Composite material structure damage identification method based on acoustic emission wavelet packet energy characteristics

Abstract

The invention relates to the technical field of composite material structure monitoring, in particular to a composite material structure damage identification method based on acoustic emission wavelet packet energy characteristics, and the method comprises the steps: S1, obtaining acoustic emission signal data of a composite material, and carrying out the preprocessing; s2, performing wavelet packet decomposition on the denoised composite material acoustic emission signal data by using a wavelet basis function and discrete wavelet inverse transformation; s3, acquiring the processed acoustic emission signal data of the composite material, and extracting the characteristic energy of each piece of acoustic emission signal data; and S4, performing clustering analysis by using a clustering method to complete damage identification of the composite material structure. According to the method, the acoustic emission signal is used as an analysis basis, wavelet transform and k-means clustering analysis are used for accurately realizing structural damage identification and judgment of the composite material, dynamic continuous monitoring can be realized, the method is not influenced by structural complexity, and the method has wide applicability and is suitable for various composite materials.

Description

Technical Field

[0001] The invention belongs to the technical field of composite material structure monitoring, and in particular relates to a composite material structure damage identification method based on acoustic emission wavelet packet energy characteristics. Background Art

[0002] Composite materials are now widely used in the aerospace industry due to their advantages such as lightweight, corrosion resistance, superior mechanical properties and high design flexibility. Their usage has become an important indicator of the advancement of aircraft structures and their market competitiveness. However, composite material connection structures have the characteristics of complex failure mechanisms, coupled failure modes, diverse damage extension pathways and large scale of adhesive layer damage areas. These characteristics make it difficult to analyze the strength and damage of composite material connection structures, which has caused great obstacles to the reliability analysis of composite materials. As an indispensable key component of composite material structures, nearly 80% of composite material failures are currently caused by damage to the connection structure. Therefore, in order to avoid sudden structural damage to composite material connection structures, it is urgent to develop intelligent health monitoring technology and non-destructive damage detection technology for them.

[0003] Acoustic emission refers to the phenomenon that when a material is subjected to external loading, a certain area inside the material undergoes a small change and generates transient elastic waves. As a non-destructive testing technology, acoustic emission technology is used to monitor the tiny sounds released by materials and structures during loading, which are caused by processes such as microcracks and crack extension inside the material. These sounds are captured by placing sensors on the surface of the material, and their characteristics are analyzed to identify and evaluate the damage state of the material. This technology is of great significance in structural health testing such as damage diagnosis, location, identification and prediction of composite adhesive joint structures.

[0004] At present, damage identification of composite material connection structures is mainly carried out by extracting traditional univariate statistical signal features such as rise time, peak frequency, center of gravity frequency, weighted frequency, ringing count, threshold voltage, etc. from the acoustic emission signal to analyze the acoustic emission signal group. However, since the existing univariate statistical features have not been deeply understood, it is still very difficult to effectively distinguish and identify various damage modes based on arbitrary univariate statistical acoustic emission time domain features and spectrum features. At the same time, there are still many problems with the clustering analysis results based on traditional acoustic emission features. First, the peak frequency has a great influence on clustering, but the extraction of peak frequency is easily interfered by noise, which will greatly affect the accuracy of clustering acoustic emission events of different damage modes. Therefore, it is necessary to design a new acoustic emission signal feature extraction method to obtain clustering results with better explanatory power and optimize the composite material damage identification method.

[0005] In this context, this patent uses the adhesive connection structure of composite materials with screw holes as the test object, uses acoustic emission technology, and combines the wavelet packet analysis method to extract the characteristics of the acoustic emission signal to explore the damage mechanism and failure mode of the structure under tensile conditions. In addition, the cluster analysis method is introduced to realize the classification and identification of the damage type of the specimen structure. Summary of the invention

[0006] In view of the shortcomings of the prior art, the present invention provides a composite material structure damage identification method based on the energy characteristics of acoustic emission wavelet packets. The method takes the acoustic emission signal as the analysis basis, uses wavelet transform and k-means clustering analysis to accurately realize the structural damage identification and judgment of the composite material. The method has high sensitivity, can realize dynamic monitoring and continuous monitoring, is not affected by structural complexity, and has wide applicability and is suitable for a variety of composite materials.

[0007] To achieve the above object, the present invention discloses the following technical solution:

[0008] A composite material structure damage identification method based on acoustic emission wavelet packet energy characteristics, comprising:

[0009] S1: Acquire the acoustic emission signal data of the composite material and pre-process the acoustic emission signal data;

[0010] An acoustic emission sensor is installed on the structural surface of the composite material, a force loading test is performed on the composite material using a universal testing machine, and acoustic emission signal data is collected; the acoustic emission signal data is preprocessed, a power spectrum of the acoustic emission signal data is calculated using the Welch power spectrum method, the peak frequency of the acoustic emission signal data is extracted, a peak frequency threshold is set, and the acoustic emission signal data with a peak frequency lower than the threshold is filtered out as noise to obtain the denoised composite material acoustic emission signal data f(t);

[0011] S2: using wavelet basis functions and discrete wavelet inverse transform equations to perform wavelet packet decomposition on the composite material acoustic emission signal data after denoising in step S1;

[0012] S21: Using wavelet basis function, the composite material acoustic emission signal data after denoising in step S1 is decomposed into:

[0013]

[0014] Wherein, f(t) is the processed acoustic emission signal data; F(ω) is the frequency domain function of the acoustic emission signal data after Fourier transform of f(t); is an imaginary unit; ω is the angular frequency of the acoustic emission signal data; t is time; WT(a,τ) is the time-frequency function of the acoustic emission signal data after wavelet transformation; τ is the time translation of the acoustic emission signal data; a is the time expansion of the acoustic emission signal data; ψ is the mother wavelet function;

[0015] S22: Performing a discrete wavelet inverse transform on the composite material acoustic emission signal data f(t) to obtain a first component and a second component of the composite material acoustic emission signal data. The discrete wavelet inverse transform equation is:

[0016]

[0017] Among them, DWT(i,k) is the wavelet transform coefficient of the acoustic emission signal data; i is the decomposition level of the acoustic emission signal data; k is the discrete time domain translation;

[0018] S3: Obtaining the composite material acoustic emission signal data processed in step S2, and extracting the characteristic energy of each acoustic emission signal data;

[0019] The energy of the wavelet packet coefficient of each acoustic emission signal data and the energy proportion of each frequency band in the composite material acoustic emission signal data are calculated, and the relative energy distribution of each energy level of the composite material is extracted as follows:

[0020]

[0021] in, is the relative energy distribution of the energy level of the i-th layer of the composite material; E is the frequency band energy of the acoustic emission signal data; Total (t) is the energy of the wavelet packet coefficients of the acoustic emission signal data;

[0022] Each acoustic emission signal data is decomposed into the i-th layer using wavelet packets, and the acoustic emission signal data is decomposed into the 2 i-th layer. i The wavelet packet energy ratio features form a 2 i The characteristic energy of each energy level of the composite material is:

[0023]

[0024] Among them, X i is the characteristic energy of the i-th energy level of the composite material;

[0025] S4: Use the k-means clustering method to cluster the characteristic energy of each energy level of the composite material in step S3, identify the type of structural damage of the composite material, and complete the full life cycle health monitoring of the composite material.

[0026] Preferably, in step S1, the power spectrum of the acoustic emission signal data is calculated using the Welch power spectrum method, which specifically includes the following steps:

[0027] S11: segmenting the acoustic emission signal data into N segments. In order to obtain better frequency resolution in the frequency domain and make the power spectrum smoother, a 20% overlap is set between each window.

[0028] S12: adding a window to the acoustic emission signal data, performing window function processing on each segment of data, reducing the leakage effect of spectrum estimation, and reducing the frequency resolution loss introduced by the window function;

[0029] S13: Performing Fourier transform on the signal in each acoustic emission signal data window;

[0030] S14: Calculate the power spectrum density estimation value of each frequency band of the acoustic emission signal data.

[0031] Preferably, the first component of the composite material acoustic emission signal data in step S22 is calculated using a discrete wavelet inverse transform equation, specifically:

[0032]

[0033] Wherein, AVC is the frequency bandwidth of the first component after the acoustic emission signal data is decomposed by wavelet packet; f s is the sampling rate of the acoustic emission signal data; 2 i is the number of components in the i-th layer after wavelet packet decomposition of the acoustic emission signal data.

[0034] Preferably, the second component of the composite material acoustic emission signal data emission signal in step S22 is calculated using a discrete wavelet inverse transform equation, specifically:

[0035]

[0036] Wherein, DQ is the frequency bandwidth of the second component of the acoustic emission signal data after the acoustic emission signal data is decomposed by wavelet packet.

[0037] Preferably, the discrete wavelet inverse transform equation in step S22 is determined according to a discrete wavelet transform mathematical model, specifically:

[0038]

[0039] Where c is the discrete wavelet transform coefficient.

[0040] Preferably, in step S3, the energy of the wavelet packet coefficient of each acoustic emission signal data in the composite material acoustic emission signal data and the energy proportion of each frequency band are calculated, specifically:

[0041] The energy criterion is used to distinguish the energy distribution of each composite material, and the energy level of each acoustic emission signal data is set as:

[0042]

[0043] Where j = 1, j = 1, ..., 2 i Number the jth component of the wavelet packet decomposition to the i-th layer; is the component of the i-th layer of the decomposed signal; is the frequency band energy of the acoustic emission signal data; t0 is the starting time of the acoustic emission signal data;

[0044] The energy of the wavelet packet coefficients of the acoustic emission signal data is set to:

[0045]

[0046] Among them, E Total (t) is the energy of the wavelet packet coefficients of the acoustic emission signal data;

[0047] According to the energy level of each acoustic emission signal data and the total energy of the acoustic emission signal data, the relative energy distribution of each energy level of the composite material is determined.

[0048] Preferably, in step S4, the characteristic energy of each energy level of the composite material in step S3 is clustered and analyzed using a k-means clustering method, specifically:

[0049] The characteristic energy of each energy level of n groups of composite materials is X={X1,X2…X i …X n Input the k-means clustering method and divide it into k clusters so that the distance between each characteristic energy data sample and the center point of the cluster to which it belongs is minimized. Initialize the centers of the k clusters: Calculate the Euclidean distance from each acoustic emission signal data point to each cluster center as follows:

[0050]

[0051] Among them, Dis(X i ,C p ) is the Euclidean distance from the ith acoustic emission signal data point to the pth cluster center; p is the cluster center number; X i is the characteristic energy of the i-th energy level of the composite material; C p is the cluster center of the pth class, {C1,…C p …,C k},1<p≤n;X it is the t-th characteristic energy coordinate value of the i-th acoustic emission characteristic energy data sample; Cpt is the t-th characteristic energy coordinate value of the p-th cluster center; t is the characteristic energy coordinate value number; m is the characteristic energy dimension.

[0052] Preferably, the composite material structure damage types in step S4 include matrix cracking MC, adhesive debonding AD and fiber breakage FB; the acoustic emission signal energy generated by the three types of damage is concentrated in different frequency bands, and the composite material structure damage type identification is achieved through the acoustic emission damage information contained in the wavelet packet energy;

[0053] The matrix cracking MC damage is that there is a bonding part in the composite material structure, and the fiber needs to bear pressure. The fiber at the bonding part first breaks to a certain extent, resulting in a gradual decrease in the mechanical properties of the composite material structure.

[0054] The adhesive debonding AD damage is manifested as the composite material structure is subjected to increasing pressure, the mechanical properties of the composite material structure are insufficient, the adhesive begins to debond, and the continuous debonding of the adhesive causes the fibers on the surface of the composite material structure to be gradually pulled out and broken;

[0055] The fiber breakage FB damage is manifested as complete rupture and debonding of the adhesive, and a large number of fibers are pulled out and broken, resulting in large-scale cracking of the resin matrix connected to the fibers and with weaker mechanical properties.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] (1) The present invention uses acoustic emission signals as the basis for analyzing composite material structures, has high sensitivity, can realize dynamic monitoring and continuous monitoring, is not affected by the complexity of the composite material structure, and the obtained acoustic emission signal data is real-time and can be applied to a variety of composite materials.

[0058] (2) The present invention adopts the Welch method to calculate the power spectrum of the acoustic emission signal, which can reduce the variance of the acoustic emission signal data, reduce the leakage effect, improve the frequency resolution, and better reflect the characteristics of the non-stationary signal changing with time.

[0059] (3) The present invention adopts the Meyer wavelet which is compactly supported, smooth, asymmetric and has the characteristics of multi-resolution analysis as the wavelet basis of wavelet transform. Its compact support enables it to effectively capture the local details of the acoustic emission signal, while its smoothness and asymmetry give it sensitivity to the overall characteristics and transient characteristics of the signal.

[0060] (4) The present invention uses an unsupervised k-means machine learning algorithm to perform cluster analysis on acoustic emission signals. The algorithm is simple, convenient, efficient, and scalable. It does not require the prior specification of cluster prototypes, has wide applicability, is easy to parallelize, has rigorous mathematical theory, and the results are interpretable. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is a flow chart of the composite material structure damage identification method based on acoustic emission wavelet packet energy characteristics of the present invention;

[0062] Figure 2 This is a physical picture of the tensile mechanics test of the composite material connection specimen in the laboratory of the embodiment of the present invention;

[0063] Figure 3 is the Welch spectrum of the acoustic emission signal of the present invention;

[0064] Figure 4 Schematic diagram of the wavelet transform process of a signal of the present invention;

[0065] Figure 5 Schematic diagram of the wavelet packet transform process of a signal of the present invention;

[0066] Figure 6 It is a logic diagram of the process of performing wavelet packet decomposition of a signal according to the present invention;

[0067] Figure 7 It is a time domain signal diagram of the acoustic emission signal of the present invention;

[0068] Figure 8a-8h A time domain signal diagram reconstructed by wavelet packets of the present invention;

[0069] Figure 9a-9h The spectrum diagram of the wavelet packet reconstruction signal of the present invention;

[0070] Fig.10 The extracted wavelet packet energy ratio histogram of the present invention;

[0071] Fig.11 It is a flow chart of the k-means clustering algorithm of the present invention;

[0072] Fig.12 A visualized three-dimensional graph of the k-means clustering analysis process of the acoustic emission signal of the present invention;

[0073] Fig.13a and Fig.13b A first projection diagram and a second projection diagram are visualized for the k-means clustering analysis process of the acoustic emission signal of the present invention;

[0074] Fig.14 It is a damage mode evolution diagram of the present invention based on the k-means cluster analysis of the wavelet packet energy ratio feature;

[0075] Figure 15a-Figure 15c The wavelet packet energy ratio histogram of the centers of clusters of different damage types of the present invention;

[0076] Fig.16aand Fig.16b The output of the composite material CFRP matrix cracking damage result diagram of the present invention;

[0077] Fig.17a and Fig.17b It is a composite material CFRP adhesive debonding damage result diagram outputted by the present invention;

[0078] Fig.18a and Fig.18b This is a graph showing the fracture damage results of the composite material CFRP fiber outputted by the present invention. DETAILED DESCRIPTION

[0079] The exemplary embodiments, features and aspects of the present invention will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0080] The embodiment of the present invention uses composite material connectors for verification analysis and provides a composite material structure damage identification method based on acoustic emission wavelet packet energy characteristics, such as Figure 1 As shown, the acoustic emission signal data of the composite material is obtained and preprocessed; the denoised composite material acoustic emission signal data is decomposed by wavelet packet using wavelet basis function and discrete wavelet inverse transform; the processed composite material acoustic emission signal data is obtained, and the characteristic energy of each acoustic emission signal data is extracted; cluster analysis is performed using a clustering method to complete the composite material structure damage identification; which includes:

[0081] Step S1: Acquire acoustic emission signal data of the composite material connector and pre-process the acoustic emission signal data.

[0082] When composite material connectors are subjected to external forces or produce slight deformations, tiny cracks or damage will occur, accompanied by the release of energy. Part of this energy release will propagate in the form of elastic waves and reach the surface of the material, causing tiny vibrations. The acoustic emission sensor is installed on the structural surface of the composite material.

[0083] These weak vibration signals are converted into electrical signals, and after being amplified and digitized by the acquisition system, they are collected as acoustic emission signal data. A universal testing machine is used to perform a force loading test on the composite connector to collect the acoustic emission signal data. Taking the laboratory test as an example, the entire test system includes an acoustic emission sensor 1, a preamplifier 2, an acoustic emission receiver 3, a composite connection specimen 4 and a universal testing machine 5. Figure 2As shown, the acoustic emission sensor is arranged on the surface of the composite material connection specimen through a coupling agent. When the universal testing machine applies force to the specimen, the acoustic emission sensor amplifies the collected acoustic emission signal through a preamplifier and continuously transmits it to the acoustic emission receiver. The signal is recorded after being preprocessed and digitized.

[0084] Preprocess the acoustic emission signal data. Since the friction between the surface of the composite connector and the force loading device or external particles will also generate acoustic emission signals, as noise, it can be converted to the frequency domain. By setting the frequency domain threshold, the internal damage signals of these non-composite connectors can be filtered out. The power spectrum of the acoustic emission signal data is calculated using the Welch power spectrum method, the peak frequency is extracted, and the noise signals with too low and too high peak frequencies are filtered out. The specific steps include the following:

[0085] Step S11: segment the acoustic emission signal data into N segments. In order to obtain better frequency resolution in the frequency domain and make the power spectrum smoother, a 20% overlap is set between each window.

[0086] Step S12: Add a window to the acoustic emission signal data, and perform window function processing on each segment of data to reduce the leakage effect of spectrum estimation and the frequency resolution loss introduced by the window function.

[0087] Step S13: Perform Fourier transform on the signal in each acoustic emission signal data window.

[0088] Step S14: Calculate the power spectrum density estimation value of each frequency band of the acoustic emission signal data.

[0089] The peak frequency of the acoustic emission signal data is extracted, a peak frequency threshold is set, and the acoustic emission signal data with a peak frequency lower than the threshold is filtered out as noise to obtain the denoised composite connector acoustic emission signal data f(t).

[0090] Step S2: using wavelet basis functions and discrete wavelet inverse transform equations to perform wavelet packet decomposition on the composite material connector acoustic emission signal data after denoising in step S1.

[0091] Step S21: wavelet transform of a finite length attenuated wavelet basis is like a movable signal microscope that can simultaneously obtain the frequency and time information of the signal; the denoised composite connector acoustic emission signal data in step S1 is decomposed using the wavelet basis function into:

[0092]

[0093] Wherein, f(t) is the processed acoustic emission signal data; F(ω) is the frequency domain function of the acoustic emission signal data after Fourier transform of f(t); is the imaginary unit; ω is the angular frequency of the acoustic emission signal data; t is the time; WT(a,τ) is the time-frequency function of the acoustic emission signal data after wavelet transform; τ is the time shift of the acoustic emission signal data; a is the time expansion of the acoustic emission signal data; ψ is the mother wavelet function.

[0094] Step S22: Use discrete wavelet transform (DWT) to perform inverse discrete wavelet transform on the composite connector acoustic emission signal data f(t), transform the original waveform, and obtain the first component and the second component of the composite connector acoustic emission signal data, the first component represents the approximate component of the composite acoustic emission signal data, and the second component represents the detail component of the composite acoustic emission signal data; the inverse discrete wavelet transform equation is:

[0095]

[0096] Among them, DWT(i,k) is the wavelet transform coefficient of the acoustic emission signal data; i is the decomposition level of the acoustic emission signal data; k is the discrete time domain translation.

[0097] The embodiment of the present invention selects Meyer wavelet as the mother wavelet. Meyer wavelet is a compactly supported, smooth, asymmetric wavelet function with the characteristics of multi-resolution analysis. Its compact support enables it to effectively capture the local details of the signal, while its smoothness and asymmetry give it sensitivity to the overall characteristics and transient characteristics of the signal. As a tool for multi-resolution analysis, Meyer wavelet can provide frequency information at different scales, and its wavelet packet energy characteristics extracted in this study are the most obvious. However, when processing acoustic emission signals, the wavelet transform will not further decompose the high-frequency detail signals. Figure 3 As shown in , it only continuously decomposes the low-frequency part of the acoustic emission signal, which will result in the loss of a large amount of high-frequency detail information of the signal. For the acoustic emission random signal with a high sampling rate, its decomposition and representation capabilities are relatively limited. Figure 4 The wavelet decomposition process is shown in Figure 1. Wavelet packet transform is a further development of wavelet transform. It not only retains the time-frequency analysis characteristics of wavelet transform, but also can realize the simultaneous decomposition of the low-frequency part and high-frequency part of the signal. By selecting different resolutions at different frequency levels, the original acoustic emission signal can be fully decomposed in the entire frequency band. Figure 5 The wavelet packet decomposition process shown in FIG. 1 decomposes and reconstructs the signal at different scales, making it very effective in processing acoustic emission signals with multi-scale characteristics.

[0098] The first component of the acoustic emission signal data of the composite connector is calculated using the discrete wavelet inverse transform equation, specifically:

[0099]

[0100] Wherein, AVC is the frequency bandwidth of the first component after the acoustic emission signal data is decomposed by wavelet packet; f s is the sampling rate of the acoustic emission signal data; 2 i is the number of components in the i-th layer after wavelet packet decomposition of the acoustic emission signal data.

[0101] The second component of the acoustic emission signal data of the composite connector is calculated using the discrete wavelet inverse transform equation, specifically:

[0102]

[0103] Wherein, DQ is the frequency bandwidth of the second component of the acoustic emission signal data after the acoustic emission signal data is decomposed by wavelet packet.

[0104] The discrete wavelet inverse transform equation is determined according to the discrete wavelet transform mathematical model, specifically:

[0105]

[0106] Where c is the discrete wavelet transform coefficient.

[0107] by Figure 2 Taking the tensile mechanical test of the composite material connection specimen shown in the figure as an example, the set frequency is 2MHz. According to the sampling theorem, the frequency range at the wavelet packet decomposition node (0, 0) should be 0-1MHz. In the early stage of single variable statistical feature analysis, from the acoustic emission signal data of the acoustic emission impact group and the distribution density of the center of gravity frequency, the energy of the acoustic emission signal is mainly concentrated in the frequency range of 80-500kHz. Figure 6 The figure shows the process logic diagram of the present invention for performing wavelet packet decomposition on the signal, performing 4-layer wavelet packet decomposition on the acoustic emission signal, and collecting the wavelet packet coefficients of the first 8 nodes of the fourth layer: (4,0), (4,1), (4,2), (4,3), (4,4), (4,5), (4,6), (4,7) for wavelet packet signal reconstruction. The 8 frequency bands corresponding to these 8 nodes are: 0-62.5kHz, 62.5-125kHz, 125-187.5kHz, 187.5-250kHz, 250-312.5kHz, 312.5-375kHz, 375-437.5kHz, 437.5-500kHz; Figure 7 The time domain signal diagram of the acoustic emission signal of the present invention is shown as the original signal diagram. The wavelet packet decomposition process is to Figure 7 It is decomposed into Figure 8 and Figure 9, as shown in the following figure.

[0108] After completing the wavelet packet decomposition of the acoustic emission signal, the wavelet packet coefficients are rearranged in the order of frequency bands, starting from 0kHz and increasing by 62.5kHz, so as to ensure that the 8 frequency bands corresponding to these 8 nodes are: 0-62.5kHz, 62.5-125kHz, 125-187.5kHz, 187.5-250kHz, 250-312.5kHz, 312.5-375kHz, 375-437.5kHz, 437.5-500kHz. After completing the order rearrangement, the wavelet packet coefficients obtained by decomposition are used to reconstruct the acoustic emission impact signal in the time domain, which is divided into 8 components. The time domain signal is shown in Figure 8. Figure 8a The time domain signal diagram reconstructed by wavelet packets in the frequency band 0-62.5kHz; Figure 8b The time domain signal diagram reconstructed by wavelet packets in the frequency band 62.5-125kHz; Figure 8c This is the time domain signal diagram reconstructed by wavelet packets in the frequency band 125-187.5kHz; Figure 8d The time domain signal diagram of wavelet packet reconstruction in the frequency band 187.5-250kHz; Figure 8e This is the time domain signal diagram reconstructed by wavelet packets in the frequency band 250-312.5kHz; Figure 8f This is the time domain signal diagram reconstructed by wavelet packets in the frequency band 312.5-375kHz; Figure 8g This is the time domain signal diagram reconstructed by wavelet packets in the frequency band 375-437.5kHz; Figure 8h The time domain signal diagram of wavelet packet reconstruction in the frequency band 437.5-500kHz. The frequency domain signal is shown in Figure 9. Figure 9a The spectrum diagram of the wavelet packet reconstructed signal in the frequency band 0-62.5kHz;

[0109] Figure 9b The spectrum diagram of the wavelet packet reconstructed signal in the frequency band 62.5-125kHz; Fig.9c The spectrum diagram of the wavelet packet reconstructed signal in the frequency band 125-187.5kHz; Figure 9d The spectrum diagram of the wavelet packet reconstructed signal in the frequency band 187.5-250kHz; Fig.9e The spectrum diagram of the wavelet packet reconstructed signal in the frequency band 250-312.5kHz; Figure 9f The spectrum diagram of the wavelet packet reconstructed signal in the frequency band 312.5-375kHz; Figure 9g The spectrum diagram of the wavelet packet reconstructed signal in the frequency band 375-437.5kHz; Figure 9h This is the spectrum diagram of the wavelet packet reconstructed signal in the frequency band 437.5-500kHz.

[0110] Step S3: Acquire the acoustic emission signal data of the composite material connector after being processed in step S2, and extract the characteristic energy of each acoustic emission signal data.

[0111] Calculate the energy of the wavelet packet coefficient of each acoustic emission signal data and the energy proportion of each frequency band in the acoustic emission signal data of the composite connector, specifically:

[0112] The energy criterion is used to distinguish the energy distribution of each composite connector, and the energy level of each acoustic emission signal data is set as:

[0113]

[0114] Where j = 1,...,2 i Number the jth component of the wavelet packet decomposition to the i-th layer; is the component of the i-th layer of the decomposed signal; is the frequency band energy of the acoustic emission signal data; t0 is the starting time of the acoustic emission signal data.

[0115] The energy of the wavelet packet coefficients of the acoustic emission signal data is set to:

[0116]

[0117] Among them, E Total (t) is the energy of the wavelet packet coefficients of the acoustic emission signal data.

[0118] According to the energy level of each acoustic emission signal data and the total energy of the acoustic emission signal data, the relative energy distribution of each energy level of the composite connector is determined for:

[0119]

[0120] in, is the relative energy distribution of the energy level of the i-th layer of the composite connector; E is the frequency band energy of the acoustic emission signal data; Total (t) is the energy of the wavelet packet coefficients of the acoustic emission signal data.

[0121] Each acoustic emission signal data is decomposed into the i-th layer using wavelet packets, and the acoustic emission signal data is decomposed into the 2 i-th layer. i The wavelet packet energy ratio features form a 2 i The characteristic energy of each energy level of the composite material connector is:

[0122]

[0123] Among them, X i is the characteristic energy of the i-th energy level of the composite connector.

[0124] Since the acoustic emission signal energy generated by different damages of composite material connectors is concentrated in different frequency bands, for example, matrix-shaped damage is generally concentrated in low frequencies, and fiber-shaped damage is generally concentrated in high frequencies, the wavelet packet energy characteristics actually contain a lot of acoustic emission damage information. This information can be mined through unsupervised machine learning algorithms to achieve the identification of different acoustic emission damage signals. In the embodiment of the present invention, for the acoustic emission signal generated by the damage of the composite material connection structure, if the sampling rate is 1Mhz, the wavelet packet is used to decompose to the i=3th layer to form a 2 3 =8-dimensional feature vector.

[0125] The embodiment of the present invention is Figure 2 Taking the tensile mechanical test of composite material connection specimens as an example, the square root of each element is calculated, the energy of the wavelet packet signal of each node is calculated, the total energy is summed up and the wavelet packet energy proportion of each node is calculated, and a bar chart is drawn as shown in Fig.10 As shown, nodes 1 to 8 correspond to 8 frequency bands: F1 (0-62.5kHz), F2 (62.5-125kHz), F3 (125-187.5kHz), F4 (187.5-250kHz), F5 (250-312.5kHz), F6 (312.5-375kHz), F7 (375-437.5kHz), F8 (437.5-500kHz). These 8 energy proportions form an 8-dimensional feature vector.

[0126] Step S4: Use the k-means clustering method to cluster and analyze the characteristic energy of each energy level of the composite connector in step S3. The k-means clustering algorithm is a simple, easy to implement, fast calculation speed, wide applicability, and strong interpretability unsupervised learning method (Unsupervised Learning), which belongs to a type of machine learning. Since the algorithm only involves the distance calculation between samples and cluster centers, it can achieve efficient clustering analysis under appropriate data structure and algorithm optimization, which is especially suitable for large-scale data sets. It can flexibly select the number of clusters k and is suitable for various types of data. The clustering results it produces are intuitive and easy to understand, and have a rigorous mathematical theoretical basis. The goal of the k-means clustering algorithm is to divide the samples in the data set into k clusters so that the distance between each sample and the center point of the cluster to which it belongs is minimized. The algorithm flow is as follows Fig.11 As shown in the figure, the data points are iteratively assigned to k clusters so that the distance between each data point and the center of the cluster to which it belongs is minimized. The cluster center is continuously updated until the convergence condition is met. Its goal is to minimize the variance of the data points within the cluster and maximize the difference between clusters, so as to cluster data points with similar characteristics together; specifically: the characteristic energy of each energy level of n groups of composite connectors X={X1,X2…X i …Xn Input the k-means clustering method and divide it into k clusters so that the distance between each characteristic energy data sample and the center point of the cluster to which it belongs is minimized. Initialize the centers of the k clusters: Calculate the Euclidean distance from each acoustic emission signal data point to each cluster center as follows:

[0127]

[0128] Among them, Dis(X i ,C p ) is the Euclidean distance from the ith acoustic emission signal data point to the pth cluster center; p is the cluster center number; X i is the characteristic energy of the energy level of the i-th layer of the composite connector; C p is the cluster center of the pth class, {C1,…C p …,C k},1<p≤n;X it is the t-th characteristic energy coordinate value of the i-th acoustic emission characteristic energy data sample; C pt is the t-th characteristic energy coordinate value of the p-th cluster center; t is the characteristic energy coordinate value number; m is the characteristic energy dimension. Taking the sampling rate of 1Mhz and the wavelet packet decomposition to the third layer as an example, m=8, =8, t=1,…,8.

[0129] Input the feature vector into the clustering algorithm for clustering analysis, and obtain the k-means clustering result visualization as follows Fig.12 As shown in Figure 2, three types of three-dimensional clustering analysis results are shown. Fig.13a It is a visualized first projection diagram of the k-means clustering analysis process of the acoustic emission signal of the present invention, showing the clustering analysis results of three types of peak amplitudes (dB); Fig.13b It is a second projection diagram for visualization of the k-means clustering analysis process of the acoustic emission signal of the present invention, showing clustering analysis results of three types of peak frequencies (kHz).

[0130] According to the clustering results, the acoustic emission events are divided into different clusters. Taking the composite connector data of the tensile test as an example, this method identifies three damage modes that occur in the composite connector from integrity to fracture, such as Fig.14 This is the damage pattern evolution diagram of the k-means cluster analysis based on the wavelet packet energy ratio feature of the present invention; among them, type 1 is matrix crack MC, type 2 is adhesive debonding AD, and type 3 is fiber breakage FB. For the composite material bonding specimen in this experiment, according to the damage signal identified by this method, its destruction process is:

[0131] 1. Due to the existence of the bonding part, the fiber is subjected to a large amount of pressure, and the fiber at the bonding part first breaks to a certain extent, resulting in a gradual decrease in the mechanical properties of the specimen; Fig.15a The histogram of wavelet packet energy ratio at the cluster center of damage type (1).

[0132] 2. As the tensile test progresses, the mechanical properties of the specimen are insufficient, and the adhesive begins to debond. The continuous debonding of the adhesive causes the fibers on the surface to be gradually pulled out and broken; Fig.15b The histogram of wavelet packet energy ratio at the cluster center of damage type (2).

[0133] 3. At the last moment, the adhesive completely breaks and debonds, and a large number of fibers are pulled out and broken, resulting in large-scale cracking of the resin matrix connected to the fibers and with weaker mechanical properties. This method can, to a certain extent, monitor the structural health of the damage of composite connectors throughout their life cycle. Fig.15c The histogram of wavelet packet energy ratio at the cluster center of damage type (3).

[0134] Using k-means cluster analysis based on the wavelet packet energy ratio characteristics of the acoustic emission signal, three types of composite material connection specimen damage modes were identified: matrix cracking MC (Matrix Cracking), adhesive debonding AD (Adhesive Debonding), and fiber breakage FB (Fiber Breakage). As shown in Figures 16, 17, and 18, the three types of damage modes were observed using an electron microscope to output the damage type of each acoustic emission event. Fig.16a This is a graph of the crack damage detection result of the composite material CFRP matrix output by the embodiment of the present invention. Fig.16b This is a diagram showing the matrix cracking damage effect observed using an electron microscope in the present invention.

[0135] like Fig.17a FIG. 1 is a graph showing the debonding damage detection result of a composite material CFRP adhesive outputted by an embodiment of the present invention. Fig.17b This is a diagram showing the adhesive debonding damage observed by the present invention using an electron microscope. Fig.18a FIG. 1 is a graph showing the fracture damage detection result of the composite material CFRP fiber outputted by the embodiment of the present invention. Fig.18b This is a diagram showing the effect of fiber breakage observed using an electron microscope in the present invention.

[0136] The embodiment of the present invention takes the composite material connection structure as the analysis object, takes the real-time monitored acoustic emission signal as the analysis basis, adopts the Welch method to calculate the signal power spectrum, realizes the extraction of a series of traditional acoustic emission features such as rise time, peak amplitude, peak frequency, etc., and understands the signal distribution by performing univariate statistical analysis on the acoustic emission signal. The Mayer wavelet is used as the wavelet basis to perform wavelet packet decomposition on the acoustic emission signal, and realizes the extraction of wavelet packet energy characteristics of the acoustic emission signal. An 8-dimensional feature vector is provided for the multivariate statistical analysis of the acoustic emission signal. A k-means clustering analysis algorithm model is established. The extracted feature vector is input into the k-means clustering analysis algorithm model, and finally the three types of damage of the composite material connection specimen are identified and observed under an electron microscope through acoustic emission feature analysis: matrix cracking, adhesive debonding and fiber breakage.

[0137] The beneficial effects of the present invention are as follows: the present invention provides a structural damage identification method for composite material connectors based on acoustic emission wavelet packet energy characteristics, which takes acoustic emission signals as the analysis basis, has high sensitivity, can realize dynamic monitoring, continuous monitoring, is not affected by structural complexity, has real-time data, and is applicable to a variety of composite materials; uses the Welch method to calculate the signal power spectrum, which can reduce variance, reduce leakage effect, improve frequency resolution, and better reflect the characteristics of non-stationary signals changing over time; uses wavelet packet decomposition to extract the energy characteristics of acoustic emission signals, which not only retains the time-frequency analysis characteristics of wavelet transform, but also can realize the simultaneous analysis of the low-frequency part and the high-frequency part of the signal. Solution, by selecting different resolutions at different frequency levels, the original acoustic emission signal can be completely decomposed in the entire frequency band; using the Mayer wavelet with compact support, smoothness, asymmetry, and multi-resolution analysis characteristics as the wavelet basis of the wavelet transform, its compact support enables it to effectively capture the local details of the signal, while its smoothness and asymmetry give it sensitivity to the overall characteristics and transient characteristics of the signal; using the unsupervised k-means machine learning algorithm to cluster the acoustic emission signals, it is simple to understand, highly efficient, and highly scalable. It does not require the pre-specified prototype of the cluster, has wide applicability, is easy to parallelize, has rigorous mathematical theory, and its results are interpretable. Finally, the present invention can better classify and identify the damage of composite materials through acoustic emission signals.

[0138] The embodiments described above are only descriptions of the preferred implementation modes of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A composite material structure damage identification method based on acoustic emission wavelet packet energy characteristics, characterized in that: It includes: S1: Acquire the acoustic emission signal data of the composite material and pre-process the acoustic emission signal data; An acoustic emission sensor is installed on the structural surface of the composite material, a force loading test is performed on the composite material using a universal testing machine, and acoustic emission signal data is collected; the acoustic emission signal data is preprocessed, a power spectrum of the acoustic emission signal data is calculated using the Welch power spectrum method, the peak frequency of the acoustic emission signal data is extracted, a peak frequency threshold is set, and the acoustic emission signal data with a peak frequency lower than the threshold is filtered out as noise to obtain the denoised composite material acoustic emission signal data f(t); S2: using wavelet basis functions and discrete wavelet inverse transform equations to perform wavelet packet decomposition on the composite material acoustic emission signal data after denoising in step S1; S21: Using wavelet basis function, the composite material acoustic emission signal data after denoising in step S1 is decomposed into: Wherein, f(t) is the processed acoustic emission signal data; F(ω) is the frequency domain function of the acoustic emission signal data after Fourier transform of f(t); is an imaginary unit; ω is the angular frequency of the acoustic emission signal data; t is time; WT(a,τ) is the time-frequency function of the acoustic emission signal data after wavelet transformation; τ is the time translation of the acoustic emission signal data; a is the time expansion of the acoustic emission signal data; ψ is the mother wavelet function; S22: Performing a discrete wavelet inverse transform on the composite material acoustic emission signal data f(t) to obtain a first component and a second component of the composite material acoustic emission signal data. The discrete wavelet inverse transform equation is: Among them, DWT(i,k) is the wavelet transform coefficient of the acoustic emission signal data; i is the decomposition level of the acoustic emission signal data; k is the discrete time domain translation; S3: Obtaining the composite material acoustic emission signal data processed in step S2, and extracting the characteristic energy of each acoustic emission signal data; The energy of the wavelet packet coefficient of each acoustic emission signal data and the energy proportion of each frequency band in the composite material acoustic emission signal data are calculated, and the relative energy distribution of each energy level of the composite material is extracted as follows: Among them, P i j (t) is the relative energy distribution of the energy level of the i-th layer of the composite material; E is the frequency band energy of the acoustic emission signal data; Total (t) is the energy of the wavelet packet coefficients of the acoustic emission signal data; Each acoustic emission signal data is decomposed into the i-th layer using wavelet packets, and the acoustic emission signal data is decomposed into the 2 i-th layer. i The wavelet packet energy ratio features form a 2 i The characteristic energy of each energy level of the composite material is: Among them, X i is the characteristic energy of the i-th energy level of the composite material; S4: Use the k-means clustering method to cluster the characteristic energies of each energy level of the composite material in step S3 to identify the type of structural damage of the composite material, thereby monitoring the health of the composite material throughout its life cycle.

2. The composite material structure damage identification method based on acoustic emission wavelet packet energy characteristics according to claim 1 is characterized by: In step S1, the power spectrum of the acoustic emission signal data is calculated using the Welch power spectrum method, which specifically includes the following steps: S11: segmenting the acoustic emission signal data into N segments. In order to obtain better frequency resolution in the frequency domain and make the power spectrum smoother, a 20% overlap is set between each window. S12: adding a window to the acoustic emission signal data, performing window function processing on each segment of data, reducing the leakage effect of spectrum estimation, and reducing the frequency resolution loss introduced by the window function; S13: Performing Fourier transform on the signal in each acoustic emission signal data window; S14: Calculate the power spectrum density estimation value of each frequency band of the acoustic emission signal data.

3. The composite material structure damage identification method based on acoustic emission wavelet packet energy characteristics according to claim 1 is characterized by: The first component of the composite material acoustic emission signal data in step S22 is calculated using the discrete wavelet inverse transform equation, specifically: Wherein, AVC is the frequency bandwidth of the first component after the acoustic emission signal data is decomposed by wavelet packet; f s is the sampling rate of the acoustic emission signal data; 2 i is the number of components in the i-th layer after wavelet packet decomposition of the acoustic emission signal data.

4. The composite material structure damage identification method based on acoustic emission wavelet packet energy characteristics according to claim 1 is characterized by: The second component of the composite material acoustic emission signal data emission signal in step S22 is calculated using the discrete wavelet inverse transform equation, specifically: Wherein, DQ is the frequency bandwidth of the second component of the acoustic emission signal data after the acoustic emission signal data is decomposed by wavelet packet.

5. The composite material structure damage identification method based on acoustic emission wavelet packet energy characteristics according to claim 1 is characterized by: The discrete wavelet inverse transform equation in step S22 is determined according to the discrete wavelet transform mathematical model, specifically: Where c is the discrete wavelet transform coefficient.

6. The composite material structure damage identification method based on acoustic emission wavelet packet energy characteristics according to claim 1 is characterized by: In step S3, the energy of the wavelet packet coefficient of each acoustic emission signal data in the composite material acoustic emission signal data and the energy proportion of each frequency band are calculated, specifically: The energy criterion is used to distinguish the energy distribution of each composite material, and the energy level of each acoustic emission signal data is set as: Where j = 1,...,2 i is the number of the jth component of the wavelet packet decomposition to the i-th layer; f i j is the component of the i-th layer of the decomposed signal; is the frequency band energy of the acoustic emission signal data; t0 is the starting time of the acoustic emission signal data; The energy of the wavelet packet coefficients of the acoustic emission signal data is set to: Among them, E Total (t) is the energy of the wavelet packet coefficients of the acoustic emission signal data; According to the energy level of each acoustic emission signal data and the total energy of the acoustic emission signal data, the relative energy distribution P of each energy level of the composite material is determined. i j .

7. The composite material structure damage identification method based on acoustic emission wavelet packet energy characteristics according to claim 1 is characterized by: In step S4, the characteristic energy of each energy level of the composite material in step S3 is clustered and analyzed using the k-means clustering method, specifically: The characteristic energy of each energy level of n groups of composite materials is X={X1,X2…X i …X n Input the k-means clustering method and divide it into k clusters so that the distance between each characteristic energy data sample and the center point of the cluster to which it belongs is minimized. Initialize the centers of the k clusters: Calculate the Euclidean distance from each acoustic emission signal data point to each cluster center as follows: Among them, Dis(X i ,C p ) is the Euclidean distance from the ith acoustic emission signal data point to the pth cluster center; p is the cluster center number; X i is the characteristic energy of the i-th energy level of the composite material; C p is the cluster center of the pth class, {C1,…C p …,C k },1<p≤n;X it is the t-th characteristic energy coordinate value of the i-th acoustic emission characteristic energy data sample; C pt is the t-th characteristic energy coordinate value of the p-th cluster center; t is the characteristic energy coordinate value number; m is the characteristic energy dimension.

8. The composite material structure damage identification method based on acoustic emission wavelet packet energy characteristics according to claim 1 is characterized by: The types of composite material structure damage in step S4 include matrix cracking MC, adhesive debonding AD and fiber breakage FB; the acoustic emission signal energy generated by the three types of damage is concentrated in different frequency bands, and the composite material structure damage type identification is achieved through the acoustic emission damage information contained in the wavelet packet energy; The matrix cracking MC damage is that there is a bonding part in the composite material structure, and the fiber needs to bear pressure. The fiber at the bonding part first breaks to a certain extent, resulting in a gradual decrease in the mechanical properties of the composite material structure. The adhesive debonding AD damage is manifested as the mechanical properties of the composite material structure decrease as the pressure on the composite material structure increases, and the adhesive begins to debond. The continuous debonding of the adhesive causes the fibers on the surface of the composite material structure to be gradually pulled out and broken. The fiber breakage FB damage is manifested as cracking and debonding of the adhesive, and a large number of fibers are pulled out and broken, resulting in cracking of the resin matrix connected to the fibers and with weaker mechanical properties.

Images