Eddy current testing signal depth feature and thermal process parameter based backstepping optimization method

Abstract

The application discloses a back-stepping optimization method based on vortex detection signal depth features and thermal process parameters, relates to the technical field of vortex signal processing, and comprises the following steps: collecting vortex signals of a valve, and performing denoising processing on the vortex signals; performing principal component analysis on the denoised vortex signals by adopting a multi-scale principal component analysis method to obtain a principal component feature matrix; extracting depth feature characteristics and bispectrum characteristics from the denoised vortex signals; combining the extracted depth features, bispectrum characteristics and the principal component feature matrix into a comprehensive feature vector; constructing a valve parameter inversion model based on multivariate linear regression, taking the comprehensive feature vector as input, and inversely outputting valve parameters; establishing a relationship model based on the inversely output valve parameters and thermal process parameters, and inversely optimizing the thermal process parameters by an optimization algorithm. The application realizes accurate inversion of process parameters through vortex signal processing and a machine learning model.

Description

Technical Field

[0001] This invention relates to the field of eddy current signal processing technology, and specifically to a reverse optimization method based on the depth characteristics of eddy current detection signals and thermal process parameters. Background Technology

[0002] Valve is a crucial component of aero-engine piston engines. Hardness testing of valves reflects their resistance to external elastic deformation, plastic deformation, and damage, allowing for the prediction of their overall mechanical properties and helping to reduce the probability of valve train and engine failure during operation. Typically, valves undergo tempering, quenching, and surface hardening treatments during manufacturing to obtain sufficient mechanical and overall mechanical properties. Hardness, as a direct and multi-dimensional physical quantity describing the elasticity, plasticity, strength, and toughness of a component, allows for the assessment of valves' overall mechanical properties through hardness testing, reflecting their resistance to external elastic deformation, plastic deformation, and damage.

[0003] In valve manufacturing, the traditional hardness testing method uses a load test. However, the load test method has problems: first, it requires stopping the machine to take samples during point pressure application and requires specialized fixtures to complete the test, resulting in a long testing cycle and low sorting efficiency; second, the load can damage the surface quality of the sample, leading to low sample reuse rate and safety hazards. Compared with traditional testing methods, eddy current testing has gained wider application due to its advantages of short testing cycle, high efficiency, low cost, and no destructive effect on the surface quality of components.

[0004] Meanwhile, thermal processing parameters affect the microstructure of materials, thereby influencing specific parameters such as electrical conductivity and magnetic permeability. Therefore, a method is needed to inversely deduce these parameters from eddy current signals and optimize thermal processing parameters accordingly. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to achieve a complete mapping from eddy current signals to process parameters through eddy current signal feature extraction and thermal process parameter optimization.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] The inverse optimization method based on the depth characteristics of eddy current detection signals and thermal process parameters includes the following steps:

[0008] Step S1: Acquire the vortex signal of the valve and perform denoising processing on the vortex signal based on wavelet threshold denoising.

[0009] Step S2: Perform principal component analysis on the denoised eddy current signal using multi-scale principal component analysis to obtain the principal component feature matrix;

[0010] Step S3: Extract depth features and bispectral features from the denoised eddy current signal;

[0011] Step S4: Combine the extracted depth features, bispectral features, and principal component feature matrix into a comprehensive feature vector;

[0012] Step S5: Construct a valve parameter inversion model based on multiple linear regression, using the comprehensive feature vector as input, and invert the valve parameters.

[0013] Step S6: Based on the valve parameters output by inversion, establish a relationship model with the thermal process parameters, and optimize the thermal process parameters by back-calculating through optimization algorithms.

[0014] Furthermore, step S1 specifically includes the following steps:

[0015] Step S1.1: Acquire valve vortex signals at different frequencies;

[0016] Step S1.2: Select the Daubechies wavelet as the basis function, set the number of decomposition levels, perform wavelet decomposition on the signal, and obtain the wavelet coefficients of each level;

[0017] Step S1.3: Calculate the adaptive threshold based on the signal-to-noise ratio;

[0018] Step S1.4: Apply the soft thresholding function to process the wavelet coefficients;

[0019] Step S1.5: Reconstruct the eddy current signal using the denoised wavelet coefficients to obtain the denoised eddy current signal.

[0020] Further, in step S1.3, the formula for calculating the adaptive threshold is:

[0021]

[0022] in, This represents the threshold of the j-th level wavelet decomposition. This represents the wavelet decomposition level index. This represents the noise standard deviation estimate of the wavelet coefficients at the j-th level. Indicates signal length. This represents the signal-to-noise ratio estimate for the j-th layer. This represents the signal-to-noise ratio adjustment coefficient;

[0023] In step S1.4, the formula for calculating the soft threshold function is as follows:

[0024]

[0025] in, Represents the wavelet coefficients after denoising. Represents wavelet coefficients, Indicates the coefficient position index. Represents a symbolic function.

[0026] Furthermore, step S2 specifically includes the following steps:

[0027] Step S2.1: Standardize the denoised eddy current signal matrix to obtain a standardized matrix;

[0028] Step S2.2: Construct a multi-scale covariance matrix, which is obtained by weighted summation of covariance matrices at different scales;

[0029] Step S2.3: Perform eigenvalue decomposition on the multi-scale covariance matrix to obtain eigenvalues ​​and eigenvectors;

[0030] Step S2.4: Select the top performers based on their cumulative contribution rate. The eigenvectors corresponding to the largest eigenvalues ​​form the projection matrix;

[0031] Step S2.5: Multiply the normalized matrix and the projection matrix to obtain the principal component feature matrix after dimensionality reduction.

[0032] Furthermore, step S3 specifically includes the following steps:

[0033] Step S3.1: Set a standard signal as a reference signal;

[0034] Step S3.2: Calculate the cross-correlation function between the reference signal and the denoised eddy current signal;

[0035] Step S3.4: Perform bispectral analysis to extract bispectral features, including the mean and variance of the bispectral amplitudes.

[0036] Step S3.4: Perform bispectral analysis and extract bispectral features, wherein the bispectral features include the mean and variance of the bispectral amplitudes.

[0037] Furthermore, step S5 specifically includes the following steps:

[0038] Step S5.1: Establish independent multiple linear regression models for each valve parameter;

[0039] Step S5.2: Train the multiple linear regression model using the training sample set, and estimate the regression coefficients using the least squares method;

[0040] Step S5.3: Use the trained multiple linear regression model to invert valve parameters.

[0041] Furthermore, the valve parameters include electrical conductivity and magnetic permeability.

[0042] Furthermore, step S6 specifically includes the following steps:

[0043] Step S6.1: Establish a relationship model between valve parameters and thermal process parameters, wherein the thermal process parameters include: heat treatment temperature and heat treatment time;

[0044] Step S6.2: Construct an optimization function based on the aforementioned relationship model, with the objective of minimizing valve parameter deviation;

[0045] Step S6.3: Solve the optimization function using an optimization algorithm and output the optimal set of thermal process parameters.

[0046] Further, in step S6.2, the calculation formula for the optimization function is:

[0047]

[0048] in, This indicates the total number of valve parameters that need to be optimized. Indicates valve parameter index, Represents the optimization function. Represents a vector of thermal process parameters. This represents the weight of the i-th valve parameter. This represents the value of the i-th valve parameter output by the relational model. Indicates the target value of the valve parameters. This represents the numerical stability constant.

[0049] Furthermore, in step S6.3, the optimization algorithm employs a sequential quadratic programming method.

[0050] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0051] 1. This invention employs a multi-scale covariance matrix analysis and cross-correlation-bispectral feature fusion strategy, which can stably extract sensitive features characterizing changes in the microstructure of materials even under strong noise environments.

[0052] 2. This invention establishes a nonlinear mapping model of material parameters and thermal process parameters based on physical mechanisms, and realizes the reverse decision-making from eddy current detection signals to thermal process parameters through a multi-objective optimization algorithm.

[0053] 3. This invention constructs a high signal-to-noise ratio signal feature set by integrating improved wavelet denoising, multi-scale principal component analysis and deep feature extraction, and constructs an adaptive threshold function so that the system can automatically adjust the processing parameters according to the signal characteristics.

[0054] 4. This invention achieves quantitative back-inference from detection signals to material properties and then to process parameters by modeling the correlation between deep features and thermal process parameters, providing reliable decision support for intelligent manufacturing. Attached Figure Description

[0055] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0056] Figure 1 This is a flowchart illustrating an embodiment of the present invention;

[0057] Figure 2 This is a schematic diagram of the data flow in an embodiment of the present invention;

[0058] Figure 3 This is a schematic diagram of the eddy current detection principle according to an embodiment of the present invention. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0060] like Figure 1 As shown, the back-calculation optimization method based on the depth characteristics of eddy current detection signals and thermal process parameters includes the following steps:

[0061] Step S1: Acquire the vortex signal of the valve and perform denoising processing on the vortex signal based on wavelet threshold denoising.

[0062] Step S2: Perform principal component analysis on the denoised eddy current signal using multi-scale principal component analysis to obtain the principal component feature matrix;

[0063] Step S3: Extract depth features and bispectral features from the denoised eddy current signal;

[0064] Step S4: Combine the extracted depth features, bispectral features, and principal component feature matrix into a comprehensive feature vector;

[0065] Step S5: Construct a valve parameter inversion model based on multiple linear regression, using the comprehensive feature vector as input, and invert the valve parameters.

[0066] Step S6: Based on the valve parameters output by inversion, establish a relationship model with the thermal process parameters, and optimize the thermal process parameters by back-calculating through optimization algorithms.

[0067] The core idea of ​​wavelet transform is to decompose a signal into multiple scales using a wavelet function with adjustable scale and variable position, thereby analyzing the different frequency components and local features of the signal.

[0068] This invention selects the Daubechies wavelet as the basis function, which has better smoothness and orthogonality, and is widely used in signal analysis and image compression. The basis functions of the Daubechies wavelet can be obtained by solving the wavelet equation.

[0069] The Daubechies wavelet's vanishing moment is set to order 4, which can capture transient changes and local features in eddy current signals, while achieving a good balance between computational complexity and time-frequency localization; the support length is usually twice the vanishing moment. The number of decomposition layers J is determined by the main frequency components of the signal. For eddy current detection signals, the number of decomposition layers is set to 5 or 6, which can cover the entire frequency band from high-frequency noise to low-frequency trends.

[0070] Step S1 specifically includes the following steps:

[0071] Step S1.1: Acquire valve vortex signals at different frequencies;

[0072] Step S1.2: Select the Daubechies wavelet as the basis function, set the number of decomposition levels, perform wavelet decomposition on the signal, and obtain the wavelet coefficients of each level;

[0073] Step S1.3: Calculate the adaptive threshold based on the signal-to-noise ratio;

[0074] Step S1.4: Apply the soft thresholding function to process the wavelet coefficients;

[0075] Step S1.5: Reconstruct the eddy current signal using the denoised wavelet coefficients to obtain the denoised eddy current signal.

[0076] In step S1.3, the formula for calculating the adaptive threshold is:

[0077]

[0078] in, This represents the threshold of the j-th level wavelet decomposition. This represents the wavelet decomposition level index. This represents the noise standard deviation estimate of the wavelet coefficients at the j-th level. Indicates signal length. This represents the signal-to-noise ratio estimate for the j-th layer. This represents the signal-to-noise ratio adjustment coefficient;

[0079] In step S1.4, the formula for calculating the soft threshold function is as follows:

[0080]

[0081] in, Represents the wavelet coefficients after denoising. Represents wavelet coefficients, Indicates the coefficient position index. Represents a symbolic function.

[0082] Step S2 specifically includes the following steps:

[0083] Step S2.1: Standardize the denoised eddy current signal matrix to obtain a standardized matrix;

[0084] Step S2.2: Construct a multi-scale covariance matrix. The multi-scale covariance matrix is ​​obtained by weighted summation of the covariance matrices at different scales. The specific formula is as follows:

[0085]

[0086] in, Represents the multi-scale covariance matrix. Let represent the covariance matrix at the s-th scale. Indicates scale index. Indicates the total number of scales. The weight of the s-th scale is specifically the ratio of the largest eigenvalue of the covariance matrix at the s-th scale to the sum of the largest eigenvalues ​​of the covariance matrices at all scales.

[0087] The scale is determined by the number of wavelet decomposition levels. A signal at one scale corresponds to the number of wavelet decomposition levels. After performing J levels of wavelet decomposition, J approximate coefficients at different scales are obtained. These, along with the original signal, result in a total of J+1 scales, which is the total number of scales. ;

[0088] For the s-th scale, represent the signals of all samples at that scale as a matrix, and independently calculate the covariance matrix of the signal matrix at each scale, thus obtaining... ;

[0089] Step S2.3: Perform eigenvalue decomposition on the multi-scale covariance matrix to obtain eigenvalues ​​and eigenvectors;

[0090] Step S2.4: Select the top performers based on their cumulative contribution rate. The eigenvectors corresponding to the largest eigenvalues ​​form the projection matrix, and the specific formula is as follows:

[0091]

[0092] in, Indicates the number of principal components selected. represents the candidate variable, and represents the number of principal components assumed to be selected in the formula. Represents all conditions that are met. The set, This represents the total number of original features. This represents the x-th eigenvalue. Indicates the feature index;

[0093] The purpose of the above formula is to find the minimum number of principal components. This makes the former The cumulative variance contribution rate of each principal component reaches or exceeds the preset threshold, i.e., 95%;

[0094] Step S2.5: Multiply the normalized matrix and the projection matrix to obtain the dimensionality-reduced principal component score matrix.

[0095] Step S3 specifically includes the following steps:

[0096] Step S3.1: Set a standard signal as a reference signal;

[0097] The standard signal is an ideal impedance signal calculated by establishing a physical model of the interaction between the vortex field and the material, such as the Dodd-Deeds model, with known material parameters and geometric dimensions as input. The reference signal is used to provide a benchmark, and the difference between the standard signal and the reference signal reflects the deviation of the valve under test from the material properties or microstructure.

[0098] Step S3.2: Calculate the cross-correlation function between the reference signal and the denoised eddy current signal;

[0099] Cross-correlation function is a concept in signal analysis that represents the degree of correlation between two time series, that is, it describes the degree of correlation between the values ​​of a signal at any two different times.

[0100] Step S3.3: Extract the peak position and peak amplitude from the cross-correlation function as the depth features of the eddy current signal, wherein the peak amplitude reflects the maximum similarity of the signal and the peak position reflects the phase difference information of the signal;

[0101] Step S3.4: Perform bispectral analysis to extract bispectral features, including the mean and variance of the bispectral amplitudes. The specific formula is as follows:

[0102]

[0103] in, This represents the bispectral value, reflecting the phase coupling strength between two frequencies. and These represent the frequencies of the preprocessed eddy current signals, and These represent the preprocessed eddy current signals at frequencies of [frequency range missing]. and The Fourier transform result at the point, This indicates the preprocessed eddy current signal at a frequency of and The complex conjugate of the Fourier transform at the sum of the terms. Represents the mathematical expectation;

[0104] Bispectral analysis specifically includes: traversing all possible frequency combinations between 0 and the Nyquist frequency with a fixed frequency resolution, and calculating the bispectral value for each combination according to the above formula.

[0105] Step S5 specifically includes the following steps:

[0106] Step S5.1: Establish independent multiple linear regression models for each valve parameter;

[0107] Step S5.2: Train the multiple linear regression model using the training sample set, and estimate the regression coefficients using the least squares method;

[0108] Step S5.3: Use the trained multiple linear regression model to invert valve parameters.

[0109] The valve parameters include electrical conductivity and magnetic permeability.

[0110] The specific formula for the multiple linear regression model is as follows:

[0111]

[0112] in, , and These represent electrical conductivity, magnetic permeability, and excitation frequency, respectively. Indicates the first The intercept term of the model, Indicates the first In the model, the first The regression coefficients of each feature, Represents the first in the comprehensive feature vector One characteristic, Indicates the first The random error term of the model, This represents the total dimension of the integrated feature vector.

[0113] Using a training dataset containing sample integrated feature vectors and corresponding real valve parameter values, the regression coefficients are directly estimated using the least squares method.

[0114] Step S6 specifically includes the following steps:

[0115] Step S6.1: Establish a relationship model between valve parameters and thermal process parameters, wherein the thermal process parameters include: heat treatment temperature and heat treatment time;

[0116] Step S6.2: Construct an optimization function based on the aforementioned relationship model, with the objective of minimizing valve parameter deviation;

[0117] Step S6.3: Solve the optimization function using an optimization algorithm and output the optimal set of thermal process parameters.

[0118] The calculation formula for the relational model is as follows:

[0119]

[0120]

[0121] in, This represents the electrical conductivity of the material at heat treatment temperature T and time t. Indicates reference conductivity. This represents the maximum change in conductivity. This represents the time constant of temperature-dependent conductivity changes. The Avrami exponent represents the change in electrical conductivity. This represents the material's magnetic permeability at heat treatment temperature T and time t. Indicates the reference permeability. This represents the maximum change in permeability. Indicates the strength of the applied magnetic field. This represents the coercive force field related to thermal process parameters. Indicates the shape parameters of the magnetization curve;

[0122] The specific formula for temperature-dependent conductivity is as follows:

[0123]

[0124] in, The pre-exponential factor representing the change in conductivity. The activation energy represents the change in conductivity. Represents the gas constant. This indicates the heat treatment temperature.

[0125] The formula for calculating the coercive force field related to thermal process parameters is as follows:

[0126]

[0127] in, Indicates the reference coercive field strength. This represents the coefficient of influence of temperature on coercivity. Indicates the Curie temperature of the material. Indicates the temperature dependence index of coercivity. The time constant represents the characteristic change in magnetic properties. It represents the kinetic index of changes in magnetic properties.

[0128] Temperature parameters such as heat treatment temperature Curie temperature of materials All units are Kelvin.

[0129] The above parameters need to be obtained by fitting actual thermal process experimental data. Among them, the electrical conductivity model is based on diffusion-controlled phase transition dynamics, and the magnetic permeability model is based on magnetic domain theory and microstructure evolution.

[0130] In step S6.2, the calculation formula for the optimization function is as follows:

[0131]

[0132] in, This indicates the total number of valve parameters that need to be optimized. Indicates valve parameter index, Represents the optimization function. Represents a vector of thermal process parameters. This represents the weight of the i-th valve parameter. This represents the value of the i-th valve parameter output by the relational model. Indicates the target value of the valve parameters. This represents the numerical stability constant, typically taken as 10. -10 .

[0133] Weighting coefficient This reflects the importance of each valve parameter in optimization, determined through sensitivity analysis. Specifically, this includes: obtaining the degree of impact of valve parameter changes on product performance indicators through experiments or simulations, i.e., sensitivity; and assigning weights based on sensitivity, with parameters having higher sensitivity receiving greater weights. For example, if conductivity's sensitivity to product performance is twice that of permeability, then conductivity's weight is set to twice that of permeability. Specific weight values ​​can be adjusted according to actual process requirements, and the sum of the weights of all valve parameters is 1.

[0134] To ensure that the thermal process parameters meet the actual production conditions and process safety requirements, the constraints mainly include the allowable range constraints of the thermal process parameters and the process stability constraints.

[0135] Among them, the allowable range constraint of thermal process parameters limits the boundary values ​​of the two process parameters, heat treatment temperature and time, to ensure that they are within an industrially feasible, safe and effective range.

[0136] The process stability constraint limits the instantaneous change rate of material conductivity and magnetic permeability caused by changes in heat treatment temperature, thus ensuring stable temperature changes during heat treatment.

[0137] The specific values ​​of the constraints are set according to the actual production conditions and process safety requirements.

[0138] In step S6.3, the optimization algorithm adopts the sequential quadratic programming method.

[0139] The sequential quadratic programming method is a numerical iterative algorithm for solving optimization problems with nonlinear constraints. The specific steps of this algorithm include:

[0140] 1. Perform algorithm initialization: Set the initial process parameter values ​​for thermal processing temperature and time, and select appropriate initial Lagrange multipliers;

[0141] 2. Enter the iterative solution loop: In each iteration, construct the Lagrangian function at the current iteration point. This function integrates the original objective function and all constraints. The original nonlinear objective function is approximated by a quadratic function, and the nonlinear constraints are approximated by linear constraints.

[0142] 3. Perform step size calculation: Use the Armijo line search technique to determine the appropriate step size along the search direction to ensure that the objective function has sufficient descent.

[0143] 4. Perform a convergence check: Determine whether the current solution meets the convergence conditions, including whether the change in the objective function, the change in the variables, and the degree of constraint violation are all less than the preset tolerance; if the convergence conditions are met, output the current thermal process parameters as the optimal solution; otherwise, proceed to the next iteration.

[0144] The entire iterative process continues until the optimal combination of thermal process parameters that satisfies all process constraints and makes the valve parameters closest to the target value is found, or until the maximum number of iterations is reached.

[0145] like Figure 2 As shown, the original signal acquired by the multi-frequency eddy current detector is processed by a soft thresholding function based on an adaptive threshold to effectively separate noise from useful signals, resulting in a high-quality denoised signal. Then, parallel three-way feature extraction is performed:

[0146] Multi-scale principal component analysis: The denoised signal is subjected to multi-scale covariance matrix construction and eigenvalue decomposition to generate a dimensionality-reduced principal component score matrix;

[0147] Cross-correlation analysis: Using the standard signal as a reference, calculate the cross-correlation function between the denoised signal and the reference signal, and extract the peak position and peak amplitude as depth features;

[0148] Bispectral analysis: Analyzing a signal to extract its bispectral features;

[0149] The final step is feature fusion, where principal component features, depth features, and bispectral features are standardized and then concatenated into a comprehensive feature vector.

[0150] The comprehensive feature vector is input into two pre-trained multiple linear regression models, one model responsible for inverting the electrical conductivity of the material and the other for inverting the magnetic permeability. The inverted electrical and magnetic permeabilities are used as inputs and substituted into a pre-established physical mechanism-based relational model. With the relational model as the core, an optimization function is constructed to make the predicted material parameters approximate the target value. The sequential quadratic programming method is used to search within the feasible region of the heat treatment parameters to find the optimal combination of thermal process parameters that minimizes the objective function. The optimal heat treatment temperature and time parameters output by the system can be directly used to guide and adjust the actual production process, thereby achieving precise and proactive control of valve product quality.

[0151] The examples described herein are merely preferred embodiments of the invention and are not intended to limit the concept and scope of the invention. Any modifications and improvements made by those skilled in the art to the technical solutions of the invention without departing from the design concept of the invention should fall within the protection scope of the invention.

Claims

1. A back-calculation optimization method based on the depth characteristics of eddy current detection signals and thermal process parameters, characterized in that, Includes the following steps: Step S1: Acquire the vortex signal of the valve and perform denoising processing on the vortex signal based on wavelet threshold denoising. Step S2: Perform principal component analysis on the denoised eddy current signal using multi-scale principal component analysis to obtain the principal component feature matrix; Step S3: Extract depth features and bispectral features from the denoised eddy current signal; Step S4: Combine the extracted depth features, bispectral features, and principal component feature matrix into a comprehensive feature vector; Step S5: Construct a valve parameter inversion model based on multiple linear regression, using the comprehensive feature vector as input, and invert the valve parameters. Step S6: Based on the valve parameters output by inversion, establish a relationship model with the thermal process parameters, and optimize the thermal process parameters by back-calculating through optimization algorithms; Specifically, step S3 includes the following steps: Step S3.1: Set a standard signal as a reference signal; Step S3.2: Calculate the cross-correlation function between the reference signal and the denoised eddy current signal; Step S3.3: Extract the peak position and peak amplitude from the cross-correlation function as the depth features of the eddy current signal; Step S3.4: Perform bispectral analysis to extract bispectral features, including the mean and variance of the bispectral amplitudes; Step S6 specifically includes the following steps: Step S6.1: Establish a relationship model between valve parameters and thermal process parameters, wherein the thermal process parameters include: heat treatment temperature and heat treatment time; Step S6.2: Construct an optimization function based on the aforementioned relationship model, with the objective of minimizing valve parameter deviation; Step S6.3: Solve the optimization function using an optimization algorithm and output the optimal set of thermal process parameters; In step S6.2, the calculation formula for the optimization function is as follows: in, This indicates the total number of valve parameters that need to be optimized. Indicates valve parameter index, Represents the optimization function. Represents a vector of thermal process parameters. This represents the weight of the i-th valve parameter. This represents the value of the i-th valve parameter output by the relational model. Indicates the target value of the valve parameters. This represents the numerical stability constant.

2. The method according to claim 1, characterized in that, Step S1 specifically includes the following steps: Step S1.1: Acquire valve vortex signals at different frequencies; Step S1.2: Select the Daubechies wavelet as the basis function, set the number of decomposition levels, perform wavelet decomposition on the signal, and obtain the wavelet coefficients of each level; Step S1.3: Calculate the adaptive threshold based on the signal-to-noise ratio; Step S1.4: Apply the soft threshold function to process the wavelet coefficients; Step S1.5: Reconstruct the eddy current signal using the denoised wavelet coefficients to obtain the denoised eddy current signal.

3. The method according to claim 2, characterized in that, In step S1.3, the formula for calculating the adaptive threshold is: in, This represents the threshold of the j-th level wavelet decomposition. This represents the wavelet decomposition level index. This represents the noise standard deviation estimate of the wavelet coefficients at the j-th level. Indicates signal length. This represents the signal-to-noise ratio estimate for the j-th layer. This represents the signal-to-noise ratio adjustment coefficient; In step S1.4, the formula for calculating the soft threshold function is as follows: in, Represents the wavelet coefficients after denoising. Represents wavelet coefficients, Indicates the coefficient position index. Represents a symbolic function.

4. The method according to claim 3, characterized in that, Step S2 specifically includes the following steps: Step S2.1: Standardize the denoised eddy current signal matrix to obtain a standardized matrix; Step S2.2: Construct a multi-scale covariance matrix, which is obtained by weighted summation of covariance matrices at different scales; Step S2.3: Perform eigenvalue decomposition on the multi-scale covariance matrix to obtain eigenvalues ​​and eigenvectors; Step S2.4: Select the top performers based on their cumulative contribution rate. The eigenvectors corresponding to the largest eigenvalues ​​form the projection matrix; Step S2.5: Multiply the normalized matrix and the projection matrix to obtain the principal component feature matrix after dimensionality reduction.

5. The method according to claim 4, characterized in that, Step S5 specifically includes the following steps: Step S5.1: Establish independent multiple linear regression models for each valve parameter; Step S5.2: Train the multiple linear regression model using the training sample set, and estimate the regression coefficients using the least squares method; Step S5.3: Use the trained multiple linear regression model to invert valve parameters.

6. The method according to claim 5, characterized in that, The valve parameters include electrical conductivity and magnetic permeability.

7. The method according to claim 6, characterized in that, In step S6.3, the optimization algorithm adopts the sequential quadratic programming method.

Images