Window multi-frequency ECT for multi-coils in a RFEC system for FBR SG tubes

1. Introduction

Multi-frequency (MF) algorithms, based on eddy currents (EC), used in the in-service inspection (ISI) of steam generator (SG) tubes of nuclear reactors is a well-known applied technique used to improve the signal/noise (S/N) ratio of defect detection [1–3]. When it is applied to the SG tubes of fast breeder reactors (FBR) cooled by sodium, the high sodium electrical conductivity changes the EC distribution around tube defects [4]. For ferromagnetic SG tubes of FBR, a remote field eddy current (RFEC) technique [5,6] is sensitive to outer tubes defects (OD) but when it is applied to a small multi-coils sensors system, its sensitivity is a fine balance between sensors size, numbers and their orientations. For detection of defects located under SG tubes support plates (SP), the authors developed in the past a new algorithm named windows multi-frequency (WMF), that was shown it can reduce both sodium and SP signal. Application of WMF was previous investigated [4,7,9] by the authors using bobbin coils type with either FEM simulations or experimental measurements in the absence or presence of sodium. Also, it was validated its application in detecting smaller defects when using multi-coils at mixed 150–450 Hz frequencies, but by using only 3D FEM simulations of various sources of noises (lift-off, sodium filling defects volume) and each of them separated and not combined together [7–9]. In the present paper, it is extended the validations of WMF algorithms with the use of multi-coil RFEC sensors, by using either only experimental measurements in the WMF algorithm, or combined numerical FEM simulation-measurements, to enhance the S/N ratio for defect detection before SG tubes were soaked in sodium or even after SG tubes were drained of sodium. The optimized frequency set to be used with WMF is determined based on validation with experimental measurements.

2. 3D FEM simulations of the RFEC signal for helical FBR SG tubes

The 3D FEM simulations are based on a code previously developed/validated in-house [7–10] and conducted for SG tubes connected to a large support plate (SP) as shown in Fig. 1a. C-scan signals, mapping the SG tube outer surface, were computed in the past by authors based on an in-house developed code 3D-RFEC and using large scale parallel computer (up to 5192 CPUs) (Fig. 1b). The results of FEM simulations were compared previously by authors [10] and validated with experimental measurements of ECT signal from multi-coil RFEC sensors with 10 channels (1 channel being a differential connected pair of two coils), shown in Fig. 1c, for three helical SG tubes connected to a large SP, but at single eddy currents frequencies.

OPEN IN VIEWER

Fig. 1.

(a) View of SG tubes and support plate; (b) Comparison of 3D FEM simulations with measurements of ECT signal from SG tube support; (c) 10-channel multicoils sensors in the RFEC system.

Outer tube defects are located near tube SP: an outer tube circumferential groove 10 mm wide and depths 20%tw (from tube wall thickness) and a partial outer tube defect 30%tw and 90 degrees circumferential spanned are located between tube SP and SG tube. After tubes are soaked and drained of sodium, the defects volume might be filled with sodium in amounts ranging from 0% to 100%. These defects were machined in different locations, far from each other along helical tubes, and supported by identical SP structures. An example of SP and location defects near SP is shown in Fig. 1.

The parameters of SG tubes and of the RFEC multi-coil detection and bobbin excitation system in the RFEC sensor are presented in Table 1.

In previous papers [4,7–9], several new variants of WMF eddy current algorithms were validated for helical tubes using FEM simulated sodium deposits inside defects or using experimental measurements but only in the absence of sodium. In the WMF algorithm a standard MF algorithm is applied to a smaller subset data, called “window”, of the ECT data. The parameters of WMF in the “window” are previously optimally found after using a brute force attack on all data length to ensure the maximum S/N ratio. In this particular case, the “noise” is considered to be the SP signal combined with the sodium signal while the “signal” is the defect signal.

The present paper uses the WMF algorithm, which is found out by a brute force attack in a limited “window area”. For each “window” S/N ratio is maximized through adjustment of algorithm parameters. An example of WMF output as ECT sensor moves along SG tube is shown in Fig. 2.

OPEN IN VIEWER

Fig. 2.

Schematic of dynamic WMF algorithm applied to the eddy current signal.

In the dynamical WMF algorithm, previously described in [7], (see Eq. 1), it is used a fixed and smaller subset data of length “w” from signal of length “n” at two frequencies S ₁(1, n) and S ₂(1, n). These signals are combined using a standard MF algorithm (defined by amplification 𝛼 and rotation matrix R (𝜙)), for each window “w” resulting in the signal S _i . The dynamic WMF signal S _w (i) at each “i” point of the signal is defined by the maximum ratio S _i ∕N _i where N _i is the signal of noise. (1)

C-scan signals are computed using an in-house developed FEM code (3D-RFEC code) for various combined noise sources (lift-off and sodium filling the defects). Application of WMF to the computed ECT signals for the multi-coil RFEC unit, investigated for combined noises (lift-off and sodium filling the defect volume) is shown in Fig. 3a. Large S/N ratios are obtained even when combined effect of noises is taken into account for the detection of both partial OD30%tw (with depth 30% from tube wall thickness) and full circumferential groove OD20%tw located under tube SP.

OPEN IN VIEWER

Fig. 3.

(a) WMF using 3D FEM simulations of ECT signals; (b) The effect of window band size on defect signal.

In previous work, various WMF algorithms were investigated by 3D FEM simulations at frequencies between 150 and 450 Hz to determine the optimum window band size [9]. Figure 3a shows the application of dynamic WMF using 150–250 Hz frequency set for various “window” sizes (from 3 to 14 mm) and when sodium fills the circumferential groove OD20%tw or partial OD30%tw volume from 0% to 100%.

A minimum “S/N > 2” criteria was chosen in order to have a clear distinction between defect signal and noise. The noise signal is represented by the maximum signal amplitude in the area excluding the SP and defect. The most optimal “window band” size (7–14 mm), as determined through FEM simulations, is a value closer to the RFEC detection coils size (13 mm), showing that the optimal value of dynamic WMF is actually connected to the size of RFEC multi-coil detection system.

When WMF is applied directly to experimental measurements from multi-coils RFEC unit, the S/N ratio of detection of partial OD30%tw located under tube SP varies with the band size and the frequency of the low-pass filter used to remove high frequency noise. Figure 4a shows that high peak of S/N ratio (larger then 8) can be obtained for window band size between 11–15 mm, in accord with previous 3D FEM simulations. Therefore, all measurements were filtered with a low-pass filter at 60 Hz and the operation area of RFEC multi-coil unit is shown in Fig. 4a. Further investigation of WMF algorithms using only experimental measurements for combination of all 2-frequencies sets among (150, 250, 350 and 450 Hz signals) showed in Fig. 3b that the peak S > N ratio for detection of OD30%tw under tube SP is higher when lower frequency sets are used as in next pairs: 150–250 Hz, 150–350 Hz, 150–450 Hz or 250–450 Hz.

OPEN IN VIEWER

Fig. 4.

(a) Influence of window band in dynamic WMF algorithm using measurements with a low-pass filter using 2-frequencies; (b) Peak of S/N using 2 frequencies in dynamic WMF for measurements of partial OD30%tw near tube SP.

Each RFEC measurement was recorded separately at each frequency with the RFEC unit inserted inside SG tubes each time. Therefore, slightly rotation of coils when passing the defect area or variation of distance between signals of SP adds additional noise in the measurements However, in order to demonstrate the feasibility of the WMF algorithm, the input data were not stretched, but only aligned to the same starting point, therefore the SP signal is visible at 120 mm and 220 mm mark. Figure 5 shows the measurement of ECT signal of RFEC multi-coil unit at 150 Hz and 250 Hz for partial OD30%tw or circumferential groove OD20%tw under tube SP. In successive measurements frequency scans (1st at 150 Hz, 2nd at 250 Hz and so on) of the defect area, the location of multi-coils is different rotated relative to defect area and cannot be controlled as is passing the defect area, therefore different lift-offs are expected for each coil in the multi-coil unit. Also, because of speed variations as the probe moves inside SG tubes, the peak of SP and defect signals might not be well aligned when the 150 Hz and 250 Hz signals are compared, for example.

OPEN IN VIEWER

Fig. 5.

Measurements of multi-coils ECT signal of defects near tube SP at 150 Hz: (a) partial OD30%tw; (b) groove OD20%tw or at 250 Hz: (c) partial OD30%tw; (d) groove OD20%tw.

The application of WMF directly to experimental measurements from SG tubes, before they were soaked in sodium is shown in Fig. 6 for both partial OD30%tw and circumferential groove OD20%tw using the 150–250 Hz frequencies set. Despite all the additional noises in measurements (data not aligned well, various lift-offs of coils at each frequency) the defect area is properly identified in the dynamic WMF algorithm with a high S/N ratio (>6).

OPEN IN VIEWER

Fig. 6.

WMF applied to measurements before sodium adhering to SP and SG tubes (a) partial OD30%tw; (b) groove OD20%tw.

The application of dynamic WMF to the measurements from tubes after they were soaked and drained of sodium is presented in Fig. 7 and they also show a proper suppression of SP and sodium signal, but with a 6 ∼ 19% smaller value of S/N ratio then before tubes were soaked in sodium, confirming the feasibility of WMF algorithms and also being in agreement with previous 3D FEM simulations.

OPEN IN VIEWER

Fig. 7.

WMF applied to measurements after sodium was drained from SP and SG tubes: (a) partial OD30%tw; (b) groove OD20%tw.

While in experimental measurements in mock-up (ECT signal from SG tubes) both signals from defects near tube SP or SP free of defects are known (and can be combined in the WMF algorithm), in the actual inspection of SG tubes of FBR only one measurement is available: the ECT data from coils scanning the SG tube. Therefore, it is not possible to know if the data is either a signal from a SP with sodium residue or a SP with a defect located under it. The authors shows that the computed SP signal using 3D FEM simulations can be used as a reference signal, as long as the simulations can accurately compute the ECT signal from multi-coil RFEC unit, as was shown in Fig. 1b. Therefore, in the next was investigated the feasibility of using WMF algorithm but with mixed data: numerical simulations of SP (free of defect) and experimental measurements (without knowing if there is a defect hidden under tube SP). Figure 8a presents the validation of the WMF using the above approach, before SG tubes were soaked in sodium, and shows that SP signal is removed. If a defect is located under SP, then the WMF algorithm is able to pick-up the defect location (Fig. 8b). The same results is shown in Fig. 9 for SG tubes with both partial OD30%tw and groove OD20%tw near tube SP after the tubes were soaked and drained of sodium, confirming the feasibility of WMF algorithms using mixed simulation-measurements data.

OPEN IN VIEWER

Fig. 8.

Validation of WMF using mixed FEM simulations-measurements to suppress SP (before sodium adherence): (a) tube SP; (b) partial OD30%tw near SP.

OPEN IN VIEWER

Fig. 9.

Validation of WMF using mixed FEM simulations-measurements to suppress SP and sodium signal (after sodium drain): (a) groove OD20%tw near SP; (b) partial OD30%tw near SP.

Based on numerical 3D FEM simulations, the impact of various noises on WMF algorithms can be assessed for their maximum combined effect and optimization of WMF band. When WMF algorithms are applied to the measurements ECT signals from small multi-coils, optimum defect detection and multi-coil RFEC sensor operation area can be estimated. Through validations of FEM simulations with measurements, in the presence of sodium, WMF algorithm performance can be confirmed to be used in ISI of FBR SG tubes. The paper shows the applicability of using WMF algorithm directly with experimental measurements (if both SP and SP with defect signal are known, in mock-up) or with mixed FEM simulations-measurements (to be used in an actual in-service inspection) to confirm the location of outer tube defects (partial OD30%tw or groove OD20%tw) near tube SP before SG tubes were soaked in sodium or after sodium was drained.