Data processing method based on whale optimization algorithm for optimizing variational mode combined with dual-tree complex wavelet

Abstract

The invention belongs to the field of data signal processing, and provides a de-noising method based on combination of a whale optimization algorithm (NSCWOA-VMD, variational mode decomposition) and DTCWT (dual-tree complex wavelet transform) aiming at the problem that noise interference exists in an eddy current detection signal. The method comprises the following steps: firstly, optimizing VMD parameters by using an NSCWOA algorithm, and finding an optimal parameter combination penalty factor alpha and a decomposition modulus K; then, DTCWT is applied to each IMF (intrinsic mode function) component obtained through decomposition, detail coefficients of a real part and an imaginary part of the IMF component are obtained, and signals and noise are distinguished through phase information of complex wavelet coefficients. Through self-adaptive thresholding processing, in combination with local characteristics and noise levels of the signals, noise is effectively suppressed, and key characteristics of the signals are reserved. And finally, reconstructing a final de-noised eddy current signal through inverse transformation of the DTCWT. Compared with a traditional denoising method, the method has the advantages that the signal-to-noise ratio is effectively improved, the root-mean-square error is reduced, and the purpose of accurately denoising the eddy current detection signal is achieved.

Description

Technical Field

[0001] The invention relates to the technical field of eddy current detection data processing systems, and in particular to a denoising signal processing method. Background Art

[0002] As a non-destructive testing technology, eddy current testing has the advantages of non-contact, fast and efficient, applicable to a variety of conductive materials, and high sensitivity. It can detect surface and near-surface defects of ferromagnetic and non-ferromagnetic conductive materials. In addition, it also has good repeatability and reliability, and can provide accurate test results. Therefore, it is widely used in aerospace, electric power, petrochemical and other industries.

[0003] Compared with other detection signals, eddy current signals have rich frequency components, rapid amplitude and phase changes, and obvious transient characteristics. These characteristics make eddy current signals susceptible to various noises, including electromagnetic interference, mechanical vibration, equipment noise, and environmental factors, which leads to a decline in detection quality, difficulty in extracting subtle information, and affects detection accuracy.

[0004] Dual-tree complex wavelet transform (DTCWT) is an improved technology in the field of wavelet analysis. Compared with traditional wavelet transform, DTCWT has the advantages of excellent directional selectivity, approximate translation invariance, approximate suitability for a variety of noise types, good frequency resolution, etc. These advantages enable DTCWT to better retain the key features of the image or signal while removing the eddy current signal noise, thereby effectively eliminating interference, extracting effective signals, and ensuring accurate detection of materials. Summary of the invention

[0005] The present invention aims to solve the above-mentioned difficulties in eddy current detection signal processing, and proposes a data processing method based on the whale optimization algorithm to optimize the variational mode decomposition (VMD) and combined with the dual-tree complex wavelet transform (DTCWT). Specifically, the technical solution of the present invention is as follows: the whale optimization algorithm (NSCWOA) is used to optimize the decomposition layer number and penalty factor of VMD to obtain the best modal component; the optimal parameter combination is found through the optimization algorithm to ensure that the VMD decomposition can more accurately separate the different modes of the signal.

[0006] The optimized modal components are denoised using DTCWT, and the noise is removed by processing the complex wavelet coefficients; the complex wavelet coefficients generated by DTCWT contain the real and imaginary information of the signal. By thresholding these complex wavelet coefficients, the noise can be effectively removed while retaining useful signal features.

[0007] Specifically, the specific steps of the present invention are as follows: S1: Improvement of the whale optimization algorithm, by changing the whale search mechanism and search path to improve the algorithm's global search capability; S2: Use the whale optimization algorithm to optimize the variational modal parameters, the variational modal parameter penalty factor α and the number of decomposition layers K; S3: Denoising by dual-tree complex wavelet; DTCWT uses two independent wavelet transform trees (real tree and imaginary tree) to decompose the original signal into multi-scale wavelet coefficients, which can be expressed as amplitude and phase, separate the real and imaginary parts from the wavelet coefficients, perform adaptive thresholding on each, remove or weaken those coefficients that are considered to be noise, and merge the processed real and imaginary parts into complex wavelet coefficients.

[0008] S1 specifically includes: S11: Improvement of search mechanism: using nonlinear decreasing search factor to fuse sine and cosine strategies; S12: Parameter adjustment: Change shape constants.

[0009] S2 is used to find the optimal parameter combination penalty factor α and decomposition modulus K. S2 specifically includes: S21: Use multi-scale permutation entropy as the fitness function to initialize the optimal fitness; S22: The initial settings for establishing the whale algorithm are as follows: the population size is set to 100; the maximum number of iterations is set to 30; the value range of the penalty factor α is set to [100,3500]; the value range of the decomposition level K is set to [2,12]; S23: For each whale in the population (combination of K, α), use the VMD algorithm to perform modal decomposition and calculate the minimum multi-scale permutation entropy; S24: Update the next generation fitness value according to the optimal fitness value experienced during initialization. If it is higher than the original corresponding whale fitness value, substitute it as the current optimal value position; S25: compare its fitness value with the global optimal value, and if the fitness value is better, replace it again; S26: The iteration stops and the optimal whale position is output as the optimal K and α.

[0010] S3 specifically includes: S31: Perform dual-tree complex wavelet decomposition on each IMF component obtained by VMD decomposition to obtain detail coefficients of the real and imaginary parts and retain the approximation coefficients, and use the phase information of the complex wavelet coefficients to distinguish between noise and useful signal components; S32: Separate the real part and the imaginary part, perform adaptive thresholding on the real part and the imaginary part respectively, the adaptive threshold is dynamically adjusted based on the local characteristics of the signal and the noise level, removes or weakens the coefficients related to the noise, and finally merges the processed real part and imaginary part into the logarithmic wavelet coefficients.

[0011] S4: Reconstruct the denoised eddy current signal; perform inverse DTCWT transformation on the wavelet coefficients after thresholding to reconstruct the denoised eddy current signal.

[0012] The present invention is a denoising method designed according to the characteristics of eddy current detection signals. It applies several advantages of double-tree complex wavelets and corresponds to the characteristics of eddy current detection signals to better eliminate the influence of noise on eddy current signals. Compared with other denoising methods, the denoising effect is more obvious, which lays an important foundation for the subsequent identification of eddy current detection defects. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 A flow chart of the data processing method for optimizing variational mode joint dual-tree complex wavelet based on the whale optimization algorithm; Figure 2 To optimize the whale algorithm, the IMF components of the decomposed eddy current signal of VMD are optimized; Figure 3 This is a comparison chart of soft threshold denoising, hard threshold denoising, DTCWT denoising, and NSCWOA+DTCWT denoising signals. DETAILED DESCRIPTION

[0014] The present invention is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0015] like Figure 1 As shown, this embodiment provides a data processing method based on whale optimization algorithm to optimize variational mode joint dual-tree complex wavelet, including the following steps: S1: Improvement of whale optimization algorithm; The analysis found that the whale algorithm still has certain defects in convergence speed and global search ability. The algorithm's optimization ability can be improved by combining the improvement of the search mechanism and parameter adjustment. In terms of search mechanism improvement, a nonlinear decreasing search factor is used to fuse the sine and cosine strategies. The variable spiral position is changed in the parameter adjustment position; the b constant controls the spiral shape. The parameter b is redesigned to increase the whale's ability to explore unknown areas, thereby improving the algorithm's global search capability; S2: Use the whale optimization algorithm to optimize the variational mode decomposition and find the optimal parameter combination penalty factor α and decomposition modulus K; S2 specific steps: S21: Use multi-scale permutation entropy as the fitness function to initialize the optimal fitness; S22: The initial settings for establishing the whale algorithm are as follows: the population size is set to 100; the maximum number of iterations is set to 30; the value range of the penalty factor α is set to [100,3500]; the value range of the decomposition level K is set to [2,12]; S23: For each whale in the population (combination of K, α), use the VMD algorithm to perform modal decomposition and calculate the minimum multi-scale permutation entropy; S24: Update the next generation fitness value according to the optimal fitness value experienced during initialization. If it is higher than the original corresponding whale fitness value, substitute it as the current optimal value position; S25: compare its fitness value with the global optimal value, and if the fitness value is better, replace it again; S26: The iteration stops and the optimal whale position is output as the optimal K and α; Figure 2 To obtain the optimal K and α values ​​of VMD, decompose the IMF components; S3: Denoising by dual-tree complex wavelet thresholding method; S3 specifically includes: S31: performing dual-tree complex wavelet decomposition on each IMF component obtained by VMD decomposition to obtain detail coefficients of the real part and the imaginary part and retain the approximation coefficients; S32: Separate the real part and the imaginary part, perform adaptive thresholding on the real part and the imaginary part respectively, calculate a suitable threshold (λ), calculate the adaptive threshold by estimating the standard deviation of the noise (σ), and calculate the adaptive threshold according to the length of the signal (N) and a constant (α), process the wavelet coefficients by combining soft thresholding and hard thresholding, remove or weaken the coefficients related to the noise, and merge the real part and the imaginary part; S3 specific steps: S31: setting the number of dual-tree complex wavelet decomposition layers and dual-tree wavelet basis, applying DTCWT, and obtaining multiple high-frequency parts of the low-frequency part; S32: Ensure that the IMF length is an even number; S33: performing dual-tree complex wavelet decomposition on the IMF component to decompose the original signal into complex wavelet coefficients; S34: Noise standard deviation estimation; Use the median absolute deviation (MAD) method to estimate the noise standard deviation of the current IMF component; S34: Calculate the consistency measure of the phase; extract the phase information of the first layer high-frequency wavelet coefficients, and calculate its standard deviation as a measure of phase consistency; S35: Adaptive threshold adjustment: dynamically adjust the threshold according to the result of phase consistency measurement; S36: Signal reconstruction: Use the inverse dual-tree complex wavelet transform to reconstruct the denoised IMF components; S4: Reconstruction of denoised eddy current signal; The thresholded detail coefficients and the retained approximation coefficients are used to perform the inverse transform of DTCWT to obtain the denoised components; the final denoised eddy current signal is reconstructed. Figure 3This is a comparison chart of different effects of soft threshold denoising, hard threshold denoising, DTCWT denoising, and NSCWOA+DTCWT denoising on the same original signal.

[0016] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A data processing method based on whale optimization algorithm to optimize variational mode joint dual-tree complex wavelet, characterized in that: The following steps are involved: S1: Improvement of whale optimization algorithm; S2: Optimize the variational mode decomposition parameters using the whale optimization algorithm; S3: Denoising by dual-tree complex wavelet thresholding method; S4: Reconstruct the denoised eddy current signal.

2. According to claim 1, a data processing method based on whale optimization algorithm to optimize variational mode combined with dual-tree complex wavelet is characterized in that Enhance global search capability and convergence speed. S1 specifically includes: S11: Improvement of search mechanism: using nonlinear decreasing search factor to fuse sine and cosine strategies; S12: Parameter adjustment: variable spiral position update.

3. According to claim 1, a data processing method based on whale optimization algorithm to optimize variational mode combined with dual-tree complex wavelet is characterized in that Find the optimal parameter combination penalty factor α and decomposition modulus K, S2 specific steps: S21: Use multi-scale permutation entropy as the fitness function to initialize the optimal fitness; S22: Establish the initial settings of the whale algorithm, including the population size; the value range of the penalty factor α; the value range of the decomposition level K; S23: For each whale in the population (combination of K, α), use the VMD algorithm to perform modal decomposition and calculate the minimum multi-scale permutation entropy; S24: Update the next generation fitness value according to the initialized optimal fitness value. If it is higher than the original corresponding whale fitness value, substitute it as the current optimal value position; S25: compare its fitness value with the global optimal value, and if the fitness value is better, replace it again; S26: The iteration stops and the optimal whale position is output as the optimal K and α.

4. According to claim 1, a data processing method based on whale optimization algorithm to optimize variational mode joint dual-tree complex wavelet is characterized in that: S3 specifically includes: S31: perform dual-tree complex wavelet decomposition on each IMF component obtained by VMD decomposition, obtain detail coefficients of the real part and the imaginary part and retain the approximation coefficients, and calculate the consistency measure of the phase; S32: Separate the real part and the imaginary part, dynamically adjust the threshold according to the result of the phase consistency measurement, perform adaptive thresholding processing on the real part and the imaginary part respectively, and remove or weaken the coefficients related to the noise.

5. According to claim 1, a data processing method based on whale optimization algorithm to optimize variational mode joint dual-tree complex wavelet is characterized in that: S4 specifically includes: S41: using the thresholded detail coefficients and the retained approximation coefficients to perform an inverse transform of the DTCWT to obtain a denoised component; S42: Reconstruct the final denoised eddy current signal.

Images