Defects Sizing Using a Pulsed Eddy Current Testing Method for Local Wall-Thinning Evaluation
公開日:
カテゴリ: 第7回
Graduate School of Engineering, Tohoku University Shejuan XIE Student Member
Institute of Fluid Science, Tohoku University Toshiyuki TAKAGI Member
Institute of Fluid Science, Tohoku University Tetsuya UCHIMOTO Member
Introduction
In nuclear power plants, there may happen local all-thinning on the inner surface of a pipe due to flowcelerated corrosion (FAC) and liquid droplet opingement (LDI) of the coolant inside the pipe. ecause of the existence of thick insulators outside the pe, inspecting the wall-thinning which is located at the ner surface of a pipe by ultrasonic testing method comes very difficult. Pulsed eddy current testing ulsed ECT) technology is one method developed in cent years [1, 2]. Because of its un-necessity of contact tween the probe and the inspected specimen, also its ch frequency components and applicability of large ectric current [3-5], pulsed ECT method may showomising capability of detecting and evaluating of all-thinning in thick-walled piping with lift-off. As the -st step of this study, the aim of this paper is to discuss e feasibility of detection and evaluation of large area all-thinning and small area wall-thinning on the inner rface of a pipe using a pulsed ECT method with 1mm-off. - Pulsed ECT system1 Specimens -1 SpecimensTo simulate the large area wall-thinning on the ner surface of a pipe, six AISI316 austenitic stainless eel flat plates with different thicknesses were preparedour experiments. The size of the plates is various icknesses: 2, 3, 4, 5, 7mm and 10mm, 100mm in length2.2cons120mmmmDepth10mmDefectg.1. Plate specimen with a slot defect simulatingsmall area wall-thinning -2 Experiment setupPulsed ECT experiment system we established onsists of a function generator (WF1945, NF), a power mplifier (BP4610, NF), a scanning stage, an AD board ad a PC etc. In the experiments, square wave pulse was enerated from the function generator and then amplifiedthe power amplifier by which the output current could - controlled instead of the output voltage. Then the 300 -mplified current signal of square wave pulse was pplied to the exciting coil as the exciting signal. A Hall ensor, one kind of magnetic sensor, is located at the ottom center of the exciting coil as the pick-up sensor, whose sensitivity is 5mV/G, and the manufacturer isllegro Microsystems. Magnetic flux density of vertical Eirection is measured with the magnetic sensor. The ulsed ECT experiment setup was shown in Figure 2.Function generatorPower amplifierScanning stageHACoil, pick-up sensor and specimenFig.2. Pulsed ECT experiment setup2.3 Simulation methodFinite Element Method was applied and ANSYS software was chosen as the simulation tool to simulate the transient eddy current problem. For the simulation with the specimens of the first group we can simply use the axisymmetric model because the plate-specimens are only various in different thicknesses and the exciting coil is pancake shape [6-9).3. Results and discussion of large area vall-thinning detectionIn this study, the thickness of a plate is to be valuated from the pick-up signal using pulsed ECT nethod in experiments and simulation, respectively.3.1 Original pick-up signal validationIn the experiment, lift-off is 1mm and the parameters of the exciting coil are inner diameter 30mm, outer diameter 40mm, height 15mm, wire diameter 1mm and turns 60. The pulse exciting current is exponentially decrease from high level to low level (see Figure 3) and he magnitude of the DC part is 7.8A, the period is 0.01s, he duty is 50%. The pick-up signal collected by the Hall sensor was averaged over 100 cycles of the transient output for duration of 1.0s, to reduce noise.Typical 1-D transient outputs at the measured point of 4mm thickness plate and 10mm thickness plate are shown in Figure 4 for both experimental and numerical results. We can see that the experimental results and numerical simulation results match well quantitatively with each other. So the validity of our pulsed ECT experiment setup was verified.-:- Exciting current: both in experimentsand simulationExciting current (A)0.0048 0.0049 0.0050 0.0051 0.0052Time (s) Fig.3. Exciting signals both in experiments andsimulation20061Experimental results: o Som: Signal of plate thickness 4mm a?￠ Sommi Signal of plate thickness 10mmSimulation results:-:- SammSommMagnetic induction, Bz (G)OoooooooooDORADQ00000Bococooooooooooooooooo0000000000oodood dodaca89cosog000080.0050.00560.0052 0.0054Time(s)Fig.4. Comparison of original pick-up signals betweenexperiments and simulation3.2 New feature extractionFigure 5 and Figure 6 show the experimental and simulation results of the differential signal of magnetic flux density based on the above pick-up signals, respectively, where the reference signal is the signal measured on a plate of 10mm thickness. And Figure 7 shows the typical conventional feature extraction from the differential signal, that is, peak value and peak time. Peak value is the magnitude of the peak point and peak time is the time to the peak point [10].As we know the two conventional features extracted from the differential signal both are the property of the peak point, so we can simply imagine that they strongly rely on only the peak point, thus they should be easily affected by the occasional error in real application. Here another new feature was extracted from the differential signal, ““area““, which is the area between the differential signal curve and the time axis. We could imagine that even though the peak point was affected by the occasional error a little the area should not change a lot because it contains the information of all the points in pick-up signal, not only one point (peak point).301Experiment resultsDifferential magnetic induction, BZ (G)m7mm0.0050 0.0052 0.0054 0.0056 0.0058Time (s) Fig.5. Differential signal of magnetic flux densityin experimentsSimulation resultsm2mmSonun 3mmS. 10mm-S.4mmSomsmmDifferential magnetic induction, Bz (G)0.0050 0.0052 0.0054 0.0056 0.0058Time (s) Fig.6. Differential signal of magnetic flux densityin simulation- Experiment, Sonn-SammSimulation, Somm-S2mmDifferential magnetic induction Bz (G)Peak valuePeak time0.0050.00560.0052 0.0054Time(s)Fig. 7. Typical conventional feature extractionin differential signalRepeated experiments were carried out five times for every plate, the averaged signal was utilized as the experimental pick-up signal to avoid the occasional error which we mentioned above. The three features (peak value, peak time, area) were extracted from the experimental results and simulation results and shown in Figure 8, 9 and 10, respectively. We can see that all of them can give us good agreements with simulation results. Thus the thickness of a plate can be successfully evaluated from the characteristics of the differential pick-up signal.- - Experiment peak value -.- Simulation peak valuePeak value in differential signal, Bz (G)273 4 5 6 Thickness of plate (mm)Fig.8. Relationship between peak value and thicknessof plate both in experiments and simulation- Experiment peak time - Simulation peak time-Peak time in differential signal Bz (us)23 4 5 6 Thickness of plate (mm)7Fig. 9. Relationship between peak time and thicknessof plate both in experiments and simulation6000- Experiment area - Simulation area5000Area (Gauss x us)200010002A13 4 5 6 Thickness of plate (mm)Fig.10. Relationship between area and thickness ofplate both in experiments and simulation3.3 Stability investigation of featuresAs we know, stability of the extracted features is very important for thickness evaluation in practical application. So here three features were extracted from the above five times repeated experimental results of every plate, respectively. Results of 2mm thickness plate were investigated to check the stability of the extracted features. Table 1 shows the values of the three features in every experimental result and their errors in which the value of mean square error divided by mean value was applied.302Table 1 Stability comparison of the three features inrepeated experimental resultsEvaluation itemsPeak time(us)Experiment 1 Experiment 2 Experiment 3 Experiment 4Experiment 5 Mean square error /mean valuePeak value (Gauss) 29.865 29.895 29.865 30.011 30.223 0.51%76 73 76 71Area (Gaussxus)5182.9 5172.0 5161.8 5130.2 5161.8 0.38%730.029From Table 1 we can see that the error of peak time is the biggest, so peak time is the most not stable feature in the repeated experimental results. Inversely, the errors of peak value and area are very small, so they are relative stable features. Thus we could conclude that in practical application, if possible, feature of peak value or area had better been applied as the characteristic to evaluate the defect, not peak time.4. Results and discussion of small area wall-thinning detectionIn this study, the depth of a slot defect on the bottom side of a plate is to be evaluated from the pick-up signal using feature of peak value in experiments.In the experiment, lift-off is 1mm and the parameters of the exciting coil are inner diameter 30mm, outer diameter 50mm, height 15mm, wire diameter 1mm and turns 129. The magnitude of the DC part of the pulse exciting current is 10A, the period is 0.02s, the duty is 50%. The pick-up signal collected by the Hall sensor was also averaged over 100 cycles of the transient output for duration of 2.0s, to reduce noise.Specimen and inspection condition are shown in Figure 11. In experiments, scanning signal were obtained using scanning stage and the point which is 27mm far from the center of the defect was chosen as the origin. The bottom center of the exciting coil (also the position of Hall sensor locates at) started from the origin point and moved along the scanning direction where the interval of every step is 2mm.120mmHall sensorExciting chaExciting coil““Scanning directionScanning directionLift-off: Imm27mm10mmOrigin10mmDefectFig. 11. Inspection and scanning modeThe peak value at every scanning point was extracted from the differential signal where the reference signal is the signal of the first point (signal of origin point) to detect the slot defect of the plate. Figure 12 shows the results of five different specimens with different depth of slot defect. We can see that the positions of the absolute maximum value in all scanningsignals are the same which are located at the center of the slot defect. In addition, the absolute maximum value of the scanning signal becomes larger when the depth of the defect becomes bigger. Figure 13 shows the relationship between the absolute maximum value of scanning signal and the depth of defect.lift-off: 1mmPeak value of differential signal (G)d: depth of defect - -d=8mm - -d=7mm - A-d=6mm - -d=5mmd=2.5mm05010 20 30 40 Position of coil center (mm)Fig. 12. 1-D scanning signal of five differentspecimensMaximum value of the scanning signal (G)09 101 2 3 4 5 6 7 8Depth of slit defect (mm)Fig.13. Relationship between the absolute maximumvalue of scanning signal and depth of defectThus we can conclude that the depth of a slot defect can be detected and evaluated through the relationship between the absolute maximum value of scanning signal and depth of defect.Here another point that we need pay attention to is the defect of 2.5mm depth could not be detected because its peak value of defect signal is close to the noise level of Hall sensor which is about 1.0Gauss [11]. So for the future work, more sensitive magnetic sensor combined with corresponding novel exciting mode will be considered to increase the capability of detection of small defect in thicker specimen with bigger lift-off [12-13].5. ConclusionA novel and more stable feature was extracted from303the differential signal as one new characteristic to evaluate the wall-thinning in pulsed ECT method. Numerical and experimental results demonstrate that the difference in the thickness of a plate (thinner than 10mm) and the depth of a slot defect in a plate (thickness is 10mm) can be obviously evaluated from the characteristics of the differential signals. So the large and small area wall-thinning of the pipes in nuclear power plants both can be detected using pulsed ECT method. For the future work, high sensitive magnetic sensor combined with corresponding novel exciting mode will be considered to increase the capability of detection of small defect in thicker specimen with bigger lift-off. AcknowledgmentsThis work was conduced as a part of Nuclear and Industrial Safety Agency (NISA) project on Enhancement of Ageing Management and Maintenance of Nuclear Power Plants in Japan and supported by the Grant-in-Aid for the Global COE Program, ““World Centre of Education and Research for Trans-Disciplinary Flow Dynamics““, from the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan. The authors would like to thank Mr. Takeshi Sato of Institute of Fluid Science, Tohoku University, for the preparation of the specimens and the fabrication of the probe. G.Y. Tian, A. Sophian. Defect classification using a new feature for pulsed eddy current sensors. NDT&E International 38 (2005) 77-82. J. Kim, G. Yang, L. Udpa, S. Udpa. Classification of pulsed eddy current GMR data on aircraft structures. NDT&E International 43 (2010) 141-144. M. Fan et al. Analytical modeling for transient probe response in pulsed eddy current testing. NDT&E International 42 (2009) 376a?“383. T. Chen et al. Feature extraction and selection for defect classification of pulsed eddy current NDT. NDT&E International 41 (2008) 467-476. M. Morozov, G.Y. Tian and D. Edgar. Comparison of PEC and SFEC NDE Techniques. NDT&E International 24 (2009) 153-164. Y. Li et al. Fast analytical modelling for pulsed eddy current evaluation. NDT&E International 41 (2008) 477a?“483. J.R. Bowler. Pulsed eddy current response to a conducting half-space. IEEE TRANSACTIONS ON MAGNETICS 33 (1997) 2258-2264. R. Ludwig. Numerical and analytical modeling of pulsed eddy currents in a conducting half-space. IEEE TRANSACTIONS ON MAGNETICS 26 (1990) 299-307. F.W. Fu and J.R. Bowler. Transient eddy current References [1] G.Y. Tian, A. Sophian. Defect classification using anew feature for pulsed eddy current sensors.NDT&E International 38 (2005) 77-82. [2] J. Kim, G. Yang, L. Udpa, S. Udpa. Classificationof pulsed eddy current GMR data on aircraft structures. NDT&E International 43 (2010)141-144. [3] M. Fan et al. Analytical modeling for transientprobe response in pulsed eddy current testing.NDT&E International 42 (2009) 376a?“383. [4] T. Chen et al. Feature extraction and selection fordefect classification of pulsed eddy current NDT.NDT&E International 41 (2008) 467-476. [5] M. Morozov, G.Y. Tian and D. Edgar. Comparisonof PEC and SFEC NDE Techniques. NDT&EInternational 24 (2009) 153a?“164. [6] Y. Li et al. Fast analytical modelling for pulsededdy current evaluation. NDT&E International 41(2008) 477-483. [7] J.R. Bowler. Pulsed eddy current response to aconducting half-space. IEEE TRANSACTIONSON MAGNETICS 33 (1997) 2258-2264. [8] R. Ludwig. Numerical and analytical modeling ofpulsed eddy currents in a conducting half-space. IEEE TRANSACTIONS ON MAGNETICS 26(1990) 299-307. [9] F.W. Fu and J.R. Bowler. Transient eddy current- 304 - response due a conductive cylindrical rod. Review of Quantitative Nondestructive Evaluation 26(2007) 332-339. [10] A. Sophian, G.Y. Tian, D. Taylor, J. Rudlin. Afeature extraction techniques for pulsed eddy current NDT. NDT&E International 36 (2003)37-41 [11] G.Y. Tian and A. Sophian. Study of magneticsensors for pulsed eddy current techniques. Insight47 (2005) 277-279. [12] J. Kim, G. Yang, L. Udpa, S. Udpa. Classificationof pulsed eddy current GMR data on aircraft structures. NDT&E International 43 (2010)141-144. [13] S. Gartner, H.J. Krause. Non-destructiveevaluation of aircraft structures with a multiplexed HTS rf SQUID magnetometer array. Physica C 372a?“376 (2002) 287a?“290.on43“ “Defects Sizing Using a Pulsed Eddy Current Testing Method forLocal Wall-Thinning Evaluation“ “Shejuan XIE Student,Toshiyuki TAKAGI,Tetsuya UCHIMOTO