Lithium battery defect signal extraction method based on improved WOA-VMD

Abstract

The invention discloses a lithium battery defect signal extraction method based on improved WOA-VMD, and belongs to the technical field of lithium battery nondestructive testing and signal processing. According to the method, aiming at the problems of low signal-to-noise ratio after ultrasonic signal penetration and difficulty in defect feature extraction caused by an internal multi-layer structure during air coupling ultrasonic detection of a lithium battery, a cosine convergence factor and a nonlinear inertia weight strategy are introduced to improve a traditional whale optimization algorithm; key parameters of variational mode decomposition are optimized in a self-adaptive mode; variational mode decomposition is carried out on the noisy ultrasonic signal by using the optimized parameters, and dominant intrinsic mode function components are screened out according to a correlation coefficient threshold value; and carrying out wavelet threshold denoising on the screened components, and finally reconstructing to obtain a defect characteristic signal with a high signal-to-noise ratio. The method has high adaptability and robustness, is suitable for defect detection of lithium batteries of different specifications, and provides an effective technical means for safety and reliability evaluation of the lithium batteries.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of nondestructive testing of lithium batteries, in particular to a lithium battery defect signal extraction method based on improved WOA-VMD. BACKGROUND

[0002] In recent years, lithium batteries have become the core components of new energy storage systems due to their high energy density, long service life and good environmental compatibility, and are involved in fields such as portable electronic devices, new energy vehicles, national defense and military industry, aerospace and the like. However, in the manufacturing and use process of lithium batteries, problems such as electrolyte unsoaking, separator failure and internal gas production seriously affect the energy conversion efficiency and service life of lithium batteries, and more seriously, may cause safety accidents such as explosion and deflagration, so fault diagnosis of internal defects of the battery is particularly important. Air-coupled ultrasonic testing as a non-contact and nondestructive testing method is suitable for characterization of internal defects of lithium batteries. However, due to the multi-layer winding structure inside the lithium battery, the ultrasonic signal is easily disturbed by scattering and noise during penetration, resulting in low signal-to-noise ratio of the received signal and difficulty in extracting defect features.

[0003] Traditional noise reduction methods such as Fourier transform and adaptive filtering perform relatively well when processing stationary signals, but when facing typical non-stationary, nonlinear and complex signals such as air-coupled ultrasonic signals, it is often difficult to balance noise reduction effect and signal detail preservation. Although wavelet analysis can use a wavelet function that can be translated and scaled to effectively suppress noise components in sudden signals, its performance is largely dependent on the selection of the wavelet basis function and its threshold parameters, and improper selection may lead to problems such as difficulty in signal recognition and loss of effective components. Empirical mode decomposition and its improved algorithm CEEMD do not rely on pre-set basis functions and can decompose complex signals into a series of intrinsic mode functions with different frequencies, but are sensitive to noise and sampling, and are prone to modal aliasing when decomposing high-frequency noise or similar frequency components, resulting in boundary distortion and introduction of false frequency components. Variational mode decomposition can effectively suppress modal aliasing by constraining the variational model to decompose signals, but its performance is affected by the number of modes K and the penalty factor, and traditional methods rely on empirical settings, which is difficult to achieve optimal performance.

[0004] Therefore, there is an urgent need for a method that can adaptively optimize VMD parameters, effectively extract lithium battery defect signals and suppress noise. SUMMARY

[0005] The purpose of the present application is to provide a lithium battery defect signal extraction method based on improved WOA-VMD to solve the problems of low signal-to-noise ratio of air-coupled ultrasonic signals and difficulty in feature extraction.

[0006] To this end, the application provides a lithium battery defect signal extraction method based on improved WOA-VMD, comprising the following steps: S1, collecting the air-coupled ultrasonic penetration signal of the lithium battery; S2, based on the collected ultrasonic penetration signal, using the improved whale optimization algorithm WOA to optimize the parameters K and of the variational mode decomposition VMD, taking the minimum envelope entropy as the fitness evaluation criterion; S3, using the optimized parameters K and to perform VMD decomposition on the input signal to obtain a plurality of intrinsic mode function IMF components; S4, calculating the correlation coefficient of each IMF component and the original signal, and dividing the components into effective components and irrelevant components according to a preset threshold; S5, performing wavelet threshold denoising on the screened effective components; S6, reconstructing the denoised components to obtain the processed defect signal.

[0007] Preferably, in the step S2, the improved whale optimization algorithm WOA introduces a cosine convergence factor and an inertia weight strategy in the traditional whale optimization algorithm: The expression of the cosine convergence factor is: , wherein t is the current iteration number, and T is the maximum iteration number; The expression of the inertia weight strategy is: .

[0008] Preferably, the initial whale population is set, the population size is 50, the maximum iteration number is 30, the VMD parameters K and are defined as the position vectors of the whale individuals, the whale individual positions are updated by iteration, and the parameter combination that minimizes the envelope entropy is found .

[0009] Preferably, in the improved whale optimization algorithm, the fitness function is the minimum envelope entropy of each modal component after variational mode decomposition, and the calculation formula of the envelope entropy is , wherein , is the envelope signal after Hilbert demodulation, and N is the signal length.

[0010] Preferably, in the step S4, the preset correlation coefficient threshold is 0.6, the components with a correlation coefficient greater than the threshold are regarded as effective components, and the rest are regarded as noise components.

[0011] Preferably, in the step S5, the wavelet threshold denoising adopts a soft threshold or a hard threshold function, and the threshold is adaptively determined according to the signal variance and length, and the expressions of the soft threshold and the hard threshold function are respectively: ​​, wherein is a decomposition scale, is a coefficient label, and is a threshold value, is a variance, and is a signal length.

[0012] Preferably, the wavelet threshold denoising specifically comprises the following steps: (501) performing wavelet multi-scale decomposition on each dominant signal component to obtain corresponding wavelet coefficients; (502) calculating an adaptive threshold value based on the wavelet coefficients; (503) performing threshold processing on the wavelet coefficients using a soft threshold function or a hard threshold function; (504) performing inverse wavelet transformation on the processed wavelet coefficients to obtain a denoised modal component.

[0013] Preferably, the input signal is a lithium battery defect signal collected by an air-coupled ultrasonic detection system, and the lithium battery defect signal is a transmission signal when a lithium battery has an internal defect. The air-coupled ultrasonic detection system integrates an air-coupled ultrasonic excitation, reception, signal collection, processing and analysis of an automatic nondestructive testing platform, which is used for non-contact, non-destructive detection of internal defects of lithium ion batteries (especially soft package batteries), such as bubbles, delamination, electrolyte non-wetting, and electrode sheet wrinkles. The core of the system is based on an improved whale optimization algorithm to optimize the signal processing method of the variational mode decomposition, aiming to extract the waveform information reflecting the defect characteristics from the original ultrasonic signal with extremely low signal-to-noise ratio with high fidelity.

[0014] Preferably, the method further comprises the steps of feature extraction and defect recognition of the reconstructed signal, specifically comprising calculating the time domain features, frequency domain features or time-frequency features of the reconstructed signal, and realizing the recognition of the presence or absence, type or size of the defect based on the feature vectors through a classifier or a threshold value.

[0015] Preferably, the collection parameters of the ultrasonic signal include a transmission frequency of 200 kHz, a sampling frequency of not less than 10 MHz, a probe-battery surface distance of 3 cm, and a signal gain of 20 dB.

[0016] The improved WOA-VMD-based lithium battery defect signal extraction method has the following beneficial effects: The improved WOA-VMD-based lithium battery defect signal extraction method has the following beneficial effects: The cosine convergence factor is introduced to replace the linear convergence factor, so that the algorithm can quickly explore globally at the beginning of iteration, and then develop locally finely, adapt to the nonlinear optimization process, add nonlinear inertia weight, dynamically balance the influence of the current optimal solution and random individuals on position update, further improve the convergence precision and search efficiency, compared with the traditional WOA-VMD, the method can guarantee the optimization effect while having faster convergence speed, and a good balance between denoising performance and calculation efficiency is achieved; The method does not rely on fixed parameters, can automatically adjust the VMD parameters according to the characteristics of the input signal, adapts to different battery models, capacities, defect types and signal intensity changes, and can effectively extract defect features in the measured signals of multiple batches and different specifications of lithium batteries, and has strong universality and system robustness, and is suitable for various engineering application scenarios from qualitative defect detection to quantitative defect size evaluation; The system module is clear, including signal acquisition, parameter optimization, decomposition screening, denoising reconstruction and the like, which is convenient for embedding into an existing air-coupled ultrasonic detection platform, the system module is clear, including signal acquisition, parameter optimization, decomposition screening, denoising reconstruction and the like, which is convenient for embedding into an existing air-coupled ultrasonic detection platform, through efficient extraction of defect signals, high-quality data basis is provided for subsequent defect identification, classification, positioning and size inversion, which is helpful to realize quality monitoring in the lithium battery manufacturing process, safety warning and life evaluation in use, and improve the overall safety and reliability of the battery system. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the drawings needed to be used in the embodiment or related art description will be briefly introduced. Obviously, for those skilled in the art, other drawings can also be obtained without creative labor based on these drawings.

[0018] Figure 1 The flowchart of the present application; Figure 2 The schematic diagram of the detection system of the present application; Figure 3 The physical diagram of the flat soft package polymer lithium battery of the present application; Figure 4 The air-coupled transmission signal diagram of the lithium battery of the present application; Figure 5 The time-frequency diagram of each IMF component of the present application; Figure 6 The denoising effect comparison diagram of four algorithms of the present application; Figure 7 The physical diagram of the lithium battery to be tested of the present application. DETAILED DESCRIPTION

[0019] In order to make the purposes, technical solutions and beneficial technical effects of the present application clearer, the present application is further described in detail below in combination with the drawings and specific embodiments. It should be understood that the specific embodiments described in the specification are only for the purpose of explaining the present application, and are not intended to limit the present application.

[0020] Embodiment: Please refer to Figures 1-4 The present application provides a lithium battery defect signal extraction method based on improved WOA-VMD, which is carried out in a MATLAB environment. The effectiveness, accuracy and efficiency of the method are verified by constructing a simulation signal containing target features and typical interference.

[0021] In order to verify the application effect of the algorithm in actual signal processing, a set of lithium battery nondestructive testing system is constructed according to the detection principle of air-coupled ultrasonic penetration method. The system hardware includes air-coupled ultrasonic emission subsystem, air-coupled ultrasonic receiving subsystem, signal excitation and acquisition subsystem, mechanical scanning and positioning subsystem, and main control and data processing computer subsystem. The overall architecture works cooperatively to realize automatic excitation, reception, conditioning and digitization of signals.

[0022] The core component of the air-coupled ultrasonic emission subsystem is an air-coupled ultrasonic emission probe. The center frequency is selected between 50 kHz and 400 kHz to adapt to the penetration ability and resolution requirement of the multi-layer structure of the lithium battery. The probe is designed based on piezoelectric composite material or capacitive principle to realize good acoustic impedance matching with air medium. The core component of the air-coupled ultrasonic receiving subsystem is an air-coupled ultrasonic receiving probe, which matches the center frequency of the emission probe to capture extremely weak penetration signals and convert the ultrasonic waves after penetrating the battery into weak electrical signals. The signal excitation and acquisition subsystem includes a signal generator / pulse emission card, a preamplifier, and a data acquisition card synchronous trigger. The signal generator / pulse emission card is used to generate electrical pulses to excite the emission probe. It can be a high-voltage narrow pulse (used for pulse echo mode) or a modulated continuous wave or burst wave. The preamplifier is used immediately after the receiving probe to amplify the received microvolt-level weak signals to overcome the influence of input noise of the subsequent acquisition card. The data acquisition card is used for high-speed and high-precision analog-to-digital conversion of the amplified analog signals. The synchronous trigger is used to ensure that the emission and acquisition actions are strictly synchronized to ensure that the time reference of each acquisition waveform is consistent. The mechanical scanning and positioning subsystem adopts a high-precision X-Y-Z three-axis electric translation stage, the X-Y axis is used to realize two-dimensional plane scanning between the probe and the battery, C-scan imaging is performed, and a gray-scale image or a pseudo-color image reflecting the defect distribution of the entire battery surface is generated, and the Z axis is used to accurately adjust the distance between the transmitting probe and the receiving probe and the surface of the battery to optimize the sound field coupling and signal strength. The transmitting and receiving probes are fixed on the moving platform, and the relative position and angle thereof are accurately adjustable; The main control and data processing computer is a high-performance industrial computer, which is equipped with sufficient computing power and memory to run complex signal processing algorithms, implement improved WOA-VMD joint wavelet threshold denoising algorithm, and subsequent feature extraction and defect analysis. The user interface provides parameter setting panel, real-time waveform display window, C-scan image display area, processing result report generation and other functions.

[0023] The working process of the system is as follows: the user sets the hardware parameters such as ultrasonic frequency, gain, sampling rate, scanning range, scanning step, sets the parameters of the improved WOA algorithm (population size, iteration number, etc.) and the initial parameter range of VMD and wavelet threshold processing. The motion platform moves point by point according to the preset path, at each scanning point, the signal transmission and data acquisition are triggered synchronously, and the original time domain waveform x(t) and its position coordinates are stored; For each collected signal, the improved WOA algorithm (including cosine convergence factor and dynamic inertia weight) is called, the sum of the envelope entropy of each IMF component after VMD decomposition is taken as the fitness function, and the optimal VMD parameter combination is searched in the set parameter space. Use [K,] to perform VMD decomposition on the original signal x(t), obtain K IMF components, calculate the correlation coefficient of each IMF component and the original signal, select the components with high correlation coefficient as the main signal components, and regard the rest as noise for rejection. For the selected main IMF components, wavelet threshold denoising is performed, the denoised main components are reconstructed, and the final denoised feature signal of the scanning point is obtained. The features that can be used for defect judgment are extracted from the feature signal, such as signal peak amplitude, signal energy, arrival time or specific frequency band energy. The feature values (such as amplitude attenuation) corresponding to all points in the entire scanning region are mapped to gray or color to generate a defect C-scan image, which directly displays the position, shape and relative size of the defect. The user interface provides image contrast, profile line analysis, feature statistics and other functions, generates a detection report containing defect position, size estimation and signal quality index.

[0024] Referring to Figure 3 In the experiment, the test piece is a flat soft packaging polymer lithium ion battery with a length of 275 mm, a width of 185 mm, and a height of 5 mm, and there are circular defects with diameters of 15 mm, 10 mm and 5 mm inside.

[0025] Referring to Figure 4The transmitting and receiving probes are arranged at a distance of 3 cm from the lithium battery; the transmitting frequency is 200 kHz; the sampling frequency is 10 MHz; the sampling point number is 2048; and the signal gain is 20 dB. Figure 4 (a) shows the collected internal defect-free air-coupled ultrasonic signal, and it can be observed that the signal amplitude is large and accompanied by a large amount of interference noise, Figure 4 (b) is the air-coupled ultrasonic signal at the position of the bubble defect, and it can be seen that the transmission signal amplitude is low and the effective characteristic information is almost submerged in noise, and it is difficult to judge the defect size according to the signal characteristics.

[0026] The optimal parameter combination obtained by the improved WOA-VMD decomposition processing of the signal with defects is: and the best fitness is 2.8637. Figure 5 The time-domain graph and the corresponding frequency spectrum graph of each IMF component obtained by the optimal parameter combination decomposition of the noisy signal with defects are shown in the figure, and it can be seen that the main frequency of the signal is 200 kHz, and the correlation coefficients of each component are 0.5517, 0.2945, 0.2900 and 0.2782 respectively. It can be judged that IMF1 is a useful component, and the remaining components are noise components and need to be removed.

[0027] Referring to Figure 6 The preliminary decomposition result shows that the main characteristic component IMF1 representing the original signal has been successfully extracted, but it is still interfered by residual noise. Although this result can meet the qualitative identification of defects, the current denoising effect is still slightly insufficient when facing engineering applications that require precise quantitative evaluation of defect size, and further processing is needed. The main component IMF1 is denoised by wavelet thresholding, and the denoised signal using this algorithm is compared with the denoised results of other algorithms.

[0028] The signal processed by wavelet thresholding performs well in noise suppression, but there is still a small amount of high-frequency noise remaining. The selection of the threshold is critical, and a small threshold setting will lead to incomplete denoising, while a large threshold will damage the original waveform characteristics. Figure 6 (b) is the signal denoised by the EMD algorithm, and the effect of this algorithm in suppressing noise is limited, and the parameters need to be adjusted manually multiple times during the decomposition process, and the overall processing procedure is relatively complicated. Figure 6 (c) is the result processed by EMD combined with wavelet thresholding, and compared with the single EMD method, the noise is further reduced, but the problem of adaptive selection of parameters is still not solved. Figure 6 (d) is the signal denoised by the method in this paper, and the noise components have been basically removed, the denoised signal waveform is more stable, the fluctuation is significantly reduced and the signal characteristics are preserved.

[0029] To further verify the universality of the method, a lithium battery of different batches, different capacities and different sizes was selected for verification test. The battery is 110 mm long, 85 mm wide and 5 mm high, with a capacity of 5100 mAh, as shown in Fig. 1. Figure 7 (a). Figure 7 (b) is the corresponding ultrasonic C-scan image of the battery, where the blue part is the defect part, and it can be seen that there is a circular natural defect with a diameter of about 8 mm inside. The same air-coupled ultrasonic penetration detection system as the above experimental parameters was used. The improved WOA algorithm found different optimal VMD parameter combinations through iteration, which proved that the algorithm was not dependent on a fixed set of parameters, but had the ability to adaptively adjust according to the characteristics of the input signal. Although the original signal characteristics were slightly different due to individual differences in the battery, the denoised signal still clearly extracted the defect waveform, which was smooth and had distinct features, verifying the universality and high robustness of the algorithm.

[0030] Aiming at the problems of serious noise interference in air-coupled ultrasonic signal detection of lithium battery and insufficient parameter adaptive ability of traditional denoising algorithm, a denoising method based on improved WOA optimized VMD parameters combined with wavelet threshold processing is proposed. The cosine convergence factor and inertia weight strategy are introduced to improve the WOA algorithm, which improves the optimization ability in the global range and the convergence precision in the local range. Compared with wavelet threshold and EMD algorithm, the signal-to-noise ratio is relatively improved by 29.45% and 17.70%, the waveform similarity coefficient is relatively improved by 2.47% and 1.35%, and the root mean square error is reduced by 49.10% and 26.51% respectively, which verifies that the method has the best signal processing result in different evaluation indexes, and has important significance for analyzing the relationship between defect size and signal characteristics and detecting the safety of lithium battery.

[0031] The above describes only the preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any skilled person in the art can make equivalent replacements or changes to the technical solutions and inventive concepts of the present application within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A method for extracting defect signals from lithium batteries based on improved WOA-VMD, characterized in that: Includes the following steps: S1. Acquire the air-coupled ultrasonic penetration signal of the lithium battery; S2. Based on the acquired ultrasonic penetration signal, the improved Whale Optimization Algorithm (WOA) is used to optimize the parameters K and VAR in Variational Mode Decomposition (VMD). Optimization is performed, using minimum envelope entropy as the fitness evaluation criterion; S3, using the optimized parameters K and The input signal is decomposed using VMD to obtain multiple intrinsic mode function (IMF) components. S4. Calculate the correlation coefficient between each IMF component and the original signal, and divide the components into effective components and irrelevant components according to the preset threshold. S5. Perform wavelet threshold denoising on the selected effective components; S6. Reconstruct the denoised components to obtain the processed defect signal.

2. The lithium battery defect signal extraction method based on improved WOA-VMD according to claim 1, characterized in that: In step S2, the improved Whale Optimization Algorithm (WOA) introduces a cosine convergence factor and an inertia weight strategy into the traditional Whale Optimization Algorithm: The expression for the cosine convergence factor is: In the formula, t is the current iteration number. This represents the maximum number of iterations. The expression for the inertia weighting strategy is: .

3. The lithium battery defect signal extraction method based on improved WOA-VMD according to claim 2, characterized in that: The initialization of the whale population involves setting the population size to 50 and the maximum number of iterations to 30. The VMD parameters K and K' represent the position vectors of individual whales. By iteratively updating the positions of individual whales, the goal is to find the parameter combination that minimizes the envelope entropy. .

4. The lithium battery defect signal extraction method based on improved WOA-VMD according to claim 3, characterized in that: In the improved whale optimization algorithm, the fitness function is the minimum envelope entropy of each modal component after variational mode decomposition, and the formula for calculating the envelope entropy is: In the formula , is the envelope signal after Hilbert demodulation, and N is the signal length.

5. The lithium battery defect signal extraction method based on improved WOA-VMD according to claim 1, characterized in that: In step S4, the preset correlation coefficient threshold is 0.

6. Components with correlation coefficients greater than the threshold are considered as valid components, and the rest are considered as noise components.

6. The lithium battery defect signal extraction method based on improved WOA-VMD according to claim 1, characterized in that: In step S5, the wavelet thresholding denoising employs either a soft thresholding or a hard thresholding function. The threshold is adaptively determined based on the signal variance and length. The expressions for the soft thresholding or hard thresholding functions are as follows: In the formula, represents the decomposition scale, represents the coefficient label, and represents the threshold. , where is the variance and is the signal length.

7. The lithium battery defect signal extraction method based on improved WOA-VMD according to claim 6, characterized in that: The wavelet threshold denoising specifically includes the following steps: (501) Perform wavelet multi-scale decomposition on each dominant signal component to obtain the corresponding wavelet coefficients; (502) Calculate the adaptive threshold based on the wavelet coefficients; (503) Use a soft thresholding function or a hard thresholding function to perform thresholding on the wavelet coefficients; (504) Perform inverse wavelet transform on the processed wavelet coefficients to obtain the denoised modal components.

8. The lithium battery defect signal extraction method based on improved WOA-VMD according to claim 1, characterized in that: The input signal is a lithium battery defect signal acquired by an air-coupled ultrasonic testing system, and the lithium battery defect signal is a transmission signal when there are defects inside the lithium battery.

9. The lithium battery defect signal extraction method based on improved WOA-VMD according to claim 1, characterized in that: The method also includes the steps of feature extraction and defect identification of the reconstructed signal, specifically including calculating the time-domain features, frequency-domain features or time-frequency features of the reconstructed signal, and identifying the presence, type or size of defects based on the feature vector by classifier or threshold judgment.

10. A method for extracting lithium battery defect signals based on improved WOA-VMD according to claim 1, characterized in that: The ultrasonic signal acquisition parameters include a transmission frequency of 200kHz, a sampling frequency of not less than 10MHz, a probe-to-battery distance of 3cm, and a signal gain of 20dB.

Images