Circumferential Current Field Testing System with TMR Sensor Array for Non-contact Detection and Estimation of Cracks on Power Plant Piping

Abstract

Piping is a critical component in power plant, which is subject to surface thermal fatigue cracks. Due to the high temperature and the presence of coatings, conventional non-destructive testing methods are unsuitable to detect the thermal fatigue cracks on power plant piping. This paper presents a novel circumferential uniform current field testing system for the non-contact detection and estimation of cracks on power plant piping. A 3D dynamic finite element method model is developed to obtain the characteristic signals and optimize the inducing frequency. The circumferential uniform current field testing system with tunnel magneto resistance (TMR) sensor array is set up and the crack detection experiments are carried out. The results show that the circumferential uniform current field testing system with TMR sensor array can achieve a non-contact, visual and efficient detection of cracks on piping in a one pass scan. The depth and length of cracks can be estimated by the proposed calibration method.

1 Introduction

Piping is a critical component in power plant, such as the steam piping, hot reheat steam piping, cold reheat steam piping, feed water piping, cooling water piping and auxiliary piping. The piping serves in a harsh environment due to the noticeable cyclical pressure and thermal stress [1,2,3]. The pressure can reach 30 MPa and the temperature is more than 500 ℃ inside the piping. The high pressure and temperature can cause initial thermal fatigue cracks on the outer surface of piping [4, 5]. What’s more, the initial thermal fatigue cracks gather and grow quickly by the cyclical stress, which cause the leakage and failure of the piping [6, 7]. Therefore, it is a critical issue to detect and estimate cracks on power plant piping before the failure happens [8, 9].

There are various non-destructive testing (NDT) methods for the detection of cracks on piping. Because of the closed piping system, the inner detection techniques, such as pipeline pigs, cannot be applied for the detection of cracks on power plant piping during in-service time. The ultrasonic testing (UT) needs coupling medium, which makes it an inadequate method in high temperature environment [10]. The infrared testing is a good method for surface cracks detection. However, the temperature changes constantly on the surface of piping. Thus it is hard to perform the infrared testing on power plant piping [11]. The magnetic flux leakage testing (MFL) is not sensitive to the narrow thermal fatigue cracks because of the little leakage of magnetic field on stainless steel piping [12]. The eddy current testing (ET) is sensitive to lift-off, which cannot penetrate thick coatings [13]. The microwave waveguide imaging technique is mainly used for corrosion inspection on plate [14]. What’s more, most of the methods as mentioned above need multiple scans to achieve a full 360° inspection of the piping surface.

The magnetic particle testing (MT) and penetrant testing (PT) are traditional NDT methods for surface cracks detection on power plant piping. However, most processes of MT and PT needs manual operation in the harsh environment. These methods just give surface information of the crack but no depth information which is more important to estimate the residual life of the piping. What’s more, the coatings need to be removed or the attachments on the piping should be cleaned before NDT operation. The magnetic particles and penetrants should be cleaned and the coatings should be re-painted after the inspection, which is time-consuming and costly [15].

The current field perturbation NDT methods, such as alternating current field measurement (ACFM), alternating current potential drop (ACPD) technique, are promising techniques to inspect defects on conductive material [16, 17]. For the detection of cracks on piping, the coaxial solenoid loaded with sinusoidal excitation signal can induces the current field on the surface of piping in circumferential direction. In the center of the coaxial solenoid, the induced circumferential current field on piping can be regard as an approximate uniform field. When the crack is presented, the circumferential current field will be disturbed. The space magnetic field can be picked up by sensors around the piping without contact [18, 19]. The defects can be detected and evaluated by the distorted magnetic field.

The circumferential uniform current field testing system is presented for non-contact detection and estimation of surface cracks on the power plant piping in this paper. The circumferential current is induced on the surface of piping by a coaxial encircling excitation coil. The inducing frequency is optimized to balance the penetration depth and detection sensitivity. The circumferential uniform current field testing system can cover a full 360° area on the surface of piping by TMR sensor array in a one pass scan. All the cracks on the surface of piping can be identified and evaluated visually and efficiently by the space magnetic field without contact using the circumferential uniform current field testing system.

This paper is organized as follows. In Sect. 2, the finite element method (FEM) model is set up and the inducing frequency is optimized by the simulation model. In Sect. 3, the circumferential uniform current field testing system is developed and the crack detection experiments are carried out. In Sect. 4, the cracks are estimated by the distorted space magnetic field. In Sect. 5, the conclusion and further work are delivered.

2 Circumferential Current Field FEM Model

2.1 Model Set up

The 3D FEM model of the circumferential current field is set up by ANSYS software, as shown in Fig. 1a. The model consists of excitation coil, piping and axial crack. The excitation coil is coaxial with the piping. The axial crack lies on the outer surface of piping symmetrically. To simulate the detection process, a dynamic FEM model is developed. The excitation coil is driven at 0.1 mm step-size to scan along the piping. The material of excitation coil is copper and the piping is stainless steel. As shown in Fig. 1b, the lift-off of the excitation coil is 10 mm and the lift-off of sensors is 7 mm, which gives enough space for the coatings. The size of the crack is 20 mm in length, 0.3 mm in width, 2 mm in depth. A sinusoidal excitation signal (frequency 20 kHz, amplitude 2 V) is loaded on the excitation coil.

Fig. 1

3D FEM model of circumferential current field testing probe. a FEM model. b Size of the FEM model

When the excitation coil is above the crack, the current field around the crack is extracted on the surface of pipe string. As shown in the bottom right corner of Fig. 1a, the circumferential current field is uniform when the crack is not presented. The uniform current field gathers and turns around at the tips of the axial crack. The disturbed uniform current field perturbs the space magnetic field above the crack [20].

To get the distorted magnetic field caused by disturbed uniform current field, the magnetic field in axial direction (Bx) and radial direction (Bz) are picked up in the center of the excitation coil at the lift-off of 7 mm. Thus, as the excitation coil moves one step, the Bx and Bz are plotted once at each position, as shown in Fig. 2. The Bx shows a trough in the center of the crack due to the decrease of current density in the depth of the crack, which contains the depth information. Meanwhile, the Bz plots a positive and negative peak at the tips of the crack. The distance between the two peaks reflects the length of the crack.

Fig. 2

Bx and Bz

2.2 Frequency Optimization

Due to the skin effect, inducing frequency determines the penetration depth on piping, which has a great significant influence on detection sensitivity of crack depth [21, 22]. The skin thickness is given in Eq. (1), where $u_{{\text{r}}}$ is the relative magnetic permeability, $u_{0}$ is the magnetic permeability of free space, and $\sigma$ is the electrical conductance, $f$ is the inducing frequency. When applying a high frequency excitation signal on the excitation coil, the current field tends to concentrate in a thin layer flowing in circumferential direction, which goes against sizing crack depth [23]. When the inducing frequency is low, the current field has a large penetration depth, which is good for sizing crack depth. However, when the inducing frequency is too low, the induced current density is very weak on the surface of piping. Thus, the characteristic signals of crack are weak, which can be covered by noise easily. Therefore, it is necessary to optimize the inducing frequency to balance the detection sensitivity and penetration depth [24].

$$\delta = {1 \mathord{\left/ {\vphantom {1 {\left( {\pi u_{{\text{r}}} u_{0} \sigma f} \right)}}} \right. \kern-0pt} {\left( {\pi u_{{\text{r}}} u_{0} \sigma f} \right)}}^{{{1 \mathord{\left/ {\vphantom {1 2}} \right. \kern-0pt} 2}}}$$

(1)

As mentioned above, the Bx contains the depth information of crack. Hence, the effect of inducing frequency on Bx is analyzed using the FEM model. The background magnetic field is extracted at different frequencies (50 Hz–60 kHz), as shown in Fig. 3a. The background magnetic field rises steeply as the frequency increases from 50 Hz to 20 kHz and becomes a slow climb after 20 kHz. It suggests that when the inducing frequency reaches 20 kHz, the current field saturates nearly on the surface of piping. So to get a deeper penetration depth, the inducing frequency should be lower.

Fig. 3

Frequency optimization. a Background of Bx with different frequencies. b Bx with different frequencies. c Sensitivity of Bx with different frequencies

The Bx with different inducing frequency is simulated, as shown in Fig. 3b. Initially the troughs of Bx are flat and becomes obvious after 1 kHz. The sensitivity of Bx is a critical parameter for the estimation of crack depth, which is defined as follows [25].

$$S_{Bx} = Bx_{\max } /Bx_{0}$$

(2)

where $\Delta Bx_{\max }$ is the maximum distortion of Bx, $Bx_{0}$ is the background of Bx.

A high sensitive Bx signal will greatly improve signal to noise ratio and carry more crack depth information. As shown in Fig. 3c, the sensitivity of Bx increases steeply initially and reaches top at 20 kHz and then drops at last. Above all, the simulation results suggest that 20 kHz is the appropriate inducing frequency for high sensitive detection of crack depth.

3 Circumferential Uniform Current Field Testing System

3.1 Testing System

In practice, the piping is fixed and the probe is driven by mechanical devices or robots to scan the piping. In the laboratory, for convenience the probe is fixed on the scanner and the piping is driven to pass through the probe. The circumferential uniform current field testing system includes probe, excitation module, signal processing module, acquisition card, scanner and personal computer (PC), as shown in Fig. 4. The scanner is controlled by the programmable logic controller (PLC) in control cabinet. The piping is driven by the scanner to pass through the probe at a constant speed. The displacement information of the piping is recorded by the encoder. The probe induces a uniform current field on the surface of piping by the excitation coil. When a crack is presented, the uniform current field will be disturbed. The disturbed current field perturbs the space magnetic field. The magnetic sensors on the probe pick up the distorted magnetic field which is sent to signal processing box for signal processing and acquisition. The software in the computer shows the imaging of space magnetic field visually and estimates the size of the crack.

Fig. 4

Block diagram of circumferential uniform current field test system

The probe consists of sensor module array, excitation coil and detachable nylon yoke, as shown in Fig. 5. The sensor module is made of two high-precision tunnel magneto resistance (TMR) magnetic sensors, whose operating temperature can reach 125 ℃ [26,27,28]. The two TMR sensors are sealed on each side of one common printed circuit board (PCB) and the sensitive axis of the sensors is orthogonal, which is used to measure the Bx and Bz. On the PCB, there are two primary amplifier chips (AD620) for Bx and Bz amplification (Bx 5 times and Bz 10 times). To achieve a full 360°detection of the piping surface, the sensor modules are installed on the yoke with an equal space as sensor array. In this paper, the sensor array is with 20° to each other in the circumferential direction of the piping [29]. There are 5 close TMR sensor arrays in the probe. The space magnetic field can be measured visually for imaging under the sensor array [30]. To avoid removing the coatings on the piping, the lift-off of sensor modules is 7 mm. Because the piping is closed, the yoke is separated in half symmetrically. Thus the yoke and sensors can be disassembled and installed to avoid the elbow pipes and flanges. When the two part yokes are encapsulated, the excitation coil (copper wire whose diameter is 1 mm) is wound on the yoke with 50 turns.

Fig. 5

Probe of circumferential uniform current field testing system. a TMR sensor module. b Dvided yokes and sensors. c Photo of probe

There are four main parts: power system, excitation module, signal processing module and acquisition card in the signal processing box. The power system is a rechargeable lithium ion battery which provides power for each module. The excitation model transmits sinusoidal signal with 20 kHz frequency and 2 V amplitude, as shown in Fig. 6a. The signal processing module includes second amplifying circuit (Bx 10 times and Bz 10 times), band-pass filtering (10–30 kHz) and zeroing circuit, as shown in Fig. 6b. Thus the Bx and Bz is amplified 50 times and 100 times respectively as a whole. The Bx and Bz are filtered by the band-pass filtering and then calibrated by the zeroing circuit to keep the same zero point and scale. The Bx and Bz are captured by the acquisition card and then sent to the PC. The circumferential uniform current field testing system is set up, as shown in Fig. 6c.

Fig. 6

Circumferential uniform current field testing system. a Excitation module. b Signal processing module. c Photo of testing system

4 Experiments and Discussions

4.1 Crack Depth Estimation

The samples are two stainless steel pipes (external diameter 65 mm, inner diameter 47), as shown in Fig. 7. On No.1 sample, the length of cracks is the same (30 mm) and the depth is different (2, 4, 6, 8 mm and a through crack). On No. 2 sample, the length of cracks is different (55, 50, 45, 40, 35, 30 mm) and the depth is the same (4 mm). The width of all cracks is 0.3 mm.

Fig. 7

Samples. a No. 1 sample with different depth cracks. b No. 2 sample with different length cracks

The first sample (No. 1 simple) is driven by the scanner to pass through the probe at the speed of 10 mm/s (from 2 mm depth crack to through crack). The space magnetic field Bx around the crack on the No. 1 sample is shown in PC, as shown in Fig. 8. There are five troughs in Bx, while there are five opposite peaks in Bz at the same location. The characteristic of the Bx and Bz are in accord with the simulation results. The Bx and Bz are highlighted in the background magnetic field at the lift-off of 7 mm, which helps to recognize the cracks easily without contact. Especially, the maximum distortion values of Bx troughs and Bz peaks both appear in channel 4. Thus the location of the cracks is confirmed on piping according to the position of sensors in channel 4 o in the probe. Because the background of Bz approximates to zero, the distortion of Bz is more smooth and outstanding than Bx.

Fig. 8

Space magnetic field Bx and Bz of cracks on no. 1 sample

The Bx and Bz are selected in channel 4, as shown in Fig. 9. The troughs of Bx increase as the crack depth grows and the peaks of Bz goes up at the same time. The $S_{Bx}$ with different depth cracks are shown in Fig. 10a. The maximum distortion values of Bz ($\Delta Bz$) with different depth cracks are shown in Fig. 10b.

Fig. 9

Signals Bx and Bz in channel 4 of no. 1 samples. a Bx. b Bz

Fig. 10

a $S_{Bx}$ with different depth cracks. b $\Delta Bz$ with different depth cracks

The $S_{Bx}$ and $\Delta Bz$ almost increase linearly as the crack depth grows from 2 to 8 mm. However, there is a distortion point at the through crack. This is due to the material discontinuity in the through crack and the uniform current could not flow at the bottom of the crack. The current density drops sharply in the X direction and gathers more seriously at the tips of the crack.

Because there is a so well linear relationship between $S_{Bx}$ and the crack depth (and the similar liner relationship between $\Delta Bz$ and the crack depth), the calibration method is proposed to estimate the crack depth. The first two shallower cracks are set as the calibrated crack to predict the last three cracks. According to the first two cracks, the depth of the other three cracks can be calibrated by Eqs. (3) and (4) fitted by Fig. 10a, b respectively:

$$D_{s} = 119S_{Bx} - 0.2432$$

(3)

$$D_{\Delta } = 0.070098\Delta Bz + 0.8661$$

(4)

where $D_{s}$ is the crack depth estimated by $S_{Bx}$, $D_{\Delta }$ is the crack depth estimated by $\Delta Bz$.

Thus the last three cracks can be estimated by Eqs. (3) and (4). The predicted depths (PD) and relative errors are shown in Table 1. The 6 mm depth crack can be estimated by $S_{Bx}$ and $\Delta Bz$. The relative errors (RE) are 13.0% and 2.6% respectively. The 8 mm depth and through crack are regarded as a break according the estimation by $S_{Bx}$. The 8 mm depth is estimated by $\Delta Bz$ and the relative error is 8.6%. The through crack is identified as a break by $S_{Bx}$ and $\Delta Bz$. If the first three cracks are set as calibrated crack, the relative errors of 8 mm depth crack are both less than 5% by $S_{Bx}$ and $\Delta Bz$. We can make a conclusion that the crack depth can be estimated by $S_{Bx}$ and $\Delta Bz$ using the calibration method. Thus the residual thickness of power plant piping can be evaluated by periodic detection using circumferential uniform current field testing system.

Table 1 Estimated results of crack depth

4.2 Crack Length Estimation

The second sample (No. 2 sample) is driven to pass through the probe at the same speed (from 55 mm length crack to 30 mm length crack). As shown in Fig. 11, the space magnetic field around the cracks on the No. 2 sample is plotted. There are 6 troughs in Bx and 6 peaks in Bz at the same location. Because the depth of the cracks on No.2 sample is the same, there are no obvious changes in the troughs of Bx and the peaks of Bz. Similarly, the troughs and peaks are located at channel 4.

Fig. 11

Testing results of no. 2 sample

The Bx and Bz are selected in channel 4 and the encoder records the displacement information in the X coordinate, as shown in Fig. 12. Because the peaks of Bz locate at the tips of crack, the distance between the two opposite peaks of Bz ($\Delta L$) reflects the length of cracks. As shown in Fig. 12b, the $\Delta L$ becomes narrow as the crack length decreases. However, the current field deflects clockwise at one tip of the crack and deflects anticlockwise at the other tip of the crack. The maximum magnetic field locates in the center of the deflected current field, which makes the $\Delta L$ less than the crack length.

Fig. 12

Signals Bx and Bz in channel 4 of no. 2 samples

The $\Delta L$ with different length cracks is shown in Fig. 13. $\Delta L$ goes up linearly as the crack length grows. $\Delta L$ is less than the actual length of the crack and the relative errors become unacceptable for short crack, as shown in Table 2. The calibration method is also proposed to estimate the length of the cracks. The first two cracks are set as calibrated crack to predict the last four cracks. The calibration equation of crack length (L) is given by Eq. (5) fitted by Fig. 13.

$$L = 1.042\Delta L + 2.917$$

(5)

Fig. 13

$\Delta L$ with different length cracks

Table 2 Testing results of crack length

As shown in Table 2, the relative errors of the crack length drop sharply by the calibration equation. The calibrated crack length equals the actual length of the crack except the 45 mm length crack which has a tiny relative error. Thus the length of the crack on piping can be estimated by the distance between the peaks of Bz using the calibration method. What’s more, the development of the crack length on power plant piping can be monitored and warned by periodic detection and estimation.

5 Conclusion and Further Work

This paper presents a novel circumferential uniform current field testing system for surface cracks detection and estimation on power plant piping. The dynamic FEM model is developed to extract the characteristic signals of the crack. The inducing frequency is optimized by the FEM model for sensitive and accurate detection of the crack. In the end, the circumferential uniform current field testing system is set up and the crack detection experiments are carried out. The results show that the cracks on piping can be detected and estimated visually and efficiently without contact by the circumferential uniform current field testing system with TMR sensor array using calibration method in a one pass scan. The depth of the crack can be estimated by the sensitivity of Bx and the maximum distortion of Bz. Meanwhile, the length of the crack can be evaluated by the distance between the opposite peaks of Bz. What’s more, the circumferential uniform current field testing system provides a new method for non-contact, visual and efficient detection and estimation of cracks on power plant piping as an alternative technique to MT and PT. Further work will focus on monitoring the development of the crack using circumferential uniform current field testing system. Thus the propagation of the crack and the residual wall thickness can also be monitored and estimated.