Concrete beam damage classification identification method and system based on acoustic emission signals

Abstract

The application discloses a concrete beam damage classification identification method and system based on an acoustic emission signal, and comprises the following steps: acquiring acoustic emission signal data of concrete beams containing different damage categories; extracting acoustic signal characteristic parameters and constructing a characteristic parameter dataset; constructing a characteristic matrix based on the characteristic parameter dataset and normalizing the characteristic matrix into a normalized matrix; adopting a jackal optimization algorithm, taking the maximum fitness under kernel K-Means clustering as an optimization target, and determining a kernel function of the normalized matrix; performing kernel K-Means clustering on the normalized matrix based on the kernel function, obtaining clustering centers, and determining the damage category corresponding to each clustering center; acquiring an acoustic emission signal to be analyzed, calculating the membership degree of the acoustic emission signal to be analyzed and each clustering center, and determining the damage category of the acoustic emission signal to be analyzed; and the application can accurately identify the load, freeze-thaw and corrosion damage categories of the concrete beam, and provide a basis for structural damage early warning.

Description

Technical Field

[0001] This application relates to the field of concrete beam structure monitoring technology, and in particular to a method and system for classifying and identifying damage in concrete beams based on acoustic emission signals. Background Technology

[0002] Under the influence of various factors such as load, corrosion, and freeze-thaw cycles, concrete beams will develop different types of damage. Corrosion and freeze-thaw damage mainly affect structural durability, while damage caused by load is often structural damage that affects the inherent safety of the structure. Distinguishing between different types of damage is of great significance for scientifically evaluating the working condition and operational risks of reinforced concrete beams.

[0003] Acoustic emission (AE) technology, as a passive non-destructive testing method, possesses unique advantages and is widely used in concrete structure monitoring, enabling effective detection of cracks within the micrometer range. However, existing AE signal identification methods largely rely on direct judgment of key indicators, resulting in strong subjectivity and poor adaptability. Furthermore, AE signals often exhibit complex nonlinear characteristics and non-spherical distributions, making it difficult for traditional clustering methods (such as standard K-Means) based directly on Euclidean distance in the original feature space to accurately distinguish different damage categories. In addition, the choice of kernel function type and its parameters significantly impacts clustering results; relying on manual settings often fails to guarantee optimal performance. Moreover, AE signals often contain mixed damage information such as load, freeze-thaw cycles, and corrosion, making it difficult to distinguish damage categories and affecting the reliability of structural damage early warning. Therefore, accurately distinguishing and identifying different damage categories through AE signals has become a key challenge in current AE signal processing technology. Summary of the Invention

[0004] To address the aforementioned technical issues, this application provides a method and system for classifying and identifying damage to concrete beams based on acoustic emission signals. This method can accurately identify damage categories of concrete beams, such as load, freeze-thaw cycles, and corrosion, providing a reliable basis for early warning of structural damage.

[0005] In a first aspect, this application provides a method for classifying and identifying damage in concrete beams based on acoustic emission signals, including: Acquire acoustic emission signal data of concrete beams containing different damage categories; Determine the acoustic emission events corresponding to each damage category in the acoustic emission signal data, extract the acoustic signal feature parameters of each acoustic emission event, and construct a feature parameter dataset; A feature matrix is ​​constructed based on the feature parameter dataset, and the feature matrix is ​​normalized to form a normalized matrix; The coyote optimization algorithm is used to determine the kernel function of the normalized matrix with the goal of maximizing fitness under kernel K-Means clustering. Based on the determined kernel function, kernel K-Means clustering is performed on the normalized matrix to obtain the cluster centers of the normalized matrix, and the damage category corresponding to each cluster center is determined. Acquire the acoustic emission signal to be analyzed from the target concrete beam, calculate the membership degree between the acoustic emission signal to be analyzed and each cluster center of the normalized matrix, and determine the damage category corresponding to the cluster center with the largest membership degree as the damage category of the acoustic emission signal to be analyzed.

[0006] In one possible implementation of the first aspect, the acoustic signal characteristic parameters include at least one of rise time, ring count, energy, duration, amplitude, average frequency, and peak count.

[0007] In one possible implementation of the first aspect, the step of constructing a feature matrix based on a feature parameter dataset and normalizing the feature matrix to form a normalized matrix includes: Define the feature parameter dataset as , This represents the feature vector of the i-th acoustic emission event. , This represents the characteristic parameter of the j-th acoustic signal in the i-th acoustic emission event. , n is the number of acoustic emission events, and m is the number of acoustic signal characteristic parameters in each acoustic emission event; Based on feature parameter dataset Constructing the feature matrix , Represents an n x m real matrix; The feature matrix is ​​normalized column by column, and each element in the feature matrix is ​​linearly mapped to the interval [0,1] to form a normalized matrix; The expression for the normalization process is as follows: ; For the first The first acoustic emission event One normalized feature parameter; This represents the minimum value of the acoustic signal characteristic parameter in the j-th column of the characteristic matrix; This represents the maximum value of the acoustic signal characteristic parameter in the j-th column of the characteristic matrix.

[0008] In one possible implementation of the first aspect, the kernel function is one of the following: RBF kernel function, polynomial kernel function, and linear kernel function.

[0009] In one possible implementation of the first aspect, the coyote optimization algorithm is used to determine the kernel function of the normalized matrix with the goal of maximizing fitness under kernel K-Means clustering, including: Encode the kernel function type and its kernel parameters into a three-dimensional vector. The three-dimensional vector is used as the gene vector of an individual; where t is the kernel function type encoding. , For kernel parameters; Using the gene vector to be optimized as an individual, the total population size is set to [value missing]. Each group contains For each individual, a random initialization is performed to generate a gene vector for that individual. , Indicates the individual index number; For each individual, the kernel K-Means is used to normalize the matrix. Perform clustering and calculate intra-cluster deviations. : ; For the number of clusters, The index number is the number of clusters, r∈[1,K]; For the first A set of samples from each cluster; For the first The number of samples in each cluster; , They are respectively The first in The, the One normalized feature vector; It is an implicit mapping function. Distance calculation is represented by a kernel function. Based on normalized matrix The minimum and maximum values ​​of each normalized feature parameter are used to generate... Groups of uniformly distributed samples; for each group of uniformly distributed samples, use kernel K-Means clustering with the same kernel function configuration, and the number of clusters is preset to [value missing]. Calculate the first Intra-cluster deviation of uniformly distributed samples : ; For the first After clustering the uniformly distributed samples, the first The sample set of each cluster, b=1,2,…,B; For the first The first uniformly distributed sample in the group The number of samples in each cluster; , They are respectively The first in The, the One normalized feature vector; Calculate the fitness of each individual. : ; The global optimal solution is iteratively updated based on the criterion of maximizing fitness. When the fitness of the global optimal solution does not increase within a preset number of iterations, the iteration is terminated, and the three-dimensional vector corresponding to the global optimal solution is output as the optimal kernel function configuration.

[0010] In one possible implementation of the first aspect, the coyote optimization algorithm iteratively updates the global optimum by maximizing fitness according to the following sub-steps: Within each population, identify the individual with the highest fitness within the population, and denote its gene vector as . Based on the gene vectors of all individuals in the population, the median of each of the three gene dimensions is extracted and combined into a new vector, which serves as the median individual of the population. The gene vector of this median individual is denoted as [insert vector here]. ; For each individual in the population, randomly select two different individuals in the population to generate the gene vector of the new individual. : ; This represents the gene vector of the current individual; , Gene vectors of two randomly selected different individuals; , A random factor is randomly generated in the interval [0,1]. In each population, two individuals are randomly selected and designated as the first parent individual and the second parent individual, respectively, for crossover to generate new individuals; for each gene dimension of the new individual, a random crossover probability is generated, and a preset threshold is used as the boundary to determine whether the gene of the new individual in that dimension comes from the first parent individual or the second parent individual; Calculate the fitness of the new individual and compare it with the individual with the lowest fitness in the group. If the new individual has better fitness, replace the individual with the lowest fitness. With a preset exchange probability, two different groups are randomly selected, and one individual is randomly selected from each group to exchange the positions of the two individuals. This generates a new generation of population. The fitness of this new generation is repeatedly calculated and the global optimum is updated; this continues until the termination condition is met, at which point the three-dimensional vector corresponding to the global optimum is output. .

[0011] In one possible implementation of the first aspect, the step of performing kernel K-Means clustering on the normalized matrix based on the determined kernel function to obtain the cluster centers of the normalized matrix includes: Based on the three-dimensional vector corresponding to the global optimal solution Determine the optimal kernel function configuration, where Used to determine the kernel function type. , Used to determine the corresponding kernel parameters; Using the determined optimal kernel function configuration, the normalized matrix is... Each normalized feature vector in Mapped to a high-dimensional feature space; Kernel K-Means clustering is performed in the high-dimensional feature space, with the number of clusters preset to K, dividing the normalized feature vector into K clusters; the cluster center of each cluster is calculated. : ; For the first Cluster centers; For the first A set of samples from each cluster; For the first The number of samples in each cluster.

[0012] In one possible implementation of the first aspect, determining the damage category corresponding to each cluster center includes: Multiple precast concrete beams were subjected to damage tests according to different damage categories, and test signals corresponding to each damage category were collected during the damage tests. The acoustic emission characteristic parameters of the test signals are extracted, and the mean characteristic vectors of the test signals for each damage category at the original characteristic scale are constructed. ; Cluster centers at normalized scale Restoring to the original feature scale yields the cluster feature mean vector. The cluster feature mean vector Cluster center Representation at the original feature scale; Calculate the characteristic mean vector of the test signal With cluster feature mean vector The Euclidean distance between the cluster centers is used to determine the damage type. Based on the principle of minimum distance, each cluster is matched to the damage type with the minimum Euclidean distance, thus completing the correspondence between the cluster centers and the damage types.

[0013] In one possible implementation of the first aspect, based on the determined kernel function, the membership degree between the acoustic emission signal to be analyzed and each cluster center of the normalized matrix is ​​calculated, including: Calculate the acoustic emission signal to be analyzed and the first Distance between cluster centers : ; Here is the normalized eigenvector of the acoustic emission signal to be analyzed; For the first A set of samples from each cluster; For the first The number of samples in each cluster; , , They are respectively The first in The, the The, the One normalized feature vector; The determined kernel function; Calculate the acoustic emission signal to be analyzed and the first Membership of each cluster center : ; For the number of clusters, The index number represents the number of clusters. The index number is used to traverse the number of clusters, r,k∈[1,K]; To prevent division by zero of constants.

[0014] Secondly, this application provides a system for performing the aforementioned method for classifying and identifying damage in concrete beams based on acoustic emission signals, comprising: The sample signal acquisition module is used to acquire acoustic emission signal data of concrete beams containing different damage categories; The dataset construction module is used to determine the acoustic emission events corresponding to each damage category in the acoustic emission signal data, extract the acoustic signal feature parameters of each acoustic emission event, and construct a feature parameter dataset. The normalization module constructs a feature matrix based on the feature parameter dataset and normalizes the feature matrix to form a normalized matrix. The kernel function optimization module uses the coyote optimization algorithm to determine the kernel function of the normalized matrix with the goal of maximizing the fitness under kernel K-Means clustering. The damage category matching module performs kernel K-Means clustering on the normalized matrix based on the determined kernel function to obtain the cluster centers of the normalized matrix, and matches the corresponding damage category for each cluster center. The damage identification module is used to acquire the acoustic emission signal to be analyzed from the target concrete beam, calculate the membership degree between the acoustic emission signal to be analyzed and each cluster center of the normalized matrix, and determine the damage category corresponding to the cluster center with the largest membership degree as the damage category of the acoustic emission signal to be analyzed.

[0015] Compared with the prior art, this application has the following beneficial effects: 1. This application utilizes unsupervised clustering of acoustic emission signals, eliminating the need for pre-labeled training data and manually preset discrimination rules. It can automatically discover the inherent clustering structure in acoustic emission signals, and complete the automated classification of different damage types such as load, freeze-thaw, and corrosion. This significantly improves the adaptability, timeliness, and objectivity of the classification process. Furthermore, it is unaffected by changes in data distribution under different engineering scenarios and has a wide range of applications. 2. This application introduces a kernel function to map signal features to a high-dimensional feature space, transforming nonlinear and non-spherical data that are difficult to separate in the original space into linearly separable or approximately spherical structures, thereby enabling effective clustering in the high-dimensional space, significantly improving the classification accuracy of complex acoustic emission signals, and solving the problem that traditional clustering methods are difficult to handle nonlinear data; 3. This application employs the Coyote optimization algorithm for global optimization search, with the optimization objective being to maximize the fitness (Gap Statistic) under kernel K-Means clustering, and automatically selects the optimal kernel function type and its parameter combination. This mechanism can adaptively adjust the kernel function configuration according to the characteristics of different datasets, ensuring that the clustering results always maintain optimal performance in various complex environments, thus possessing good stability and generalization ability, while improving the accuracy and reliability of clustering. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating a method for classifying and identifying damage in concrete beams based on acoustic emission signals, provided for an embodiment of this application; Figure 2 A flowchart illustrating an iterative optimization search method for optimal kernel function configuration provided in this application embodiment; Figure 3 A structural block diagram of a concrete beam damage classification and identification system based on acoustic emission signals provided for embodiments of this application; Figure 4 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0018] like Figure 1 As shown, this application provides a method for damage classification and identification of concrete beams based on acoustic emission signals, including: S100: Retrospectively collect or randomly select acoustic emission signal data of concrete beams containing multiple damage categories during actual work or simulation tests; the damage categories include, but are not limited to, load damage, freeze-thaw damage, and corrosion damage.

[0019] S200: Based on pre-calibrated test data (such as precast single-damaged beams) or field monitoring data of known damage types, determine the characteristic parameters of acoustic emission events corresponding to each damage category in the acoustic emission signal data, thereby establishing the correspondence between damage categories and acoustic emission events; extract the acoustic signal characteristic parameters of each acoustic emission event and construct a characteristic parameter dataset; It should be noted that the acoustic emission event refers to an independent physical process in which a concrete beam, under load, freeze-thaw, or corrosion, instantaneously releases elastic strain energy due to local damage and generates transient elastic waves, which is collected by an acoustic emission sensor. Characteristic parameters such as rise time, ring count, and energy can be extracted from the acoustic emission event. In this application, the acoustic emission signal data contains multiple acoustic emission events, each corresponding to a local damage, which belongs to a certain damage category among load, freeze-thaw, or corrosion, thereby establishing a correspondence between acoustic emission events and damage categories. For example, if the damage category is freeze-thaw damage, then one acoustic emission event corresponds to one local damage process caused by freeze-thaw action. Specifically, a pre-calibration test can be used to determine the acoustic emission events corresponding to each damage category in the acoustic emission signal data. For example, by prefabricating concrete beams that only bear load, only withstand freeze-thaw cycles, and only withstand corrosion, acoustic emission signals of each beam can be collected throughout the damage process. Since each beam has only a single damage type, the acoustic emission events collected correspond to that damage category. Alternatively, signal segments can be selected from the field monitoring data of known damage stages to calibrate the acoustic emission events. For example, m acoustic signal feature parameters are extracted from each acoustic emission event. These acoustic signal feature parameters include, but are not limited to, rise time, ring count, energy, duration, amplitude, average frequency, and peak count. The first acoustic emission event The characteristic parameters of the acoustic signal are denoted as follows: , , n represents the number of acoustic emission events, and m represents the number of acoustic signal feature parameters in each acoustic emission event; construct the feature vector of the i-th acoustic emission event. , The feature vectors are aggregated to construct a feature parameter dataset. .

[0020] S300: Construct a feature matrix based on the feature parameter dataset, and normalize the feature matrix to form a normalized matrix.

[0021] S400: The coyote optimization algorithm is used to determine the kernel function of the normalized matrix with the goal of maximizing fitness under kernel K-Means clustering.

[0022] S500: Based on the determined kernel function, perform kernel K-Means clustering on the normalized matrix to obtain the cluster centers of the normalized matrix and determine the damage category corresponding to each cluster center.

[0023] S600: Obtain the acoustic emission signal to be analyzed from the target concrete beam, calculate the membership degree between the acoustic emission signal to be analyzed and each cluster center of the normalized matrix, and determine the damage category corresponding to the cluster center with the largest membership degree as the damage category of the acoustic emission signal to be analyzed.

[0024] In one optional implementation, step S300 includes the following: S310: Define the feature parameter dataset as follows , This represents the feature vector of the i-th acoustic emission event. , This represents the characteristic parameter of the j-th acoustic signal in the i-th acoustic emission event. , n represents the number of acoustic emission events, and m represents the number of acoustic signal characteristic parameters in each acoustic emission event; S320: Based on feature parameter dataset Constructing the feature matrix , Represents an n x m real matrix; S330: Normalize the feature matrix column-wise, linearly mapping each element (i.e., acoustic signal feature parameters) in the feature matrix to the interval [0,1], forming a normalized matrix, denoted as... , For the first The first acoustic emission event The normalized feature parameter will be the first normalized feature parameter. The normalized eigenvectors of the acoustic emission events are represented as follows: ; The expression for the normalization process is as follows: ; in, Let be the element in the i-th row and j-th column of the characteristic matrix, i.e., the element in the i-th row and j-th column. The first acoustic emission event Each acoustic signal characteristic parameter; Let be the element in the i-th row and j-th column of the normalized matrix, i.e., the element in the i-th row and j-th column. The first acoustic emission event One normalized feature parameter; This represents the minimum value of the acoustic signal characteristic parameter in the j-th column of the characteristic matrix; This represents the maximum value of the acoustic signal characteristic parameter in the j-th column of the characteristic matrix.

[0025] In one optional implementation, the kernel function mentioned in step S400 is one of a radial basis function (RBF), a polynomial kernel function, or a linear kernel function; specifically: The RBF kernel function is expressed as follows: ; The polynomial kernel function is expressed as follows: ; The linear kernel function is expressed as follows: ; in, , They represent the first The, the Normalized feature vectors of individual acoustic emission events express and Euclidean distance, express and The inner product; The width parameter of the RBF kernel function. , These are the bias constant and degree parameter of the polynomial kernel function, respectively.

[0026] like Figure 2 As shown, in one optional implementation, step S400 includes the following: S410: Encode the kernel function type and its kernel parameters into a three-dimensional vector. The three-dimensional vector is used as the gene vector of an individual, which contains three parameters to be optimized; where t is the kernel function type encoding. , For kernel parameters; For example, t can take the values ​​1, 2, or 3; when t=1, the kernel function type is RBF kernel function. The width parameter representing the RBF kernel function , Invalid; at t=2, the kernel function type is a polynomial kernel function. The bias constant representing the polynomial kernel function , The degree parameter representing the polynomial kernel function At t=3, the kernel function type is a linear kernel function. and All are invalid; S420: Using the gene vector containing three parameters to be optimized as an individual, set the total population size to be... Each group contains For each individual, a random initialization is performed to generate a gene vector for that individual. For example, the first is defined The individual (i.e., the first individual) The gene vectors (with candidate kernel function configurations) are subscript Indicates the individual index number. During random initialization, In the set Randomly select a value from the middle. In the interval Randomly select a value from the middle. In the interval Randomly select a value from the middle; S430: For each individual, use kernel K-Means to normalize the matrix. Perform clustering and calculate the intra-cluster deviations using the following formula. : ; During clustering, the normalization matrix Each normalized feature vector in For each sample, each sample is assigned to a unique cluster, and these clusters co-cluster into K clusters. Each cluster will subsequently be associated with a damage category. Cluster denoted as r=1,2,…,K; in, For the first The set of samples from the nth cluster, i.e., all those assigned to the nth cluster. The set of normalized eigenvectors of a cluster; For the first The number of samples in the i-th cluster, i.e., the number of samples in the j-th cluster. The number of normalized feature vectors in each cluster; , They belong to the same cluster The two samples in the sample belong to The first in The, the One normalized feature vector; It is an implicit mapping function that satisfies the kernel function property for distance calculation. Replaced by kernel functions; S440: Based on normalized matrix The minimum and maximum values ​​of each normalized feature parameter are used to generate... Using a set of uniformly distributed samples as reference data, the range of values ​​for the normalized feature parameters in each set of uniformly distributed samples is similar to that of the original normalization matrix. The normalized feature parameters have the same range of values; for each group of uniformly distributed samples, kernel K-Means is used for clustering with the same kernel function configuration, and the number of clusters is preset to [value missing]. Calculate the number using the following formula Intra-cluster deviation of uniformly distributed samples : ; in, For the first After clustering the uniformly distributed samples, the first The sample set of each cluster, b=1,2,…,B; For the first The first uniformly distributed sample in the group The number of samples in each cluster; , They belong to the same cluster The first in The, the Normalized feature vectors, distance calculation Replaced by kernel functions; It should be noted that the "same kernel function configuration" refers to the kernel function type and its kernel parameters being the same as those used in step S430 for the normalized matrix. The kernel function used for clustering is configured consistently. S450: Iteratively update the global optimal solution based on maximizing fitness; the global optimal solution is the individual with the highest fitness among all individuals from the first generation to the current generation; after each generation iteration, if the fitness of the current generation's best individual is greater than the fitness of the global optimal solution saved in the previous generation, then update the global optimal solution to the current generation's best individual. Specifically: Before the iteration begins, initialize the global optimal solution to an empty or invalid value; In each iteration, the fitness of each individual is calculated using the following formula. : ; in, Represents the natural logarithm; The individual with the highest fitness in each generation is taken as the best individual in the current generation. The fitness of the current generation's best individual is compared with the fitness of the saved global optimal solution. If the former is greater, the current generation's best individual is updated to the new global optimal solution; otherwise, the original global optimal solution is retained. When the fitness of the global optimal solution does not increase within a preset number of iterations (e.g., 10 consecutive iterations), the iteration is terminated, and the three-dimensional vector corresponding to the global optimal solution is output as the optimal kernel function configuration.

[0027] Furthermore, step S450 updates the global optimum iteratively using the criterion of maximizing fitness, following these sub-steps: S451: In each population, identify the individual with the highest fitness within the population, and denote its gene vector as... Based on the gene vectors of all individuals in the population, the medians of the three gene dimensions (t, p1, p2) are extracted and combined into a new vector, which serves as the median individual of the population. The gene vector of this median individual is denoted as [insert vector here]. ; For each individual in the population, randomly select two different individuals from the population and generate the gene vector of the new individual according to the following formula. : ; in, This represents the gene vector of the current individual; , Gene vectors of two randomly selected different individuals; , A random factor is randomly generated in the interval [0,1]. S452: In each population, two individuals are randomly selected and designated as the first parent individual and the second parent individual, respectively, for crossover to generate new individuals; for each gene dimension of the new individual, a random crossover probability is generated, and a preset threshold is used as the boundary to determine whether the gene of the new individual in that dimension comes from the first parent individual or the second parent individual; For example, the preset threshold can be Genes for each dimension of the new individual are generated according to the following rules: ; in, The value of the u-th dimension of the new individual; The random crossover probability of the u-th gene is generated randomly in the interval [0,1], where u is the index number of the gene dimension, and u=1,2 or3; , These are the u-th gene values ​​of the first and second parent individuals, respectively. like Then the u-th dimension of the new individual's gene is inherited from the u-th dimension of the first parent individual; if If the u-th gene of the new individual is inherited from the u-th gene of the second parent individual, then a complete new individual is generated. Calculate the fitness of the new individual and compare it with the worst individual in the group (i.e., the individual with the lowest fitness). If the new individual has better fitness, replace the worst individual. S453: Randomly select two different groups, and randomly select one individual from each group, then swap the positions of these two individuals; for example, a preset swap probability is used. ; S454: At this point, a new generation of population is generated. The fitness of this new generation is repeatedly calculated and the global optimum is updated; until the termination condition is met, the three-dimensional vector corresponding to the global optimum is output. .

[0028] In an optional implementation, step S500, which involves performing kernel K-Means clustering on the normalized matrix based on the determined kernel function to obtain the cluster centers of the normalized matrix, specifically includes the following: S510: Based on the three-dimensional vector corresponding to the global optimal solution Determine the optimal kernel function configuration, where Used to determine the kernel function type. , Used to determine the kernel parameters corresponding to the kernel function type; S520: Using the determined optimal kernel function configuration, each normalized eigenvector in the normalized matrix Y is... Mapped to a high-dimensional feature space; S530: Perform kernel K-Means clustering in this high-dimensional feature space, with the number of clusters preset to K, and normalize the feature vectors. The data is divided into K clusters, and the cluster centers of each cluster are calculated using the following formula. : ; in, For the first There are 1 cluster center, and each cluster center is used to characterize the typical features of the sample under the corresponding damage state, that is, the feature mean vector of all samples under the damage state. For the first A set of samples from each cluster; For the first The number of samples in each cluster.

[0029] In an optional implementation, step S500 clearly identifies the damage category corresponding to each cluster center through pre-testing and the minimum distance principle, thereby determining the damage category corresponding to each cluster center, specifically including the following: S540: Precast multiple test concrete beams and conduct damage tests according to different damage categories such as load, freeze-thaw, and corrosion. Collect the acoustic emission signals of the test concrete beams during the damage test process as the test signals corresponding to each damage category. S550: Extract the acoustic emission characteristic parameters of the test signal and construct the mean feature vector of the test signal for each damage category at the original feature scale. ; S560: Cluster centers at normalized scale Restoring to the original feature scale yields the cluster feature mean vector. The cluster feature mean vector Cluster center Representation at the original feature scale; S570: Calculate the characteristic mean vector of the test signal. With cluster feature mean vector The Euclidean distance between them; the formula is as follows: ; in, and These are the characteristic mean vectors of the test signal. and cluster feature mean vector The One portion, The number of characteristic parameters of the acoustic signal, i.e., the mean vector of the experimental signal features. and cluster feature mean vector The dimension; Based on the principle of minimum distance, each cluster is matched to the damage type with the minimum Euclidean distance, thus completing the correspondence between cluster centers and damage categories.

[0030] In an optional implementation, step S600, calculating the membership degree between the acoustic emission signal to be analyzed and each cluster center of the normalized matrix, includes: S610: The target concrete beam is the object to be analyzed that needs to be damaged. Acoustic emission technology is used to obtain the acoustic emission signal of the target concrete beam to be analyzed. S620: Calculate the acoustic emission signal to be analyzed and the... Distance between cluster centers : ; in, Here is the normalized eigenvector of the acoustic emission signal to be analyzed; For the first A set of samples from each cluster; For the first The number of samples in each cluster; , , They belong to the same cluster Any three samples in the list belong to The first in The, the The, the One normalized feature vector; The determined kernel function; S630: Calculate the acoustic emission signal to be analyzed and the... Membership of each cluster center : ; in, For the number of clusters, The index number represents the number of clusters, and in this formula, it corresponds to the index number used to calculate the current membership degree. The index number is used to traverse the number of clusters, r,k∈[1,K]; To prevent division by zero of constants; Membership degree is used to represent the degree to which the acoustic emission signal to be analyzed belongs to different cluster centers (i.e., different damage categories). In this application, the distance between the signal to be analyzed and each cluster center in the high-dimensional feature space is first calculated by a kernel function, and then the distance is converted into membership degree: the smaller the distance, the higher the membership degree; the higher the membership degree, the stronger the similarity between the acoustic emission signal to be analyzed and the corresponding cluster center, and thus the more likely it is to belong to the damage category represented by the cluster; the sum of the membership degrees of all damage categories is normalized to 1.

[0031] Based on the same inventive concept, this application also provides a system for implementing the above-mentioned method for classifying and identifying damage in concrete beams based on acoustic emission signals. The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations in one or more system embodiments provided below can be found in the limitations of the concrete beam damage classification and identification method described above, and will not be repeated here.

[0032] In one exemplary embodiment, such as Figure 3 As shown, a damage classification and identification system for concrete beams based on acoustic emission signals is provided, including: a sample signal acquisition module, a dataset construction module, a normalization module, a kernel function optimization module, a damage category matching module, and a damage identification module, wherein: The sample signal acquisition module is used to acquire acoustic emission signal data of concrete beams containing different damage categories; The dataset construction module is used to determine the acoustic emission events corresponding to each damage category in the acoustic emission signal data, extract the acoustic signal feature parameters of each acoustic emission event, and construct a feature parameter dataset. The normalization module constructs a feature matrix based on the feature parameter dataset and normalizes the feature matrix to form a normalized matrix. The kernel function optimization module uses the coyote optimization algorithm to determine the kernel function of the normalized matrix with the goal of maximizing the fitness under kernel K-Means clustering. The damage category matching module performs kernel K-Means clustering on the normalized matrix based on the determined kernel function to obtain the cluster centers of the normalized matrix, and matches the corresponding damage category for each cluster center. The damage identification module is used to acquire the acoustic emission signal to be analyzed from the target concrete beam, calculate the membership degree between the acoustic emission signal to be analyzed and each cluster center of the normalized matrix, and determine the damage category corresponding to the cluster center with the largest membership degree as the damage category of the acoustic emission signal to be analyzed.

[0033] The modules in the aforementioned concrete beam damage classification and identification system based on acoustic emission signals can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0034] In one exemplary embodiment, this application also provides a computer device, which may be a server, and its internal structure diagram may be as follows: Figure 4 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores acoustic emission signal data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a method for classifying and identifying damage in concrete beams.

[0035] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0036] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. The databases involved in the embodiments provided in this application can include at least one of relational and non-relational databases. Non-relational databases can include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application can be general-purpose processors, central processing units, graphics processors, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0037] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0038] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A method for classifying and identifying damage in concrete beams based on acoustic emission signals, characterized in that, include: Acquire acoustic emission signal data of concrete beams containing different damage categories; Determine the acoustic emission events corresponding to each damage category in the acoustic emission signal data, extract the acoustic signal feature parameters of each acoustic emission event, and construct a feature parameter dataset; A feature matrix is ​​constructed based on the feature parameter dataset, and the feature matrix is ​​normalized to form a normalized matrix; The coyote optimization algorithm is used, with the goal of maximizing fitness under kernel K-Means clustering, to determine the kernel function of the normalized matrix; specifically, it includes: Encode the kernel function type and its kernel parameters into a three-dimensional vector. The three-dimensional vector is used as the gene vector of an individual; where t is the kernel function type encoding. , For kernel parameters; Using the gene vector to be optimized as an individual, the total population size is set to [value missing]. Each group contains For each individual, a random initialization is performed to generate a gene vector for that individual. , Indicates the individual index number; For each individual, the kernel K-Means is used to normalize the matrix. Perform clustering and calculate intra-cluster deviations. : ； For the number of clusters, The index number is the number of clusters, r∈[1,K]; For the first A set of samples from each cluster; For the first The number of samples in each cluster; , They are respectively The first in The, the One normalized feature vector; It is an implicit mapping function. Distance calculation is represented by a kernel function. Based on normalized matrix The minimum and maximum values ​​of each normalized feature parameter are used to generate... Groups of uniformly distributed samples; for each group of uniformly distributed samples, K-Means clustering is performed using the same kernel function configuration, with the number of clusters preset to [value missing]. Calculate the first Intra-cluster deviation of uniformly distributed samples : ； For the first After clustering the uniformly distributed samples, the first The sample set of each cluster, b=1,2,…,B; For the first The first uniformly distributed sample in the group The number of samples in each cluster; , They are respectively The first in The, the One normalized feature vector; Calculate the fitness of each individual. : ； The global optimal solution is iteratively updated based on the criterion of maximizing fitness. When the fitness of the global optimal solution does not increase within a preset number of iterations, the iteration is terminated, and the three-dimensional vector corresponding to the global optimal solution is output as the optimal kernel function configuration. Based on the determined kernel function, kernel K-Means clustering is performed on the normalized matrix to obtain the cluster centers of the normalized matrix and determine the damage category corresponding to each cluster center. Acquire the acoustic emission signal to be analyzed from the target concrete beam, calculate the membership degree between the acoustic emission signal to be analyzed and each cluster center of the normalized matrix, and determine the damage category corresponding to the cluster center with the largest membership degree as the damage category of the acoustic emission signal to be analyzed.

2. The method for classifying and identifying damage in concrete beams based on acoustic emission signals according to claim 1, characterized in that, The acoustic signal characteristic parameters include at least one of rise time, ring count, energy, duration, amplitude, average frequency, and peak count.

3. The method for classifying and identifying damage in concrete beams based on acoustic emission signals according to claim 1, characterized in that, The process of constructing a feature matrix based on a feature parameter dataset and normalizing the feature matrix to form a normalized matrix includes: Define the feature parameter dataset as , This represents the feature vector of the i-th acoustic emission event. , This represents the characteristic parameter of the j-th acoustic signal in the i-th acoustic emission event. , n is the number of acoustic emission events, and m is the number of acoustic signal characteristic parameters in each acoustic emission event; Based on feature parameter dataset Constructing the feature matrix , Represents an n x m real matrix; The feature matrix is ​​normalized column by column, and each element in the feature matrix is ​​linearly mapped to the interval [0,1] to form a normalized matrix; The expression for the normalization process is as follows: ； For the first The first acoustic emission event One normalized feature parameter; This represents the minimum value of the acoustic signal characteristic parameter in the j-th column of the characteristic matrix; This represents the maximum value of the acoustic signal characteristic parameter in the j-th column of the characteristic matrix.

4. The method for classifying and identifying damage in concrete beams based on acoustic emission signals according to claim 1, characterized in that, The kernel function is one of the following: RBF kernel function, polynomial kernel function, or linear kernel function.

5. The method for classifying and identifying damage in concrete beams based on acoustic emission signals according to claim 1, characterized in that, The coyote optimization algorithm updates the global optimal solution iteratively according to the following sub-steps, based on the criterion of maximizing fitness: Within each population, identify the individual with the highest fitness within the population, and denote its gene vector as . Based on the gene vectors of all individuals in the population, the median of each of the three gene dimensions is extracted and combined into a new vector, which serves as the median individual of the population. The gene vector of this median individual is denoted as [insert vector here]. ; For each individual in the population, randomly select two different individuals in the population to generate the gene vector of the new individual. : ； This represents the gene vector of the current individual; , Gene vectors of two randomly selected different individuals; , A random factor is randomly generated in the interval [0,1]. In each population, two individuals are randomly selected and designated as the first parent individual and the second parent individual, respectively, for crossover to generate new individuals; for each gene dimension of the new individual, a random crossover probability is generated, and a preset threshold is used as the boundary to determine whether the gene of the new individual in that dimension comes from the first parent individual or the second parent individual; Calculate the fitness of the new individual and compare it with the individual with the lowest fitness in the group. If the new individual has better fitness, replace the individual with the lowest fitness. With a preset exchange probability, two different groups are randomly selected, and one individual is randomly selected from each group to exchange the positions of the two individuals. This generates a new generation of population. The fitness of this new generation is repeatedly calculated and the global optimum is updated; this continues until the termination condition is met, at which point the three-dimensional vector corresponding to the global optimum is output. .

6. The method for classifying and identifying damage in concrete beams based on acoustic emission signals according to claim 5, characterized in that, Based on the determined kernel function, kernel K-Means clustering is performed on the normalized matrix to obtain the cluster centers of the normalized matrix, including: Based on the three-dimensional vector corresponding to the global optimal solution Determine the optimal kernel function configuration, where Used to determine the kernel function type. , Used to determine the corresponding kernel parameters; Using the determined optimal kernel function configuration, the normalized matrix is... Each normalized feature vector in Mapped to a high-dimensional feature space; Kernel K-Means clustering is performed in the high-dimensional feature space, with the number of clusters preset to K, dividing the normalized feature vector into K clusters; the cluster center of each cluster is calculated. : ； For the first Cluster centers; For the first A set of samples from each cluster; For the first The number of samples in each cluster.

7. The method for classifying and identifying damage in concrete beams based on acoustic emission signals according to claim 6, characterized in that, Determining the damage category corresponding to each cluster center includes: Multiple precast concrete beams were subjected to damage tests according to different damage categories, and test signals corresponding to each damage category were collected during the damage tests. The acoustic emission characteristic parameters of the test signals are extracted, and the mean characteristic vectors of the test signals for each damage category at the original characteristic scale are constructed. ; Cluster centers at normalized scale Restoring to the original feature scale yields the cluster feature mean vector. The cluster feature mean vector Cluster center Representation at the original feature scale; Calculate the characteristic mean vector of the test signal With cluster feature mean vector The Euclidean distance between the cluster centers is used to determine the damage type. Based on the principle of minimum distance, each cluster is matched to the damage type with the minimum Euclidean distance, thus completing the correspondence between the cluster centers and the damage types.

8. The method for classifying and identifying damage in concrete beams based on acoustic emission signals according to claim 1, characterized in that, Based on the determined kernel function, the membership degree between the acoustic emission signal to be analyzed and each cluster center of the normalized matrix is ​​calculated, including: Calculate the acoustic emission signal to be analyzed and the first Distance between cluster centers : ； Here is the normalized eigenvector of the acoustic emission signal to be analyzed; For the first A set of samples from each cluster; For the first The number of samples in each cluster; , , They are respectively The first in The, the The, the One normalized feature vector; The determined kernel function; Calculate the acoustic emission signal to be analyzed and the first Membership of each cluster center : ； For the number of clusters, The index number represents the number of clusters. The index number is used to traverse the number of clusters, r,k∈[1,K]; To prevent division by zero of constants.

9. A system for performing the concrete beam damage classification and identification method based on acoustic emission signals as described in any one of claims 1-8, characterized in that, include: The sample signal acquisition module is used to acquire acoustic emission signal data of concrete beams containing different damage categories; The dataset construction module is used to determine the acoustic emission events corresponding to each damage category in the acoustic emission signal data, extract the acoustic signal feature parameters of each acoustic emission event, and construct a feature parameter dataset. The normalization module constructs a feature matrix based on the feature parameter dataset and normalizes the feature matrix to form a normalized matrix. The kernel function optimization module uses the coyote optimization algorithm to determine the kernel function of the normalized matrix with the goal of maximizing the fitness under kernel K-Means clustering. The damage category matching module performs kernel K-Means clustering on the normalized matrix based on the determined kernel function to obtain the cluster centers of the normalized matrix, and matches the corresponding damage category for each cluster center. The damage identification module is used to acquire the acoustic emission signal to be analyzed from the target concrete beam, calculate the membership degree between the acoustic emission signal to be analyzed and each cluster center of the normalized matrix, and determine the damage category corresponding to the cluster center with the largest membership degree as the damage category of the acoustic emission signal to be analyzed.

Images