Composite laminated plate damage identification method, device, equipment, storage medium and computer program product

Abstract

This application relates to the field of material structure health monitoring technology, and in particular to a method, device, equipment, storage medium, and computer program product for damage identification of composite laminates. The method involves acquiring structural description information of the laminate to be tested, constructing an analysis model of the laminate, and determining a parameter space; solving for multiple natural frequencies for different combinations of damage parameters within the parameter space, and establishing a vibration feature sample library that pairs natural frequency features with damage parameters; employing a hybrid intelligent modeling strategy based on the sample library, utilizing swarm intelligence to optimize neural network parameters, and training candidate identification models with multiple natural frequencies as input and damage parameters as output; selecting a target identification model based on preset engineering requirements constraints, collecting vibration signals, and extracting multiple natural frequencies consistent with the sample library to form the vibration features to be tested; after random noise processing and frequency offset correction of the vibration features to be tested, inputting them into the target identification model, and outputting the damage parameter identification results.

Description

Technical Field

[0001] This application relates to the field of material structure health monitoring technology, and in particular to a method, device, equipment, storage medium and computer program product for identifying damage to composite laminates. Background Technology

[0002] Composite laminates are widely used in structural components such as aerospace and wind turbine blades due to their lightweight and high strength. However, during service, they are prone to internal damage such as delamination at the interlaminar interfaces. This type of damage is hidden and can cause changes in the structural dynamics. Therefore, the industry often adopts a vibration response-based identification approach: by collecting vibration signals, extracting modal features such as multi-order natural frequencies, and establishing a correspondence between "damage parameters and vibration features" to determine the damage state. However, existing technologies still have shortcomings in engineering implementation: on the one hand, it is difficult to efficiently generate sufficiently comprehensive multi-order natural frequency samples within a reasonable damage parameter space and form a vibration feature sample library that can be used for learning; on the other hand, neural network-based mapping modeling is sensitive to parameter initialization and training optimization, lacks a stable modeling mechanism that combines with swarm intelligence optimization, and lacks a unified logic for selecting the target model from candidate models under different engineering requirements. Furthermore, the natural frequency features extracted on-site are easily affected by random noise and frequency shifts, and there is a lack of noise processing and frequency shift correction procedures consistent with the sample library, resulting in poor accuracy of the output composite laminate damage identification results. Therefore, improving the accuracy of damage identification results for composite laminates has become an urgent technical problem to be solved. Summary of the Invention

[0003] The main objective of this application is to provide a method, apparatus, device, storage medium, and computer program product for identifying damage in composite laminates, aiming to solve the technical problem of how to improve the accuracy of damage identification results in composite laminates.

[0004] To achieve the above objectives, this application provides a method for identifying damage in composite laminates, the method comprising the following steps: Obtain the structural description information of the laminate to be tested, and construct a laminate analysis model including damage parameter characterization based on the structural description information to determine the parameter space used to characterize the damage state; Based on the parameter space, the multi-order natural frequencies corresponding to different combinations of damage parameters are solved to obtain a vibration feature sample library. The vibration feature sample library establishes a sample correspondence between multi-order natural frequency features and the damage parameters. Based on the vibration feature sample library, a damage parameter prediction model is constructed using a preset hybrid intelligent modeling strategy. The hybrid intelligent modeling strategy includes optimizing the parameters of the neural network model using a swarm intelligence optimization strategy, and training a candidate recognition model with the multi-order natural frequency features as input and the damage parameters as output. Based on the preset engineering requirements constraints, the target recognition model is determined from the candidate recognition models, and the vibration signal of the laminate to be detected is collected to extract the multi-order natural frequency features consistent with the vibration feature sample library, so as to obtain the vibration features to be tested. The vibration characteristics to be tested are subjected to random noise processing and frequency offset correction to obtain corrected vibration characteristics to be tested. The corrected vibration characteristics to be tested are then input into the target recognition model to obtain the damage parameter recognition result of the laminate to be tested.

[0005] In one embodiment, the step of obtaining structural description information of the laminate to be tested and constructing a laminate analysis model including damage parameter characterization based on the structural description information to determine the parameter space used to characterize the damage state includes: The layup information, geometric information, material parameter information, and boundary constraint information of the laminate to be tested are obtained, and the layup information, geometric information, material parameter information, and boundary constraint information are aggregated to obtain the structural description information; Based on the structural description information, a partitioned structural model including interlayer delamination is established, the laminate to be tested is divided into multiple sub-layer regions, and damage parameters for characterizing delamination length and delamination location are defined in the partitioned structural model so that the damage parameters are associated with the sub-layer regions. Based on the partitioned structure model, the displacement field equations of each sub-layer region are established using classical laminated plate theory, and the displacement continuity constraints of the delamination region are set in combination with the damage parameters to obtain the laminated plate analysis model, thereby determining the parameter space used to characterize the damage state.

[0006] In one embodiment, the step of solving for the multiple natural frequencies corresponding to different combinations of damage parameters based on the parameter space to obtain a vibration feature sample library includes: Based on the parameter space, multiple damage parameter combinations are generated, and a partition structure description corresponding to each damage parameter combination is determined. Based on the description of each partition structure, a displacement function that satisfies the boundary constraints and the displacement continuity requirement of the delamination region is constructed using Chebyshev polynomials, and the system stiffness matrix and mass matrix corresponding to the displacement function are established based on the Rayleigh-Ritz energy analysis algorithm. The vibration equation is solved based on the system stiffness matrix and the mass matrix to obtain the multi-order natural frequencies corresponding to each combination of damage parameters. A sample association relationship is established between each multi-order natural frequency and the corresponding combination of damage parameters to obtain a vibration feature sample library.

[0007] In one embodiment, the step of constructing a damage parameter prediction model based on the vibration feature sample library using a preset hybrid intelligent modeling strategy, wherein the hybrid intelligent modeling strategy includes optimizing the parameters of the neural network model using a swarm intelligence optimization strategy, and training a candidate recognition model using the multi-order natural frequency features as input and the damage parameters as output, includes: Based on the vibration feature sample library, a training sample set is constructed, such that each sample in the training sample set includes multi-order natural frequency features corresponding to the same damage parameter as input samples, and the damage parameter as supervision labels. The training sample set is then subjected to feature unification processing to obtain standardized training data. Based on the standardized training data, the network structure and parameter set of the neural network model are set, and a swarm intelligence optimization strategy is used to generate a candidate parameter set. Based on each candidate parameter in the candidate parameter set, the neural network model is driven to make predictions on the standardized training data and determine the prediction error. The prediction error is used as the fitness and the candidate parameter set is iteratively updated to obtain the target parameters. The target parameters are loaded into the neural network model, and the loaded neural network model is trained based on the standardized training data, so that the neural network model forms a mapping relationship from the multi-order intrinsic frequency features to the damage parameters, and a candidate recognition model is obtained through training.

[0008] In one embodiment, the step of determining the target recognition model from the candidate recognition models based on preset engineering requirement constraints, and collecting the vibration signal of the laminate to be tested to extract multi-order natural frequency features consistent with the vibration feature sample library to obtain the vibration features to be tested includes: The preset engineering requirement constraints are obtained and converted into evaluation criteria for recognition accuracy requirements and / or response latency requirements. Based on the evaluation criteria, the output error and computational overhead of the candidate recognition models on preset verification data are evaluated. Based on the evaluation results, the target recognition model that meets the preset engineering requirement constraints is determined from the candidate recognition models. Based on the target recognition model, the vibration acquisition configuration and feature extraction configuration corresponding to the vibration feature sample library are determined, and the excitation and response acquisition of the laminate to be tested are performed based on the vibration acquisition configuration to obtain the vibration signal characterizing the dynamic response of the laminate to be tested. Based on the feature extraction configuration, frequency domain analysis and modal parameter extraction are performed on the vibration signal to obtain multi-order natural frequency features consistent with the vibration feature sample library, and the multi-order natural frequency features are collected to obtain the vibration feature to be measured.

[0009] In one embodiment, the step of performing random noise processing and frequency offset correction on the vibration feature to be tested to obtain a corrected vibration feature to be tested, and inputting the corrected vibration feature to be tested into the target recognition model to obtain the damage parameter recognition result of the laminate to be tested, includes: The vibration characteristics to be tested are obtained, and random noise superposition processing is performed on the vibration characteristics to be tested according to the preset noise distribution and preset noise amplitude rules to obtain the noisy vibration characteristics to be tested. Obtain the reference natural frequency corresponding to each natural frequency feature in the noisy vibration feature to be tested, and determine the frequency offset of each order based on the reference natural frequency and the noisy vibration feature to be tested. Based on the frequency offset, perform frequency offset correction on the noisy vibration feature to be tested to obtain the corrected vibration feature to be tested. The modified vibration characteristics to be tested are input into the target recognition model to obtain the damage parameter recognition results of the laminate to be tested.

[0010] Furthermore, to achieve the above objectives, this application also proposes a composite laminate damage identification device, the composite laminate damage identification device comprising: The parameter space module is used to acquire the structural description information of the laminate to be tested, and to construct a laminate analysis model including damage parameter characterization based on the structural description information, so as to determine the parameter space used to characterize the damage state. The vibration feature module is used to solve for the multi-order natural frequencies corresponding to different combinations of damage parameters based on the parameter space, and obtain a vibration feature sample library. The vibration feature sample library establishes a sample correspondence between multi-order natural frequency features and the damage parameters. The candidate model module is used to construct a damage parameter prediction model based on the vibration feature sample library and using a preset hybrid intelligent modeling strategy. The hybrid intelligent modeling strategy includes using a swarm intelligence optimization strategy to optimize the parameters of the neural network model and training the candidate recognition model with the multi-order natural frequency features as input and the damage parameters as output. The constraint feature module is used to determine the target recognition model from the candidate recognition models based on preset engineering requirement constraints, and to collect the vibration signal of the laminate to be tested to extract multi-order natural frequency features consistent with the vibration feature sample library to obtain the vibration features to be tested. The target module is used to perform random noise processing and frequency offset correction on the vibration characteristics to be tested to obtain corrected vibration characteristics to be tested, and input the corrected vibration characteristics to be tested into the target recognition model to obtain the damage parameter recognition result of the laminate to be tested.

[0011] Furthermore, to achieve the above objectives, this application also proposes a composite laminate damage identification device, the device comprising: a memory, a processor, and a composite laminate damage identification program stored in the memory and executable on the processor, the composite laminate damage identification program being configured to implement the steps of the composite laminate damage identification method as described in any of the above embodiments.

[0012] In addition, to achieve the above objectives, this application also proposes a storage medium storing a composite laminate damage identification program, which, when executed by a processor, implements the steps of the composite laminate damage identification method as described above.

[0013] In addition, to achieve the above objectives, this application also proposes a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the composite laminate damage identification method described above.

[0014] This application obtains the structural description information of the laminate to be tested and constructs an analysis model of the laminate including damage parameter characterization based on the structural description information to determine the parameter space used to characterize the damage state. Based on the parameter space, the multi-order natural frequencies corresponding to different combinations of damage parameters are solved to obtain a vibration feature sample library. The vibration feature sample library establishes a sample correspondence between multi-order natural frequency features and damage parameters. Based on the vibration feature sample library, a damage parameter prediction model is constructed using a preset hybrid intelligent modeling strategy. The hybrid intelligent modeling strategy includes using a swarm intelligence optimization strategy to optimize the parameters of the neural network model and training a candidate recognition model with multi-order natural frequency features as input and damage parameters as output. Based on preset engineering requirement constraints, a target recognition model is determined from the candidate recognition models, and the vibration signal of the laminate to be tested is collected to extract multi-order natural frequency features consistent with the vibration feature sample library to obtain the vibration features to be tested. The vibration features to be tested are subjected to random noise processing and frequency offset correction to obtain corrected vibration features to be tested. The corrected vibration features to be tested are then input into the target recognition model to obtain the damage parameter recognition result of the laminate to be tested. This application first constructs an analytical model containing damage parameter representation based on structural description information and determines the parameter space. Within this parameter space, it solves for the multi-order natural frequencies corresponding to different combinations of damage parameters, forming a vibration feature sample library that establishes a correspondence between "multi-order natural frequency features and damage parameters," thus solidifying the supervised mapping required for identification from the source. Then, based on the sample library, it uses a hybrid intelligent modeling strategy that includes swarm intelligence optimization to optimize the parameters of the neural network model and train it to obtain candidate identification models, making the model parameters more consistent with the mapping relationship defined in the sample library to reduce fitting errors. Finally, in the inference stage, it selects the target identification model from the candidate models according to engineering requirements constraints, and extracts the multi-order natural frequency features consistent with the sample library as the vibration features to be measured. After random noise processing and frequency offset correction, these features are then input into the target identification model, thereby reducing feature mismatch and system bias caused by noise and offset, and thus improving the accuracy of the output damage parameter identification results. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the first embodiment of the composite material laminate damage identification method of this application; Figure 2 This is a schematic diagram of a sub-process in the second embodiment of the composite material laminate damage identification method of this application; Figure 3 This is a schematic diagram of a sub-process in the third embodiment of the composite material laminate damage identification method of this application; Figure 4 This is a schematic diagram of the module structure of the composite laminate damage identification device according to an embodiment of this application; Figure 5This is a schematic diagram of the equipment structure of the hardware operating environment involved in the composite laminate damage identification method in this application embodiment.

[0016] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of this application.

[0018] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0019] It should be noted that composite laminates are widely used in structural components such as aerospace and wind turbine blades due to their lightweight and high strength. However, during service, they are prone to internal damage such as delamination at the interlayer interfaces. This type of damage is hidden and can cause changes in the structural dynamics. Therefore, the industry often adopts a vibration response-based identification approach: by collecting vibration signals, extracting modal features such as multi-order natural frequencies, and establishing a correspondence between "damage parameters and vibration features" to determine the damage state. However, existing technologies still have shortcomings in engineering implementation: on the one hand, it is difficult to efficiently generate sufficiently comprehensive multi-order natural frequency samples within a reasonable damage parameter space and form a vibration feature sample library that can be used for learning; on the other hand, neural network-based mapping modeling is sensitive to parameter initialization and training optimization, lacks a stable modeling mechanism that combines with swarm intelligence optimization, and lacks a unified logic for selecting the target model from candidate models under different engineering requirements. Furthermore, the natural frequency features extracted on-site are easily affected by random noise and frequency shifts, and there is a lack of noise processing and frequency shift correction procedures consistent with the sample library, resulting in poor accuracy of the output composite laminate damage identification results. Therefore, improving the accuracy of damage identification results for composite laminates has become an urgent technical problem to be solved.

[0020] The main solution of this application is as follows: First, obtain the structural description information of the laminate to be tested, and construct a laminate analysis model including damage parameter characterization based on the structural description information to determine the parameter space used to characterize the damage state. Second, based on the parameter space, solve for the multi-order natural frequencies corresponding to different combinations of damage parameters to obtain a vibration feature sample library. This library establishes a sample correspondence between multi-order natural frequency features and damage parameters. Third, based on the vibration feature sample library, construct a damage parameter prediction model using a pre-defined hybrid intelligent modeling strategy. This strategy includes optimizing the parameters of the neural network model using a swarm intelligence optimization strategy, and training a candidate recognition model using multi-order natural frequency features as input and damage parameters as output. Fourth, based on pre-defined engineering requirement constraints, determine the target recognition model from the candidate recognition models, and collect the vibration signal of the laminate to be tested to extract multi-order natural frequency features consistent with the vibration feature sample library, obtaining the vibration features to be tested. Fifth, process the vibration features to be tested with random noise and perform frequency offset correction to obtain corrected vibration features to be tested. Finally, input the corrected vibration features to be tested into the target recognition model to obtain the damage parameter recognition result of the laminate to be tested.

[0021] This application first constructs an analytical model containing damage parameter representation based on structural description information and determines the parameter space. Within this parameter space, it solves for the multi-order natural frequencies corresponding to different combinations of damage parameters, forming a vibration feature sample library that establishes a correspondence between "multi-order natural frequency features and damage parameters," thus solidifying the supervised mapping required for identification from the source. Then, based on the sample library, it uses a hybrid intelligent modeling strategy that includes swarm intelligence optimization to optimize the parameters of the neural network model and train it to obtain candidate identification models, making the model parameters more consistent with the mapping relationship defined in the sample library to reduce fitting errors. Finally, in the inference stage, it selects the target identification model from the candidate models according to engineering requirements constraints, and extracts the multi-order natural frequency features consistent with the sample library as the vibration features to be measured. After random noise processing and frequency offset correction, these features are then input into the target identification model, thereby reducing feature mismatch and system bias caused by noise and offset, and thus improving the accuracy of the output damage parameter identification results.

[0022] It should be noted that the execution subject of the method in this embodiment can be a computing service device with data processing, network communication, and program execution functions, or it can be the aforementioned composite laminate damage identification device with the same or similar functions. This embodiment and the following embodiments will be described using a composite laminate damage identification device as an example.

[0023] Based on this, a first embodiment of the composite laminate damage identification method of this application is proposed. Please refer to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the composite material laminate damage identification method of this application.

[0024] In this embodiment, the method for identifying damage in composite laminates includes the following steps: S1: Obtain the structural description information of the laminate to be tested, and construct a laminate analysis model including damage parameter characterization based on the structural description information, so as to determine the parameter space used to characterize the damage state; S2: Based on the parameter space, the multi-order natural frequencies corresponding to different combinations of damage parameters are solved to obtain a vibration feature sample library. The vibration feature sample library establishes a sample correspondence between multi-order natural frequency features and the damage parameters. It should be noted that the laminate to be tested is a composite laminate object requiring damage state identification. Structural description information is a set of fundamental information used to characterize the laminate's structure and material. Damage parameters are a set of parameters used to characterize the laminate's damage state. The laminate analysis model is a structural analysis model built upon the structural description information, capable of solving for dynamic characteristics (such as natural frequencies) under given damage parameter conditions. The parameter space is a set consisting of the range of values ​​for damage parameters and their combinations. Multiple natural frequencies are the different orders of natural frequency characteristics obtained from the analysis model. Ritz-type semi-analytical solving algorithms are a class of semi-analytical methods used to solve for natural frequencies. The vibration feature sample library is a set of multiple natural frequencies obtained by solving for different combinations of damage parameters within the parameter space. The sample correspondence is the pairing relationship of "multi-order natural frequency characteristic damage parameters" in the sample library.

[0025] Specifically, the structural description information of the laminate to be tested is obtained. This structural description information is used to characterize at least the laminate / ply configuration, geometric dimensions, material parameters, and boundary constraints of the laminate, and is organized into a set of structural parameters that can be used for computational modeling. Subsequently, an analytical model of the laminate is established based on the structural description information. Damage parameters are introduced into this analytical model as a dimension representing the damage state, enabling the damage parameters to participate in the subsequent dynamic solution process as input conditions of the model. This provides the analytical model with the basis for solving under different damage state conditions.

[0026] Furthermore, after introducing the damage parameters into the analysis model, a parameter space for characterizing the damage state is determined based on the analysis model, and multiple different combinations of damage parameters are generated within this parameter space. For each combination of damage parameters, it is loaded as a calculation condition for the analysis model. Then, a Ritz-type semi-analytical solution algorithm is used to solve for the multi-order natural frequencies under the corresponding damage parameter combinations. This includes constructing a displacement function that satisfies the boundary constraints using the Chebyshev-Ritz method and establishing the vibration solution equation using the Rayleigh-Ritz method to obtain the multi-order natural frequencies. Finally, each combination of damage parameters and its corresponding solved multi-order natural frequencies are written into a vibration feature sample library, thereby forming a sample correspondence between multi-order natural frequency features and damage parameters in the sample library.

[0027] By first establishing a laminate analysis model containing damage parameter representations based on structural description information and determining the parameter space, and then solving for the corresponding multi-order natural frequencies for different combinations of damage parameters within this parameter space and constructing a sample library, the correspondence between "damage parameters and multi-order natural frequency features" is systematically established and solidified during the sample construction stage. On the one hand, the parameter space ensures that the samples cover the range of combinations of target damage states; on the other hand, the Ritz-class semi-analytical solution provides matching multi-order natural frequency features for each set of damage parameters, thus forming consistent feature-label pairing data. When conducting subsequent identification modeling based on this sample correspondence, the model can learn around a stable "input (multi-order natural frequency features) - output (damage parameters)" mapping, thereby reducing identification bias caused by missing sample pairings or unclear correlation between features and damage states, and providing a data and mapping foundation for improving the accuracy of damage identification results.

[0028] S3: Based on the vibration feature sample library, a damage parameter prediction model is constructed using a preset hybrid intelligent modeling strategy. The hybrid intelligent modeling strategy includes using a swarm intelligence optimization strategy to optimize the parameters of the neural network model, and using the multi-order natural frequency features as input and the damage parameters as output to train and obtain a candidate recognition model. It should be noted that the vibration feature sample library is a set of samples obtained within the damage parameter space. The preset hybrid intelligent modeling strategy is a pre-defined set of modeling schemes used to construct a damage parameter prediction model. The swarm intelligence optimization strategy is a strategy that optimizes the target parameters through swarm search and iterative updates. The neural network model is a function mapping model that takes multi-order natural frequency features as input and damage parameters as output. The candidate identification model is one or more model instances trained under the above modeling strategies that can be used to output damage parameters.

[0029] Specifically, based on the vibration feature sample library, a training data set for learning is constructed, such that each training sample contains: the multi-order natural frequency feature as input, and the damage parameter corresponding to the input sample as output label; on this basis, the neural network model type and network structure for modeling are determined according to the preset hybrid intelligent modeling strategy, so that the neural network model has the mapping ability of taking the multi-order natural frequency feature as input and the damage parameter as output.

[0030] Furthermore, after determining the neural network model framework, the swarm intelligence optimization strategy is invoked to optimize the parameters of the neural network model: the set of parameters to be optimized (including network parameters that affect model prediction) is used as the search object of the group individuals, the neural network model is driven to generate prediction output based on training data, and a fitness evaluation is constructed with the prediction error, thereby obtaining target parameters that match the fitness during the iterative update process; subsequently, the target parameters are loaded into the neural network model and training is completed, so that the neural network model forms a mapping relationship from the multi-order intrinsic frequency features to the damage parameters, thereby training the candidate recognition model.

[0031] By constructing supervised training data of "multi-order natural frequency features - damage parameters" based on the vibration feature sample library, and using a hybrid intelligent modeling method that includes a swarm intelligence optimization strategy to optimize and retrain the neural network model parameters to obtain candidate recognition models, the parameter values ​​of the candidate recognition models are determined by "sample library-driven fitness evaluation and swarm iterative optimization" rather than relying solely on single initialization or a fixed training path. This makes the neural network model more closely fit the input-output correspondence established in the sample library, reduces prediction errors in the mapping learning process, and provides a more stable fitting basis for subsequent output damage parameters.

[0032] S4: Based on the preset engineering requirements constraints, determine the target recognition model from the candidate recognition models, and collect the vibration signal of the laminate to be detected to extract the multi-order natural frequency features consistent with the vibration feature sample library to obtain the vibration features to be tested; S5: Perform random noise processing and frequency offset correction on the vibration feature to be tested to obtain the corrected vibration feature to be tested, and input the corrected vibration feature to be tested into the target recognition model to obtain the damage parameter recognition result of the laminate to be tested.

[0033] It should be noted that the preset engineering requirement constraints are model selection constraints pre-defined for different engineering application scenarios. The candidate recognition model is one or more usable model instances trained using a vibration feature sample library and a hybrid intelligent modeling strategy. The target recognition model is the model selected from the candidate recognition models for actual recognition output, provided that the preset engineering requirement constraints are met. The vibration features to be measured are a set of multi-order natural frequency features extracted from the vibration signal. Random noise processing is the process of adding random noise to the acquired / extracted natural frequency features. Frequency offset correction is the process of performing offset correction on the noisy natural frequency features.

[0034] Specifically, the preset engineering requirement constraints are obtained and used as selection criteria for candidate recognition models. Based on this, the candidate recognition models are adapted to meet the engineering requirements to determine the target recognition model that satisfies the preset engineering requirement constraints. Subsequently, the vibration signal acquisition unit is activated to acquire vibration signals from the laminate to be tested, obtaining vibration signals characterizing the dynamic response of the laminate. Next, the vibration signals are processed based on a preset feature extraction process to extract multi-order natural frequency features that are consistent with the vibration feature sample library in terms of feature form and order composition. These multi-order natural frequency features are then aggregated to form the vibration features to be tested.

[0035] Furthermore, after obtaining the vibration characteristics to be tested, random noise processing is performed on the vibration characteristics to obtain noisy vibration characteristics containing random disturbances; then, frequency offset correction is performed on the noisy vibration characteristics based on a preset frequency offset correction rule to obtain corrected vibration characteristics. Finally, the corrected vibration characteristics are input into the target recognition model, and the target recognition model outputs the damage parameter identification result of the laminate to be tested.

[0036] By first determining the target recognition model from candidate recognition models based on preset engineering requirements constraints during the recognition stage, the model used for subsequent inference is matched with the engineering scenario constraints. Simultaneously, by collecting vibration signals of the laminate to be detected and extracting multi-order natural frequency features consistent with the vibration feature sample library to form the vibration features to be tested, the online input features and the features in the training sample library maintain the same feature definition. Furthermore, before inputting the vibration features to be tested into the target recognition model, random noise processing and frequency offset correction are performed on them. This allows random disturbances and frequency offsets to be processed and corrected at the feature level, thereby reducing the impact of feature mismatch on model inference. As a result, the target recognition model can output the damage parameter recognition results of the laminate to be detected based on more consistent input features.

[0037] This embodiment obtains the structural description information of the laminate to be tested and constructs an analysis model of the laminate including damage parameter representation based on the structural description information to determine the parameter space used to characterize the damage state. Based on the parameter space, the multi-order natural frequencies corresponding to different combinations of damage parameters are solved to obtain a vibration feature sample library. The vibration feature sample library establishes a sample correspondence between multi-order natural frequency features and damage parameters. Based on the vibration feature sample library, a damage parameter prediction model is constructed using a preset hybrid intelligent modeling strategy. The hybrid intelligent modeling strategy includes using a swarm intelligence optimization strategy to optimize the parameters of the neural network model and training a candidate recognition model with multi-order natural frequency features as input and damage parameters as output. Based on preset engineering requirement constraints, a target recognition model is determined from the candidate recognition models, and the vibration signal of the laminate to be tested is collected to extract multi-order natural frequency features consistent with the vibration feature sample library to obtain the vibration features to be tested. The vibration features to be tested are subjected to random noise processing and frequency offset correction to obtain corrected vibration features to be tested. The corrected vibration features to be tested are then input into the target recognition model to obtain the damage parameter recognition result of the laminate to be tested. This embodiment first constructs an analysis model containing damage parameter representations based on structural description information and determines the parameter space. Within this parameter space, it solves for the multi-order natural frequencies corresponding to different combinations of damage parameters, forming a vibration feature sample library that establishes a correspondence between "multi-order natural frequency features and damage parameters," thus solidifying the supervised mapping required for identification from the source. Then, based on the sample library, it uses a hybrid intelligent modeling strategy that includes swarm intelligence optimization to optimize the parameters of the neural network model and train it to obtain candidate identification models. This makes the model parameters more closely match the mapping relationship defined in the sample library, thereby reducing fitting errors. Finally, in the inference stage, it selects the target identification model from the candidate models according to engineering requirements constraints and extracts the multi-order natural frequency features consistent with the sample library as the vibration features to be measured. After random noise processing and frequency offset correction, these features are then input into the target identification model, thereby reducing feature mismatch and system bias caused by noise and offset, and thus improving the accuracy of the output damage parameter identification results.

[0038] Based on the first embodiment described above, a second embodiment of the composite laminate damage identification method of this application is proposed. Please refer to... Figure 2 , Figure 2 This is a schematic diagram of a sub-process in the second embodiment of the composite material laminate damage identification method of this application.

[0039] like Figure 2 As shown, in this embodiment, step S1 includes: S11: Obtain the ply information, geometric information, material parameter information, and boundary constraint information of the laminate to be tested, and combine the ply information, geometric information, material parameter information, and boundary constraint information to obtain the structural description information; S12: Based on the structural description information, establish a partitioned structural model including interlayer delamination, divide the laminate to be tested into multiple sub-layer regions, and define damage parameters in the partitioned structural model to characterize the delamination length and delamination position, so that the damage parameters are associated with the sub-layer regions; S13: Based on the partitioned structure model, the displacement field equations of each sub-layer region are established using classical laminated plate theory, and the displacement continuity constraints of the delamination region are set in combination with the damage parameters to obtain the laminated plate analysis model, so as to determine the parameter space used to characterize the damage state.

[0040] It should be noted that ply information is structural information used to describe the composition and lamination sequence of each ply in a laminate. Geometric information is information used to describe the geometric elements of the laminate, such as its external dimensions and thickness direction. Material parameter information is parameter information used to describe the mechanical / physical properties of the materials in each ply. Boundary constraint information is information used to describe the constraint form of the laminate under engineering boundary conditions. The partitioned structural model is a structural partitioning model built on the structural description information. Interlaminar delamination is the separation / damage mode that occurs between adjacent plies of a laminate. Sublayer regions are regional units obtained by partitioning the laminate based on the partitioned structural model. Classical laminate theory is the theoretical framework used to describe the mechanical response of laminates. Displacement continuity constraints are displacement continuity requirements applied at delamination regions and their boundaries. The laminate analysis model is established by integrating structural description information, partitioned structural model, damage parameters, displacement field, and continuity constraints, and is used to carry out subsequent analysis and solution under given damage parameters.

[0041] Specifically, the layup information, geometric information, material parameter information, and boundary constraint information of the laminate to be tested are obtained, and this information is uniformly organized and aggregated to form structural description information that can be used for modeling. Then, based on the structural description information, a partitioned structural model including interlayer delamination is established, dividing the laminate to be tested into multiple sub-layer regions. Damage parameters characterizing delamination length and location are defined in the partitioned structural model, establishing a correlation between the damage parameters and the corresponding sub-layer regions, so that the subsequent analysis process can express the damage state within the "region partitioning-damage parameter" framework.

[0042] Furthermore, after obtaining the partitioned structural model and its damage parameter definitions, displacement field equations are established for each sub-layer region using classical laminated plate theory, so that the partitioned structural model has a basic expression that can be used to describe the structural displacement response. Further, displacement continuity constraints are set in the delamination region in conjunction with the damage parameters, so that the displacement compatibility conditions of the delamination region and its adjacent regions can be consistently applied at the model level. Finally, the displacement field equations and displacement continuity constraints are incorporated into the partitioned structural model to obtain the laminated plate analysis model, and a parameter space for characterizing the damage state is determined based on the laminated plate analysis model.

[0043] First, information such as ply, geometry, material parameters, and boundary constraints is collected to form structural description information. Based on this, a partitioned structural model including delamination is constructed, so that the delamination length and delamination location are explicitly introduced in the form of damage parameters and associated with the sub-layer regions. Then, displacement field equations are established for each sub-layer region based on classical laminated plate theory, and displacement continuity constraints are applied to the delamination region in combination with damage parameters. This allows the laminated plate analysis model to express the consistent relationship of "partitioned structure - damage parameters - displacement response - continuity conditions" under the same modeling framework. In this way, the definition of the parameter space has clear damage state characterization boundaries and combination basis, providing a unified and reusable model foundation for subsequent natural frequency solving and sample correspondence establishment in the parameter space.

[0044] Based on the first embodiment described above, in this embodiment, step S2 includes: S21: Based on the parameter space, generate multiple damage parameter combinations, and determine the partition structure description corresponding to each damage parameter combination. S22: Based on the description of each partition structure, a displacement function that satisfies the boundary constraints and the displacement continuity requirement of the delamination region is constructed using Chebyshev polynomials, and the system stiffness matrix and mass matrix corresponding to the displacement function are established based on the Rayleigh-Ritz energy analysis algorithm. S23: Solve the vibration equation based on the system stiffness matrix and the mass matrix to obtain the multi-order natural frequencies corresponding to each combination of damage parameters, and establish a sample association relationship between each multi-order natural frequency and the corresponding combination of damage parameters to obtain a vibration feature sample library.

[0045] It should be noted that the displacement function is constructed using Chebyshev polynomials as basis functions. The Rayleigh-Ritz energy analysis algorithm is a Ritz-class solution method based on energy principles, used to derive and establish the system's stiffness and mass matrices from the displacement function. The mass matrix is ​​the system matrix established by Rayleigh-Ritz energy analysis using the displacement function.

[0046] Specifically, multiple sets of damage parameter combinations are generated based on the parameter space, and for each set of damage parameter combinations, a corresponding partition structure description is determined so that the partition structure description can reflect the partition relationship between the delaminated and non-delaminated regions and the connection relationship between the regions under the damage state. In this way, the "damage state enumeration" in the parameter space is transformed into "structural partition input conditions" that can be directly called by the subsequent solution process.

[0047] Furthermore, for each of the described partition structures, a Chebyshev polynomial is used to construct a displacement function, ensuring that the displacement function simultaneously satisfies boundary constraints and the displacement continuity requirement of the delamination region. Based on this, a system stiffness matrix and mass matrix corresponding to the displacement function are established using the Rayleigh-Ritz energy analysis algorithm. The vibration equation is then solved based on the system stiffness matrix and the mass matrix to obtain the multi-order natural frequencies corresponding to each combination of damage parameters. Finally, a sample correlation is established between each of the multi-order natural frequencies and the corresponding combination of damage parameters, and these are written into a dataset to obtain a vibration feature sample library.

[0048] By generating multiple sets of damage parameter combinations in the parameter space and determining a corresponding partitioned structural description for each set, different damage states can be systematically enumerated under unified structural partitioning input conditions. Furthermore, a displacement function that simultaneously satisfies boundary constraints and delamination displacement continuity requirements is constructed using Chebyshev polynomials. The system stiffness matrix and mass matrix are established using Rayleigh-Ritz energy analysis to solve the vibration equation and obtain multiple natural frequencies. This ensures that each set of damage parameter combinations yields a corresponding set of natural frequency characteristics that satisfy the constraint conditions. Based on this, the multiple natural frequencies and damage parameter combinations are correlated to form a vibration feature sample library, creating stable and consistent "vibration feature-damage parameter" pairing samples. This provides a more reliable mapping basis for subsequent supervised modeling based on the sample library, thereby reducing identification bias caused by missing sample pairings or inconsistent features.

[0049] This embodiment obtains the structural description information of the laminate to be tested and constructs an analysis model of the laminate including damage parameter representation based on the structural description information to determine the parameter space used to characterize the damage state. Based on the parameter space, the multi-order natural frequencies corresponding to different combinations of damage parameters are solved to obtain a vibration feature sample library. The vibration feature sample library establishes a sample correspondence between multi-order natural frequency features and damage parameters. Based on the vibration feature sample library, a damage parameter prediction model is constructed using a preset hybrid intelligent modeling strategy. The hybrid intelligent modeling strategy includes using a swarm intelligence optimization strategy to optimize the parameters of the neural network model and training a candidate recognition model with multi-order natural frequency features as input and damage parameters as output. Based on preset engineering requirement constraints, a target recognition model is determined from the candidate recognition models, and the vibration signal of the laminate to be tested is collected to extract multi-order natural frequency features consistent with the vibration feature sample library to obtain the vibration features to be tested. The vibration features to be tested are subjected to random noise processing and frequency offset correction to obtain corrected vibration features to be tested. The corrected vibration features to be tested are then input into the target recognition model to obtain the damage parameter recognition result of the laminate to be tested. This embodiment first constructs an analysis model containing damage parameter representations based on structural description information and determines the parameter space. Within this parameter space, it solves for the multi-order natural frequencies corresponding to different combinations of damage parameters, forming a vibration feature sample library that establishes a correspondence between "multi-order natural frequency features and damage parameters," thus solidifying the supervised mapping required for identification from the source. Then, based on the sample library, it uses a hybrid intelligent modeling strategy that includes swarm intelligence optimization to optimize the parameters of the neural network model and train it to obtain candidate identification models. This makes the model parameters more closely match the mapping relationship defined in the sample library, thereby reducing fitting errors. Finally, in the inference stage, it selects the target identification model from the candidate models according to engineering requirements constraints and extracts the multi-order natural frequency features consistent with the sample library as the vibration features to be measured. After random noise processing and frequency offset correction, these features are then input into the target identification model, thereby reducing feature mismatch and system bias caused by noise and offset, and thus improving the accuracy of the output damage parameter identification results.

[0050] Based on the second embodiment described above, a third embodiment of the composite material laminate damage identification method of this application is proposed. Please refer to... Figure 3 , Figure 3 This is a schematic diagram of a sub-process in the third embodiment of the composite material laminate damage identification method of this application.

[0051] In this embodiment, step S3 includes: S31: Based on the vibration feature sample library, construct a training sample set, such that each sample in the training sample set includes multi-order natural frequency features corresponding to the same damage parameter as input samples, and the damage parameter as supervision labels, and perform feature unification processing on the training sample set to obtain standardized training data. S32: Based on the standardized training data, set the network structure of the neural network model and the set of parameters to be optimized, and use a swarm intelligence optimization strategy to generate a set of candidate parameters. Based on each candidate parameter in the set of candidate parameters, drive the neural network model to make predictions on the standardized training data and determine the prediction error. Use the prediction error as the fitness and iteratively update the set of candidate parameters to obtain the target parameters. S33: Load the target parameters into the neural network model, and train the loaded neural network model based on the standardized training data, so that the neural network model forms a mapping relationship from the multi-order intrinsic frequency features to the damage parameters, and train to obtain a candidate recognition model.

[0052] It should be noted that the supervisory label is the annotation information corresponding to the input sample, used to supervise the training of the neural network to learn the mapping relationship from frequency features to damage parameters. The neural network model is a learning model used to establish the mapping relationship between "multi-order intrinsic frequency features → damage parameters". The set of parameters to be optimized is the set of parameters in the neural network model that needs to be determined through optimization. The swarm intelligence optimization strategy is an optimization strategy used to search for the target parameters of the neural network in the candidate parameter space. The target parameters are the parameter values ​​determined after iterative search and used to load into the neural network model. The candidate recognition model is the model obtained after loading the target parameters and completing training.

[0053] Specifically, a training sample set is constructed using the vibration feature sample library as the data source, ensuring that each sample in the training sample set is organized according to the same pairing logic: multi-order natural frequency features corresponding to the same damage parameter constitute the input sample, and the damage parameter is written as a supervision label into the corresponding sample record, thereby forming a "feature-label" sample structure that can be used for supervised learning. Subsequently, the training sample set undergoes feature unification processing, standardizing the input samples in terms of feature expression and data organization to meet the data input specifications required for neural network training, thus obtaining standardized training data.

[0054] Further, the process involves: standardized training data → network structure / parameters to be optimized → swarm optimization → loading target parameters for training → candidate recognition model. After obtaining standardized training data, the network structure of the neural network model is set, and the set of parameters to be optimized is determined. A swarm optimization strategy is used to search for network parameters, for example, using particle swarm optimization to optimize the initial weights, biases, and learning rate, or using whale optimization to optimize the initial weights and thresholds. Subsequently, the swarm optimization strategy is used to generate a set of candidate parameters, and the fitness of each candidate parameter in the set is evaluated: the candidate parameters are loaded into the neural network model to drive it to predict the standardized training data, the prediction error is calculated, and the prediction error is used as fitness to guide the iterative update of the candidate parameters, thereby obtaining the target parameters. Finally, the target parameters are loaded into the neural network model, and the loaded neural network model is trained based on the standardized training data to form a mapping relationship from the multi-order intrinsic frequency features to the damage parameters, thereby training a candidate recognition model.

[0055] By organizing the vibration feature sample library into a training sample set with "multi-order natural frequency features as input and damage parameters as supervision labels," and standardizing the training samples through feature unification, the input feature expressions received by the neural network model during the training phase are kept consistent, thereby reducing training bias caused by differences in feature expressions. Furthermore, a swarm intelligence optimization strategy is introduced before training to search for the set of parameters to be optimized in the neural network model, and the prediction error is used as the fitness to drive the iterative update of candidate parameters, so that the model training starts from a more suitable parameter starting point. This enables the neural network model to learn the mapping relationship of "multi-order natural frequency features → damage parameters" more stably and form a candidate recognition model, thus providing a more reliable model foundation for subsequent damage parameter recognition.

[0056] Based on the second embodiment described above, in this embodiment, step S4 includes: S41: Obtain the preset engineering requirement constraints, and convert the preset engineering requirement constraints into evaluation criteria for recognition accuracy requirements and / or response delay requirements. Based on the evaluation criteria, evaluate the output error and computational overhead of the candidate recognition model on the preset verification data. Based on the evaluation results, determine the target recognition model that meets the preset engineering requirement constraints from the candidate recognition models. S42: Based on the target recognition model, determine the vibration acquisition configuration and feature extraction configuration corresponding to the vibration feature sample library, and based on the vibration acquisition configuration, excite and acquire the response of the laminate to be tested to obtain a vibration signal characterizing the dynamic response of the laminate to be tested; S43: Based on the feature extraction configuration, frequency domain analysis and modal parameter extraction are performed on the vibration signal to obtain multi-order natural frequency features consistent with the vibration feature sample library, and the multi-order natural frequency features are collected to obtain the vibration feature to be measured.

[0057] It should be noted that the preset verification data is a dataset used to evaluate candidate recognition models. Output error is a measure of the deviation between the damage parameters output by the candidate recognition model on the preset verification data and the corresponding annotations. The target recognition model is a model selected from the candidate recognition models that meets preset engineering requirements and is used for actual recognition output. Vibration acquisition configuration is a set of configurations related to vibration signal acquisition. Feature extraction configuration is a set of configurations related to feature extraction. Excitation and response acquisition is the process of applying excitation to the laminate to be tested and acquiring its dynamic response to obtain vibration signals. Modal parameter extraction is the process of extracting modal-related parameters from the vibration signal (or its frequency domain representation).

[0058] Specifically, pre-defined engineering requirements constraints are obtained and transformed into evaluation criteria for recognition accuracy and / or response latency requirements, enabling these requirements to be used for model selection in a comparable and measurable manner. Subsequently, based on the evaluation criteria, candidate recognition models are evaluated on pre-defined validation data: on the one hand, the output error of the candidate models is statistically analyzed to characterize their recognition accuracy level; on the other hand, their computational overhead is statistically analyzed to characterize their response latency performance. Finally, based on the evaluation results, the target recognition model that meets the pre-defined engineering requirements constraints is determined from the candidate models.

[0059] Furthermore, after determining the target recognition model, the vibration acquisition configuration and feature extraction configuration corresponding to the vibration feature sample library are determined based on the target recognition model, so that the subsequently acquired test features are consistent with the sample library in feature form. Subsequently, according to the vibration acquisition configuration, excitation and response acquisition are performed on the laminate to be tested to obtain vibration signals characterizing its dynamic response; then, according to the feature extraction configuration, frequency domain analysis and modal parameter extraction are performed on the vibration signals to obtain multi-order natural frequency features consistent with the vibration feature sample library, and the multi-order natural frequency features are collected to form the test vibration features.

[0060] By converting pre-defined engineering requirements into evaluation criteria and simultaneously assessing the output error and computational overhead of candidate recognition models on pre-defined verification data, the determination of the target recognition model is based on the constraint matching of "accuracy requirements and time delay requirements," avoiding a disconnect between model selection and engineering needs. Simultaneously, based on the target recognition model, the vibration acquisition configuration and feature extraction configuration corresponding to the vibration feature sample library are determined, and multi-order natural frequency features consistent with the sample library are extracted from the vibration signal to form the vibration features to be measured. This ensures that the online input features and the training sample library features are consistent in definition and organization, thereby reducing the risk of feature mismatch caused by inconsistencies in acquisition methods or feature expression, and providing a more stable input foundation for subsequent damage parameter recognition results based on the target recognition model.

[0061] Based on the second embodiment described above, in this embodiment, step S5 includes: S51: Obtain the vibration feature to be tested, and perform random noise superposition processing on the vibration feature to be tested according to the preset noise distribution and preset noise amplitude rules to obtain the noisy vibration feature to be tested; S52: Obtain the reference natural frequency corresponding to each natural frequency feature in the noisy vibration feature to be tested, and determine the frequency offset of each natural frequency based on the reference natural frequency and the noisy vibration feature to be tested, and perform frequency offset correction on the noisy vibration feature to be tested based on the frequency offset to obtain the corrected vibration feature to be tested. S53: Input the modified vibration characteristics to be tested into the target recognition model to obtain the damage parameter recognition result of the laminate to be tested.

[0062] It should be noted that the preset noise distribution is a statistical distribution used to generate random noise. The preset noise amplitude rule is a pre-defined rule used to limit the range of random noise superposition amplitude. Random noise superposition processing is a process of randomly perturbing and superimposing the vibration characteristics to be measured according to the preset noise distribution and noise amplitude rule to obtain noisy characteristics. The noisy vibration characteristics to be measured are the vibration characteristic data to be measured obtained after random noise superposition processing. The frequency offset is the deviation between the reference natural frequency and the corresponding order natural frequency in the noisy vibration characteristics to be measured.

[0063] Specifically, the vibration characteristics to be measured are acquired, and a preset noise distribution and a preset noise amplitude rule are determined. The preset noise distribution defines the random generation method of the noise, and the preset noise amplitude rule defines the amplitude range of the noise superposition. Subsequently, random noise is generated according to the preset noise distribution, and the random noise is superimposed on the vibration characteristics to be measured according to the preset noise amplitude rule, thereby obtaining noisy vibration characteristics to be measured, so that the vibration characteristics to be measured have a representational form of random disturbances before being input into the model.

[0064] Furthermore, after obtaining the noisy vibration characteristics to be tested, a reference natural frequency corresponding to each natural frequency characteristic in the noisy vibration characteristics to be tested is obtained, and the frequency offset of each order is determined based on the reference natural frequency and the noisy vibration characteristics to be tested; subsequently, frequency offset correction is performed on the noisy vibration characteristics to be tested according to the frequency offset, to obtain the corrected vibration characteristics to be tested. Finally, the corrected vibration characteristics to be tested are input into the target recognition model, and the target recognition model outputs the damage parameter recognition result of the laminate to be tested.

[0065] By randomly superimposing noise onto the vibration features under test according to a preset noise distribution and preset noise amplitude rules, the vibration features under test can reflect random disturbances before entering the recognition model. Then, using the reference natural frequency as a correction benchmark, the frequency offset of each natural frequency of the noisy vibration features under test is determined and frequency offset correction is performed accordingly. This corrects the noisy input features into corrected vibration features that can be used for recognition inference. On this basis, the corrected vibration features under test are input into the target recognition model, which can reduce the impact of random disturbances and frequency offsets on the consistency of input features. This allows the target recognition model to output damage parameter recognition results based on more consistent features, thereby improving the stability and reliability of the recognition process in noisy environments.

[0066] This embodiment obtains the structural description information of the laminate to be tested and constructs an analysis model of the laminate including damage parameter representation based on the structural description information to determine the parameter space used to characterize the damage state. Based on the parameter space, the multi-order natural frequencies corresponding to different combinations of damage parameters are solved to obtain a vibration feature sample library. The vibration feature sample library establishes a sample correspondence between multi-order natural frequency features and damage parameters. Based on the vibration feature sample library, a damage parameter prediction model is constructed using a preset hybrid intelligent modeling strategy. The hybrid intelligent modeling strategy includes using a swarm intelligence optimization strategy to optimize the parameters of the neural network model and training a candidate recognition model with multi-order natural frequency features as input and damage parameters as output. Based on preset engineering requirement constraints, a target recognition model is determined from the candidate recognition models, and the vibration signal of the laminate to be tested is collected to extract multi-order natural frequency features consistent with the vibration feature sample library to obtain the vibration features to be tested. The vibration features to be tested are subjected to random noise processing and frequency offset correction to obtain corrected vibration features to be tested. The corrected vibration features to be tested are then input into the target recognition model to obtain the damage parameter recognition result of the laminate to be tested. This embodiment first constructs an analysis model containing damage parameter representations based on structural description information and determines the parameter space. Within this parameter space, it solves for the multi-order natural frequencies corresponding to different combinations of damage parameters, forming a vibration feature sample library that establishes a correspondence between "multi-order natural frequency features and damage parameters," thus solidifying the supervised mapping required for identification from the source. Then, based on the sample library, it uses a hybrid intelligent modeling strategy that includes swarm intelligence optimization to optimize the parameters of the neural network model and train it to obtain candidate identification models. This makes the model parameters more closely match the mapping relationship defined in the sample library, thereby reducing fitting errors. Finally, in the inference stage, it selects the target identification model from the candidate models according to engineering requirements constraints and extracts the multi-order natural frequency features consistent with the sample library as the vibration features to be measured. After random noise processing and frequency offset correction, these features are then input into the target identification model, thereby reducing feature mismatch and system bias caused by noise and offset, and thus improving the accuracy of the output damage parameter identification results.

[0067] This application also provides a composite laminate damage identification device. Please refer to... Figure 4 , Figure 4 This is a schematic diagram of the module structure of the composite laminate damage identification device according to an embodiment of this application. The composite laminate damage identification device includes: The parameter space module 401 is used to acquire the structural description information of the laminate to be tested, and to construct a laminate analysis model including damage parameter characterization based on the structural description information, so as to determine the parameter space used to characterize the damage state. The vibration feature module 402 is used to solve for the multi-order natural frequencies corresponding to different combinations of damage parameters based on the parameter space, and obtain a vibration feature sample library. The vibration feature sample library establishes a sample correspondence between multi-order natural frequency features and the damage parameters. The candidate model module 403 is used to construct a damage parameter prediction model based on the vibration feature sample library and using a preset hybrid intelligent modeling strategy. The hybrid intelligent modeling strategy includes using a swarm intelligence optimization strategy to optimize the parameters of the neural network model and training the candidate recognition model with the multi-order natural frequency features as input and the damage parameters as output. The constraint feature module 404 is used to determine the target recognition model from the candidate recognition models based on the preset engineering requirement constraints, and to collect the vibration signal of the laminate to be tested to extract multi-order natural frequency features consistent with the vibration feature sample library to obtain the vibration features to be tested. The target module 405 is used to perform random noise processing and frequency offset correction on the vibration feature to be tested to obtain the corrected vibration feature to be tested, and input the corrected vibration feature to be tested into the target recognition model to obtain the damage parameter recognition result of the laminate to be tested.

[0068] The composite laminate damage identification device provided in this application, employing the composite laminate damage identification method described in the above embodiments, can solve the technical problem of how to improve the accuracy of composite laminate damage identification results. Compared with the prior art, the beneficial effects of the composite laminate damage identification device provided in this application are the same as those of the composite laminate damage identification method provided in the above embodiments, and other technical features in the composite laminate damage identification device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0069] This application provides a composite laminate damage identification device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the composite laminate damage identification method in the above embodiments.

[0070] The following is for reference. Figure 5 , Figure 5 This is a schematic diagram of the hardware operating environment of the composite laminate damage identification method in the embodiments of this application, showing a schematic diagram of the structure of the composite laminate damage identification device suitable for implementing the embodiments of this application. Figure 5The composite laminate damage identification device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this application.

[0071] like Figure 5 As shown, the composite laminate damage identification device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the composite laminate damage identification device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the composite laminate damage identification device to communicate wirelessly or wiredly with other devices to exchange data. Although composite laminate damage identification devices with various systems are shown in the figures, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems may be implemented alternatively.

[0072] In particular, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. When the computer program is executed by the processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0073] The composite laminate damage identification device provided in this application, employing the composite laminate damage identification method described in the above embodiments, can solve the technical problem of how to improve the accuracy of composite laminate damage identification results. Compared with the prior art, the beneficial effects of the composite laminate damage identification device provided in this application are the same as those of the composite laminate damage identification method provided in the above embodiments, and other technical features of this composite laminate damage identification device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0074] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0075] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0076] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the composite laminate damage identification method in the above embodiments.

[0077] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by the composite laminate damage identification device, the composite laminate damage identification device: acquires structural description information of the laminate to be inspected, and constructs a laminate analysis model including damage parameter characterization based on the structural description information to determine the parameter space used to characterize the damage state; based on the parameter space, solves for the multi-order natural frequencies corresponding to different combinations of damage parameters to obtain a vibration feature sample library, in which a sample correspondence between multi-order natural frequency features and damage parameters is established; based on the vibration feature sample library, a preset hybrid intelligent modeling strategy is used to construct... A damage parameter prediction model is constructed. The hybrid intelligent modeling strategy includes optimizing the parameters of the neural network model using a swarm intelligence optimization strategy, and training a candidate recognition model with multi-order natural frequency features as input and damage parameters as output. Based on preset engineering requirement constraints, a target recognition model is determined from the candidate recognition models, and vibration signals of the laminate to be tested are collected to extract multi-order natural frequency features consistent with the vibration feature sample library, thus obtaining the vibration features to be tested. Random noise processing and frequency offset correction are applied to the vibration features to be tested to obtain corrected vibration features to be tested, which are then input into the target recognition model to obtain the damage parameter recognition result of the laminate to be tested. Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—and conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer through any type of network—including a local area network (LAN) or a wide area network (WAN)—or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0078] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0079] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0080] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described composite laminate damage identification method, thereby solving the technical problem of how to improve the accuracy of composite laminate damage identification results. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the composite laminate damage identification method provided in the above embodiments, and will not be repeated here.

[0081] This application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the composite laminate damage identification method described above.

[0082] The computer program product provided in this application can solve the technical problem of how to improve the accuracy of damage identification results for composite laminates. Compared with the prior art, the beneficial effects of the computer program product provided in the embodiments of this application are the same as the beneficial effects of the composite laminate damage identification method provided in the above embodiments, and will not be repeated here.

[0083] The above are merely preferred embodiments of this application and do not limit the scope of protection of this application. Any equivalent structural or procedural transformations made based on the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the scope of this application.

Claims

1. A method for identifying damage in composite laminates, characterized in that, The method includes: Obtain the structural description information of the laminate to be tested, and construct a laminate analysis model including damage parameter characterization based on the structural description information to determine the parameter space used to characterize the damage state; Based on the parameter space, the multi-order natural frequencies corresponding to different combinations of damage parameters are solved to obtain a vibration feature sample library. The vibration feature sample library establishes a sample correspondence between multi-order natural frequency features and the damage parameters. Based on the vibration feature sample library, a damage parameter prediction model is constructed using a preset hybrid intelligent modeling strategy. The hybrid intelligent modeling strategy includes optimizing the parameters of the neural network model using a swarm intelligence optimization strategy, and training a candidate recognition model with the multi-order natural frequency features as input and the damage parameters as output. Based on the preset engineering requirements constraints, the target recognition model is determined from the candidate recognition models, and the vibration signal of the laminate to be detected is collected to extract the multi-order natural frequency features consistent with the vibration feature sample library, so as to obtain the vibration features to be tested. The vibration characteristics to be tested are subjected to random noise processing and frequency offset correction to obtain corrected vibration characteristics to be tested. The corrected vibration characteristics to be tested are then input into the target recognition model to obtain the damage parameter recognition result of the laminate to be tested.

2. The method as described in claim 1, characterized in that, The step of obtaining structural description information of the laminate to be tested and constructing a laminate analysis model including damage parameter characterization based on the structural description information to determine the parameter space used to characterize the damage state includes: The layup information, geometric information, material parameter information, and boundary constraint information of the laminate to be tested are obtained, and the layup information, geometric information, material parameter information, and boundary constraint information are aggregated to obtain the structural description information; Based on the structural description information, a partitioned structural model including interlayer delamination is established, the laminate to be tested is divided into multiple sub-layer regions, and damage parameters for characterizing delamination length and delamination location are defined in the partitioned structural model so that the damage parameters are associated with the sub-layer regions. Based on the partitioned structure model, the displacement field equations of each sub-layer region are established using classical laminated plate theory, and the displacement continuity constraints of the delamination region are set in combination with the damage parameters to obtain the laminated plate analysis model, thereby determining the parameter space used to characterize the damage state.

3. The method as described in claim 1, characterized in that, The step of solving for the multiple natural frequencies corresponding to different combinations of damage parameters based on the parameter space to obtain a vibration feature sample library includes: Based on the parameter space, multiple damage parameter combinations are generated, and a partition structure description corresponding to each damage parameter combination is determined. Based on the description of each partition structure, a displacement function that satisfies the boundary constraints and the displacement continuity requirement of the delamination region is constructed using Chebyshev polynomials, and the system stiffness matrix and mass matrix corresponding to the displacement function are established based on the Rayleigh-Ritz energy analysis algorithm. The vibration equation is solved based on the system stiffness matrix and the mass matrix to obtain the multi-order natural frequencies corresponding to each combination of damage parameters. A sample association relationship is established between each multi-order natural frequency and the corresponding combination of damage parameters to obtain a vibration feature sample library.

4. The method as described in claim 1, characterized in that, The step of constructing a damage parameter prediction model based on the vibration feature sample library using a preset hybrid intelligent modeling strategy, wherein the hybrid intelligent modeling strategy includes optimizing the parameters of the neural network model using a swarm intelligence optimization strategy, and training a candidate recognition model using the multi-order natural frequency features as input and the damage parameters as output, includes: Based on the vibration feature sample library, a training sample set is constructed, such that each sample in the training sample set includes multi-order natural frequency features corresponding to the same damage parameter as input samples, and the damage parameter as supervision labels. The training sample set is then subjected to feature unification processing to obtain standardized training data. Based on the standardized training data, the network structure and parameter set of the neural network model are set, and a swarm intelligence optimization strategy is used to generate a candidate parameter set. Based on each candidate parameter in the candidate parameter set, the neural network model is driven to make predictions on the standardized training data and determine the prediction error. The prediction error is used as the fitness and the candidate parameter set is iteratively updated to obtain the target parameters. The target parameters are loaded into the neural network model, and the loaded neural network model is trained based on the standardized training data, so that the neural network model forms a mapping relationship from the multi-order intrinsic frequency features to the damage parameters, and a candidate recognition model is obtained through training.

5. The method as described in claim 1, characterized in that, The steps of determining the target recognition model from the candidate recognition models based on preset engineering requirement constraints, and collecting the vibration signal of the laminate to be tested to extract multi-order natural frequency features consistent with the vibration feature sample library, to obtain the vibration features to be tested, include: The preset engineering requirement constraints are obtained and converted into evaluation criteria for recognition accuracy requirements and / or response latency requirements. Based on the evaluation criteria, the output error and computational overhead of the candidate recognition models on preset verification data are evaluated. Based on the evaluation results, the target recognition model that meets the preset engineering requirement constraints is determined from the candidate recognition models. Based on the target recognition model, the vibration acquisition configuration and feature extraction configuration corresponding to the vibration feature sample library are determined, and the excitation and response acquisition of the laminate to be tested are performed based on the vibration acquisition configuration to obtain the vibration signal characterizing the dynamic response of the laminate to be tested. Based on the feature extraction configuration, frequency domain analysis and modal parameter extraction are performed on the vibration signal to obtain multi-order natural frequency features consistent with the vibration feature sample library, and the multi-order natural frequency features are collected to obtain the vibration feature to be measured.

6. The method as described in claim 1, characterized in that, The steps of performing random noise processing and frequency offset correction on the vibration characteristics to be tested to obtain corrected vibration characteristics, and inputting the corrected vibration characteristics to be tested into the target recognition model to obtain the damage parameter recognition result of the laminate to be tested, include: The vibration characteristics to be tested are obtained, and random noise superposition processing is performed on the vibration characteristics to be tested according to the preset noise distribution and preset noise amplitude rules to obtain the noisy vibration characteristics to be tested. Obtain the reference natural frequency corresponding to each natural frequency feature in the noisy vibration feature to be tested, and determine the frequency offset of each order based on the reference natural frequency and the noisy vibration feature to be tested. Based on the frequency offset, perform frequency offset correction on the noisy vibration feature to be tested to obtain the corrected vibration feature to be tested. The modified vibration characteristics to be tested are input into the target recognition model to obtain the damage parameter recognition results of the laminate to be tested.

7. A device for identifying damage in composite laminates, characterized in that, The device includes: The parameter space module is used to acquire the structural description information of the laminate to be tested, and to construct a laminate analysis model including damage parameter characterization based on the structural description information, so as to determine the parameter space used to characterize the damage state. The vibration feature module is used to solve for the multi-order natural frequencies corresponding to different combinations of damage parameters based on the parameter space, and obtain a vibration feature sample library. The vibration feature sample library establishes a sample correspondence between multi-order natural frequency features and the damage parameters. The candidate model module is used to construct a damage parameter prediction model based on the vibration feature sample library and using a preset hybrid intelligent modeling strategy. The hybrid intelligent modeling strategy includes using a swarm intelligence optimization strategy to optimize the parameters of the neural network model and training the candidate recognition model with the multi-order natural frequency features as input and the damage parameters as output. The constraint feature module is used to determine the target recognition model from the candidate recognition models based on preset engineering requirement constraints, and to collect the vibration signal of the laminate to be tested to extract multi-order natural frequency features consistent with the vibration feature sample library to obtain the vibration features to be tested. The target module is used to perform random noise processing and frequency offset correction on the vibration characteristics to be tested to obtain corrected vibration characteristics to be tested, and input the corrected vibration characteristics to be tested into the target recognition model to obtain the damage parameter recognition result of the laminate to be tested.

8. A device for identifying damage in composite laminates, characterized in that, The device includes: a memory, a processor, and a composite laminate damage identification program stored in the memory and executable on the processor, the composite laminate damage identification program being configured to implement the steps of the composite laminate damage identification method as described in any one of claims 1 to 6.

9. A storage medium, characterized in that, The storage medium stores a composite laminate damage identification program, which, when executed by a processor, implements the steps of the composite laminate damage identification method as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the composite laminate damage identification method as described in any one of claims 1 to 6.

Images