Component defect detection method and device, electronic equipment and storage medium

Abstract

The invention discloses a component defect detection method and device, electronic equipment and a storage medium. The method comprises the following steps: constructing a three-dimensional model of a component; collecting laser ultrasonic data of the three-dimensional model, and collecting a thermal imaging image and a polarized light image of the assembly; based on the laser ultrasonic data, the thermal imaging image and the polarized light image, predicting to obtain a first defect category and a first defect position corresponding to a defect area of the assembly; based on the defect category and the defect position of the defect area, reconstructing three-dimensional defect data of the component; and on the basis of the three-dimensional defect data, predicting to obtain an expansion direction and an expansion range of the defect area.

Description

Technical Field

[0001] The present application relates to the field of component detection, and specifically to a component defect detection method, device, electronic device and storage medium. Background Art

[0002] In recent years, with the continued rise in global demand for clean energy, photovoltaic power generation, as a key pillar of renewable energy, has seen its technological innovation and application scope continue to expand. As the core component of a photovoltaic system, the performance and reliability of modules directly impact the power generation efficiency and service life of the entire substation. Therefore, comprehensive and efficient module testing is crucial to ensuring stable substation operation.

[0003] Currently, a basic system of module defect detection technologies has been established, including electroluminescent (EL) testing, infrared thermal imaging, visual inspection, and electrical performance testing. However, existing module defect detection technologies primarily focus on identifying defects immediately, making it difficult to predict and implement effective preventive measures during long-term module operation, thereby increasing substation operation and maintenance risks and costs. Summary of the Invention

[0004] The present application provides a component defect detection method, device, electronic device and storage medium, which effectively integrates feature information in different types of data, can more accurately output defect categories and defect locations, that is, improve the accuracy of defect detection, and reconstruct three-dimensional defects based on the defect categories and defect locations, predict the reconstructed three-dimensional defects, output the expansion direction and expansion range of the defect area, accurately predict its future development trend, provide a more scientific and reliable basis for substation operation and maintenance decisions, and effectively reduce operation and maintenance risks and costs.

[0005] In a first aspect, the present application provides a component defect detection method, the method comprising:

[0006] Build a 3D model of the component;

[0007] Acquire laser ultrasonic data of the 3D model, as well as thermal and polarized images of the component;

[0008] Predicting a first defect category and a first defect position corresponding to a defective area of ​​the component based on the laser ultrasonic data, the thermal imaging image, and the polarized light image;

[0009] Reconstruct the three-dimensional defect data of the component based on the defect type and defect location of the defect area;

[0010] Based on the three-dimensional defect data, the expansion direction and range of the defect area are predicted.

[0011] In a second aspect, the present application provides a component defect detection device, the device comprising: an acquisition unit and a processing unit;

[0012] a processing unit for constructing a three-dimensional model of the component;

[0013] an acquisition unit for acquiring laser ultrasonic data of the three-dimensional model, and acquiring thermal imaging images and polarized light images of the components;

[0014] The processing unit is further configured to predict a first defect category and a first defect location corresponding to a defective region of the component based on the laser ultrasonic data, the thermal imaging image, and the polarized light image;

[0015] The processing unit is further used to reconstruct three-dimensional defect data of the component based on the defect category and defect location of the defect area;

[0016] The processing unit is further used to predict the expansion direction and expansion range of the defect area based on the three-dimensional defect data.

[0017] In a third aspect, the present application provides an electronic device comprising: a processor and a memory, the processor being connected to the memory, the memory being used to store computer programs, and the processor being used to execute the computer programs stored in the memory, so that the electronic device performs the method of the first aspect.

[0018] In a fourth aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method of the first aspect is performed.

[0019] In a fifth aspect, the present application provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, performs the method of the first aspect.

[0020] The implementation of this application has the following beneficial effects:

[0021] In an embodiment of the present application, a three-dimensional model of a component is first constructed; then, laser ultrasonic data of the three-dimensional model is collected, as well as thermal imaging images and polarized light images of the component; then, based on the laser ultrasonic data, thermal imaging images and polarized light images, a first defect category and a first defect position corresponding to a defect area of ​​the component are predicted; then, based on the defect category and defect position of the defect area, the three-dimensional defect data of the component is reconstructed; then, based on the three-dimensional defect data, the expansion direction and expansion range of the defect area are predicted. That is to say, by integrating laser ultrasonic data, as well as collecting thermal imaging images and polarized light images of components, the characteristic information in different types of data is effectively integrated, and the defect category and defect location can be output more accurately, which improves the accuracy of defect detection. In addition, the three-dimensional defects are reconstructed according to the defect category and defect location, so that the shape and characteristics of the defects are intuitively presented in three-dimensional form. The reconstructed three-dimensional defects are then predicted, and the expansion direction and expansion range of the defect area are output. Compared with traditional single detection technology, it can more comprehensively and accurately discover deep internal defects of components and accurately predict their future development trends, providing a more scientific and reliable basis for substation operation and maintenance decisions, effectively reducing operation and maintenance risks and costs, and improving the power generation efficiency and service life of substations. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0023] Figure 1 A schematic diagram of a flow chart of a component defect detection method provided in an embodiment of the present application;

[0024] Figure 2 A training method for a first model provided in an embodiment of the present application;

[0025] Figure 3 A schematic diagram of a process for reconstructing three-dimensional defects of a component based on the defect type and defect location of the defect area provided in an embodiment of the present application;

[0026] Figure 4 A schematic diagram of a process for predicting the expansion direction and expansion range of a defect area based on a three-dimensional defect provided in an embodiment of the present application;

[0027] Figure 5 A schematic diagram of a component defect detection system provided in an embodiment of the present application;

[0028] Figure 6A block diagram of the functional units of a component defect detection device provided in an embodiment of the present application;

[0029] Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0030] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0031] The terms "first," "second," "third," and "fourth," etc., in the specification, claims, and drawings of this application are used to distinguish between different objects, not to describe a particular order. In addition, the terms "including," "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or elements is not limited to the listed steps or elements, but may optionally include steps or elements not listed, or may optionally include other steps or elements inherent to the process, method, product, or apparatus.

[0032] References herein to "embodiments" mean that a particular feature, result, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0033] See Figure 1 , Figure 1 This is a flow chart of a component defect detection method provided in an embodiment of the present application. The method is applied to a component defect detection device, and the method includes but is not limited to steps S101-S105:

[0034] S101. Construct a three-dimensional model of the component.

[0035] In an embodiment of the present application, first, a surface roughness test is performed on the component to obtain a surface reflection test result. For example, by entangled photons and quantum interference, a subwavelength surface roughness test is performed on a light source with a wavelength within a first preset range to obtain a surface reflection test result. Among them, entangled photons refer to the existence of quantum correlation between two or more photons, and their states are interdependent. Even if they are far apart, the change in the state of one photon will instantly affect the state of another photon. This characteristic makes entangled photons have extremely high accuracy and sensitivity in measurement. Quantum interference utilizes the interference effect of two entangled photons on a beam splitter. When two identical photons are incident on the beam splitter at the same time, they will be emitted from the same output port at the same time with a certain probability. By measuring this interference effect, the characteristics of the photons, including their phase, polarization, etc., can be accurately analyzed. In surface roughness detection, the interaction between photons and the surface will change their phase and polarization. By analyzing the interference signal, the microscopic structure information of the surface can be inferred. Traditional surface roughness detection methods are limited by the optical diffraction limit and cannot achieve subwavelength-level accuracy. Therefore, the present invention breaks through this limitation by utilizing entangled photons and quantum interference technology, and realizes the accurate detection of subwavelength-level roughness of component surfaces.

[0036] Then, according to the surface reflection detection results, the scanning path of the component is adjusted to obtain the adjusted scanning path. For example, a microlens array is deployed to simulate the dragonfly compound eye structure, and the scanning path of the component is adjusted according to the surface reflection detection results to obtain the adjusted scanning path. Among them, the microlens array that simulates the dragonfly compound eye structure can disperse the incident light to multiple microlenses, each microlens corresponds to a local area, thereby achieving rapid scanning of a large area. This embodiment can obtain information from multiple areas at the same time by simulating the microlens array of the dragonfly compound eye structure, which greatly shortens the scanning time, and the real-time optimization of the scanning path further improves the targeted scanning and avoids wasting time and resources in defect-free areas. In addition, according to the surface reflection detection results, the surface roughness distribution and reflection characteristics are analyzed. For areas with higher reflectivity or greater roughness (such as areas with reflectivity greater than or equal to the first threshold), the scanning density can be increased; for areas with lower reflectivity or relatively flat surfaces (such as areas with reflectivity less than the first threshold), the scanning density can be reduced to improve scanning efficiency and accuracy. The optimized scanning path ensures full coverage of key areas of the component, provides accurate location information for the subsequent collection of multispectral point cloud data and local capacitive sensing data, and helps to improve the quality and integrity of data collection.

[0037] Then, based on the adjusted scanning path, the multispectral point cloud data and local capacitance sensing data of the component are collected. Then, based on the multispectral point cloud data and local capacitance sensing data, the internal defects of the component are determined. Among them, the multispectral acquires the three-dimensional coordinate information of the component and the reflection characteristics at different wavelengths by emitting laser pulses of different wavelengths. At this time, the multispectral point cloud data contains rich spatial and spectral information, which can reflect the surface and shallow structural characteristics of the component. The multispectral point cloud data provides the geometric information of the component, and the local capacitance sensing data provides the change information of the dielectric constant. The combination of the two can more accurately identify defects such as hidden cracks inside the component, overcoming the limitations of a single data source. This application combines multispectral point cloud data and local capacitance sensing data, and uses data analysis algorithms to fuse and analyze the data. By identifying the abnormal area of ​​dielectric constant, the shape, position and size of internal defects such as hidden cracks are reconstructed. The identified internal defect information provides an important data basis for the subsequent construction of a three-dimensional model of the component, which can accurately reflect the actual structure of the component.

[0038] Then, based on the impact of the component's internal defects and adhesive layer aging on light propagation, a three-dimensional model of the component is constructed. For example, by using the lattice Boltzmann algorithm to simulate the effect of adhesive layer aging on light propagation characteristics, the constructed three-dimensional model more accurately reflects the component's actual state in use, facilitating a more precise analysis of component performance and defects. Based on the component's internal defects and the impact of adhesive layer aging on light propagation characteristics simulated using the lattice Boltzmann algorithm, a three-dimensional model of the component is constructed using modeling algorithms in computer graphics, such as voxel modeling or triangulation algorithms. This 3D model not only includes the component's geometry but also characterizes the impact of the component's internal defects and adhesive layer aging on light propagation.

[0039] It can be seen that in the embodiments of the present application, laser pulses of different wavelengths are emitted by multispectral LiDAR (Light Detection and Ranging), and the reflected signals are received to obtain the three-dimensional coordinate information of the component and the reflection characteristics at different wavelengths. Lasers of different wavelengths have different responses to the material, surface state, etc. of the object, and can provide richer target information. In this embodiment, the collected laser reflection signals are processed and converted into point cloud data, and each point contains information such as three-dimensional coordinates and reflection intensity. Using these point cloud data, modeling algorithms in computer graphics, such as triangulation algorithms, are used to connect discrete points into surfaces to construct a three-dimensional model of the component. The three-dimensional model constructed in this embodiment can accurately present the geometric shape, size and spatial position relationship of the component, thereby improving the efficiency and accuracy of data acquisition.

[0040] S102 , collecting laser ultrasonic data of the three-dimensional model, and collecting thermal imaging images and polarized light images of the components.

[0041] In an embodiment of the present application, laser ultrasound technology can be used to excite ultrasound waves in a component using a laser, and defects within the material can be detected by detecting the propagation characteristics of the ultrasound waves. Also, during component operation, defects can cause local temperature anomalies. Therefore, by collecting thermal imaging images, differences in temperature distribution on the component surface can be captured, thereby discovering potential defects. Furthermore, different materials and surface conditions have different effects on the polarization state of light. By analyzing polarized light images, changes in the polarization state when light is reflected or transmitted through the surface of an object can be obtained, and information about the surface microstructure and material properties of the object can be obtained, which helps detect some surface or near-surface defects. In other words, in an embodiment of the present application, multiple data sources reflect the defect characteristics of the component from different perspectives. The laser ultrasound data reveals the defects within the component, the thermal imaging images reflect the temperature anomalies caused by the defects, and the polarized light images provide surface and near-surface microscopic information. Combining these data allows for more comprehensive and accurate detection of component defects, helping to improve the accuracy of predictions.

[0042] S103 : Based on the laser ultrasonic data, the thermal imaging image, and the polarized light image, predict the defect type and defect location corresponding to the defective area of ​​the component.

[0043] In an embodiment of the present application, the laser ultrasonic data may include first scanning data and second scanning data, wherein the first scanning data is generated by point scanning (such as A-scan, the English name is Amplitude Scan, which is not limited in this application), reflecting the time-amplitude information of the ultrasonic wave propagating in the component; the second scanning data is generated by line scanning (such as B-scan, the English name is Brightness Scan, which is not limited in this application), which is a two-dimensional image obtained by performing A-scan at different positions, and can intuitively display the position and shape of the defect.

[0044] In an embodiment of the embodiment of the present application, before executing step S103, the laser ultrasonic data can also be enhanced, and then subsequent processing is performed based on the enhanced laser ultrasonic data. Specifically: first, an adaptive decomposition tree is constructed according to the signal-to-noise ratio of the laser ultrasonic data. For example, a wavelet basis is used to dynamically determine the decomposition depth according to the signal-to-noise ratio of the laser ultrasonic data and construct an adaptive decomposition tree, thereby avoiding the problem of over-decomposition or insufficient decomposition that may be caused by a fixed decomposition depth, and improving the pertinence and effectiveness of signal decomposition.

[0045] Then, the wavelet packet node energy distribution in the adaptive decomposition tree is mapped to the Ising model spin state. For example, a quantum annealing algorithm is used to map the wavelet packet node energy distribution in the decomposition tree to the Ising model spin state. The quantum annealing algorithm can effectively handle complex combinatorial optimization problems such as wavelet packet node energy distribution optimization. By mapping the problem to the Ising model spin state, the advantages of quantum computing can be fully utilized to improve solution efficiency and accuracy. This embodiment also establishes a connection between the wavelet packet node energy distribution and the Ising model spin state, providing a feasible method for subsequently obtaining the optimal frequency band combination (i.e., the target frequency band combination) by solving the Ising model.

[0046] Then, the Ising model spin state is solved to obtain the target frequency band combination. For example, the optimal frequency band combination, or target frequency band combination, is obtained by solving it through a quantum computing platform. Compared with traditional computing methods, the quantum computing platform has obvious advantages in handling large-scale combinatorial optimization problems, greatly improving the solution efficiency. The optimal frequency band combination contains the target frequency component information in the laser ultrasound data, providing a key basis for subsequent signal enhancement.

[0047] Then, based on the target frequency band combination, the laser ultrasonic data is enhanced, which can effectively remove noise interference, improve the signal-to-noise ratio and clarity of the signal, and make the effective information in the signal more prominent. The enhanced signal can provide more accurate and reliable data support for subsequent defect detection, material performance evaluation and other applications, thereby improving the accuracy and reliability of the entire detection and analysis process.

[0048] Therefore, illustratively, feature extraction can be performed on the first scan data to obtain time-domain defect features. For example, depthwise separable convolution can be used to extract features from the first scan data to obtain multi-scale time-domain features corresponding to the first scan data. The multi-scale time-domain features are then focused on the defect echoes through a gated attention mechanism to obtain time-domain defect features in the first scan data. Feature extraction can then be performed on the second scan data to obtain frequency-domain entanglement features. For example, the second scan data can be mapped to a quantum state through a quantum Fourier transform, and then a variational quantum algorithm can be used to extract frequency-domain entanglement features. Of course, other methods can also be used to extract time-domain defect features and frequency-domain entanglement features, which are not limited in this application.

[0049] Then, feature extraction is performed on the thermal imaging image to obtain the first image feature, and feature extraction is performed on the polarized light image to obtain the second image feature. For example, according to the preset window size and sliding step size, the thermal imaging image is divided into multiple first image blocks; each first image block is then mapped to a high-dimensional embedding space through linear transformation to obtain the embedded features of each first image block; the embedded features of each first image block are then self-attention processed to obtain the first feature vector of each first image block; then, based on the offset window size, the multiple first image blocks are window-shifted to obtain multiple second image blocks, wherein the multiple first image blocks and the multiple second image pairs have a one-to-one correspondence; then, attention processing is performed based on the first feature vector of each first image block and the multiple second image blocks to obtain each second image block. The second eigenvector corresponding to the image block; then, based on the first eigenvector of each first image block and the second eigenvector of each second image block, a third eigenvector corresponding to the thermal imaging image is obtained, for example, by splicing, weighted fusion, etc., which is not limited in this application; then, the first image feature is determined based on the third eigenvector, for example, the third eigenvector can be directly determined as the first image feature, or the third eigenvector can be repeatedly executed once or multiple times by the subsequent steps corresponding to the above "performing self-attention processing on the embedded features of each first image block to obtain the first eigenvector of each first image block" to obtain the first image feature, which is not limited in this application.

[0050] Then, based on the time domain defect feature, the frequency domain entanglement feature, the first image feature and the second image feature, a fusion feature is obtained. For example, the correlation strength between the features is determined. Specifically, for the first feature, the correlation strength between the first feature and the second feature is determined, wherein the correlation strength between the first feature and the second feature can be the entanglement entropy between the first feature and the second feature, the first feature is any one of the time domain defect feature, the frequency domain entanglement feature, the first image feature and the second image feature, and the second feature is the feature other than the first feature among the time domain defect feature, the frequency domain entanglement feature, the first image feature and the second image feature; then the correlation strength between the first feature and each second feature is summed to obtain the first correlation strength corresponding to the first feature; then, based on the first correlation strength corresponding to the first feature, the weight corresponding to the first feature is determined. Similarly, the weight corresponding to each first feature can be obtained, that is, the weight corresponding to the time domain defect feature, the frequency domain entanglement feature, the first image feature and the second image feature are obtained respectively; then the weight corresponding to each first feature, the time domain defect feature, the frequency domain entanglement feature, the first image feature and the second image feature are weighted to obtain the fusion feature. Of course, other fusion methods can also be used, which are not limited in this application. Then, based on the fused features, the defect category and defect location are predicted. For example, the fused features are predicted based on a convolutional neural network to obtain the defect category and defect location corresponding to the defect area of ​​the component.

[0051] In an optional embodiment, step S103 may also be performed by a first model, which may include an input layer, a time-frequency domain dual-flow sub-attention layer, an image fusion layer, a cross-modal association layer, a convolutional neural network layer, an output layer, etc. Therefore, refer to Figure 2 , Figure 2 The present invention provides a first model training method. The first model training method may include but is not limited to the following steps S201-S207:

[0052] S201 . Input first sample data into an input layer, wherein the first sample data includes laser ultrasound sample data of a three-dimensional model of a component, a thermal imaging sample image of the component, and a polarized light sample image.

[0053] In an embodiment of the present application, a generative adversarial network can be used to generate multi-source sample data of a three-dimensional model of a component (i.e., laser ultrasound sample data, thermal imaging sample images of the component, and polarized light sample images). At this time, each first sample data can include multi-source sample data, and the true label corresponding to each first sample data (i.e., the defect category and defect position of the defect area of ​​the component corresponding to each first sample data) is determined. It should be noted that this application does not limit the number of first sample data, and here it is mainly explained by taking one as an example.

[0054] S202. Through the time-frequency domain dual-flow quantum attention layer, feature extraction is performed on the first scanning sample data in the laser ultrasonic sample data to obtain time domain defect sample features, and feature extraction is performed on the second scanning sample data in the laser ultrasonic sample data to obtain frequency domain entanglement sample features.

[0055] It should be noted that the specific principles will not be described here in detail, and reference can be made to the corresponding explanations above.

[0056] S203 , performing feature extraction on the thermal imaging sample image through the image fusion layer to obtain first image sample features, and performing feature extraction on the polarized light sample image to obtain second image sample features.

[0057] It should be noted that the specific principles will not be described here in detail, and reference can be made to the corresponding explanations above.

[0058] S204 , fusing the time domain defect sample features, the frequency domain entanglement sample features, the first image sample features, and the second image sample features through a cross-modal association layer to obtain a fused sample feature.

[0059] S205: Process the fused sample features through a convolutional neural network to obtain prediction features.

[0060] S206 , activating the prediction features through the output layer to obtain a first defect category and a first defect position corresponding to the first sample data.

[0061] S207. Determine a first loss based on the first defect category and the first defect location, and the true label corresponding to the first sample data, and update the first model based on the first loss to obtain a trained first model.

[0062] In an embodiment of the present application, a first sub-loss, such as a cross-entropy loss, can be determined based on the first defect category and the first defect location, as well as the true label corresponding to the first sample data; and a second sub-loss, such as a mean square error loss, can be determined based on the first defect category and the first defect location, as well as the true label corresponding to the first sample data; and then a first loss can be obtained based on the first sub-loss and the second sub-loss, such as a weighted summation, which is not limited in this application. Furthermore, an optimization algorithm, such as the Adam algorithm, can be used to backpropagate and update the parameters of the first model according to the first loss until the first loss is within a preset range, thereby obtaining a trained first model.

[0063] It can be seen that in the embodiments of the present application, the laser ultrasonic data is enhanced by the wavelet packet-blind source separation joint algorithm, and the time-frequency domain dual-flow quantum attention module and the image fusion module are combined to extract the time domain defect features, the frequency domain entanglement features and the corresponding image features respectively. Then, the cross-modal correlation layer measures the correlation strength between different modal features through quantum entanglement entropy to achieve efficient fusion of cross-modal features, which significantly improves the accuracy and robustness of defect identification. This method of multi-source data fusion and cross-modal feature extraction enables the detection system to comprehensively analyze the defect information of the component from multiple dimensions, effectively avoiding the limitations of a single detection technology.

[0064] S104: Reconstruct the three-dimensional defects of the component based on the defect category and defect location of the defect area.

[0065] For example, see Figure 3 , Figure 3 The present invention provides a schematic diagram of a process for reconstructing three-dimensional defects of a component based on the defect type and defect location of the defect area, which specifically includes but is not limited to the following steps S301-S304:

[0066] S301 : Predict the color and density of the defect area based on a first defect category and a first defect position of the defect area.

[0067] For example, if the first defect category and the first defect position corresponding to the defect area of ​​the component are in the first viewing direction, then multiple second defect categories and multiple second defect positions in the first viewing direction can be obtained; then, based on the multiple second defect categories and the multiple second defect positions, multiple defect information groups are obtained, wherein the multiple second defect categories, the multiple second defect positions and the multiple defect information groups correspond one to one, and each defect information group includes a second defect category and a second defect position; then, a first similarity between the first defect category and the second defect category in each defect information group is determined, and a second similarity between the first defect position and the second defect position in each defect information group is determined; then, based on the first similarity and the second similarity corresponding to each defect information group, a third similarity corresponding to each defect information group is determined; then, based on the third similarity corresponding to each defect information group, a first defect information group is determined from the multiple defect information groups; then, based on the first mapping relationship between the defect category, the defect position in the defect information group and the color and density of the defect area, the color and density of the defect area corresponding to the first defect information group are determined.

[0068] In an optional embodiment, step S301 may also be performed by a second model, wherein the second model may be, for example, a neural radiance field model based on a multi-layer perceptron. Fields, NeRF), which is not limited in this application; the training method of the second model may include but is not limited to the following: first, using a generative adversarial network to generate defect category samples and defect location samples of the component; then, with the defect location samples of the component as the center, determining multiple viewing directions; then determining the fourth defect category and fourth defect location corresponding to the component in each viewing direction, and constructing a first training set based on the fourth defect category and fourth defect location corresponding to the component in each viewing direction, the first training data set includes multiple second sample data, the multiple second sample data correspond one-to-one to the multiple viewing directions, and each second sample data includes the fourth defect category and fourth defect location in each viewing direction; then inputting each second sample data into the second model to predict the color and density of the defect area corresponding to the second sample data in each viewing direction; then, based on the color and density of the defect area corresponding to the second sample data in each viewing direction and the true label, determining a second loss, such as a cross entropy loss, a mean square error loss, etc., which is not limited in this application; then, training the second model based on the second loss until the second model converges, and further, based on the learned correspondence between each sample data and the color and density of the defect area in each viewing direction, a first mapping relationship can be generated.

[0069] S302 : Generate a plurality of first images for representing the appearance of the defective area based on the color and density of the defective area.

[0070] Exemplarily, the color and density of the defect area are integrated according to the rays in each of the multiple projection directions (such as numerical integration, Monte Carlo integration, etc., which are not limited in this application) to obtain the color value corresponding to each ray; then, based on the color values ​​corresponding to the rays, multiple first images are generated, wherein the first image is a two-dimensional image, and the multiple first images correspond one-to-one to the multiple projection methods.

[0071] S303 : Generate three-dimensional voxel data of the defect area based on the multiple first images.

[0072] In an embodiment of the present application, feature extraction is performed on each first image to obtain features corresponding to each first image, such as by using Scale-Invariant Feature Transform (SIFT), Speeded-Up Robust Features (SURF), etc., which are not limited in the present application; then, feature matching is performed based on the features corresponding to each first image to obtain matched feature point pairs, wherein the feature point pairs are feature points of the same object corresponding to the first images in different projection directions; then, based on each feature point pair, the disparity corresponding to each feature point in the feature point pair is determined, wherein the disparity refers to the difference in pixel position of the same object point under different viewing angles (projection directions); then, based on the disparity of each feature point, the three-dimensional coordinates of each feature point are determined; then, based on the three-dimensional coordinates of each feature point, three-dimensional reconstruction is performed to obtain three-dimensional voxel data, such as a three-dimensional reconstruction algorithm including voxel gridding, surface reconstruction, etc., which are not limited in the present application.

[0073] S304: Reconstruct the three-dimensional defects of the component based on the three-dimensional voxel data.

[0074] In the embodiments of the present application, the three-dimensional voxel data can be reconstructed to obtain the three-dimensional shape and structure of the defect area, thereby achieving accurate representation of the three-dimensional defect.

[0075] S105. Based on the three-dimensional defect, predict the expansion direction and expansion range of the defect area.

[0076] For example, see Figure 4 , Figure 4 A schematic diagram of a process for predicting the expansion direction and expansion range of a defect area based on a three-dimensional defect provided in an embodiment of the present application, specifically including but not limited to steps S401-S404:

[0077] S401 , extracting features from three-dimensional defect data to obtain three-dimensional defect features.

[0078] S402: Obtain a mapping relationship between defect geometric parameters and environmental stress tensors.

[0079] In the embodiments of the present application, the defect geometric parameters mainly describe the characteristics of the defect in three-dimensional space, such as shape, size and position, including but not limited to the following aspects: (1) Shape parameters: may include crack length and width, pore diameter and depth, inclusion shape (such as spherical, elliptical, etc.) and size (such as diameter, volume, etc.), etc., which are not limited in this application; (2) Position parameters: may include the coordinate position of the defect in three-dimensional space, the depth information of the defect, etc., which are not limited in this application; (3) Direction parameters: may include crack direction (i.e., the extension direction of the crack, which can affect the crack expansion rate and direction), defect inclination angle (which can affect its mechanical properties and expansion behavior), etc., which are not limited in this application.

[0080] The environmental stress tensor describes the stress state of an object in three-dimensional space. It reflects the stress distribution at each point within the object due to external forces and can influence the rate and direction of defect growth. For example, under tensile stress, a crack may more likely propagate in the tensile direction; under shear stress, a crack may more likely deflect or bifurcate. This mapping relationship between defect geometric parameters and the environmental stress tensor can be studied, allowing prediction of defect growth behavior under different stress states and providing a scientific basis for defect prevention and control.

[0081] S403 : Based on the mapping relationship and the three-dimensional defect feature, extract the historical extension feature corresponding to the three-dimensional defect feature.

[0082] In an embodiment of the present application, a temporal convolutional network can be constructed according to the mapping relationship between the defect geometric parameters and the environmental stress tensor, and then based on the temporal convolutional network, historical expansion features corresponding to the three-dimensional defect features can be extracted, such as the historical expansion direction and range, etc., which are not limited in this application.

[0083] S404 , predicting the expansion direction and expansion range of the defect area based on historical expansion features, laser ultrasonic data, thermal imaging images, and polarized light images.

[0084] In an embodiment of the present application, based on the physical constraints of the components, laser ultrasonic data, thermal imaging images and polarized light images are screened to obtain first data, which is one or more of the laser ultrasonic data, thermal imaging images and polarized light images; then, feature extraction is performed on each first data to obtain a first feature corresponding to each first data; then, the historical extension feature and the first feature corresponding to each first data are fused to obtain a first fused feature; then, prediction is performed based on the first fused feature, for example, using a long short-term memory network (LSTM) to predict the first fused feature to obtain the expansion direction and expansion range of the defect area.

[0085] In an optional embodiment, in step S105, based on the three-dimensional defect, the prediction of the expansion direction and expansion range of the defect area can also be performed by a third model, wherein the third model can include a physical information neural network (PINN) and a long short-term memory network, and the training method of the third model can specifically include but is not limited to the following: first, obtain three-dimensional defect sample data, the number of three-dimensional defect sample data can be one or more, and this application mainly uses one as an example to illustrate; then the three-dimensional defect sample data is input into the physical information neural network and the long short-term memory network in the third model for processing, and the expansion prediction direction and expansion prediction range of the defect area are output; then based on the expansion prediction direction and expansion prediction range of the defect area and the true label, the third loss is determined, wherein the principle of the third loss is similar to that of the first loss and will not be repeated here; then based on the third loss, the third model is trained, for example, an optimization algorithm such as the Adam algorithm can be used to backpropagate and update the parameters of the first model according to the third loss until the third loss is within a preset range to obtain a trained third model.

[0086] It can be seen that in the embodiments of the present application, by combining the LSTM network and the PINN network, combining time series analysis and physical knowledge, it is possible to predict the future expansion direction and range of the defective area, provide a decision-making basis for the maintenance and replacement of components, help reduce operation and maintenance costs, and improve the power generation efficiency and service life of the substation.

[0087] For further information, see Figure 5 , Figure 5 A schematic diagram of a component defect detection system provided in an embodiment of the present application. The system includes a component defect detection device, a laser ultrasonic device, a thermal imaging device, and a polarized light imaging device; the component defect detection device can be a terminal device or a server, and the terminal device can be a smart phone, tablet computer, laptop computer, desktop computer, smart TV, desktop computer, smart watch, smart car and other smart terminals, but is not limited to this, and this application does not make any specific restrictions; the server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, as well as basic cloud computing services such as big data and artificial intelligence platforms, and this application does not make any specific restrictions. Figure 5 The embodiment is mainly described by taking the component defect detection device as a terminal device as an example.

[0088] Therefore, the component defect detection device constructs a three-dimensional model of the component; and the laser ultrasonic equipment collects laser ultrasonic data of the three-dimensional model, the thermal imaging equipment collects thermal imaging images of the component, and the polarized light imaging equipment collects polarized light images; then the component defect detection device predicts the first defect category and the first defect position corresponding to the defect area of ​​the component based on the laser ultrasonic data received from the laser ultrasonic equipment, the thermal imaging image received from the thermal imaging equipment, and the polarized light image received from the polarized light imaging equipment; then the component defect detection device reconstructs the three-dimensional defect data of the component based on the defect category and defect position of the defect area; then the component defect detection device predicts the expansion direction and expansion range of the defect area based on the three-dimensional defect data. The specific implementation principle will not be repeated here, and the corresponding explanation of the above embodiment can be referred to.

[0089] Of course, if the component defect detection device is a server, after predicting the expansion direction and expansion range of the defect area, the component defect detection device can also send the expansion direction and expansion range of the defect area to the terminal device, so that the operation and maintenance personnel can take preventive measures based on the expansion direction and expansion range of the defect area.

[0090] It should be noted that Figure 5 The component defect detection device in the embodiment can also perform other steps corresponding to the above embodiment and achieve the same technical effects, which will not be described here one by one.

[0091] See Figure 6 , Figure 6 This is a block diagram of the functional units of a component defect detection device provided in an embodiment of the present application. The component defect detection device 600 includes: an acquisition unit 601 and a processing unit 602;

[0092] a processing unit for constructing a three-dimensional model of the component;

[0093] an acquisition unit for acquiring laser ultrasonic data of the three-dimensional model, and acquiring thermal imaging images and polarized light images of the components;

[0094] The processing unit is further configured to predict a first defect category and a first defect location corresponding to a defective region of the component based on the laser ultrasonic data, the thermal imaging image, and the polarized light image;

[0095] The processing unit is further used to reconstruct three-dimensional defect data of the component based on the defect category and defect location of the defect area;

[0096] The processing unit is further used to predict the expansion direction and expansion range of the defect area based on the three-dimensional defect data.

[0097] In one embodiment of the present application, the laser ultrasonic data includes first scanning data and second scanning data, the first scanning data is generated by a point scanning method, and the second scanning data is generated by a line scanning method. In predicting the defect type and defect location corresponding to the defective area of ​​the component based on the laser ultrasonic data, the thermal imaging image, and the polarized light image, the processing unit 602 is specifically configured to:

[0098] Performing feature extraction on the second scanning data to obtain frequency domain entanglement features;

[0099] Performing feature extraction on the thermal imaging image to obtain a first image feature, and performing feature extraction on the polarized light image to obtain a second image feature;

[0100] Obtaining fusion features based on the time domain defect features, the frequency domain entanglement features, the first image features, and the second image features;

[0101] Based on the fused features, the defect category and defect location are predicted.

[0102] In one embodiment of the present application, in terms of obtaining a fusion feature based on the time domain defect feature, the frequency domain entanglement feature, the first image feature, and the second image feature, the processing unit 602 is specifically configured to:

[0103] Determine the strength of associations between features;

[0104] Based on the correlation strength between features, time domain defect features, frequency domain entanglement features, first image features and second image features, a fusion feature is obtained.

[0105] In one embodiment of the present application, in terms of reconstructing a three-dimensional defect of a component based on a defect type and a defect location of a defect area, the processing unit 602 is specifically configured to:

[0106] Based on the defect category and defect location of the defect area, the color and density of the defect area are predicted;

[0107] generating a plurality of first images for representing the appearance of the defect area based on the color and density of the defect area;

[0108] generating three-dimensional voxel data of the defect area based on the plurality of first images;

[0109] Reconstruct the 3D defects of components based on 3D voxel data.

[0110] In one embodiment of the present application, in terms of predicting the expansion direction and expansion range of the defect area based on the three-dimensional defect data, the processing unit 602 is specifically configured to:

[0111] Perform feature extraction on three-dimensional defect data to obtain three-dimensional defect features;

[0112] Obtaining the mapping relationship between defect geometric parameters and environmental stress tensor;

[0113] Based on the mapping relationship and the three-dimensional defect features, the historical extension features corresponding to the three-dimensional defect features are extracted;

[0114] Based on historical expansion characteristics, laser ultrasonic data, thermal imaging images and polarized light images, the expansion direction and range of the defect area are predicted.

[0115] In one embodiment of the present application, in terms of constructing a three-dimensional model of a component, the processing unit 602 is specifically configured to:

[0116] Perform surface roughness testing on the components to obtain surface reflection test results of the components;

[0117] Adjusting the scanning path of the component according to the surface reflection detection result to obtain an adjusted scanning path;

[0118] Based on the adjusted scanning path, multispectral point cloud data and local capacitive sensing data of the component are collected;

[0119] Identify internal defects of components based on multispectral point cloud data and local capacitive sensing data;

[0120] Build a 3D model of the component based on the impact of internal defects of the component and aging of the adhesive layer on light propagation.

[0121] In one embodiment of the present application, the processing unit 602 is specifically configured to:

[0122] According to the signal-to-noise ratio of laser ultrasonic data, an adaptive decomposition tree is constructed;

[0123] Mapping the wavelet packet node energy distribution in the adaptive decomposition tree to the Ising model spin state;

[0124] Solve the Ising model spin state to obtain the target frequency band combination;

[0125] Laser ultrasound data is enhanced based on target frequency band combinations.

[0126] In a specific implementation, the acquisition unit 601 and the processing unit 602 described in the embodiment of the present invention may also execute other implementation methods described in the embodiments of other component defect detection methods provided by the embodiment of the present invention, which will not be repeated here.

[0127] See Figure 7 , Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 7As shown, electronic device 700 includes transceiver 701, processor 702 and memory 703. They are connected via bus 704. Memory 703 is used to store computer programs and data, and can transmit data stored in memory 703 to processor 702.

[0128] The processor 702 is configured to read the computer program in the memory 703 and perform the following operations:

[0129] Build a 3D model of the component;

[0130] Controlling the transceiver 701 to collect laser ultrasonic data of the three-dimensional model, as well as thermal imaging images and polarized light images of the components;

[0131] Predicting a first defect category and a first defect position corresponding to a defective area of ​​the component based on the laser ultrasonic data, the thermal imaging image, and the polarized light image;

[0132] Reconstruct the three-dimensional defect data of the component based on the defect type and defect location of the defect area;

[0133] Based on the three-dimensional defect data, the expansion direction and range of the defect area are predicted.

[0134] In one embodiment of the present application, the laser ultrasonic data includes first scanning data and second scanning data, the first scanning data is generated by a point scanning method, and the second scanning data is generated by a line scanning method. In terms of predicting the defect type and defect location corresponding to the defect area of ​​the component based on the laser ultrasonic data, the thermal imaging image, and the polarized light image, the processor 702 is configured to perform the following operations:

[0135] Performing feature extraction on the first scanning data to obtain time domain defect features;

[0136] Performing feature extraction on the second scanning data to obtain frequency domain entanglement features;

[0137] Performing feature extraction on the thermal imaging image to obtain a first image feature, and performing feature extraction on the polarized light image to obtain a second image feature;

[0138] Obtaining fusion features based on the time domain defect features, the frequency domain entanglement features, the first image features, and the second image features;

[0139] Based on the fused features, the defect category and defect location are predicted.

[0140] In one embodiment of the present application, in terms of obtaining a fusion feature based on the time domain defect feature, the frequency domain entanglement feature, the first image feature, and the second image feature, the processor 702 is configured to perform the following operations:

[0141] Determine the strength of associations between features;

[0142] Based on the correlation strength between features, time domain defect features, frequency domain entanglement features, first image features and second image features, a fusion feature is obtained.

[0143] In one embodiment of the present application, in terms of reconstructing a three-dimensional defect of a component based on a defect category and a defect location of a defect region, the processor 702 is configured to perform the following operations:

[0144] Based on the defect category and defect location of the defect area, the color and density of the defect area are predicted;

[0145] generating a plurality of first images for representing the appearance of the defect area based on the color and density of the defect area;

[0146] generating three-dimensional voxel data of the defect area based on the plurality of first images;

[0147] Reconstruct the 3D defects of components based on 3D voxel data.

[0148] In one embodiment of the present application, in terms of predicting the expansion direction and expansion range of the defect area based on the three-dimensional defect data, the processor 702 is configured to perform the following operations:

[0149] Perform feature extraction on three-dimensional defect data to obtain three-dimensional defect features;

[0150] Obtaining the mapping relationship between defect geometric parameters and environmental stress tensor;

[0151] Based on the mapping relationship and the three-dimensional defect features, the historical extension features corresponding to the three-dimensional defect features are extracted;

[0152] Based on historical expansion characteristics, laser ultrasonic data, thermal imaging images and polarized light images, the expansion direction and range of the defect area are predicted.

[0153] In one embodiment of the present application, in terms of constructing a three-dimensional model of a component, the processor 702 is configured to perform the following operations:

[0154] Perform surface roughness testing on the components to obtain surface reflection test results of the components;

[0155] Adjusting the scanning path of the component according to the surface reflection detection result to obtain an adjusted scanning path;

[0156] Based on the adjusted scanning path, multispectral point cloud data and local capacitive sensing data of the component are collected;

[0157] Identify internal defects of components based on multispectral point cloud data and local capacitive sensing data;

[0158] Build a 3D model of the component based on the impact of internal defects of the component and aging of the adhesive layer on light propagation.

[0159] In one embodiment of the present application, the processor 702 is configured to perform the following operations:

[0160] According to the signal-to-noise ratio of laser ultrasonic data, an adaptive decomposition tree is constructed;

[0161] Mapping the wavelet packet node energy distribution in the adaptive decomposition tree to the Ising model spin state;

[0162] Solve the Ising model spin state to obtain the target frequency band combination;

[0163] Laser ultrasound data is enhanced based on target frequency band combinations.

[0164] In a specific implementation, the transceiver 701 and the processor 702 described in the embodiment of the present invention may also execute other implementations described in the embodiment of the component defect detection method provided by the embodiment of the present invention, which will not be described in detail here.

[0165] Specifically, the transceiver 701 can be Figure 6 The acquisition unit 601 of the component defect detection device 600 of the embodiment, the above-mentioned processor 702 can be Figure 6 The processing unit 602 of the component defect detection device 600 of the embodiment.

[0166] It should be understood that an embodiment of the present application also provides a computer-readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement part or all of the steps of any component defect detection method recorded in the above method embodiments.

[0167] An embodiment of the present application also provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to enable a computer to execute part or all of the steps of any component defect detection method described in the above method embodiments.

[0168] It should be noted that for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that this application is not limited by the order of the actions described, because according to this application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all optional embodiments, and the actions and modules involved are not necessarily required by this application.

[0169] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0170] In the several embodiments provided in this application, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0171] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0172] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or in the form of software program modules.

[0173] If the integrated unit is implemented in the form of a software program module and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a memory and includes a number of instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned memory includes various media that can store program codes, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0174] Those skilled in the art will appreciate that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program. The program can be stored in a computer-readable memory, and the memory can include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0175] The above is a detailed introduction to the embodiments of the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application. At the same time, for those skilled in the art, based on the ideas of the present application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A component defect detection method, characterized in that: The method comprises: Build a 3D model of the component; collecting laser ultrasonic data of the three-dimensional model, and collecting thermal imaging images and polarized light images of the component; Predicting a first defect category and a first defect position corresponding to a defective area of ​​the component based on the laser ultrasonic data, the thermal imaging image, and the polarized light image; reconstructing three-dimensional defect data of the component based on the defect type and defect position of the defect area; Based on the three-dimensional defect data, the expansion direction and expansion range of the defect area are predicted.

2. The method according to claim 1, characterized in that The laser ultrasonic data includes first scanning data and second scanning data, the first scanning data is generated by a point scanning method, and the second scanning data is generated by a line scanning method; the defect type and defect location corresponding to the defect area of ​​the component are predicted based on the laser ultrasonic data, the thermal imaging image, and the polarized light image, including: performing feature extraction on the first scanning data to obtain time-domain defect features; performing feature extraction on the second scanning data to obtain frequency domain entanglement features; performing feature extraction on the thermal imaging image to obtain a first image feature, and performing feature extraction on the polarized light image to obtain a second image feature; Obtaining a fusion feature based on the time domain defect feature, the frequency domain entanglement feature, the first image feature, and the second image feature; Based on the fused features, the defect category and the defect location are predicted.

3. The method according to claim 2, characterized in that The obtaining of a fusion feature based on the time domain defect feature, the frequency domain entanglement feature, the first image feature, and the second image feature includes: Determine the strength of associations between features; The fusion feature is obtained based on the correlation strength between the features, the time domain defect feature, the frequency domain entanglement feature, the first image feature and the second image feature.

4. The method according to any one of claims 1 to 3, characterized in that The reconstructing the three-dimensional defects of the component based on the defect type and defect position of the defect area includes: Predicting the color and density of the defect area based on the defect type and defect location of the defect area; generating a plurality of first images for representing the appearance of the defect area based on the color and density of the defect area; generating three-dimensional voxel data of the defect area based on the plurality of first images; Based on the three-dimensional voxel data, a three-dimensional defect of the component is reconstructed.

5. The method according to any one of claims 1 to 4, characterized in that: The predicting of the expansion direction and expansion range of the defect area based on the three-dimensional defect data includes: performing feature extraction on the three-dimensional defect data to obtain three-dimensional defect features; Obtaining the mapping relationship between defect geometric parameters and environmental stress tensor; Extracting a historical extension feature corresponding to the three-dimensional defect feature based on the mapping relationship and the three-dimensional defect feature; Based on the historical expansion characteristics, the laser ultrasonic data, the thermal imaging image and the polarized light image, the expansion direction and expansion range of the defect area are predicted.

6. The method according to any one of claims 1 to 5, characterized in that The three-dimensional model of the building component includes: Performing a surface roughness test on the component to obtain a surface reflection test result of the component; Adjusting the scanning path of the component according to the surface reflection detection result to obtain an adjusted scanning path; Based on the adjusted scanning path, collecting multispectral point cloud data and local capacitance sensing data of the component; determining internal defects of the component based on the multispectral point cloud data and the local capacitance sensing data; Based on the influence of internal defects of the component and aging of the adhesive layer on light propagation, a three-dimensional model of the component is constructed.

7. The method according to any one of claims 1 to 6, characterized in that The method further comprises: constructing an adaptive decomposition tree according to the signal-to-noise ratio of the laser ultrasound data; Mapping the wavelet packet node energy distribution in the adaptive decomposition tree to an Ising model spin state; Solving the Ising model spin state to obtain a target frequency band combination; The laser ultrasound data is enhanced based on the target frequency band combination.

8. A component defect detection device, characterized in that: The device comprises: an acquisition unit and a processing unit; a processing unit for constructing a three-dimensional model of the component; an acquisition unit, configured to acquire laser ultrasonic data of the three-dimensional model, and acquire thermal imaging images and polarized light images of the component; The processing unit is further configured to predict a first defect category and a first defect position corresponding to a defective area of ​​the component based on the laser ultrasonic data, the thermal imaging image, and the polarized light image; The processing unit is further configured to reconstruct three-dimensional defect data of the component based on the defect type and defect position of the defect area; The processing unit is further configured to predict the expansion direction and expansion range of the defect area based on the three-dimensional defect data.

9. An electronic device, characterized in that: include: A processor and a memory, the processor is connected to the memory, the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the electronic device performs the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and the computer program is executed by a processor to implement the method according to any one of claims 1 to 7.

Images