NDT&E International 43 (2010) 409 415 Contents lists available at ScienceDirect NDT&E International journal homepage: www.elsevier.com/locate/ndteint Defect edge identification with rectangular pulsed eddy current sensor based on transient response signals Yunze He n, Feilu Luo, Mengchun Pan, Xiangchao Hu, Bo Liu, Junzhe Gao College of Mechatronics and Automation, National University of Defense Technology, Changsha 410073, China article info Article history: Received 5 November 2009 Received in revised form 4 March 2010 Accepted 24 March 2010 Available online 30 March 2010 Keywords: Pulsed eddy current (PEC) Rectangular coil Edge identification Feature extraction Transient response signal abstract The Pulsed Eddy Current (PEC) testing is an increasingly emerging nondestructive testing & evaluation (NDT&E) technique. The main purpose of this study is to improve the performance of defect edge identification of C-scan imaging technique utilizing the rectangular PEC sensor. When sensor scans along the defect, peak waves of response signals always present a crest and a trough in direction of magnetic induction flux, while present different shapes in direction of exciting current. The maximum and minimum values of peak waves in direction of magnetic induction flux are corresponding to the moment of sensor entering and leaving the length edge of defect, which provides us a way to evaluate the length edge of defect. To evaluate the width edge of defect, we obtain and analyze the C-scan imaging results in direction of magnetic induction flux. For improving the identification of width edge of defect, we proposed news features from response signals and differential response signals. Experiment results have shown that the width edge of defects on surface can be identified effectively by selecting and normalizing the appropriate features in time domain. Therefore, both length edge and width edge of defect can be evaluated effectively in direction of magnetic induction flux. The rectangular PEC sensor is helpful for C-scan imaging inspection technique and has a good prospect in field of nondestructive testing & evaluation. Crown Copyright & 2010 Published by Elsevier Ltd. All rights reserved. 1. Introduction The pulsed eddy current (PEC) nondestructive testing is an effective technology that has been demonstrated to be capable of quantifying defect in the aging aircraft structure [1 3]. PEC testing possesses many advantages against the conventional single frequency eddy current testing, including more extended detection depth, richer information about defects and higher robustness of anti-interference [4]. Meanwhile, pulses can be easily generated and controlled and the other advantage of PEC method over multi-frequency eddy current method is less expensive instrumentation. Therefore, the PEC testing has been particularly developed and devised for sub-surface crack measurements, crack reconstruction, and depth estimation [5]. In 1994, Rose et al. [6] described the defect detection in time domain for both air-core and ferrite-core PEC probes with one flat coil. In 1996, Moulder et al. [1] developed a new scanned pulsed eddy current instrument for nondestructive inspection of aging aircraft and Tai et al. [7] determined the thickness and conductivity of conductive coatings on metal plates. In 2001, Giguere et al. [8] n Corresponding author. Tel.: +86 13467698133. E-mail address: hejicker@gmail.com (Y. He). illustrated three features adopted in PEC testing to quantify defects and exemplify their application. In 2002, Yang and Tai [9] determined the thickness or conductivity of metallic coatings on a metal substrate for the case when either the coating or substrate is magnetic. In 2003, Sophian et al. [10] introduced the application of principal component analysis in extracting information from PEC response. In 2004, Tian et al. [11] investigated the dynamic behavior of pulsed eddy current techniques and the feasibility of their use in an on-line inspection system. In 2005, Tian and Sophian [12] also presented a new feature called as rising point time to identify the different defect types and lift-off and Tian et al. [13] presented a new approach for defect classification and quantification by using pulsed eddy current sensors and integration of principal component analysis and wavelet transform for feature based signal interpretation. In 2007, Li et al. [4] introduced development of differential probes in pulsed eddy current testing. In 2008, Chen et al. [14] extracted and selected appropriate features for defect classification of pulse eddy current. In 2009, Fan et al. [15] made a research about analytical modeling for transient probe response in pulsed eddy current testing. In 2009, the defects in riveted structures are detected effectively utilizing the differential hall/coil probe in our laboratory [16]. In all these studies, the cylindrical driving coil is excited by repeated pulses and the response signal is measured with a 0963-8695/$ - see front matter Crown Copyright & 2010 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.ndteint.2010.03.007
410 Y. He et al. / NDT&E International 43 (2010) 409 415 sensor, which may be the driving coil, single pick-up coil, differential coil or a hall-effect sensor [17]. The Alternating Current Field Measurement (ACFM) technique is an electromagnetic technique capable of detecting and sizing defects in metallic components. It was developed within the NDE Centre, department of mechanical engineering at university college London [18] and is an extension of the Alternating Current Potential Drop (ACPD) crack detection and measurement technique. It is based on the thin-skin theory developed by Lewis, Michael, Lugg and Collins (LMLC theory) [19] and is widely applied for the detection of near-surface defects. The ACFM probe usually incorporate an induction coil for the generation of surface current and the pick-up coils for the measurement Bx and Bz in the vicinity of a defect [20]. Associated with the current flowing in the surface, there is a magnetic field above the surface. The magnetic field will be disturbed in the presence of a defect and the pick-up coils will induce the disturbed magnetic field [21,22]. In our laboratory, rectangular exciting coil which was widely used in ACFM method was proposed and researched in PEC testing. The experimental results have indicated that it is also capable of detecting and evaluating defects. In 2006, Yang et al. [23,24] proposed peak value and over-zero time of response signal, respectively, to measure the length and depth of defect in aircraft multi-layered structure. In 2007, they also did a research on edge identification of a defect using pulsed eddy current based on principal component analysis [25]. In our previous studies, defects on surface can be identified and evaluated effectively utilizing the 2D butterfly-shape graph in both directions of sensor scanning in 2009 [26]. The main purpose of this paper is to detect the edge of defects on surface and to enhance the accuracy of edge identification based on rectangular exciting coil. The rest of the paper is arranged as follows. Firstly, experimental set-up including hardware, software, and specimen is established in Section 2. Next, different directions of rectangular sensor scanning are introduced and peak waves in different directions are analyzed in Section 3. Then, the results of C-scan imaging are displayed and features from response signals are extracted to detect the edge of defect. Experiment results of edge identification are shown in Section 4. Finally, conclusions and further work are outlined in Section 5. 2. Experimental set-up and specimen The PEC experimental set-up used in this research consists of pulse generator, power amplifier, signal conditioning module, data acquisition, X Y displacement system, and PC-based software. The generator module is used to generate the exciting pulse, whose frequency range is 10 Hz 1 MHz and minimum spacing adjustment is 10 Hz. The power amplifier is employed to enhance the power of exciting signal, which is used to excite the rectangular coil. Then, the response signals from pick-up coil are amplified and sampled by data acquisition module with 100 khz sampling rate. Operation of the set-up is controlled by a Windows XP-based program, which is programmed by Microsoft Visual C++ 6.0, combined with Matlab 7.0 [16,26]. In the experiment of defect edge identification, the amplitude of the exciting pulse is 10 V, the repetition rate of the excitation is 100 Hz and the pulse duration is 5 ms, whose schematic diagram is shown in authors another paper Ref [16]. An aluminum specimen whose thickness is 2 mm is designed to verify performance of the proposed method. On the surface, two slots (named defect one and defect two) are manufactured to simulate corrosion type of defects in real situation. The length width depth of defect one and defect two, respectively are 15 5 1.5 and 15 2.5 1.5 mm 3. 3. Rectangular pulsed eddy current sensor The probe we designed in our research consists of one rectangular exciting coil and one pick-up coil. The length, width and height of rectangular exciting coil, respectively, are 50, 45 and 45 mm. The number of turns is 400. The pick-up coil is located orthogonally in the centre at the bottom of the exciting coil, which are reeled with inductive coil to induce the change of magnetic fields along the scanning path. The turn of pick-up coil is 1000 [26]. In our previous studies, peak waves of magnetic field are analyzed in different directions of sensor scanning. One is the direction of magnetic induction flux; the other is the direction of exciting current. As shown in Fig. 1, considering the orientation of the coil to the sample, a Cartesian coordinate system is introduced. The direction of magnetic induction flux is parallel to X-axis and the direction of exciting current is parallel to Y-axis. In the course of probe scanning, the response signal of each pickup coil is sampled in real time and the periodic peak voltage of each signal is extracted at the same time. Consequently, peak waves form after peak voltages are connected [26]. To observe the difference of peak waves in different directions, a defect is detected utilizing the PEC sensor in both directions. As shown in Figs. 2 and 3, there are some kinds of scanning forms when sensor scans along the defect in each direction. The arrows point the scanning direction of rectangular sensor. Peak waves of response signals on different forms in each direction are shown in Figs. 4 and 5, respectively. The horizontal coordinates represent the scanning time; the vertical coordinates represent the amplitude of peak scanning waveforms. It can be seen from the plots in Figs. 4 and 5 that the peak waves are distorted as the defect is scanned. When sensor scans along the defect in the direction of magnetic induction flux, peak waves caused by defect in specimen are shown in Fig. 4. As shown in Fig 2 (5), when eddy currents in specimen induced by exciting pulse are disturbed by defect whose resistance is bigger than that of aluminum specimen, they will flow to the two ends and bottom of the defect. However, the flow directions of eddy currents in two ends are contrary. One is clockwise, while the other is anti-clockwise. Consequently, the magnetic force induced by changed eddy currents in two ends is also opposite. Hence, the currents in pick-up coil will change adversely when sensor is on the different ends of defect. In other Fig. 1. The diagram of PEC probe and scanning direction.
Y. He et al. / NDT&E International 43 (2010) 409 415 411 Fig. 2. The scanning forms of sensor in direction of magnetic induction flux: (1) and (2) pick-up coil is on the right of defect; (3) pick-up coil is on the centre of defect; (4) and (5) pick-up coil is on the left of defect. words, as the probe scans along the defect, a crest and a trough always appear on the peak waves [27]. In addition, when the pickup coil is away from the defect, such as the form 1 and 5, peak waves still present a crest and a trough. When sensor scans in the direction of exciting current, peak waves caused by defects are shown in Fig. 5. If eddy currents in specimen are disturbed by defect, they will flow to the two sides and bottom of the defect, which is shown in Fig. 3 (5). However, the flow directions of eddy currents in two sides are contrary. Consequently, the magnetic force induced by changed eddy currents in two sides is also opposite. So the induced currents in pick-up coil will change adversely when sensor is on the different sides of defect. In other words, when defect is on the left of pick-up coil (form 1 and 2), a broad crest appears on peak waves. In contrast, when defect is on the right of pick-up coil (form 4 and 5), a broad trough appears on peak waves. When defect is on the centre of sensor (form 3), because the eddy current flow to the bottom of defect, the magnetic force induced by eddy current is vertical with the normal of the pick-up coil. So, the variation of induced current in pick-up coil is small and the voltage of peak wave is approximate constant [27]. Furthermore, it can be seen that even the pick-up coil is away from the defect, peak waves in the form 1 and 5 still present a crest or a trough. Comparing the results in Figs. 4 and 5, we can find that even though the pick-up coil is not on the defect, the response signals still changed. Fortunately, in Fig. 4, the maximum and minimum values of peak waves are corresponding to the moment of sensor entering and leaving the length edge of defect [18 21]. In other words, the length edge of defect can be evaluated in magnetic induction flux. If we can identify the width edge in magnetic induction flux, the defect edge can be detected. Therefore, in next section, we analyze the C-scan imaging results in direction of magnetic induction flux. Fig. 3. The scanning forms of sensor in direction of exciting current: (1) and (2) pickup coil is on the right of defect; (3) pickup coil is on the centre of defect; (4) and (5) pickup coil is on the left of defect. Fig. 4. Peak waves on different forms in direction of magnetic induction flux. 4. Results and discussion 4.1. C-scan imaging results The specimen designed in Section 2 is used in the experiment of C-scan imaging detection. As shown in Fig. 6, the arrows point
412 Y. He et al. / NDT&E International 43 (2010) 409 415 Fig. 5. Peak waves on different forms in direction of exciting current. Fig. 7. The C-scan imaging results of defect one. Fig. 8. The C-scan imaging results of defect two. As the same in Fig. 8, the estimated width edge (from form 3 to 13, 10 mm) of defect two is also clearly bigger than actual width (2.5 mm). In other words, the width edge of defect in C-scan imaging results is overestimated. So, the new method is needed to improve the estimation of the width edge of defects. Fig. 6. The scanning forms of sensor against defect. the scanning direction of rectangular sensor. We make a research of 15 scanning forms when sensor scans along the defect in direction of magnetic induction flux. The step distance between two adjacent forms is 1 mm, which is controlled by X Y displacement system. The C-Scan imaging results of defects are shown in Figs. 7 and 8. It can be seen that the length of defect can be evaluated effectively. Unfortunately, the width of defects is evaluated large than the real width. As shown in Fig. 7, the average voltage of defect-free is 420 mv. If selecting the 40% (168 mv) of average voltage as the threshold voltage (588 mv), the width edge of defect one will appear between form 2 and 14. And then, the estimated width edge (from form 2 to 14) is 12 mm, which is obviously bigger than actual width (5 mm) of defect one. 4.2. Features from defect response signal As shown in Fig. 9, the regular features of response signals in time domain are as follows: time to rise, time to peak, the rising time, and peak amplitude [16,27]. The peak amplitude and the time to peak are the main features used in extracting information from PEC response signals. It is found that the peak amplitude is valuable to identify the edge of defect in experiments. However, the only one feature is not enough to identify the edge of defects. 4.3. Features from differential response signal In Section 4.1, the peak amplitude is proposed to identify the edge of defects in C-scan imaging results. For improving and enhancing the performance of edge identification, we extract three new features from differential response signals, which are
Y. He et al. / NDT&E International 43 (2010) 409 415 413 Fig. 9. The features of defect response signal in time domain. Fig. 11. The waves of peak and differential peak of defect one. computed by subtracting a defect-free signal from the defect signals. Defect-free signal is obtained when the probe is located on a known defect-free sample [16]. As shown in Fig. 10, the first feature is called as differential peak. The second feature is differential time to peak, which is the time to peak amplitude of the differential response signal. The last feature is the differential time to zero, which is also the time when the defect response signal (blue line in Fig. 10) intersects with defect-free response signal (black line in Fig. 10). In experiment, it is found that the three features of differential response signals can be used to identify the edge of defect. Therefore, the three new features are extracted and combined with peak amplitude to identify the edge of defect in direction of magnetic induction flux in next subsection. 4.4. Edge identification Fig. 10. The features of differential response signal. The main purpose of this subsection is to identify the width edge of defects on surface by exacting the appropriate features in time domain. The four features proposed in Sections 4.1 and 4.2 (peak, differential peak, differential time to peak, and differential time to zero) are used and processed. The four features on Fig. 12. The waves of differential time to peak and differential time to zero of defect one. maximum point of peak waves from 15 scanning forms (1 15 in Fig. 6) of defect one are shown in Figs. 11 and 12. The horizontal coordinates represent the state 1 15; the vertical coordinates represent the volt amplitude and time, respectively. Obviously, the waves of peak and differential peak present a crest. On the contrary, the waves of differential time to peak and differential time to zero present a trough. If normalizing the four feature waves, they will intersect at some time each other, which may be useful to identify the width edge of defect. The results of edge identification of defect one after the normalization in direction of magnetic induction flux are shown in Fig. 13. The horizontal coordinates represent the scanning form 1 15; the vertical coordinates represent the normalization amplitude. There is no doubt that the differential peak wave (blue line) and the wave of differential time to zero (black line) intersect at two junctions. One is between form 5 and 6; the other is between form 10 and 11. The distance of two junctions is 5 mm, which is the same to the width of defect one. Hence, the width edge is evaluated correctly, which is better than estimated width (12 mm) in Section 4.1.
414 Y. He et al. / NDT&E International 43 (2010) 409 415 Fig. 13. The features waves of defect one. present the same shape in direction of magnetic induction flux, while present different shapes in direction of exciting current. The maximum and minimum values of peak waves in direction of magnetic induction flux are related to the moment of sensor entering and leaving the length edge of defect, which gives us a way to estimate the length edge of defect. Therefore, we put our emphasis on the C-scan imaging results in direction of magnetic induction flux. Unfortunately, the estimated width edge of C-scan imaging results in direction of magnetic induction flux is unperfected. For improving the identification accuracy of width edge, we proposed news features from response signal and differential response signal. Experiment results show that the width edge of defects can be identified effectively by selecting and normalizing the appropriate features in time domain. Thus, the length edge and width edge can be evaluated in direction of magnetic induction flux. To sum up, the PEC rectangular sensor is valuable for inspection and identification of defects in field of nondestructive testing and evaluation. Future research of the authors will include the real-time defect identification, in situ defects evaluation, and defects reconstruction. Acknowledgements The authors would like to extend their appreciation to BinFeng Yang, Ping Xu, and TingTing Feng for contributions when they were in National University of Defense Technology. And the authors wish to thank the Subject Editor, Gerd Dobmann and the reviewers, for these very valuable comments and suggestions. The authors also thank National University of Defense Technology, China, for funding the study. References For validating the performance of method, the peak waves of defect two are processed as the defect one. The results of edge identification are shown in Fig. 14 and the estimated width is 3 mm. Although it is slightly bigger than the actual width of defect two (2.5 mm), it is further better than the estimated width (10 mm) in Section 4.1. Comparing the results in Figs. 13 and 14 with the results in Figs. 7 and 8 in Section 4.1, we can find that the identification of width edge is improved remarkably by proposed method. So, the method provides us a better way to identify the width edge of defect. Combining with the length edge, which has been evaluated by the crest and trough on peak waves in Section 3, the edge of defect can be identified effectively in direction of magnetic induction flux. 5. Conclusions Fig. 14. The features waves of defect two. In this paper, the further study of PEC rectangular probe proposed in author s previous work has been made to identify the edge of defects on surface. When sensor is on the different positions against the defect, peak waves of response signals [1] Moulder JC, Bieber JA, Ward WW, et al. Scanned pulsed eddy current instrument for non-destructive inspection of aging aircraft. SPIE 1996;2945: 2 13. [2] Smith RA, Hugo GR. Transient eddy current NDE for aging aircraft-capabilities and limitations. Insight 2001;43(1):14 25. [3] Smith RA, Hugo GR. Deep corrosion and crack detection in aging aircraft using transient eddy current NDE. Review of Progress in Quantitative NDE 1999: 401 1408. [4] Li S, Huang SL, Zhao W, et al. Study of pulse eddy current probes detecting cracks extending in all directions. Sensors and Actuators A 2008;141:13 9. [5] Tian GY, Sophian A, Taylor D, et al. Multiple sensors on pulsed eddy current detection for 3-D subsurface crack assessment. IEEE Sensors Journal 2005; 5(1):90 6. [6] Rose J.H., Uzal E. and Moulder J.C.. Pulsed eddy current characterization of corrosion in aircraft lap splices: quantitative modeling. SPIE;2160:164 176. [7] Tai CC, Rose JH, Moulder JC. Thickness and conductivity of metallic layers from pulsed eddy-current measurements. Review of Science Instrument 1996;67:3965 72. [8] Giguere S, Lepine BA, Dubois JMS. Pulsed eddy current technology: characterizing material loss with gap and lift off variations. Research in Nondestructive Evaluation 2001;13:119 29. [9] Yang HC, Tai CC. Pulsed eddy-current measurement of a conducting coating on a magnetic metal plate. Measurement and Science Technology 2002;13: 1259 1265. [10] Sophian A, Tian GY, Taylor D, et al. A feature extraction technique based on principal component analysis for pulsed Eddy current NDT. NDT&E International 2003;36:37 41. [11] Tian GY, Sophian A, Taylor D, et al. Pulsed eddy current system for dynamic inspection of defects. Insight 2004;46:256 9. [12] Tian GY, Sopian A. Defect classification using a new feature for pulsed eddy current sensors. NDT&E International 2005;38:77 82. [13] Tian GY, Sophian A, Taylor D, et al. Wavelet-based PCA defect classification and quantification for pulsed eddy current NDT. IEE Proceeding-Science, Measurement and Technology 2005;152(4):141 8. [14] Chen TL, Tian GY, Sophian A, et al. Feature extraction and selection for defect classification of pulse eddy current NDT. NDT&E International 2008;41: 467 476. [15] Fan MB, Huang PJ, Ye B, et al. Analytical modeling for transient probe response in pulsed eddy current testing. NDT&E International 2009;42: 376 383.
Y. He et al. / NDT&E International 43 (2010) 409 415 415 [16] He YZ, Luo FL, Pan MC, et al. Pulsed eddy current technique for defect detection in aircraft riveted structures. NDT&E International 2009;43: 176 181. [17] Tian GY, Yong L, Catalin M. Study of lift-off invariance for pulsed eddycurrent signals. IEEE Transaction on Magnetics 2009;45:184 91. [18] Collins R, Dover WD, Michael DH. The use of AC field measurements for nondestructive testing.. London: Academic Press; 1985. [19] Lewis AM, Michael DH, Lugg MC, Collins R. Thin-skin electromagnetic fields around surface-breaking. Journal of Applied Physics 1988;64(8): 3777 3784. [20] Knight MJ, Brennan FP, Dover WD. Effect of residual stress on ACFM crack measurements in drill collar threaded connections. NDT&E International 2004;37:337 43. [21] Dobmann G. Defect detection and sizing by eddy-current imaging with electric-current-perturbation (ECM) method. In: European Society for Nondestructive Testing: Proceedings of the third European Conference on Nondestructive testing, Florence, October, 1984 (3). [22] Dover WD, Collins R, Michael DH. The use of AC-field measurements for crack detection and sizing in air and underwater. Philosophical Transactions of the Royal Society London 1986;320A:271 83. [23] Yang BF, Luo FL, Zhang YH. Quantification and classification of cracks in aircraft multi-layered structure. Chinese Journal of Mechanical Engineering 2006;42(2):63 7. [24] Yang BF, Luo FL, Han D. Pulsed eddy current technique used for nondestructive inspection of aging aircraft. Insight 2006;48(7):411 4. [25] Yang BF, Luo FL, Han D. Research on edge identification of a defect using pulsed eddy current based on principal component analysis. NDT&E International 2007;40:294 9. [26] He YZ, Luo FL, Hu XC, et al. Defect identification and evaluation based on three-dimensional magnetic field measurement of pulsed eddy current. Insight 2009;51(6):310 4. [27] He YZ, Luo FL, Pan MC, et al. Defect classification based on rectangular pulsed eddy current sensor in different directions. Sensors and Actuators A 2010; 157:26 31.