New energy material microdefect identification method and system based on computer vision

Through multimodal data fusion and adaptive learning methods, the problem that feature extraction modules in wind power blade defect detection is difficult to capture multimodal information, and the accurate identification and prediction of microscopic defects of wind power materials is achieved, which improves the accuracy of detection and adaptability.

CN119048698AActive Publication Date: 2024-11-29山东省科院易达信息科技有限公司

Patent Information

Application Number
CN202411546344.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-01
Publication Date
2024-11-29
Estimated Expiration
2044-11-01

AI Technical Summary

Technical Problem

In the detection of wind power blade defects, the feature extraction module is difficult to capture multimodal information. The detection model has low accuracy for small defects and a single data source, which limits the accuracy and adaptability of the detection.

Method used

Using multimodal data fusion and adaptive learning methods, the combined processing and dynamic optimization of multimodal features can be used to accurately identify and predict microscopic defects of wind power materials, including multimodal optical interference, photoacoustic imaging scanning, sparse representation Adaptive signal decoupling, multi-scale iterative inversion, spatiotemporal perturbation simulation and multi-level feedback optimization.

Benefits of technology

It improves the robustness and accuracy of defect detection, enhances the detection sensitivity and three-dimensional reconstruction accuracy of microscopic defects, and realizes all-round analysis and defect evolution prediction of wind power blade materials.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119048698A_ABST
    Figure CN119048698A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of computer vision and material science, in particular to a new energy material microdefect identification method and system based on computer vision, and the method comprises the steps: firstly, collecting data through a multi-mode optical interference and photoacoustic imaging technology, and carrying out feature alignment and fusion; self-adaptive signal decoupling is carried out on the preprocessed data, and a defect area is reconstructed through multi-scale iteration inversion and grid optimization; finally, model parameters are dynamically adjusted through transfer learning and multi-level feedback optimization, and the recognition precision and adaptability of the system are improved. The method can accurately identify the microdefects in the wind power material, and provides a scientific basis for the maintenance decision of equipment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the fields of computer vision and material science technology, and in particular to a method and system for identifying microscopic defects of new energy materials based on computer vision. Background Art

[0002] Wind power materials (such as epoxy resin and carbon fiber on wind turbine blades) are prone to microscopic defects such as cracks, sand holes, and delamination in actual use due to environmental factors (such as wind erosion, salt spray, lightning strikes, etc.). These defects will affect the structural integrity of wind turbine blades, leading to equipment performance degradation and even serious accidents. Therefore, effective defect detection and identification is of great significance to ensure the safe operation of wind power equipment.

[0003] The existing technology (Chinese invention patent, publication number: CN114565571A, name: A method and device for detecting defects in wind turbine blades based on computer vision) mainly uses deep learning models and non-destructive testing methods (such as infrared thermal imaging, acoustic emission monitoring, etc.) to detect defects in wind turbine blades. These methods usually process image data by constructing multi-scale feature fusion and deep convolutional networks. However, these methods have certain limitations: the feature extraction module is difficult to fully capture multimodal information; the detection model has low detection accuracy for small defects; and the data source is single (such as using only optical images), which limits the accuracy and adaptability of detection. Summary of the invention

[0004] In view of the many problems existing in the above-mentioned prior art, the present invention provides a method and system for identifying microscopic defects of new energy materials based on computer vision. The present invention is based on multimodal data fusion and adaptive learning methods, and realizes accurate identification and prediction of microscopic defects of wind power materials through joint processing and dynamic optimization of multimodal features. Its core principles include the collection, fusion, adaptive signal decoupling, three-dimensional defect reconstruction and multi-level feedback optimization of multimodal feature data, which improves the robustness and accuracy of detection.

[0005] A method for identifying microscopic defects of new energy materials based on computer vision comprises the following steps: Perform multimodal optical interference and photoacoustic imaging scanning on new energy materials, collect phase interference data, polarization light data and photoacoustic data, and generate multimodal response data through multimodal feature alignment and multimodal feature fusion, and then denoise and preprocess the multimodal response data to obtain preprocessed multimodal data; Performing sparse representation of adaptive signal decoupling on preprocessed multimodal data to generate abnormal signal data, and performing nonlinear enhancement processing and feature normalization processing on the abnormal signal data to obtain clustered abnormal signal data; The defect area of ​​clustered abnormal signal data is reconstructed through multi-scale iterative inversion, and adaptive grid optimization and dynamic regulation are performed in combination with the material physical model to generate the final three-dimensional reconstruction data; Perform spatiotemporal disturbance simulation on the final 3D reconstructed data, extract dynamic features, and perform multi-level feedback optimization based on correlation inversion to obtain defect evolution prediction data; Based on the defect evolution prediction data, combined with optical, acoustic and mechanical characteristics, through the multi-level feedback optimization algorithm and cross-domain adaptive learning of the neural network model, the weights and algorithm parameters of the neural network model are dynamically adjusted to achieve system update and iterative optimization.

[0006] Preferably, in the multimodal optical interference and photoacoustic imaging joint scanning step, by adjusting the phase angle and polarization angle, phase interference data and polarization light data are collected at different angles to cover the surface features and internal structural features of the new energy material; the multimodal feature alignment and multimodal feature fusion include aligning the phase interference data, polarization light data and photoacoustic data, so that the phase interference data, polarization light data and photoacoustic data are aligned in the same coordinate system, and the characteristic values ​​of the phase interference data, polarization light data and photoacoustic data are fused by a weighted average method to generate multimodal response data.

[0007] Preferably, in the adaptive signal decoupling step of the sparse representation, the preprocessed multimodal data is decomposed into normal signal data and abnormal signal data, wherein the sparse coefficients of the preprocessed multimodal data are solved using a sparse representation method, and the abnormal signal data is extracted by minimizing the reconstruction error, and the abnormal signal data is used for nonlinear enhancement processing.

[0008] Preferably, the multi-scale iterative inversion step comprises: According to the clustered abnormal signal data, a material physical model is constructed, and initial meshing is performed based on the stress and strain characteristics of the material to generate an initial defect area model; the meshing adopts an adaptive algorithm to make the mesh in the high stress concentration area finer by adjusting the mesh density; In each iteration, the inversion error is calculated and the regularization parameters in the inversion algorithm are dynamically adjusted to optimize the error term in the inversion process. The defect area model is gradually updated in an iterative manner to obtain more accurate three-dimensional reconstruction data.

[0009] Preferably, the regularization parameter The adjustment is based on the following expression: in, Indicates The regularization parameter in the iteration; Indicates The regularization parameter in the iteration; Represents the error change in the current iteration, calculated as the difference between the error of this iteration and the error of the previous iteration; Represents the baseline error value of the initial iteration, which is used to dynamically control the regularization strength during the inversion process.

[0010] Preferably, in the spatiotemporal disturbance simulation step, a variety of disturbance conditions are applied to the final three-dimensional reconstructed data, including successively adjusting the temperature field and stress field according to a preset disturbance scheme, and solving the spatiotemporal evolution equation of the material by numerical methods to extract dynamic characteristics under different disturbance conditions, and the dynamic characteristics are used to reflect the changes in the defect area.

[0011] Preferably, the correlation inversion processes the multi-scale spatiotemporal feature data by means of joint optimization, wherein the weight coefficient in the weighted convolution algorithm is Calculated according to the following expression: in, Indicates The weight coefficient of each feature layer; Indicates The feature importance coefficient of each feature layer; Represents the total number of all feature layers; Represents the weight normalization factor of all feature layers.

[0012] Preferably, in the cross-domain adaptive learning step of the multi-level feedback optimization algorithm and the neural network model, the weights and algorithm parameters of the neural network model are adjusted by transfer learning by combining the feedback results of the optical features, acoustic features and mechanical features, wherein the transfer learning process achieves dynamic adjustment by comparing the similarities between the optical features, acoustic features and mechanical features and updating the weights of the neural network model to optimize the overall performance of the system.

[0013] Preferably, in the transfer learning process, a cross-domain feature fusion model of optical features, acoustic features and mechanical features is constructed, the characteristic values ​​of the optical features, acoustic features and mechanical features are normalized, and the similarities between the optical features, acoustic features and mechanical features are calculated, and the weight update strategy of the neural network model is adjusted according to the similarities.

[0014] A system for implementing the method for identifying microscopic defects of new energy materials based on computer vision, the system comprising: Multimodal data acquisition module, used to perform multimodal optical interference and photoacoustic imaging scanning on new energy materials, and collect phase interference data, polarization data and photoacoustic data; A data preprocessing module, connected to the multimodal data acquisition module, is used to perform feature alignment and multimodal feature fusion on the multimodal data to generate multimodal response data, and to denoise and preprocess the multimodal response data to obtain preprocessed multimodal data; The signal decoupling and anomaly detection module is connected to the data preprocessing module and is used to perform adaptive signal decoupling of sparse representation on the preprocessed multimodal data, generate abnormal signal data, and perform nonlinear enhancement and feature normalization processing on the abnormal signal data to obtain clustered abnormal signal data; The multi-scale inversion and 3D reconstruction module is connected to the signal decoupling and anomaly detection module, and is used to perform multi-scale iterative inversion on the clustered anomaly signal data, and perform adaptive grid optimization and dynamic regulation in combination with the material physical model to generate the final 3D reconstruction data; The spatiotemporal evolution modeling module connects the multi-scale inversion and 3D reconstruction modules to simulate the spatiotemporal disturbance of the final 3D reconstruction data, extract dynamic features, and perform multi-level feedback optimization based on the correlation inversion to obtain defect evolution prediction data; The feedback optimization and adaptive learning module is connected to the spatiotemporal evolution modeling module. It is used to perform cross-domain adaptive learning of multi-level feedback optimization algorithms and neural network models based on defect evolution prediction data, combined with optical, acoustic and mechanical characteristics, and dynamically adjust the weights and algorithm parameters of the neural network model to achieve system update and iterative optimization.

[0015] Compared with the prior art, the advantages and beneficial effects of the present invention are: The present invention collects phase, polarization and acoustic data through multimodal optical interference and photoacoustic imaging, and combines multimodal feature alignment and fusion technology to achieve higher-precision defect detection.

[0016] The present invention decouples preprocessed data through sparse representation, effectively separates abnormal signals, and enhances the detection sensitivity of defects.

[0017] The present invention adopts a mesh optimization technology based on a material physical model to improve the three-dimensional reconstruction accuracy of the defect area.

[0018] The present invention uses transfer learning to dynamically adjust model parameters and combines feedback from multimodal data to optimize recognition performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of the process of the present invention; Figure 2 Schematic diagram of the multimodal data fusion process in the present invention; Figure 3 It is a schematic diagram of the multi-scale iterative inversion and adaptive optimization process in the present invention; Figure 4A flow chart of dynamic feedback optimization and transfer learning in the present invention; Figure 5 This is a system structure block diagram of the present invention. DETAILED DESCRIPTION

[0020] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is obvious that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0021] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0022] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0023] like Figure 1 As shown, a method for identifying microscopic defects of new energy materials based on computer vision includes the following steps: Perform multimodal optical interference and photoacoustic imaging scanning on new energy materials, collect phase interference data, polarization light data and photoacoustic data, and generate multimodal response data through multimodal feature alignment and multimodal feature fusion, and then denoise and preprocess the multimodal response data to obtain preprocessed multimodal data; Preferably, in the multimodal optical interference and photoacoustic imaging joint scanning step, by adjusting the phase angle and polarization angle, phase interference data and polarization light data are collected at different angles to cover the surface features and internal structural features of the new energy material; the multimodal feature alignment and multimodal feature fusion include aligning the phase interference data, polarization light data and photoacoustic data, so that the phase interference data, polarization light data and photoacoustic data are aligned in the same coordinate system, and the characteristic values ​​of the phase interference data, polarization light data and photoacoustic data are fused by a weighted average method to generate multimodal response data.

[0024] like Figure 2As shown, in the present invention, new energy materials (mainly wind turbine blade materials, such as epoxy resin and carbon fiber) are first detected through joint scanning of multimodal optical interference and photoacoustic imaging to achieve efficient identification and analysis of material microscopic defects.

[0025] This method takes advantage of the combination of multimodal optical interference and photoacoustic imaging technology to collect detailed information on the surface and interior of wind power materials. Optical interference technology can provide high-resolution microscopic features of the material surface by analyzing the phase difference between the incident light and the reflected light; while photoacoustic imaging stimulates ultrasonic waves on the surface of the material through laser irradiation, and obtains the structural characteristics of the material based on the time and intensity changes of the ultrasonic waves propagating inside the material. Therefore, these two imaging methods complement each other and can simultaneously detect defects on the surface and inside of the material.

[0026] In specific operations, the system will scan wind power materials at multiple angles, and collect phase interference data and polarization light data at different angles by adjusting the phase angle and polarization angle of optical interference. This multi-angle scanning method can cover the surface microscopic features of the material (such as scratches and cracks on the surface) and the structural features inside the material (such as internal gaps and delamination). In this way, the comprehensiveness and accuracy of the detection can be significantly improved. In addition, photoacoustic imaging supplements the internal information that cannot be detected by optical interference, making the scanning results more complete.

[0027] The collected data include phase interference data, polarization data and photoacoustic data. In order to ensure the accuracy of multimodal data, these data need to be subjected to multimodal feature alignment and multimodal feature fusion processing. Specifically, multimodal feature alignment refers to mapping the data of each modality to the same coordinate system so that different modal data have a consistent reference standard in space. This step is achieved through registration technology, which aligns the phase interference data, polarization data and photoacoustic data to the same coordinate system to ensure the correlation between the data in subsequent processing.

[0028] The aligned multimodal data needs to be fused. The purpose of multimodal feature fusion is to integrate the useful information from data of different modes into a whole to generate more accurate multimodal response data. In the present invention, the characteristic values ​​of phase interference data, polarization light data and photoacoustic data are weighted fused by weighted averaging. The weight setting of weighted fusion is adjusted according to the data quality and modal importance. For example, in the inspection of wind turbine blade materials, photoacoustic data may be more important for the description of internal defects of the material. Therefore, the weight of photoacoustic data can be appropriately increased during weighting to improve the sensitivity of internal defect detection of the material. The multimodal response data generated after fusion contains a comprehensive description of the surface and internal characteristics of the material, providing more comprehensive data support for subsequent defect analysis.

[0029] In order to further improve the quality of multimodal response data, the fused multimodal response data needs to be denoised and preprocessed. Denoising mainly uses filters to eliminate noise interference that may occur during the acquisition process, such as the impact of external environment vibration or measurement equipment errors on the data; while preprocessing includes normalization, scale adjustment and other operations to make the data more suitable for subsequent signal processing and defect analysis steps in terms of numerical range and distribution.

[0030] This step realizes the comprehensive detection of wind turbine blade materials and the comprehensive processing of multimodal data. It can not only identify tiny cracks and defects on the surface of the material, but also detect structural defects inside the material (such as voids, delamination or debonding). The multi-angle scanning method and multimodal feature fusion significantly improve the resolution and accuracy of the detection, allowing the present invention to effectively deal with complex defect situations. In addition, through the alignment and fusion of multimodal data, a full-scale analysis of the microstructure of the material can be achieved, especially for the composite materials of wind turbine blades, which can accurately distinguish the defect distribution and characteristics between different material layers.

[0031] For example, if there is an incompletely bonded area between the epoxy resin layer and the carbon fiber layer of a wind turbine blade, it may appear as a slight amplitude change on the surface in the phase interference data, while photoacoustic imaging may detect local abnormal sound wave propagation speed. Through the multimodal feature fusion processing in the present invention, these two abnormal features can be considered and combined simultaneously to accurately locate the location and range of the defect and evaluate its impact. Finally, through data denoising and preprocessing, the reliability and accuracy of the defect detection results are further ensured.

[0032] Performing sparse representation of adaptive signal decoupling on preprocessed multimodal data to generate abnormal signal data, and performing nonlinear enhancement processing and feature normalization processing on the abnormal signal data to obtain clustered abnormal signal data; Preferably, in the adaptive signal decoupling step of the sparse representation, the preprocessed multimodal data is decomposed into normal signal data and abnormal signal data, wherein the sparse coefficients of the preprocessed multimodal data are solved using a sparse representation method, and the abnormal signal data is extracted by minimizing the reconstruction error, and the abnormal signal data is used for nonlinear enhancement processing.

[0033] In the present invention, the pre-processed multimodal data is subjected to sparse representation adaptive signal decoupling to extract and identify abnormal signals in the material, thereby realizing the detection and analysis of possible microscopic defects in wind turbine blade materials (such as epoxy resin and carbon fiber layers). This step uses sparse representation technology to process signal data, thereby highlighting abnormal features in massive data and improving the accuracy and reliability of defect identification.

[0034] The preprocessed multimodal data is adaptively decoupled by sparse representation. Sparse representation is a signal processing technology, the core idea of ​​which is to represent the signal as a linear combination of a few elements in a dictionary, so that most irrelevant or redundant information is ignored, thereby retaining only important feature information. In the present invention, the goal of sparse representation is to decompose the preprocessed multimodal data into normal signal data and abnormal signal data. Normal signal data reflects the background information of the material and the characteristics of the defect-free area, while abnormal signal data represents the potential defect area in the material.

[0035] Specifically, the adaptive signal decoupling step of sparse representation first uses sparse coding method to perform sparse processing on the preprocessed multimodal data. The sparse processing process can be expressed as solving the sparse coefficients to minimize the reconstruction error of the signal. The mathematical expression of this step is usually: in, Represents the preset dictionary matrix, represents a sparse coefficient vector, is the preprocessed multimodal data, is a balance parameter used to control sparsity and reconstruction error. By solving the above optimization problem, we can get the sparse coefficient vector , and then decompose the abnormal signal data and normal signal data. The abnormal signal data usually shows a significant increase in the sparse coefficient at certain locations, which indicates that these signal features may correspond to potential defects in the material.

[0036] In order to further highlight the characteristics of abnormal signal data, the present invention introduces nonlinear enhancement processing. The purpose of nonlinear enhancement is to amplify the weak features in the abnormal signal data so that they can be more easily identified and separated in subsequent analysis. This step usually uses logarithmic transformation or exponential enhancement methods to stretch the amplitude of the signal to improve the contrast. For example, assuming that the detected abnormal signal data has a weak response in certain frequency bands, logarithmic enhancement can be applied to the signals in these frequency bands so that smaller signal changes appear more significant in the enhanced signal. This is particularly important for detecting tiny cracks or delaminations in wind turbine blades, because these defects may appear as very low amplitude signals in the original data.

[0037] After completing the nonlinear enhancement processing, the abnormal signal data is subjected to feature normalization. The purpose of feature normalization is to map data of different modes and scales to a unified numerical range to eliminate deviations caused by different signal amplitudes or units, and to ensure that subsequent clustering analysis can be performed based on the same standard. Normalization can be achieved through linear normalization (such as minimum-maximum normalization) or standardization (such as zero mean normalization). The normalized abnormal signal data will be used as clustered abnormal signal data for further defect area clustering and analysis.

[0038] Adaptive signal decoupling through sparse representation can effectively separate the normal and abnormal features of wind turbine blade materials, significantly improving the ability to detect microscopic defects. After nonlinear enhancement and feature normalization processing, the abnormal signal data can significantly enhance the identifiability of tiny defects, making the present invention more sensitive when detecting microcracks, delamination or other hidden defects. For example, if there is an area between the epoxy resin layer and the carbon fiber layer of a wind turbine blade that is not fully bonded, the sparse representation can separate the abnormal signal data in the area, and the nonlinear enhancement processing will amplify the signal amplitude. After normalization processing, it can ensure that the signal has significant distinctive features in subsequent analysis.

[0039] This method has important application value in the daily inspection and maintenance of wind turbine blades. By timely detecting and identifying microscopic defects, the expansion of defects can be prevented and the service life of the blades can be extended. In addition, the adaptability of sparse representation technology also enables the present invention to be flexibly adjusted according to different material properties and detection environments, thereby improving the versatility and adaptability of detection.

[0040] The defect area of ​​clustered abnormal signal data is reconstructed through multi-scale iterative inversion, and adaptive grid optimization and dynamic regulation are performed in combination with the material physical model to generate the final three-dimensional reconstruction data; Preferably, Figure 3 As shown, the multi-scale iterative inversion step includes: According to the clustered abnormal signal data, a material physical model is constructed, and initial meshing is performed based on the stress and strain characteristics of the material to generate an initial defect area model; the meshing adopts an adaptive algorithm to make the mesh in the high stress concentration area finer by adjusting the mesh density; In each iteration, the inversion error is calculated and the regularization parameters in the inversion algorithm are dynamically adjusted to optimize the error term in the inversion process. The defect area model is gradually updated in an iterative manner to obtain more accurate three-dimensional reconstruction data.

[0041] In the present invention, clustered abnormal signal data is used to perform three-dimensional reconstruction of defect areas through a multi-scale iterative inversion method, and adaptive grid optimization and dynamic regulation are performed in combination with the physical model of the material to generate high-precision three-dimensional reconstruction data. This process uses clustered abnormal signal data to reflect the defect area inside the material, continuously improves the inversion accuracy through iterative optimization, and performs grid division and optimization in combination with the stress and strain characteristics of the material, ultimately forming an accurate reconstruction of the microscopic defects of the wind turbine blade.

[0042] Multiscale iterative inversion is a method that gradually updates and optimizes the target model through multiple iterations, and is particularly suitable for material defect reconstruction problems. In the present invention, the defect area of ​​clustered abnormal signal data is reconstructed through multiscale iterative inversion, and a corresponding physical model is established in combination with the physical properties of the material to guide and optimize the inversion process.

[0043] Specifically, the initial physical model of the material is first constructed based on the clustered abnormal signal data. The model uses the stress-strain characteristics of the wind turbine blade material for initial meshing, that is, according to the stress concentration degree at different locations in the material, the density of the mesh is automatically adjusted to make the mesh in the high stress concentration area more refined. This ensures that the inversion algorithm has a higher resolution in key areas, thereby improving the accuracy of defect recognition.

[0044] Next, multi-scale iterative inversion is performed. In each iteration, the inversion algorithm calculates the inversion error based on the clustered abnormal signal data and the current defect area model, and optimizes the error term by dynamically adjusting the regularization parameter. The regularization parameter is used to balance the relationship between data fit and model complexity, ensuring that the inversion result not only conforms to the known data, but also has good smoothness and robustness. During the inversion process, dynamic regulation of the regularization parameter can further optimize the accuracy and stability of the inversion result, so that the defect area model gradually approaches the actual defect morphology after each iteration.

[0045] In the specific implementation, the following process can be adopted: 1. Initial meshing: Based on the material properties of wind turbine blades, such as the composite structure of epoxy resin and carbon fiber, the initial meshing will be differentiated according to the stress-strain behavior of these different material layers. For example, in the carbon fiber layer on the surface of the blade, the high stress area may require a higher mesh density to capture tiny cracks or delamination, while in the epoxy resin layer, a relatively coarse mesh can be used.

[0046] 2. Multi-scale inversion iteration: The inversion algorithm is performed at different scales, starting with a coarser grid and gradually refining the grid. During each iteration, the current inversion error is calculated, and the regularization parameter is dynamically adjusted according to the change in the error.

[0047] 3. Adaptive mesh optimization: During the inversion process, the mesh density will be dynamically adjusted as the inversion progresses. For example, for surface cracks on wind turbine blades, as the inversion accuracy improves, the local mesh can be further refined to capture more details, while for larger-scale delamination defects, a coarser mesh can be maintained to save computing resources.

[0048] Through multi-scale iterative inversion and adaptive mesh optimization, the high-precision three-dimensional morphology of the defective area can be reconstructed in a short computing time. This method has a significant effect on the identification of microscopic defects in wind turbine blade materials, especially in the multi-layer structure of composite materials, the specific location and range of defects can be accurately located. For example, if there are fine delamination or debonding areas in the carbon fiber layer under the epoxy resin layer, this method can accurately reconstruct the defective area and provide clear three-dimensional reconstruction data. This is of great significance for evaluating the structural integrity of wind turbine blades and performing preventive maintenance.

[0049] Preferably, the regularization parameter The adjustment is based on the following expression: in, Indicates The regularization parameter in the iteration; Indicates The regularization parameter in the iteration; Represents the error change in the current iteration, calculated as the difference between the error of this iteration and the error of the previous iteration; Represents the baseline error value of the initial iteration, which is used to dynamically control the regularization strength during the inversion process.

[0050] The core of regularization parameter adjustment is to dynamically control the regularization strength of the inversion through the change of error. Regularization is used to prevent overfitting in the inversion problem and improve the stability of the result by smoothing the inversion solution.

[0051] During each iteration, the system calculates the inversion error, which is the difference between the model prediction and the actual observed value. If the error of the current iteration changes significantly compared to the previous iteration (i.e. Larger), indicating that the model's fitting effect on the data has changed significantly. In this case, increasing the regularization strength (increasing ) to prevent the risk of overfitting; on the contrary, if the error change is small, the regularization strength can be appropriately weakened to enhance the model's ability to fit details. In this way, the regularization parameter The dynamic adjustment of achieves a balance between fitting accuracy and model smoothness during the inversion process.

[0052] In the present invention, the dynamic regulation of the regularization parameter is mainly used in the process of multi-scale iterative inversion to achieve the reconstruction of microscopic defects of wind turbine blade materials (such as epoxy resin and carbon fiber). The specific operation process is as follows: 1. Initial meshing and inversion initialization: Before the multi-scale iterative inversion begins, the initial meshing is performed based on the material physical model. For wind turbine blade materials, the initial meshing is performed based on the stress-strain characteristics of different material layers (such as carbon fiber layers and epoxy resin layers) to ensure a higher mesh density in high stress concentration areas to improve the inversion accuracy. Initial regularization parameters is set to a moderate value to balance the quality of the initial inversion solution.

[0053] 2. Error calculation and parameter update: In each inversion iteration, the system calculates the inversion error of the current iteration and compares it with the error of the previous iteration to obtain .based on and the initial error value , adjust the value of the regularization parameter according to the above formula. This update process ensures that the inversion algorithm can dynamically adapt to different error changes, thereby optimizing the accuracy and stability of the inversion solution. For example, if the inversion error of the carbon fiber layer suddenly increases in a certain iteration, it means that the current model cannot fit the defects in this area well. At this time, the system will increase the regularization strength to suppress the drastic fluctuations of the inversion solution, thereby ensuring the smoothness of the reconstruction.

[0054] 3. Gradually refined multi-scale inversion: As iterations proceed, the inversion algorithm gradually refines the grid based on the coarser grid, and performs more detailed defect reconstruction for local areas. This multi-scale strategy helps improve the algorithm's computational efficiency while ensuring global and local reconstruction accuracy. By dynamically adjusting the regularization parameters, the inversion process can maintain good fitting quality and solution stability at different scales.

[0055] Termination conditions and 3D reconstruction result generation: When the iterative process meets the preset termination conditions (such as the error change is less than a certain threshold or the number of iterations reaches the upper limit), the inversion process stops and the final 3D reconstruction data is generated. The reconstructed data at this time can accurately reflect the defect area and its distribution characteristics inside the wind turbine blade.

[0056] By dynamically adjusting the regularization parameters, the inversion algorithm can adaptively control the inversion strength at different iteration stages to balance the fitting accuracy and model smoothness. This method enables the present invention to show higher flexibility and adaptability when detecting microscopic defects in wind turbine blades. For example, in the defect identification at the junction of the carbon fiber layer and the epoxy resin layer, since such interface areas usually have complex stress concentration phenomena, the regularization strength needs to be frequently adjusted during the inversion process to capture detailed changes. The dynamically controlled regularization parameters enable the system to better adapt to these changes, and the generated three-dimensional reconstruction data is more accurate and delicate.

[0057] Perform spatiotemporal disturbance simulation on the final 3D reconstructed data, extract dynamic features, and perform multi-level feedback optimization based on correlation inversion to obtain defect evolution prediction data; In the present invention, dynamic features are extracted by performing spatiotemporal disturbance simulation on the final 3D reconstruction data, and multi-level feedback optimization is performed based on the correlation inversion technology to predict the evolution trend of defects and generate defect evolution prediction data. This process uses 3D reconstruction data to simulate the response characteristics of materials under different external conditions, thereby deeply analyzing and predicting the changes of material microscopic defects under different time and space conditions, providing a reliable basis for preventive maintenance of defects.

[0058] Spatiotemporal perturbation simulation is a technique that simulates material response by applying different temporal and spatial perturbation conditions. This method is used in the present invention to analyze the dynamic behavior of wind turbine blade materials (such as epoxy resin and carbon fiber layers) under various external conditions, thereby capturing the evolution characteristics of their defect areas. Specifically, spatiotemporal perturbation conditions can include changes in factors such as temperature changes, stress loading and vibration, which will affect the microscopic defects inside the material, causing their position, size and shape to change over time.

[0059] In the specific implementation process, firstly, spatiotemporal disturbance conditions are applied to the three-dimensional reconstructed data. Taking wind turbine blades as an example, when performing spatiotemporal disturbance simulation, different temperature changes (such as temperature difference between day and night) and stress loading (such as wind load on blades) can be applied to the reconstructed defective area. These disturbance conditions are applied to the three-dimensional model through numerical simulation methods to generate corresponding material response data. The response data usually includes changes in stress field, strain field and other physical quantities inside the material, which reflect the dynamic evolution trend of defects inside the material to a certain extent.

[0060] Next, extract the dynamic features under time-space disturbance. The extraction of dynamic features is achieved by analyzing the change pattern of material response data under time-space disturbance conditions. The specific implementation method can use feature decomposition techniques, such as principal component analysis (PCA) or Fourier analysis in the time-space domain, to extract feature vectors representing the dynamic behavior of the material. For example, for small cracks in the carbon fiber layer, by analyzing the crack expansion trend under temperature changes and stress loading conditions, the crack expansion rate, direction, and crack shape change characteristics can be extracted. These dynamic features can be used for subsequent defect evolution prediction.

[0061] After extracting the dynamic features, the present invention performs multi-level feedback optimization based on correlation inversion. Correlation inversion is a method that combines multiple physical and mathematical models to perform reverse reasoning and analysis on complex data. The multi-level feedback optimization here includes multiple feedback loops, which are optimized for different dynamic characteristics of the material to improve the accuracy of defect evolution prediction. During the feedback optimization process, the system will repeatedly adjust parameters (such as the elastic modulus, thermal conductivity, etc. of the material) to make the simulation results consistent with the actual observed dynamic characteristics, thereby correcting the errors of the prediction model and ultimately generating more accurate defect evolution prediction data.

[0062] Preferably, in the spatiotemporal disturbance simulation step, a variety of disturbance conditions are applied to the final three-dimensional reconstructed data, including successively adjusting the temperature field and stress field according to a preset disturbance scheme, and solving the spatiotemporal evolution equation of the material by numerical methods to extract dynamic characteristics under different disturbance conditions, and the dynamic characteristics are used to reflect the changes in the defect area.

[0063] In the present invention, multiple disturbance conditions are applied to the final 3D reconstruction data through spatiotemporal disturbance simulation to extract dynamic features and reflect the changes in the defect area. This process is intended to simulate the response behavior of materials under different external environmental conditions, so as to predict and analyze the evolution trend of material defects. For the complex structure of wind turbine blade materials (such as epoxy resin and carbon fiber layers), by adjusting disturbance conditions such as temperature field and stress field, we can have a more comprehensive understanding of the material degradation process and defect expansion under actual working conditions.

[0064] The core of spatiotemporal perturbation simulation is to simulate the response characteristics of materials by applying different external conditions (perturbation conditions), and to solve the spatiotemporal evolution equations of materials by numerical methods, thereby extracting dynamic characteristics. Perturbation conditions can include temperature changes, mechanical stress, vibration, and humidity changes, etc. These factors will cause changes in the internal stress field, strain field, and temperature field of the material, thereby affecting the microscopic defect behavior of the material. In the present invention, the implementation steps of spatiotemporal perturbation simulation are as follows: 1. Apply multiple disturbance conditions: Apply different disturbance conditions to the 3D reconstructed data, such as changes in temperature field and stress field. For the material of wind turbine blades, different disturbance conditions such as day and night temperature changes, wind load changes, and humidity changes in the marine environment can be considered separately. Specifically, temperature field disturbance can simulate the thermal expansion effect of blades under high and low temperature conditions by setting different temperature distribution models, such as linear temperature rise model or periodic temperature fluctuation model; while stress field disturbance can simulate the stress condition of blades under different wind conditions and study the influence of stress concentration on defect expansion.

[0065] 2. Numerical solution of time-space evolution equations: After applying disturbance conditions, it is necessary to establish the time-space evolution equations of the material to describe the response behavior of the material under disturbance conditions. These equations usually include heat conduction equations, stress-strain equations, and momentum balance equations. By solving these equations through the finite element method (FEM) or other numerical solution techniques, the time and space response data of the material under different disturbance conditions can be obtained. Taking epoxy resin and carbon fiber layers as an example, the heat conduction equation can be used to describe the changes in the temperature field and simulate the thermal expansion and contraction of the material in temperature fluctuations; the stress-strain equation can be used to describe the changes in the stress concentration area when the blade is subjected to wind load, helping to analyze the expansion trend of the crack.

[0066] 3. Extract dynamic features: After obtaining the response data of the material, analyze these data to extract dynamic features. Dynamic features include the rate of change, direction, and range of the defect area, which are used to reflect the evolution trend of the defect under different disturbance conditions. The method of extracting dynamic features can use time-frequency domain analysis techniques, such as short-time Fourier transform (STFT) or wavelet transform, to analyze the change pattern of the defect area on different time scales. For example, in the process of crack propagation in the carbon fiber layer, if the crack propagation rate under high stress conditions is found to increase significantly through space-time disturbance simulation, the risk of structural strength reduction in the area can be predicted.

[0067] Through spatiotemporal perturbation simulation, the present invention can effectively evaluate the defect evolution behavior of wind turbine blade materials under various environmental conditions, thereby providing important technical support for preventive maintenance and structural health monitoring of materials. For epoxy resin and carbon fiber composite material layers, spatiotemporal perturbation simulation can reveal the combined effects of temperature changes and wind loads on material degradation. For example, in the simulation, it was found that with the periodic fluctuations in temperature, the stress concentration area of ​​the carbon fiber layer gradually expanded and peeled off from the interface with the epoxy resin layer. By extracting dynamic feature data, the aggravation of interface peeling can be further predicted, and the time of blade failure can be estimated.

[0068] Preferably, the correlation inversion processes the multi-scale spatiotemporal feature data by means of joint optimization, wherein the weight coefficients in the weighted convolution algorithm are Calculated according to the following expression: in, Indicates The weight coefficient of each feature layer; Indicates The feature importance coefficient of each feature layer; Represents the total number of all feature layers; Represents the weight normalization factor of all feature layers.

[0069] In the present invention, multi-scale spatiotemporal feature data are processed by correlation inversion to optimize the accuracy and robustness of defect detection. Correlation inversion is a reverse reasoning technology that combines multiple physical and mathematical models to infer the specific characteristics of defects based on the response data of the material. In this process, a weighted convolution algorithm is used to jointly optimize multi-scale features, and the weight coefficient of the weighted convolution is calculated based on the importance of the feature to ensure that the contribution of different feature layers to the final result is reasonably distributed.

[0070] The core of correlation inversion is to process multi-scale spatiotemporal feature data to accurately reflect the microscopic defect distribution and evolution trend of wind turbine blade materials. Multi-scale spatiotemporal feature data includes feature information extracted from different scales and time resolutions, such as temperature changes, stress concentration, and crack growth rate. In order to comprehensively analyze these features, a weighted convolution algorithm is used.

[0071] In practice, the feature importance coefficient The determination can be made by statistical analysis based on the performance of the feature in multiple experiments. For example, in the composite structure of a wind turbine blade, if it is found that the temperature variation data at a certain scale has a greater impact on the accuracy of crack propagation prediction, then the feature importance coefficient of the feature layer is It will be correspondingly higher, thereby increasing the weight in the weighted convolution.

[0072] First, different feature layers are extracted from multi-scale spatiotemporal data, which may include different time resolutions, spatial resolutions, and changes in physical fields (such as temperature fields and stress fields). For the detection of wind turbine blade materials, feature data can be extracted from multiple angles, such as stress concentration under high temperature conditions, material shrinkage behavior under low temperature, and dynamic strain changes under wind loads. The extracted feature layer data is used to calculate the feature importance coefficient according to its importance. , and then calculate the weight coefficient of each feature layer according to the above expression .

[0073] In obtaining the weight coefficient After that, all feature layer data are weighted convolution processed, that is, the data of each feature layer are weighted and summed according to the corresponding weight. In this way, the influence of different feature layers on the final reconstruction result will be adjusted according to the size of their weights. For example, in wind turbine blade materials, if the strain change of the carbon fiber layer under high stress is more important for predicting crack propagation, then the weight of the corresponding feature layer in the weighted convolution will be relatively high, making the final correlation inversion result more dependent on the feature layer.

[0074] The characteristic data after weighted convolution is used for correlation inversion analysis, and multi-level joint optimization is performed on this basis. The optimization process includes repeatedly adjusting the model parameters to make the inversion results more consistent with the experimental data. This multi-level feedback optimization can continuously improve the prediction accuracy by analyzing the dynamic characteristics of defect evolution, and ultimately generate detailed three-dimensional reconstruction and evolution trend data of the defects.

[0075] Through the joint optimization of correlation inversion and weighted convolution, the present invention can more accurately reconstruct the defect distribution in wind turbine blade materials and predict its evolution trend. The use of weighted convolution algorithm can fully utilize the advantages of multi-scale feature data, and the complementary effects between different feature layers are effectively exerted. For example, in the interface area of ​​epoxy resin and carbon fiber, if stress concentration under different temperature conditions has a significant effect on the expansion of cracks, the weight calculation method of weighted convolution can automatically give higher weights to these important feature layers, thereby giving priority to the contribution of these features in the correlation inversion process. This method can better capture the various dynamic behaviors of materials in complex environments, significantly improving the accuracy and robustness of defect identification.

[0076] Based on the defect evolution prediction data, combined with optical, acoustic and mechanical characteristics, through the multi-level feedback optimization algorithm and cross-domain adaptive learning of the neural network model, the weights and algorithm parameters of the neural network model are dynamically adjusted to achieve system update and iterative optimization.

[0077] Preferably, Figure 4 As shown, in the cross-domain adaptive learning step of the multi-level feedback optimization algorithm and the neural network model, the weights and algorithm parameters of the neural network model are adjusted by transfer learning by combining the feedback results of the optical features, acoustic features and mechanical features, wherein the transfer learning process achieves dynamic adjustment by comparing the similarities between the optical features, acoustic features and mechanical features and updating the weights of the neural network model to optimize the overall performance of the system.

[0078] The multi-level feedback optimization algorithm combines the feedback information of multiple modes to continuously iterate and adjust the neural network model so that it has higher adaptability to data of different characteristic modes. In the present invention, the optical characteristics, acoustic characteristics and mechanical characteristics come from visual detection, ultrasonic detection and stress-strain analysis of the material surface respectively. These characteristic information can respectively reflect the surface morphology, internal structural state and mechanical response behavior of the material. On the basis of multimodal data integration, by comparing the similarity between different modal features, the transfer learning process is used to update the weights of the neural network model so that the model can better capture the correlation between these features.

[0079] During the transfer learning process, the system first calculates the similarity between optical features, acoustic features, and mechanical features. The similarity can be calculated using methods such as the Euclidean distance or cosine similarity between feature vectors to measure the degree of similarity between different modalities. For example, when an abnormal reflected light intensity is found in a certain area during optical inspection and the ultrasonic reflection of the acoustic feature changes significantly, the feature similarity between the two modalities will be high, indicating that they may reflect the presence of the same defect. Based on this similarity calculation result, transfer learning will adjust the weights of the neural network model so that the model can more accurately integrate information from different modalities.

[0080] Through a multi-level feedback optimization algorithm, the characteristic feedback results of different modes are used as input to continuously iterate and adjust the weights and algorithm parameters of the neural network. In specific implementation, when the optical features change frequently and are consistent with the strain changes of the mechanical features, the weights of these two modal features in the model can be increased, thereby enhancing the model's sensitivity to these features. This weight update strategy can be implemented through the error backpropagation technology. After each update, the system will recalculate the model's error to determine whether the performance of the adjusted model on the new data set has improved.

[0081] Cross-domain adaptive learning enables neural network models to transfer knowledge between different modal fields (such as optics, acoustics, and mechanics). The core of this process is to use a pre-trained neural network model (a model trained on a certain modal feature) to fine-tune the data of other modalities so that the model can adapt to the feature distribution of the new modality. For example, assuming that a detection model has been trained on optical feature data, the model can be fine-tuned using acoustic feature data, and its adaptability to ultrasonic detection data can be enhanced through an appropriate transfer learning process, ultimately achieving accurate identification of potential defects in the carbon fiber layer.

[0082] Through multi-level feedback optimization and transfer learning, the present invention can significantly improve the defect detection and identification accuracy of wind turbine blade materials. The similarity calculation between multimodal features enables the model to effectively utilize data from different sensors and improve the multi-dimensional information fusion capability of composite materials. For example, in the carbon fiber layer below the epoxy resin layer, if anomalies in the optical and acoustic features are detected and the change trends of the two are similar, the system can strengthen the reconstruction and identification of this area through transfer learning, making the final defect prediction results more accurate. In addition, through cross-domain adaptive learning, the system can realize flexible knowledge transfer and parameter adjustment between different modal data, further improving the ability to cope with complex defect situations.

[0083] Preferably, in the transfer learning process, a cross-domain feature fusion model of optical features, acoustic features and mechanical features is constructed, the characteristic values ​​of the optical features, acoustic features and mechanical features are normalized, and the similarities between the optical features, acoustic features and mechanical features are calculated, and the weight update strategy of the neural network model is adjusted according to the similarities.

[0084] In the present invention, transfer learning improves the adaptability of the neural network model to multimodal data by constructing a cross-domain feature fusion model of optical features, acoustic features, and mechanical features. The cross-domain feature fusion model is used to normalize and calculate the similarity of features under different physical modes, so that the weight update strategy of the neural network model can be adjusted according to the similarity during the transfer learning process, thereby optimizing the overall performance of the system. This process is of great significance for defect detection of wind power materials (such as epoxy resin and carbon fiber composite materials) because it can better utilize the complementarity of multimodal data and enhance the accurate identification of defects.

[0085] The core of transfer learning is to process feature data from different modalities through feature fusion models, so that the neural network model can adapt to the complex distribution of multimodal data. In the present invention, optical features, acoustic features and mechanical features correspond to the visual features, ultrasonic detection features and stress-strain behavior features of the material, respectively. In order to make full use of these feature information, it is necessary to first normalize the feature values ​​to make the data scales of each modality consistent, so as to perform effective fusion and similarity calculation.

[0086] The purpose of eigenvalue normalization is to eliminate the differences in the numerical ranges between different modal features so that they can be compared on the same scale. The specific method can be standard normalization (subtracting the mean from the eigenvalue and dividing it by the standard deviation) or minimum-maximum normalization (mapping the eigenvalue to the interval of 0 to 1). For example, for the surface reflected light intensity obtained by optical detection, there may be a large range of fluctuations, while the ultrasonic reflection intensity obtained by acoustic detection may be concentrated in a smaller numerical range. After normalization, these feature data can be compared and fused on the same scale.

[0087] The cross-domain feature fusion model is constructed to integrate the features of different modes into a unified representation space. In the model, the normalized data of each modal feature is used as input, and the feature extraction layer (such as convolution layer or fully connected layer) generates further feature representations for these data. The feature fusion layer can use weighted averaging, attention mechanism or deep feature fusion technology to integrate the feature information of different modes. For example, in the crack detection of wind turbine blades, the feature vectors of optical features (surface reflected light intensity), acoustic features (ultrasonic signals) and mechanical features (stress distribution) can be input into the fusion model. The model will integrate these feature information to obtain a cross-domain comprehensive feature representation for subsequent analysis.

[0088] In the output layer of the feature fusion model, the process of transfer learning is guided by calculating the similarity between optical features, acoustic features, and mechanical features. The similarity can be calculated using Euclidean distance, cosine similarity, or other metrics to measure the similarity between different modal features. For features with high similarity, the system will assume that these modes may reflect the same physical phenomenon or defect characteristics, thereby appropriately increasing the weight of the relevant features during transfer learning. For example, if similar change patterns of optical and acoustic features are detected in the interface area between the carbon fiber layer and the epoxy resin layer of a wind turbine blade, the weights of these two modes can be increased to make the model pay more attention to the multimodal feature fusion results in this area.

[0089] Based on the results of similarity calculation, transfer learning optimizes the model training process by dynamically adjusting the weight update strategy of the neural network. Weight update can be achieved by using the gradient descent method combined with error back propagation. For each training iteration, the learning rate or weight adjustment amplitude is adjusted according to the similarity, so that features with high similarity contribute more to the model training, while features with low similarity have less impact.

[0090] By constructing a cross-domain feature fusion model and combining the similarity adjustment weight update strategy in transfer learning, the present invention can significantly improve the accuracy and robustness of micro defect detection of wind turbine blade materials. Normalization processing ensures that different modal data are compared on the same scale, effectively eliminating the impact of differences in numerical ranges. The construction of the feature fusion model enables information from different physical modes to be comprehensively analyzed in a unified feature space, thereby capturing more correlation information between multimodal data.

[0091] For example, in the carbon fiber layer of a wind turbine blade, the expansion of a crack may simultaneously cause a change in the surface optical reflection intensity and the attenuation of the ultrasonic signal. The feature fusion model of the present invention can capture these changes at the same time and combine the similarity between the two to enhance the detection capability of the crack area. In this way, the system can not only identify existing defects, but also predict the potential defect expansion trend, thereby providing a more scientific basis for the maintenance of wind turbine blades.

[0092] like Figure 5 As shown, a system for implementing the method for identifying microscopic defects of new energy materials based on computer vision, the system comprises: The multimodal data acquisition module is used to perform multimodal optical interference and photoacoustic imaging scanning on new energy materials, and collect phase interference data, polarization data, and photoacoustic data; multimodal optical interference and photoacoustic imaging technology are used to scan wind power materials (such as epoxy resin and carbon fiber composite materials) and collect phase interference data, polarization data, and photoacoustic data. These data can reflect the different physical properties of the materials and help obtain more comprehensive defect information.

[0093] The data preprocessing module is connected to the multimodal data acquisition module and is used to perform feature alignment and multimodal feature fusion on the multimodal data to generate multimodal response data, and to denoise and preprocess the multimodal response data to obtain preprocessed multimodal data; feature alignment and fusion are performed on the collected multimodal data to generate multimodal response data, and through denoising and preprocessing, redundant information is removed and effective features are retained to obtain preprocessed multimodal data. This process enhances the availability of data and lays the foundation for subsequent analysis.

[0094] The signal decoupling and anomaly detection module is connected to the data preprocessing module, and is used to perform sparse representation of the preprocessed multimodal data for adaptive signal decoupling, generate abnormal signal data, and perform nonlinear enhancement and feature normalization on the abnormal signal data to obtain clustered abnormal signal data; the sparse representation method is used to perform adaptive signal decoupling on the preprocessed multimodal data to separate the abnormal signal data. Subsequently, the abnormal signal data is subjected to nonlinear enhancement and feature normalization to obtain clustered abnormal signal data, thereby more clearly identifying potential defects.

[0095] The multi-scale inversion and three-dimensional reconstruction module is connected to the signal decoupling and anomaly detection module, which is used to perform multi-scale iterative inversion on the clustered abnormal signal data, and perform adaptive grid optimization and dynamic regulation in combination with the material physical model to generate the final three-dimensional reconstruction data; multi-scale iterative inversion on the clustered abnormal signal data, and perform adaptive grid optimization and dynamic regulation in combination with the material physical model to generate three-dimensional reconstruction data. This process improves the accuracy of defect location through multi-scale analysis.

[0096] The spatiotemporal evolution modeling module connects the multi-scale inversion and 3D reconstruction modules to simulate the spatiotemporal disturbance of the final 3D reconstruction data, extract dynamic features, and perform multi-level feedback optimization based on the correlation inversion to obtain defect evolution prediction data; perform spatiotemporal disturbance simulation based on the 3D reconstruction data to extract dynamic features. Combined with the correlation inversion technology, multi-level feedback optimization is performed to generate defect evolution prediction data and provide the possibility of future development of material defects.

[0097] The feedback optimization and adaptive learning module is connected to the spatiotemporal evolution modeling module. It is used to perform cross-domain adaptive learning of multi-level feedback optimization algorithms and neural network models based on defect evolution prediction data, combined with optical features, acoustic features, and mechanical features, and dynamically adjust the weights and algorithm parameters of the neural network model to achieve updated iterative optimization of the system. According to the defect evolution prediction data, combined with the feedback results of multimodal features, the multi-level feedback optimization algorithm and cross-domain adaptive learning of the neural network model are used to dynamically adjust the weights and algorithm parameters of the model to improve the iterative optimization effect of the system.

[0098] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware.

[0099] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for identifying microscopic defects of new energy materials based on computer vision, characterized in that: The following steps are involved: Perform multimodal optical interference and photoacoustic imaging scanning on new energy materials, collect phase interference data, polarization light data and photoacoustic data, and generate multimodal response data through multimodal feature alignment and multimodal feature fusion, and then denoise and preprocess the multimodal response data to obtain preprocessed multimodal data; Performing sparse representation of adaptive signal decoupling on preprocessed multimodal data to generate abnormal signal data, and performing nonlinear enhancement processing and feature normalization processing on the abnormal signal data to obtain clustered abnormal signal data; The defect area of ​​clustered abnormal signal data is reconstructed through multi-scale iterative inversion, and adaptive grid optimization and dynamic regulation are performed in combination with the material physical model to generate the final three-dimensional reconstruction data; Perform spatiotemporal disturbance simulation on the final 3D reconstructed data, extract dynamic features, and perform multi-level feedback optimization based on correlation inversion to obtain defect evolution prediction data; Based on the defect evolution prediction data, combined with optical, acoustic and mechanical characteristics, through the multi-level feedback optimization algorithm and cross-domain adaptive learning of the neural network model, the weights and algorithm parameters of the neural network model are dynamically adjusted to achieve system update and iterative optimization.

2. The computer vision-based new energy material micro-defect identification method according to claim 1, characterized in that: In the multimodal optical interference and photoacoustic imaging joint scanning step, by adjusting the phase angle and polarization angle, phase interference data and polarization light data are collected at different angles to cover the surface characteristics and internal structure characteristics of the new energy material; The multimodal feature alignment and multimodal feature fusion include aligning the phase interference data, polarization light data and photoacoustic data so that the phase interference data, polarization light data and photoacoustic data are aligned in the same coordinate system, and fusing the characteristic values ​​of the phase interference data, polarization light data and photoacoustic data by a weighted average method to generate multimodal response data.

3. The method for identifying microscopic defects of new energy materials based on computer vision according to claim 1, characterized in that: In the adaptive signal decoupling step of the sparse representation, the preprocessed multimodal data is decomposed into normal signal data and abnormal signal data, wherein the sparse coefficients of the preprocessed multimodal data are solved using a sparse representation method, and the abnormal signal data is extracted by minimizing the reconstruction error, and the abnormal signal data is used for nonlinear enhancement processing.

4. The method for identifying microscopic defects of new energy materials based on computer vision according to claim 1, characterized in that: The multi-scale iterative inversion step comprises: According to the clustered abnormal signal data, a material physical model is constructed, and initial meshing is performed based on the stress and strain characteristics of the material to generate an initial defect area model; the meshing adopts an adaptive algorithm to make the mesh in the high stress concentration area finer by adjusting the mesh density; In each iteration, the inversion error is calculated and the regularization parameters in the inversion algorithm are dynamically adjusted to optimize the error term in the inversion process. The defect area model is gradually updated in an iterative manner to obtain more accurate three-dimensional reconstruction data.

5. The method for identifying microscopic defects of new energy materials based on computer vision according to claim 4, characterized in that: The regularization parameter The adjustment is based on the following expression: in, Indicates The regularization parameter in the iteration; Indicates The regularization parameter in the iteration; Represents the error change in the current iteration, calculated as the difference between the error of this iteration and the error of the previous iteration; Represents the baseline error value of the initial iteration, which is used to dynamically control the regularization strength during the inversion process.

6. The method for identifying microscopic defects of new energy materials based on computer vision according to claim 4, characterized in that: In the spatiotemporal disturbance simulation step, a variety of disturbance conditions are applied to the final three-dimensional reconstructed data, including successively adjusting the temperature field and stress field according to a preset disturbance scheme, and solving the spatiotemporal evolution equation of the material by numerical methods to extract dynamic characteristics under different disturbance conditions, and the dynamic characteristics are used to reflect the changes in the defect area.

7. The computer vision-based new energy material micro-defect identification method according to claim 6, characterized in that: The correlation inversion processes multi-scale spatiotemporal feature data by means of joint optimization, wherein the weight coefficients in the weighted convolution algorithm Calculated according to the following expression: in, Indicates The weight coefficient of each feature layer; Indicates The feature importance coefficient of each feature layer; Represents the total number of all feature layers; Represents the weight normalization factor of all feature layers.

8. The method for identifying microscopic defects of new energy materials based on computer vision according to claim 1, characterized in that: In the cross-domain adaptive learning step of the multi-level feedback optimization algorithm and the neural network model, the weights and algorithm parameters of the neural network model are adjusted by transfer learning by combining the feedback results of the optical features, acoustic features and mechanical features. The transfer learning process achieves dynamic adjustment by comparing the similarities between the optical features, acoustic features and mechanical features and updating the weights of the neural network model to optimize the overall performance of the system.

9. The method for identifying microscopic defects of new energy materials based on computer vision according to claim 8, characterized in that: In the transfer learning process, a cross-domain feature fusion model of optical features, acoustic features and mechanical features is constructed, the feature values ​​of the optical features, acoustic features and mechanical features are normalized, and the similarities between the optical features, acoustic features and mechanical features are calculated, and the weight update strategy of the neural network model is adjusted according to the similarities.

10. A system for implementing the computer vision-based new energy material micro-defect recognition method according to any one of claims 1 to 9, characterized in that: The system includes: Multimodal data acquisition module, used to perform multimodal optical interference and photoacoustic imaging scanning on new energy materials, and collect phase interference data, polarization data and photoacoustic data; A data preprocessing module, connected to the multimodal data acquisition module, is used to perform feature alignment and multimodal feature fusion on the multimodal data to generate multimodal response data, and to denoise and preprocess the multimodal response data to obtain preprocessed multimodal data; The signal decoupling and anomaly detection module is connected to the data preprocessing module and is used to perform adaptive signal decoupling of sparse representation on the preprocessed multimodal data, generate abnormal signal data, and perform nonlinear enhancement and feature normalization processing on the abnormal signal data to obtain clustered abnormal signal data; The multi-scale inversion and 3D reconstruction module is connected to the signal decoupling and anomaly detection module, and is used to perform multi-scale iterative inversion on the clustered anomaly signal data, and perform adaptive grid optimization and dynamic regulation in combination with the material physical model to generate the final 3D reconstruction data; The spatiotemporal evolution modeling module connects the multi-scale inversion and 3D reconstruction modules to simulate the spatiotemporal disturbance of the final 3D reconstruction data, extract dynamic features, and perform multi-level feedback optimization based on the correlation inversion to obtain defect evolution prediction data; The feedback optimization and adaptive learning module is connected to the spatiotemporal evolution modeling module. It is used to perform cross-domain adaptive learning of multi-level feedback optimization algorithms and neural network models based on defect evolution prediction data, combined with optical, acoustic and mechanical characteristics, and dynamically adjust the weights and algorithm parameters of the neural network model to achieve system update and iterative optimization.

Citation Information

Patent Citations

  • Integrated multi-mode imaging method fusing ultrasonic imaging, photoacoustic imaging and optical coherence tomography and system employing same

    CN103048294A

  • Multi-mode signal detection method and multi-mode signal detection system for rail defects

    CN105092705A

  • Combined imaging method and system based on intravascular ultrasound-photoacoustic-OCT

    CN108703774A

  • An ultrasonic image reconstruction method based on sparse reconstruction

    CN109087246A

  • Laser ultrasonic vibration detection device and method

    CN109444265A

Cited By

  • Detection method and system for intermediate annealing of multi-pass rolling of copper wire

    CN119614844A

  • Graphene film surface performance detection device

    CN120085101A

  • A graphene film surface performance detection device

    CN120085101B

  • Method and system for detecting defects of super-soft knitted gauze based on image recognition

    CN120198392A

  • Image recognition-based ultra-soft knitted gauze fabric defect detection method and system

    CN120198392B