Pulsed eddy current testing method for pipeline wall thickness under large lift-off based on new signal processing

Abstract

A pulse eddy current detection method for pipeline wall thickness under large lift-off based on new signal processing consists of two parts: a filtering method based on independent component analysis and Gaussian denoising and a feature adaptive extraction method based on Hough transform. When implementing the method, independent component analysis is first performed on the pulse eddy current detection signal to separate the effective component from the power frequency noise and eliminate the typical power frequency interference component in the pulse eddy current detection signal. Then, Gaussian filtering is used to eliminate the random noise of the pulse eddy current signal after filtering out the power frequency interference. Subsequently, Hough transform is used to adaptively extract the straight line segment of the later stage of the pulse eddy current detection signal after filtering, and the wall thickness angle corresponding to the straight line is used as a feature quantity to characterize the pipeline wall thickness. Finally, the feature and the corresponding calibration curve are calculated to obtain the pipeline wall thickness value at the measurement point. The method of the present invention proposes a new adaptive feature quantity and can greatly improve the signal-to-noise ratio of the detection signal, and has high practical engineering application value.

Description

Technical Field

[0001] The present invention relates to a pulsed eddy current detection method for pipeline wall thickness under large lift-off based on signal processing, and in particular to a filtering method based on independent component analysis and Gaussian denoising and a feature extraction method based on Hough transform. Background Art

[0002] Coated pipes are widely used in industries such as oil, natural gas, and nuclear power, playing an important role in transporting oil, natural gas, and high-temperature steam. However, due to the corrosive nature of the transported medium and the high-temperature, high-pressure service environment, the pipe wall is often corroded, resulting in a reduction in wall thickness. When the pipe wall thickness falls below the safety threshold, the pipe may rupture, causing a major production safety accident. Therefore, high-precision non-destructive measurement of pipe wall thickness is of great practical significance. When using traditional non-destructive testing methods to measure pipe wall thickness, it is often necessary to first remove the outer coating before testing. This significantly increases the economic and time costs of testing. More seriously, the downtime caused by the extended testing cycle will result in huge economic losses for the owner. Therefore, it is crucial to develop advanced non-destructive testing methods and technologies that do not require the removal of the coating.

[0003] Since the 1970s, pulsed eddy current testing (PEC) technology has been a popular research method in the field of nondestructive testing. Due to its advantages, such as non-contact testing, it has received considerable attention in recent years for the inspection of coated pipes. When inspecting coated pipes, PEC requires the probe to be placed at a significant elevation above the pipe. However, when the probe is at a significant elevation, the PEC signal becomes very weak and is easily affected by noise. Therefore, developing novel signal processing methods to improve the signal-to-noise ratio (SNR) of PEC signals is crucial. Furthermore, feature extraction of PEC signals is affected by the pipe wall thickness, which is unknown. Therefore, adaptive extraction of PEC signal features is difficult. Therefore, developing novel adaptive signal feature extraction methods and signal feature quantities is crucial for the application of PEC testing methods for pipe wall thickness inspection at significant elevations. Summary of the Invention

[0004] In order to achieve the goal of quickly and accurately detecting the pipe wall thickness under large lift-off, the present invention provides a pulse eddy current detection method for pipe wall thickness under large lift-off based on a new signal processing method. The method is a signal processing method based on independent component analysis and Gaussian filtering and a pulse eddy current detection signal adaptive feature extraction method based on Hough transform. The method consists of an independent component analysis method, Gaussian filtering and Hough transform. When implementing the method, a pulse eddy current detection probe is placed at the pipe to be detected, and appropriate excitation parameters are set for the pulse eddy current detection experimental device to obtain a corresponding original detection signal under certain lift-off conditions. Then, the independent component analysis method is used to separate the power frequency interference in the original signal. After the separation is completed, the Gaussian white noise in the detection signal is suppressed by a one-dimensional Gaussian filter. Finally, the Hough transform is used to adaptively extract the characteristic quantity of the pulse eddy current detection signal, and the corresponding pipe wall thickness is calculated in combination with the calibration curve. The method of the present invention can filter out power frequency interference and Gaussian noise in the pulsed eddy current detection signal, greatly improve the signal-to-noise ratio of the detection signal, and adaptively extract the characteristic quantity of the signal, which provides a fast and accurate method for detecting the wall thickness of the pipeline under large lifting distance, and has high practical engineering application value and good application and development prospects.

[0005] In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0006] A pulsed eddy current testing method for pipeline wall thickness under large lift-off based on novel signal processing includes the following steps:

[0007] Step 1: Build a pulsed eddy current testing experimental device, which includes a pulsed eddy current detection probe, a high-power excitation source and a differential amplifier connected to the pulsed eddy current detection probe, and a signal acquisition card connected to the differential amplifier; and place the pulsed eddy current detection probe at a preset lift-off position above the pipeline to be tested. Set the corresponding excitation signal excitation frequency (0.5-16 Hz) and excitation amplitude (10-20 V) for the high-power excitation source according to the designed wall thickness of the pipeline, and apply the excitation signal to the excitation coil of the pulsed eddy current detection probe. Then, the detection coil of the pulsed eddy current detection probe will capture the corresponding detection signal. After the detection signal is amplified by the differential amplifier and quantized by the signal acquisition card, the original pulsed eddy current detection signal u0(t) corresponding to the pipeline to be tested is obtained;

[0008] Step 2: Use the orthogonal sine basis and its higher harmonics to construct the reference power frequency interference signal n(t), which is expressed as shown in formula (1).

[0009]

[0010] Where f represents the fundamental frequency of the power frequency interference, t represents time, and k represents the highest order of the power frequency interference harmonics. k is determined by Fourier transform of the original pulsed eddy current detection signal u0(t);

[0011] Step 3: Use the original pulsed eddy current detection signal u0(t) and the reference power frequency interference signal n(t) to construct an independent component analysis model, which is expressed as shown in formula (2).

[0012]

[0013] Where: s represents the independent component matrix, which contains k+1 components: s1(t) s2(t) … s k+1 (t), A represents the mixing matrix of independent components;

[0014] Step 4: Use the fast independent component analysis method to solve equation (2) to obtain the independent component matrix s, and calculate the correlation coefficient R between each component of the independent component matrix s and the original pulsed eddy current detection signal u0(t) i , its expression is shown in formula (3),

[0015]

[0016] Where cov() represents covariance and var() represents variance; next, find the subscript i corresponding to the largest correlation coefficient. max , s imax (t) is the pulsed eddy current detection signal u1(t) after eliminating the power frequency interference;

[0017] Step 5: Filter u1(t) using the Gaussian filtering method to obtain the final noise-filtered pulsed eddy current detection signal u(t). The expression for Gaussian filtering is shown in formula (4).

[0018]

[0019] In the formula Represents convolution operation; g represents a Gaussian convolution kernel with a mean of zero;

[0020] Step 6: Take the common logarithm of u(t) and use Hough transform to transform each sampling point (t i ,lg[u i ]) is transformed into Hough space, where the subscript i represents the sampling point number. The transformation expression is shown in formula (5).

[0021] ρ ij =t i cos(θ j )+lg[u i ]sin(θ j) i=1,2,…,mj=1,2,…,n (5)

[0022] Where θ j represents the angle between the perpendicular line of any straight line passing through the sampling point i and the time axis, ρ ij Indicates that the sampling point (t i ,lg[u i ]) and the angle between its vertical line and the time axis is θ j The vertical distance between the straight line and the origin of the coordinate system, m represents the number of signal sampling points, and n represents θ j The number of

[0023] The Hough space is constructed with θ and ρ as the horizontal and vertical coordinates respectively. After transformation, m×n points in the Hough space are obtained. The Hough space is then divided into several sub-regions. The number of points in each sub-region is counted, and the horizontal coordinate θ0 of the center point of the sub-region with the largest number of points is extracted as the characteristic value of the pulsed eddy current detection signal at the pipeline to be tested. This characteristic value is related to the wall thickness of the pipeline, so it is called the wall thickness angle.

[0024] Step 7: Prepare pipe calibration specimens with different wall thicknesses. Use the calibration curve d=f(θ) measured on the calibration specimen in steps 1 to 6 above, where d represents the pipe wall thickness and θ represents the adaptive characteristic value wall thickness angle. Substitute the characteristic value θ0 extracted at the pipe test point into the calibration curve d=f(θ). After calculation, the wall thickness value at the test point is obtained.

[0025] Preferably, in step 6, θ j The value is a series of discrete values. Since the pulsed eddy current signal shows a downward trend in the late stage, θ j The value range of is 1° to 89°, and the sampling point step is adjusted according to the required accuracy.

[0026] Preferably, in step 5, the length of g is adjusted according to the excitation frequency and the sampling frequency, and the variance is adjusted according to the required filtering effect.

[0027] Compared with the prior art, the advantages of the present invention are as follows:

[0028] 1) By combining a power frequency interference elimination method based on independent component analysis with a random noise elimination method based on Gaussian filtering, the method of the present invention can simultaneously remove power frequency interference and random noise from the pulsed eddy current detection signal, greatly improving the signal-to-noise ratio of the detection signal. In addition, thanks to the excellent suppression effect of Gaussian filtering on random noise, the method of the present invention does not require average filtering preprocessing compared with traditional filtering methods, thereby significantly shortening the detection time and significantly improving the detection efficiency;

[0029] 2) Since the Hough transform converts the straight line identification problem in the original space (t, u) into a statistical peak problem in the Hough space (θ, ρ), the method of the present invention does not require relevant prior knowledge of the pulsed eddy current detection signal waveform and is insensitive to the noise in the detection signal. It is an adaptive and stable feature extraction method. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a schematic diagram and a flow chart of each step when performing wall thickness detection at large lift-off according to the present invention.

[0031] Figure 2a and Figure 2b They are schematic diagrams of the Hough transform principle when a point in the original space and several points in the original space are collinear and used in the present invention.

[0032] Figure 3 This is a diagram showing the effect of the present invention on the detection of ferromagnetic step tube wall thickness under large lift-off. DETAILED DESCRIPTION

[0033] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0034] The primary sources of noise in pulsed eddy current (PEC) testing signals are power frequency interference and normally distributed Gaussian white noise. The power frequency interference has a much greater amplitude than the Gaussian white noise, and the distributions of these two types of noise are independent of the PEC signal. Based on this fact, independent component analysis (ICA) can be developed to separate the power frequency interference from the noisy PEC signal. Then, Gaussian filtering can be applied to the separated signal, potentially significantly improving the signal-to-noise ratio (SNR) and detection accuracy.

[0035] When performing nondestructive measurement of pipe wall thickness at high lift, traditional methods extract the slope of the late, straight-line portion of the pulsed eddy current (PEC) signal. However, the start time of this straight-line portion is affected by the pipe wall thickness and cannot be adaptively extracted, which directly affects the accuracy and efficiency of traditional wall thickness measurement methods. The present invention develops a Hough transform algorithm to overcome this difficulty, adaptively identifying this straight-line portion and extracting the corresponding feature value, enabling high-precision and high-efficiency PEC testing of pipe wall thickness at high lift.

[0036] For Figure 1 The specific implementation steps of the method of the present invention are as follows: first, Figure 1As shown in the figure, a square wave excitation signal is generated by a high-power excitation source and applied to the excitation coil of the pulsed eddy current detection probe. Then, the detection coil of the pulsed eddy current detection probe will capture the corresponding detection signal. The detection signal is amplified by a differential amplifier and quantized by a signal acquisition card to obtain the original pulsed eddy current detection signal. Then, based on independent component analysis and one-dimensional Gaussian filtering, the power frequency interference and Gaussian white noise in the original detection signal are filtered out respectively, the signal-to-noise ratio of the original signal is improved, and a smoother pulsed eddy current detection signal is obtained. Finally, as shown in the figure, Figure 1 As shown, using Hough transform (its principle diagram is as follows Figure 2a and Figure 2b (As shown in the figure), the method adaptively identifies the late straight line segment of the pulsed eddy current test signal and extracts the angle between the perpendicular line through the origin and the time axis as a characteristic value. After calculation in combination with the corresponding calibration curve, the wall thickness value at the test point is obtained. The method of the present invention can filter out power frequency interference and Gaussian noise in the pulsed eddy current test signal, significantly improving the signal-to-noise ratio of the test signal and adaptively extracting the signal's characteristic value. This provides a fast and accurate method for detecting pipe wall thickness under large lift-off conditions and has promising application and development prospects.

[0037] The following combination Figure 1 , Figure 2 and Figure 3 The steps and effects of the present invention are further described in detail with reference to the accompanying drawings and specific embodiments.

[0038] A pulsed eddy current detection method for pipeline wall thickness under large lift-off based on novel signal processing comprises the following steps:

[0039] Step 1: If Figure 1 As shown, a pulsed eddy current detection experimental device is built, including a high-power signal source that can generate arbitrary waveforms, a low-noise differential amplifier and a high-precision signal acquisition card, a pulsed eddy current detection probe, including an excitation coil placed on the inside and a detection coil placed on the outside, the high-power excitation source and the differential amplifier are connected to the pulsed eddy current detection probe, the signal acquisition card is connected to the differential amplifier, and the pulsed eddy current detection probe is placed above the pipeline to be detected. According to the designed wall thickness of the pipeline, the pulsed high-power excitation source is set to a square wave excitation signal with a corresponding excitation frequency (0.5-16 Hz) and excitation amplitude (10-20 V), and the square wave excitation signal is applied to the excitation coil of the pulsed eddy current detection probe. Then, the detection coil of the pulsed eddy current detection probe will capture the corresponding detection signal. After the detection signal is amplified by the differential amplifier and quantized by the signal acquisition card, the original pulsed eddy current detection signal u0(t) corresponding to the pipeline to be detected is obtained;

[0040] Step 2: Use the orthogonal sine basis and its higher harmonics to construct the reference power frequency interference signal n(t), which is expressed as shown in formula (1).

[0041]

[0042] Where f represents the fundamental frequency of the power frequency interference, which is generally 50 Hz, t represents time, and k represents the highest order of the power frequency interference harmonics, which is determined by Fourier transform of the original pulsed eddy current detection signal u0(t);

[0043] Step 3: Based on the characteristic that the independent component analysis method can unmix several mixed independent signals, the independent component analysis model is constructed using the original pulsed eddy current detection signal u0(t) and the reference power frequency interference signal n(t). Its expression is shown in formula (2).

[0044]

[0045] Where: s represents the independent component matrix, which contains k+1 components: s1(t) s2(t) … s k+1 (t), A represents the mixing matrix of independent components;

[0046] Step 4: Use the fast independent component analysis method to solve equation (2) to obtain the independent component matrix s, and calculate the correlation coefficient R between each component of the independent component matrix s and the original pulsed eddy current detection signal u0(t) i , its expression is shown in formula (3),

[0047]

[0048] Where cov() represents covariance and var() represents variance; next, find the subscript i corresponding to the largest correlation coefficient. max , That is the pulsed eddy current detection signal u1(t) after eliminating the power frequency interference;

[0049] Step 5: Filter u1(t) using the Gaussian filtering method to obtain the final noise-filtered pulsed eddy current detection signal u(t). The expression for Gaussian filtering is shown in formula (4).

[0050]

[0051] In the formula Represents the convolution operation, g represents the Gaussian convolution kernel with zero mean, its length should be adjusted according to the excitation frequency and sampling frequency, and its variance should be adjusted according to the required filtering effect;

[0052] Step 6: Figure 2a As shown, each point in the original space corresponds to a sine curve in the Hough space after the Hough transform. Further, as Figure 2bAs shown in the figure, when several points in the original space are collinear, the curves transformed from the points in the original space in the corresponding Hough space will intersect at one point. It is only necessary to find this intersection point to identify the straight line segment in the signal. Therefore, this step first takes the common logarithm of u(t) and uses the Hough transform to transform each sampling point (t i ,lg[u i ]) is transformed into Hough space, where the subscript i represents the sampling point number. The transformation expression is shown in formula (5).

[0053] ρ ij =t i cos(θ j )+lg[u i ]sin(θ j ) i=1,2,…,mj=1,2,…,n (5)

[0054] Where θ j It represents the angle between the vertical line of any straight line passing through the sampling point and the time axis, which takes a series of discrete values. Since the pulsed eddy current signal shows a downward trend in the late stage, θ j The value range is 1° to 89°, and the sampling point step is adjusted according to the required accuracy. ij Indicates that the sampling point (t i ,lg[u i ]) and the angle between its vertical line and the time axis is θ j The vertical distance between the straight line and the origin of the coordinate system, m represents the number of signal sampling points, and n represents θ j The number of

[0055] The Hough space is constructed with θ and ρ as the horizontal and vertical coordinates respectively. After the above transformation, m×n points in the Hough space are obtained. Next, the Hough space is divided into several sub-regions. The number of points in each sub-region is counted, and the horizontal coordinate θ0 of the center point of the sub-region with the largest number of points is extracted as the characteristic value of the pulsed eddy current detection signal at the pipeline to be tested. This characteristic value is related to the wall thickness of the pipeline, so it is called the wall thickness angle.

[0056] Step 7: Prepare pipe calibration specimens with different wall thicknesses. Use the calibration curve d=f(θ) measured on the calibration specimen in steps 1 to 6 above, where d represents the pipe wall thickness and θ represents the adaptive characteristic wall thickness angle. Substitute the characteristic value θ0 extracted at the pipe test point into the calibration curve d=f(θ). After calculation, the wall thickness value at the test point can be obtained.

[0057] In the embodiment, the method of the present invention is applied to a 160 cm long ferromagnetic stepped tube, which comprises four wall thickness steps, each step length is 40 cm, and the wall thicknesses are 25 mm, 15 mm, 10 mm and 5 mm, respectively, that is, the residual wall thicknesses are 100%, 60%, 40% and 20%, respectively. Figure 3 The pulsed eddy current detection probe is placed above the pipe at a height of 40 mm and scanned with a step length of 5 cm. The method of the present invention is applied to the detection signal collected at each sampling point, and the residual wall thickness value obtained by detection is shown as follows: Figure 3 As shown in the black dotted line graph in Figure 1, the detected residual wall thickness is consistent with the actual residual wall thickness. Therefore, the method of the present invention can effectively and accurately detect the wall thickness of pipes under large lifting conditions and has good application and development prospects.

Claims

1. A pulsed eddy current testing method for pipeline wall thickness under large lift-off conditions based on novel signal processing, characterized by: The steps include: Step 1: Build a pulsed eddy current testing experimental device, which includes a pulsed eddy current detection probe, a high-power excitation source and a differential amplifier connected to the pulsed eddy current detection probe, and a signal acquisition card connected to the differential amplifier; and place the pulsed eddy current detection probe at a preset lift-off position above the pipeline to be tested. Set the corresponding excitation signal excitation frequency and excitation amplitude for the high-power excitation source according to the designed wall thickness of the pipeline, and apply the excitation signal to the excitation coil of the pulsed eddy current detection probe. Then, the detection coil of the pulsed eddy current detection probe will capture the corresponding detection signal. After the detection signal is amplified by the differential amplifier and quantized by the signal acquisition card, the original pulsed eddy current detection signal u0(t) corresponding to the pipeline to be tested is obtained; Step 2: Use the orthogonal sine basis and its higher harmonics to construct the reference power frequency interference signal n(t), which is expressed as shown in formula (1). Where f represents the fundamental frequency of the power frequency interference, t represents time, and k represents the highest order of the power frequency interference harmonics. k is determined by Fourier transform of the original pulsed eddy current detection signal u0(t); Step 3: Use the original pulsed eddy current detection signal u0(t) and the reference power frequency interference signal n(t) to construct an independent component analysis model, which is expressed as shown in formula (2). Where: s represents the independent component matrix, which contains k+1 components: s1(t) s2(t)…s k+1 (t), A represents the mixing matrix of independent components; Step 4: Use the fast independent component analysis method to solve equation (2) to obtain the independent component matrix s, and calculate the correlation coefficient R between each component of the independent component matrix s and the original pulsed eddy current detection signal u0(t) i , its expression is shown in formula (3), Where cov() represents covariance and var() represents variance; next, find the subscript i corresponding to the largest correlation coefficient. max , s imax (t) is the pulsed eddy current detection signal u1(t) after eliminating the power frequency interference; Step 5: Filter u1(t) using the Gaussian filtering method to obtain the final noise-filtered pulsed eddy current detection signal u(t). The expression for Gaussian filtering is shown in formula (4). In the formula Represents convolution operation; g represents a Gaussian convolution kernel with a mean of zero; Step 6: Take the common logarithm of u(t) and use Hough transform to transform each sampling point (t i ,lg[u i ]) is transformed into Hough space, where the subscript i represents the sampling point number. The transformation expression is shown in formula (5). r ij =t i cos(θ j )+lg[u i ]sin(θ j ) i=1,2,…,mj=1,2,…,n (5) Where θ j represents the angle between the perpendicular line of any straight line passing through the sampling point i and the time axis, ρ ij Indicates that the sampling point (t i ,lg[u i ]) and the angle between its vertical line and the time axis is θ j The vertical distance between the straight line and the coordinate origin, m represents the number of signal sampling points, n represents θ j The number of The Hough space is constructed with θ and ρ as the horizontal and vertical coordinates respectively. After transformation, m×n points in the Hough space are obtained. The Hough space is then divided into several sub-regions. The number of points in each sub-region is counted, and the horizontal coordinate θ0 of the center point of the sub-region with the largest number of points is extracted as the characteristic value of the pulsed eddy current detection signal at the pipeline to be tested. This characteristic value is related to the wall thickness of the pipeline, so it is called the wall thickness angle. Step 7: Prepare pipe calibration specimens with different wall thicknesses. Use the calibration curve d=f(θ) measured on the calibration specimen in steps 1 to 6 above, where d represents the pipe wall thickness and θ represents the adaptive characteristic value wall thickness angle. Substitute the characteristic value θ0 extracted at the pipe test point into the calibration curve d=f(θ). After calculation, the wall thickness value at the test point is obtained.

2. The pulsed eddy current detection method for pipeline wall thickness under large lift-off based on novel signal processing according to claim 1 is characterized in that: In step 6, θ j The value is a series of discrete values. Since the pulsed eddy current signal shows a downward trend in the late stage, θ j The value range of is 1° to 89°, and the sampling point step is adjusted according to the required accuracy.

3. The pulsed eddy current testing method for pipeline wall thickness under large lift-off based on novel signal processing according to claim 1 is characterized in that: In step 5, the length of g is adjusted according to the excitation frequency and the sampling frequency, and the variance is adjusted according to the required filtering effect.

4. The pulsed eddy current testing method for pipeline wall thickness under large lift-off based on novel signal processing according to claim 1 is characterized in that: In step 1, the excitation signal excitation frequency set for the high-power excitation source is 0.5-16 Hz and the excitation amplitude is 10-20 V according to the designed wall thickness of the pipeline.

Images