Fiber laminate damage acoustic emission signal processing method, equipment and medium

Abstract

The invention discloses a fiber laminate damage acoustic emission signal processing method, equipment and a medium, and relates to the technical field of signal processing.The method comprises the steps that a fiber laminate is fixed to a test device, signals are collected through an acoustic emission sensor, filtering and preprocessing are carried out, features are extracted through an improved ReliefF algorithm, and a fiber laminate damage acoustic emission signal is obtained; randomly selecting samples, comparing the samples with nearest neighbors, updating feature weights, and screening most relevant features. An expectation maximization algorithm and a Gaussian mixture model are combined to establish a damage identification model, different types of damage are automatically classified, and a Gaussian distribution model is obtained through iterative optimization. And by utilizing a clustering analysis and feature selection algorithm, the damage of the laminated plate with different layering angles is identified, and different damage types are accurately distinguished. According to the method, high-precision damage classification is realized through feature extraction in combination with an expectation maximization algorithm and a Gaussian mixture model, laminated plate damage analysis at different layering angles is dynamically adapted, and feature extraction efficiency, classification accuracy and model adaptability are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of signal processing, and particularly to a method, device and medium for processing acoustic emission signals of fiber laminated plate damage. Background Art

[0002] Due to its higher specific strength and stiffness than metals, glass fiber reinforced plastic (GFRP) has been widely used in critical structural applications. The initiation and propagation of delamination damage in fiber reinforced composites is an extremely complex process, such as matrix cracking within or between layers, fiber fracture, etc. A large number of experiments and analyses are dedicated to studying the failure mode recognition of multi-directional fiber reinforced composites.

[0003] When using acoustic emission to collect structural damage signals, especially for fiber reinforced composites, a high-dimensional data set will be formed. There are a large number of redundant and irrelevant features in the high-dimensional data set, which increases the difficulty of damage recognition. Therefore, in many studies, data dimensionality reduction is used as a preprocessing step to improve the calculation and learning accuracy. Although researchers have conducted a large number of studies on the association between damage and acoustic emission features, there is still a large research space for the most suitable feature selection or the best way of damage characterization. Summary of the Invention

[0004] Based on the above-mentioned disadvantages of the prior art, the purpose of the present invention is to provide a method, device and medium for processing acoustic emission signals of fiber laminated plate damage to solve the above technical problems.

[0005] To achieve the above purpose, the present invention provides the following technical solutions: A method for processing acoustic emission signals of fiber laminated plate damage, including:

[0006] S1: Fix the fiber laminated plate on the test device, and collect the acoustic emission signals generated during the direct shear test through the acoustic emission sensors installed on the surface of the device to obtain data containing time domain and frequency domain information;

[0007] S2: Clean the collected acoustic emission signals, use filtering and preprocessing methods to remove irrelevant noises, and select the effective signals within the main frequency band based on the frequency range;

[0008] S3: Use the improved ReliefF algorithm to extract feature parameters from the high-dimensional data set of acoustic emission signals. By randomly selecting samples and comparing them with the nearest neighbor samples, gradually update the weights of the feature parameters, and screen out the most relevant feature parameters;

[0009] S4: Combine the expectation maximization algorithm and the Gaussian mixture model to establish an automated damage recognition model. According to the screened out most relevant feature parameters, automatically identify and classify different types of damage, and obtain the Gaussian distribution model for damage classification through iterative solution of the expectation maximization algorithm;

[0010] S5: Utilize the clustering analysis results, through the feature selection algorithm and the Gaussian distribution clustering algorithm, to identify the damage of laminates with different ply angles, and compare the characteristic parameters of the acoustic emission signals under different damage modes to accurately distinguish the damage types.

[0011] The present invention is further configured such that the characteristic parameters include duration / amplitude, amplitude / initial frequency, amplitude / center frequency, rise time / amplitude, rise time / duration, amplitude / average frequency, peak factor, waveform factor, and amplitude mean value.

[0012] The present invention is further configured such that in step S3, the improved ReliefF algorithm is used to extract features from the high-dimensional dataset of acoustic emission signals, including:

[0013] S31: According to the collected damage acoustic emission signals, process or delete the error domain, and label the data to obtain the input data with labels.

[0014] S32: Input the input data into the training set D, and set the number of sampling times m, the feature weight threshold δ, and the number of nearest neighbor samples k.

[0015] S33: Set all feature weights to 0.

[0016] S34: For i = 1:m, randomly select a sample R from the training set D, find k nearest neighbors H j (j = 1, 2,..., k) of R from the set of samples of the same class as R, and find k nearest neighbors M j (C) from each set of samples of different classes, and obtain Repeat the above steps N times until the final correlation vector of the features is obtained, and filter out the most relevant feature parameters according to the preset feature weight threshold δ, where W'(A) is the weight of feature A after iterative update, W(A) is the weight of feature A, which is used to measure the importance of feature A for the classification task, with an initial value of 0, diff(A, R, H j ) is the difference between sample R and H j on feature A, p(C) is the prior probability of class C, class(R) is the class to which sample R belongs, M j (C) is the j-th nearest neighbor sample in the class, which describes the distribution relationship between R and samples of other classes, and class C is a certain class different from the class to which R belongs.

[0017] The present invention is further configured such that, in combination with the expectation maximization algorithm and the Gaussian mixture model, an automated damage identification model is established, including:

[0018] According to the Gaussian mixture model, the probability density function is calculated as follows: where p(x) is the probability density function, representing the probability of the given data point x occurring. T is the number of mixture components, and N(x|μ t ,σ t ) is the probability density of the t-th component in the mixture model, and π t is the mixing coefficient, and it satisfies 0 ≤ π t ≤ 1, μ t and σ t are the Gaussian parameters of the t-th component;

[0019] The likelihood function is calculated based on the probability density function: lnP(x|π, μ, σ) = where lnP(x|π, μ, σ) is the likelihood function.

[0020] The present invention is further configured to obtain the parameters π t , μ t and σ t through the expectation-maximization algorithm, specifically including:

[0021] Based on the assumed values of the parameters, an expected estimate of the unknown variable is given, and the membership degree of the observed data to each cluster T is calculated through the variable γ(z it ): γ(z it ) represents the probability that the sample y j belongs to the t-th category, and is used to measure the membership degree of the data point y j belonging to each Gaussian distribution. P(z = t, y j ∣μ, σ) is the joint probability that the sample y j belongs to the t-th category;

[0022] Maximize the log-likelihood function based on the current parameters to solve for the new round of iterative parameters:

[0023] Repeat the iteration until convergence to obtain the final parameters, and substitute them into the likelihood function lnP(x|π, μ, σ) for solution.

[0024] The present invention is further configured to include the clustering analysis results:

[0025] Evaluate and compare the performance of the damage identification model under different numbers of clusters by using the Akaike information criterion and Bayesian information criterion indicators, where the different numbers of clusters include 2 or more;

[0026] Input the most relevant characteristic parameter data into the damage identification model for training. After iterative calculation by the expectation-maximization algorithm, stable parameters are obtained and clustering is performed. Among them, the characteristics of different clusters are related to the damage mechanism;

[0027] Finally, by combining the load, acoustic emission curve, and microstructure changes, analyze the relationship between the damage mechanism of laminated plates with different interfacial fiber orientations and the clustering results, verify the rationality of the clustering analysis results, and show that it can reflect the damage characteristics of the material.

[0028] The present invention is further configured to identify the damage of laminated plates with different ply angles, and compare the characteristic parameters of acoustic emission signals under different damage modes to accurately distinguish the damage types, including:

[0029] In the mode I cracking experiment of glass / epoxy laminated plates with different ply angles, acoustic emission signals are collected by acoustic emission sensors, and the original waveforms are converted into AE hits. Each AE hit contains multiple basic parameters, and the basic parameters include rise time, count, energy, duration, amplitude, average frequency, back-calculated frequency, center frequency, and peak frequency. Composite parameters are constructed based on the basic parameters, and the basic parameters and composite parameters are set as a set of characteristic parameters to form a data set containing time-domain and frequency-domain parameters. Among them, different ply angles include 0°, 30°, 45°, and 60°;

[0030] Apply the Relief F algorithm to the data set for feature selection, select the characteristic parameters, use principal component analysis to reduce the data dimension to two dimensions, use the first principal component as the X-axis and the second principal component as the Y-axis, perform Gaussian mixture clustering on the new coordinates, and determine the optimal number of clusters to be 3 through the Akaike information criterion and Bayesian information criterion, and finally obtain the clustering results of different damage mechanisms, so as to realize the distinction of different damage mechanisms.

[0031] The present invention is further configured to integrate the damage identification model into a real-time monitoring system. The real-time monitoring system analyzes the acoustic emission signal data automatically and immediately issues an alarm when damage is detected. At the same time, the real-time monitoring system provides detailed information on the damage location and type to provide guidance for maintenance and repair.

[0032] The present invention also provides an electronic device, which includes:

[0033] One or more processors;

[0034] A storage device for storing one or more programs, which when executed by the one or more processors, enable the electronic device to implement a method for processing acoustic emission signals of fiber laminated plate damage as described in any one of the above.

[0035] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor of a computer, the computer is caused to execute a method for processing acoustic emission signals of a fiber laminate as described in any one of the above.

[0036] The present invention provides a method, device and medium for processing acoustic emission signals of a fiber laminate. The method fixes the fiber laminate on a test device, and collects acoustic emission signals generated during a direct shear test through an acoustic emission sensor installed on the surface of the device to obtain data containing time-domain and frequency-domain information; cleans the collected acoustic emission signals, uses filtering and preprocessing methods to remove irrelevant noises, and selects effective signals within the main frequency band based on the frequency range; extracts characteristic parameters from the high-dimensional data set of the acoustic emission signals by using an improved ReliefF algorithm, gradually updates the weights of the characteristic parameters by randomly selecting samples and comparing them with the nearest neighbor samples, and screens out the most relevant characteristic parameters; combines the expectation maximization algorithm and the Gaussian mixture model to establish an automated damage recognition model, automatically recognizes and classifies different types of damage according to the screened most relevant characteristic parameters, and obtains a Gaussian distribution model for damage classification through iterative solution of the expectation maximization algorithm; uses the clustering analysis results, through a feature selection algorithm and a Gaussian distribution clustering algorithm, to identify the damage of laminates with different ply angles, and compares the characteristic parameters of the acoustic emission signals under different damage modes to accurately distinguish the damage types. The beneficial effects produced include:

[0037] 1. Improve the efficiency and accuracy of feature extraction: By using the improved ReliefF algorithm to extract characteristic parameters from the high-dimensional acoustic emission signal data set, the features most relevant to the damage mechanism can be screened out, redundant data can be reduced, and the efficiency and accuracy of feature extraction are improved. By screening out the key characteristic parameters, the data dimension is effectively reduced and the computational complexity is reduced;

[0038] 2. Enhance the damage recognition ability: By combining the expectation maximization algorithm and the Gaussian mixture model, the damage signals are automatically classified. By using the Gaussian distribution model to model and analyze different damage mechanisms, high-precision damage classification is achieved. By using the Gaussian mixture clustering algorithm, the balance between the inter-class distance and the intra-class dispersion of the model is ensured, and the accuracy of distinguishing damage mechanisms is significantly improved;

[0039] 3. Precise adaptation to different ply angles: For laminates with different ply angles, through data dimensionality reduction and clustering analysis, the damage modes under different ply angles are accurately distinguished, providing an effective method for damage analysis of complex fiber structures. By combining the Akaike information criterion and the Bayesian information criterion, the optimal number of clusters is dynamically selected, making the model more adaptable under different ply angles.

[0040] The above description is only an overview of the technical solution of this application. In order to be able to understand the technical means of this application more clearly, it can be implemented in accordance with the content of the specification. And in order to make the above and other purposes, features and advantages of this application more obvious and understandable, the following specifically illustrates the specific implementation manners of this application. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings. In the drawings:

[0042] Figure 1 is a flowchart of a method for processing acoustic emission signals of fiber laminate damage shown in an exemplary embodiment of the present invention;

[0043] Figure 2 is a schematic diagram of a DCB test device and an acoustic emission data acquisition device for a method for processing acoustic emission signals of fiber laminate damage shown in an exemplary embodiment of the present invention;

[0044] Figure 3 is a trained Gaussian mixture model represented by principal components, a damage identification model for four types of angle specimens (a - d) shown in an exemplary embodiment of the present invention;

[0045] Figure 4 is a multivariate normal distribution contour map of four types of angle damage identification models (a - d) shown in an exemplary embodiment of the present invention;

[0046] Figure 5 is a curve graph showing the change of Akaike information criterion and Bayesian information criterion with the number of clustering components shown in an exemplary embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] The following will illustrate the embodiments of the present invention with reference to the drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific implementation manners. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention, rather than for limiting the protection scope of the present invention.

[0048] It should be noted that the illustrations provided in the following embodiments only schematically illustrate the basic concept of the present invention. Therefore, only the components related to the present invention are shown in the drawings, rather than being drawn according to the number, shape, and size of the components in actual implementation. The types, quantities, and proportions of the components in actual implementation can be arbitrarily changed, and the component layout type may also be more complex.

[0049] In the following description, a large number of details are explored to provide a more thorough explanation of the embodiments of the present invention. However, it is obvious to those skilled in the art that the embodiments of the present invention can be implemented without these specific details. In other embodiments, well-known structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present invention difficult to understand.

[0050] A method for processing acoustic emission signals of fiber laminates, as Figure 1 shown, includes:

[0051] S1: Fix the fiber laminate on the test device, and collect the acoustic emission signals generated during the direct shear test through the acoustic emission sensors installed on the surface of the device to obtain data containing time-domain and frequency-domain information;

[0052] S2: Clean the collected acoustic emission signals, use filtering and preprocessing methods to remove irrelevant noise, and select the effective signals within the main frequency band based on the frequency range;

[0053] S3: Use the improved ReliefF algorithm to extract characteristic parameters from the high-dimensional data set of acoustic emission signals. By randomly selecting samples and comparing them with the nearest neighbor samples, gradually update the weights of the characteristic parameters, and screen out the most relevant characteristic parameters;

[0054] S4: Combine the expectation maximization algorithm and the Gaussian mixture model to establish an automated damage identification model. According to the most relevant characteristic parameters screened out, automatically identify and classify different types of damage, and obtain the Gaussian distribution model for damage classification through iterative solution of the expectation maximization algorithm;

[0055] S5: Utilize the clustering analysis results, through the feature selection algorithm and the Gaussian distribution clustering algorithm, identify the damage of laminates with different ply angles, and compare the characteristic parameters of the acoustic emission signals under different damage modes to accurately distinguish the damage types.

[0056] Specifically, in step S1, the laminate is fixed to an electronic universal testing machine through a hinge. Two AE sensors are installed on the upper surface of the specimen, and the laminate is monitored and data is collected through an acoustic emission sensor and equipment. The acquired data includes impact signals and continuous waveform signals during tensile testing, ensuring that the collected data contains sufficient time-domain and frequency-domain information to reduce the phenomenon of internal information loss in the signals.

[0057] The specific process is as follows:

[0058] 1. After preparing the laminate specimen, in order to eliminate factors such as matrix cracking at the specimen edges, it is polished with 120-mesh sandpaper.

[0059] 2. Prepare the sensor and data acquisition system: Select appropriate sensors and data acquisition systems to obtain acoustic emission signals. Piezoelectric sensors are used as sensors.

[0060] 3. Install the sensor: Install the sensor on the object to be measured. The position and quantity of the sensors should be planned according to requirements to ensure that accurate acoustic emission signals can be obtained.

[0061] 4. Calibrate the sensor: Calibrate the sensor to ensure the accuracy and reliability of its measurement results. The calibration process includes sensitivity calibration, frequency response calibration, etc.

[0062] 5. Set the data acquisition system: Set the data acquisition system according to requirements, including parameters such as sampling rate and filter. The selection of these parameters will depend on the characteristics of the object to be measured and the required signal resolution.

[0063] 6. Start data acquisition: Start the data acquisition system and start collecting acoustic emission signals and continuous waveform data. Figure 2 Shows a schematic diagram of a DCB delamination experimental test device and an acoustic emission data acquisition device according to an exemplary embodiment of the present disclosure. The DCB test device and the AE data acquisition device are as Figure 2 shown. The DCB test device consists of two main data acquisition systems. The electronic universal testing machine records data such as load displacement, load time, and stress-strain curves of the crosshead. A high-optical zoom digital camera is set up to record crack initiation and propagation during the test.

[0064] 7. Data recording and storage: Record the acquired data and perform appropriate storage and backup. The data can be stored on a computer, a data collector, or other storage media for subsequent analysis and processing.

[0065] 8. Data analysis and interpretation: The acoustic emission sensor continuously receives the acoustic emission signals generated by the damage source, analyzes and interprets the collected data, and extracts useful information. Based on the collected acoustic emission signals, the relationship with damage is established to achieve the intelligent diagnosis of laminate damage. By using the hit definition, the original waveform is converted into AE hits. The time-domain and frequency-domain characteristics of acoustic emission together represent the characteristics of the damage source. Therefore, the parameter set combines time-domain and frequency-domain parameters to analyze the damage degree of the specimen and the types of internal damage.

[0066] The present invention is further configured such that the characteristic parameters include duration / amplitude, amplitude / initial frequency, amplitude / center frequency, rise time / amplitude, rise time / duration, amplitude / average frequency, peak factor, waveform factor, and amplitude mean value. Specifically, duration / amplitude represents the ratio of the duration of the acoustic emission signal to its amplitude, reflecting the comprehensive relationship between signal intensity and time characteristics; amplitude / initial frequency represents the ratio of the signal amplitude to its initial frequency (the main frequency at the start of the signal), reflecting the relationship between signal intensity and initial frequency characteristics; amplitude / center frequency represents the ratio of the signal amplitude to its center frequency, reflecting the relationship between the central position of signal energy distribution and intensity; rise time / amplitude represents the ratio of the time required for the signal to reach the maximum amplitude from the starting point to the amplitude, reflecting the relationship between signal rise speed and intensity; rise time / duration represents the ratio of the signal rise time to the entire duration, reflecting the proportion of the signal rise stage in the entire signal; amplitude / average frequency represents the ratio of the signal amplitude to its average frequency, reflecting the relationship between signal intensity and frequency characteristics; peak factor represents the ratio of the maximum amplitude of the signal to its root mean square value (RMS), reflecting the transient peak intensity of the signal; waveform factor represents the ratio of the root mean square value of the signal to its absolute average value, reflecting the waveform characteristics of the signal; amplitude mean value represents the average value of the amplitudes of all sampling points of the signal over the entire duration. In previous literature, in the laminate tensile experiment, the frequency between 350 - 500 kHz refers to fiber fracture, matrix cracking is locked between 80 - 120 kHz, and delamination damage is between these two frequency bands. Therefore, the data was preprocessed before running Relief N to remove irrelevant noise and select the effective signals within the main frequency band based on the frequency range.

[0067] The present invention is further configured such that in step S3, the improved ReliefF algorithm is used to extract features from the high-dimensional dataset of acoustic emission signals, including:

[0068] S31: According to the collected damage acoustic emission signals, process or delete the error domain, and label the data to obtain the input data with labels.

[0069] S32: Input the input data into the training set D, and set the number of sampling times m, the feature weight threshold δ, and the number of nearest neighbor samples k.

[0070] S33: Set all feature weights to 0.

[0071] S34: For i = 1:m, randomly select a sample R from the training set D, and find k nearest neighbors H of R from the set of samples of the same class as R j (j = 1, 2,..., k), and find k nearest neighbors M from each set of samples of different classes j (C), and obtain Repeat the above steps N times until the final correlation vector of the features is obtained, and filter out the most relevant feature parameters according to the preset feature weight threshold δ. Among them, W'(A) is the weight of feature A after iterative update, W(A) is the weight of feature A, which is used to measure the importance of feature A to the classification task, and the initial value is 0. diff(A, R, H j ) is the difference between sample R and H j on feature A, p(C) is the prior probability of class C, class(R) is the class to which sample R belongs, and M j (C) is the j-th nearest neighbor sample in the class, which describes the distribution relationship between R and samples of other classes. The class C is a certain class different from the class to which R belongs. Specifically, the ReliefF algorithm was first extended by Kononeill on the basis of the Relief algorithm and can handle multi-class problems. It is one of the commonly used feature selection methods in data mining. Each time a sample R is randomly selected from the sample set, and then k nearest neighbor samples (near Hits) of R are found from the set of samples of the same class as R, and k nearest neighbor samples (near Misses) are found from each set of samples of different classes of R, and then the weight of each feature is updated; in the above formula, diff(A, R1, R2) represents the difference between samples R1 and R2 on feature A, and M j (C) represents the j-th nearest neighbor sample in the class. As shown in the following formula:

[0072] In the calculation logic of the present invention, by improving the ReliefF algorithm and introducing the problem of imbalance in the number of class samples is balanced, so that the algorithm can still accurately identify important features in the case of class imbalance.

[0073] The present invention is further configured to establish an automated damage identification model in combination with the expectation maximization algorithm and the Gaussian mixture model, including:

[0074] According to the Gaussian mixture model, the probability density function is calculated as follows: where p(x) is the probability density function, representing the probability of the given data point x occurring, T is the number of mixture components, and N(x|μ t ,σ t ) is the probability density of the t-th component in the mixture model, π t is the mixing coefficient, and satisfies 0 ≤ π t ≤ 1, μ t and σ t are the Gaussian parameters of the t-th component; specifically, determining the class of a point in a cluster can actually be divided into two steps: (1) randomly select any class, and the probability of each class being selected is π t ; (2) separately consider selecting a point from the distribution in this class.

[0075] The likelihood function is calculated according to the probability density function: where lnP(x|π, μ, σ) is the likelihood function.

[0076] Since the collected data has no labels, the exact number of various data points cannot be determined, and there are latent variables in the solution process. To obtain the parameters π t , μ t and σ t , the expectation maximization algorithm needs to be introduced. The present invention is further configured to obtain the parameters π t , μ t and σ t through the expectation maximization algorithm, specifically including:

[0077] According to the assumed values of the parameters, give the expected estimate of the unknown variables, and calculate the membership degree of the observed data to each cluster T through the variable γ(z it ): γ(z it ) represents the probability that the sample y j belongs to the t-th class, and is used to measure the membership degree of the data point y j belonging to each Gaussian distribution. P(z = t, y j ∣μ, σ) is the joint probability that the sample y j belongs to the t-th class;

[0078] Maximize the log-likelihood function based on the current parameters to solve for the new round of iterative parameters:

[0079] Repeat the iteration until convergence to obtain the final parameters, and substitute them into the likelihood function lnP(x|π, μ, σ) for solution. Specifically, please refer to Figure 3 ,Figure 3 The trained Gaussian mixture model of the principal component representation shown, (a-d) damage identification models for four types of angle specimens. The T components of the Gaussian mixture model indicate how many Gaussian distribution models the model consists of. Determining the optimal number of clusters is often the first problem to be solved in cluster analysis, and the choice of T will directly affect the performance of clustering.

[0080] The present invention is further configured such that the cluster analysis results include:

[0081] Evaluating and comparing the performance of the damage identification model under different numbers of clusters by using the Akaike information criterion and the Bayesian information criterion metrics, wherein the different numbers of clusters include 2 or more;

[0082] Inputting the most relevant feature parameter data into the damage identification model for training, obtaining stable parameters through iterative expectation-maximization algorithm, and performing clustering, wherein the characteristics of different clusters are related to the damage mechanism;

[0083] Finally, combining the load, acoustic emission curve and microstructure change, analyzing the relationship between the damage mechanism of the laminated plate with different interface fiber orientations and the clustering results, and verifying the rationality of the cluster analysis results, indicating that it can reflect the damage characteristics of the material.

[0084] The present invention is further configured to identify the damage of laminated plates with different ply angles, and compare the characteristic parameters of acoustic emission signals under different damage modes to accurately distinguish the damage types, including:

[0085] In the mode I cracking experiment of glass / epoxy laminated plates with different ply angles, collecting acoustic emission signals through acoustic emission sensors, converting the original waveform into AE hits, each AE hit containing multiple basic parameters, the basic parameters including rise time, count, energy, duration, amplitude, average frequency, back-calculated frequency, center frequency and peak frequency, constructing composite parameters according to the basic parameters, setting the basic parameters and composite parameters as the feature parameter set, forming a data set containing time-domain and frequency-domain parameters, wherein the different ply angles include 0°, 30°, 45° and 60°; specifically, determining the parameters π t , μ t and σ t , to complete the training of the Gaussian mixture model. For laminated plates with different ply angles, the Gaussian mixture model is represented by a multivariate normal distribution as Figure 4, it can be observed from these distributions that the data structures clearly represent different damage mechanisms. The more compact the distribution of each cluster, the higher the similarity of the data, indicating that most of the data in this type of damage have similar parameter characteristics or are easy to determine the value range. There are clusters with a large concentration of data in the Gaussian density maps of the four angles, and the distribution range is small. The signal caused by matrix cracking in Figure a, because for unidirectional laminates, matrix cracking accounts for the main part of the overall signal volume. The other two clusters show different distributions and shapes in the four figures, which is due to the differences in data characteristics of delamination and fiber fracture events caused by different fiber orientations.

[0086] Apply the Relief F algorithm to the data set for feature selection, select feature parameters, use principal component analysis to reduce the data dimension to two dimensions, use the first principal component as the X-axis and the second principal component as the Y-axis, perform Gaussian mixture clustering on the new coordinates, and determine the optimal number of clusters to be 3 through the Akaike information criterion and the Bayesian information criterion. Finally, obtain the clustering results of different damage mechanisms. For example, different clusters show obvious differences in amplitude and peak frequency, thus realizing the distinction of different damage mechanisms, as Figure 5 shown. Specifically, using the clustering analysis results, through the feature selection algorithm and the clustering algorithm based on the Gaussian mixture model, identify the damage of laminates with different ply angles, assign higher weights to the low-frequency characteristics of matrix cracking signals, and assign stronger weights to the high-frequency characteristics of fiber fracture signals. By comparing the acoustic emission parameter characteristics under different damage modes, use the Gaussian mixture model to perform soft clustering on the selected features. Each data point is assigned to a different damage category, and the damage type is judged by the characteristic parameters of the probability density function of each cluster, which has a good effect on the discrimination of damage mechanisms.

[0087] The present invention is further configured to integrate the damage recognition model into a real-time monitoring system. The real-time monitoring system automatically analyzes the acoustic emission signal data and immediately issues an alarm when damage is detected. At the same time, the real-time monitoring system provides detailed information on the damage location and type, providing guidance for maintenance and repair.

[0088] Through the above steps, the present invention provides an efficient and accurate method for monitoring and identifying damage to composite laminates, greatly reducing the data dimension, reducing the computational amount, being able to better identify different damage mechanisms, and providing strong technical support for the health monitoring and maintenance of composite structures.

[0089] An embodiment of the present application further provides an electronic device, including: one or more processors; a storage device for storing one or more programs, which when executed by the one or more processors, cause the electronic device to implement a method for processing acoustic emission signals of fiber laminate damage provided in each of the above embodiments.

[0090] An embodiment of the present application further provides a computer system of an electronic device. It should be noted that the computer system of the electronic device is only an example and should not impose any limitations on the functions and usage scope of the embodiments of the present application.

[0091] Specifically, the computer system includes a Central Processing Unit (CPU), which can perform various appropriate actions and processes according to the program stored in the Read-Only Memory (ROM) or the program loaded from the storage part into the Random Access Memory (RAM), such as executing the method described in the above embodiments. In the RAM, various programs and data required for system operation are also stored. The CPU, ROM, and RAM are connected to each other via a bus. The Input / Output (I / O) interface is also connected to the bus.

[0092] The following components are connected to the I / O interface: an input part including a keyboard, a mouse, etc.; an output part including such as a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), etc. and a speaker; a storage part including a hard disk, etc.; and a communication part including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication part performs communication processing via a network such as the Internet. A drive is also connected to the I / O interface as needed. Removable media, such as disks, optical discs, magneto-optical discs, semiconductor memories, etc., are installed on the drive as needed so that the computer programs read from them can be installed into the storage part as needed.

[0093] Particularly, according to the embodiments of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part, and / or installed from the removable media. When the computer program is executed by the Central Processing Unit (CPU), various functions defined in the system of the present application are executed.

[0094] It should be noted that the computer-readable medium shown in the embodiments of the present application may be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. A computer-readable storage medium may be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of the computer-readable storage medium may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable computer program. Such a propagated data signal may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable medium may be transmitted by any appropriate medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0095] The units involved in the embodiments described in the present application may be implemented in software or in hardware, and the described units may also be provided in a processor. Among them, the names of these units do not, in some cases, constitute a limitation on the unit itself.

[0096] On the other hand, the present application also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements a method for processing acoustic emission signals of a fiber laminate as described above. The computer-readable storage medium may be included in the electronic device described in the above embodiments, or may exist separately without being assembled into the electronic device.

[0097] Another aspect of the present application also provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes a method for processing acoustic emission signals of a fiber laminate provided in each of the above embodiments.

[0098] The above embodiments are only used to exemplarily illustrate the principles and effects of the present invention, rather than to limit the present invention. Any person familiar with this technology can modify or change the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or changes completed by those with ordinary knowledge in the technical field without departing from the spirit and technical idea disclosed by the present invention should still be covered by the claims of the present invention.

Claims

1. A method for processing acoustic emission signals of fiber laminate damage, characterized in that: include: S1: Fix the fiber laminate on the test device, collect the acoustic emission signal generated during the direct shear test through the acoustic emission sensor installed on the surface of the device, and obtain data containing time domain and frequency domain information; S2: Clean the collected acoustic emission signals, remove irrelevant noises by filtering and preprocessing, and select effective signals in the main frequency band based on the frequency range; S3: The improved ReliefF algorithm is used to extract feature parameters from the high-dimensional data set of acoustic emission signals. By randomly selecting samples and comparing them with the nearest neighbor samples, the weights of the feature parameters are gradually updated to screen out the most relevant feature parameters. S4: Combine the expectation maximization algorithm and the Gaussian mixture model to establish an automated damage identification model. By screening out the most relevant characteristic parameters, different types of damage can be automatically identified and classified. The Gaussian distribution model for damage classification is obtained through iterative solution of the expectation maximization algorithm. S5: Using the cluster analysis results, the damage of laminates with different ply angles is identified through feature selection algorithm and Gaussian distribution clustering algorithm, and the characteristic parameters of acoustic emission signals under different damage modes are compared to accurately distinguish the damage types.

2. A fiber laminate damage acoustic emission signal processing method according to claim 1, characterized in that: The characteristic parameters include duration / amplitude, amplitude / initial frequency, amplitude / center frequency, rise time / amplitude, rise time / duration, amplitude / average frequency, peak factor, waveform factor and amplitude mean.

3. A fiber laminate damage acoustic emission signal processing method according to claim 2, characterized in that: In step S3, the improved ReliefF algorithm is used to extract features from the high-dimensional data set of the acoustic emission signal, including: S31: processing or deleting the error domain according to the collected damage acoustic emission signal, and labeling the data to obtain input data with labels; S32: Input the input data into the training set D, set the sampling times m, the feature weight threshold δ, and the number of nearest neighbor samples k; S33: Set all feature weights to 0; S34: For i=1:m, randomly select a sample R from the training set D, and find the k nearest neighbors H of R from the same sample set as sample R. j (j=1,2,...,k), find k nearest neighbors M from each different class sample set j (C), we get Repeat the above steps N times to obtain the final correlation vector of the feature, and filter out the most relevant feature parameters according to the preset feature weight threshold δ, where W'(A) is the weight of feature A after iterative update, and W(A) is the weight of feature A, which is used to measure the importance of feature A to the classification task. The initial value is 0, and diff(A,R,H j ) are samples R and H j The difference on feature A, p(C) is the prior probability of category C, class(R) is the category to which sample R belongs, M j (C) is a class The jth nearest neighbor sample in describes the distribution relationship between R and samples of other classes. Class C is a class different from the class to which R belongs.

4. A fiber laminate damage acoustic emission signal processing method according to claim 3, characterized in that: Combining the expectation maximization algorithm and the Gaussian mixture model, an automated damage identification model is established, including: According to the Gaussian mixture model, the probability density function is calculated: Among them, p(x) is the probability density function, which represents the probability of a given data point x appearing, T is the number of mixed components, N(x|μ t ,σ t ) is the probability density of the tth component in the mixture model, π t is the mixing coefficient and satisfies 0≤π t ≤1,μ t and σ t is the Gaussian parameter of the tth component; The likelihood function is calculated based on the probability density function: Among them, lnP(x|π,μ,σ) is the likelihood function.

5. A fiber laminate damage acoustic emission signal processing method according to claim 4, characterized in that: The parameter π is obtained by the expectation maximization algorithm t , μ t and σ t , specifically including: According to the assumed values ​​of the parameters, the expected estimate of the unknown variable is given by the variable γ(z it ) Calculate the membership of the observation data to each cluster T: γ(z it ) represents sample y j The probability of belonging to the tth category, which is used to measure the probability of data point y j The membership degree of each Gaussian distribution, P(z=t,y j |μ,σ) is the sample y j The joint probability of belonging to the tth category; Based on the current parameters, maximize the log-likelihood function and solve the parameters for the next round of iterations: Repeat the iteration until convergence to obtain the final parameters, and substitute them into the likelihood function lnP(x|π,μ,σ) for solution.

6. A fiber laminate damage acoustic emission signal processing method according to claim 5, characterized in that: The cluster analysis results include: The performance of the damage identification model under different numbers of clusters is evaluated and compared by using the Akaike information criterion and the Bayesian information criterion indicators, wherein the different numbers of clusters include 2 or more; The most relevant characteristic parameter data is input into the damage identification model training, and the stable parameters are obtained through the iteration of the expectation maximization algorithm, and clustering is performed, where different cluster characteristics are related to the damage mechanism; Finally, combined with the load, acoustic emission curve and microstructure changes, the relationship between the damage mechanism and clustering results of laminates with different interface fiber orientations was analyzed, the rationality of the clustering analysis results was verified, and it was shown that it can reflect the characteristic performance of material damage.

7. A fiber laminate damage acoustic emission signal processing method according to claim 6, characterized in that: Identify the damage of laminates with different ply angles, and compare the characteristic parameters of acoustic emission signals under different damage modes to accurately distinguish the damage types, including: In the type I cracking experiment of glass / epoxy resin laminates with different ply angles, acoustic emission signals are collected by acoustic emission sensors, and the original waveforms are converted into AE impacts. Each AE impact contains multiple basic parameters, including rise time, count, energy, duration, amplitude, average frequency, back-calculated frequency, center frequency and peak frequency. Composite parameters are constructed based on the basic parameters, and the basic parameters and composite parameters are set as characteristic parameter sets to form a data set containing time domain and frequency domain parameters. The different ply angles include 0°, 30°, 45° and 60°. The Relief F algorithm is applied to the data set for feature selection. The feature parameters are selected, and the data is reduced to two dimensions using principal component analysis. The first principal component is used as the X-axis and the second principal component is used as the Y-axis. Gaussian mixture clustering is performed on the new coordinates. The optimal number of clusters is determined to be 3 through the Akaike information criterion and the Bayesian information criterion. Finally, the clustering results of different damage mechanisms are obtained, thereby realizing the distinction of different damage mechanisms.

8. The method for processing fiber laminate damage acoustic emission signals according to claim 1, characterized in that: The damage identification model is integrated into the real-time monitoring system, which automatically analyzes the acoustic emission signal data and immediately issues an alarm when damage is detected. At the same time, the real-time monitoring system provides detailed information on the location and type of damage to provide guidance for maintenance and repair.

9. An electronic device, characterized in that: The electronic device comprises: one or more processors; A storage device for storing one or more programs, which, when executed by the one or more processors, enables the electronic device to implement a method for processing acoustic emission signals of fiber laminate damage as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor of a computer, the computer is caused to execute a method for processing acoustic emission signals of fiber laminate damage according to any one of claims 1 to 8.

Images