A method and system for detecting internal defects in copper wire with eddy current

Abstract

The present application relates to the technical field of defect detection, in particular to a copper wire internal defect detection method and system combined with eddy current. The method comprises: applying a composite excitation superimposed by multi-frequency harmonic signals and collecting eddy current response signals, obtaining an initial time-frequency spectrum matrix through short-time Fourier transform; performing singular value decomposition on the matrix, reconstructing the lift-off correction spectrum matrix after eliminating the main singular value component related to the lift-off effect; calculating the maximum Lyapunov exponent and spectral kurtosis value of the correction matrix as the first and second defect characteristic quantities respectively; constructing a two-dimensional feature plane based on the two characteristic quantities, and calculating the angle correction amount using the energy of the eliminated component to adjust the defect judgment region, and finally combining the Manhattan distance and the adjusted angle interval to judge the defect. That is, the scheme of the present application can effectively suppress the interference of the lift-off effect, construct a dynamic criterion, and realize high-precision and high-reliability detection of the internal weak defects of the copper wire.

Description

Technical Field

[0001] This invention relates to the field of defect detection technology. More specifically, this invention relates to a method and system for detecting internal defects in copper wires using eddy currents. Background Technology

[0002] Eddy current testing, as a non-contact, high-efficiency, and highly sensitive non-destructive testing method, is widely used in defect detection of metallic conductors such as copper wires. During online or offline testing of copper wires, a slight distance change inevitably occurs between the probe and the surface of the copper wire, resulting in a lift-off effect. This physical phenomenon generates interference signals similar in characteristics to defect signals, but with much larger amplitudes than the actual defect signals. This creates strong background noise that drowns out the weak internal defect signals, which is the root cause of the problem affecting the signal-to-noise ratio and accuracy of the test.

[0003] To address the aforementioned issues, existing technologies have made some attempts. On the one hand, methods such as multi-frequency detection and phase analysis have been employed to directly suppress the lift-off effect; however, these methods have limited suppression effectiveness or struggle to effectively distinguish specific types of defects from lift-off signals. On the other hand, at the signal processing and defect identification level, traditional methods often rely on analyzing amplitude or phase information at a single frequency, resulting in limited information dimensions and weak anti-interference capabilities. To obtain richer information, time-frequency analysis methods such as short-time Fourier transform have been introduced, capable of simultaneously displaying the signal's distribution characteristics in both the time and frequency domains; however, background noise caused by the lift-off effect remains the dominant component in the time spectrum.

[0004] In summary, existing technologies face the following pressing problems. First, even with obtained time-frequency spectra, directly extracting conventional features such as energy and peak values ​​is problematic because these features are highly dependent on signal-to-noise ratios and lack robustness. They are easily affected by signals contaminated by lift-off effects, leading to feature instability. Second, existing discrimination criteria have poor adaptability and cannot dynamically adjust according to the strength of the lift-off effect. This results in false alarms when the lift-off effect is large, while small defects may be missed when the lift-off effect is small. This is especially true during high-speed detection, where the lift-off effect changes more drastically and irregularly, posing a significant challenge to stable detection. Therefore, extracting robust features that stably characterize internal defects from complex time-frequency information and establishing intelligent defect judgment criteria that can adapt to lift-off changes are current technical challenges. Summary of the Invention

[0005] The purpose of this invention is to propose a method and system for detecting internal defects in copper wires by combining eddy currents, in order to solve the technical problems of existing technologies having a single information dimension, weak anti-interference ability, and difficulty in fully reflecting the intrinsic characteristics of defects; to this end, this invention provides solutions in the following two aspects.

[0006] In a first aspect, the present invention provides a method for detecting internal defects in copper wires using eddy currents, comprising the following steps: A composite excitation signal consisting of multiple preset frequency harmonic signals is applied to a copper wire using a probe coil, and the corresponding eddy current response signal is acquired. Based on the time-domain data of the eddy current response signal, an initial time-frequency spectrum matrix is ​​obtained through short-time Fourier transform. Singular value decomposition is performed on the initial time-frequency spectrum matrix. The first principal singular value component with the largest amplitude related to the lift-off effect is identified and removed, and the energy of the removed component is used as the lift-off energy index. The matrix is ​​reconstructed based on the remaining singular value components to obtain the lift-off correction spectrum matrix. The spectral vectors at each time point in the lift-off correction spectrum matrix are used to construct a high-dimensional phase space trajectory, and the high-dimensional phase space trajectory is calculated. The maximum Lyapunov exponent of the spatial trajectory is used as the first defect feature; simultaneously, the kurtosis value of the time-averaged spectral energy of the lift-off correction spectrum matrix with respect to the frequency distribution is calculated as the second defect feature; a two-dimensional feature plane is constructed based on the first and second defect features; according to the lift-off energy index, the angle correction amount used to adjust the judgment region is calculated; when the data points on the feature plane satisfy the condition that the Manhattan distance from the data point to the origin is greater than a preset distance threshold, and the angle between the data point and the coordinate axis of the first defect feature is within the preset defect interval adjusted by the angle correction amount, it is determined that there is a defect inside the copper wire.

[0007] Preferably, the probe coil is a through-type differential probe coil.

[0008] Preferably, the step of applying a composite excitation signal composed of multiple preset frequency harmonic signals to the copper wire using the probe coil includes: acquiring five harmonic signal frequencies, namely 5kHz, 10kHz, 20kHz, 50kHz and 100kHz; linearly superimposing the five frequency harmonic signals with equal amplitude to generate the composite excitation signal and outputting it to the probe coil by the signal generator.

[0009] Preferably, performing singular value decomposition on the initial time spectrum matrix includes: decomposing the initial time spectrum matrix... Decomposed into ,in and It is a unitary matrix. It is a diagonal matrix, and the diagonal elements are singular values ​​arranged in descending order. The first principal singular value component is the same as the maximum singular value. Corresponding components , unitary matrix The first column vector, unitary matrix The first column vector.

[0010] Preferably, the calculation of the maximum Lyapunov exponent of the high-dimensional phase space trajectory as the first defect feature quantity includes: using the Wolf algorithm to calculate, selecting an initial point on the phase space trajectory, finding the nearest neighbor of the initial point, and calculating the initial Euclidean distance between the two points. ; a fixed time step for trajectory evolution Calculate the new distance between the two points after the evolution. Preserve the orientation of the initial point after evolution, find a new nearest neighbor within a predetermined cone angle range and replace the previous neighbor, repeat the evolution process, and estimate the maximum Lyapunov exponent by averaging the logarithmic ratios of the new distance to the initial Euclidean distance in all evolution steps along the set length trajectory.

[0011] Preferably, the step of calculating the angle correction amount for adjusting the judgment region based on the lift-off energy index includes: obtaining a reference lift-off energy. and a linear adjustment coefficient The angle correction amount Calculated using the following formula: ,in The lift-off energy index; when the angle between the data point on the feature plane and the coordinate axis of the first defect feature quantity. satisfy Within the adjusted preset defect range, and The initial lower and upper limits.

[0012] Preferably, the short-time Fourier transform includes using a Hanning window with a length of 256 sampling points, and setting the overlap length between window functions to 128 sampling points.

[0013] In the second aspect, the copper wire internal defect detection system incorporating eddy currents includes the following units: The signal acquisition unit applies a composite excitation signal, consisting of multiple preset frequency harmonic signals superimposed, to a copper wire using a probe coil, and acquires the corresponding eddy current response signal. Based on the time-domain data of the eddy current response signal, an initial time-frequency spectrum matrix is ​​obtained through short-time Fourier transform. The spectrum correction unit performs singular value decomposition on the initial time-frequency spectrum matrix, identifies and removes the first principal singular value component with the largest amplitude related to the lift-off effect, and uses the energy of the removed component as the lift-off energy index. Based on the remaining singular value components, the matrix is ​​reconstructed to obtain the lift-off correction spectrum matrix. The feature extraction unit constructs a high-dimensional phase space trajectory from the spectral vectors of each time point in the lift-off correction spectrum matrix. The maximum Lyapunov exponent of the high-dimensional phase space trajectory is calculated as the first defect feature quantity; simultaneously, the kurtosis value of the time-averaged spectral energy of the lift-off correction spectrum matrix with respect to the frequency distribution is calculated as the second defect feature quantity; the defect determination unit constructs a two-dimensional feature plane based on the first and second defect feature quantities; according to the lift-off energy index, the angle correction amount used to adjust the determination region is calculated; when the data points on the feature plane satisfy the condition that the Manhattan distance from the data point to the origin is greater than a preset distance threshold, and the angle between the data point and the coordinate axis of the first defect feature quantity is within the preset defect interval adjusted by the angle correction amount, it is determined that there is a defect inside the copper wire.

[0014] Preferably, the probe coil is a through-type differential probe coil.

[0015] Preferably, the step of applying a composite excitation signal composed of multiple preset frequency harmonic signals to the copper wire using the probe coil includes: acquiring five harmonic signal frequencies, namely 5kHz, 10kHz, 20kHz, 50kHz and 100kHz; linearly superimposing the five frequency harmonic signals with equal amplitude to generate the composite excitation signal and outputting it to the probe coil by the signal generator.

[0016] The beneficial effects of this invention are as follows: By performing singular value decomposition on the signal time spectrum, this invention can separate and eliminate strong interference energy caused by the lift-off effect (change in distance between the probe and the copper wire surface), allowing weak defect information hidden under strong background noise to be clearly presented. Based on interference suppression, by calculating the maximum Lyapunov exponent and spectral kurtosis value, a feature combination more sensitive to defect disturbances is obtained. Simultaneously, the eliminated lift-off energy information is used to dynamically correct the judgment region in the two-dimensional feature plane, constructing a more rigorous defect criterion. This overcomes the shortcomings of traditional fixed threshold methods, which are prone to misjudgment or missed detection under different lift-off states, thereby achieving higher accuracy and reliability in detecting internal defects in copper wires. Attached Figure Description

[0017] Figure 1The flowchart illustrating the steps of the copper wire internal defect detection method incorporating eddy currents in this embodiment is shown schematically. Detailed Implementation

[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0019] like Figure 1 As shown in this embodiment, a method for detecting internal defects in copper wires using eddy currents includes the following steps: Step S1: Apply a composite excitation signal consisting of multiple preset frequency harmonic signals superimposed on the copper wire using a probe coil, and collect the corresponding eddy current response signal; based on the time domain data of the eddy current response signal, obtain the initial time spectrum matrix through short-time Fourier transform.

[0020] Specifically, three sinusoidal signals of 10kHz, 20kHz, and 30kHz are generated by a signal generator and linearly superimposed to form a composite excitation signal. This composite excitation signal drives a through-type differential probe coil after being amplified by a power amplifier. When a copper wire passes through the probe at a constant speed, the induced voltage acquired by the probe is the eddy current response signal. The analog signal is sampled and converted from analog to digital by a data acquisition card to obtain a discrete time-domain data sequence. The time-domain data sequence is segmented, using a Hanning window with a length of 256 sampling points and an overlap length of 128 sampling points between window functions. A Fast Fourier Transform is performed on each windowed data segment. The transformation results of all data segments are arranged in chronological order to form an initial time-spectrum matrix. The rows of the initial time-spectrum matrix represent frequency components, the columns represent time points, and the element values ​​in the matrix are the energy amplitudes at the corresponding time and frequency points.

[0021] In one embodiment, a composite excitation signal consisting of multiple preset frequency harmonic signals is applied to a copper wire using a probe coil, including: acquiring five harmonic signal frequencies, namely 5kHz, 10kHz, 20kHz, 50kHz and 100kHz; linearly superimposing the five frequency harmonic signals with equal amplitude to generate the composite excitation signal, which is then output to the probe coil by a signal generator.

[0022] The purpose of multi-frequency excitation is to utilize the skin effect in eddy current testing, where signals of different frequencies penetrate copper wires to varying depths. For example, a low-frequency signal of 5 kHz can penetrate deeper and is more sensitive to defects inside the copper wire, while a high-frequency signal of 100 kHz is mainly concentrated on the surface of the copper wire, providing better detection of micro-cracks. By combining these frequencies, information about the copper wire at different depths from the surface to the interior can be obtained simultaneously, thus enabling a more comprehensive assessment of the copper wire quality.

[0023] In practice, the signal generator produces five independent standard sinusoidal signals, each with a frequency corresponding to a preset value. To ensure that each frequency component has a significant weight in the detection, their amplitudes are identical, for example, all 1 volt. The five signals are linearly summed through an adder circuit to form a non-sinusoidal periodic waveform. The composite excitation signal is then fed to the probe coil, inducing a composite eddy current field containing multiple frequency components in the copper wire.

[0024] Step S2: Perform singular value decomposition on the initial time spectrum matrix; identify and remove the first principal singular value component with the largest amplitude related to the lift-off effect, and use the energy of the removed component as the lift-off energy index; reconstruct the matrix based on the remaining singular value components to obtain the lift-off correction spectrum matrix.

[0025] Specifically, performing singular value decomposition on the initial time-frequency spectrum matrix includes: decomposing the initial time-frequency spectrum matrix... Decomposed into ,in and It is a unitary matrix. It is a diagonal matrix, and the diagonal elements are singular values ​​arranged in descending order. The first principal singular value component is the same as the maximum singular value. Corresponding components , unitary matrix The first column vector, unitary matrix The first column vector.

[0026] Identification and Maximum Singular Values Corresponding components The first principal singular value component is selected and removed; the remaining singular value components, i.e., those from... arrive of The summation is performed to reconstruct the lift-off correction spectrum matrix.

[0027] During eddy current testing, minute fluctuations in the distance between the probe and the copper wire, known as the lift-off effect, generate an interference signal with a much stronger intensity than the defect signal. This interference occurs in the time-spectrum matrix. In this context, the lift-off disturbance manifests as a component with concentrated energy and slow change. Singular value decomposition can transform the matrix... Different signal modes are separated into different singular value components according to their energy levels. Typically, the largest singular value... For example, the largest singular value is 95.8, while the second largest singular value is... The value dropped to 4.2, indicating the first singular value component. It captured the vast majority of the signal energy, which mainly corresponds to the lift-off effect.

[0028] Therefore, by removing the first principal singular value component with the highest energy from the original decomposition, it is equivalent to removing the lift-off interference at the data level. The remaining singular value components, i.e., those from... arrive All corresponding components are summed to reconstruct a new time-frequency spectrum matrix. The reconstructed matrix is ​​the lift-off correction spectrum matrix, which preserves the weak signal characteristics caused by the defect while eliminating the strong background noise of the lift-off effect.

[0029] Step S3: Construct a high-dimensional phase space trajectory from the spectral vectors at each time point in the lift-off correction spectrum matrix, and calculate the maximum Lyapunov exponent of the high-dimensional phase space trajectory as the first defect feature quantity; at the same time, calculate the kurtosis value of the time-averaged spectral energy of the lift-off correction spectrum matrix with respect to the frequency distribution as the second defect feature quantity.

[0030] Specifically, each column of the extracted correction spectrum matrix, i.e., the spectrum vector at each time point, is taken as a state point in the high-dimensional phase space; the state points at all time points are connected in chronological order to form a phase space trajectory; the Wolf algorithm is used to track the average exponential divergence rate of adjacent points on the phase space trajectory over time, and the maximum Lyapunov exponent is calculated. As the first defect feature quantity; the average value of the lift-off correction spectrum matrix is ​​calculated along the time axis, i.e., column by column, to obtain a single spectrum vector, which represents the average spectrum energy distribution within the detection time period; the kurtosis value K of the average spectrum vector is calculated according to the statistical formula and used as the second defect feature quantity.

[0031] In one embodiment, calculating the maximum Lyapunov exponent of the high-dimensional phase space trajectory as a first defect feature includes: using the Wolf algorithm to select an initial point on the phase space trajectory, finding the nearest neighbor of the initial point, and calculating the initial Euclidean distance between the two points. ; a fixed time step for trajectory evolution Calculate the new distance between the two points after the evolution. Preserve the orientation of the initial point after evolution, find a new nearest neighbor within a predetermined cone angle range and replace the previous neighbor, repeat the evolution process, and estimate the maximum Lyapunov exponent by averaging the logarithmic ratios of the new distance to the initial Euclidean distance in all evolution steps along the set length trajectory.

[0032] Specifically, the time-series signal from eddy current detection is transformed into a motion trajectory in a high-dimensional phase space. The signal generated by defect-free copper wire is deterministic and periodic, and its phase space trajectory exhibits a stable limiting cycle, with neighboring trajectory points not diverging, thus the maximum Lyapunov exponent approaches zero. In contrast, defect signals have abrupt and nonlinear characteristics, and their phase space trajectory exhibits a hybrid nature, meaning that small initial differences are amplified exponentially over time, resulting in a positive maximum Lyapunov exponent.

[0033] In the calculation process, it is assumed that a point is selected from the phase space trajectory. Find the point closest to it. Initial distance The value is 0.005. Let each of these two points evolve along the trajectory for 10 time steps to reach a new position. and Their new distance at this time It becomes 0.015. Find a new neighboring point that satisfies the directional constraints. The evolution process is repeated. This process continues along the entire trajectory, for example, iterating 10,000 times, recording the logarithmic ratio of the distance each time. The average of all these ratios yields the maximum Lyapunov exponent. An exponent value of 0.85 indicates chaotic characteristics in the system, suggesting a defect, while a value close to 0.02 indicates stable signal, corresponding to a defect-free state.

[0034] Step S4: Construct a two-dimensional feature plane based on the first defect feature quantity and the second defect feature quantity; calculate the angle correction amount for adjusting the judgment area according to the lift-off energy index; when the data point on the feature plane satisfies the condition that the Manhattan distance from the data point to the origin is greater than a preset distance threshold, and the angle between the data point and the coordinate axis of the first defect feature quantity is within the preset defect range adjusted by the angle correction amount, it is determined that there is a defect inside the copper wire.

[0035] Specifically, establish a system based on the maximum Lyapunov index. The x-axis represents a two-dimensional feature plane with kurtosis K as the y-axis, and each detection time corresponds to a data point on the plane. In one embodiment, the angle correction amount... By multiplying the extraction energy index by a preset scaling factor Calculations show that a basic defect angle range, such as 30° to 60°, is pre-obtained; an angle correction amount is then added to both the upper and lower limits of the range. This yields the adjusted defect range; simultaneously, a distance threshold is obtained. For data points on the feature plane, calculate the Manhattan distance, i.e. The sum of the absolute values ​​of and K, if the Manhattan distance is less than . If it is greater than 1, it is judged to be defect-free; if it is greater than 1, it is judged to Then, further calculate the angle between the line connecting the data point and the origin and the positive direction of the horizontal coordinate axis. Determine the included angle If the defect falls within the adjusted defect range, it is determined that there is a defect inside the copper wire; otherwise, it is determined that there is no defect.

[0036] In one embodiment, calculating the angle correction amount for adjusting the judgment region based on the lift-off energy index includes: obtaining a reference lift-off energy. and a linear adjustment coefficient The angle correction amount Calculated using the following formula: ,in The lift-off energy index; When the angle between the data points on the feature plane and the coordinate axis of the first defect feature quantity satisfy Within the adjusted preset defect range, and The initial lower and upper limits.

[0037] Although singular value decomposition removes most of the lift-off effect, residual or varying lift-off can still slightly affect the phase angle of defective data points on the feature plane, potentially causing data points to deviate from the fixed decision region boundary. To address this issue, this invention employs an adjustment mechanism for the lift-off energy index. It is calculated using the energy of the first principal singular value component that was removed, reflecting the actual lift-off magnitude at the current detection point. In practical applications, a baseline lift-off energy is obtained by detecting at a standard lift-off distance. For example, for 100 units, the initial defect angle range is obtained as 25° to 45°. Linear adjustment coefficient. Determined based on experimental data, for example Set to 0.1 degrees per energy unit. In a single test, if the measured lift-off energy... The value is 150 units, higher than the baseline, indicating that the probe is farther from the copper wire. The angle correction amount calculated at this point... The angle is 5°. Therefore, the angle range for defect detection is adjusted to 30° to 50°. Consequently, even if the phase of the defect signal shifts slightly due to the lift-off change, it can still be accurately detected.

[0038] This invention also provides an online inspection system for surface defects in automotive parts. It includes the following units: The signal acquisition unit applies a composite excitation signal consisting of multiple preset frequency harmonic signals superimposed on a copper wire using a probe coil, and acquires the corresponding eddy current response signal; based on the time-domain data of the eddy current response signal, the initial time-spectrum matrix is ​​obtained through short-time Fourier transform; The spectrum correction unit performs singular value decomposition on the initial spectrum matrix; identifies and removes the first principal singular value component with the largest amplitude related to the lift-off effect, and uses the energy of the removed component as the lift-off energy index; and reconstructs the matrix based on the remaining singular value components to obtain the lift-off correction spectrum matrix. The feature extraction unit constructs a high-dimensional phase space trajectory from the spectral vectors at each time point in the lift-off correction spectrum matrix, calculates the maximum Lyapunov exponent of the high-dimensional phase space trajectory as the first defect feature quantity, and simultaneously calculates the kurtosis value of the time-averaged spectral energy of the lift-off correction spectrum matrix with respect to the frequency distribution as the second defect feature quantity. The defect determination unit constructs a two-dimensional feature plane based on the first defect feature quantity and the second defect feature quantity; calculates the angle correction amount for adjusting the determination area according to the lift-off energy index; when the data point on the feature plane satisfies the condition that the Manhattan distance from the data point to the origin is greater than a preset distance threshold, and the angle between the data point and the coordinate axis of the first defect feature quantity is within the preset defect range adjusted by the angle correction amount, it determines that there is a defect inside the copper wire.

[0039] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, compact disc read-only memory (CD-ROM), optical storage, etc.) containing computer-usable program code.

[0040] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented by computer-readable / executable instructions stored or otherwise maintained on such a computer-readable medium.

[0041] In the description of this specification, "multiple" means at least two, such as two, three or more, etc., unless otherwise expressly and specifically defined.

[0042] While various embodiments of the invention have been shown and described in this specification, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention.

Claims

1. A method for detecting internal defects in copper wires using eddy currents, characterized in that, Includes the following steps: A composite excitation signal consisting of multiple preset frequency harmonic signals is applied to a copper wire using a probe coil, and the corresponding eddy current response signal is acquired. Based on the time-domain data of the eddy current response signal, the initial time-frequency spectrum matrix is ​​obtained through short-time Fourier transform. Perform singular value decomposition on the initial spectrum matrix; Identify and remove the first principal singular component with the largest amplitude associated with the lift-off effect, and use the energy of the removed component as the lift-off energy index; reconstruct the matrix based on the remaining singular components to obtain the lift-off correction spectrum matrix; The spectral vectors at each time point in the lift-off correction spectrum matrix are used to construct a high-dimensional phase space trajectory. The maximum Lyapunov exponent of the high-dimensional phase space trajectory is calculated as the first defect feature. At the same time, the kurtosis value of the time-averaged spectral energy of the lift-off correction spectrum matrix with respect to the frequency distribution is calculated as the second defect feature. A two-dimensional feature plane is constructed based on the first and second defect feature quantities; Based on the lift-off energy index, calculate the angle correction amount used to adjust the judgment area, including: Obtain a baseline lift-off energy and a linear adjustment coefficient ; Angle correction amount Calculated using the following formula: ， in To improve the energy index; When the angle between the data points on the feature plane and the coordinate axis of the first defect feature quantity satisfy The data point is determined to fall within the adjusted preset defect range, where and These are the initial lower and upper limits, respectively; When a data point on the feature plane satisfies the condition that the Manhattan distance from the data point to the origin is greater than a preset distance threshold, and the angle between the data point and the coordinate axis of the first defect feature quantity is within the preset defect range adjusted by the angle correction amount, it is determined that there is a defect inside the copper wire.

2. The method for detecting internal defects in copper wires using eddy currents according to claim 1, characterized in that, The probe coil is a through-type differential probe coil.

3. The method for detecting internal defects in copper wires using eddy currents according to claim 1, characterized in that, The method of applying a composite excitation signal, consisting of multiple preset frequency harmonic signals superimposed, to the copper wire using a probe coil includes: Five harmonic signal frequencies were obtained: 5kHz, 10kHz, 20kHz, 50kHz, and 100kHz. The five harmonic signals of different frequencies are linearly superimposed with equal amplitude to generate the composite excitation signal, which is then output to the probe coil by the signal generator.

4. The method for detecting internal defects in copper wires using eddy currents according to claim 1, characterized in that, The singular value decomposition of the initial time spectrum matrix includes: The initial time-frequency matrix Decomposed into ,in and It is a unitary matrix. It is a diagonal matrix, and the diagonal elements are singular values ​​arranged in descending order. ; The first principal singular value component is the same as the maximum singular value. Corresponding components , unitary matrix The first column vector, unitary matrix The first column vector.

5. The method for detecting internal defects in copper wires using eddy currents according to claim 1, characterized in that, The calculation of the maximum Lyapunov exponent of the high-dimensional phase space trajectory, as the first defect characteristic quantity, includes: The Wolf algorithm is used for calculation. An initial point is selected on the phase space trajectory, the nearest neighbor of the initial point is found, and the initial Euclidean distance between the two points is calculated. ; a fixed time step for trajectory evolution Calculate the new distance between the two points after the evolution. ; Preserving the orientation of the initial point after evolution, a new nearest neighbor is found within a predetermined cone angle range and replaced with the previous neighbor. The evolution process is repeated, and the maximum Lyapunov exponent is estimated by averaging the logarithmic ratios of the new distance to the initial Euclidean distance in all evolution steps along the set length trajectory.

6. The method for detecting internal defects in copper wires using eddy currents according to claim 1, characterized in that, The short-time Fourier transform includes using a Hanning window with a length of 256 sampling points and setting the overlap length between window functions to 128 sampling points.

7. A system for implementing the copper wire internal defect detection method combined with eddy currents as described in any one of claims 1 to 6, characterized in that, Includes the following units: The signal acquisition unit applies a composite excitation signal consisting of multiple preset frequency harmonic signals superimposed on a copper wire using a probe coil, and acquires the corresponding eddy current response signal; based on the time-domain data of the eddy current response signal, the initial time-spectrum matrix is ​​obtained through short-time Fourier transform; The spectrum correction unit performs singular value decomposition on the initial spectrum matrix; identifies and removes the first principal singular value component with the largest amplitude related to the lift-off effect, and uses the energy of the removed component as the lift-off energy index; and reconstructs the matrix based on the remaining singular value components to obtain the lift-off correction spectrum matrix. The feature extraction unit constructs a high-dimensional phase space trajectory from the spectral vectors at each time point in the lift-off correction spectrum matrix, calculates the maximum Lyapunov exponent of the high-dimensional phase space trajectory as the first defect feature quantity, and simultaneously calculates the kurtosis value of the time-averaged spectral energy of the lift-off correction spectrum matrix with respect to the frequency distribution as the second defect feature quantity. The defect determination unit constructs a two-dimensional feature plane based on the first defect feature quantity and the second defect feature quantity; Calculate the angle correction amount used to adjust the judgment area based on the lift-off energy index; When a data point on the feature plane satisfies the condition that the Manhattan distance from the data point to the origin is greater than a preset distance threshold, and the angle between the data point and the coordinate axis of the first defect feature quantity is within the preset defect range adjusted by the angle correction amount, it is determined that there is a defect inside the copper wire.

Images