Abstract
The paper presents the progress in the developments of an electro-magnetic acoustic transducer (EMAT) based on a Halbach magnet system optimized for detection of defects in thick plates with application to the inspection of fast breeder reactors (FBR) vessel. The defects are located either far or near weld lines in austenitic stainless steel plates (50 mm thick), on the opposite plate side where are located the EMAT sensors. Performance of single or double units EMATs are investigated for working in industrial conditions with very long cables (up to 45 m). Experimental measurements are focused on an increased signal/noise ratio for detection of signal from defects located near weld lines of FBR vessel. A fast and novel way of EMAT measurements and signal processing (named dynamic scanning) is introduced and its performance is confirmed for austenitic stainless steel plate defects with depths ranging from 5% to 50% from the plate thickness.
Introduction
Electro-magnetic acoustic transducers (EMAT) are based on the propagation of ultrasonic waves in metallic materials, those waves being created by Lorentz force oscillations when eddy currents formed in metals are subjected to the static field of permanent magnets [1]. The waveform signal can be pick-up either by the EMAT itself through variation of its coil impedance or by additional piezoelectric transducers [2]. However, their applicability can be also extended to online monitoring of thickness of metallic structures [3]. In the in-service inspection of fast breeder reactor (FBR) vessel, due to the high temperature (200 degrees), EMAT sensors main advantages is that they do not require a coupling between sensor and vessel when producing ultrasonic waves, therefore reliable operating at high temperature for longer periods of time [4–7]. Using samarium-cobalt magnets, it is possible for EMAT to retain excellent performance even at temperatures above 200 degrees without a cooling system. Previous work showed promising results when using EMAT for both, austenitic stainless steel and dissimilar metal welds [8]. However, the EMAT performance might varies when its application moves from laboratory controlled conditions of small samples to large industrial samples (due to neighboring of other large scale metals structures, running of other large industrial equipment, long cables, temperature variations and others), by experiencing a reduction in the signal/noise (S/N) ratio of defect detections.
In the present paper, we focus on the application of EMAT for inspection of FBR vessel, using a system based on Halbach magnet array, previously developed in our laboratory. Optimization of the EMAT sensor performance included transition from a single to a double unit. Experimental measurements were focused on increased detection of defects far from weld or heat-affected zone (HAZ), in regard to the operation in large structures, and validated in a reactor vessel (RV) mock-up. The ways and methods of achieving the right EMAT sensors performance when transitioning from test pieces in laboratory to large plates in a RV mock-up, as experienced by the authors, is explained and detailed in the paper. Also, a novel way, to present the EMAT signal not in a temporal but in a spatial C-scan distribution of amplitude, in assessment of sensor performance, that can be used in reactor in-service inspection, is also introduced and tested in the paper.
Experimental set-up of the EMAT system
The main parameters of EMAT system and test pieces used in the experimental measurements are detailed in Tables 1 and 2. Inspection of thick stainless steel 304 test pieces is employed using an EMAT based on a Halbach magnets structure. In the paper it is evaluated the performance of EMAT signal to check for outer defects in three small test samples (20 × 400 mm2) as well as in a large round plate mounted in the RV mock-up. The EMAT signal is generated at frequency of 680 kHz, being able to penetrate the 50 mm in thickness.
The samples No. 1 and No. 2 (shown in Fig. 1), are similar to the ones used in a FBR vessel, made of stainless steel 304 with a defect near weld (HAZ), while the sample No. 3 is free of welds.
Experimental measurements were conducted using the RITEC RPR-4000 to both generate the tone burst pulse of EMAT and also to receive the signal due to various defects (5% to 50% depth from plate thickness). The schematics of the experimental set-up, shown in Fig. 1, includes also a XY stage controller that controls the movement (with a 0.1 mm precision) of EMAT sensor that is in close contact with the surface of the test piece, an oscilloscope that display in real time the average frames in time domain of EMAT sensor, and a video recorder unit. In order to improve the acquired EMAT S/N ratio, the oscilloscope is set to an average of 256 frames, triggered by the RPR-4000 with a burst rate of 200 samples/secs. While the averaging is performed continuously in real time by the oscilloscope with the latest 256 frames, the video records with a 30 frames/secs the display of oscilloscope. The EMAT signal at each frame of the acquired video image is extracted, calibrated and digitized; therefore 30 frames of EMAT signals/secs are available as the sensor scan the surface of sample. If, for example, the EMAT sensors moves with a speed of 4 mm/s, then an EMAT signal will be acquired at every step of 0.13 mm.
The EMAT sensor is based on either a single or double units approach using a Halbach magnet system, for increasing magnetic field amplitude in the metallic plate subjected to be analyzed (see Fig. 2 left).
EMAT test parameter
EMAT test parameter
Parameters of test piece
Schematic view of EMAT testing and three test samples.
Schematic view of the test sample in the RV mock-up and double or single unit EMAT.
The large size round plate (800 × 1000 mm2), 50 mm thick is mounted in the RV mock-up (see Fig. 2 right) and has slits along a weld line, located either before or after weld as they are seen from the EMAT sensor location. The performance of EMAT sensors is analyzed experimentally for either small plates in laboratory or large round plate mounted in the RV mock-up, at room temperature.
EMAT operation point and acquisition of definition of signal: Static versus dynamic scanning
The operation point of EMAT sensors was chosen in accord with data presented in Table 1. These parameters change significantly the S/N ratio of the defect detection. Therefore, experimental measurements were conducted specifically for 50 mm plate thickness to measure the S/N ratio and not increasing only the signal S amplitude by varying the following parameters: frequency (400 ∼ 1300 kHz), number of input pulse cycle (3 ∼ 13), amplification signal in the excitation system (control = 10 ∼ 90), amplification signal in the detection system varies (gain = 20 ∼ 80 dB), burst rate of the input pulse (10 ∼ 500 Hz). Large values of amplification signal in either excitation or detection system, while indeed increases the EMAT signal amplitude, it was found that decreases usually the S/N ratio due to a nonlinear amplification effect in the EMAT equipment. Also, it was found that larger values of pulse cycle (beyond 11) increase the S/N ratio.
When EMAT sensor is located above a metallic plate the typical EMAT signal is presented in Fig. 3. Because the signal is high polluted with noise, multiple signals acquired at each burst, previously named frames (256 or larger) are required to be averaged.
In a “static C-scan” of a metallic plate the sensor is moved from point to point and while is position does not changed (is fixed) the signal is acquired, averaged and saved. In a “dynamic C-scan” the EMAT sensor moves continuously with a low speed. While the static scanning takes a very long time (30–40 hours for test sample No. 1 with a 2 mm pitch), the dynamic scanning is much faster (1–2 hours) but its sensitivity depends on the speed and scanning direction. Therefore, the dynamic C-scan procedure can be used in the inspection of RV but its actual performance has to be validated against static C-scan.
A novel way of acquiring the C-scan image is based on a spatial representation of EMAT signal and is newly introduced in the paper. At each point in the C-scan of sample the full EMAT time-transient signal acquired (Fig. 3 right) is replaced by one point data in order to build a C-scan EMAT representation. Therefore a simpler visual interpretation of the EMAT signal is associated to the inspected surface favouring more the defect signal while minimizing other signals (as weld signal). One point in the C-scan represents the amplitude at a defined moment of time t0 (fixed for all points), as defined in the formula in Fig. 3. The interval of time Δt, to calculate maximum amplitude, is related to the frequency 𝜈 of the input EMAT signal. The authors found that the value Δt = 2∕𝜈 provide a good representation of the signal in the C-scan. Larger values seems to decrease the S/N ratio, while smaller values then Δt =1∕𝜈, generates a more noisy signal. The path of moving EMAT in the dynamic scanning is also shown in Fig. 3 left.
While static scanning does not depends on the direction of scanning (parallel or perpendicular with the defect) the dynamic scanning has to be performed on a path parallel with the defect (see Fig. 3). If the direction of the defect is not known (as in a real inspection) then two scans 90 degrees rotated can overcame the above difficulty, but with a double time increased in scanning. Figure 4 shows also the signal scan along a line, parallel with the defect (at x = 60 mm). Because the signal is averaged continuously (last 256 frames), there is a non-alignment of signals when a double unit EMAT sensor scans the defect even on the same line but first from left → right and then again from right → left. The size of the non-alignment depends on the EMAT speed (4 mm/s or 1 mm/s). At higher speed the EMAT amplitude (defined by formula in Fig. 3) decreases due to additional spatial integration over a larger spanned area, but also the noise decreases, and therefore not changing significantly the S/N ratio. In case of static scanning, both scanning left → right or right → left are identically, since the sensor is not moved while the data frames are averaged and saved.
Figure 5 left and right shows the effect of number of frames averaging, for double unit EMAT sensor along line L in Fig. 3 (x = 60 mm in Fig. 6 left for 25%t) for static and dynamic scanning, respectively (with speed 1 mm/s). Lower noise in the dynamic scanning (Fig. 5 right) was observed and is due to the additional sensor signal spatial integration as sensor slowly moves in time. Therefore it was confirmed that at 256 average frames, dynamic scanning (at low speed) signal and S/N ratio is very similar with the one of static scanning. There is also a smaller influence of the averaging frames during dynamic scanning than static scanning.
A comparison of C-scan of test sample No. 1 using either static scan (left image) or dynamic scan at speed 4 mm/s (right image) is presented in Fig. 6 In the dynamic scan each line scanning has to be re-aligned (middle image), due to the reasons explained before. However, if the procedure of scanning is changed and each line is scanned in the same direction (the sensor is returned all the time to the y = 0 and then advanced to the next X position) there is no need for alignment, but the time of scanning will almost double due to additional sensor movements. Figure 7 shows a 3D plot of the same comparison between static and dynamic scan (4 mm/s) for detection of slits with either 25% or 50% depth. Therefore, it was confirmed that the dynamic scan provide enough resolution to detect a 25% depth slit with similar S/N ratio as static scan.
EMAT signal in time domain; definition of signal C (x, y)- a point in the C-scan.
Non-alignment and speed effect during dynamic scanning of a 25%t slit (256 averaging frames); Path of scanning in C-scan.
Effect of averaging frames (left-static scanning, right-dynamic scanning) on double unit EMAT signal.
C-scan for No. 1 test sample in 50%t slit before weld and 25%t slit after weld: (a) static scan; (b–c) dynamic scan.
Result of static and dynamic C-scan (4 mm/s) for 50%t and 25%t slit with double unit EMAT sensor.
In the dynamic scan, a faster speed of EMAT sensor is desirable as long as there is a high S/N ratio for detection of defects, located either before or after welds (as seen from the point of detection of EMAT sensor). Figure 8 shows the C-scan signal of double unit EMAT sensor using a dynamic scan at two speeds for sample No. 1 and confirmed that the sensor performance at 4 mm/s, even for 25%t slits located after welds.
While in laboratory tests the cables connected the EMAT sensors to the EMAT equipment are short (2 m), in the inspection of RV of FBR the cables are much longer (up to 45 m). The change of impedance due to cables has to be taken into considerations and the EMAT sensors (including cables) and designed to be as close to the 50-ohm impedance required at the amplification unit of EMAT equipment. Experimental measurements (shown in Fig. 9), demonstrated that when using longer cables, the S/N ratio of EMAT signal in the dynamic scanning can actually provide a better C-scan image than when using short cables, because of a better impedance matching between EMAT sensors and EMAT equipment. In the dynamic scan with 4 mm/s of sample No. 2 from B → A (as in Fig. 1) the S/N ratio of detection for a 12%t slit is near 4 (even when located after weld line) while S/N ratio for a 6%t slit is closer to 2.
EMAT scanning speed 1 mm/s and 4 mm/s (No. 1; 50%t, 25%t slit).
Dynamic C-scan (No. 2; from B → A, 12%t, 6%t slits) comparison between short (left) and long cable (right).
Dynamic C-scan (No. 2; from A → B 12%t, 6%t slits) comparison between double (left) and single (right) unit EMAT.
The C-scan procedure using the dynamic scanning and the proposed signal definition were also investigated for both a single unit EMAT. Figure 1 shows the results when using long cables (45 m) and 4 mm/s speed scanning of sample No. 2 from A → B. The single unit EMAT shortcomings are smaller S/N ratio as well as more noise from the edge/back of the test samples, confirm the advantages of double unit EMAT to be used with the above procedures in the dynamic scanning.
While in Fig. 10, the 12%t slit was located before weld line, in Fig. 9 the 12%t slit is located after the weld line, decreasing slightly the S/N ratio. However, the double unit EMAT sensor performance, using the developed C-scan procedure was limited in the detection of 6%t slit before weld line and could not “see” the same 6%t slit when located after weld line, showing the limits of double unit EMAT in detecting smaller slits next to weld line.
Experimental confirmation of smaller slits using the developed dynamic scanning procedure was investigated for sample No. 3, in the absence of weld lines and for 4 mm/s speeds and double unit EMAT sensor. The dynamic C-scan is shown in Fig. 11 when scanning the sample in both directions A → B and B → A, with 2 m cables. The tests also confirm the repeatability and consistency of the dynamic C-scan procedure and detection of slits up to 10%t when far from welds with S/N ratio close to 4.
In the inspection of a large sample mounted in the RV mock-up, long cable (45 m) where used with the double unit EMAT. In the present paper the performance of the sensor was investigated using only a static procedure for the EMAT signal, due to the difficulties associated with a controlled scan using a XY controller in the reduced gap (30 cm) between the RV and guard vessel. The main purpose of the experiment was to confirm the feasibility, quality and accuracy of the signal from the double unit EMAT even when located in a fixed point, but in a large metallic environment to provide a high S/N ratio for slits when located near reactor weld lines. Figure 12 displays experimental results of double unit EMAT signal of large sample (800 × 1000 mm2) in a RV mock-up, for 20% depth slit from plate thickness, located before or after weld lines, in the HAZ area, as compared to the signal from the weld line itself, free of defects.
Dynamic C-scan of sample No. 3 from A → B (left) and B → A (right). (5, 10, 20 50%t slits).
Time-scale for double unit EMAT signal in a large plate in the RV mock-up with long cables (45 m).
The measurements confirmed the high S/N ratio of signal and the feasibility of detecting up to 20%t slits with double unit EMAT with long cables, in the large sample mounted in the RV mock-up. Also, the weld signal is close to the EMAT noise signal. The present results in the RV mock-up were obtained at room temperature and showed that by carefully tweaking of double unit EMAT parameters is it possible to obtain high S/N ratio from defects located near reactor weld lines. The S/N ratio of detecting defects in both small samples in controlled laboratory conditions and large sample mounted in RV mock-up even when using long cables, was high and similar to each other. Therefore, it shows the feasibility of further applying the dynamic C-scan procedure to the inspection of welds in the RV if remote controlled robots carrying the EMAT sensor can move inside the gap between reactor and guard vessel.
Highly accurate, performant EMAT signals based on double unit EMAT and Halbach permanent magnet were demonstrated using experimental measurements with small (200 × 400 mm2) metallic samples and further confirmed in a large (800 × 1000 mm2) sample mounted in a FBR vessel mock-up. The paper showed that slits up to 10 ∼ 12%t near HAZ or far from weld lines could be confirmed with a high S/N ratio when using a new procedure of dynamic C-scan of surface small samples to visualize EMAT signal. The paper shows that carefully EMAT design, optimization and measurement methodology can resulting in highly accurate and repetitive measured signals, even with long cables (45 m) in both small samples in laboratory conditions or large sample mounted in a RV mockup.