL OUT

PC COMPUTER

USB

A/C

251

POWER AMPLIFIER Figure 2. Block diagram of measuring system.

Induced voltage in the pickup coils are amplified by a programmable gain amplifier in lock-in LIA-BVD-150-H module [2]. It is a dual phase lock-in amplifier with digital phase shifter and detection of real component, imaginary component and amplitude of input signal. Such arrangement enables significant noise suppression. Lock-in amplifiers are commonly used to detect and measure very small AC signals - all the way down to a few nanovolts. Accurate measurements can be made even when the small signal is obscured by noise sources many thousands of times larger. Lock-in amplifiers use a technique known as phase-sensitive detection in order to separate real and imaginary components of a signal at a specific reference frequency and phase. Noise signals, at frequencies other than the reference frequency, are rejected and do not affect the measurement. Signals from lock-in module are then supplied to A/D DAQ module.

Figure 3. View of the LabView® environment software application.

252


The proposed system works with specialized software operating in the National Instruments' LabView environment. The software interface is shown in Figure 3. The application is used to drive the DAQ converter in order to control excitation instruments, configure PLL operating parameters (such as sensitivity and time constant) and acquire measuring signals (a real and imaginary part of measuring signal). All parameters of lock-in amplifier, DAQ and excitation signal can be introduced to the system through a configuration file. The application enables external and internal triggering of acquisition events. During the measurements, real time results of acquired signal, its absolute value, real and imaginary part, phase shift and finally plot of real versus imaginary part are presented to an operator. In the system there were three processing algorithms implemented utilizing statistical analysis, gradient and lowpass filtering. They allows to eliminate trends in the signal caused by the lift off and surface roughness. The algorithms compute correction values for a signal basing on value measured at present step and past ones and can be applied in-situ or after the measurements. 1.2. The transducer The transducer used in the system is presented in Figure 4. In order to make possible defect detection located 5 mm beneath the sample surface its dimensions (Figure 4b) and excitation frequency were optimized using FEM (Finite Element Method) and Comsol software. The transducer is built using an E-type ferrite core with excitation coils wound on exterior columns of the ferrite core and a pick-up coil on a central one. The excitation coils generate contrary directed fluxes in the pick-up coil. Therefore, the flux flowing through pick-up coil is close to zero in equilibrium state. If a flaw appearing close to one of the excitation coils distorts the generated flux flow, the signal different from zero is induced in the pick-up coil. Such configuration of the transducer allows us for the optimal usage of the dynamic range of the A/D converter module. In order to achieve the greatest equilibrium the excitation coils of the transducers were driven by separate power amplifier.

a)

b)

1 3

c) 4

44

1

5

2

25

3

2

Figure 4. Photo and view with dimensions of the transducer: a) photo of the transducer, b) 3D view of the transducer, c) photo of the transducers head; all dimensions are in [mm]; 1 ferrite core, 2 excitation coils, 3 pick up coil, 4 transducer head, 5 sample


253

Figure 5. Photo and view of the titanium plate: a) photo, b) drawing of the notch, c) cross section drawing of the notch; x axis is the scanning axis

2. The verifying experiments The system is designated for the detection of minor outer flaws OF (casing discontinuity in the surface at the opposite site of the plate to the scanning side) in thick titanium specimens. In order to verify the performance of the proposed system several experiments were carried out. 2.1. Object of the experiment A titanium alloy plate having artificial EDM notches of different depth d was used (Figure 5.). The depth was ranging from 10 to 100% of the plate thickness (6.9 mm). All measurements were carried out for outer flaws. 2.2. Results of the experiment Results presented in the paper were obtained for 40% notch. First, the A-scan measurements using different exciting frequencies were carried out in order to find the optimal one. Figure 6 shows the results of real and imaginary component as well as amplitude and phase of the signal acquired during the experiments versus excitation frequency. From the set of achieved spectrograms one can notice that the frequency range between 8 and 16 kHz gives the greatest chance to detect the flaw. Considering result obtained for the real component of the signal (Figure 6a) it is clearly visible that frequency equal to 11.2 kHz is the optimal one. In the second stage of the experiments C-scan measurements were done. The two dimensional distribution of the signals amplitude measured in both scanning direction over selected area of 40% OF was presented in Figure 7a. In order to better visualize changes of signal caused by the flaw, plot of amplitude and phase shift as well as real and imaginary component of a signal measured in selected y line were shown in Figure 7b and Figure 8. The presented results allow easily to detect the flaw.

254


b)

f [kHz]

f [kHz]

a)

x [mm]

x [mm]

d)

f [kHz]

f [kHz]

c)

x [mm]

x [mm]

Figure 6. Multi frequency results of measurements obtained for 40% OF: a) real part of measured signal Re(Usig), b) imaginary part of measured signal Im(Usig), c) amplitude of measured signal |Usig|, d) phase of measured signal arctg(Im(Usig)/Re(Usig)).

b)

y [mm]

Usig [V]

a)

x [mm]

x [mm]

Figure 7. Single frequency results obtained for 40% OF: a) amplitude of signal |Usig| measured in selected area in both scanning direction, b) amplitude of signal |Usig| measured in selected y line.

3. Conclusions The newly developed ECT system for observation of minor defects in titanium alloy was presented. The system uses lock-in technique which allows to detect signals having the amplitude lower than the noise level by taking into consideration only that component of a signal which frequency is equal to the reference one. In consequence it


b)

d)

x [mm]

arctg(Im(U)/ Re(U)) [rad]

x [mm]

Im(U) [V]

c)

Im(U) [V]

Re(U) [V]

a)

255

Re(U) [V]

x [mm]

Figure 8. Single frequency results of 1D scan obtained for 40% OF: a) real part of measured signal Re(Usig), b) imaginary part of measured signal Im(Usig), c) amplitude of measured signal |Usig|, d) phase of measured signal arctg(Im(Usig)/Re(Usig)).

leads to increase the possible depth penetration. The results of experiments with the titanium alloy plate having artificial notches confirm the possibility of using this system in industrial application. In the future, further experiments will be carried out for titanium discs with artificial flaws.

Acknowledgement This work was supported in part by Polish Ministry of Science and Higher Education and by European Commission sponsored project QualiTi which is a collaboration between the following organizations: I.S.O.TEST Engineering s.r.l, West Pomeranian University of Technology (ZUT), Tecnitest Ingenieros S.L., TIMET UK Ltd, TWI Ltd and Vermon SA. The project is co-ordinated and managed by TWI Ltd and is partly funded by the EC under the Research for the Benefit of Specific Groups Project (ref: FP7-SME-2007-1-GA-222476.)

References [1] National Transportation Safety Board Aircraft Accident Report NTSB AAR 98/01 [2] http://www.femto.de/datasheet/FEMTO_Product_Overview.pdf

256


Giant Magneto-Resistive Sensor based Magnetic Flux Leakage Technique for Inspection of Track Ropes W. Sharatchandra SINGH, B.P.C. RAO1, S. MAHADEVAN, T. JAYAKUMAR and Baldev RAJ Nondestructive Evaluation Division, Indira Gandhi Centre for Atomic Research, Kalpakkam, TN-603102, India

Abstract. Giant magneto resistive (GMR) sensor based magnetic flux leakage (MFL) technique is proposed for detection of defects in 64 mm diameter track ropes. Helmholtz coil is used for magnetization and tangential component of leakage flux from defects is measured using a GMR sensor. This technique is able to detect local flaws and loss of metallic cross sectional area type defects in the track rope. Keywords. GMR sensor, magnetic flux leakage, track rope, finite element model

Introduction Steel wire ropes are used for material handling in mines and hauling of men in ski-lift operations [1]. In wire ropes, corrosion and wear are the main causes for damage such as formation of local flaws (LF) and loss of metallic cross-sectional area (LMA) [2-4]. LFs are external and internal discontinuities such as broken wires, cracks and corrosion pitting. Reasons for wire breaking can be due to fatigue, inter-strand nicking or martensitic embrittlement. LMAs are distributed defects such as missing of wires caused by corrosion, abrasion and wear resulting in loss of cross-sectional area. Periodic inspection of wire ropes is important to assess the structural integrity and to take corrective actions. Nondestructive inspection of wire ropes is challenging due to the heterogeneous structure of wire ropes, multiplicity, uncertainty of broken wires and hostile working environment. Visual and magnetic flux leakage (MFL) techniques are widely used for nondestructive inspection of wire ropes [4]. A variety of procedures that use different types of sensors and magnetizing devices have been employed for detection of defects in wire ropes by MFL technique [6-10]. Jomdecha et al. [7] used printed circuit-shaped coils connected in series as field sensors and solenoid as magnetization unit for inspection of wire ropes and reported detection of 2 mm deep surface defects in a 38 mm diameter wire rope. Kalwa et al. [9] developed an MFL system comprising of magnetic concentrators, Hall sensors and sensing coils and suggested that measurement of tangential component is more versatile than normal component for detection of multiple defects in wire ropes. 1

Corresponding Author: Dr. B.P.C. Rao, Head, EMSI Section, NDE Division, Indira Gandhi Centre for Atomic Research, Kalpakkam, TN 603102, India, E mail: [email protected].

W.S. Singh et al. / GMR Sensor Based MFL Technique for Inspection of Track Ropes

257

In this paper, we propose MFL technique for detection of defects on the outer surface of 64 mm diameter track rope used for transportation of coal. The track ropes are stationary and are rigidly supported at periodic intervals. They are operated for about 10 hours every day transporting nearly 3000 tons of coal with the help of 256 numbers of stationary buckets each carrying nearly 1.6 tons of coal. The schematic of cross-section of the track rope is shown in Figure 1. The track rope has 8 layers of stranded wires of different diameters as detailed in Table 1. The width of the outer surface of the Z wire is 6.45 mm and the gap width between two outer Z wires is 0.76 mm. The carriage wheels of the bucket are in contact with the track rope and thus causing damage to the outer surface of the Z-wire. Service induced surface flaws are the major causes of failure in these ropes. When more than two Z wires of the outer layer are broken, they will be separated from the adjacent layers. Table 1 Track rope cross sectional details. Layer Wire details 1 2 3 4 5 6 7 8

1 centre wire 6 round wires 6 filler wires 12 round wires 18 round wires 13 round wires 32 Z wires 34 Z wires

Wire diameter or thickness, mm 4.00 4.66 1.93 4.33 4.27 4.95 5.50 6.48 Figure 1. Cross section of track rope.

Recently, GMR sensors are being widely used in MFL testing for detection of defects in carbon steel plates [10], pipelines [11] in view of their high sensitivity for low magnetic fields, good signal-to-noise ratio (SNR) and high spatial resolution. GMR sensors are attractive for measurement of feeble magnetic fields from shallow surface defects and deeply located sub-surface defects. They are also useful for detection of stress and fatigue damage [12]. This paper discusses MFL technique proposed using Helmholtz coil and GMR sensors for detection of defects in the track ropes. The results of the developmental studies carried out on artificial LFs and LMA defect in track rope are discussed. In order to interpret the measured MFL signal from LMA defect in the track rope, 3-D finite element modeling is performed and results are discussed.

1. GMR based MFL Technique 1.1. Measurement Set-up The schematic of the proposed MFL technique is shown in Figure 2. It consists of magnetization device, variable DC power supply, GMR sensor, track rope with defects, amplifier, and personal computer for data acquisition. The magnetization device consists of a Helmholtz coil with two circular coils of 70 turns each separated by 50 mm. The width and height of each coil are 20 mm and 15 mm respectively. The use of Helmholtz coil ensures uniform axial magnetization of the track rope. GMR sensor is

258


used for detection of tangential component (along the scan direction) of leakage flux from defects and is kept at the middle of the Helmholtz coil. Measurements are made by moving the GMR sensor and Helmholtz coil together as a single unit over the track rope in steps of 1 mm and at each axial location GMR sensor is scanned circumferentially. A constant lift off of 0.3 mm is maintained between the GMR sensor and the track rope to avoid physical damage to the sensor. In order to enhance the sensitivity of GMR sensor, its output is first amplified using a low-noise amplifier consisting of a differential amplifier, notch rejection filter at 50 Hz, 100 kHz low-pass filter and a single-ended variable gain amplifier with DC suppression. The variable gain amplifier is set such that it amplifies the sensor output 10 times. The amplified GMR sensor output is digitized using a 2-channel data acquisition system (16 bit) and stored in the computer for analysis. &RPSXWHU '&SRZHUVXSSO\

%LDVVXSSO\$PSOLILHU

PP 6FDQGLUHFWLRQ

*05 6HQVRU

:LUH5RSH

PP +HOPKROW]FRLO

Figure 2. GMR sensor based MFL testing set up.

1.2 GMR Sensor GMR sensor works on giant magneto-resistive effect in which there is a large drop in electrical resistance of the multilayer for an incident magnetic field due to spin dependent scattering of electrons at the interface between two ferromagnetic layers (few nm thick) separated by a nonmagnetic layer. The GMR bridge sensors (AA00302) manufactured by M/s NVE Associates are used. These sensors measure differential output voltage and ensure high stability with low noise (sensitivity ~260 VT -1 at 5V bias voltage). They exhibit linear response in 0.2-1.3 mT range followed by saturation. 1.3 Artificial Defects Localized flaws are simulated by introducing electro-discharge machined (EDM) notches on the outer surface of the track rope. Four axial (A, B, C, D) notches and four circumferential (E, F, G, H) notches of 5.5 mm length, 2 mm width and of different depths are machined as shown in Figure 3. The depths of the EDM notches measured using replica technique are given in Table 2. The distance between the notches is maintained at 80 mm which is slightly more than the length of Helmholtz magnetization unit. One LMA type defect is simulated by removing material (42.0 mm length, 9.2 mm width and 3.0 mm depth) from the outer surface of the Z wire and this is around 46% of the outer Z-wire (6.48 mm diameter).

259


PP

PP

PP

$

PP

%

PP

PP

&

PP

'

(

PP

PP

)

PP

*

PP

PP

+

PP PP

Figure 3. Schematic of track rope having axial and circumferential machined artificial notches.

1.4 Magnetizing Current The magnetizing current in the Helmholtz coil is optimized such that detectable leakage field comes from all the notches. Current is optimized by studying the MFL signal peak amplitudes for shallow axial notches (A and C) and shallow circumferential notches (E and G). The signal amplitude is found to increase up to 6 A, beyond that it is decreased due to the penetration of significant amount of magnetic flux into the inner layer of the rope, as shown in Figure 4. Hence, a magnetizing current of 6 A is considered optimum and is used in this study.

GMR Signal Amplitude, V

2.5

2.05 mm deep axial notch (A) 5.86 mm deep axial notch (C) 1.94 mm deep circum. notch (E) 5.90 mm deep circum. notch (G)

2.0

1.5

1.0

0.5

0.0 0

2

4 6 8 Magnetizing Current, A

10

Figure 4. Optimization of magnetizing current.

2. Results and Discussion 2.1. LF defects GMR signals (tangential component of the leakage flux) of axial notches A, B, C and D are shown in Figure 5. The GMR sensor output for the shallowest axial notch (2.05 mm deep) is approximately 2.0 times the background noise that comes mainly from 0.76 mm wide stranded structure of the track rope. As can be observed all the four axial notches are detected by the technique with two distinct peaks correlated to the edges of axial notches. To access the detection and sizing capability of the technique, peak amplitude and Full Width at Half Maximum (FWHM) are determined after Gaussian fitting and interpolation and the results are shown in Table 2. The error in the

260


determination of FWHM is ±0.1. As can be noted the amplitude of the MFL signals is found to increase with increase in notch depth. The FWHM is found to be nearly the same for all the axial notches, because of the constant notch length (5.5 mm). MFL signals of circumferential notches viz. E, F, G, and H are shown in Figure 6. The GMR response of circumferential notches is seen sharp with a single peak and the amplitude of 1.94 mm deep notch is 3.2 times the background noise. As can be seen from Figure 6, all the four notches are detected by this technique. The signal amplitude increases up to a depth of 5.90 mm and then decreases while the FWHM remains nearly constant (refer Table 2). It is observed from Table 2 that the amplitude of the circumferential notches is 2 to 3 times higher than that of axial notches of similar depth. However, the FWHM of circumferential notches is found to be nearly 2 to 3 times lower than that of the axial notches. These observations are due to the axial magnetization of the track rope causing circumferential notches to leak out higher flux in comparison to the extended axial notches. Using a simple threshold FWHM, it is possible to identify axial and circumferential defects in the track rope. Alternately, a set of array sensors covering the circumference can be used. 1 .2

C

A x ia l N o tc h e s

D

1 .0 GMR Output, V

0 .8

B

0 .6

A

0 .4 0 .2 0 .0 -0 .2 -0 .4 0

50

100

150

200

250

300

S c a n D is ta n c e , m m

Figure 5. MFL signals of axial notches. 2.5 Circumferential Notches F

GMR Output, V

2.0

H

G

E

1.5 1.0 0.5 0.0 -0.5 0

50

100

150

200

250

Scan Distance, mm

Figure 6. MFL signals of circumferential notches.

300

261


Table 2. The MFL signal amplitude and FWHM of axial and circumferential notches. Notch Notch Depth, mm Amplitude, V FWHM, mm A (axial) 2.05 0.58 4.8 B (axial) 4.11 0.91 5.5 C (axial) 5.86 1.22 4.8 D (axial) 7.91 1.25 4.6 E (circumferential) 1.94 1.86 1.8 F (circumferential) 3.88 2.33 1.6 G (circumferential) 5.90 2.56 2.0 H (circumferential) 8.24 2.50 1.7

The performance of the MFL technique is assessed for detection of a fine circumferential saw cut that simulates a fatigue crack in the track rope. The saw cut (length 15 mm, width 0.5 mm and depth 2 mm) and corresponding GMR sensor output are shown in Figure 7. Clearly GMR sensor has detected the saw cut. The signal amplitude nearly matches with the MFL signals of circumferential notches of Figure 6. The signal amplitude is found to be slightly less than the signal amplitude of circumferential notch E of similar depth due to smaller width of the former. Thus, this technique can be used for early detection of localized cracks in track ropes. In order to assess the resolution to multiple cracks, 5 saw cuts (C1, C2, C3, C4 and C5) are simulated with each of length 23 mm, width 1 mm and depth 2 mm, separated by a distance of 13.5 mm, 10.0 mm, 5.5 and 3.2 mm and the GMR sensor output is shown in Figure 8. As can be seen, the technique is able to detect all the 5 saw cuts and resolve the saw cuts separated by 3.2 mm distance. 2.2. LMA defects Photograph of LMA defect and its GMR sensor response are shown in Figure 9. The depth distribution (profile) of the LMA defect is measured using optical method. As can be observed, the GMR sensor has detected the LMA with good SNR.

1.2 1.0

GMR Output, V

0.8 0.6 0.4 0.2 0.0 -0.2 2

4

6

8

10

Scan Distance, mm

Figure 7. Photograph of a 2 mm deep saw cut in track rope and the measured GMR sensor output.

12

262

W.S. Singh et al. / GMR Sensor Based MFL Technique for Inspection of Track Ropes 2.0 C2

C1

C3

C4 C5

1.5

GMR Output, V

1.0 0.5 0.0 -0.5 -1.0 0

10

20 30 40 Scan Distance, mm

50

Figure 8. Photograph of series of 2mm deep saw cuts machined in track rope and the corresponding measured GMR sensor output.

The LMA signal is found, as expected, extended as compared to the LF signal. There is a decrease in sensor output with respect to the baseline background, observed for the LMA defect. In order to understand this aspect, 3-D finite element modeling has been performed using COMSOL 3.4 Multiphysics software package. The model predicted tangential component of leakage field profile is also shown in Figure 9. In the model, the relative magnetic permeability and magnetization of the track rope are assumed as 100 and 106 A/m respectively. There is a good agreement between the measured and the model predicted signals and the decrease in signal amplitude for LMA defect appears to be due to the pulling of magnetic field lines into the track rope. Distance along LMA, mm 0

5

10

15

20

25

30

35

40

45

2.6 2.4 GMR Output, V

50 1

GMR response (experimental) GMR response (model) LMA depth distribution

2.2

0

-1

2.0 -2

1.8

LMA Depth, mm

2.8

-5

1.6 -3 1.4 1.2 0

10

20

30 40 50 Scan Distance, mm

60

70

-4 80

Figure 9. Photograph of LMA (42.0 mm length, 9.2 mm width and 3.0 mm depth) machined in the track rope, GMR sensor response scanned along dotted line direction and optically measured depth profile.


263

3. Conclusion GMR sensor based MFL technique has been proposed for detection of LF and LMA type of defects on the outer surface of 64 mm diameter track ropes. Performance of this technique has been validated by detecting a few saw-cuts and EDM notches simulating cracks in the track rope. For LMA type defects, a decrease in signal amplitude has been observed and confirmed by 3-D finite element model. This proposed MFL technique can be used for detection of both LF and LMA types of defects in track ropes. In order to further enhance the detection sensitivity and to implement the technique in field, work on the use of split type coils and array GMR sensors is underway.

4. Acknowledgements Authors thank Mr. S. Vaidyanathan and Mr. P. Krishnaiah, Electromagnetics, Modeling, Sensors and Imaging Section, NDE Division, Metallurgy and Materials Group, IGCAR, Kalpakkam for their help during the experiments.

References [1] S.S. Udpa and M.O. Patrick, ‘Nondestructive Testing Handbook - Electromagnetic Testing’, ASNT, 3rd Edition, Vol. 5, 2004, p. 437. [2] D. Basak, S. Pal and D. Chandra Patranabis, ‘In situ non destructive assessment of a haulage rope in a monocable zigback passenger ropeway’, Insight, Vol. 50, No. 3, March 2008, pp. 136 137. [3] Herbert R. Weischedel and R. P. Ramsey, ‘Electromagnetic testing, a reliable method for the inspection of wire ropes in service’, NDT International, June 1989, pp. 155 161. [4] Herbert R. Weischedel, ‘The Inspection of Wire Ropes in Service: A Critical Review’, Materials Evaluation, Vol. 43, December 1985, pp. 1592 1605. [5] N. Sumyong, A. Prateepasen and P. Kaewtrakulpong, ‘Influence of Scanning Velocity and Gap Distance on Magnetic Flux Leakage Measurement’, ECTI Transactions on Electrical Engineering, Electronics and Communications, Vol. 5, No. 1, February 2007. [6] E. Kalwa and K. Piekarski, ‘Design of Hall effect sensors for magnetic testing of steel ropes’, NDT International, Vol. 22, No. 5, October 1987, pp. 295 301. [7] C. Jomdecha and A. Prateepasen, ‘Design of modified electromagnetic main flux for steel wire rope inspection’, NDT&E International, Vol. 42, 2009, pp. 77 83. [8] Wang Hong yao, Hua Gang and Tian Jie, ‘Research on Detection Device for Broken Wires of Coal Mine Hoist Cable’, Journal of China University Mining & Technology, Vol. 17, No. 3, pp. 376 381. [9] E. Kalwa and K. Piekarski, ‘Design of inductive sensors for magnetic testing of steel ropes’, NDT International, Vol. 20, No. 6, December 1987, pp. 347 353. [10] W. Sharatchandra Singh, B. P. C. Rao, S. Vaidyanathan, T. Jayakumar and Baldev Raj, ‘Detection of Leakage Magnetic Flux from near side and far side Defects in Carbon Steel Plates using Giant magnetoresistive Sensor’, Measurement Science and Technology, Vol. 19, 2008 015702 (8pp). [11] L. Chen, P.W. Que and T. Jin, ‘A Giant Magnetoresistance Sensor for Magnetic Flux Leakage Nondestructive Testing of a Pipeline’, Russian Journal of NDT, Vol. 41, No.7, 2005, pp. 462 465. [12] T. Chady and G. Psuj, ‘Data fusion from multidirectional remanent flux leakage transducers for NDT of stress and fatigue loaded steel samples’, IEEE Trans. on Magn., Vol. 44, Nov. 2008, pp. 3285 3288.


Application of Electromagnetic Nondestructive Techniques



267

Pipe Wall Thickness Measurement by Electro-Magnetic Acoustic Transducer using Band Exciting Method Daigo KOSAKA a, Fumio KOJIMA b,1 and Kosuke UMETANI c Graduate School of Engineering, Kobe University, Kobe, Japan b Organization of Advanced Science and Technology, Kobe University, Kobe, Japan c Graduate School of System Informatics, Kobe University, Kobe, Japan a

Abstract. This paper is concerned with a quantitative nondestructive evaluation of pipe wall thinning using an Electro-Magnetic Acoustic Transducer (EMAT). First, two methods for sizing wall thickness, pulse-echo method (PEC) and electromagnetic resonance measurement technique (EMAR) are examined using calibration specimen of SS400. It was shown that EMAR is superior to PEC for its sizing accuracy. Secondly, we discuss a feasible measurement strategy based on EMAR technique. Finally, the feasibility studies of the method are summarized for evaluating mock Flow-Accelerated Corrosion (FAC) test samples. Keywords. Nondestructive test, Condition monitoring, Signal processing, Flowaccelerated corrosion

Introduction Phenomena of pipe wall thinning are critical issues on ageing management of nuclear power plants [1-7]. Pipe wall thinning management in nuclear power plants is aimed at providing a life management process ensuring replacement or repair prior to in-service failure. The main objective of pipe wall inspection is to identify the location of maximum thinning, to ascertain the extent and depth of the thinning, and to evaluate the wear rate. Currently the conventional technique is ultrasonic testing (UT) which provides the accurate resolution for the pipe wall thickness measurement. However, in UT, we must remove insulation from the piping area at each in-service inspection. There also exist so many inaccessible locations for the in-situ inspection by surveyors in practical plants. Therefore, continuous surveillance during operation is much more important than the periodical inspection by UT. Recently, we propose a monitoring technique of pipe wall thinning using Electro-Magnetic Acoustic Transducer (EMAT) [8]. EMAT is a non-contacting inspection device that generates an ultrasonic pulse in the sample inspected [9-13]. The advantage of our method is that measurements can be remotely implemented and the exciting test signals can be automatically reproduced by electrical circuits. EMAT consists of a magnet and a coil of wire and relies on electro-magnetic acoustic interaction for elastic wave generation. 1

Corresponding Author: Fumio KOJIMA, Organization of Advanced Science and Technology, Kobe University, 1-1 Rokkodai-cho, Nada-ku,Kobe 657-8501; Japan, E-mail: kojima@koala kobe-u.ac.jp

268

D. Kosaka et al. / Pipe Wall Thickness Measurement by EMAT Using Band Exciting Method

Figure 1 illustrates our experimental apparatus. It is composed of EMAT, a burst power supply, an oscilloscope and a computer. A using sensor is shown in Figure 2. The sensor is a vertical type of EMAT which produces SH wave. The magnets generate a static magnetic field. The coil generates a dynamic magnetic field that induces eddy currents in a surface of a sample specimen. An ultrasonic wave is generated in the surface by the static magnetic field and the eddy currents. This type of sensor fits thickness measurement of pipe. There are two thickness measurement techniques for EMAT. Pulse-echo method (PEC) is a fundamental technique based on time domain approach. PEC is ease of use and rapid processing. However, unlike the conventional UT technique, the detecting signal by EMAT is not so large. Therefore, the use of PEC is not adequate for pipe wall thickness measurement. Electromagnetic resonance method (EMAR) is another thickness measurement technique based on frequency domain [14-15]. The method can estimate wall thickness from seeking a sequence of peak spectrums of detecting signals corresponding to sweeping frequencies of exciting test signals. Comparing that PEC method use one single frequency for exciting, measurement process by EMAR requires tedious treatment of signal processing. Nevertheless EMAR has advantages on the accurate resolution for sizing thickness of pipe. This paper is organized as follows: First, two measurement methods were compared in the point of thickness measurement method. Second, we propose a measurement method that uses an exciting voltage with a wide band. A receiving voltage of a wide band exciting voltage has information of with wide band frequency. This means this method is able to obtain the resonant frequency at one time. Finally, the method was applied to evaluate mock Flow-Accelerated Corrosion sample specimens.

Figure 1. Experimental setup.

Figure 2. Top view of EMAT.

1. Thickness measurement by EMAT Our experimental apparatus is composed of an EMAT, a pulser receiver, an oscilloscope, an auto stage and a computer. Burst waveform of the excitation voltage of 800Vpp is added to EMAT by pulsar. Two methods for sizing wall thickness are examined using calibration plate sample specimen of SS400.


269

1.1. Pulse-echo method The pulsar signal was taken by two pulses with 2MHz. The reflective waveforms were recorded by a computer coupled to an oscilloscope. Using PEC, thickness calculation can be represented by

T=

vΔt 2

(1)

where T , v and ǻt denote thickness, sound velocity and time of flight (TOF). Sound velocity of SS400 is 3.216×103m/s. Detecting waveforms are shown in Figure 3. Figure 3(a) is a waveform of a plate sample specimen with 8mm thickness. Noting that the time of flight (TOF) ǻt was 4.9ȝs and from Eq. (1), the estimated thickness was taken as 7.9mm of which error became 0.1mm. Figure 3(b) is a waveform of a plate sample specimen with 6mm thickness. Since TOF was ǻt = 3.6ȝs, the estimated thickness was 5.8mm where the error was 0.2mm. Taking into account that TOF ǻt becomes proportional to the thickness, it is more difficult to recognize TOF ǻt for a smaller thickness.

Figure 3. Detecting waveforms using pulse-echo method.

270


1.2. Resonance method EMAR utilizes collections of resonance frequencies by sweeping exciting voltages with wide frequency range. In our experiment, exciting time is 100ȝs, and exciting frequency range is 0.6MHz to 2.6MHz. The interval of the sweeping frequency was taken as 1kHz. For measurements by EMAR, the estimate of thickness can be derived from

T=

v 2Δf

(2)

where ǻf denotes a peak interval with respect to the frequency. The detecting waveforms of 4.7mm and 4.9mm thickness are shown in Figure 4. The ǻf is very clear that, if T is smaller, then the ǻt of pulse-echo method becomes smaller interval while the ǻf becomes larger interval. The ǻf of 4.7mm thickness and the ǻf of 4.9mm thickness are clearly distinguishable. This implies that EMAR method is more accurate measurement technique than PEC method. The ǻf of the detecting waveform of 4.7mm and 4.9mm thickness are 0.343MHz and 0.330MHz. Detecting thickness of the sample specimens is 4.69mm and 4.87mm. Figure 5 depicts the total resolution of the method presented here. In Figure 5, we admit the detecting accuracy was up to 0.1mm for SS400. On the other hand, measurement becomes time-consuming because EMAR requires sweeping exciting frequencies with wide range. In our experiment, each measurement spent about 2 minutes for obtaining sizing estimate.

Figure 4. Resonance frequency using EMAR method


271

Figure 5. Thickness resolution of resonance method.

2. Measurement method using band exciting in EMAR In this section, we discuss a feasible measurement strategy for implementing EMAR method. As already suggested in the previous section, EMAR is time consuming in order to seek resonant state with wide frequency ranges. Figure 6 and 7 demonstrate one typical example for measurement strategy using an exciting waveform with a large frequency band. Figure 6, a frequency band of the exciting waveform was chosen from 0.9MHz to 1.8MHz. Then it was found that the detecting waveform in Figure 7 has some resonance frequencies at one shot. Hence we can reduce the measurement time by using a set of exciting signals with wide band.

Figure 6. Exciting waveform with wide band.

272


Figure 7. Spectrums waveforms of band exciting method.

3. Application to pipe wall thickness measurement In this section, the band exciting resonance method is applied to evaluate a mock FAC sample specimen. As shown in Figure 8, a 6B-SS400 pipe was tested and the corrosion shape was fabricated to simulate flow accelerated corrosion (FAC). The extent of FAC was taken as 200 mm with the maximum thinning depth b = 1.0mm. Figure 9 shows estimating sizing results of the sample specimen by our measurement method. For the purpose of the comparative discussions, the conventional UT techniques were also performed.

Figure 8. Sample specimen of FAC model (a=160mm, b=1.0mm)


273

Figure 9. Estimated sizing results for mock Flow-Accelerated Corrosion sample specimen.

4. Conclusions Two measurement methods for pipe wall thickness measurements were examined

using calibration specimen of SS400. It was shown that EMAR is superior to PEC for its sizing accuracy. It was shown that the measurement strategy using the exciting voltage with a wide band has achieved the rapid processing for thickness measurements. The feasibility studies of the method are summarized for evaluating mock Flow-Accelerated Corrosion (FAC) test samples. Detecting results of the method agreed well with detecting results of ultrasonic testing.

Acknowledgments This work has been performed as a part of the National Research Project for Enhancement of Measures against the Ageing of Nuclear Power Plants sponsored by the Nuclear and Industrial Safety Agency (NISA). The authors also gratefully acknowledge the Institute of Nuclear Safety System Inc. (INSS).

References [1] C. Schefski, J. Pietralik, T. Dyke, M. Lewis, A physical model to predict wear sites engendered by flowassisted corrosion, US DOE Rep, pp. 149-154, 1995. [2] B. Chexal et al., Flow-accelerated corrosion in power plants, EPRI report, TR-106611, 1996. [3] R. B. Dooley, V. K. Chexal, Flow-accelerated corrosion of pressure vessels in fossil plants, International Journal of Pressure Vessels and Piping, Volume 77, Issues 2-3, pp. 85-90, 2000. [4] Flow-accelerated corrosion in nuclear power plants: Application of CHECWORKS at Darlington, [5] T.R. Allen, P J. King, and L. Nelson: Flow Accelerated Corrosion and Cracking of Carbon Steel Piping in Primary Water Operating Experience at the Point Lepreau Generating Station, Proceedings of the 12th International Conference, pp. 773-784, 2005. [6] K. D. Efird, Flow accelerated corrosion testing basics, Pap Corros, p. 16, 2006. [7] S. Uchida, M. Naitoh, Y. Uehara, H. Okada, N. Hiranuma, W. Sugino, S. Kishizuka and D. H. Lster: Evaluation of Wall Thinning Rate with the Coupled Model of Static Electrochemical Analysis and Dynamic Double Oxide Layer Analysis, Journal of Nuclear Science and Technology Vol. 46, No. 1, pp. 31-40, 2009. [8] D. Kosaka, F. Kojima, H. Yamaguchi and K. Umetani, Monitoring System for Pipe Wall Thinning Management using Electromagnetic Acoustic Transducer, Vol.2, No. 1, pp.34-42, 2010.

274


[9] H.M. Fros, Electromagnetic-ultrasonic transducer: Principles, practice and applications, in: Physical Acoustics, Vol. 14, W.P. Mason and R.N. Thurston eds., Academic Press, New York, pp. 179-275, 1979. [10] B.W. Maxfield and C M. Fortunko, The design and use of electromagnetic acoustic wave transducers, in: Material Evaluation, Vol.41, pp. 1399-1408, 1983. [11] R.B. Thompson, Physical principles of measurements with EMAT transducers, in: Physical Acoustics, Vol.19, Academic Press, New York, pp. 157-200, 1990. [12] K. Mirkhani et al. Optimal design of EMAT transmitters, in: NDT&E International, Vol. 37, pp. 181– 193, 2004. [13] D. MacLauchlan, S. Clark, B. Cox, T. Doyle, and B. Grimmett, Recent advancements in the application of EMATs to NDE, in: Proc. of the 16th WCNDT 2004, Montreal, Canada. [14] K. Kawashima, Very high frequency EMAT for resonant measurement, in: Proc IEEE Ultrasonic Symposium, No. 2, pp.1111-1119, 1994. [15] H. Ogi, “Field dependence of coupling efficiency between electromagnetic field and ultrasonic bulk waves”, J. Appl. Phys. 82 (8), 15, pp. 3940-3949, 1997.


275

Applicability of Magnetic Flux Leakage Method for Wall Thinning Monitoring in Nuclear Power Plants Hiroaki KIKUCHI1, Isamu SHIMIZU, Katsuyuki ARA, Yasuhiro KAMADA and Satoru KOBAYASHI NDE & SR-C, Faculty of Engineering, Iwate University, 4-3-5 Ueda, Morioka, Iwate, 020-8551, Japan Abstract. This paper presents an applicability of magnetic flux leakage (MFL) method for a nondestructive evaluation (NDE) of wall thinning in nuclear power plants. Since MFL method has already been applied to other industry field, this method should be an effective tool in nuclear power plants. The ancillary yoke poles were employed to adapt to the piping shape, and higher sensitivity was achieved to estimate the depth in a slit on piping. The distribution of magnetic flux density corresponding with profile of wall thinning depth was also obtained. Additionally, an applicability of MFL method for probing a slit fabricated on superposed specimens was investigated. Keywords. Wall thinning, magnetic flux leakage (MFL), ancillary yoke pole, reinforcing plate, dual layer

1. Introduction An ageing management for nuclear power plants is quite important issue due to its long-term operation in Japan. One of ageing problem is wall thinning on pipe. Wall thinning occurs at orifices, elbows and under reinforcing plates, etc., through FlowAccelerate Corrosion and Liquid Droplet Impingment erosion. Although several methods like ultrasonic, eddy current are available tools, the magnetic flux leakage (MFL) [1-4] also must be an effective nondestructive evaluation (NDE) technique for the wall thinning, because this method has already been applied to industry field such as gas pipeline. It is difficult to assess wall thinning under a reinforcing plate, where steels are superposed for reinforcements, using ultrasonic and eddy current, because of the reflection of signal and the wall thickness. Therefore MFL method is expected to apply for an evaluation of wall thinning under the reinforcing plate. The fundamental experiment of MFL using a single-yoke was examined and it was clarified that sizing of a rectangular slit fabricated on steel plates is possible [5]. However, a development of a probe feasible for piping and a feasibility study of MFL for application in a reinforcing plate are required for practical uses. Therefore, in this study, an applicability of MFL method for wall thinning monitoring in nuclear power plants was investigated, and a possibility to evaluate wall thinning under reinforcing 1

Corresponding Author: Hiroaki Kikuchi, NDE & Science Research Center, Faculty of Engineering, Iwate University, 4 3 5 Ueda, Morioka 020 8551, Japan; Phone:+81 19 621 6890; Email: hkiku@iwate u.ac.jp

276

H. Kikuchi et al. / Applicability of Magnetic Flux Leakage Method for Wall Thinning Monitoring

plates using MFL was examined. The ancillary yoke poles were utilized to solve the problem in contact between the magnetic-yoke and the pipe surface and to achieve higher sensitivity for estimation of slit depth on piping. Additionally, steels superposed were prepared and a slit modeled wall thinning was formed in underlayer, then an evaluation of the slit was performed using MFL method.

2. Utilization of ancillary yoke pole 2.1 FEM analysis Firstly, to confirm an effect of ancillary yoke pole, two-dimensional (2D) FEM analysis was done. Figure 1(a) shows the analysis model for the calculation. Since a conventional magnetic-yoke has a flat plane on contact part, there is a large air gap between the yoke and the surface of the pipe. The magnetic flux density distribution was calculated along with the dot line shown in Figure 1(a) when the pipe has a slit with 10 mm width and 2 mm depth and when the pipe has also no slit. The calculation was done for both x- and z-components of magnetic flux density, Bx, By. An applied current to the excitation coil was 1 A. Figure 2 shows the calculated results. The results show the same behavior in both cases being independent of existence of slit. This means detection of a slit on pipe is difficult if there is a large air gap between the yoke and the pipe surface. Figure 1(b) shows the FEM analysis model when ferromagnetic material exists between the yoke and the surface of the pipe as ancillary yoke poles. The distribution of both x- and z-components, Bx, By, of magnetic flux density was calculated along with the dot line shown in Figure 1(b). Figure 3 shows those calculated field distributions. One can see the notable difference in both distribution profiles. These calculated results indicate the ancillary yoke pole is very useful to achieve higher sensitivity. 2.2 Experimental procedure Figure 4(a) shows the dimension of the magnetic-yoke used for experimental investigating the utilization of ancillary yoke poles. The geometry and the size of the specimen are shown in Figure 5. The diameter is 60.5 mm and thickness is 5.2 mm, and carbon steel STPT is used for their material. A slit with 10 mm width is fabricated at Conventional yoke Coil: 1000 T

Air gap

Conventional yoke Coil: 1000 T

Ferromagnetic material

Silicon Iron

Silicon Iron 50

Coil

50

Coil

60 Unit: mm 9

13

Bz

S15C

60 Unit: mm

Bx

13

x

0

Pipe Slit

9

S15C

Bz

Bx

x

0

Pipe Slit

(a) with air gap (b) using ancillary yokes Figure 1. 2D FEM analysis model.

H. Kikuchi et al. / Applicability of Magnetic Flux Leakage Method for Wall Thinning Monitoring 800

Leakage flux density Bz (Gauss)

Leakage flux density Bx (Gauss)

0 -200 -400 -600 -800 -1000 -1200 without slit with slit

-1400 -1600 -16

277

-12

-8

-4

0

4

8

12

400 200 0 -200 -400 -600 -800 -16

16

without slit with slit

600

-12

-8

Position x (mm)

-4

0

4

8

12

16

Position x (mm)

(a) x component, Bx (b) z component, Bz Figure 2. Leakage magnetic flux density distribution with air gap between yoke and pipe. 100 without slit with slit

Leakage flux density Bz (Gauss)

Leakage flux density Bx (Gauss)

0 -20 -40 -60 -80 -100 -120 -140 -160 -16

-12

-8

-4

0

4

Position x (mm)

8

12

16

without slit with slit 50

0

-50

-100 -16

-12

-8

-4

0

4

8

12

16

Position x (mm)

(a) x component, Bx (b) z component, Bz Figure 3. Leakage magnetic flux density distribution with ancillary yokes between yoke and pipe.

the center part of the pipes, and its depths are 0.5, 1, 4, 4.5 mm, respectively. The coordinate for measurement system is defined as shown in Figure 4(a). The center part of slit is the origin of x-direction (horizontal), and the vertical direction to horizontal is z-direction (the surface of specimen is z = 0). Figure 4 (b) shows the geometry and the size of ancillary yoke pole adjusted for the pipe shown in Figure 5. The material for this pole is pure iron. To obtain soft magnetic properties, the pole was annealed at 700 ƷC for 1 hour in the atmosphere after its machining. After annealing, the poles were polished to remove oxide layer. 2.3 Results Figure 6 plots x- and z- component of magnetic flux density, Bx, Bz against position x (Here, center of the slit is x = 0). On the measurement, the applied current to the excitation coil was 3 A, and the magnetic-yoke was fixed as the center of the yoke corresponding with the center of the slit, and a gauss meter (Lakeshore 460) scanned the pipe surface in x-direction at z = 0.5 mm. The pipes with slit depth of 4 and 4.5 mm were used here. The results obtained with ancillary yoke poles are shown as compared with measurement without ancillary yoke poles. It is clarified that the higher sensitivity was obtained when the ancillary yoke poles were used. As for the pipe with slit depth of 4 mm, the intensity of Bx has no change and Bz has no peak at the edge of slit in the case without ancillary yoke poles, while the profile using ancillary yoke poles shows large increment in Bx profile and has a peak in Bz at the edge of slit (x = 5 mm). For the

278


Figure 4. Measuring system for MFL and dimensions of magnetic yoke and ancillary yoke pole. Wall thinning

d

w

10 mm

60.5mm

t = 5.2mm

# 2345mm

Wall thinning

60mm

Figure 5. Dimension of pipe with slit. 200 Magnetic flux density Bz (Gauss)

Magnetic flux density Bx (Gauss)

-20 -40 -60 -80 -100 -120

4mm NAY 4 5mm NAY 4mm AY 4 5mm AY

-140 -160

100 0 -100 4mm NAY 4 5mm NAY 4mm AY 4 5mm AY

-200 -300

-2

0

2

4 6 8 10 Position x (mm)

12

14

-2

0

2

4 6 8 10 Position x (mm)

12

14

(a) x component, Bx (b) z component, Bz Figure 6. Experimental leakage magnetic flux density distribution with and without ancillary yoke poles between yoke and pipe (AY: with ancillary yoke, NAY: without ancillary yoke).

pipe with slit depth of 4.5 mm, the changes in Bx and the peak intensity in Bz at x = 5 mm were enhanced using ancillary yoke poles. Figure 7 plots x- and z- component of magnetic flux density, Bx, Bz against position x, when the probe includes the magnetic-yoke, ancillary yoke poles and the gauss meter scanned the pipe surface in x-direction (In this case, gauss meter and magnetic-yoke scanned together). The depth of slit changes from 0.5 to 4.5 mm. The xcomponent Bx of flux has rapid increase at x = 5 mm, and its rate of change increases with increasing slit depth. On the other hand, the z-component Bz of flux has a peak at x = 5 mm, and its intensity increases with increasing depth of slit. The position x = 5 mm is consistent with the edge of slit.

3. Evaluation of wall thinning on pipe Pipe specimens with modeled wall thinning being similar to the practical cases were

H. Kikuchi et al. / Applicability of Magnetic Flux Leakage Method for Wall Thinning Monitoring 300 Magnetic flux density Bz (Gauss)


-20

279

-40 -60 -80 -100 -120 d = 0.5 mm d = 1 mm d = 4 mm d = 4.5 mm

-140 -160 -180

d = 0.5 mm d = 1 mm d = 4 mm d = 4.5 mm

250 200 150 100 50 0 -50

-5

0

5 10 15 Position x (mm)

20

25

-5

0

5 10 15 Position x (mm)

20

25

(a) x component, Bx (b) z component, Bz Figure 7. Experimental leakage magnetic flux density distribution with ancillary yoke poles, when slit depth d changes. 60

20

14

57

Figure 8. Dimension of pipe with wall thinning. a is maximum depth, b width of slit and ' central angle of wall thinning part.

63.4㫦

Unit: mm

Figure 9. Dimension of ancillary yoke pole.

prepared, and MFL method using ancillary yoke poles was examined to assess wall thinning of those specimens, here. The configuration and the size of pipe are shown in Figure 8. Its diameter is 114.3 mm and its thickness is 8.6 mm, respectively. The material of the pipe is carbon steel STPG. We prepared two specimens, A and B. The value of a, b and ' shown in Figure 8 are 4.4 mm, 100 mm, 130 deg., respectively for specimen A, and those values for specimen B are 2.15 mm, 50 mm, 90 deg., respectively. The magnetic-yoke used here was the same as mentioned in Figure 1 (a), while another ancillary yoke poles adjusted for the pipe shown in Figure 8 were prepared. Figure 9 shows the geometry and the size of ancillary yoke pole. The gauss meter was fixed at the center part of the magnetic-yoke, and the magnetic-yoke with the ancillary yoke poles scanned the pipe surface in x-direction. The x- component of magnetic flux density, Bx was measured with application of 3 A current to the excitation coil. The scans were performed along with the line passing through the deepest wall thinning point in both A and B specimens and were also performed along with the line passing through out of wall thinning area; this means scanning excludes wall thinning part. Figure 10 shows the distribution of x- component of magnetic flux density Bx against position x. For both specimen A and B, magnetic flux density is constant on the line passing through out of wall thinning area, while the distributions on the line passing through deepest wall thinning point reflect the depth profiles of wall thinning in each specimen (the cases of A: a = 4.4 mm and B: b = 2.15 mm). The magnetic flux density at deepest wall thinning point in specimen A is larger than that of specimen B. At deeper wall thinning point, larger magnetic flux density is obtained.

280



40 Specimen A: Wall thinning area 35

30 Specimen B: Wall thinning area

25

A A B B

20 -100

-50

a 4 4 mm line out of slit a 2 15 mm line out of slit

0 50 Position x (mm)

100

Figure 10. Experimental leakage magnetic flux density distribution for pipe with modeled wall thinning.

Excitation coil 150 turns

Unit: mm 80

20

Gauss meter 6

45

z Bz

Scan Bx x

0

Specimen

10

d

6

Figure 11. Measurement setup for investigation of dual layered specimen.

4. Evaluation of slit on dual layered specimen Here, we investigated an applicability of MFL method to an evaluation of wall thinning under the reinforcing plate. As a feasibility study, we tried to detect a slit fabricated at the underlayer of superposed plates using MFL method. Figure 11 shows the measurement system for evaluation on dual layered specimen (superposed plates). The magnetic-yoke the same as described in section 2 was located on the superposed specimen with a slit (10 mm width) on its underlayer. Each plate had 6 mm thick, and total thickness was 12 mm. The yoke and the gauss meter scanned together in x-direction (See Figure 11), and measured the x- and zcomponent of magnetic flux density, Bx, Bz, on the surface of specimen when the slit depth d changed from 1 to 3 mm. Figure 12 plots the value of Bx, Bz against position x. For all specimens, Bx has the maximum at x = 0 and a rapid change rate in Bx was obtained at x = 5 mm. The intensity of Bx at x = 0 and change rate in Bx at x = 5 mm increase with increasing slit depth. On the other hand, Bz has a peak at 5 mm where it is consistent with the edge of slit and the peak intensity increases with increasing slit depth. These results indicate MFL method is applicable to detection of slit formed in the superposed specimen, that is, applicable to evaluation of wall thinning under reinforcing plate.

H. Kikuchi et al. / Applicability of Magnetic Flux Leakage Method for Wall Thinning Monitoring 40 no slit d 1 mm d 2 mm d 3 mm

55 50

Magnetic flux density Bz (Gauss)


60

45 40 35 30 25 20 -20 -15 -10

281

-5 0 5 10 Position x (mm)

15

20

30 20

no slit d 1 mm d 2 mm d 3 mm

10 0 -10 -20 -30 -40 -20 -15 -10

-5 0 5 10 Position x (mm)

15

20

(a) x component, Bx (b) z component, Bz Figure 12. Relations between magnetic flux density and position x, when slit depth d changes.

5. Conclusion In order to accommodate the shell structure of pipe, ancillary yoke poles are adopted, and probing of slit and wall thinning on pipe was performed. Additionally, detection of slit fabricated on dual layered specimen was examined as a feasibility study of evaluation of wall thinning under reinforced plate. The following results were obtained. (1) It was clarified the evaluation for wall thinning on pipe using MFL method is possible. (2) The higher sensitivity was achieved with ancillary yoke poles. (3) The capability of evaluation of wall thinning on dual layer was performed; this indicates MFL method can be applied for evaluation of wall thinning under reinforcing plate. (4) Output distributions corresponding with wall thinning depth on specimen being similar with actual cases were obtained.

Acknowledgment This work is supported by Nuclear and Industrial Safety Agency Project on Enhance of Ageing Management and Maintenance of Nuclear Power Plants.

References [1] C. E. Edwards, S. B. Palmer, The magnetic leakage field of surface breaking cracks, Journal of Physics D: Applied Physics, 19, pp. 657 673 (1986). [2] K. Sekine, Present Status and Some Problems in Defect Evaluation Using Magnetic Inspection, Tetsu- toHagane, 74, pp. 2231 2238 (1988). [3] T. Suzuki, Electromagnetic Simulation Technique for Magnetic Leakage Flux Testing, Journal of the Japanese Society for Non-destructive Inspection, 47, pp. 92 97 (1998). [4] Y. Zhang, G. Yan, Detection of Gas Pipe Wall Thickness Based on Electromagnetic Flux Leakage, Russian Journal of Nondestructive Testing, 43, pp. 123 132 (2007). [5] H. Kikuchi, Y. Kurisawa, Y. Kamada and S. Kobayashi, K. Ara, International Journal of Applied Electromagnetics and Mechanics, 33, pp. 1087 1094 (2010).

282


Stress Corrosion Cracks Evaluation in 316 Austenitic Stainless Steel Plate Tomasz CHADY1, Jacek KOWALCZYK, Grzegorz PSUJ Department of Electrical and Computer Engineering, West Pomeranian University of Technology, al. Piastow 17, 70-310 Szczecin, Poland

Abstract. In this paper a ECT system was used to inspect stress corrosion cracks (SCC). The system utilizes lock in amplifier, which enables observation of signals having the amplitude lower than the noise level. The sample with three SCCs examined during test was made of 16 mm thick 316 austenitic stainless steel. Keywords. Eddy current testing, stress corrosion cracks

Introduction Nowadays, evaluation of stress corrosion cracks (SCC) in real construction is one of the most important problems. The SCC strongly affects the structural integrity of the material. Furthermore, there are many differences between SCC and artificial notches in their responses to nondestructive testing (NDT) methods. Therefore the notches cannot be used as a substitute of SCC during evaluation of the NDT system performance. This creates strong and obvious need of fabricating cracks more equivalent to real one. On this background, the Japan Society of Maintenology has started round-robin tests under SCC NDT database Project [1]. Some works on the methodology of artificial SCC preparation and observation have been published [2-4]. The round–robin test creates opportunity to evaluate quality of the measurements achieved with various techniques. Objective of this paper is to present results obtained by using ECT multi-frequency technique and system equipped with a lock-in amplifier. The advantages of the lock-in amplifiers like high noise immunity, precise estimation of real and imaginary part of the measured signals allows to detect defects with a higher precision.

1. Experimental setup In the experiments lock-in ECT system was used and the three stress corrosion cracks were examined. The details of the system and the sample tested will be given below. The utilized system is an integrated unit. It allows to monitor amplitude and phase responses of a transducer. The system can operate using single frequency ranging from 1 kHz up to 100 kHz. It consists of excitation, pick-up and control subsystem.

1

Tomasz Chady, email: [email protected]

T. Chady et al. / Stress Corrosion Cracks Evaluation in 316 Austenitic Stainless Steel Plate

283

Figure 1. Photo and view of the transducer: a) 3D view of the transducer, c) photo of the transducers.

The excitation subsystem uses D/A converter and high speed power amplifier (TDA 7294) for generation and gaining sinusoidal current. The main part of the system is a lock-in amplifier based pick-up system. Voltage induced in the pickup coils are amplified by a programmable gain amplifier in the lock-in LIA-BVD-150-H module. The LIA-BVD-150-H is dual phase lock-in amplifier with digital phase shifter and detection of real component, imaginary component and amplitude of input signal. Signals from the lock-in module are then supplied to an A/D module. Both subsystems are controlled by a PC class computer with software written in LabView® environment. The transducer was moved over the sample using a X, Y, Z scanning device. The transducer used during the measurements consists of a ferrite core (6) with five symmetrically placed columns [5]. The view and photo of the transducer is shown in Figure1. Pick-up coil (5) is wounded on the central column. Four excitation coils (1, 2, 3 and 4) placed on remaining columns are divided into two sections perpendicular to each other. Each section consists of two excitation coils connected in series and produce in the pickup coil opposite (to another pair) directed fluxes, which induce in the pick-up coil a signal close to zero in equilibrium state. The flaw occurring close to one of the excitation coils distorts that state and therefore a signal different from zero appears in the pick-up coil. 1.1. SCC samples The sample TP01 used in the experiment belong to the round robin test organized by the Japan Society of Maintenology. It is made of 316 austenitic stainless steel. The dimensions of TP01 are 160 mm long, 100 mm width and 16 mm thick. The sample is having three stress corrosion cracks: SCC01-1, SCC01-2, and SCC01-3. Detailed description of SCC fabrication can be found in [1-4]. The distance between two neighboring cracks is about 30 mm. Surface lengths of the cracks evaluated using penetrant testing (PT) technique are 26, 28 and 29 mm. Detail information about the specimen can be found elsewhere [1].

2. Results of Experiments The experiment was divided into two stages. During the first one, stress corrosion cracks were scanned along one line. Different excitation frequencies were used for each scan. The frequency range was between 70 kHz and 88 kHz with step of 2kHz. Figure 2 and Figure 3 show the results of multi-frequency experiments. The transmitter was moved perpendicular to the direction of the defects and above their centers.

284


x [mm]

f)

x [mm]

f [kHz]

x [mm]

f [kHz]

e)

d)

f [kHz]

x [mm]

f [kHz]

c)

f [kHz]

b)

f [kHz]

a)

x [mm]

x [mm]

h)

U [V]

arctg(Im(U)/Re(U)) [rad]

g)

x [mm]

x [mm]

Figure 2. Multi frequency results of measurements obtained for SCC01 1(a) amplitude and b) phase of the measured signal), SCC01 2 (c) amplitude and d) phase of the measured signal) and SCC01 3 (e) amplitude and f) phase of the measured signal); plot of amplitude (g) and phase (h) of signals obtained for all SCCs using exciting frequency equal to 82 kHz .


x [mm]

f)

x [mm]

f [kHz]

x [mm]

f [kHz]

e)

d)

f [kHz]

x [mm]

f [kHz]

c)

f [kHz]

b)

f [kHz]

a)

285

x [mm]

x [mm]

h)

Im(U) [V]

Re(U) [V]

g)

x [mm]

x [mm]

Figure 3. Multi frequency results of measurements obtained for SCC01 1(a) real and b) imaginary component of the measured signal), SCC01 2 (c) real and d) imaginary component of the measured signal) and SCC01 3 (e) real and f) imaginary of the measured signal); plot of real (g) and imaginary component (h) of signals obtained for all SCCs using exciting frequency equal to 82 kHz .

286


y [mm]

One can see that all results allow proper indication of the SCC in the sample. There are only slight differences in measured amplitude and phase of signals as well as in real and imaginary components of signals obtained for each frequency.

x [mm] Figure 4. Amplitude of signal |Usig| measured in selected area of SCC01 1 in both scanning direction using exciting frequency equal to 82 kHz.

b)

U [kHz]

arctg(Im(U)/Re(U)) [rad]

a)

x [mm]

x [mm]

d)

Im(U) [V]

Re(U) [V]

c)

x [mm]

x [mm]

Figure 5. Results of measurements obtained for selected line scan of SCC01 1 using exciting frequency of 82 kHz: a) amplitude of measured signal |Usig|, b) phase of measured signal arctg(Im(Usig)/Re(Usig)), c) real part of measured signal Re(Usig), d) imaginary part of measured signal Im(Usig).


287

Nevertheless, the highest amplitude was observed in case of the exciting frequency of 82 kHz. It creates the greatest possibility of proper indication of SCC. From the Figure 2 g-h and Figure3 g-h one can observe that response of SCC01-1 is the highest and of SCC01-3 is the smallest, which is confirmed by the results achieved by other research teams [1]. In the second stage of experiments the measurements for SCC01-1 in both directions x and y using exciting frequency equal to 82 kHz were carried out. The achieved results are presented in Figure 4 and Figure 5. Figure 4 shows a twodimensional plot of amplitude of the measured signal and Figure 5 plots of amplitude and phase of a signal as well as real and imaginary component of a signal for selected line scan.

3. Conclusions The results of the experiments show the possibility of usage of the lock-in ECT system for evaluation of SCC in stainless steels. The highest signals parameters and signals components were achieved for SCC01-1. The results are not yet verified by destructive tests, however similar conclusions were presented by other institutions examining the sample.

Acknowledgements The authors would like to thank Dr Noritaka Yusa for providing the specimens that belong to the round robin test organized by the Japan Society of Maintenology (http://jsm.or.jp/at/scc/). This work was supported in part by Polish Ministry of Science and Higher Education and by European Commission sponsored project QualiTi which is a collaboration between the following organizations: I.S.O.TEST Engineering s.r.l, West Pomeranian University of Technology (ZUT), Tecnitest Ingenieros S.L., TIMET UK Ltd, TWI Ltd and Vermon SA. The project is co-ordinated and managed by TWI Ltd and is partly funded by the EC under the Research for the Benefit of Specific Groups Project (ref: FP7-SME-2007-1-GA-222476.)

References [1] SCC NDT database Project, The Japan Society of Maintenology, online access, http://jsm.or.jp/jsm/at/scc/index_eng.htm [2] N. Yusa, S. Perrin, K. Miya, Eddy current data characterizing less volumetric stress corrosion cracking in nonmagnetic materials, Matterials Letters 61 (2007), 827 829 [3] N. Yusaa, H. Hashizume, Evaluation of stress corrosion cracking as a function of its resistance to eddy currents, Nuclear Engineering and Design 239 (2009) 2713 2718 [4] N. Yusa, K. Miya, Discussion on the equivalent conductivity and resistance of stress corrosion cracks in eddy current simulations, NDT&E International 42 (2009) 9 15 [5] T. Chady, Evaluation of Stress Loaded Steel Samples Using Selected Electromagnetic Methods, Review of Progress in Quantitative Nondestructive Evaluation, vol. 23A, D.O. Thompson, D.E.Chimenti, American Institute of Physics, 2004, pp. 1296 1302

288


Multi-frequency eddy current NDE of the distance between parts in aeronautical assemblies Thanh Long CUNG, Pierre Yves JOUBERT and Eric VOURC’H SATIE, ENS Cachan, CNRS, UniverSud, 61 avenue du Président Wilson, 94235 Cachan Cedex, France

Abstract. This paper deals with the eddy current nondestructive evaluation of the distance between parts in metallic assemblies. An experimental study of the coupling of a magnetic cup core coil sensor with a metallic layered structure is carried out and confirmed by finite element modeling simulations. This study allows behavioral models of the sensor/structure interactions to be built, and two estimation methods to be proposed. The first one is based on the inversion of a polynomial forward model, and the second one on an inverse model. Both techniques are implemented, characterized and compared. The second one is proved to be more accurate when the signal to noise ratio of the eddy current data decreases.

Keywords. Non-destructive evaluation, eddy currents, multi-frequency approach, air gap estimation, multilayered aeronautical structure, behavioral inverse model

Introduction The eddy current technique (EC) is widely used for the NDE of electrically conducting parts, such as those that may be found in the aeronautic industry. Indeed, it is easy to implement, sensitive, robust and eco-aware. Under certain conditions, it allows defects [1-3] or geometrical/physical features [4] of the inspected parts to be evaluated with satisfactory accuracy. However, the quantitative evaluation of parts parameters starting from the raw EC data is generally known as an "ill-posed" problem, among other reasons because of the incompleteness of the available EC data [5]. This remark particularly applies to the evaluation of multilayered structures such as aeronautical assemblies, which is known to be particularly difficult [6]. To solve this particular problem, several methods have been proposed, such as the inversion of a physical model of the impedance of an air coil, using the least mean square criterion. Another solution relies on the use of gradient techniques or neural networks [7]. Another possibility consists in an experimental approach in which the experimental data are directly differentiated from measurements simultaneously carried out on a series of calibrated mock-ups [6]. In this paper, we propose to build multi-frequency behavioral models, based on experimental analyses of the interactions between a cup-core EC sensor and a multilayered assembly. This technique has the advantage of being applicable to any type of EC sensor and to be easy to implement, once the model developed. Moreover, the multi-frequency approach allows the available EC data to be enriched so as to increase the reliability and the accuracy of the NDE [1]. In this study,

T.L. Cung et al. / Multi-Frequency Eddy Current NDE of the Distance

289

we consider the problem of evaluating the thickness of an air layer comprised between two conductive plates, representative of a metallic coating assembled to ribs or spars in aircraft wings. Section 1 both reports on the experimental setup and on the analysis of the interactions between the used sensor and a layered structure. In Section 2, a polynomial based modeling is derived from this analysis and inversion results are presented. In Section 3, we propose and implement a behavioral inverse model associated to a multi-frequency evaluation algorithm. Evaluations results are compared and concluding remarks are given in section 4.

1. Experimental study of the problem 1.1. Experimental setup In the experimental setup, the wing coating is represented by a plate of constant thickness tc = 1.5 mm, whereas the rib which in practice is a piece of variable thickness, is represented by a series of plates featuring thicknesses comprised between 1.5 mm and 25 mm. Both the coating and the rib are made of an aluminum alloy featuring an electrical conductivity σ = 17 MS/m and a unitary relative magnetic permeability. To adjust the distance between the conductive layers, calibrated isolating sheets are used. They feature a unitary relative magnetic permeability and a unitary relative dielectric permittivity. The resulting distance between the conductive layers, or air-gap thickness, denoted t, is comprised in the [0 μm 500 μm] range. The EC NDE is performed by means of a magnetic cup-core coil probe electromagnetically coupled to the assembly, as shown in Figure 1. The coil features 110 turns and an outer diameter of 35 mm. Its free impedance (obtained when the sensor is not coupled with the target) is equivalent to a resistance R0 = 4.6 Ω in series with an inductance L0 = 3.04 mH. EC data are provided through the measurement of the sensor impedance via a PC controlled impedance analyzer (HP4192A). More precisely, we consider the sensor normalized impedance denoted Zn and defined as [4]:

Zn = Rn + j.X n = (Zt − R0 ) / X 0

(1)

where Zt is the impedance of the sensor coupled with the assembly featuring a given air gap t, and X0 is the reactance of the uncoupled sensor.

Figure 1. Eddy current cup core coil sensor coupled with the inspected multilayered assembly.

290


Figure 2. Universal impedance diagrams of the EC sensor: a) general view, and b) detailed view for Znt with t ranging from 0 to 500 μm.

1.2. Sensor and target interactions The properties of the universal impedance diagram (UID), i.e. the evolution of Zn in the complex impedance plane (Rn, Xn), have been studied [4, 8] in the case of a bulk target. It is used here to evaluate the assembly. As an example, the UIDs obtained for two assemblies featured by t = 0 μm (close to a bulk target) and t = 500 μm, are presented in Figure 2a over an 80 Hz to 30 kHz frequency range. These UIDs show that there exists a frequency range in which the properties of the measured EC data are likely to allow t to be evaluated. The UIDs plotted in that frequency range are represented in Figure 2b, for parts separated from 0 to 500 μm. Indeed, the locus of the points obtained at the same frequency but for different distances t (tc and tr being fixed) is a linear curve. This result led us to analyze the relationship between t and a normalized impedance distance (NID) defined as (2). NID f i = Z nt ( f i ) − Z n 0 ( f i )

(2)

where fi is the considered EC frequency, Zn0 and Znt are the normalized impedance of the sensor for t = 0 and t 0 respectively. Experimentally, it appears that a linear equation relates NID to t (3), provided that the frequency f does not exceed a maximum value fmax, fmax being such that the skin depth δ of the induced EC approximately equals 2/3rds of tc.

NID ( t ) = a ( f , tc , tr ) t

(3)

The slopes of the linear characteristics (3) depend on the frequency, the coating and the rib thicknesses, as shown in Figure 3 for the following configurations: f1 = 680 Hz, f2 = 6600 Hz, tr1 = 1.5 mm, tr2 = 25 mm, and tc = 1.5 mm. As a result, equation (4), which is established at a single frequency f1, is not sufficient to estimate t for an unknown value of the thickness tr. Therefore, a multi-frequency approach is required.


291

Figure 3. Linear relationship between NID and the air-gap thickness t −1

ti = a f i .NID f i

(4)

2. Estimation method relying on a polynomial forward model

2.1. Inversion of a polynomial forward model (IPFM) In order to build a multi-frequency forward model of NID as a function of t, tc, and tr, we use a second order polynomial of the variables t and tr expressed in (5). Such a small order is chosen with inversion accuracy in view [9].

NID f i (t , t r ) = t ( c0, f i + c1, f i t r + c2, f i t r2 )

(5)

where the coefficients {c0,fi , ... c2,fi} are such that for a fixed value of tr, NIDfi be a linear function of t, according to (2), with NIDfi (0,tr) = 0. To determine the coefficients of (5), we use the data resulting from a series of FE simulations of the sensor coupled with a layered structure featuring tc = 1.5 mm, tr ∈ {1.5, 2.0, 2.5, 3.0, 3.5}, given in mm, t ∈ {100, 200, 300, 400, 500}, given in μm and with a set of five frequencies fi (i∈{1,2,3,4,5}) given in Hz: f1 = 680, f2 = 1060, f3 = 1440, f4 = 1820, f5 = 2200. These data constitute the learning set used to determine the polynomials of the forward model at each considered fi frequency, by the means of a least mean square criterion. The multi-frequency forward modeling of NID then writes:

n = Ct

(6)

where n is the column vector of the NIDfi elements, C is the matrix of the {c0,fi , ... c2,fi} coefficients, and t = [t, ttr , ttr2 ]T . Consequently the estimation of t and tr is given by (7).

292


t = (CT ⋅ C ) ⋅ CT ⋅ n −1

(7)

2.2. Estimation results The estimation based on the IPFM is implemented and evaluated using computed and experimental data. In order to better evaluate the estimation performances, additional white and Gaussian noise of various powers have been added to the simulated data so as to introduce some uncertainty representative of instrumental noise and sensor mispositioning errors [11]. Table 1. RAE and RPE of the air layer thickness estimation obtained using the IPFM and different frequency combinations. (*) Meaningless values of RAE or RPE (> 100 %). Assembly configurations tc = 1.5 mm ; tr = 1.5 mm ; t ∈ {0, 50, 100, 150, ...500) in μm IPFM using fi IPFM using fi IPFM using fi Data i∈ {2,3,4,5} i∈ {1,2,3,4,5} i∈ {2,3,4} (RPE% ; RAE%) (RPE% ; RAE%) (RPE% ; RAE%) Comput.(SNR = 60 dB) 0.77 ; (*) 13.20 ; 1.20 1.24 ; 0.42 Comput. (SNR = 33 dB) 17.25 ; (*) 75.00 ; (*) 28.02 ; -0.20 Experim.(SNR=33 dB) 40.44 ; (*) (*) ; (*) 49.16 ; -19.21 Assembly configurations tc = 1.5 mm ; tr = 3.5 mm ; t ∈ {0, 50, 100, 150, ...500) in μm IPFM using fi IPFM using fi IPFM using fi Data i∈ {2,3,4,5} i∈ {1,2,3,4,5} i∈ {2,3,4} (RPE%-RAE%) (RPE%-RAE%) (RPE%-RAE%) Comput.(SNR = 60 dB) 0.65 ; -81.81 16.35 ; 2.47 1.56 ; 0.08 Comput.(SNR = 33 dB) 14.20 ; -81.81 75.16 ; (*) 34.13 ; 2.43

Figure 4. Estimation of the air-gap thickness using a 5-frequency 2nd order polynomial forward model for a tc = 1.5 mm; tr = 1.5 mm configuration. a) simulated data (SNR = 33dB), b) experimental data (SNR = 33 dB)

The simulated test set is constituted by the configurations described in section 2.1., but with t taking 11 values linearly arranged within the 0 to 500 μm range. For each of these configurations, 1000 random noise trials are used. Regarding the experimental data, the test set is constituted by data relative to configurations featured by t ranging from 0 to 500 μm, and tc = 1.5 mm (which constitute the worst case for the estimation). For each considered configuration, each measurement was repeated 12 times (with sensor dismounting) so as to take uncertainty into account. Finally, in order to

293


characterize the estimation results, we consider the relative accuracy error (RAE) and the relative precision error (RPE), defined in (8), where tˆ is the estimate of t, and where mean(.) and std(.) denote the mean value and the standard deviation, respectively.

§ tˆ − t · RAE % = mean ¨100. ¸ t ¹ ©

and

§ std (tˆ) · ¸ RPE % = mean ¨¨ 100 . mean (tˆ ) ¸¹ ©

(8)

The estimation of the distance between parts using 3, 4 and 5 excitation frequencies was implemented according to (7). Estimation errors obtained using simulated data featured by a 60 dB and a 33 dB signal to noise ratio (SNR) and experimental data of 33 dB SNR are gathered in Table 1. The estimation errors significantly decrease when using a 5-frequency forward modeling (down to the order of 1 % for a 60 dB SNR). However, the estimation method exhibits a lower accuracy when the SNR falls down to 33 dB (Table 1, Figure 4), due to the poor conditioning of the forward matrix (6). Furthermore, the estimation errors are increased when obtained from the experimental data, due to the imperfect agreement between experimental and simulated data relative to the same configurations.

3. Estimation method relying on an inverse behavioral model

3.1. Inverse behavioral model (IBM) In order to enhance the estimation performances, we propose a second estimation technique, based on an IBM. Here, we aim at exploiting the linear characteristics (4). We build such sets of characteristics from simulated data obtained for discrete values of tr, considering the same assembly configuration and frequency sets as in section 2.1. Then, using 5th-order polynomials, we extrapolate the values of the slopes a −f 1 of i

equations (4) for any tr value comprised in the considered rib thickness range:

a −f i1 (tr ) = k5, f i t r5 + k 4, f i tr4 + k3, f i t r3 + k 2, f i t r2 + k1, f i tr + k0, f i

(9)

Provided the coefficients {k0,fi , ... k5,fi}, it is possible to estimate the air gap as follows. For every linear curve (4) corresponding to a given rib thickness trl, the measured NIDfi corresponds to a til air gap abscissa which is a possible solution of the problem:

til = a −f i1 (trl ). NID f i with tr1 ≤ trl ≤ trL and l ∈ {1,2,..., L}

(10)

Among the solutions provided by the equations (10), only those corresponding to the actual rib thickness, denoted tr l , will be true and also equal to each other. As a act

consequence, tr l may be estimated as the one such that the equations (10) satisfy: act tˆrl act = t rlˆ

act

so that

( ) = min{std (t ), std (t

std tilˆ

act

i1

i2

),..., std (tiL )}

(11)

294


Finally, the air gap is estimated at every fi according to (12):

1 tˆ = N

N

¦t i =1

ilâct

(12)

3.2. Estimation results and comparison The simulated and experimental data which the IPFM was applied to in section 2.2 are now used to implement the IBM. The estimation results obtained using simulated as well as experimental data featured by tc = tr = 1.5 mm and SNR = 33 dB (that is the worst estimation case considered) are given in Figure 5. For the simulated data, the RPE and the RAE are about 9.45 % and -10.38 % respectively, while for the experimental data they are 6.60 % and -10.03 % respectively. Results from simulated and experimental data are thus in good agreement and exhibit both satisfactory precision and accuracy. Furthermore, Table 2 compares the precision and the accuracy of the IPFM and of the IBM based estimation methods, considering a 5-frequency approach. The estimation relying on the IBM appears to be significantly more precise than that relying on the IPFM, whatever the considered case. In particular, with the IBM, in the worst cases considered the RPE does not exceed 10 % when it reaches 50 % with the IPFM. With regards to the accuracy, the two methods performances are in the same order.

Figure 5. Estimation of the air-gap thickness using a IBM: a) simulated data with the SNR = 33 dB, b) experimental data. Assembly configurations : tc = 1.5 mm ; tr = 1.5 mm ; t ∈ {0, 50, 100, 150, ...500) in μm Table 2. RAE and RPE of the air layer thickness estimations obtained using the IPFM and the IBM and a 5frequency approach. Assembly configurations tc = 1.5 mm ; tr = 1.5 mm ; t ∈ {0, 50, 100, 150, ...500) in μm Data IBM using fi i∈ {1,2,3,4,5} IPFM using fi i∈ {1,2,3,4,5} (RPE% ; RAE%) (RPE% ; RAE%) Comput.(SNR = 60 dB) 1.24 ; 0.42 0.93 ; 0.36 Comput. (SNR = 33 dB) 28.02 ; -0.20 9.45 ; -10.38 Experim.(SNR=33 dB) 49.16 ; -19.21 6.60 ; -10.03 Assembly configurations tc = 1.5 mm ; tr = 3.5 mm ; t ∈ {0, 50, 100, 150, ...500) in μm Data IBM using fi i∈ {1,2,3,4,5} IPFM using fi i∈ {1,2,3,4,5} (RPE% ; RAE%) (RPE% ; RAE%) Comput.(SNR = 60 dB) 1.56 ; 0.08 0.06 ; -0.03 Comput.(SNR = 33 dB) 34.13 ; 2.43 2.84 ; 1.10


295

4. Conclusions

An experimental study of the coupling between a cup core coil eddy current sensor and a layered conducting assembly (similar to some aeronautical wing assemblies) has been reported. The analysis of the universal impedance diagram of the sensor has shown that the normalized impedance distance could be used with a view to the eddy current NDE of the distance between the conductive parts. These results, which have been confirmed by finite elements simulation results, led us to propose behavioral models of the coupling between the eddy current sensor and the layered structure. Furthermore, two methods for the estimation of the thickness of the distance between parts have been proposed and implemented. On the one hand, the estimation relied on the inversion of a polynomial forward behavioral model (IPFM), and on the other hand, an inverse behavioral model (IBM) was used. As far as the actual problem of the NDE of the assembly of a coating on a wing rib is concerned, two unknowns are to be estimated: the air layer thickness and the rib thickness, which may vary from one position on the inspected structure to another. As a consequence a multi-frequency estimation approach is necessary. The overall performances of the IBM based estimation method have proved to be more precise and as accurate as those of the IPFM based method. Further works will focus on the optimisation of the used EC sensor so as to increase the SNR of the EC data, as well as considering the elaboration of more accurate learning data set, so as to reduce the accuracy error. Also, dedicated sensor configurations will be designed so as to be used on actual aeronautical assembly geometries.

References [1] Y. Le Diraison, P.-Y. Joubert, D. Placko, Characterization of subsurface defects in aeronautical riveted lap-joints using multi-frequency eddy current imaging, NDT&E international 42 (2009), 133–140. [2] M. Wrzuszczak, J. Wrzuszczak, Eddy current flaw detection with neural network applications, Measurement 38 (2005), 132-136. [3] L. S. Rosado, T. G. Santos, M. Piedade, P. M. Ramos, P. Vilaça, Advanced technique for non-destructive testing of friction stir welding of metals, Measurement, In Press, Corrected Proof, Available online 13 February 2010. [4] S.N. Vernon, The universal impedance diagram of the ferrite pot core eddy current transducer, IEEE trans magn 25(3):2639–45 (1999). [5] J. Pavo, S. Gyimothy, Adaptative inversion database for electromagnetic nondestructive evaluation, NDT & E International 40 (2007), 192–202. [6] P. Huanga, G. Zhanga, Z. Wub, J. Caia, Z. Zhou, Inspection of defects in conductive multi-layered structures by an eddy current scanning technique: simulation and experiments, NDT&E international 39 (2006), 578–584. [7] I.T. Renakos, T.P. Theodoulidis, S.M. Panas, T.D. Tsiboukis, Impedance inversion in eddy current testing of layered planar structures via neural networks, NDT&E international 30 (1997), 69-74. [8] Y. Le Bihan, Study on the Transformer Equivalent Circuit of Eddy Current Nondestructive Evaluation, NDT&E International 36 (2003), 297–302. [9] C.V. Dood C V, W.E. Deeds, Eddy-current multiple property measurements, International Advances in Nondestructive Testing 9 (1981), 317-333. [10] H. L. Libby, Introduction to electromagnetic nondestructive test methods, Wiley, New York, 1971. [11] P. Y. Joubert, Y. Le bihan, Eddy Current Data Fusion for the Enhancement of Defect Detection in Complex Metallic Structures, International Journal of Applied Electromagnetics and Mechanics 19 (2004), 647 – 651.

296


Artificial Heart Valve Testing Using Electromagnetic Method Tatiana STRAPACOVA1, Klara CAPOVA, Ladislav JANOUSEK, Milan SMETANA Department of Electromagnetic and Biomedical Engineering, Faculty of Electrical Engineering, University of Zilina, Univerzitna 1, 010 26 Zilina, Slovak Republic

Abstract. The paper focuses on non-destructive inspection of BjĘrk-Shiley Convexo Concave artificial mechanical heart valves. A new approach for the inspection based on sweep frequency eddy current testing is proposed. The standard self inductance pancake probe positioned in proximity of the valve drives the eddy currents and pick-ups the response signal. The frequency of exciting current is changed in a wide range. Numerical simulations are carried out to verify the usefulness of the proposed technique. The gained results presented in the paper show that there is a clear difference in the amplitude frequency spectrum of the response signal between the intact valve and the other with a single leg fracture of the outlet strut. Keywords. non-destructive inspection, eddy currents, sweep frequency, outlet strut fractures, BjĘrk- Shiley Convexo Concave prosthetic heart valve

Introduction A dysfunction of the heart valve is a common complication after a heart valve disease. When the symptoms became intolerable with normal human lifestyle, the malfunction heart valve is replaced by an artificial one. Artificial heart valves are engineered mechanical metallic devices or they are made from a biological tissue.

Figure 1. Typical design of the BSCC heart valve and the fracture of the outlet strut.

1

Corresponding Author: Tatiana Strapacova, Department of Electromagnetic and Biomedical Engineering, Faculty of Electrical Engineering, University of Zilina, Univerzitna 1, 010 26 Zilina, Slovak Republic; E-mail: [email protected]

T. Strapacova et al. / Artificial Heart Valve Testing Using Electromagnetic Method

297

One of the artificial mechanical heart valves is the BjĘrk-Shiley Convexo Concave (BSCC) valve, shown in Fig. 1, which was developed in order to improve the hemodynamic and to reduce the risk of thromboembolism. BSCC was used to replace the aortic valves or the mitral ones. The BSCC valves have a carbon occluder disc held in the place by two metallic struts. One of the two struts, the inlet one is integrated into the valve suture ring, while the outlet one is welded into the suture ring. The suture ring made of Teflon is sewn to the cardiac tissue to hold the valve in the place. Approximately 120 thousand patients have received the BSCC [1]; however, in number of cases (619 patients, [1]) a malfunction of the valve has occurred because of the outlet strut fractures. The main reason of the outlet strut fractures has not been determined yet. The process of cyclic slamming open and shut of the occluder disc during heart function subjects the valve to percussive impact stress, which can cause a fatigue failure. These failures have been observed at the outlet strut flange junction near or at the weld. The failure of both struts results in dislodgement of the occluder disc and embolization of the disc. A dual strut failure results in abrupt onset of dyspnea, loss of consciousness, or cardiovascular collapse due to embolization of the disc and acute severe valvular regurgitation. A patient with the dual strut fracture of aortic prosthesis dies in minutes, but those with strut fracture of mitral prosthesis may survive long enough to undergo the valve replacement. Several types of mechanical BSCC heart valve failures have been reported in literature [1]: 1. an outlet strut fracture (OSF), fracture of both legs of the outlet strut, the strut is completely separated from the flange, 2. a single leg fracture (SLF), fracture of one leg of the outlet strut, both ends grate against each other, 3. a single leg separation (SLS), fracture of one leg of the outlet strut, both ends of the fractured strut are separated as displayed in Fig. 1. Although the function of the valve in the cases 2, 3 can continue under these conditions, the stress concentration on the intact end of the welded strut is increased and it is not clear how long the other end of the strut can remain intact. The period of time between the first and the second fractures is highly variable [1]. Due to this fact there is a considerable interest, therefore, to develop methods for assessing the state of the valve in general condition of the outlet strut weld by minimally invasive nondestructive inspection method. Electromagnetic methods represent a good candidate [2], [3] for the inspection. Eddy current non-destructive testing is widely utilized in a variety of industrial applications. An eddy current probe scans over an inspected surface and the response signals are evaluated accordingly. However, the conditions of considered application do not allow such scanning. A new approach for the non-destructive inspection of BSCC using eddy current technique is therefore proposed in the paper. It is based on sweep frequency technique. Numerical simulations are carried out to evaluate the proposal and the gained results are presented and discussed in the study.

1. Numerical model Possibilities of the sweep frequency eddy current inspection of BSCC are investigated in the paper using numerical means. The commercially available software OPERA-3D based on finite element method is utilized for the purpose.

298


Standard self-inductance pancake probe, shown in Fig. 2, is used for the inspection. Dimensions of the probe are set up according to the dimensions of heart valve. Dimensions of the catheter which would be used for encapsulations of the coil are considered as well. The probe is positioned 1 mm above the weld of the outer strut where a crack is considered to appear. Current density of the exciting signal is kept constant during the inspection at a value of J = 1 A/mm2, while its frequency is changed in a wide range starting from 10 kHz up to 500 kHz.

Figure 2. Dimensions of the pancake probe.

A complex model of the BSCC shown in Fig. 3 is developed. The dimensions and the electromagnetic properties of model are adjusted according to the dimensions and the material properties of real BSCC heart valve. The materials commonly used for the BSCC heart valve replacement are Stainless steels, Titanium alloys and Co-Cr alloys. The developed model considers the electromagnetic parameters of a Co-Cr alloy Hayness 25; the conductivity is adjusted to ı = 1.14·106 S/m and the relative permeability to ȝr = 1.

Figure 3. Dimensions of the BSCC heart valve model.


299

A considered defect is localized at one end of the outlet strut, Fig.3. The defect is non-conductive, i.e. fatigue crack, and it represents the single leg fracture (SLF) of the BSCC. Dimensions, an orientation and a depth of the defect are adjusted according to real conditions that may appear in the outlet strut of BSCC heart valve. A width of the modelled defect is set to w = 0.1 mm and its depth d is varied from 0.1 up to 1.9 mm with a step of 0.2 mm in a direction of the leg diameter. The single leg separation (SLS) is not considered in this study as the attempt of inspection should be early alarm of fatigue crack development. The numerical calculations were run for the intact outlet strut (IOS) for comparison as well. The modelled system considers also real environmental conditions, i.e. the blood and the tissue of myocardium as it is shown in Fig. 4. Electromagnetic properties of the blood and the myocardium are adjusted according to their real values presented in Tab. 1. The myocardium tissue is positioned around suture ring and the blood is inside the BSCC heart valve. Table. 1. Properties of the myocardium tissue and the blood for different frequencies.

Figure 4. Model configuration of the cardiovascular system

Numerical simulations of the developed model are carried out to calculate the probe response signal. The finite element mesh of the simulated problem has approximately 3 million elements of two types, linear and quadratic ones. The following section presents the results.

300


2. Numerical results The results of numerical simulations of the model introduced in the previous section are presented here. Figure 5 shows amplitude frequency spectrum of the probe response signal in absolute values for four cases. The curve denoted as IOS represents the sweep frequency response signal for the intact outlet strut. The other three dependences are the response signals for the single leg fracture (SLF) with different depth of the defect, i.e. 0.1 mm, 0.5 mm and 1.9 mm. Similar results for the IOS and the SLF with the defect depth of 1.9 mm but in relative values are shown in Fig. 6. The inductive component of the probe response signal is dominant and thus the phase frequency spectrum does not show almost any difference between the IOS and the SLF.

Figure 5. Amplitude frequency spectrum of the probe response signal in absolute values

Figure 6. Amplitude frequency spectrum of the probe response signal in relative values

It is evident from the presented results that there is a clear difference in the amplitude frequency spectrum between the IOS and the SLF. It should be noted that the difference is more notable when the depth of defect is increasing. Difference in the dependences between the IOS and the SLF is largest in a frequency range from 70 kHz up to 150 kHz.


301

It can be concluded that the sweep frequency technique is very promising for the eddy current inspection of the BSCC outer strut. The technique helps to overcome one problem connected with the eddy current inspection concerning the given conditions. The scanning over an inspected surface is not needed.

Conclusion The paper proposed a new approach for a defect detection in BjĘrk-Shiley Convexo Concave artificial heart valve using sweep frequency eddy current technique. This method detects perturbations in eddy current flow due to a defect in the outlet strut in a wide frequency range. A model of the valve was developed and the proposed technique was verified using numerical simulations. Presented results clearly showed that there is a clear difference in the amplitude frequency spectrum between the intact outlet strut and the strut with a defect. Experimental measurements are going to be carried out to confirm the numerical results.

Acknowledgement This work was supported by the Slovak Research and Development Agency under the contract No. APVV-0194-07. This work was also supported by grant of the Slovak Grant Agency VEGA, project No. 1/0308/08. The authors express their thanks to Prof. R. Grimberg for lending the valve prototypes.

References [1] Paul Van Neer: The BjĘrk- Shiley valve: Detecting broken struts using standards diagnostic ultrasound instruments, MSc Thesis, 2005 [2] Chan Shiu C., Yue Li, Udpa Lalita, Udpa Satish S.: Electromagnetic techniques for detecting strut failures in artificial heart valve, Studies in Applied Electromagnetics and Mechanics Electromagnetic, Vol. 26, Nondestructive Evaluation, Ed. G. Dobmann, ISBN 1- 58603-594-0 [3] Grimberg R, Lalita Udpa, Adriana Savin, Schiu C. Chan, Rozina Steigmann, Satish S. Udpa : Noninvasive evaluation of Bjork-Shiley convexo concave prosthetic heart valves, NDT&E, Int.(2009), doi: 10.1016/j.ndtteint.2009.01.013

302


Conceptual Design of an Industrial System for Automatic Radiogram Analysis Ryszard SIKORA, Tomasz CHADY, Piotr BANIUKIEWICZ, Marcin CARYK, Przemysław ŁOPATO, Lech NAPIERAŁA, Tomasz PIETRUSEWICZ and Grzegorz PSUJ Department of Electrical and Computer Engineering, Faculty of Electrical Engineering, West Pomeranian University of Technology, al. Piastów 17, 70-310 Szczecin, Poland

Abstract. In this work a conceptual design of an Intelligent System for Radiogram Analysis (ISAR) is presented. ISAR is a complete system, containing both hardware and software solutions, that aims to support radiologists in their work. On the hardware side, we have a PC station with digital radiography scanner. The software side consists of application, that supports variety of scanner standards and implements number of functions designed to augment human-based defect recognition. ISAR also evaluates welds quality and returns information about welding defects to the database. Classification of defects is based on fuzzy logic standards. In this paper, hardware integration and implemented functions are described. Keywords. Radiogram, Non-Destructive Inspection, IQI, image processing

Introduction Non-destructive techniques of radiography inspection are widely used for damage detection in ship, car and aircraft metal structures. Among many kinds of defects and damages, it is also possible to detect weld defects using of industrial radiography technique [1]. The process of welded joints quality assessment must be very fast and accurate. Furthermore, type, size and acceptance level must be also determined for any heterogeneity in weld. The aim of ISAR project (Intelligent System for Radiogram Analysis, in Polish: Inteligentny System Analizy Radiogramów) is a creation of an intelligent expert system of recognition and rating of welds defects. There are various systems for radiography inspection of welding joints, including automatic analysis [2 – 5]. The ISAR is dedicated for shipbuilding industry in Poland. Integration between software and hardware is an important part of a concept of ISAR. Our system is composed of many different elements, all of which have to interoperate with designed software. It contains many image processing, analysis and recognition modules, full list of which can be seen in Section 2.2. All components of our system, both hardware and software, are integrated via single computer program, being core of our system. Software modules are easily integrated due to certain programmistic regime: all functions are in separate C++ files and written with use of an uniform template. This ensures easy integrability, as those

R. Sikora et al. / Conceptual Design of an Industrial System for Automatic Radiogram Analysis

303

can be both copied line by line to source code of core application, or compiled and included as libraries.

1. Hardware The ISAR system requires hardware and software of highest quality, which allows one to obtain high-quality results. On the hardware side we have: a PC computer with 6core 2.8GHz processor, 24 GB DDR3 RAM and graphical card with CUDA support; iCR 3600 image plate scanner [6]; Rad-Icon Shad-o-Box 1024 x-ray image sensor; Eltec P3I3 frame grabber; any TWAIN-interface scanner. Each of these components needed a different approach when it came to integration with designed software. The image plate scanner iCR 3600 was integrated using dedicated iCR software. Our program calls certain iCR program to handle all technical issues and leave the result in a tiff file. This is then read from hard drive by our software, which is transparent from the user’s point of view. The Rad-Icon Shad-o-Box 1024 was integrated via software development kit provided by Eltec P3I3 frame grabber’s manufacturer. The TWAIN interface was implemented with use of a TWAIN Group’s documentation and include files [7]. Figure 1 presents hardware used by ISAR.

2. Software 2.1. Requirements of ISAR-software ISAR application is build of many self-contained functions written in C++ environment and integrated in one software solution to create complete intelligent system. All algorithms implemented in our software take advantage of parallel computing whenever it’s applicable. The application was build for Windows x64 platform. The independence of functions gives the possibility of applications of different image processing procedure for each radiogram separately. We can also test all functions in Matlab software before using them in ISAR complete system. In Section 2.2 a select set of ISAR-functions are presented. 2.2. Implemented functions The Software side consists of application, that supports variety of scanner standards and implements a number of functions designed to augment human-based defect recognition. Those can be separated into few groups, namely: linear and nonlinear filtering algorithms (preliminary image processing), thresholding algorithms, quality of

Figure 1. The view of the hardware used by ISAR.

304


image evaluation, as well as some specialized functions, such as IQI search function (that automatically detects and removes image quality indicators according to the standard). After the preliminary processing the image is shown and ISAR decides on welds quality and returns information about welding defects to the data base. 2.2.1. Preliminary image processing Image processing plays an important role in detecting weld flaws in radiograms. The quality of the images are enhanced by linear and nonlinear filters. The ISAR software implements high pass and low pass convolution filters (in spatial and frequency domain, with changeable mask size), Gauss, Laplace, LoG, Kirsch, Sobel, Prewitt filters. Figure 5b in Section 2.2.2 shows the effect of high pass filtering. Our system is also able to perform morphological operations. In our system, a weighted median filters with definable mask shape are implemented. This type of filters may be used for noise removal (smaller mask size) or background image calculation (bigger mask size). In order to obtain background image or determine center and other regions of the weld - an approximation procedure was carried out. Each line of the weld profile is approximated using the following or similar formula (Figure 2):

§ § ( x − b) · 2 · ¸ + dx + e f ( x) = a ⋅ exp¨ − ¨ ¨ © c ¸¹ ¸ © ¹

(1)

where: a, …, e – approximation coefficients, x – normalized position, f(x) – approximated intensity of image. For minimalization of Root Mean Square Error MSE between signal and intensity approximation, multidimensional Rosenbrock optimization algorithm was used with equation (1). Exemplary results are shown in Figure 3. The image obtained by subtracting the approximated image from the original one is also presented. Higher

Figure 2. Approximation of weld profile.


305

Figure 3. Results of X-ray weld image approximation.

contrast and visibility of defects can be observed. To normalize radiographic images of welded parts the authors propose the use of image normalization function according to histogram point Ibase. The gray level value Ibase is equal to a point for which histogram has maximum value or is calculated as a mean value on determined image area. The value Inew base corresponds to gray level value of welding part (background). The procedure of image normalization is shown on Figure 4. An important parameter describing the radiogram taken into analysis is the signalto-noise ratio. This parameter is also defined by the norm - EN-14784-1. According to this norm SNR is evaluated between background of radiographic image and useful signal which is, in most cases, the signal from the specimen. There are some restrictions on choosing those regions. Firstly, the size of ROI should be at least 20x55 pixels, which gives 1100 values. Moreover, selected ROIs have to be the best representatives of background and specimen – for example the most important part of the specimen should be used for calculating the SNR. Selected ROI is assumed to be series of 55 datasets (groups) containing 20 values each. For every group i, the mean

Figure 4. Image normalization procedure: a) histogram after moving Ibase to Inew base, b) algorithm.

306


value Imean i is evaluated as well as standard deviation ıi. Values of ıi have to be corrected by coefficient following from statistical analysis of filtration obtained as a result of data grouping (2).

σ corr = 1.0179 ⋅ σ i i

(2)

Then, the final value of Imean is calculated as median value of all Imean i. In the same way, the final value of ı is evaluated. This value should be related to spatial resolution of digital radiogram. Thus, ı yields to:

σ SR = σ ⋅ ( SRmax / 88.6 )

(3)

Where SRmax stands for spatial resolution (in μm) measured perpendicularly and horizontally to scanning direction. Finally, the normalized SNR is evaluated as:

SNR =

I mean

σ SR

(4)

2.2.2. Thresholding Thresholding is an important step in almost every algorithm designed for image analysis, radiogram defect detection being no exception. In our software a variety of thresholding algorithms, both local and global, have been implemented. Among those: - Iterative thresholding – computes global threshold in iterative way. It’s starting threshold is located in the middle of image’s dynamic range. In each iteration new threshold is computed using the average of mean values of background and foreground found with previous threshold. Algorithm stops when difference between nth and n+1st threshold is sufficiently small [8]. - Otsu thresholding – computes global threshold by minimizing weighted sum of within-class variance of background and foreground [8]. - Niblack thresholding – computes local threshold according to local mean and local standard deviation over a specific window with use of following formula [9]:

T (i, j ) = m (i, j ) + k ⋅ σ (i, j )

(5)

- Sauvola thresholding – is an expansion of Niblack’s thresholding, uses modified formula [10]:

ª σ (i, j ) º ½ T (i, j ) = m(i, j ) + ®1 + k ⋅ « − 1» ¾ ¬ R ¼¿ ¯ Figure 5c shows the result of Sauvola thresholding.

(6)


307

2.2.3. IQI detection IQIsn function detects how many wires from single IQI are visible on radiographic image according to CNR (Contrast to Noise Ratio) [11, 12]. First the background of part radiographic image containing IQI is generated. This is done by using a median filter with the longitudinal mask. Then the background is subtracted from input image, and the CNR is calculated for each wire. The number of visible wires depends on the minimum calculated value of CNR. Figure 6 shows an example of an image before and after the operation. Table 1 shows results for example image where function detected 5 wires.

Figure 5. Example of implemented function’s work: a) original image, b) results of high pass filtering, c) results of Sauvola thresholding.

Figure 6. IQI detection example: a) original image, b) after IQI detection preprocessing.

308


Table 1. Results of IQI detection on image from Figure 6. Wire MidPos Thicknes[mm] CNR

1 76 0,4 12,2

2 195 0,32 10,52

3 310 0,25 7,17

4 421 0,2 6,79

5 531 0,16 1,99

3. Conclusions The conception of ISAR system is prepared to apply in industrial non-destructive testing. Presented functions do not encompass entire functionality of presented system, as it is freely expandable due to described programmistic approach. Final effect of ISAR project will be a complete intelligent system. Functions described in this text can effectively improve the quality of X-ray radiograms and assist in the extraction of information concerning defects in the weld.

Acknowledgments This work was supported by the Ministry of Education and Science, Poland, under grant N R01 0037 06 (2009–2012).

References [1] Alaknanda, R.S. Anand, Pradeep Kumar, Flaw detection in radiographic weld images using morphological approach, NDT&E International 39 (2006), 29–33. [2] Dong Du, Guo-rui Cai, Yuan Tian, Run-shi Hou, Li Wang, Automatic Inspection of Weld Defects with X-Ray Real-Time Imaging, Lecture Notes in Control and Information Sciences, Vol. 362/2007, 359366 (2007) [3] R. D. Bowman, B. A. Bennett, M. E. Stevenson, Radiographic inspection in failure investigations, Journal of Failure Analysis and Prevention. Vol. 3, Number 3, 73-77 [4] GE Sensing & Inspection Technologies. Inspection Technologies Productivity through inspection solutions. Advanced nondestructive imaging, http://www.ge-mcs.com/download/it-common/GEIT10012EN ndt-brochure.pdf [5] T. Warren Liao, Yueming Li, An automated radiograp.hic NDT. system for weld tnspection:. Part IIFlaw detection,.NDT&E International, Vol. 31, No. 3, pp. 183-192 (1998). [6] http://www.icrcompany.com/Computed-Radiography/3600.html [7] http://www.twain.org/ [8] M. Sezgin, B. Sankur, Survey over image thresholding techniques and quantitative performance evaluation, Journal of Electronic Imaging 13(1), 146–165 (January 2004). [9] Naveed Bin Rais , M. Shehzad Hanif and Imtiaz A. Taj, Adaptive Thresholding Technique for Document Image Analysis, 8th IEEE International Multitopic Conference (INMIC), Lahore, Pakistan, December, 2004 [10] J. Sauvola, M. Pietikainen, Adaptive Document Image Binarization, Pattern Recognition, vol. 33, 110.2, pp. 225- 236, 2000 [11] U. Ewert, U. Zscherpel, K. Bavendiek, Strategies for Film Replacement in Radiography - a comparative study, IV Conferencia Panamericana de END Buenos Aires – Octubre 2007. [12] U. Ewert, U. Zscherpel, C. Bellon, Gerd-Rüdiger Jaenisch, J. Beckmann, M. Jechow, Flaw Size Dependent Contrast Reduction and Additional Unsharpness by Scattered Radiation in Radiography Film and Digital Detectors in Comparison, 17th World Conference on Nondestructive Testing, 25-28 Oct 2008, Shanghai, China.

Material Characterization



311

Quantification of Sigma Phase Precipitation by Magnetic Non Destructive Testing João M. A. REBELLOa,1, Rodrigo SACRAMENTOa, Maria C. L. AREIZAa, K. SANTOS DE ASSISa a Non-destructive Testing, Corrosion and Welding Laboratory – Department of Metallurgical and Materials Engineering COPPE/UFRJ – Federal University of Rio de Janeiro – Rio de Janeiro/RJ - Brazil Abstract. Duplex stainless steels (DSS) are two-phase materials the microstructure of which consists of grains of ferrite (į) and austenite (Ȗ). DSS exhibit a combination of attractive properties: strength, toughness and ductility, stress corrosion cracking resistance and low cost, when compared to the conventional stainless steel. DSS are increasingly used in the oil refinery industry, mainly due to their economical combination of strength and corrosion resistance even in H2S containing environments. However, DSS attain these attractive properties after a controlled solution heat treatment carried out during manufacture, leading to an equal and balanced proportion of the volume fractions of į and Ȗ phases. Regarding the DSS magnetic properties, it is known that į is ferromagnetic and Ȗ is paramagnetic. Any further heat treatment carried out after manufacture, welding operation for example, will change the original 50% į and Ȗ base metal phase proportion. So the weld heat affected zone microstructure, where temperatures reach values between 300°C and 1000°C give origin to non-balanced į and Ȗ phases. Besides, exposition to these high temperatures may lead to precipitation of a deleterious phase, named sigma (ı), and when this occurs the material becomes brittle and its corrosion resistance to H2S is severely impaired. In the present work, the degradation of the material was intentionally obtained by thermally treating DSS specimens, at temperatures ranging from 750°C up to 800°C, after solution heat treatment at 1200°C giving rise to different ı phase contents. These specimens were analyzed by optical microscopic techniques to determine the volumetric fraction of the ı phase. The results showed that samples heat treated in this way, exhibited volumetric fraction of ı phase ranging from zero to 15.63%. In parallel, eddy currents (EC) testing and swept frequency analyses were employed to detect ı phase. A probe consisting of one coil, with cylindrical geometry, 1000 wire turns and a ferrite core of 10mm diameter was designed for the implementation of this technique. The probe exhibited its higher sensitivity at low frequencies around 4 kHz and its inductance was 22.42 mH. The results showed that magnetic testing is a reliable technique to detect and quantify the ı phase in DSS. Keywords. Eddy Current, duplex stainless steel, sigma phase

Introduction DSS present a good combination of mechanical properties and high corrosion resistance due to the presence of two phases (į and Ȗ) in their microstructure [1], and 1

Metallurgy and Materials Dept.Federal University of Rio de Janeiro. COPPE/EE/UFRJ. PO BOX 68505 CEP 21941-972. Rio de Janeiro, RJ BRAZIL E-mail: [email protected]

312

J.M.A. Rebello et al. / Quantification of Sigma Phase Precipitation

are widely used in the marine and petrochemical industries. Such microstructure consisting of two different phases, γ and δ, with very different physical properties (austenite presents high electric conductivity and low magnetic permeability and ferrite presents low electric conductivity and high magnetic permeability) renders the inspection through non-destructive testing techniques, difficult. DSS best properties are obtained in solution with approximately equal proportions of phases į and Ȗ [2,3]. Welding operation may lead to microstructural changes of the base metal and heat affected zone (HAZ), changing the balance of į and Ȗ phases and/or causing the precipitation of deleterious phases due to steel exposition to temperatures ranging from 300º C to 1000º C. The most harmful of the deleterious phases that can be originated in the material microstructure is ı phase, because it presents higher volumetric fraction than others and its precipitation causes chromium depletion in the adjacent regions, impairing mechanical and corrosion properties of the material. The presence of ı phase causes a change of the electromagnetic properties of DSS. Ferrite is ferromagnetic while austenite and σ phase are paramagnetic. Thus an increase in σ phase percent and the resulting decrease in δ phase volumetric fraction render the material behavior more paramagnetic. This characteristic makes the EC technique an interesting non destructive tool for the detection of material degradation caused by an increase in ı phase percent.

1. Theoretical review 1.1. Duplex Stainless Steels Commercial DSS present chromium content, which is an element that favors the formation of į phase, varying from 17% to 30%, nickel content, which is an element that favors the formation of Ȗ phase, varying from 3% to 13% and molybdenum content, which is also a ferrite former element, varying from 2% to 4%, while manganese and silicon are used for alloy deoxidation [4]. Due to their good combination of properties, DSS have been increasingly used in the chemical, power generation and on off-shore industries [1]. In the off-shore industry, DSS are used both, in oil and gas production and processing and transportation, in the production of water and water systems at the sea, in the manufacture of subsea umbilical and submerged components without cathodic protection, among others [5]. These steels have also been used successfully in the petroleum refining industry in pipes, plates, pressure vessels and heat exchangers in corrosive environments containing H2S and water with chloride [6,7]. Maehara [9] reported the relationship between the toughness at room temperature and the volumetric fraction of σ phase. A decrease in the impact energy absorbed by the material even for a small increase in the σ phase volume fraction was observed. These measurements were made at room temperature.

313


1.2. Magnetic properties As already said, the two phases present in DSS microstructure show very different magnetic properties. Austenite, Ȗ phase presents high electric conductivity and low magnetic permeability and į phase presents low electric conductivity and high magnetic permeability [3]. Besides, į phase shows a ferromagnetic behavior while Ȗ phase shows a paramagnetic behavior [10, 11]. When both phases are in equal proportion or with a higher į phase percent DSS show a ferromagnetic behavior. The presence of σ phase causes a change in the balance between į and Ȗ, and consequently the magnetic properties of the steel also change. Some authors mention ı phase as non-magnetic [8, 10, 11, 12], so an increase in its volumetric fraction causes a decrease in the ferromagnetic behavior of the material. Some papers [12,13] show that magnetic saturation of DSS is proportional to δ volumetric fraction, and thus its decrease due to an increase in ı phase causes a change of the magnetic properties of the material. Mészáros [12] obtained magnetization curves for DSS with different į phase percents using the vibrating sampler magnometer (VSM) technique. From those magnetization curves, a quantitative correlation between the values of magnetization in the saturation region of the curves and į phase percent in different samples was obtained. It was observed a good correlation between saturation induction and ferrite content obtained with VSM and į phase content [12]. The magnetic properties of phases į and Ȗ are well described in literature, however the magnetic properties of ı phase, and its influence on the properties of DSS need to be further studied. The only mention founded in the literature about this matter was from Todorov [11], who showed that a qualitatively correlation exists between ı phase percent and conventional EC parameters. 2. Materials and methodology 2.1. Materials - Eight DSS samples, specification UNS 31803, with length 50 mm, width 32 mm and thickness 4 mm were used. Table 1 shows the correspondingly chemical composition. Table 1. Chemical Composition of DSS (UNS S31803)

Element UNS S 31803 (%wt)

C

Si

Mn

Ni

Mo

Cr

0.034

0.283

1.78

5.58

2.87

22.02

- Heat treatment furnace with maximum temperature of 1300ºC; - Optical microscope (OM), ZEISS Axio Imager, with working station and software for image acquisition, processing and analysis (Axio Vision and Image Pro);

314


- Scanning electron microscope (SEM) of 30 kV with low vacuum capacity, ZEISS EMA-25, equipped with Energy Dispersive Spectroscopy detection system (EDS), Oxford Inca Synergy 350 integrated system; - EC equipment, Olympus Omniscan MX; - Impedance analyzer HP 4295A; - EC probe customized for material characterization; 2.2. Methodology •

Heat treatments

Initially the eight samples were submitted to a preliminary solution heat treatment aiming at obtaining a percent balance of approximately 50% of both δ and γ phases. The solution heat treatment was conducted at 1120º C during one hour, followed by water quenching. Six samples received additional aging heat treatments which introduced different amounts of σ phase volumetric fractions. Table 2 shows the correlation between the different heat treatments and the percentage of each phase obtained in all the samples used in this study. Table 2. Correlation between samples, heat treatments and phase content. Austenite Ferrite Time of Sample Temperature heat of heat treatment treatment 1 1120º C 1h 44.56% 55.44% 2 1120º C 1h 44.51% 55.49% 3 800º C ½h 43.38% 55.92% 4 800º C ½h 43.56% 55.69% 5 800º C 1h 49.10% 42.46% 6 800º C 1h 45.76% 45.88% 7 750º C 2h 47.25% 37.12% 8 750º C 2h 47.28% 37.64%

•

Sigma phase

0% 0% 0.70% 0.75% 8.44% 8.36% 15.63% 15.08%

Metallographic analysis

A small portion (15 mm long, 5 mm wide and 4 mm thick) was removed from each sample so that they could be examined through SEM, EDS and OM. For this purpose the conventional method of mechanical polishing were used followed by an electrolytic chemical attack in 30% solution of NaOH – 3V over a time varying from 5 to 60 seconds in order to reveal their microstructures. ASTM A 923 standard was followed in these procedures [14]. After preparation the samples were examined by OM and 30 images were randomly captured in different regions of the transversal section of each sample. Those images were used for counting the phases volumetric fraction in the microstructure as specified by ASTM E1245-03 [15]. Besides OM, SEM images were obtained and EDS analysis were made.

315


• Probe simulation In this study the simplest geometrical shape for probe design, e.g, a cylindrical metallic bobbin core was used, and the mathematical model adopted was an azimuthal approximation. Figure 1 shows the principal regions of the model: (A) a ferritic core, (B) a bobbin with N turns, and (C) the air. The vertical axis corresponds to the symmetrical axis, and the horizontal axis is the parameter r, so, all of the components are in the rz plane. An alternated induced current is applied on B region in the normal direction to the rz plane. The corresponding generated magnetic field is more intense in the core region. (Fig. 1, right). The mathematical calculation of the induced current follows equation 1, which is the time-harmonic formulation (Vloop) in a quasi-static operational condition. It contains the Curl of the magnetic vector potential. In this equation ω is the angular frequency, σ the conductivity, ȝ the permeability, ε the permittivity, and Vloop the voltage applied to the coil. Outside the coil and core, σ is set to zero.

& & & σ.Vloop ( jωσ − ω2ε ) Aϕ + ∇ × (μ −1∇ × Aϕ ) = Jϕe = 2πr

(1)

r(mm)

Core

Coil

Figure 1. Azimutal model for EC simulations.

The current induced, I, is calculated from the current density,

&e & .ds = I

³J

J ϕe (Eq. 2) (2)

S

The model has to define the boundary condition for the outer border, called magnetic insulation. This implies that the magnetic vector potential is zero at the boundary, corresponding to a zero magnetic flux (Eq. 3). Other boundary condition is the symmetry of the magnetic potential vector with respect to the z axis. Finally, the magnetic potential must be continuous (Eq. 4) inside A, B and C regions, e.g., the core, the copper wires, and surrounding air.

n× H = 0

(3)

316


•

n × ( H1 − H 2 ) = 0 Probe construction

(4)

A probe consisting of one coil, with cylindrical geometry, 1000 wire turns and a metallic core of 10 mm diameter was designed for the implementation of this technique. The internal radius was 5 mm and the external radius 5 mm. The material of the core was ANSI 1045 type. Swept frequency measurement was made to detect the region of frequencies with high sensitivity. The probe exhibited its higher sensitivity at low frequencies, around 4 kHz and its inductance was 22.42 mH, figure 2. But, regarding the second derivative curve, there is also a point at 2kHz where a good option for high sensitivity measurements is possible, as another alternative to be used.

Figure 2. Swept frequency measurement, probe at air.

•

Measurement of magnetic field attenuation

The magnetic field generated by the coil is attenuated when it spreads through the ferrous material. This magnetic field attenuation was measured in samples of DSS with 4 mm of thickness and (4x4) mm of area. BRelative was the parameter used for the quantitative evaluation of the interaction of the magnetic field with the material, it is defined by equation 5, where BAir is the magnetic flux density measured at air and BMaterial is the magnetic flux density measured in contact with the material. The schematic representation of the setup is shown in figure 3. The measurements of magnetic flux density were made with a FW Bell Gaussmeter (Model 6010) and the current source was the Agilent 33250A. The current of the probe was setup at 10 Vpp and 40 Hz.

BRe lative( Air ) =

(BAir − BMaterial ) B Air

(5)


317

Figure 3. Schematic setup for the measurements of magnetic attenuation. (A) Direct measurement of the magnetic field generated by the bobbin; (B) Magnetic measurements at the backside of the sample.

•

Measurement of conventional EC

In this test, the EC signal produced at a point distant from the conductive samples (signal in air, Fig. 3A) was taken as reference and then, the probe was brought in contact to each of the eight samples surface (Fig. 3B). The amplitude and phase angle of those signals were measured. In order to statistically improve the results 30 measurements were taken for each sample and at the end, the average was calculated obtaining the final value for amplitude and phase of each sample. Then, the values obtained during the EC tests were correlated with the percentage of σ phase in each sample. •

Measurement of swept-frequency

Swept frequency measurements were made in a range of 40Hz until 30kHz. Two different electrical parameters were analyzed: impedance and inductance, using the same setup described for EC. In equation 6, a parameter LRelative was defined, in order to compare the value of the inductance at one specific frequency (2 and 4 kHz). Similar parameter was defined for the electrical impedance (ZRelative).

LRe lative ( Air ) =

3.

(LAir − LMaterial ) LAir

Z Re lative ( Air ) =

(Z Air − Z Material ) Z Air

(6)

Results and discussion

3.1. Metallographic analysis Figure 4 shows the SEM image of sample 1, where phases į and Ȗ in approximately the same proportion can be observed in the DSS microstructure.

318


Figure 4. SEM image of sample 1.

Figure 5 shows a SEM micrograph of Sample 6 (S6) and its respective EDS spectra. In this figure, quantitative EDS analyses were performed in three points, each one representing one of the three focused phases. Point 1 represents į phase, demonstrated in the spectrum by the presence of chromium, smaller presence of nickel and presence of molybdenum. Point 2 represents Ȗ phase, demonstrated by higher nickel and smaller chromium presence when compared to point 1. Point 3 represents σ phase, demonstrated by chromium presence similar to point, higher presence of molybdenum and smaller nickel presence.

Figure 5. SEM image indicating the Sample 6 points where EDS analyses were conducted and their respective spectra.

3.2. Conventional EC Figure 6A shows the lift-off signals obtained in the EC test for samples 2, 4, 6 and 8. In this figure the signals are shown in the impedance plane, characteristic of EC technique. The same signals of the impedance plane are presented through the


319

components of the plane (resistive or inductive) as shown in figure 6B. A frequency of 4 kHz with a gain of 32dB and a 4V drive were used to obtain those results. Figure 6 shows that EC technique was able to differentiate the four samples tested in this study, which received different heat treatments and thus exhibited different volumetric fraction of σ phase. It is worth mentioning that the inductive component (XL , Fig. 7B) of the EC test was the most sensitive to detect small variations in σ phase contents, compared to the resistive component.

Figure 6. Lift-off signals obtained for the four samples through the EC technique with frequency of 4 kHz. (A) Impedance plane; (B) Resistive and inductive components.

3.2. Swept-Frequency Swept frequency measurements were made in a range from up to 30 kHz. Two different electrical parameters were analyzed: impedance and inductance. Figure 7 shows the good differentiation of the samples with different volumetric fraction of σ phase, for frequencies up to 9kHz.

Figure 7. Sweep frequency measurements of inductance.

320


Figure 8 shows the values of the relative parameters, LRelative and ZRelative (Eq. 6) for the two best operation frequencies, as indicated in figure 2. At 2 kHz, 28% of signal difference was obtained between a sample without σ phase and the samples 7 and 8, 15.63% and 15.08% of σ phase, respectively. A higher sensitivity was observed in the measurement with small σ phase content. For this case, we observed a change of almost 6% in the electrical parameters reported.

Figure 8. Attenuation of electrical parameters with the influence of σ phase (Equation 6)

Small contents of σ phase are enough to change the electrical impedance and inductance. This behavior is not caused by an increase in δ phase of the material, but by the presence of the deleterious σ phase. Figure 9 shows the attenuation produced by σ phase (Eq. 5), and this result was obtained after 10 measurements for each sample and the standard deviation was less than 2%. The attenuation of magnetic field is strongly dependant on volumetric fraction of σ phase. This is in agreement with the previous results of conventional EC and swept frequency. The absolute value of magnetic field attenuation, observed in figure 9, could be a function of the thickness sample, and this implies that for 4 mm we observe 10 % of difference of attenuation. But for a thicker sample a greater absorption of magnetic field by the material could be observed

Figure 9. Measurements of BRelative (Equation 5) as function of σ phase content.


321

4.

Conclusions

-

Sigma phase is responsible for the differences in the magnetic attenuation within DSS; Swept frequency measurements have shown to be an efficient tool for the characterization of DSS in the range of 0.70% to 15.63% of volume fraction of sigma phase; The eddy current technique has proved to be equally able to detect and quantify sigma phase in DSS samples; Low frequencies are more suitable for reliable material characterizations when using 2kHz and 4kHz probe frequency; Differences of 28.9% and 33.7% in the signal for the inductance (L) and impedance (Z) parameters respectively were observed for the samples containing sigma phase in relation to air.

-

-

References [1] V. CHIAVERINI, Aços e Ferros Fundidos, São Paulo, 2005. [2] H. Sieurin, Sigma phase precipitation in duplex stainless steel 2205, Materials Science and Engineering A444 (2006), 271-276. [3] K.S. de Assis, Caracterização por Correntes Parasitas da Degradação Microestrutural de Aços Duplex, 10ª COTEQ (2009). [4] C.H. Shek, Review of Temperature Indicators and the Use of Duplex Stainless Steels for Life Assessment, Materials Science & Engineering 19 (1996), 153 – 200. [5] M. Tystad, Application of Duplex Stainless Steel in the Off-Shore Industry, Stainless Steel World – KCI Publishing (1997). [6] API – American Petroleum Institute, Use of Duplex Stainless Steels in the Oil Refining Industry, API Technical Report 938-C (2005). [7] T.H. Chen, The effect of High Temperature Exposure on the Microstructural Stability and Toughness Property in a 2205 Duplex Stainless Steel, Materials Science & Engineering A338 (2002), 259 – 270. [8] C.M. Souza Jr., The ı Phase Formation in Annealed UNS S31803 Duplex Stainless Steel: Texture Aspects, Materials Characterization 59 (2007), pp. 1301 – 1306. [9] Y. Maehara, Precipitation of ı Phase in a 25Cr-7Ni-3Mo Duplex Phase Stainless Steel, Tetsu-to-Hagané 67 (1981). [10] N. Sathirachinda, Depletion effects at phase boundaries in 2205 duplex stainless steel characterized with SKPFM and TEM/EDS, Corrosion Science 51 (2009), 1850 – 1860. [11] E.I. Todorov, Correlation between NDT Measurements and Sigma Phase Contents in Duplex Stainless Steels, Review of Quantitative Nondestructive Evaluation 28 (2009), 1259 – 1266. [12] I. Mészáros, Magnetic Characterization of Duplex Stainless Steel, Physica B 372 (2006), 181 – 184. [13] I. Mészáros, Complex Magnetic and Microstructural Investigation of Duplex Stainless Steel, NDT & E International 38 (2005), 517 – 521. [14] ASTM – American Society for Testing and Materials, Standard Test Methods for Detecting Detrimental Intermetallic Phase in Duplex Austenitic/Ferritic Stainless Steels, ASTM A 923 – 03 (2003). [15] ASTM – American Society for Testing and Materials, Standard Practice for Determining the Inclusion or Second Phase Constituent Content of Metals by Automatic Image Analysis, ASTM E 1245 – 03 (2006).

322


Nondestructive Characterization of Neutron Induced Embrittlement in Nuclear Pressure Vessel Steel Microstructure by using Electromagnetic Testing I. ALTPETER1, G. DOBMANN, G. HÜBSCHEN, M. KOPP and R. TSCHUNCKY Fraunhofer-Institut für Zerstörungsfreie Prüfverfahren, Saarbrücken, Germany

Abstract. Using nuclear power for energy generation, pressure vessel walls are exposed to neutron fluences of different levels depending on the distance to the core. Hence materials undergo a change of microstructure in terms of embrittlement, to be measured as toughness reduction and shift of the Ductile–toBrittle Transition Temperature (DBTT) to higher temperatures. Normally plant safety concerning this change in microstructure is ensured by destructive testing of surveillance samples. These are standard tensile and Charpy specimens which consist of exactly the same material as the pressure vessel and its weldments, being exposed to accelerated irradiation rates within special irradiation channels allowing a pronounced ageing. During revision downtime of the plant these samples are tested destructively in standard tensile tests at 423 K and 548 K respectively or by measuring the impact energy as a function of temperature in Charpy tests to determine the shift of DBTT. It is demonstrated that electromagnetic parameters allow characterizing the changes in the microstructure generated through neutron irradiation. After a defined calibration process a quantitative characterization of the embrittlement especially in terms of DBTT is possible. This has been demonstrated for reactor pressure vessel steels according to western design as well as to eastern specifications. As testing methods 3MA (Micromagnetic, Multiparameter, Microstructure and stress Analysis) and dynamic magnetostriction using EMATs (ElectroMagnetic Acoustic Transducers) have been applied. Keywords. EMAT, 3MA, neutron embrittlement, pressure vessel

Introduction Nondestructive material characterization techniques have traditionally been employed to detect, classify and size defects in materials. However in the last two decades a significant amount of effort has been invested to develop NDT techniques which can reliably characterize materials in terms of properties describing the fitness for use. In the case of power plant components, such as pressure vessels and pipes, the fitness for use under mechanical loads is characterized in terms of the determination of mechanical properties such as mechanical hardness, yield and tensile strength, toughness, shift of Ductile-to-Brittle Transition Temperature (DBTT), fatigue strength. 1

Corresponding Author.

I. Altpeter et al. / Nondestructive Characterization of Neutron Induced Embrittlement

323

With the exception of hardness tests which are weakly invasive, all of these parameters can be determined within surveillance programs by using destructive tests on special standardized samples (Charpy specimens). The specimens are exposed in special radiation chambers near the core of the nuclear power plant (NPP) to a higher neutron flow than at the surface of the pressure vessel wall. From time to time these specimens are removed from the chambers and used for destructive tests. The number of the Charpy specimens is limited and in the future it will be very important that nondestructive methods are available to determine the mechanical material parameters on these Charpy specimens. Furthermore in situ characterization of the reactor pressure vessel inner wall through the cladding is of interest additionally to the measurements on Charpy specimens. To solve this task a combination testing method based on 3MA (Micromagnetic, Multiparameter, Microstructure and stress Analysis) [1] and dynamic magnetostriction by using an EMAT (ElectroMagnetic Acoustic Transducer) [2] was developed. Electromagnetic methods have a high potential to characterize neutron induced microstructure states.

1. Assessment of the Embrittlement In addition to the mechanical properties, the embrittlement phenomenon also influences the magnetic properties. Even the smallest changes in the materials state can affect the magnetic domain structure. Micromagnetic test methods have a high potential to detect the change of microstructure defects e.g. precipitations since they sensitively react to changes of the domain wall (which separate the magnetic domains) configuration. Therefore electromagnetic procedures could potentially be used to assess the material embrittlement. The neutron induced embrittlement results in microstructure changes. These microstructure changes lead to the generation of vacancies and precipitations of Cu-rich coherent particles (radius: 1-1.5 nm). This results in an increase of yield strength and tensile strength, a decrease of Charpy energy upper shelf value and an increase of the shift of DBTT by 41 joule (ǻT41). The potential of electromagnetic testing methods for detection of microstructure states is based on the analogy between the interaction of dislocation with microstructure states and of Bloch-walls with microstructure states. Similar to the movements of dislocations under mechanical load Bloch-walls move under magnetization. Therefore an analogy exists between mechanical technological values (e.g. hardness values) and magnetic values. So the interaction between dislocations and copper particles leads to an increase of mechanical hardness and the interaction of the copper particles and Bloch-walls leads to an increase of magnetic hardness. The potential of electromagnetic testing methods for the detection of copper precipitations was demonstrated on the Cu-rich pressure vessel and piping steel 15 NiCuMoNb 5 (1.65 weight % Cu) [3]. The analogy between mechanical and magnetic hardness is shown in the behavior of both hardness values as function of simulated service time at 673 K (see Figure 1). The hardness maximum is in the order of a service time of 1000 h.

324


Simulated service time at 673 K [h]

Figure 1. Magnetic coercivity compared with the mechanical hardness for the detection of copper precipitations in Cu-rich pressure vessel and piping steel 15 NiCuMoNb 5 (1.65 weight % Cu). Mechanical hardness: circles, magnetic hardness: squares.

Since the dynamic magnetostriction using an EMAT is sensitive for lattice defects it was assumed that a magnetostrictively excited standing wave in the pressure vessel wall also reflects the neutron embrittlement and first experiments were performed with a special designed magnetostrictive transducer at Charpy specimens in the hot cell in order to principally demonstrate the potential. Using several electromagnetic measurements at the same time, a variety of measuring quantities is derived for each measurement cycle. When combined, they achieve the desired result (e.g. material property) more efficiently compared to individual measurement. By using a calibration function or pattern recognition the desired quantity of an unknown set of samples investigated by that method can be detected nondestructively.

2. NDT Characterization of the Neutron Irradiation Induced Embrittlement Depending on the specific design of a pressure vessel – which varies in different countries – the pressure vessel material in nuclear power plants is exposed to neutron fluences in the range between 5.6×1018 n/cm2 and 86.0×1018 n/cm2. In order to characterize the neutron irradiation-induced embrittlement, Charpy specimens exposed to neutron fluence in the above mentioned range have been investigated in a hot cell at AREVA NP GmbH whereby only the 3MA and EMAT sensors were arranged within the hot cell and the electronic equipments (3MA and EMAT device) was outside (see Figure 2). These Charpy specimens (base material and weldments) of eastern and western design have been provided by AREVA NP GmbH and Research Centre Dresden-Rossendorf e.V. (see Table 1)[4].


325

3MA testing device

EMAT testing device

Neutron irradiated Charpy-specimens of reactor pressure vessel materials of western and eastern design Base material and weldments

Figure 2. Used equipment for electromagnetic measurements in the hot cell

Sample set of neutron irradiated base material Material

Short term Neutronfluence [n/cm²]

Range ΔT41 [K]

22 NiMoCr 3 7

P7

0 – 40.3 x 1018

0 - 32

20 MnMoNi 5 5

P141

0 – 10.7 x 1018

0-9

22 NiMoCr 3 7

P147

0 – 43.8 x 1018

0 - 23

ASTM A508 C1.3 (22 NiMoCr 3 7)

JFL

0 – 86.0 x 1018

0 - 78

ASTM A533B C1.1

JRQ

0 – 98.0 x 1018

0 - 221

(20 MnMoNi 5 5 )

Sample set of neutron irradiated weldments Material

Short term Neutronfluence [n/cm²]

Range ΔT41 [K]

S3NiMo3-UP

P16

0 – 11.7 x 1018

0 - 67

S3NiMo1-UP

P141

0 – 49.7 x 1018

0 - 38

S3NiMo1

P140

0 – 10.4 x 1018

0 - 21

Table 1. Overview of the set of neutron induced Charpy specimens (base material and weldments)

Furthermore Charpy tests were carried out for determination of the shift of DBTT (reference values).

326


Based on the data base consisting of EMAT and 3MA measuring quantities and the reference values pattern recognition and regression functions for calibration were developed. One part of each Charpy specimen set was used to calibrate and the other part – independently selected – was taken to test the model. The pattern recognition algorithm, mentioned above, was used to obtain approximation values of the shift of DBTT, which is a measure of the embrittlement (see Figure 3 and Figure 4).

Figure 3. Predicted shift of the DBTT for RPV-base material versus shift of DBTT determined by the Charpy test

Figure 4. Predicted shift of the DBTT for RPV-weldments versus shift of DBTT determined by the Charpy test

It was demonstrated that the prediction of the shift of the DBTT can be performed due to the micro-magnetic procedure. For the validation samples (base material) a correlation coefficient of R²=0.98 and a RMSE (root mean square error) of 10.13 K was obtained. In the case of the weldments the correlation coefficient was R²=0.99 and the RMSE value was 3.18 K.


327

With regard to in situ measurements on the reactor pressure vessel inner wall through the austenitic cladding test measurements with 3MA and the dynamic magnetostriction-technique using an EMAT were carried out at a ferritic test piece with 8 mm thick austenitic cladding. In order to ensure deep penetration of the electromagnetic fields, low-frequency EMATs ( 40 kHz) were applied, and hence it was possible to characterize the material condition in the ferritic base material directly below the austenitic cladding from the clad side. Figure 5 shows the principle of such an inspection method [5] which up to now is not yet realized for inservice inspection.

Figure 5. Principle of the magnetostrictive excitation of a standing wave in thickness direction of the pressure vessel wall by using an EMAT

The possibility of exciting and receiving ultrasonic waves in the ferritic base material through the cladding was documented by using HF-coil-configurations on clad test pieces by superimposing magnetic fields excited by u-shaped electromagnets. Values of signal-to-noise ratio of 23 dB were obtained on a ferritic test object with a wall thickness of 30 mm and with an austenitic cladding thickness of 8 mm. The magnetizing field strength was 260 A/cm and the selected frequency of the EMAT was 40 kHz.

8 mm cladding

8 mm lift-off

Figure 6. Ultrasound signals by an excitation on the austenitic clad side (left); Ultrasound signals by an excitation on the side without cladding with 8 mm lift-off (right)

328


In a second experiment it was demonstrated that ultrasonic signals can still be produced even with a coil lift-off of 8 mm which was achieved by using an 8 mm thick PVC plate. The obtained EMAT-ultrasound echo signals are presented in Figure 6. Comparing the two acoustic signals it is obvious that the cladding is influencing the excitation of the magnetostrictively generated ultrasonic wave mainly by lift-off and not by eddy-current damping. Furthermore it also was shown that magnetic Barkhausen noise signals can also be received through the austenitic cladding with a good signal to noise ratio (Figure 7). It was demonstrated that the δ-ferrite changes in the cladding as well as a butt weld beneath the cladding do not disturb the ultrasonic amplitude very much (Figure 8).

Figure 7. Barkhausen noise profile curve measured on a ferritic testing sample through the 8 mm austenitic cladding

Figure 8. Variation of the ultrasonic signal amplitude along a butt weld beneath the cladding


329

Therefore the non-destructive detection of the neutron irradiation by using 3MA techniques and the dynamic magnetostriction using an EMAT in a combination should be realized by developing an optimized transducer in the near future. The presence of the austenitic cladding is not a strong disturbance effect.

3. Conclusion By combining 3MA quantities with quantities derived from dynamic magnetostriction measurements using an EMAT the shift of DBTT on irradiated Charpy samples of base material and weldments of reactor pressure vessels of western and eastern design can be determined non-destructively. 3MA and EMAT-ultrasound signals can be measured through an austenitic cladding independent on microstructure changes in the cladding and the material beneath the cladding (base material/weldments). On the basis of these results the development of an inservice inspection technique is possible which allows determining the shift of the DBTT non-destructively directly on the inner diameter surface of the reactor pressure vessel.

Acknowledgement The funding of the Federal Ministry of Economics and Technology of Germany, reference number 1501352, is very much acknowledged. Thanks to AREVA NP GmbH and Research Center Dresden-Rossendorf e.V. for the supply of the Charpy specimens, the destructive tests and assistance at the investigations in the hot cell.

References [1] Altpeter, I., Becker, R., Dobmann, G., Kern, R., Theiner, W.A., Yashan, A.: Robust solutions of inverse problems in electromagnetic non-destructive evaluation, Inverse Problems 18, 2002, 1907-1921 [2] Wolter, B., Dobmann, G.: Micromagnetic testing for rolled steel, ECNDT 2006 Proceedings, Berlin, Sept. 25 - 29, 2006, Th.3.7.1 [3] Altpeter, I, Dobmann, G., Szielasko, K.: Nachweis von Cu-Ausscheidungen mittels mikromagnetischer Prüfverfahren. Abschlussbericht, BMWi Reaktorsicherheitsforschung, FKZ 1501219, GRS Forschungsbetreuung, Köln, IZFP-Bericht Nr. 020120-TW, 2002 [4] Hein, H., Keim, E., Schnabel, H., Seibert, T., Gundermann, A.: Final Results from the Crack Initiation and Arrest of Irradiated Steel Materials Project on Fracture Mechanical Assessments of Pre-Irradiated RPV Steels Used in German PWR, J. ASTM Int. 6, Paper ID JAI101962, 2009 [5] Altpeter I., Dobmann, G., Hübschen, G., Kröning, M., Verfahren zur Messung der Neutronenversprödung unterhalb der Plattierung von Druckbehältern, Patent Nr. WO 2008/135054

330


In-Line Thin Film Characterization Using Eddy Current Techniques Susanne HILLMANN, Marcus KLEIN, and Henning HEUER Fraunhofer Institute for Nondestructive Testing Dresden, Germany Abstract. Thin films are successfully used in many different ways as functional layers in various areas of science, technology, and medicine. The coatings are produced using different methods, such as evaporating, sputtering, or chemical vapor deposition and have different properties, such as adhesion, specific electric resistance, and diffusion coefficient. For all applications, the analysis of coating thicknesses plays an important role, because most of the operating and performance characteristics are a function of the coating thickness. Here, in particular, are recognized in-line-methods, working contact-less, very fast, and in a vacuum. For coatings with thicknesses in the nanometer range and lower micrometer range, Eddy Current systems are especially well adapted. By using a special transmission mode sensor application, the method is very well suited for in-line inspections. Through optimization of the measuring system, sensors, and algorithms it is possible to largely widen the limits of the measurement system and to achieve very high accuracy. The following report presents results obtained with an Eddy Current measurement system working in transmission mode on samples with different single-layer and multi-layer coatings in the nanometer range. Keywords: Eddy Current transmission mode, thin film technology, film thickness

Introduction The studies reported here are motivated primarily by questions coming from the solar industry. In modern coating facilities various thin film layers with high and low conductivity and thickness ranges in the nanometer to micrometer scale are coated on substrates of glass, silicon, or foils. Process-related thicknesses cannot always remain constant, and therefore the coating thickness must be monitored in-line. For the in-line monitoring of the coating thickness the non-contact Eddy Current method is suitable because of its excellent potential for automation and high sensitivity to conductivity changes. This paper presents test results of the performance and efficiency of this method.

1. Measurement Method 1.1. Sensor Arrangement and Measurement Equipment The Eddy Current method can be applied with different sensor arrangements. For inline monitoring of thin, conductive layers, the transmission mode arrangement is very appropriate. In contrast to a surface probe, which locates the transmitting and receiving coil in one sensor body and is applied from one side of the samples surface, the coils in

S. Hillmann et al. / In-Line Thin Film Characterization Using Eddy Current Techniques

331

the transmission mode are separately arranged at a fixed distance from each other. The sample must be thin and flat and must be aligned between the sensors. Figure 1.a) illustrates a schematic of this application. It utilizes for both sensors identical singlecoils with a special core and 30 windings each. The coils have a width of 25 mm and excite a relatively large electromagnetic field. The advantage of this setup is the low influence of misalignment of the sample between the sensors on the test result, making this very practical for in-line inspections. Additionally, it is possible to achieve a larger distance between sensor and sample using the transmission mode [1,2]. In the solar industry, a large distance between sensor and sample is often required, especially when the solar cells in quite massive and prone to bending carriers are passed through the coating line. Therefore, the challenge is to develop a transmission arrangement with the greatest possible measuring gap of several centimeters in thickness at very high resolution (in the nanometer range).

a)

b)

Figure 1.a) Schematic illustration of the Eddy Current method at flat samples by use of transmission mode; 1.b) EddyCus TF® Eddy Current desktop system applying the transmission mode.

Using the EddyCus TF® system developed at Fraunhofer IZFP Dresden, the sensor-sensor distance can be increased up to 60mm, still maintaining a high thickness resolution. Figure 1.b) illustrates this device. All of the following presented measurements were performed with this device. 1.2. Optimum Measurement Frequency Especially for very small thicknesses, it is important to choose the optimal measurement frequency to establish a good separation between data from sample thickness variations and from conductivity variations. This is particularly important when process-related temperature changes may alter the conductivity of the sample. Depending on the measuring frequency, both material properties can be separated from each other in varying degrees; Figure 2 illustrates this coherence. The left diagram in figure 2.a) illustrates a common impedance plane diagram. Here, the measurement results on samples with different wall thickness and different conductivity can be very clearly separated in case of high wall thickness (lower area of the diagram). In case of very low wall thickness (higher area of the diagram) the curves run nearly congruent [3]. The diagrams in figure 2.b) illustrate the determination of angle Į as an instrument for describing the separation between conductivity and wall thickness. The schematic

332


illustrates the points in the complex impedance plane for measurements at thin metal sheets of different thickness and conductivity. Point M1 is obtained by the measurement on a thin sheet with conductivity ı1 (e.g. aluminum) and the thickness d1. Measuring point M2 is generated by a measurement on a same material sheet (conductivity ı2=ı1), but higher layer thickness d2>d1. In contrast, the point M3 results from the measurement on a sheet of the primary thickness of d3=d1, but with a material of higher conductivity ı3>ı1 (e.g. copper). The left image shows this correlation at an unfavorable measurement frequency. Based on natural variations, the measurement points M2 and M3 are difficult to separate from each other. They are located very close to each other, because the impedance locus of conductivity and thickness are almost superimposed. The image on the right illustrates the complex impedance plane for the same samples at a suitable measurement frequency. In this case, points M2 and M3 are well separated from each other.

Figure 2.a) common impedance plane for different conductivity and wall thicknesses. 2.b) Schematic illustration of the complex impedance plane with an adverse measurement frequency f1, which produces poor separation of conductivity and thickness, and results in a small angle Į (left); a favorable measurement frequency f2 with a large opening angle Į is shown on the right.

A dimension for separability of both properties is the opening angle Į, located within the starting point M1 and points M2 and M3. The larger the angle Į, the better conductivity and coating thickness can be separated in the measurement signal, and small film thicknesses in the nanometer range can be obtained with high resolutions even at higher conductivity variations. Depending on the conductivity of the material, the opening angle Į generates its maximum at different measurement frequencies, whereas the frequency with the maximum angle increases with decreasing conductivity [3,4]. To determine the optimal measurement frequency for higher conductivity of around 50 MS/m, a series of experiments using thin metal sheets of aluminum (30 MS/m) and copper (56 MS/m) with different wall thicknesses of 10, 20 and 30 μm were performed. All samples were cut in a circle form from thin foils with a diameter of 100 mm. All samples were measured in transmission mode using several measurement frequencies ranging from 10 kHz to 1 MHz. The opening angle Į was calculated for each data set of different measurement frequencies. The results are illustrated in Figure 3. The maximum opening angle Į for the tested materials was established at a frequency at about 400 kHz, which can be used as the optimal


333

measurement frequency for this conductivity range. This obtained optimum frequency is determined for the described measurement system in transmission mode for the used kinds of sensors. Each sensor- and measurement system and each conductivity range has its own optimum measuring frequency, whereas the optimum measuring frequency increases with decreasing film conductivity. [4] angle between thickness and conductivity 25

Cu

,d ͘

20

angle Į [°]

Alu, d ͘ Į

15 10 5 0 0,01

0,1

1

frequency [MHz]

Figure 3. Determination of optimal measurement frequency for high conductivity by calculating the opening angle Į from measured data using thin aluminum and copper sheets with different thicknesses

2. Single Layer Systems in the Nanometer Range 2.1. Description of the Samples For the validation of the measurement system, two sample sets with different conductive layers were prepared. Conductive layers with thicknesses between 5 nm and 205 nm were applied by sputtering on silicon wafers with a diameter of 100 mm and a thickness of 500 μm. For sample set 1 copper layers were used to evaluate the performance of high conducting layers. In contrast, sample set 2 was coated with Tantalum silicon nitride (TaSiN), which has a much lower conductivity then copper. The Table 1 lists all used samples, the intended film thickness d and the sheet resistance Rs obtained via the touching four-point method. Table 1. List of samples from sample set 1 and 2 including intended thickness d and sheet resistance Rs Sample set 1 (copper layer) Nr. d [nm] Rs [ȍ/͘] Cu-1 5 57,5 Cu-2 10 11,3 Cu-3 15 5,52 Cu-4 20 2,95 Cu-5 25 2,04 Cu-6 50 0,77 Cu-7 55 0,68 Cu-8 100 0,30 Cu-9 105 0,28

Nr. Ta-1 Ta-2 Ta-3 Ta-4 Ta-5 Ta-6

Sample set 2 (TaSiN-layer) d [nm] Rs [ȍ/͘] 10 455 15 307 50 98,6 55 93,8 200 27,9 205 26,8

Eddy Current measurements were performed on each sample with the EddyCus TF desktop system using an optimal measurement frequency. The measuring gap was varied between 20 mm and 60 mm, and the measurement was repeated eight times for

334


each adjustment. In a first step, calibration curves were recorded to determine the Eddy Current parameter as a function of film thickness. Since there was an almost linear relationship between these two parameters, the calibration function was calculated using the measured values from samples with the lowest, middle and highest layer thicknesses. Using this calibration function, the film thicknesses of the remaining samples were obtained and the accuracy of the system was verified. 2.2. Results Obtained at Sample Set 1 – Copper Layer The calibration functions for sample set 1 (samples with copper layer) were obtained on samples Cu-1, Cu-5 and Cu-9. The film thickness measurement of the remaining samples using a measuring gap of 20 mm resulted in an average thickness accuracy of 1 nm. On samples with very thin copper layers of a maximum of 25nm, this high accuracy remains when the measuring gap is increased up to 60 mm. In contrast, the measurement accuracy decreases on samples with higher film thickness (starting at 50 nm) when increasing the measurement gap of up to 4 nm around the set point of the film thickness. Results from sample set 1 are shown in Figure 4. The diagram on the left shows the analyzed thickness values from all 8 repetitions obtained by the Eddy Current measurements as a function of the set point of the film thickness for a sensorto-sensor distance of 20 mm. The diagram on the right demonstrates the deviation from the set point of the film thickness for 3 sensor-to-sensor distances of 20 mm, 40 mm, and 60 mm. deviation from set point

copper layer on substrate of silicon

(copper layer)

sensor-sensor distance 20 mm deviation from set point [nm]

thickness (EC) [nm]

105 85 65 45 25 5

4,5 4,0 3,5 3,0 2,5 2,0 1,5 1,0 0,5 0,0

2 cm 4 cm 6 cm

10 5

25

45

65

85

105

15

20

50

55

100

thickness [nm]

thickness (set point) [nm]

Figure 4. Results of thickness measurements on sample set 1 (copper layer). Left diagram: Coating thickness as a function of the set point of the film thickness established by Eddy Current. Right diagram: Deviation of the film thickness set points for different sensor-to-sensor distances.

In a further experiment, the effects of position variations of the sample within the measuring gap on the established film thickness were investigated. Some of the copper samples were examined in a transmission mode system with a sensor-to-sensor distance of 40 mm. All samples were positioned in the center of the measuring gap, moved 5 mm up and tilted by 10°, whereas each measurement was repeated five times. The measurement positions are shown schematically in Figure 5.a); Figure 5.b) illustrates corresponding results and determined respective film thicknesses as a function of the film thickness set point. The measurement accuracy does not change when the sample is moved or tilted. There is a constant deviation from the desired set point of ± 3 nm for all three sample positions.


335

copper layer on substrate of silicon at different sample positions

Send moved of 5 mm centre tilted by 10°

Receive

a)

thickness (EC) [nm]

110 90 70 50 30 10 10

30

b)

50

70

90

110


Figure 5.a) Schematic illustration of the sample positions tested in the measuring gap; 5.b) results of thickness measurement at different sample positions

2.3. Results Obtained at Sample Set 2 – TaSiN-Layer For sample set 2 (TaSiN layer), measurements similar to sample set 1 were taken. At these samples the optimal measurement frequency was higher, due to a significantly lower conductivity than the conductivity of the copper layered samples. The measurement accuracy was not as good as the accuracy obtained from the copper layered samples, which was expected due to lower conductivity. Figure 6 illustrates the calibration points obtained from sample set 2. It shows the phase of the complex Eddy Current signal as a function of the film thickness set points for a sensor-to-sensor distance of 20mm. The measurements for each sample were repeated five times.

Eddy Current phase [rad]

TaSiN-layer on substrate of silicon -1,1605 -1,1610 -1,1615 -1,1620 -1,1625 -1,1630 -1,1635 -1,1640 -1,1645 -1,1650 -1,1655 0

50

100

150

200

250


Figure 6. Calibration points for Eddy Current film thickness measurements of sample set 2, demonstration of the phase of the complex Eddy Current signal as a function of the film thickness set point.

It is clearly visible that measurement values with film thickness variation of 5 nm cannot be separated. However, the measurement values have a high reproducibility and the values around 10 nm, 50 nm, and 200 nm are very well separated. Using a calibration function, the coating thickness can be established in the same way as for sample set 1. Using a sensor-to-sensor distance of 20mm could achieve a measurement accuracy of 20 nm. With increasing sensor-to-sensor distance, the measurement accuracy decreases cumulatively. Here we can see the limit of the measurement system. Coatings with very low conductivity can only be measured adequately if a very small measuring gap is used or higher film thicknesses are investigated.

336


3. Multi- Layer Systems in the Nanometer Range For the investigation of multi-layer coatings, we initially studied a system with two conductive layers. These samples consisted of a substrate of glass to which successive layers of zinc oxide (non-conductive), silver and aluminum were applied. The thicknesses of the conductive layers of silver and aluminum vary between 100 nm and 600 nm. Table 2 lists the samples and the appropriate coating thicknesses. Table 2. List of the samples of sample set 3 (multi layer) and their coating thicknesses Nr. multi-1 multi-2 multi-3 multi-4 multi-5 multi-6 multi-7

Thickness Silver [nm] 103 205 400 590 200 205 205

Thickness Aluminum [nm] 105 105 95 95 200 385 590

Total Thickness [nm] 208 310 495 685 400 590 795

When performing measurements in the transmission mode, the field always penetrates all layers as well as when using a surface probe; the entire layer will always be penetrated due to the very thin layers. This fact provided two measurement approaches: thickness measurements of the total thickness of both conducting layers, and second, thickness measurements of single layers. The results of both measurements are demonstrated in Figure 7. multi layer on substrate of glass

multi layer on substrate of glass

single layer

total thickness

500

thickness (EC) [nm]

thickness (EC) [nm]

800 700 600 500 400 300 200 200

300

400

500

600

700


800

450 400 350 300 250 200 silver aluminum

150 100 100

200

300

400

500


Figure 7. Coating thicknesses obtained by Eddy Current techniques as a function of the film thickness set point, measured on sample set 3 (multiple layers). Left: Measurements of total thickness (all layers). Right: Measurements of single layers

To determine the total thickness of the aluminum and silver layers, the calibration function was obtained using the measured values of the sample multi-1 (smallest total thickness) and multi-7 (maximum overall thickness). Using this almost linear function, the thickness values of the remaining samples were determined. A sensor-to-sensor distance of 40 mm was applied. A maximum deviation from the film thickness set point of 55 nm was achieved in this investigation. Since the amounts of aluminum and silver contents in the layers always vary, the total conductivity of the layer varies too, which explains the relatively high measurement error.


337

In each case, measurements of single layer samples were performed with a constant sample set; one thickness of both layers. The analysis of the silver layer was attained with the multi-1 to multi-4 samples, whereas the calibration function was acquired with samples multi-1 and multi-4. Analysis of the aluminum layer was performed with samples multi-2 and multi-5 to multi-7, while the calibration function was attained with multi-2 and multi-7 samples. On the silver layer, a maximum deviation of the film thickness set point of ± 17 nm was be detected, and a maximum deviation of ± 38 nm was observed on the aluminum layer samples. The concept outlined in the section 3, shows again that the measurement accuracy is increasing with increasing layer conductivity. Under the assumption that only the thickness layer of one multi-layer film system varies while the other remains constant, the varying thickness of this layer can be accurately determined.

4. Summary and Outlook It was demonstrated that a large number of conductive layers of thin film technology can be characterized by Eddy Current testing. Table 3 summarizes the accuracy of all presented results. Table 3. Accuracy of the tested samples at different sensor-to-sensor distances, obtained in the Eddy Current transmission mode (EddyCus TF®) Single Layer Coating material copper TaSiN

Coating thickness range 5nm – 100nm 10nm – 200nm

Sensor-to-Sensor distance 20mm ± 1nm ± 20nm

Sensor-to-Sensor distance 40mm ± 1.5nm ± 50nm

Sensor-to-Sensor distance 60mm ± 2nm ± 100nm

Multi Layer Coating material silver aluminum total

Coating thickness range 100nm – 600nm 100nm – 600nm 100nm – 600nm

Sensor-to-Sensor distance 20mm ± 10nm ± 21nm ± 54nm



It was also revealed that the transmission mode is quite insensitive to sample position variations and thus presents a solid in-line testing method for thin film technology. The EddyCus TF® measurement system allows thickness measurement by using a sensor-to-sensor distance of up to 60 mm for highly accurate thickness measurement of conductive layers in the nanometer range. Measurement accuracy of 1 nm in the thickness range of 5 nm to 100 nm can be achieved on in highly conductive layers, such as copper. In addition, multi-layer systems can be accurately analyzed with this system. Under the assumption that only the thickness of the layers varies, the varying layer thickness can be determined very precisely. By selecting the best possible measurement frequency for each conductivity range, the accuracy of the measurement system can be optimized and changes in the conductivity of the samples caused by temperature variations, for example, can be easily separated from thickness variations. This makes the EddyCus TF® system a precise, fast, and automatable measurement

338


system with a great potential for cost-effective thickness measurement of thin conductive layers.

References [1] [2] [3] [4]

K. Nitzsche, “Schichtmeßtechnik”, Vogel-Buchverlag, technical book, 1. Auflage, 1996, ISBN 3-80231530-8A.N. Author, Article title, Journal Title 66 (1993), 856–890. H. Heuer, S. Hillmann, M. Röllig, M. H. Schulze, K-J. Wolter, „Thin Film Characterization Using High Frequency Eddy Current Spectroscopy”, Conference on Nanotechnology, Genua, 2009 E. Hanke, “Prüfung metallischer Werkstoffe”, Band II, Zerstörungsfreie Prüfverfahren, VEB Deutscher Verlag für Grundstoffindustrie, Leipzig 1960, 2. überarbeitete Auflage A. Yashan, „Numerische Modellierung von Wirbelstromaufgaben und Lösung des inversen Problems“, ZfP-Zeitung 67, Oktober 1999


339

Feasibility of stress state assessment on the grounds of measurements of the strength of the residual magnetic field of ferromagnetics a

Maciej ROSKOSZa,1 Silesian University of Technology, Poland

Abstract. The relations between stress and the values of the components of the residual magnetic field of ferromagnetic steels exposed to static and low cycle loads were examined. For static loads the measurements of the strength of the residual magnetic field were shown under load and after unloading. For low cycle loads, on the other hand, the impact of the number and the amplitude of changes in load were presented. It was concluded on the grounds of reference literature and the obtained results that the irreversible changes in magnetization, which result from loads, could be used as diagnostic signals. The tangential component which is parallel to the direction of tensile loads is characterized by a good correlation with stress. The fact that the value of the RMF increases significantly after yield stress is exceeded makes it possible to work out a method to determine areas with plastic strain. The conclusions presented seem to be a good starting point for further and deeper studies and analyses whose aim is to employ the strength of the residual magnetic field to assess the stress state of machine and construction components. Keywords. residual magnetic field, stress, magneto mechanical effect, control magnetic methods

Introduction There is a strict relationship between the magnetic and mechanical properties of ferromagnetic materials. The influence of stress on the magnetic properties is described by the magneto-mechanical effect. If a ferromagnetic is placed in a magnetic field, its level of magnetization varies according to loads. The change in magnetization has a reversible component, which fades after unloading, and an irreversible one. The level of magnetization depends on the type of material, the strength of the magnetic field, the magnetic history, strain and temperature [1 – 3]. The strength of the (RMF) is known to have been used as a diagnostic signal in flaw detection – the Metal Magnetic Memory Method [1 – 4]. The magnetic properties of steel and their use in diagnostics are the subject of many publications [4 – 14] The presented results are a part of studies whose main objective is to develop a system of a non-destructive assessment system of the technical condition and wear of 1 Institute of Power Engineering and Turbomachinery, Faculty Of Energy And Environmental Engineering, ul. Konarskiego 18, 44 100 Gliwice, Poland

340

M. Roskosz / Feasibility of Stress State Assessment on the Grounds of Measurements

power machinery components subjected to typical loads such as low-cycle fatigue and creep. One of the tasks is an analysis of the possibility to use the changes in the magnetic properties of ferromagnetics as diagnostic signals for that purpose. The study focused on methods based on changes in the magnetization of components, and changes in the Barkhausen effect. It was assumed that the complementarity of the methods would allow a determination of stress distribution and wear that would be precise enough for industrial practice.

1. Objective and experimental details This publication presents the results of preliminary studies of changes in magnetization to determine: • The relations between stress and the values of the RMF components of ferromagnetics for static and quasi-static loads. • The impact of changes in the external magnetic field on the RMF of samples. • The impact of the stress amplitude and the number of cycles on the values of the RMF components of ferromagnetics. The research was conducted for ferromagnetic steel. The samples were not preliminarily demagnetized. The initial state of magnetization was a consequence of the history of their preparation and was not homogenous. The magnetometer TSC-1M-4 with the measuring sensor TSC-2M (Foerster type sensor) supplied by Energodiagnostika Co. Ltd Moscow was used for the measurements. As a result of the measurements, 3 values of the components of the RMF strength on the surface of the examined ferromagnetic samples were obtained: • Ht,x – tangential component measured in the direction perpendicular to the load applied, • Ht,y – tangential component measured in the direction parallel to the load applied, • Hn,z – normal component.

2. Static loads The geometrical form of the examined samples is shown in Fig. 1. The chemical constitution and the strength properties of the materials that the samples were made of are listed in Table 1.

Figure 1. Experimental sample.


341

4 measuring sections of the same size were selected on a part of the sample. They were denoted as , , , . The division into measuring sections was introduced to check the repeatability of the relation: stress – the values of the RMF components. The points with symbols corresponding to the denotations of the measuring sections are average values of the RMF components for these sections. Table 1. Chemical constitution and strength properties of the sample material (R e yield stress, Rm stress)

ultimate

C

Si

Mn

Cr

Cu

Al

V

W

P

S

Re [MPa]

Rm [MPa]

max 0,22

0,10,35

max 1,1

0,3

-

-

0,3

-

0,05

max 0,05

280

380-520

Exemplary measurements of the RMF strength under load – grey lines, and after unloading – black lines, were shown in Fig. 2 to 4 as a function of the applied load. For each of the measuring sections the dependence of the average values (from 3 samples) on stress was approximated with a polynomial. 2.1. Measurements under load The trend of the changes in the values of the RMF components under loading is not constant, not even within elastic strain. The stress σ values at which it changes are characteristic of a given material. For components Ht,x and Hn,z the trends of the changes depend on the measuring section – the measuring point. This proves that there is no correlation between stress σ and the value of the component. However, in the analyzed cases the trend of the changes in component Ht,y did not depend on the measuring point, and was identical in all measuring sections. The greatest diversification of the values of components Ht,x and Hn,z in individual measuring sections occurs with stress σ approximating yield stress Re. 2.2. Measurements after unloading The representation of the impact of the load on the values of the RMF components measured after unloading is dominated by a decisive change in their values around yield stress Re. As for elastic strain, it can be assumed that there is a steady trend of changes in the values of the RMF components. After yield stress Re is exceeded, the trend of changes in both component Ht,x and component Hn,z is not identical and depends on the measuring point – the measuring section. 2.3. Impact of changes in the external magnetic field Fig. 5 shows the results of measurements of the RMF conducted after unloading on the surface of the sample inside a strength testing machine and outside it. A substantial quantitative change was observed, but the qualitative relations: stress – magnetization remain unchanged. It is due to the fact that the solution to the problem of allowing for the impact of the external magnetic field value is decisive when considering the sense of using RMF measurements for stress assessment.

342


160 Ht,x [A/m]

.

120

80

40

0

-40 0

100

200

300

Figure 2. Relation between stress σ and component Ht,x, measurement under load after unloading black lines.

400

σ [MPa]

500

grey lines, measurement

500 Ht,y [A/m]

.

400

300

200

100

0 0

100

200

300

Figure 3. Relation between stress σ and component Ht,y, measurement under load after unloading black lines.

400

σ [MPa]

500


343

M. Roskosz / Feasibility of Stress State Assessment on the Grounds of Measurements 200 Hn,z [A/m]

0

-200

-400

. -600 0

100

200

300

400

Figure 4. Relation between stress σ and component Hn,z, measurement under load after unloading black lines.

σ [MPa]

500


600 Ht,y [A/m]

.

400

200

0

-200 0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

σ/Re

1.8

Figure 5. Impact of changes in the external magnetic field on the values of component Ht,y, measurement inside a strength testing machine black lines, measurement outside a strength testing machine grey lines.

344


3. Cyclic loads Fatigue tests were conducted in the area of elastic strain. The measurement results presented in Fig. 6 to 8 refer to one kind of ferromagnetic steel which was subjected to loads where the value of the cycle asymmetry coefficient R was equal to 0, with different maximum stress of the cycle. Analyzing the obtained results it was found out that: • the greatest change in the values of the components occurred after the first load cycle • each subsequent load cycle produces smaller and smaller changes in the values of the components • the impact of the maximum stress of a cycle is visible. 20 Ht,x [A/m] 0

100 Ht,y [A/m] 0

-20 -40

σ1/Re= 0.4

-50

σ2/Re= 0.8

-100

-60

INITIAL STATE

-150 1

10

100

N

1000

Figure 6. Impact of changes in the number of cycles and maximum stress on the values of component Ht,x, of the RMF.

INITIAL STATE

1

10

100

N

1000

Figure 7. Impact of changes in the number of cycles and maximum stress on the values of component Ht,y of the RMF.

150

H

Hn,z [A/m] 50 0 -50 -100 -150

INITIAL STATE

1

10

100

N

1000

Figure 8. Impact of changes in the number of cycles and maximum stress on the values of component Hn,z of the RMF.

N Figure 9. Schematic correlation between the permanent change in magnetization and the number of load change cycles.

On the grounds of this study, together with [7, 10, 14], a schematic correlation between the changes in the RMF and the number of load change cycles was determined (cf. Fig. 9). In the beginning phase of the load process, and also just before the sample is destroyed, significant changes in the value of the RMF can be observed. In the


345

plateau phase, depending on the steel microstructure, there is a slight increase or decrease in magnetization.

4. Summary The paper presents preliminary results of a study of changes in the RMF strength caused by applied loads. On the grounds of the results it can be concluded that: • The tangential component which is parallel to the direction of tensile loads is characterized by a good correlation with stress. • The fact that the value of the RMF increases significantly after yield stress is exceeded makes it possible to work out a method to determine areas with plastic strain. • The irreversible changes in magnetization, which result from stress, can be used as diagnostic signals which make it possible to determine the stress state of ferromagnetics. Because of the special character of the stress – component Ht,y relation, the determination of the stress value is not possible with component Ht,y only. Analyzing the dependences between stress and the values of the RMF components it can be stated that it is possible to define the local stress state on the grounds of the values of component Ht,y and gradient Hn,z. [10] presents a good correlation between the distribution of component Ht,x with the distribution of shear stress, which can also be used to assess the state of stress. • There is a correlation between the level and number of load cycles, and the state of magnetization. A dramatic change in the value of the RMF occurs in the last phase of fatigue wear [7, 10, 14], which makes it possible to anticipate the destruction of the component. • The external magnetic field has a significant impact on the values of the RMF. The solution of the problem of how to account for this significance is decisive when considering the sense of using RMF measurements for stress assessment. The conclusions presented seem to be a good starting point for further and deeper studies and analyses whose aim is to employ the strength of the RMF to assess the stress state of machine and construction components.

Acknowledgements This work was supported by the Polish National Centre for Research and Development within the strategic program „Advanced Technologies for Energy Generation”

References [1] [2]

Deputat J.: Basics of the Metal Magnetic Memory Method. Dozór Techniczny 5/2002 p. 97 105. (in Polish) Jiles D. C., Theory of the Magneto Mechanical Effect, J. Phys. D: Appl. Phys., 28 (1995) 1537 1546.

346

[3]

[4]

[5] [6] [7]

[8]

[9] [10] [11]

[12]

[13]

[14]


Dubow A.A.: Principal Features of Metal Magnetic Memory Method and Inspection Tools as Compared to Known Magnetic NDT Methods. WCNDT 2004, Montreal Canada, http://www.ndt.net/article/wcndt2004/papers/359.ntm John Wilson, Gui Yun Tian, Simon Barrans: Residual Magnetic Field Sensing for Stress Measurement and Defect Detection, Sensors and Actuators A: Physical, Volume 135, Issue 2, 15 April 2007, Pages 381 387 urek Z. H. Magnetic Contactless Detection of Stress Distribution and Assembly Defects in Constructional Steel Element NDT&E International 38 (2005) 589 595 Augustyniak B., Degauque J. Magneto Mechanical Properties Evolution of Fe C Alloy during Precipitation Process, Materials Science and Engineering A 370 (2004) 376 380 Dobmann G., Lang M. On line Monitoring of Fatigue in the LCF and HCF Range by Using Micro magnetic NDT at Plain Carbon and Austenitic Stainless Steel, 8th ECNDT Proceedings TOC European Conference on Nondestructive Testing Barcelona (Spain), June 17 21, 2002 R. Sabet Sharghi, L. Clapham, D.L. Atherton, T.M. Holden The Effect of Defect Introduction vs. Load Application Sequencing on Defect Induced Stress Distributions in Steel Samples NDT&E International 33 (2000) 201 212 M. Soultan, X. Kleber_, J. Chicois, A. Vincent Mechanical Barkhausen Noise during Fatigue of Iron NDT&E International 39 (2006) 493 498 Roskosz M. Gawrilenko P, Analysis of Changes in Residual Magnetic Field in Loaded Notched Samples, NDT&E International 41 (2008) 570 576 Roskosz M: Possibilities of the Application of the Metal Magnetic Memory Method to the Analysis of Gear Durability. 9th European Conference on Non Destructive Testing ECNDT Berlin 2006, Abstracts Part 2, P 85 Roskosz M.: C , "¡ ¢ £ ¢ ", ¤ 2009 pp. 53 61 Dong Lihong, Xu Binshi, Dong Shiyun, Chen Qunzhi, Wang Dan Variation of Stress Induced Magnetic Signals during Tensile Testing of Ferromagnetic Steels NDT&E International 41 (2008) 184 189 Yang En, Li luming, Chen Xing: Magnetic Field Aberration Induced by Cycle Stress, Journal of Magnetism and Magnetic Materials 312 (2007) 72 77


347

Magnetic Characterization of Material Degradation Using Dynamical Minor Loops Satoru KOBAYASHI a,1 , Shinichi TSUKIDATE a , Hiroyuki OKAZAKI a , Yasuhiro KAMADA a Hiroaki KIKUCHI a , and Toshihiro OHTANI b a NDE and Science Research Center, Faculty of Engineering, Iwate University, 4-3-5 Ueda, Morioka 020-8551, Japan b Department of Mechanical Systems Engineering, Shonan Institute of Technology, Fujisawa 251-8511, Japan Abstract. We have examined scaling laws in dynamical magnetic minor hysteresis loops of Cr-Mo-V steel subjected to creep and cold-rolled low carbon steel in order to investigate the applicability to online evaluation of materials degradation in ferromagnetic steels. Although scaling laws in the medium magnetization range found previously fail in the high magnetization frequency regime owing to a significant contribution of eddy currents, a scaling power law of the relation between remanence and remanence work of minor loops holds true in a very low magnetization regime, irrespective of magnetization frequency as well as investigated materials. The coefficient of the law exhibits a good correlation with creep damage and rolling reduction and is proportionally related to Vickers hardness. These observations clearly demonstrate that the analysis of dynamical minor loops enables us to evaluate materials degradation in a short measurement time with low measurement field and high sensitivity to defect density. Keywords. Dynamical hysteresis loop, Creep damage, plastic deformation, Cr-MoV steel, low carbon steel, Nondestructive evaluation

Introduction Non-destructive evaluation (NDE) of aging degradation of structural materials has been carried out for a long time. Various techniques using eddy current, ultrasonic wave, Xray, magnetic powder etc., are now mature and widely used. Generally, these techniques are effective for detecting cracks with dimension greater than 0.1 mm. Nevertheless, none of NDE techniques have been employed so far as a standard method to evaluate degradation before initiation of cracking. Magnetic method using minor hysteresis loops has several advantages to get information on lattice defects in ferromagnetic steels. Our previous studies in plastically deformed low carbon steels [1] revealed that there exist several scaling power laws between 1 Corresponding Author: Satoru Kobayashi, NDE and Science Research Center, Iwate University, 4-3-5 Ueda, Morioka 020-8551, Japan, E-mail: [email protected] .

348

S. Kobayashi et al. / Magnetic Characterization of Material Degradation

Table 1. Chemical compositions of Cr-Mo-V steel (ASTM A193-B16). (wt. %) C

Si

Mn

P

S

Cr

Mo

V

Fe

0.420

0.290

0.66

0.016

0.009

1.090

0.51

0.28

Bal

Table 2. Chemical compositions of cold-rolled low carbon steel. (wt. %) C

Si

Mn

Fe

0.16

0.20

0.44

Bal

parameters of quasi-static minor loops in a medium range of magnetization over which irreversible movement of Bloch wall dominates magnetization process. The power-law relations include

∗ nF Ma ∗ 0 WF = WF , (1) Ms WR∗

=

WR0

MR∗ MR

nR ,

(2)

where WF∗ , WR∗ , Ma∗ , and MR∗ are hysteresis loss, remanence work, maximum magnetization, and remanence of minor loops, respectively, and the definition is given in the inset of Fig. 2(c). M s and MR are saturation magnetization and remanence of the major loop, respectively. The exponents n F and nR are approximately 1.5 and universally independent of temperature, stress, and kinds of ferromagnetic materials [1,2]. The coefficients WF0 and WR0 are a sensitive indicator of internal stress and can be obtained with low magnetic fields, typically less than 2 kA/m, which is far less than saturation field of > 10 kA/m. When one applies this method to online NDE of ferromagnetic components, however, it is necessary to reduce the measurement time; the measurement time to obtain coefficients for one condition is considerably long (15-30 min) due to low magnetization frequency below 0.1 Hz. Measurement with a higher magnetization frequency is one of the solutions to reduce the measurement time. It would be therefore useful to examine the scaling rules in the high frequency regime to investigate the applicability for magnetic NDE. In this study, we have investigated scaling rules of dynamical minor hysteresis loops for Cr-Mo-V steels subjected to creep as well as cold-rolled low carbon steels, varying the level of creep damage and plastic deformation, respectively.

1. Experimental Procedure The chemical compositions of crept Cr-Mo-V ferritic steel and low carbon steel are listed in Table 1 and 2, respectively. The Cr-Mo-V steel, which is widely used for high-temperature power plant components such as steam pipes, pressure vessels, was taken from a commercial plate of ASTM A193-B16. The creep tests were performed at 923 K in air under applied tensile stress of 35 MPa. Twelve specimens with different life fractions t/t r were prepared; t and t r are creep time and rupture life, respectively. Figure 1 shows creep strain and Vickers

S. Kobayashi et al. / Magnetic Characterization of Material Degradation

349

hardness as a function of t/t r . The creep strain increases with t/t r , whereas the Vickers hardness decreases. Both properties exhibit three creep deformation regimes; primary (0≤ t/tr 9@

5RWRU $QJOH > GHJ@

Figure 10. Calculated brush voltage waveform.

Also, these results are in accord with the sparks generating phenomenon of DC motor. The reason compared with the sparking voltage of DC motor, there are few reports of the sparking voltage of an alternating current drive universal motors. Therefore, it judged that it was better to compare with the sparking voltage waveform of DC motor. As compared with a measurement result, and our new proposed FEM simulated basic voltage waveform are very good congruous. However, the spike voltage of the latter is not much congruous. This reason have addition of the contact state of a brush and commutator’s, the influence of a factors, the temperature rise of a contact surface, and vibration of a mechanical system other than the influence of an analysis step angle. Therefore, the

446

Y. Akiyama and Y. Niwa / Propositions for the Analysis of Commutation Phenomena

reproducibility of this analysis is not so good. Naturally, the waveform of an actual measurement is also always changed and is not constant. Also, there is the necessity of inquiring further from now on. However, we introduce the result of the our present level of Fig. 10.

4. Conclusions A complicated commutation phenomenon peculiar to a universal motor is treating many factors as a state variable, and high-precision analysis has been realized. The importance of analyzed these phenomena quantitatively is large. Therefore, we think that future EMC correspondence and a commutation sparks problem can be coped with. We think that these techniques will be helpful in the elucidation of the commutation sparks phenomenon.

References [1] R. Richter : ಯElektrische Maschinen ൙ರ, (1950) [2] Donald G. Schultz, James L. Melsa : “STATE FUNCTIONS AND LINEAR CONTROL SYSTEMS”, McGrwa Hill Book Company (1967) [3] Richard P. Feynman : ಯThe Feynman lectures on physicsൗರ, (1986) [4] (France) CEDRAT SA : “FLUX2D®, Users Guide Vol.3” [5] Y.Akiyama : “The comparison of the characteristics in INV drive BLDC motor and Brushed DC motor”, The paper of Technical Meeting on Rotating Machinery, IEE Japan, No.RM01 162, pp.19 23 (2001) (in Japanese) [6] Y.Akiyama, T.Takura, Y.Niwa* : “A study about Core shape Design of 200V Lines for Vacuum Cleaner Universal Motor”, The paper of Technical Meeting on Rotating Machinery, IEE Japan, No.RM03 137, pp.7 12 (2003) (in Japanese) [7] H.Koharagi, K.Tahara, Y.Ishii, S.Suzuki : “Application and Evaluation of Commutation Performance with an Exceptional Armature Winding Form for Small Universal Motors”, T.IEE Japan, Vol.115 D, No.4, pp.488 494 (1995) (in Japanese) [8] S.Suzuki, K.Kurihara, H.Nase, K.Takahashi : “RF Noise Associated with Time Varying Arc Current across Brush and Commutator in Universal Motors”, T.IEE Japan, Vol.118 D, No.6, pp.773 779 (1998) (in Japanese) [9] K.Kurihara, S.Sakamoto, : “Starting Preformance Analysis for Universal Motors by FEM”, IEEJ Trans. IA, Vol.126, No.2, pp.124 130 (2006) (in Japanese) [10] K.Tahara, T.Takura, : “Trend of High Performance Universal Motors”, The paper of Technical Meeting on Rotating Machinery, IEE Japan, No.RM98 32, pp.77 82 (1998) (in Japanese) [11] S.Wagatsuma, T.Ueno, N.Morita : “Considerations for commutation theory by residual current, using commutation testing machine ”, The paper of Technical Meeting on Rotating Machinery, IEE Japan, No.RM02 23, pp.7 12 (2003) (in Japanese) [12] Y.Niwa, T.Naruta, Y.Akiyama : “Consideration about the core shapes, generating Torque, and Commutation phenomena of an ultra high speed Universal motor”, ICEMS2006, DS3F2 01, (2006) (Nagasaki, Japan) [13] Y.Niwa, Y.Akiyama, K.Miyazawa : “Nonlinear FEM simulation of the Commutation Sparking phenomenon of an Ultra High Speed Universal Motors”, EPNC2008, pp.95 96 (2008) (Lille, France) [14] Toshio Fuji, “Study of the universal motor” (1977) (in Japanese) [15] IEC 61000-3-2 : ಯElectromagnetic compatibility (EMC) Part 3-2, Limits Limits for harmonic current emissions (equipment input current ู 16A per phase)ರThird edition (2005)amendment1 (2008) [16] JIS C 61000 3 2 : ಯElectromagnetic compatibility (EMC) Part 3 2, Limits emissions (equipment input current ู 20A per phase)ರ (2003) [17] JIS C 9108 : “Electric vacuum cleaners” (1992)

Limits for harmonic current


447

Detection of CFRP’s degradation through electromagnetic procedures Bogdan SERGHIAC1, Daniel Petrica SALAVASTRU2, Paul Doru BARSANESCU2, Adriana SAVIN3, Rozina Steigmann3 and Raimond GRIMBERGa3 1 University Al.I.Cuza, Iasi, Romania 2 Technical University Gh.Asachi, Iasi, Romania 3 National Institute of Research and Development for Technical Physics, Iasi, Romania

Abstract. Carbon fiber reinforced plastics, due their average electrical conductivity of the reinforcement, can be nondestructively evaluated by electromagnetic procedures. In the case of reinforcement from carbon fibers woven, to improve the spatial resolution of the image of the conductivity, we propose the using of a circularly aperture, very small, placed at short distance from the surface to be controlled that can diffracts the evanescent waves. By this procedure, the image of the reinforcement woven is observed, the delamination due impact with low energy being clearly distinguished. Keywords. Carbon fiber reinforced plastics, electromagnetic nondestructive evaluation, image reconstruction

Introduction Carbon fiber reinforced plastics (CFRP) have now applications from most different starting with aerospace industry and finishing with sports goods. CFRP have evolved both in reinforcements and matrix. From the point of view of the reinforcement, the tendency is to pass from pre-preg laminas at which the carbon fiber are parallel arranged to laminas that contain woven of different types of carbon fibers, leading to the improvement of formability of these composites. The carbon fibers have an average electrical conductivity around 103y104S/m and relative magnetic permeability is 1 [1]. The epoxy resin was the most usually the matrix for CFRP. Due to the facts that its mechanical properties are influenced by the moisture content and at temperature exceeding thermal destruction temperature, the resin relieves dangerous vaporous, the tendency is to be replaced it by Polyphenylene sulphide (PPS). This plastic material is less influenced by the moisture content and the quantity of noxious vaporous at high temperature is low [2]. The matrix has low electrical conductivity

Electromagnetic Nondestructive Evaluation (XIV) - Volume 35 Studies in Applied Electromagnetics and Mechanics

Electromagnetic Nondestructive Evaluation (X) (Studies in Applied Electromagnetics and Mechanics)

Electromagnetic Nondestructive Evaluation (XI) (Studies in Applied Electromagnetics and Mechanics) (No. 11)