A Method and System for Inverting Cable Alloy Defects Based on Multi-Frequency Eddy Current Impedance Spectroscopy

By generating virtual depth axes and depth domain impedance spectra, and combining standard basis functions and mapping parameters, the problem of depth aliasing of multi-frequency eddy current signals was solved, achieving high-precision quantitative inversion of cable alloy defects and improving the reliability of defect identification and the accuracy of quantitative evaluation.

CN123112232APending Publication Date: 2026-11-17江苏锡洲新材料科技有限公司
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202611621416.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-20
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies have limitations in quantitative inversion accuracy due to deep aliasing of multi-frequency eddy current signals leading to defects in cable alloys. In particular, they have limitations in signal decoupling, which makes the response of deep defects susceptible to interference from surface effects, and the quantitative inversion accuracy needs to be improved.

Method used

By acquiring initial multi-frequency eddy current detection data, calculating the phase change rate parameter, generating a virtual depth axis, mapping the electromagnetic response amplitude and impedance characteristic phase, constructing a depth domain impedance spectrum, matching the target standard basis function using a standard defect feature library, and solving the linear superposition equation by combining the amplitude weighting coefficient and spatial offset, a quantitative defect evaluation map is generated.

Benefits of technology

It achieves spatial separation of electromagnetic responses at different penetration depths on the virtual depth axis, suppresses background interference, improves the reliability of physical defect type identification and decoupling of defect size and depth, and enhances the accuracy of quantitative inversion of cable alloy defects.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123112232A_ABST
    Figure CN123112232A_ABST
Patent Text Reader

Abstract

This invention belongs to the technical field of nondestructive testing and relates to a method and system for inverting cable alloy defects based on multi-frequency eddy current impedance spectroscopy. The method includes the following steps: extracting initial multi-frequency eddy current detection data, calculating the phase change rate parameter, and performing spatial projection on the detection frequency to generate a virtual depth axis; mapping the impedance data to this axis to generate a depth-domain impedance spectrum, and matching it with a standard defect feature library to determine the target standard basis function; using this basis function and the impedance spectrum to construct and solve a linear superposition equation to obtain the mapping parameters; and decoding using a calibration lookup table to generate a quantitative defect evaluation map. This invention solves the problem in the prior art where the quantitative inversion accuracy of cable alloy defects is limited due to the deep aliasing of multi-frequency eddy current signals.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the technical field of nondestructive testing and relates to a method and system for inverting defects in cable alloys based on multi-frequency eddy current impedance spectroscopy. Background Technology

[0002] Eddy current nondestructive testing (EDT) technology plays a crucial role in ensuring the safety of conductive components such as cables in fields such as power systems. This technology enables the exploration of the interior of materials by monitoring alternating eddy current field disturbances caused by internal defects. Currently, eddy current testing technology is gradually evolving from conventional qualitative defect detection to quantitative inversion of parameters such as defect type, size, and burial depth.

[0003] For the assessment of cable defects, patent publication number CN122109212A discloses a non-destructive assessment method and system for cable microstructure based on AC impedance spectroscopy. This method measures the complex impedance values ​​at characteristic frequency points, uses fitting inversion to calculate equivalent circuit parameters, and constructs a mapping relationship to calculate indices reflecting microstructure defects, thereby determining the assessment level of the cable's microstructure condition.

[0004] While the aforementioned scheme has achieved some success in assessing microscopic conditions, it has limitations in decoupling signals at depth when handling quantitative inversion of macroscopic defects in cables. This scheme relies on the impedance response at discrete frequency points, but in multi-frequency eddy current detection, the skin effect causes a non-linear relationship between frequency and electromagnetic penetration depth. This easily leads to aliasing and superposition of defect response signals at different depths, making the response of deep defects susceptible to surface effects. Features extracted from the original frequency domain often contain coupled information from multiple physical factors, resulting in room for improvement in the accuracy of quantitative inversion.

[0005] Based on the above problems, the present invention aims to solve the technical problem that the quantitative inversion accuracy of cable alloy defects is limited due to the deep aliasing of multi-frequency eddy current signals in the prior art. Summary of the Invention

[0006] In a first aspect, the present invention provides a method for inverting defects in cable alloys based on multi-frequency eddy current impedance spectra, comprising the following steps: S1. Obtain the initial detection data of multi-frequency eddy current, extract the resistance component and reactance component from the initial detection data of multi-frequency eddy current, and calculate and generate the phase change rate parameter. S2. Obtain the preset discrete detection frequency sequence, and perform spatial projection operation on the discrete detection frequency sequence using the phase change rate parameter to generate a virtual depth axis; S3. Extract the electromagnetic response amplitude and impedance characteristic phase from the initial multi-frequency eddy current detection data, and map the electromagnetic response amplitude and impedance characteristic phase to the virtual depth axis to generate a depth domain impedance spectrum. S4. Obtain the preset standard defect feature library, match the depth domain impedance spectrum with the standard feature basis functions in the standard defect feature library, and determine the target standard basis function; S5. Using the target standard basis function, the amplitude weighting coefficient and spatial offset to be solved are introduced as unknown parameters. Together with the depth domain impedance spectrum, a linear superposition equation is constructed. The linear superposition equation is solved to obtain the converged amplitude weighting coefficient and spatial offset, which are then combined as mapping parameters. S6. Obtain a preset calibration reference table, and use the calibration reference table to perform quantitative inversion decoding on the mapping parameters to generate a quantitative defect evaluation map.

[0007] A further aspect of the present invention, step S1, includes: The impedance phase angle variable is obtained by calculating the arctangent dual-parameter function based on the resistance and reactance components. Perform phase continuity processing on the impedance phase angle variable to obtain the corrected impedance phase angle variable; Obtain the original detection frequency axis corresponding to the initial detection data of the multi-frequency eddy current, and calculate the discrete rate of change of the corrected impedance phase angle variable along the original detection frequency axis using the center difference method; Perform cubic spline interpolation on the discrete rate of change to generate a continuously varying phase rate of change parameter.

[0008] A further aspect of the present invention, step S2, includes: By comparing the amplitude distribution pattern of the phase change rate parameter, a preset phase change rate threshold is extracted. The frequency bands with an absolute value of the phase change rate parameter greater than the preset phase change rate threshold are identified as information-dense frequency bands, and the frequency bands with an absolute value of the phase change rate parameter less than or equal to the preset phase change rate threshold are identified as information-sparse frequency bands. Obtain the preset spatial coordinate stretching coefficient and waveform compression coefficient, configure the spatial coordinate stretching coefficient in the information-dense frequency band, configure the waveform compression coefficient in the information-sparse frequency band, and construct an adaptive mapping function. The discrete detection frequency sequence is imported into an adaptive mapping function for integration to calculate and obtain virtual depth nodes. Arrange all virtual depth nodes according to an increasing numerical rule to generate a virtual depth axis.

[0009] A further aspect of the present invention, step S3, includes: Extract the virtual depth nodes contained within the virtual depth axis; Based on the virtual depth nodes, the electromagnetic response amplitude is re-indexed and the impedance characteristic phase is integrated to construct a reference data sequence; The baseline data sequence is processed by a one-dimensional high-pass filtering algorithm to extract effective feature waveforms. The effective characteristic waveform is combined and mapped with the virtual depth axis to generate a depth domain impedance spectrum.

[0010] A further aspect of this invention involves processing a reference data sequence using a one-dimensional high-pass filtering algorithm to extract effective feature waveforms, including: Get the preset moving average window half-width parameter; Based on the moving average window half-width parameter, a local moving average calculation is performed on the benchmark data sequence to obtain the local baseline value; The effective amplitude data is obtained by subtracting the electromagnetic response amplitude from the reference data sequence and the local baseline value. Summarize all valid amplitude data to generate valid characteristic waveforms.

[0011] A further aspect of the present invention, step S4, includes: Perform isolated peak detection on the depth domain impedance spectrum and count the extreme values ​​of the number of isolated peaks within the depth domain impedance spectrum; Based on the extreme value of the number of isolated peaks, standard feature basis functions with mismatched topology in the standard defect feature library are eliminated, and candidate standard feature basis functions are obtained to form a candidate basis function set. Calculate the normalized cross-correlation coefficient between each standard feature basis function in the candidate basis function set and the depth domain impedance spectrum to obtain the similarity matching score; The similarity matching scores corresponding to each standard feature basis function are sorted in descending order, and the standard feature basis function with the highest score is determined as the target standard basis function.

[0012] A further aspect of the present invention, step S5, includes: Initialize the amplitude weighting coefficient and spatial offset based on the peak ratio of the depth domain impedance spectrum to the target standard basis function and the depth coordinates; Substitute the initial magnitude weighting coefficients, spatial offset, target standard basis function, and depth domain impedance spectrum into the linear superposition equation to construct an overdetermined system of equations. The overdetermined system of equations is iteratively solved using a nonlinear least squares approximation regression fitting algorithm, and the fitting residuals are calculated. The iteration is interrupted when the fitting residual is less than the preset lower limit of error. The converged amplitude weight coefficient and spatial offset are then exported and combined to generate mapping parameters.

[0013] A further aspect of the present invention, step S6, includes: Extract the defect classification labels carried by the target standard basis function and match them with the calibration reference table to obtain the physical defect types; Extract the amplitude weighting coefficient from the mapping parameters, and obtain the size conversion coefficient from the calibration reference table in combination with the type of physical defect; The absolute geometric dimensions are calculated by multiplying the amplitude weighting coefficient and the size conversion coefficient. Extract the spatial offset from the mapping parameters and convert the spatial offset into the actual burial depth; By summarizing the types of physical defects, their absolute geometric dimensions, and their actual burial depth, a quantitative defect evaluation map is generated.

[0014] A further aspect of the present invention involves acquiring initial detection data for multi-frequency eddy currents, including: Control the movement of the eddy current non-destructive testing probe on the surface of the test area of ​​the target cable alloy while maintaining a preset lift-off distance; The excitation coil inside the eddy current nondestructive testing probe outputs a sinusoidal excitation current corresponding to a preset discrete detection frequency sequence. The induced voltage signal and the excitation current reference signal are acquired by the detection coil inside the eddy current non-destructive testing probe. Digital quadrature demodulation is performed on the induced voltage signal and the excitation current reference signal to generate initial detection data for multi-frequency eddy currents.

[0015] Secondly, the present invention provides a cable alloy defect inversion system based on multi-frequency eddy current impedance spectrum, comprising the following modules: The phase change rate parameter generation module acquires the initial detection data of multi-frequency eddy currents, extracts the resistance and reactance components from the initial detection data of multi-frequency eddy currents, and calculates and generates the phase change rate parameter. The virtual depth axis construction module obtains a preset discrete detection frequency sequence, performs a spatial projection operation on the discrete detection frequency sequence using the phase change rate parameter, and generates a virtual depth axis. The impedance spectrum reconstruction module extracts the electromagnetic response amplitude and impedance characteristic phase from the initial multi-frequency eddy current detection data, maps the electromagnetic response amplitude and impedance characteristic phase to the virtual depth axis, and generates a depth domain impedance spectrum. The target standard basis function screening module obtains a preset standard defect feature library, matches the depth domain impedance spectrum with the standard feature basis functions in the standard defect feature library, and determines the target standard basis function. The parameter solving module uses the target standard basis function and introduces the amplitude weighting coefficient and spatial offset to be solved as unknown parameters. Together with the depth domain impedance spectrum, it constructs a linear superposition equation. The linear superposition equation is solved to obtain the converged amplitude weighting coefficient and spatial offset, which are then combined as mapping parameters. The quantitative inversion decoding module obtains a preset calibration reference table, uses the calibration reference table to perform quantitative inversion decoding on the mapping parameters, and generates a quantitative defect evaluation map.

[0016] In summary, the present invention has the following beneficial technical effects: 1. An adaptive mapping function is constructed by calculating the phase change rate parameter to generate a virtual depth axis, and the discrete detection frequency sequence is subjected to nonlinear spatial projection calculation. This enables spatial separation of electromagnetic responses at different penetration depths on the virtual depth axis, which helps alleviate the problem of signal superposition and mixing between shallow and deep layers caused by the skin effect, and provides a data foundation for defect signal analysis.

[0017] 2. The electromagnetic response amplitude and impedance characteristic phase are mapped to a virtual depth axis to generate a depth-domain impedance spectrum. After one-dimensional high-pass filtering to extract effective feature waveforms, similarity matching is performed with standard feature basis functions in a standard defect feature library. This suppresses background interference and determines the target standard basis function based on signal morphology, reducing the uncertainty of relying solely on amplitude features for judgment and improving the reliability of physical defect type identification.

[0018] 3. A linear superposition equation is constructed using the target standard basis function and the depth-domain impedance spectrum. The amplitude weighting coefficient and spatial offset are then used as unknown parameters for iterative fitting. Combined with a calibration reference table, the obtained mapping parameters are inverted into absolute geometric dimensions and actual burial depth. This transformation of the inversion process into parameter fitting under a known model achieves decoupling of defect size and depth, which is beneficial for improving the accuracy of quantitative inversion of cable alloy defects. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. The drawings are used to provide a further understanding of the present invention.

[0020] Figure 1 This is a flowchart illustrating an embodiment of this application.

[0021] Figure 2 This is a graph showing the changes in the real and imaginary parts of the impedance as a function of the discrete detection frequency in the embodiments of this application.

[0022] Figure 3 This is a graph showing the virtual depth axis and piecewise weight function as a function of detection frequency in the embodiments of this application.

[0023] Figure 4 This is a graph showing the distribution curves of the depth domain impedance spectrum and the target standard basis function on the virtual depth axis in the embodiments of this application.

[0024] Figure 5 This is a calibration curve showing the actual burial depth as a function of spatial offset in the embodiments of this application.

[0025] Figure 6 This is a schematic diagram of the structure in an embodiment of this application. Detailed Implementation

[0026] The following is in conjunction with the appendix Figure 1 - Figure 6 A preferred description of the present invention is provided below.

[0027] See attached document Figure 1 This invention proposes a method for inverting cable alloy defects based on multi-frequency eddy current impedance spectroscopy, comprising the following steps: S1. Obtain the initial detection data of multi-frequency eddy current, extract the resistance component and reactance component from the initial detection data of multi-frequency eddy current, and calculate and generate the phase change rate parameter. S2. Obtain the preset discrete detection frequency sequence, and perform spatial projection operation on the discrete detection frequency sequence using the phase change rate parameter to generate a virtual depth axis; S3. Extract the electromagnetic response amplitude and impedance characteristic phase from the initial multi-frequency eddy current detection data, and map the electromagnetic response amplitude and impedance characteristic phase to the virtual depth axis to generate a depth domain impedance spectrum. S4. Obtain the preset standard defect feature library, match the depth domain impedance spectrum with the standard feature basis functions in the standard defect feature library, and determine the target standard basis function; S5. Using the target standard basis function, the amplitude weighting coefficient and spatial offset to be solved are introduced as unknown parameters. Together with the depth domain impedance spectrum, a linear superposition equation is constructed. The linear superposition equation is solved to obtain the converged amplitude weighting coefficient and spatial offset, which are then combined as mapping parameters. S6. Obtain a preset calibration reference table, and use the calibration reference table to perform quantitative inversion decoding on the mapping parameters to generate a quantitative defect evaluation map.

[0028] In one embodiment of the present invention, step S1 includes the following steps: The impedance phase angle variable is obtained by calculating the arctangent dual-parameter function based on the resistance and reactance components. Perform phase continuity processing on the impedance phase angle variable to obtain the corrected impedance phase angle variable; Obtain the original detection frequency axis corresponding to the initial detection data of the multi-frequency eddy current, and calculate the discrete rate of change of the corrected impedance phase angle variable along the original detection frequency axis using the center difference method; Perform cubic spline interpolation on the discrete rate of change to generate a continuously varying phase rate of change parameter.

[0029] During implementation, a detection probe is used to scan and detect the area to be tested on the target cable alloy. The target cable alloy is a cable metal material with electrical conductivity and magnetic field response characteristics; the area to be tested is a continuous surface area on the target cable alloy that requires internal defect detection.

[0030] Before testing, the detection probe is installed on a controllable displacement mechanism, such as a linear slide rail or the end of a robotic arm. The detection probe applies an alternating electromagnetic field to the target cable alloy through an excitation coil and senses the induced voltage change caused by the change in the eddy current distribution inside the target cable alloy through the detection coil.

[0031] After installation, maintain a preset lift-off distance between the detection probe and the target cable alloy surface. This preset lift-off distance is set within the range of 2% to 15% of the average diameter of the detection probe coil, based on the probe's structure and the surface condition of the target cable alloy. This distance is maintained stable throughout the scanning process via a controllable displacement mechanism. The reason for setting this lift-off distance range is that if it is less than 2%, the detection probe is prone to friction and collision with the rough cable surface, leading to damage or mechanical vibration noise; if it is more than 15%, the alternating electromagnetic field will cause a sharp decrease in the effective magnetic flux penetrating into the target cable alloy due to spatial divergence, reducing the detection signal-to-noise ratio.

[0032] The reference response is acquired by a defect-free calibration test block. The reference response is used to compensate for the background error caused by the inherent impedance of the detection coil, the amplitude fluctuation of the excitation current, and the change in the lift-off of the detection probe. The lift-off compensation parameters are derived from the repeated measurement results of the defect-free calibration test block.

[0033] The controllable displacement mechanism moves along the axial or circumferential direction of the target cable alloy. The speed of movement is set according to the curvature of the target cable alloy, the size of the detection probe coil, and the sampling frequency, so that the displacement between adjacent spatial sampling positions is less than half the size of the effective detection area of ​​the excitation coil.

[0034] The digital signal processing module controls the excitation coil to output a sinusoidal excitation current according to a preset discrete detection frequency sequence. The discrete detection frequency sequence is arranged in ascending order of frequency value. The frequency range of the discrete detection frequency sequence is determined based on the conductivity and permeability of the target cable alloy, the size of the detection probe coil, and the expected detection depth. This ensures that the skin depth corresponding to the highest detection frequency in the discrete detection frequency sequence covers the near-surface area of ​​the target cable alloy, and the skin depth corresponding to the lowest detection frequency in the discrete detection frequency sequence covers the expected maximum detection depth.

[0035] During the implementation of the project, the number of discrete detection frequencies included in the discrete detection frequency sequence is set between 16 and 128. The reason for choosing this range is that if the number of frequencies is less than 16, it is impossible to generate sufficiently dense virtual depth nodes in subsequent steps, resulting in insufficient depth resolution and thus causing spatial aliasing; if the number of frequencies is more than 128, the total stable response time of a single scan is too long, reducing the efficiency of the project detection, and the electromagnetic response redundancy between adjacent frequencies is too high.

[0036] Adjacent discrete detection frequencies are arranged at equal intervals or logarithmic intervals. When there are deep defects in the area to be tested or the conductivity of the target cable alloy varies greatly, logarithmic intervals are preferred to increase the sampling density of smaller frequency ranges in the detection frequency sequence.

[0037] Within the stable response time corresponding to each discrete detection frequency, the induced voltage signal is acquired through the detection coil, and the excitation current reference signal is acquired simultaneously. The synchronous acquisition module samples the induced voltage signal and the excitation current reference signal synchronously, and obtains the in-phase voltage component and the quadrature voltage component respectively using a digital quadrature demodulation method.

[0038] The digital quadrature demodulation method multiplies the induced voltage signal by both the in-phase and quadrature reference carriers, and then performs a low-pass filter to obtain the resistance-related and reactance-related components. The cutoff frequency of the low-pass filter is set below 0.1 times the excitation frequency, but higher than the response change frequency caused by probe movement. The resistance and reactance components are then calculated based on the amplitude and phase relationship of the excitation current reference signal. (Unit: Ω) and reactance component (Unit: Ω) represents the real and imaginary parts of the complex impedance after background impedance compensation. The background impedance is measured by a defect-free calibration block at the same lift-off distance and the same discrete detection frequency.

[0039] The discrete detection frequencies, resistance components, reactance components, detection probe spatial positions, and acquisition times are associated and stored to form multi-frequency eddy current initial detection data, which is structured detection data composed of the aforementioned discrete detection frequencies, components, positions, and times.

[0040] The digital signal processing module reads the resistance and reactance components corresponding to each discrete detection frequency from the initial multi-frequency eddy current detection data, and calculates the impedance phase angle variable using an arctangent two-parameter function. The impedance phase angle variable is the complex impedance phase jointly determined by the resistance and reactance components, reflecting the relative change between resistive dissipation and reactive energy storage in the eddy current field. An arctangent two-parameter function is used to maintain the logical continuity of phase quadrant judgment when the resistance component approaches zero or the sign of the reactance component changes. The impedance phase angle variable is calculated according to the following formula: ; in, Indicates the first The impedance phase angle variable at a discrete detection frequency without continuous processing is measured in radians. The value range is 1 to , This represents the total number of discrete detection frequencies.

[0041] Phase continuity processing is performed on the impedance phase angle variables arranged in discrete detection frequency order to eliminate impedance phase angle variables crossing - To + Non-physical jumps occur at branch boundaries. The impedance phase angle variable after phase continuity is obtained according to the following formula: ; in, This represents the impedance phase angle variable after phase branch correction, with dimensions in rad. Integer correction amount. Indicates the first The phase branch correction integer corresponding to each discrete detection frequency, the integer correction amount Determined based on the phase difference between adjacent frequency points: ; in, Indicates the first The impedance phase angle variable after phase branching correction at a discrete detection frequency, especially when At that time, set the initial integer correction amount. ,Right now ; Pi is a constant; phase branch correction is based on the absolute value of the phase difference between adjacent frequency points not exceeding 1 / 2. This is a criterion for continuity.

[0042] After eliminating abrupt changes caused by phase branching between adjacent discrete detection frequency points, the phase change rate of the impedance phase angle variable after phase branching correction along the original detection frequency axis is calculated using the center difference method. For low signal-to-noise ratio (SNR) detection environments, the phase sequence can be low-pass smoothed with a cutoff frequency between 5% and 15% of the original frequency range before calculating the center difference. This setting is based on the fact that if it is below 5%, the signal will smoothly transition, thus erasing the abrupt changes in the actual low-frequency defect signal and causing missed detections; if it is above 15%, it cannot effectively filter out high-frequency random noise introduced by the environment and instrument, severely affecting the SNR of the subsequent differential calculation results. For non-uniform intervals between discrete detection frequencies, the difference between the corresponding discrete detection frequencies is used as the denominator of the difference, and the phase change rate parameter at the discrete detection frequency points is calculated using the center difference method. ; The calculation is performed using a one-sided difference method at the detection frequency endpoint: ; ; in, Indicates the first Discrete detection frequency The discrete phase change rate parameter at the point is expressed in rad·Hz. -1(or derive the unit rad·kHz) -1 ); , , , as well as They represent the 1st, 2nd, and 3rd respectively. The, the The and the first Impedance phase angle variable after phase branching correction at discrete detection frequencies; , , , , as well as They represent the 1st, 2nd, and 3rd respectively. The, the The, the The and the first The discrete detection frequency. The abnormal amplitude of the discrete phase change rate parameter can be determined by adding three times the standard deviation to the mean of the measurement results of the defect-free calibration block as an anomaly judgment reference. The anomaly judgment reference value is calculated based on no less than thirty sets of repeated calibration data.

[0043] The discrete phase change rate is converted into a rate of change curve over a continuous frequency range using cubic spline interpolation, generating the phase change rate parameter. (Continuous frequency location) The phase change rate parameter at that point is obtained through cubic spline interpolation: ; in, Indicates continuous frequency position The phase change rate parameter at that point, with dimensions in rad·Hz -1 ; In order to be with the first The cubic spline interpolation basis functions are corresponding to discrete detection frequencies. The cubic spline interpolation basis functions satisfy the function value constraints and first derivative continuity constraints at the corresponding nodes. The interpolation nodes are limited to adjacent discrete detection frequencies. The cubic spline interpolation adopts natural boundary conditions and does not extrapolate outside the frequency interval to avoid the rate of change distortion caused by endpoint extrapolation.

[0044] The output phase change rate parameter includes discrete phase change rate parameter values ​​at each discrete detection frequency and continuous change rate curves between each adjacent discrete detection frequency. The phase change rate parameter is synchronously saved with the corresponding frequency position and the spatial position of the detection probe. The phase change rate parameter is the rate of change of the impedance phase angle variable relative to the original detection frequency, characterizing the dynamic boundary transition rate of the capacitive characteristics of the eddy current alternating magnetic field. The larger the absolute value of the phase change rate parameter, the greater the change in the capacitive characteristics of the eddy current field caused by a unit frequency change.

[0045] In a specific embodiment, it is assumed that the target cable alloy is an aluminum alloy cable, and that the discrete detection frequency sequence contains four discrete detection frequency nodes, namely... , , and At the same spatial location of the detection probe, the digital signal processing module reads four sets of background impedance compensated resistive and reactive components from the initial multi-frequency eddy current detection data, which are respectively... , , , , , ,as well as , Substituting the first set of resistance and reactance components into the impedance phase angle calculation formula, we obtain... Substituting the second set of resistance and reactance component data into the equation yields... Substituting the data of the third set of resistance and reactance components into the equation yields... Substituting the fourth set of resistance and reactance component data into the equation yields... .

[0046] The difference between adjacent impedance phase angle variables is in to Within the range, therefore the integer values ​​for phase branch corrections are all taken as... , and Thus we obtain , , and .

[0047] According to the central difference formula, the phase change rate parameter at the second discrete detection frequency is: The phase change rate parameter at the third discrete detection frequency is: .

[0048] According to the one-sided difference formula, the phase change rate parameter at the first discrete detection frequency is: The phase change rate parameter at the fourth discrete detection frequency is .

[0049] The digital signal processing module uses the four discrete detection frequencies as interpolation nodes to perform cubic spline interpolation on the four discrete phase change rate parameters to obtain a continuously varying phase change rate parameter curve in the range of 10kHz to 80kHz. The parameter value near 20kHz reaches a positive local peak, indicating that the impedance phase angle near 20kHz changes at the largest rate with the original detection frequency. The continuous phase change rate parameter curve and its corresponding discrete detection frequency node are used together as the output result of step S1.

[0050] Combination Figure 2 As shown, by Figure 2 As can be seen from the wire harness trend, the resistance component gradually decreases nonlinearly with the increase of the detection frequency, while the reactance component shows a trend of first increasing and then decreasing, and reaches a local peak near 40kHz. The dynamic characteristics of the curve in the figure are consistent with the discrete measurement data points mentioned above.

[0051] In one embodiment of the present invention, step S2 includes the following steps: By comparing the amplitude distribution pattern of the phase change rate parameter, a preset phase change rate threshold is extracted. The frequency bands with an absolute value of the phase change rate parameter greater than the preset phase change rate threshold are identified as information-dense frequency bands, and the frequency bands with an absolute value of the phase change rate parameter less than or equal to the preset phase change rate threshold are identified as information-sparse frequency bands. Obtain the preset spatial coordinate stretching coefficient and waveform compression coefficient, configure the spatial coordinate stretching coefficient in the information-dense frequency band, configure the waveform compression coefficient in the information-sparse frequency band, and construct an adaptive mapping function. The discrete detection frequency sequence is imported into an adaptive mapping function for integration to calculate and obtain virtual depth nodes. Arrange all virtual depth nodes according to an increasing numerical rule to generate a virtual depth axis.

[0052] The digital signal processing module retrieves the continuous rate of change curve of the phase change rate parameter generated in step S1 and identifies the amplitude distribution pattern of the phase change rate parameter through statistical analysis or threshold comparison. The digital signal processing module calculates the absolute mean and standard deviation of the phase change rate parameter across the entire detection frequency range. The frequency range above the absolute mean plus one standard deviation is defined as the information-dense frequency band. This information-dense frequency band represents the location where the skin depth sweeps across the defect's geometric boundary or the interface of abrupt material property changes. At this point, the eddy current field response exhibits a gradient abrupt change with the original detection frequency, containing evaluation information for defect location and size assessment.

[0053] The digital signal processing module defines the frequency range below or equal to the absolute mean plus one standard deviation as the information sparse frequency band. The information sparse frequency band represents the region where the skin depth is far from the defect boundary. The eddy current field response changes slowly with the original detection frequency. The proportion of defect information contained in the initial detection data signal of the multi-frequency eddy current is relatively small.

[0054] The threshold mentioned above, used to distinguish between information-dense and information-sparse frequency bands, is the preset phase change rate threshold. The dimension is rad·Hz -1 The method for setting the preset phase change rate threshold is based on the detection data of defect-free samples and preset defective samples. The distribution histogram of the phase change rate parameter is statistically analyzed, and the boundary value that maximizes the difference in response between the two types of samples is selected, or the 75th percentile value of the absolute value of the phase change rate parameter of all samples is taken as the preset phase change rate threshold, so as to separate the information-dense frequency band and the information-sparse frequency band.

[0055] Configure a spatial coordinate stretching factor greater than 1 for information-intensive frequency bands. , It is a dimensionless constant used to scale the coordinates of information-dense frequency bands on the virtual depth axis; it is the spatial coordinate stretching factor. The value range is from 1.5 to 5.0. This range is limited because if the coefficient is less than 1.5, the coordinate stretching of the information-dense region is insufficient, making it impossible to effectively separate overlapping defect signals of similar depths in subsequent processing. If the coefficient is greater than 5.0, it will cause excessive local distortion of the mapped virtual depth axis, easily generating a large number of local minima when the nonlinear least squares approximation algorithm is called subsequently, causing the iterative process to fail to converge. This provides an amplified equivalent sampling resolution in the information-dense frequency band.

[0056] Configure waveform compression coefficients between 0 and 1 for information-sparse frequency bands , A dimensionless constant used to compress sparse frequency bands on the virtual depth axis; waveform compression coefficient. The value range is from 0.1 to 0.5. This range is limited because if the coefficient is below 0.1, it may completely erase extremely weak early defect signals or material gradient characteristics, leading to missed detections; if the coefficient is above 0.5, the compression of background redundancy information is insufficient, failing to effectively highlight the main peak characteristics of the defect. Therefore, a method is constructed based on the original detection frequency. Piecewise weighting function for independent variables The piecewise weight function is calculated as follows: ; in, The phase change rate parameter at continuous frequency positions generated in step S1. The piecewise weighting function then generates an adaptive mapping function with nonlinear mapping capability through integration. The adaptive mapping function is a monotonically increasing function that takes the original detection frequency as input and virtual depth coordinates as output, dynamically reconstructing the coordinate system based on the information content of the initial multi-frequency eddy current detection data signal. Adaptive mapping function The calculation formula is obtained by definite integral over the piecewise weight function, as follows: ; in, Represents frequency The virtual depth coordinates on the virtual depth axis after nonlinear spatial projection calculation; The starting frequency of the discrete detection frequency sequence.

[0057] The digital signal processing module substitutes each discrete detection frequency sequence used in the original detection process into the adaptive mapping function and performs nonlinear spatial projection operations. For each original discrete detection frequency, it calculates a new coordinate value after coordinate distortion, i.e., a virtual depth node. The virtual depth node is the mapping point of the original discrete detection frequency in the reconstructed virtual depth coordinate system. In discrete calculations, the... Discrete detection frequency Corresponding virtual depth node Calculated through numerical integration iteration: ; Among them, setting the initial virtual depth node ,and The value range is 2 to , This represents the total number of discrete detection frequencies. Virtual depth coordinates are calculated through integration. The dimensional equivalent of is Virtual depth coordinates represent the amount of accumulated phase change information, rather than time frequency.

[0058] All calculated virtual depth nodes are arranged in ascending order of value to form a non-uniformly spaced coordinate sequence. This non-uniformly spaced coordinate sequence constitutes the virtual depth axis, a one-dimensional non-uniform coordinate axis composed of all virtual depth nodes. This makes the gradient change of the initial multi-frequency eddy current detection data signal along the virtual depth axis more uniform than that along the original frequency axis. Defect responses from different depth sources are separated by coordinates on the virtual depth axis, providing a data foundation for subsequent orthogonal decomposition. The coordinate intervals within the virtual depth axis reflect the information distribution density along the original frequency axis, thus suppressing the aliasing effect of depth signals.

[0059] In a specific embodiment, the digital signal processing module acquires four discrete detection frequency points. , , and The corresponding phase change rate parameters are respectively , , and The system sets a phase change rate threshold based on historical calibration data. Set the spatial coordinate stretching factor and waveform compression coefficient .

[0060] The digital signal processing module makes point-by-point judgments, because and Frequency point and The piecewise weight function values ​​at each point are all This indicates that the two discrete detection frequency points are located in the information-dense frequency band.

[0061] because and Frequency point and The piecewise weight function values ​​at each point are all This indicates that the two discrete detection frequency points are within the information-sparse frequency band.

[0062] The virtual depth node is calculated using a numerical integration formula, and the starting virtual depth node is set. Calculate the second virtual depth node. Calculate the third virtual depth node. Calculate the fourth virtual depth node. .

[0063] The calculated virtual depth nodes {0kHz, 20.0kHz, 45.0kHz, 65.0kHz} are arranged sequentially to generate a virtual depth axis. The coordinates of the virtual depth axis correspond to the original detection frequency points of 10kHz, 20kHz, 40kHz, and 80kHz, respectively. The original uniform logarithmic frequency interval is nonlinearly transformed into virtual depth intervals of 20.0kHz, 25.0kHz, and 20.0kHz, stretching the information-dense region between 10kHz and 40kHz and compressing the information-sparse region between 40kHz and 80kHz.

[0064] See Figure 3The piecewise weighting function maintains a high weight of 2.0 in the information-dense frequency band from 10kHz to 40kHz, and then drops to 0.5 after 40kHz. Under the integral effect of different weights, the virtual depth axis has a larger slope in the dense frequency band and is stretched by the coordinates, while the slope becomes gentler in the sparse frequency band and is compressed by the coordinates, reflecting the spatial reconstruction effect of eliminating depth aliasing.

[0065] In one embodiment of the present invention, step S3 includes the following steps: Extract the virtual depth nodes contained within the virtual depth axis; Based on the virtual depth nodes, the electromagnetic response amplitude is re-indexed and the impedance characteristic phase is integrated to construct a reference data sequence; The baseline data sequence is processed by a one-dimensional high-pass filtering algorithm to extract effective feature waveforms. The effective characteristic waveform is combined and mapped with the virtual depth axis to generate a depth domain impedance spectrum.

[0066] The digital signal processing module acquires the initial multi-frequency eddy current detection data from step S1. This initial data represents the resistance and reactance components corresponding to each discrete detection frequency point. The digital signal processing module calculates the electromagnetic response amplitude information at each discrete detection frequency point. This amplitude information is the modulus of the complex impedance, characterizing the total response intensity induced by the eddy current field in the target cable alloy. Discrete detection frequency Electromagnetic response amplitude information The formula for calculating (unit: Ω) is: ; in, and The first number obtained in step S1 is respectively The impedance and reactance components are measured at discrete detection frequencies. Simultaneously, the digital signal processing module calculates the corresponding impedance characteristic phase information, which has undergone phase continuity processing. The impedance characteristic phase information is the argument of the complex impedance, characterizing the phase delay of the induced response relative to the sinusoidal excitation current signal, reflecting the energy dissipation and storage characteristics in the eddy current field. The value of the impedance characteristic phase information is the phase-continuous impedance phase angle variable calculated in step S1. .

[0067] The digital signal processing module retrieves the virtual depth axis generated in step S2. This virtual depth axis contains virtual depth nodes that correspond one-to-one with the original discrete detection frequencies. By re-indexing the original electromagnetic response amplitude information and impedance characteristic phase information according to the order of the virtual depth nodes, the impedance data is mapped to the new coordinate system, forming a reference data sequence. The reference data sequence is the expression of the original multi-frequency eddy current initial detection data in the reconstructed virtual depth coordinate system, that is, the sequence at the [missing information - likely a specific frequency]. virtual depth nodes The amplitude at that point is Phase is The mapping of impedance data to the new coordinate system is achieved through direct index remapping. Since the virtual depth nodes of the virtual depth axis are calculated based on the original discrete detection frequency, there is a one-to-one correspondence between the two, and no numerical interpolation is required.

[0068] To filter out slowly changing baseline drift introduced by material micro-inhomogeneities or electromagnetic environment fluctuations, the digital signal processing module incorporates a one-dimensional high-pass filter algorithm to process the reference data sequence. This algorithm removes low-frequency, slowly changing background trends from the signal while preserving details of defect signals with high gradients that characterize local abrupt changes. In this scenario, it is used to filter out background impedance that varies significantly along the virtual depth axis. The one-dimensional high-pass filter algorithm estimates the low-frequency baseline component by calculating the local moving average of the reference data sequence, and the calculated local baseline value... (Unit: Ω) The calculation formula is as follows: ; in, The half-width of the moving average window is a dimensionless integer. When calculating local baseline values ​​near the ends, the amplitude sequence of the reference data sequence undergoes a symmetrical extension process to meet the window calculation requirements. (The half-width of the moving average window is...) The setting is based on the total width of the moving average window. The moving average window should be larger than the node width range that the expected defect feature occupies on the virtual depth axis, but smaller than one-third of the entire virtual depth axis length. For example, if the preset defect response spans 3 to 5 virtual depth nodes, then the half-width of the moving average window... The value can be set to 3 to 5 to ensure that the moving average window covers both sides of the defect response when calculating the local baseline value, thereby separating the defect signal as an independent component.

[0069] The one-dimensional high-pass filtering algorithm achieves filtering through baseline subtraction. It subtracts low-frequency baseline components from the reference data sequence, thereby extracting defect signal features reflecting rapid local changes on the virtual depth axis and generating an effective feature waveform. The effective feature waveform is the defect response signal after subtracting local baseline values ​​from the reference data sequence. In the... Effective feature waveform values ​​at each virtual depth node The formula for calculating (unit: Ω) is: ; By combining the effective characteristic waveform with a virtual depth axis serving as its coordinate basis, a depth-domain impedance spectrum is constructed that isolates the contributions of deep physical defects from those of shallow epidermal physical defects. The depth-domain impedance spectrum is the final output data structure, consisting of the effective characteristic waveform and the virtual depth axis serving as its abscissa, providing a decoupled input signal for subsequent basis function decomposition. The filtered depth-domain impedance spectrum can be represented as a set... ,in It also underwent high-pass filtering. This represents the total number of discrete detection frequencies. The digital signal processing module outputs the depth-domain impedance spectrum as structured data to subsequent processing steps.

[0070] In a specific embodiment, the digital signal processing module acquires impedance data at four discrete detection frequency points and the corresponding virtual depth axis. Calculate the electromagnetic response amplitude at each discrete detection frequency point, and simultaneously, based on the impedance data from step S1, , , , .

[0071] The electromagnetic response amplitude information data is mapped one-to-one with the virtual depth nodes of the virtual depth axis to construct a reference data sequence. The amplitudes at virtual coordinate points 0, 20.0, 45.0 and 65.0 are 2.236Ω, 2.343Ω, 2.561Ω and 1.921Ω, respectively.

[0072] Import a one-dimensional high-pass filter algorithm and set the half-width of the moving average window. Calculate the local baseline value at the second virtual depth node. Calculate the local baseline value at the third virtual depth node. .

[0073] To simplify the engineering demonstration, this example omits the processing of symmetrical extensions at both ends of the nodes, performs only one-dimensional high-pass filtering isolation calculations on the internal nodes, and directly sets the effective characteristic waveform values ​​at the endpoints to zero. Subtracting the local baseline value from the original amplitude yields the effective characteristic waveform values ​​at the corresponding virtual depth nodes. , .

[0074] The effective characteristic waveform values ​​of the generated depth domain impedance spectrum are a set of data points at virtual coordinate points 0, 20.0, 45.0, and 65.0, corresponding to 0Ω, -0.037Ω, 0.286Ω, and 0Ω, respectively. The processing result distributes the signal peak value around the virtual depth of 45.0, indicating that the main defect features in the original multi-frequency eddy current initial detection data signal have been successfully separated.

[0075] In one embodiment of the present invention, step S4 includes the following steps: Perform isolated peak detection on the depth domain impedance spectrum and count the extreme values ​​of the number of isolated peaks within the depth domain impedance spectrum; Based on the extreme value of the number of isolated peaks, standard feature basis functions with mismatched topology in the standard defect feature library are eliminated, and candidate standard feature basis functions are obtained to form a candidate basis function set. Calculate the normalized cross-correlation coefficient between each standard feature basis function in the candidate basis function set and the depth domain impedance spectrum to obtain the similarity matching score; The similarity matching scores corresponding to each standard feature basis function are sorted in descending order, and the standard feature basis function with the highest score is determined as the target standard basis function.

[0076] The digital signal processing module accesses a pre-built standard defect feature library stored in non-volatile memory. The standard defect feature library is a database that stores standard defect response models for preset cable types and detection configurations. The library contains standard waveform data of standard characteristic basis functions corresponding to various typical cable damage categories. Each standard characteristic basis function is a data entity in the library, representing the waveform of a preset type of physical defect, such as point defects, linear cracks, or planar corrosion, after undergoing the same processing steps S1 to S3.

[0077] Standard waveform data refers to the discrete amplitude sequence of standard characteristic basis functions in standard virtual depth coordinates. The waveforms are obtained through finite element simulation or experimental measurement using standard defect blocks, under the same virtual depth axis construction rules as in steps S2 and S3. The standard waveform data of each standard characteristic basis function is translated and normalized to align the characteristic center or peak position of its response to the origin of the standard virtual depth coordinate system.

[0078] The digital signal processing module extracts all available standard feature basis functions in batches from the standard defect feature library. Each standard feature basis function contains a category label, such as "surface scratch", "internal hole" or "interlayer corrosion", as well as waveform data of the standard feature basis function in standard virtual depth coordinates.

[0079] The digital signal processing module loads the depth domain impedance spectrum data sequence to be analyzed, generated in step S3, and executes an isolated peak detection algorithm on the data sequence. The isolated peak detection algorithm determines the number and location of peaks in the depth domain impedance spectrum data sequence by calculating the first derivative of the data sequence and finding the zero-crossing point where the sign changes from positive to negative, i.e., it counts the extreme values ​​of the number of isolated peaks.

[0080] The extrema of the number of isolated peaks are used to determine the macroscopic morphology of the depth-domain impedance spectrum and to filter candidate standard characteristic basis functions for topological mismatches. The determination of the extrema of the number of isolated peaks is achieved by analyzing depth-domain impedance spectrum data. The first-order difference realization, if and Then it is determined that it is in the virtual depth node. There is a peak at that point.

[0081] If the depth-domain impedance spectrum exhibits more than one main peak, and a candidate standard characteristic basis function has a single-peak shape, then the candidate standard characteristic basis function is excluded due to topological mismatch. For example, a standard characteristic basis function with a single-peak characteristic cannot fit a measured signal containing two peaks. The preprocessing step of determining the shape by the extreme values ​​of the number of isolated peaks can reduce the number of subsequent cross-correlation calculations. After the initial screening, the remaining candidate standard characteristic basis functions constitute the candidate basis function set, that is, the dynamic set of all candidate standard characteristic basis functions after screening.

[0082] The digital signal processing module sequentially performs morphological similarity matching between each standard characteristic basis function in the candidate basis function set and the depth domain impedance spectrum. The morphological similarity matching process is implemented by calling a sparse space decomposition evaluation algorithm. The standard characteristic basis function waveform is used as a template and slid across the depth domain impedance spectrum. The inner product at each sliding position is calculated, and the maximum inner product value is the similarity matching score. The similarity matching score is the maximum value obtained from normalized cross-correlation calculation. It does not depend on the absolute amplitude and DC bias of the impedance signal, but only reflects the similarity of waveform shapes. After centering and normalization, the similarity matching score is a dimensionless pure number between -1 and 1. The closer the similarity matching score is to 1, the more similar the topological shapes of the two waveforms are.

[0083] Assume the effective amplitude sequence of the depth domain impedance spectrum is ,in For the first Effective feature waveform values ​​at each virtual depth node The total number of virtual depth nodes; the waveform of a candidate standard feature basis function is as follows: ,in The standard characteristic basis function is the first on the standard virtual depth axis. The amplitude at each virtual depth node The total number of virtual depth nodes in the standard characteristic basis function waveform, and Similarity matching score The calculation formula is: ; in, The discrete displacement of the standard characteristic basis function waveform sliding on the depth domain impedance spectrum, with values ​​ranging from 0 to... ; It is a standard characteristic basis function sequence The mean, In the current sliding discrete displacement Below, the sequence of overlapping parts with the standard characteristic basis functions in the depth-domain impedance spectrum. The mean of the denominator. In engineering implementation, when the product of the standard deviations calculated from the denominator is close to zero, the similarity matching score is determined to be zero to avoid division by zero errors and to shield invalid pure DC signals.

[0084] The digital signal processing module records the similarity matching score corresponding to each candidate standard feature basis function. After completing all traversal calculations, all similarity matching scores are sorted in descending order. The candidate basis function with the highest similarity matching score is selected. This candidate basis function refers to a specific standard feature basis function whose similarity matching score is higher than other standard feature basis functions among all candidate standard feature basis functions. The similarity matching score of this candidate basis function must be higher than a preset extraction threshold, generally set between 0.80 and 0.90, and must be at least 0.1 higher than the second-highest similarity matching score to achieve a unique match of the standard feature basis function.

[0085] The basis for setting this extraction threshold and margin is that an absolute threshold of 0.80 to 0.90 ensures that the extracted measured waveforms and theoretical basis functions have a high degree of homology in terms of physical mechanism and spatial topology; while a difference margin of 0.1 guarantees the unique determinism of the matching result under random white noise interference in the environment, preventing misclassification caused by multiple basis functions not being absolutely orthogonal. After truncating the candidate basis function with the highest matching degree, it is uniquely determined as the target standard basis function for subsequent analytical decomposition, that is, the selected standard feature basis function model.

[0086] In a specific embodiment, the digital signal processing module accesses a standard defect feature library, which contains two candidate standard feature basis functions: basis functions. Represents "surface scratches", basis function The waveform is a single-peak shape; the basis function Represents "double-layer decoupling", basis function The waveform exhibits a bimodal shape. The digital signal processing module performs peak detection on the depth domain impedance spectrum and discovers a single-peak feature near the virtual depth of 45.0.

[0087] Because the impedance spectrum signal in the depth domain to be measured has a single-peak characteristic, while the basis function The two peaks are bimodal, and their topologies do not match, therefore the basis functions... Excluded, only basis functions Enter the candidate basis function set. Assume basis functions. Waveform data in standard virtual depth coordinates is The waveform data spans three standard virtual depth coordinate nodes.

[0088] To perform cross-correlation calculations, the digital signal processing module processes the depth domain impedance spectrum sequence. With basis functions The sequence is matched. When the sliding discrete displacement... When, the overlapping subsequences are Calculate overlapping subsequences and basis functions The normalized cross-correlation coefficient. Calculate the mean. , .

[0089] The inner product of the numerator is 0.162. Then, the product of the two standard deviation terms in the denominator is approximately 0.163. Dividing these two products yields the sliding discrete displacement. The cross-correlation coefficient of the positions was calculated to be approximately 0.99. The digital signal processing module continued to try other sliding discrete displacement positions, finding that when the sliding discrete displacement... At this point, the cross-correlation coefficient reaches its maximum. This is because the candidate basis function set contains only basis functions. basis functions The highest similarity score of 0.99 obtained makes it the candidate basis function with the highest matching degree. Therefore, the digital signal processing module finally determines the basis function representing the "surface scratch". is the target standard basis function.

[0090] Combination Figure 4 As shown, after one-dimensional high-pass filtering and isolation calculation, the measured signal exhibits a distinct single-peak characteristic with an amplitude of approximately 0.286Ω near a virtual depth of 45.0kHz. The topological shape of this isolated peak is consistent with the target standard basis function shown by the dashed line in the figure, effectively removing background baseline drift and providing a clean defect signal characteristic with a high signal-to-noise ratio for subsequent fitting and inversion.

[0091] In one embodiment of the present invention, step S5 includes the following steps: Initialize the amplitude weighting coefficient and spatial offset based on the peak ratio of the depth domain impedance spectrum to the target standard basis function and the depth coordinates; Substitute the initial magnitude weighting coefficients, spatial offset, target standard basis function, and depth domain impedance spectrum into the linear superposition equation to construct an overdetermined system of equations. The overdetermined system of equations is iteratively solved using a nonlinear least squares approximation regression fitting algorithm, and the fitting residuals are calculated. The iteration is interrupted when the fitting residual is less than the preset lower limit of error. The converged amplitude weight coefficient and spatial offset are then exported and combined to generate mapping parameters.

[0092] The digital signal processing module dynamically constructs linear superposition equations in the system's running memory. The linear superposition equations describe how the measured signal can be generated from a standard signal through scaling and translation. The linear superposition equations are based on the waveform of the target standard basis function determined in step S4, and introduce amplitude weighting coefficients and spatial offsets as unknowns to be solved.

[0093] The amplitude weighting coefficient is a scalar multiplier that quantitatively characterizes the proportional relationship between the actual defect response intensity and the intensity of the target standard basis function. The magnitude of the amplitude weighting coefficient is directly related to the geometric dimensions or response intensity of the actual defect. The spatial offset is the translation distance required for the target standard basis function to match the measured signal on the virtual depth axis, indicating the center position of the actual defect response and relating to the burial depth of the actual defect. The linear superposition equation is in the... One actual discrete sampling physical stationary point Model predictions at [location] (Unit: Ω) can be expressed as: ; in, It is the amplitude weighting coefficient, which is a dimensionless scalar; It is a spatial offset, with the same dimensions as the virtual depth axis; It is a continuous function obtained by performing continuous interpolation on the target standard basis function determined in step S4. The linear superposition equation is linear with respect to the amplitude weighting coefficients, but nonlinear with respect to the spatial offset.

[0094] The digital signal processing module uses the depth domain impedance spectrum generated in step S3 as the fitting target and imports the target standard basis function as the fitting template. The module reads the actual discrete sampling physical stationary points corresponding to each data point in the depth domain impedance spectrum on the virtual depth axis. It substitutes the stationary point coordinates, the corresponding effective characteristic waveform amplitudes, and the waveform data of the target standard basis function into the linear superposition equation constraints to construct an overdetermined system of equations containing two unknowns and multiple equation constraints, i.e., the number of virtual depth node data points. Nonlinear equation systems with more than two unknown parameters, such as amplitude weighting coefficients and spatial offsets.

[0095] The digital signal processing module triggers a built-in nonlinear least squares approximation regression fitting algorithm, such as the Levenburg-Marquardt algorithm, to initiate an iterative polling mechanism to solve the overdetermined system of equations in order to find the amplitude weighting coefficients and spatial offsets that satisfy the convergence conditions. The value. In each iteration, the nonlinear least squares approximation regression fitting algorithm calculates the deviation between the current linear superposition equation model and the actual data, i.e., the fitting residual.

[0096] The fitting residual refers to the difference between the model's predicted value and the actual observed value at each virtual depth node data point. The sum of squares of the fitting residuals characterizes the fitting error of the linear superposition equation model. Let... The depth-domain impedance spectrum generated in step S3 is in the... The effective characteristic waveform amplitude of each virtual depth node, and the sum of squares of the fitted residuals. (Unit: Ω) 2 The objective function is: ; The goal of the iterative solution process is to find convergent estimates of the magnitude weighting coefficients and spatial offset. : ; The nonlinear least squares approximation regression fitting algorithm determines the gradient descent direction based on the Jacobian matrix of the fitting residuals and dynamically adjusts the values ​​of the amplitude weight coefficient and the spatial offset. The adjustment amount of the amplitude weight coefficient and the spatial offset is the gradient compensation value.

[0097] The dynamic adjustment process continues cyclically until the sum of squares of the calculated fitting residuals is less than a preset lower error threshold. (Preset lower error threshold) The measurement noise level of the measurement system is set to determine the fitting accuracy of the nonlinear least squares approximation regression.

[0098] The preset lower limit threshold is set based on the statistical mean of the sum of squared residuals obtained after the same processing procedure for defect-free samples. It is typically set to one to five times the statistical mean of the sum of squared residuals for defect-free samples. This is because if the multiple is less than one, the algorithm may overfit to background random noise, increasing meaningless iteration time and potentially causing the algorithm to fail to converge. If the multiple is greater than five, the iteration process may terminate prematurely, leading to insufficient accuracy in the nonlinear fitting of minute defect features, and consequently causing distortion in the 3D size inversion of defects. This ratio balances the system's tolerance to white noise and eliminates overfitting. The termination condition for the iteration is: ; in, For the first The magnitude weighting coefficient and spatial offset parameter values ​​for the next iteration.

[0099] The iteration process terminates when the fitting residual meets the preset lower error threshold or the number of iterations reaches the upper limit, triggering the interruption condition. The digital signal processing module derives the convergence values ​​of the amplitude weighting coefficients and spatial offset for the current iteration round. It then combines these convergence values ​​to form mapping parameters that quantitatively characterize the spatial projection relationship of the defect. These mapping parameters are the converged estimates of the amplitude weighting coefficients and spatial offset. This will be the final output of this step.

[0100] In a specific embodiment, the digital signal processing module receives a depth-domain impedance spectrum, and the effective characteristic waveform amplitude sequence of the depth-domain impedance spectrum is as follows: Virtual depth node corresponding to the virtual depth axis The received discrete waveform of the target standard basis function is as follows: .

[0101] Digital signal processing module constructs objective function When the nonlinear least squares approximation regression fitting algorithm starts, the initial values ​​of the amplitude weighting coefficients are set based on the ratio of the measured signal peak value to the target standard basis function peak value. The initial value of the spatial offset is set based on the virtual depth coordinates corresponding to the measured signal peak value, i.e. .

[0102] The preset lower limit threshold for error is set. With initial value The iteration begins, and the Levenburg-Marquardt algorithm calculates the sum of squared residuals under the current linear superposition equation model, and calculates the Jacobian matrix to determine the update direction of the magnitude weight coefficients and spatial offset parameters, as well as the step size of the gradient compensation value.

[0103] After several iterations, assuming the Levenburg-Marquardt algorithm is at the [number]th [time]... After the iteration, the amplitude weighting coefficient and the spatial offset parameter converge to and At this point, the digital signal processing module calculates the sum of squared fitting residuals under the convergence parameters; for example, it calculates... The iteration terminates because the calculated sum of squared residuals of the fit, 0.0008, is less than the preset lower limit of error, 0.001.

[0104] The digital signal processing module then derives the amplitude weighting coefficient and spatial offset convergence value, generating mapping parameters. These mapping parameters are a combination of the amplitude weighting coefficient and spatial offset values: {0.315, 44.8 kHz}. Here, 0.315 is the final determined amplitude weighting coefficient, and 44.8 kHz is the final determined spatial offset. Both of these values ​​are used as the output of this step.

[0105] In one embodiment of the present invention, step S6 includes the following steps: Extract the defect classification labels carried by the target standard basis function and match them with the calibration reference table to obtain the physical defect types; Extract the amplitude weighting coefficient from the mapping parameters, and obtain the size conversion coefficient from the calibration reference table in combination with the type of physical defect; The absolute geometric dimensions are calculated by multiplying the amplitude weighting coefficient and the size conversion coefficient. Extract the spatial offset from the mapping parameters and convert the spatial offset into the actual burial depth; By summarizing the types of physical defects, their absolute geometric dimensions, and their actual burial depth, a quantitative defect evaluation map is generated.

[0106] The digital signal processing module extracts the inherent defect classification labels of the target standard basis functions from the target standard basis functions determined in step S4. The defect classification labels are text information bound to the target standard basis functions in the pre-existing standard defect feature library, which qualitatively characterizes the defect type corresponding to the target standard basis functions, such as "surface scratches".

[0107] The digital signal processing module uses defect classification tags as search keywords to retrieve a calibration reference table stored in the system's read-only memory. This calibration reference table, a core database, stores conversion rules and calibration coefficients from signal domain features to physical domain parameters. It was established through finite element simulation and experimental measurements using standard artificial defect test blocks of known dimensions and depths. Based on this, the digital signal processing module matches and identifies the corresponding physical defect type, which represents an engineering classification of potential damage in the cable, such as cracks, corrosion, wear, and conductor breakage.

[0108] Meanwhile, the digital signal processing module extracts the amplitude weighting coefficient from the mapping parameters generated in step S5, and queries the calibration reference table according to the determined physical defect type to obtain the size conversion coefficient bound to the physical defect type.

[0109] The size conversion factor is a calibration constant obtained from experience or simulation. It establishes a linear relationship between the dimensionless amplitude weighting factor and the actual defect size with physical units. The value of the size conversion factor depends on the defect type and size measurement method of the physical defect.

[0110] The digital signal processing module multiplies the amplitude weighting coefficient by the size conversion coefficient to calculate and reconstruct the absolute geometric dimensions representing the defect's geometric scale. Absolute geometric dimensions are a quantitative description of the defect's geometric scale, such as the length and maximum width of a crack, or the average depth and area of ​​a corroded region. Let... The amplitude weighting coefficients obtained in step S5 The absolute geometric dimensions are the dimensional conversion factors associated with the corresponding physical defect types. The calculation formula is: ; In the formula, the size conversion factor is... The dimension depends on the defined dimension, such as mm, while the amplitude weighting coefficient... Dimensionless, therefore absolute geometric dimensions The dimension of is the dimension of length.

[0111] The digital signal processing module extracts the spatial offset from the mapping parameters and retrieves the skin effect attenuation constant corresponding to the corresponding cable alloy material from the calibration reference table. The skin effect attenuation constant describes the exponential decay of eddy current density with depth in the corresponding material, determining the relationship between the original detection frequency and the detection depth. Through a preset inversion function, which describes the nonlinear relationship between the virtual depth coordinates and the physical burial depth, the digital signal processing module converts the spatial offset into the actual burial depth on the radial profile of the material. The actual burial depth is the radial vertical distance from the defect center point or the main feature point of the defect to the cable surface.

[0112] Calculate the actual burial depth At that time, the adaptive mapping relationship in inversion step S2 is retrieved. In discrete computation, the obtained spatial offset is... By reverse lookup or interpolation on the virtual depth axis, the original detection frequency corresponding to the spatial offset is determined. Then, using the skin depth formula, the calculation formula is as follows: ; in, The dimensionless depth inversion correction constant is obtained by calibration based on the corresponding probe configuration and defect type. Specifically, the calibration method involves obtaining a series of test data using a stepped depth standard test block made of the same material as the target cable alloy, and fitting the ratio between the known physical depth and the theoretically calculated depth using the least squares method to obtain the optimal constant value for correcting the spatial magnetic field divergence characteristics of the probe. Pi is a constant. The units are S / m and The units are H / m, representing the conductivity and permeability of the target cable alloy, respectively. The conductivity and permeability of the target cable alloy together form the basis for calculating the skin effect attenuation constant. In practice, the conventional approach is to establish a spatial offset. Compared with the actual burial depth calibration curves between The calibration curves are obtained by measuring and fitting a series of standard defects of known depth.

[0113] The digital signal processing module combines the physical defect types obtained from the qualitative identification, the absolute geometric dimensions obtained from the quantitative calculation, and the actual burial depth obtained from the inversion calculation to form a quantitative defect evaluation conclusion that covers type judgment, geometric calibration, and spatial positioning. This conclusion is then presented in the form of a visual map, generating a quantitative defect evaluation map.

[0114] The quantitative defect evaluation map uses graphical markers on a simulated cable cross-section or 3D model to display the identified physical defect types, cross-sectional locations, and absolute geometric dimensions, completing a closed-loop operation structure from electromagnetic signals to physical defects.

[0115] In a specific embodiment, following the results of the preceding steps, the digital signal processing module obtains the target standard basis function as the basis function C1 representing the "surface scratch," and simultaneously obtains the mapping parameters {0.315, 44.8kHz}, where the amplitude weighting coefficients... Spatial offset .

[0116] The digital signal processing module extracts the defect classification label "surface scratch" from the basis function C1, and uses the defect classification label to look up the calibration reference table to determine that the physical defect type is "surface scratch".

[0117] Based on the physical defect type "surface scratch," the digital signal processing module consults the calibration reference table again to find the corresponding size conversion factor. Assuming the size conversion factor is... The size conversion factor is used to convert the dimensionless amplitude weighting factor into a scratch width with physical dimensions. The digital signal processing module calculates the absolute geometric dimensions. That is, the width of the scratch is determined to be 0.1575mm.

[0118] The digital signal processing module is based on the spatial offset. Query the calibration reference table to find the calibration curves of the calibrated virtual depth-physical depth mapping. Assuming the result is obtained through table lookup or interpolation, when... At that time, the corresponding actual burial depth was .

[0119] The digital signal processing module integrates the aforementioned results to generate a defect assessment report: the defect type is "surface scratch," the width is 0.1575 mm, and the burial depth is 0.05 mm. The defect assessment report is rendered on the cable cross-section diagram, with an arrow pointing to a mark 0.05 mm below the surface, next to which is labeled "scratch, W: 0.16 mm," completing the generation of the quantitative defect evaluation map.

[0120] Combination Figure 5 As shown, the scatter points in the figure represent known calibration point data in the preset calibration reference table, which are then used to generate a complete mapping through curve fitting. At the same time, the figure includes analytical markings for actual test cases, clearly indicating the inversion calibration result of "44.8kHz corresponds to 0.05mm", thereby realizing the quantitative conversion of the depth of small defects in cable alloys.

[0121] See appendix Figure 6 This invention also proposes a cable alloy defect inversion system based on multi-frequency eddy current impedance spectrum, which includes the following modules: The phase change rate parameter generation module acquires the initial detection data of multi-frequency eddy currents, extracts the resistance and reactance components from the initial detection data of multi-frequency eddy currents, and calculates and generates the phase change rate parameter. The virtual depth axis construction module obtains a preset discrete detection frequency sequence, performs a spatial projection operation on the discrete detection frequency sequence using the phase change rate parameter, and generates a virtual depth axis. The impedance spectrum reconstruction module extracts the electromagnetic response amplitude and impedance characteristic phase from the initial multi-frequency eddy current detection data, maps the electromagnetic response amplitude and impedance characteristic phase to the virtual depth axis, and generates a depth domain impedance spectrum. The target standard basis function screening module obtains a preset standard defect feature library, matches the depth domain impedance spectrum with the standard feature basis functions in the standard defect feature library, and determines the target standard basis function. The parameter solving module uses the target standard basis function and introduces the amplitude weighting coefficient and spatial offset to be solved as unknown parameters. Together with the depth domain impedance spectrum, it constructs a linear superposition equation. The linear superposition equation is solved to obtain the converged amplitude weighting coefficient and spatial offset, which are then combined as mapping parameters. The quantitative inversion decoding module obtains a preset calibration reference table, uses the calibration reference table to perform quantitative inversion decoding on the mapping parameters, and generates a quantitative defect evaluation map.

[0122] Each module can be implemented entirely or partially through software, hardware, or a combination thereof. It supports hardware embedded in or independent of the processor in the computer device, and also supports software stored in the memory of the computer device, so that the processor can call and execute the operations corresponding to each of the above modules.

[0123] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for defect inversion of cable alloys based on multi-frequency eddy current impedance spectroscopy, characterized in that, Includes the following steps: S1. Obtain initial multi-frequency eddy current detection data, extract the resistance component and reactance component from the initial multi-frequency eddy current detection data, and calculate and generate phase change rate parameters. S2. Obtain a preset discrete detection frequency sequence, and perform a spatial projection operation on the discrete detection frequency sequence using the phase change rate parameter to generate a virtual depth axis; S3. Extract the electromagnetic response amplitude and impedance characteristic phase from the initial multi-frequency eddy current detection data, and map the electromagnetic response amplitude and impedance characteristic phase to the virtual depth axis to generate a depth domain impedance spectrum. S4. Obtain a preset standard defect feature library, match the depth domain impedance spectrum with the standard feature basis functions in the standard defect feature library, and determine the target standard basis function; S5. Using the target standard basis function, the amplitude weighting coefficient and spatial offset to be solved are introduced as unknown parameters, and together with the depth domain impedance spectrum, a linear superposition equation is constructed. The linear superposition equation is solved to obtain the converged amplitude weighting coefficient and spatial offset, which are then combined as mapping parameters. S6. Obtain a preset calibration reference table, and use the calibration reference table to perform quantitative inversion decoding on the mapping parameters to generate a quantitative defect evaluation map.

2. The multi-frequency eddy current impedance spectroscopy based cable alloy defect inversion method of claim 1, wherein, Step S1 includes: Based on the resistance component and the reactance component, the impedance phase angle variable is obtained by calculating the arctangent two-parameter function. The impedance phase angle variable is subjected to phase continuity processing to obtain the corrected impedance phase angle variable; Obtain the original detection frequency axis corresponding to the initial detection data of the multi-frequency eddy current, and calculate the discrete rate of change of the corrected impedance phase angle variable along the original detection frequency axis using the center difference method; Perform cubic spline interpolation on the discrete rate of change to generate the continuously varying phase rate of change parameter.

3. The multi-frequency eddy current impedance spectroscopy based cable alloy defect inversion method of claim 1, wherein, Step S2 includes: By comparing the amplitude distribution pattern of the phase change rate parameter, a preset phase change rate threshold is extracted. The frequency bands where the absolute value of the phase change rate parameter is greater than the preset phase change rate threshold are determined as information-dense frequency bands, and the frequency bands where the absolute value of the phase change rate parameter is less than or equal to the preset phase change rate threshold are determined as information-sparse frequency bands. Obtain preset spatial coordinate stretching coefficients and waveform compression coefficients, configure the spatial coordinate stretching coefficients in the information-dense frequency band, configure the waveform compression coefficients in the information-sparse frequency band, and construct an adaptive mapping function; The discrete detection frequency sequence is imported into the adaptive mapping function for integration to calculate and obtain the virtual depth node; Arrange all the virtual depth nodes according to an increasing numerical rule to generate the virtual depth axis.

4. The method of claim 1, wherein the method is based on a multi-frequency eddy current impedance spectroscopy based cable alloy defect inversion method. Step S3 includes: Extract the virtual depth nodes contained within the virtual depth axis; Based on the virtual depth nodes, the electromagnetic response amplitude is re-indexed and the impedance characteristic phase is integrated to construct a reference data sequence; The reference data sequence is processed by a one-dimensional high-pass filtering algorithm to extract effective feature waveforms. The effective characteristic waveform is combined and mapped with the virtual depth axis to generate the depth domain impedance spectrum.

5. The method of claim 4, wherein the method is based on a multi-frequency eddy current impedance spectroscopy based cable alloy defect inversion method. The step of processing the reference data sequence using a one-dimensional high-pass filtering algorithm to extract effective feature waveforms includes: Get the preset moving average window half-width parameter; Based on the moving average window half-width parameter, a local moving average calculation is performed on the benchmark data sequence to obtain a local baseline value; The effective amplitude data is obtained by subtracting the electromagnetic response amplitude in the reference data sequence from the local baseline value. Summarize all the effective amplitude data to generate the effective characteristic waveform.

6. The multi-frequency eddy current impedance spectroscopy based cable alloy defect inversion method of claim 1, wherein, Step S4 includes: Perform isolated peak detection on the depth domain impedance spectrum and count the extreme values ​​of the number of isolated peaks within the depth domain impedance spectrum. Based on the extreme values ​​of the number of isolated peaks, standard feature basis functions with mismatched topology in the standard defect feature library are removed, and candidate standard feature basis functions are obtained to form a candidate basis function set. Calculate the normalized cross-correlation coefficient between each of the standard feature basis functions in the candidate basis function set and the depth domain impedance spectrum to obtain the similarity matching score; The similarity matching scores corresponding to each of the standard feature basis functions are sorted in descending order, and the standard feature basis function with the highest score is determined as the target standard basis function.

7. The cable alloy defect inversion method based on multi-frequency eddy current impedance spectrum according to claim 1, characterized in that, Step S5 includes: The amplitude weighting coefficient and the spatial offset are initialized based on the peak ratio of the depth domain impedance spectrum to the target standard basis function and the depth coordinates. Substitute the initialized amplitude weighting coefficients, spatial offset, target standard basis function, and depth domain impedance spectrum into the linear superposition equation to construct an overdetermined system of equations. The overdetermined system of equations is iteratively solved using a nonlinear least squares approximation regression fitting algorithm, and the fitting residuals are calculated. When the fitting residual is less than the preset lower limit threshold, the iteration is interrupted, and the converged amplitude weight coefficient and the spatial offset are derived and combined to generate the mapping parameters.

8. The cable alloy defect inversion method based on multi-frequency eddy current impedance spectrum according to claim 1, characterized in that, Step S6 includes: Extract the defect classification labels carried by the target standard basis function, and match them in the calibration reference table to obtain the physical defect types; Extract the amplitude weighting coefficient from the mapping parameters, and obtain the size conversion coefficient from the calibration reference table in combination with the type of physical defect; The absolute geometric dimensions are calculated by multiplying the amplitude weighting coefficient with the size conversion coefficient. Extract the spatial offset from the mapping parameters and convert the spatial offset into the actual burial depth; The quantitative defect evaluation map is generated by summarizing the types of physical defects, the absolute geometric dimensions, and the actual burial depth.

9. The cable alloy defect inversion method based on multi-frequency eddy current impedance spectrum according to claim 1, characterized in that, In step S1, acquiring the initial detection data of the multi-frequency eddy current includes: Control the movement of the eddy current non-destructive testing probe on the surface of the test area of ​​the target cable alloy while maintaining a preset lift-off distance; The excitation coil inside the eddy current nondestructive testing probe is driven to output a sinusoidal excitation current corresponding to the preset discrete detection frequency sequence. The induced voltage signal and the excitation current reference signal are acquired by the detection coil inside the eddy current non-destructive testing probe. The induced voltage signal and the excitation current reference signal are digitally quadratured to generate the initial detection data of the multi-frequency eddy current.

10. A cable alloy defect inversion system based on multi-frequency eddy current impedance spectroscopy, characterized in that, Includes the following modules: The phase change rate parameter generation module acquires the initial detection data of multi-frequency eddy currents, extracts the resistance component and reactance component from the initial detection data of multi-frequency eddy currents, and calculates and generates the phase change rate parameter. The virtual depth axis construction module obtains a preset discrete detection frequency sequence, and performs a spatial projection operation on the discrete detection frequency sequence using the phase change rate parameter to generate a virtual depth axis. The impedance spectrum reconstruction module extracts the electromagnetic response amplitude and impedance characteristic phase from the initial multi-frequency eddy current detection data, maps the electromagnetic response amplitude and impedance characteristic phase to the virtual depth axis, and generates a depth domain impedance spectrum. The target standard basis function screening module obtains a preset standard defect feature library, matches the depth domain impedance spectrum with the standard feature basis functions in the standard defect feature library, and determines the target standard basis function. The parameter solving module uses the target standard basis function, introduces the amplitude weighting coefficient and spatial offset to be solved as unknown parameters, and constructs a linear superposition equation together with the depth domain impedance spectrum. The linear superposition equation is solved to obtain the converged amplitude weighting coefficient and spatial offset, which are then combined as mapping parameters. The quantitative inversion decoding module obtains a preset calibration reference table, uses the calibration reference table to perform quantitative inversion decoding on the mapping parameters, and generates a quantitative defect evaluation map.

Citation Information

Patent Citations

  • Cable microstructure lossless evaluation method and system based on AC impedance spectroscopy

    CN122109212A