Abnormal image detection method and system for anti-ablation uniformity sample of static arc contact workpiece

Abstract

The invention provides a static arc contact workpiece anti-ablation uniformity sample abnormal image detection method and system, and the method comprises the steps: carrying out the modeling of the physical behaviors of a static arc contact workpiece under different arc conditions, and generating a multi-physical field coupling simulation model and a theoretical reference image; acquiring an actual surface image, extracting advanced features by adopting various quantitative indexes and a convolutional neural network based on the theoretical reference image, performing comparative analysis, and generating a matching degree evaluation result; performing anomaly detection on the evaluation result, identifying an abnormal region with non-uniform ablation resistance, and generating an anomaly identification result; and according to an abnormity identification result, utilizing a reverse engineering principle and combining a particle swarm optimization algorithm to reconstruct an arc behavior mode causing abnormity, inversely adjusting input parameters through a simulation model until abnormity is reproduced, obtaining preliminary cause diagnosis information, and generating a detailed cause diagnosis report. According to the invention, the capability of detecting and improving the ablation resistance uniformity of the static arc contact workpiece is improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of static arc contact workpieces, and in particular to a method and system for detecting abnormal images of samples of static arc contact workpiece anti-ablation uniformity. Background Art

[0002] In the power system, the static arc contact workpiece is a key component in the switchgear, and its anti-ablation performance is directly related to the reliability and safety of the equipment. As the demand for high voltage and high current in the power system continues to increase, the physical behavior of the static arc contact workpiece under different arc conditions becomes more and more complex.

[0003] In order to ensure that these workpieces can work stably under extreme conditions, existing technologies usually rely on experimental tests and numerical simulation methods to evaluate their anti-ablation performance. Experimental tests obtain data through tests in actual operating environments. Although intuitive, they are costly and time-consuming, and it is difficult to cover all possible working conditions. Numerical simulations use tools such as finite element analysis and computational fluid dynamics to perform preliminary physical behavior modeling. However, most of these models only consider a single physical field and cannot fully capture the coupling effects of multiple physical fields, resulting in inaccurate simulation results. In addition, traditional image processing and analysis methods such as threshold segmentation and edge detection are also often used for preliminary analysis of actual surface images.

[0004] However, these methods lack advanced feature extraction capabilities and have difficulty effectively identifying subtle differences, resulting in poor detection and improvement capabilities for the uniformity of static arc contact workpiece anti-ablation. Summary of the invention

[0005] The embodiments of the present application provide a method and system for detecting abnormal images of samples of anti-ablation uniformity of a static arc contact workpiece, so as to solve the problem of poor detection and improvement capabilities of anti-ablation uniformity of a static arc contact workpiece in the prior art.

[0006] In a first aspect, an embodiment of the present application provides a method for detecting abnormal images of a static arc contact workpiece anti-ablation uniformity sample, comprising:

[0007] Modeling the physical behavior of the static arc contact workpiece under different arc conditions to obtain a multi-physics field coupling simulation model, and generating a theoretical reference image based on the multi-physics field coupling simulation model, wherein the theoretical reference image refers to an image of the expected physical behavior of the static arc contact workpiece under specific working conditions;

[0008] Acquire an actual surface image of the static arc contact workpiece, perform comparative analysis on the actual surface image according to the theoretical reference image, adopt multiple quantitative indices, introduce a convolutional neural network to extract high-level image features, apply a scale-invariant feature transformation or a key point matching algorithm of an accelerated robust feature to perform image registration, and generate an evaluation result of the degree of matching between the actual surface image and the theoretical reference image, wherein the multiple quantitative indices include a structural similarity index and a peak signal-to-noise ratio;

[0009] Performing anomaly detection on the evaluation results to identify abnormal areas with uneven anti-ablation performance, and constructing an autoencoder or variational autoencoder unsupervised learning model based on the abnormal areas to learn normal patterns from data collected under normal operating conditions to accurately identify abnormal areas and generate abnormal identification results;

[0010] According to the abnormality identification result, the arc behavior pattern that causes the abnormal phenomenon is reconstructed by utilizing the reverse engineering principle combined with the particle swarm optimization algorithm. The input parameters are adjusted by inversion through the multi-physical field coupling simulation model until the abnormality is reproduced, thereby obtaining preliminary causal diagnosis information. The causal relationship between variables is explored using a causal reasoning framework, and a detailed causal diagnosis report is generated, so that the design or manufacturing process parameters of the static arc contact workpiece can be adjusted in a targeted manner through the detailed causal diagnosis report.

[0011] Optionally, according to the abnormality identification result, the reverse engineering principle is combined with the particle swarm optimization algorithm to reconstruct the arc behavior pattern that causes the abnormal phenomenon, and the input parameters are adjusted by the multi-physics field coupling simulation model until the abnormality is reproduced to obtain preliminary cause diagnosis information, and the causal relationship between variables is explored by applying the causal reasoning framework to generate a detailed cause diagnosis report, including:

[0012] By using the reverse engineering principle combined with the particle swarm optimization algorithm, based on the non-uniform anti-ablation performance area determined by the abnormality identification result, the key parameters affecting the arc behavior in the multi-physics field coupling simulation model are analyzed, a series of key parameters are set, and the optimal parameter combination is automatically searched through the particle swarm optimization algorithm, and the optimal parameter combination is adjusted. The multi-physics field coupling simulation model reappears the abnormal phenomenon actually detected, and a preliminary reconstruction of the arc behavior pattern is obtained;

[0013] According to the preliminary reconstruction of the arc behavior pattern, the input parameters in the multi-physics field coupling simulation model are gradually fine-tuned, and after each adjustment, the matching degree between the simulation result and the actual abnormal phenomenon is verified to obtain the influence of different parameter changes on the abnormal degree, and the simulation of the abnormal behavior pattern is refined to accurately simulate the actually observed abnormality and generate a refined abnormal behavior pattern simulation result;

[0014] Based on the detailed abnormal behavior pattern simulation results, a causal reasoning framework is applied to conduct a systematic causal relationship analysis of all involved variables, determine the factors that directly or indirectly cause the abnormality, and quantify the causal effects between the factors to obtain a detailed cause diagnosis report. All the involved variables include key parameters of arc behavior, material properties, simulation model parameters and other potential influencing factors;

[0015] Optionally, the reconstruction of the preliminary arc behavior pattern gradually fine-tunes the input parameters in the multi-physics field coupling simulation model, verifies the degree of matching between the simulation results and the actual abnormal phenomenon after each adjustment, obtains the influence of different parameter changes on the degree of abnormality, refines the simulation of the abnormal behavior pattern, accurately simulates the actually observed abnormalities, and generates refined abnormal behavior pattern simulation results, including:

[0016] Using the preliminary arc behavior pattern reconstruction, gradually fine-tuning the input parameters in the multi-physics field coupling simulation model, adjusting one or more key parameters each time to obtain an adjusted parameter combination, wherein the parameters include current intensity, voltage level, and material properties;

[0017] According to the adjusted parameter combination, the simulation model is run to generate new simulation results, and a plurality of quantitative indicators are used to compare and analyze the new simulation results with the abnormal phenomena actually detected and evaluate the matching degree, so as to obtain the matching evaluation results after each adjustment;

[0018] Based on the matching evaluation results after each adjustment, the parameter changes that affect the abnormal phenomenon are recorded and analyzed to obtain simulation results under different parameter settings, and a detailed analysis report on the impact of the parameter changes on the abnormality is obtained by comparing the simulation results under the different parameter settings;

[0019] According to the detailed analysis report, fine-tune and verify the input parameters in the simulation model until the simulation model accurately reproduces the abnormal phenomena actually observed and generates accurate abnormal behavior pattern simulation results;

[0020] Optionally, based on the matching evaluation result after each adjustment, the parameter changes that affect the abnormal phenomenon are recorded and analyzed to obtain simulation results under different parameter settings, and by comparing the simulation results under different parameter settings, a detailed analysis report on the impact of the parameter changes on the abnormality is obtained, including:

[0021] Based on the matching evaluation results after each adjustment, the key parameter changes that significantly affect the abnormal phenomenon during the simulation process are recorded in detail, and each fine-tuning and the results corresponding to each fine-tuning are accurately saved using the key parameters to obtain a detailed parameter change record, wherein the key parameters include current intensity, voltage level, and material properties;

[0022] According to the detailed parameter change records, the simulation results generated after each parameter adjustment are sorted and summarized to form a complete data set, and a summary of simulation results under different parameter settings is obtained;

[0023] Apply statistical analysis and data mining techniques to compare the simulation results under the different parameter settings, identify the impact of parameter changes on the degree of abnormality, and establish a mathematical model or a machine learning model based on the simulation results under the different parameter settings to quantify the impact of each parameter on the abnormality, and obtain a parameter impact model;

[0024] Perform sensitivity analysis using the parameter impact model to generate sensitivity analysis results, present the sensitivity analysis results in the form of charts, present the impact trend of different parameter changes on abnormal phenomena, and obtain visual charts;

[0025] According to the visualization chart and the sensitivity analysis results, a detailed analysis report of the impact of parameter changes on abnormalities is obtained, which describes the specific impact of parameter changes on abnormal phenomena and ranks the importance of each key parameter, and points out the degree of impact of each key parameter on the abnormality.

[0026] Optionally, performing abnormality detection on the evaluation results to identify abnormal areas with uneven anti-ablation performance, and constructing an autoencoder or variational autoencoder unsupervised learning model based on the abnormal areas, learning normal patterns from data collected under normal operating conditions to accurately identify abnormal areas and generate abnormality identification results, including:

[0027] Using multiple quantitative indicators in the evaluation results, anomaly detection is performed on the difference between the actual surface image of the static arc contact workpiece and the theoretical reference image, and a local outlier factor or an isolation forest algorithm is applied to identify abnormal areas indicating the presence of uneven anti-ablation performance, thereby obtaining preliminary anomaly detection results;

[0028] Based on the preliminary anomaly detection results, construct an autoencoder or variational autoencoder unsupervised learning model to learn normal patterns from data collected under normal operating conditions and generate a trained anomaly recognition model;

[0029] The preliminary anomaly detection result is input into the trained anomaly recognition model, and further analyzed and processed by the trained anomaly recognition model to accurately identify the abnormal area, and use the learning ability of the model to capture the existing minor anomalies to generate anomaly recognition results.

[0030] Optionally, the method utilizes the multiple quantitative indicators in the evaluation results to perform anomaly detection on the difference between the actual surface image of the static arc contact workpiece and the theoretical reference image, applies a local outlier factor or an isolation forest algorithm, identifies abnormal areas indicating the presence of uneven anti-ablation performance, and obtains preliminary anomaly detection results, including:

[0031] Using the actual surface image of the static arc contact workpiece, a detailed comparative analysis is performed between the actual surface image and a series of theoretical reference images covering a wide range of working parameter combinations to obtain an evaluation result;

[0032] According to multiple quantitative indicators in the evaluation result, a quantitative evaluation process is performed on the difference between the actual surface image and the theoretical reference image to generate a matching degree evaluation result, wherein the multiple quantitative indicators include a structural similarity index and a peak signal-to-noise ratio;

[0033] Based on the matching degree evaluation result, applying a local outlier factor or an isolation forest algorithm, performing anomaly detection processing on the difference between the actual surface image and the theoretical reference image, and identifying an abnormal area indicating the presence of non-uniform anti-ablation performance;

[0034] The output of the LOF or Isolation Forest algorithm is used to mark the abnormal area and generate a preliminary abnormality detection result.

[0035] Optionally, the actual surface image of the static arc contact workpiece is obtained, and the actual surface image is compared and analyzed according to the theoretical reference image, multiple quantitative indicators are used, and a convolutional neural network is introduced to extract high-level image features, and a scale-invariant feature transformation or a key point matching algorithm of accelerated robust features is applied to perform image registration, and an evaluation result of the degree of matching between the actual surface image and the theoretical reference image is generated, including:

[0036] Using high-resolution imaging equipment, the actual surface of the static arc contact workpiece is photographed to obtain high-quality surface image data and obtain the actual surface image of the static arc contact workpiece;

[0037] According to the theoretical reference image, the actual surface image is compared and analyzed to identify the similarities and differences between the theoretical reference image and the actual surface image, and a plurality of quantitative indicators are used to evaluate the similarity between the actual surface image and the theoretical reference image to obtain a preliminary result of the matching degree between the two;

[0038] Based on deep learning technology, a convolutional neural network is introduced to extract high-level features from the actual surface image to generate high-level feature representation of the image, wherein the convolutional neural network automatically learns and extracts complex patterns and features in the image;

[0039] Applying a scale-invariant feature transform or an accelerated robust feature algorithm to perform key point matching on the high-level feature representation of the image to obtain a key point matching result;

[0040] The plurality of quantitative indicators, the high-level feature representation of the image and the key point matching results are combined to generate an evaluation result of the degree of matching between the actual surface image and the theoretical reference image.

[0041] In a second aspect, an embodiment of the present application provides a detection and processing system for abnormal image of anti-ablation uniformity sample of a static arc contact workpiece, comprising:

[0042] A modeling module, used for modeling the physical behavior of the static arc contact workpiece under different arc conditions, obtaining a multi-physics field coupling simulation model, and generating a theoretical reference image based on the multi-physics field coupling simulation model, wherein the theoretical reference image refers to an image of the expected physical behavior of the static arc contact workpiece under specific working conditions;

[0043] an acquisition module, for acquiring an actual surface image of the static arc contact workpiece, performing comparative analysis on the actual surface image according to the theoretical reference image, using a variety of quantitative indices, introducing a convolutional neural network to extract high-level image features, applying a scale-invariant feature transformation or a key point matching algorithm of an accelerated robust feature to perform image registration, and generating an evaluation result of the degree of matching between the actual surface image and the theoretical reference image, wherein the multiple quantitative indices include a structural similarity index and a peak signal-to-noise ratio;

[0044] A detection module, configured to perform anomaly detection on the evaluation results, identify abnormal areas with uneven anti-ablation performance, and construct an autoencoder or variational autoencoder unsupervised learning model based on the abnormal areas, learn normal patterns from data collected under normal operating conditions, so as to accurately identify abnormal areas and generate abnormal identification results;

[0045] A reconstruction module is used to reconstruct the arc behavior pattern that causes the abnormal phenomenon based on the abnormality identification result by using the reverse engineering principle combined with the particle swarm optimization algorithm, adjust the input parameters through the multi-physical field coupling simulation model inversion until the abnormality is reproduced, obtain preliminary cause diagnosis information, apply the causal reasoning framework to explore the causal relationship between variables, and generate a detailed cause diagnosis report, so as to make targeted adjustments to the design or manufacturing process parameters of the static arc contact workpiece through the detailed cause diagnosis report.

[0046] In a third aspect, an embodiment of the present application provides a computing device, comprising a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute a method for detecting abnormal images of a static arc contact workpiece anti-ablation uniformity sample as described in any one of the first aspects.

[0047] In a fourth aspect, an embodiment of the present application provides a computer storage medium having computer program instructions stored thereon, wherein the computer program instructions, when executed by a processor, implement a method for detecting abnormal images of a static arc contact workpiece anti-ablation uniformity sample as described in any one of the first aspects.

[0048] In an embodiment of the present application, the physical behavior of a static arc contact workpiece under different arc conditions is modeled to obtain a multi-physics field coupling simulation model, and based on the multi-physics field coupling simulation model, a theoretical reference image is generated, and the theoretical reference image refers to an image of the expected physical behavior of the static arc contact workpiece under specific working conditions; an actual surface image of the static arc contact workpiece is obtained, and the actual surface image is compared and analyzed according to the theoretical reference image, a variety of quantitative indicators are used, and a convolutional neural network is introduced to extract high-level image features, and a scale-invariant feature transformation or a key point matching algorithm of an accelerated robust feature is applied to perform image registration, and an evaluation result of the degree of matching between the actual surface image and the theoretical reference image is generated, and the multiple quantitative indicators include a structural similarity index and a peak signal. Noise ratio; perform anomaly detection on the evaluation results to identify abnormal areas with uneven anti-ablation performance, and construct an autoencoder or variational autoencoder unsupervised learning model based on the abnormal areas, learn normal patterns from data collected under normal operating conditions to accurately identify abnormal areas and generate abnormal identification results; based on the abnormal identification results, use reverse engineering principles in combination with a particle swarm optimization algorithm to reconstruct the arc behavior pattern that causes the abnormal phenomenon, invert and adjust input parameters through the multi-physical field coupling simulation model until the anomaly is reproduced, obtain preliminary cause diagnosis information, apply a causal reasoning framework to explore the causal relationship between variables, and generate a detailed cause diagnosis report, so as to make targeted adjustments to the design or manufacturing process parameters of the static arc contact workpiece through the detailed cause diagnosis report.

[0049] The technical solution of this application has the following beneficial effects:

[0050] This application models the physical behavior of the static arc contact workpiece under different arc conditions, obtains a multi-physical field coupling simulation model, and generates a theoretical reference image. This method ensures the accurate simulation of complex physical phenomena, provides high-quality theoretical reference, and lays a solid foundation for subsequent anomaly detection. After obtaining the actual surface image, high-level features are extracted through a variety of quantitative indicators and convolutional neural networks, and image registration is performed by combining scale-invariant feature transformation or key point matching algorithm of accelerated robust features, so as to achieve a comprehensive evaluation of the degree of matching between the actual surface image and the theoretical reference image. This step not only takes into account the consistency of the overall structure, but also captures subtle local differences. Based on the preliminary matching degree evaluation results, an autoencoder or variational autoencoder unsupervised learning model is used to learn normal patterns from data collected under normal operating conditions, thereby accurately identifying abnormal areas with uneven anti-ablation performance. This method improves the accuracy of anomaly detection and reduces the possibility of false alarms. According to the anomaly identification results, the reverse engineering principle combined with the particle swarm optimization algorithm is used to reconstruct the arc behavior pattern that causes the abnormal phenomenon. The input parameters are adjusted through the multi-physics field coupling simulation model until the anomaly is reproduced. The causal relationship between variables is explored using the causal reasoning framework, and finally a detailed cause diagnosis report is generated. This process not only reveals the specific cause of the anomaly, but also provides suggestions for improvement. Based on the detailed cause diagnosis report, the design or manufacturing process parameters of the static arc contact workpiece are adjusted in a targeted manner, and a closed-loop feedback control system is established to monitor the key parameters in the production process in real time and adjust them dynamically. This method ensures the continuous optimization of the production process and improves the uniformity of the product's anti-ablation resistance and overall quality.

[0051] Furthermore, by combining the reverse engineering principle with the particle swarm optimization algorithm, the best parameter combination can be automatically searched to accurately reconstruct the arc behavior pattern that causes the abnormal phenomenon. This process not only improves the accuracy of abnormal reconstruction, but also reduces the need for manual intervention. The causal reasoning framework is applied to systematically analyze the causal relationship of all involved variables, determine the factors that directly or indirectly cause the abnormality, and quantify the causal effects between these factors. This makes the cause diagnosis more scientific and comprehensive, and provides a clear direction for subsequent improvements. The input parameters in the multi-physics field coupling simulation model are gradually fine-tuned, and the degree of match between the simulation results and the actual abnormal phenomenon is verified after each adjustment, and the simulation of the abnormal behavior pattern is refined. This method ensures the high accuracy of the simulation results and can accurately simulate the actual observed anomalies. The entire process realizes closed-loop feedback control from anomaly identification to cause diagnosis, greatly improving the efficiency and reliability of anomaly detection and shortening the time cycle for problem discovery and resolution. The detailed cause diagnosis report generated in the end not only reveals the specific cause of the anomaly, but also provides improvement suggestions to help engineers make targeted adjustments to the design or manufacturing process parameters of the static arc contact workpiece, optimize the production process, and improve product quality.

[0052] Furthermore, one or more key parameters are adjusted each time to obtain the adjusted parameter combination, and the simulation model is run to generate new simulation results. By comparing and analyzing the simulation results under different parameter settings, the parameter changes that affect the abnormal phenomenon are recorded and analyzed in detail, and a detailed analysis report on the impact of parameter changes on the abnormality is obtained. This step enhances the understanding of parameter sensitivity and abnormal mechanism. A variety of quantitative indicators are used to compare and analyze the new simulation results with the actual detected abnormal phenomena and evaluate the degree of matching, ensuring that the matching evaluation results after each adjustment are strictly verified. This method improves the credibility of the simulation results and ensures the high accuracy of the simulation model. According to the detailed analysis report, the input parameters in the simulation model are fine-tuned and verified until the simulation model accurately reproduces the actual observed abnormal phenomena. This process realizes the continuous optimization of the simulation model, ensuring that it can truly reflect the actual working conditions and provide a solid foundation for subsequent improvement measures. Through the detailed simulation results of abnormal behavior patterns, engineers can more intuitively understand the occurrence mechanism of abnormal phenomena, so as to make more scientific and reasonable decisions, guide actual production and design adjustments, and further improve the anti-ablation uniformity and overall quality of products.

[0053] These and other aspects of the present application will become more clearly understood in the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. 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 paying any creative work.

[0055] Figure 1 A flow chart of a method for detecting abnormal images of a static arc contact workpiece anti-ablation uniformity sample provided in an embodiment of the present application;

[0056] Figure 2 A schematic diagram of the structure of a system for detecting and processing abnormal images of samples of anti-ablation uniformity of a static arc contact workpiece provided in an embodiment of the present application;

[0057] Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0058] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0059] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0060] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0061] Figure 1 A flowchart of a method for detecting abnormal image of a static arc contact workpiece anti-ablation uniformity sample is provided for an embodiment of the present application, such as Figure 1 As shown, the method includes:

[0062] Step 101: Modeling the physical behavior of a static arc contact workpiece under different arc conditions to obtain a multi-physics field coupling simulation model, and generating a theoretical reference image based on the multi-physics field coupling simulation model;

[0063] In this step, the theoretical reference image refers to the image of the expected physical behavior of the static arc contact workpiece under specific working conditions; the multi-physics field coupling simulation model represents the simulation of the physical behavior of the static arc contact workpiece under different arc conditions through technologies such as finite element analysis and computational fluid dynamics. The model takes into account the interaction between multiple physical fields such as electromagnetic fields, heat conduction, and material deformation to obtain high-precision simulation results; the theoretical reference image represents the image of the expected physical behavior of the static arc contact workpiece generated under specific working conditions based on the multi-physics field coupling simulation model. These images are important benchmarks for subsequent abnormality detection and cause diagnosis.

[0064] In the embodiment of the present application, in an electric power equipment manufacturing plant, engineers use commercial simulation software to model the static arc contact workpiece. Different arc conditions are set, such as current intensity from 5kA to 20kA, voltage level from 10kV to 35kV, and combined with the actual material properties, the multi-physics field coupling simulation model is run to generate a series of theoretical reference images covering a wide range of working parameter combinations. These images describe in detail the temperature distribution, stress and strain conditions, and possible ablation areas on the surface of the static arc contact workpiece under different arc conditions.

[0065] The embodiment of the present application provides a high-quality theoretical reference image, which provides an accurate benchmark for subsequent anomaly detection. Through multi-physical field coupling simulation, the comprehensiveness and accuracy of the model are ensured, and the cost and time of experimental testing are reduced.

[0066] Step 102: obtaining an actual surface image of the static arc contact workpiece, performing comparative analysis on the actual surface image according to the theoretical reference image, using multiple quantitative indicators, introducing a convolutional neural network to extract high-level image features, applying a scale-invariant feature transformation or a key point matching algorithm of an accelerated robust feature to perform image registration, and generating an evaluation result of the degree of matching between the actual surface image and the theoretical reference image;

[0067] In this step, multiple quantitative indicators include structural similarity index and peak signal-to-noise ratio; actual surface image acquisition means using high-resolution imaging equipment to capture the actual surface image of the static arc contact workpiece to ensure that the image can fully reflect the details and characteristics of the workpiece surface; comparative analysis and quantitative index means using multiple quantitative indicators, such as structural similarity index and peak signal-to-noise ratio, to evaluate the similarity between the actual surface image and the theoretical reference image, and provide quantitative matching degree evaluation results; advanced feature extraction and image registration means introducing convolutional neural networks to extract advanced features in the image, such as texture, edge, and shape, and applying scale-invariant feature transformation or accelerated robust feature algorithm for key point matching to ensure the accuracy of image registration.

[0068] In an embodiment of the present application, continuing with the above embodiment, after completing the generation of the theoretical reference image, the factory uses a high-resolution industrial camera to collect actual surface images of multiple static arc contact workpieces under different operating conditions. Then, the technicians use the OpenCV library in MATLAB or Python to implement image processing and analysis.

[0069] The embodiments of the present application improve the accuracy and reliability of image comparison analysis, enhance the sensitivity to subtle differences, utilize advanced image processing technology, ensure the accuracy of image registration, and provide a solid foundation for subsequent anomaly detection.

[0070] Step 103: performing anomaly detection on the evaluation results to identify abnormal areas with uneven anti-ablation performance, and constructing an autoencoder or variational autoencoder unsupervised learning model based on the abnormal areas to learn normal patterns from data collected under normal operating conditions to accurately identify abnormal areas and generate abnormal identification results;

[0071] In this step, anomaly detection means applying statistical anomaly detection methods, such as local outlier factor LOF or isolation forest algorithm, based on the matching degree evaluation results to identify outliers that may have uneven anti-ablation performance; unsupervised learning model means building an autoencoder or variational autoencoder model to learn normal patterns from data collected under normal operating conditions to improve the accuracy of anomaly recognition; anomaly recognition results refer to further analyzing and processing the preliminary anomaly detection results through the trained model, accurately identifying abnormal areas, and generating detailed anomaly recognition reports.

[0072] In the embodiment of the present application, based on the aforementioned matching degree evaluation results, the factory technicians used the Scikit-Learn library in Python to implement the local outlier factor algorithm and identify preliminary outliers. Subsequently, they built a PyTorch-based autoencoder model to learn normal patterns from a large amount of data under normal operating conditions. After sufficient training, the model can effectively distinguish between normal and abnormal situations, and ultimately generate detailed anomaly recognition results. For example, the model successfully identified abnormal ablation areas on the surface of certain workpieces under high current conditions.

[0073] The embodiments of the present application improve the accuracy and robustness of anomaly detection, reduce the possibility of false alarms, and enhance the ability to understand and process complex image features through the learning ability of unsupervised learning models.

[0074] Step 104: Based on the abnormality identification result, the reverse engineering principle is combined with the particle swarm optimization algorithm to reconstruct the arc behavior pattern that causes the abnormal phenomenon, and the input parameters are adjusted through the multi-physics field coupling simulation model inversion until the abnormality is reproduced to obtain preliminary cause diagnosis information, and the causal relationship between variables is explored using a causal reasoning framework to generate a detailed cause diagnosis report, so as to make targeted adjustments to the design or manufacturing process parameters of the static arc contact workpiece through the detailed cause diagnosis report.

[0075] In this step, reverse engineering and particle swarm optimization means that based on the anomaly identification results, the reverse engineering principle is combined with the particle swarm optimization algorithm to determine the key parameters affecting the arc behavior, and automatically search for the best parameter combination so that the simulation model can reproduce the abnormal phenomenon; the causal reasoning framework means that the causal reasoning framework is used to conduct a systematic causal relationship analysis on all involved variables, determine the factors that directly or indirectly cause the anomaly, and quantify the causal effects between these factors; the cause diagnosis report refers to the generation of a detailed cause diagnosis report to provide a scientific basis for the design of the static arc contact workpiece or the adjustment of the manufacturing process parameters.

[0076] In the embodiment of the present application, based on the abnormality identification results, the factory's technical team first used the reverse engineering principle combined with the particle swarm optimization algorithm to reconstruct the arc behavior pattern that caused the abnormal phenomenon. By adjusting the input parameters in the multi-physics field coupling simulation model, such as current intensity, voltage level, and material properties, they gradually fine-tuned until the simulation model accurately reproduced the actual observed abnormalities. Then, a causal reasoning framework was applied to conduct a systematic analysis of all the variables involved, and it was determined that the main factors causing the anomaly were excessive current intensity and insufficient thermal conductivity of the material. Finally, a detailed cause diagnosis report was generated, suggesting that materials with higher thermal conductivity be selected during the design phase and that current intensity be strictly controlled during the manufacturing process.

[0077] The embodiment of the present application realizes closed-loop feedback control from abnormality identification to cause diagnosis, greatly improving the efficiency of problem discovery and resolution. Through a detailed cause diagnosis report, clear improvement suggestions are provided for the design and manufacture of static arc contact workpieces, optimizing the production process and improving product quality.

[0078] Through the above four steps, this method realizes the closed-loop control of the whole process from multi-physics field coupling simulation modeling to anomaly detection, cause diagnosis and process optimization, and significantly improves the detection and improvement capabilities of the anti-ablation uniformity of static arc contact workpieces. First, by high-precision modeling of the physical behavior of static arc contact workpieces under different arc conditions, theoretical reference images covering a wide range of working parameter combinations are generated, providing an accurate benchmark for subsequent analysis. Then, high-resolution imaging equipment is used to obtain actual surface images, and combined with a variety of quantitative indicators and advanced image processing techniques, such as convolutional neural networks, scale-invariant feature transform (SIFT) or accelerated robust features (SURF) algorithms, comparative analysis is performed to generate detailed matching degree evaluation results, ensuring high sensitivity and accuracy to subtle differences.

[0079] Optionally, according to the abnormality identification result, step 104 uses the reverse engineering principle combined with the particle swarm optimization algorithm to reconstruct the arc behavior pattern that causes the abnormal phenomenon, adjusts the input parameters through the multi-physics field coupling simulation model inversion until the abnormality is reproduced, obtains preliminary cause diagnosis information, applies the causal reasoning framework to explore the causal relationship between variables, and generates a detailed cause diagnosis report, including: using the reverse engineering principle combined with the particle swarm optimization algorithm, based on the anti-ablation performance inhomogeneous area determined by the abnormality identification result, analyzes the key parameters affecting the arc behavior in the multi-physics field coupling simulation model, sets a series of key parameters and automatically searches for the best parameter combination through the particle swarm optimization algorithm, adjusts the best parameter combination, and the multi-physics field coupling simulation model reappears the abnormal phenomenon actually detected, and obtains to a preliminary reconstruction of the arc behavior pattern; based on the preliminary reconstruction of the arc behavior pattern, gradually fine-tune the input parameters in the multi-physics field coupling simulation model, verify the degree of matching between the simulation results and the actual abnormal phenomenon after each adjustment, obtain the influence of different parameter changes on the degree of abnormality, refine the simulation of the abnormal behavior pattern, accurately simulate the actual observed abnormalities, and generate refined abnormal behavior pattern simulation results; based on the refined abnormal behavior pattern simulation results, apply a causal reasoning framework to conduct a systematic causal relationship analysis of all involved variables, determine the factors that directly or indirectly cause the abnormality, and quantify the causal effects between the factors, and obtain a detailed cause diagnosis report, wherein all involved variables include key parameters of arc behavior, material properties, simulation model parameters and other potential influencing factors;

[0080] In this step, the reverse engineering principle and particle swarm optimization (PSO) algorithm are used to reconstruct the arc behavior pattern that causes the abnormal phenomenon. Reverse engineering refers to the inference of possible causes through known results, while the particle swarm optimization algorithm is a global optimization technology used to automatically search for the best parameter combination. This process involves analyzing the key parameters that affect the arc behavior in the multi-physics field coupling simulation model, such as current intensity, voltage level, material properties, etc., setting a series of key parameters, and automatically finding the best parameter combination that enables the simulation model to reproduce the actual anomaly through the PSO algorithm. Finally, by adjusting these parameters, the simulation model can reproduce the actual detected anomaly, thereby obtaining a preliminary reconstruction of the arc behavior pattern.

[0081] In the embodiment of the present application, according to the above concept, the whole scheme process is as follows: First, based on the non-uniform area of ​​anti-ablation performance determined by the abnormal identification results, the engineer conducts a detailed analysis of the key parameters affecting the arc behavior in the multi-physics field coupling simulation model. Then, a series of initial parameters are set, and the optimal parameter combination is automatically searched through the particle swarm optimization algorithm. After finding the optimal parameter combination, these parameters are adjusted to ensure that the multi-physics field coupling simulation model can reproduce the actual detected abnormal phenomena. Then, based on the preliminary reconstruction of the arc behavior pattern, the input parameters in the simulation model are gradually fine-tuned, and the degree of match between the simulation results and the actual abnormal phenomena is verified after each adjustment, so as to refine the simulation of the abnormal behavior pattern and accurately simulate the actual observed abnormalities. Finally, the causal reasoning framework is applied to conduct a systematic causal relationship analysis of all involved variables, generate a detailed cause diagnosis report, and guide the subsequent design and manufacturing process parameter adjustment.

[0082] In a power equipment manufacturing plant, technicians identified abnormal areas with uneven anti-ablation performance on the surface of some static arc contact workpieces under high current conditions based on the results of anomaly identification. In order to further explore the causes of these anomalies, they first used the principle of reverse engineering and combined with the particle swarm optimization algorithm to analyze the key parameters affecting arc behavior in the multi-physics field coupling simulation model, including current intensity, voltage level and material properties. The technicians set an initial parameter set and automatically searched for the best parameter combination through the particle swarm optimization algorithm. After multiple iterations, they found a set of parameters that enabled the simulation model to successfully reproduce the actual detected anomalies. Subsequently, based on the preliminary reconstruction of the arc behavior pattern, they gradually fine-tuned the input parameters in the simulation model. After each adjustment, they verified the degree of match between the simulation results and the actual anomaly, and refined the simulation of the abnormal behavior pattern. Finally, a causal reasoning framework was applied to conduct a systematic analysis of all the variables involved, and it was determined that excessive current intensity and insufficient material thermal conductivity were the main causes of the anomalies, and the causal effects between these factors were quantified. The generated detailed cause diagnosis report recommends selecting materials with higher thermal conductivity during the design phase and strictly controlling the current intensity during the manufacturing process, which significantly improves the anti-ablation uniformity and overall quality of the product.

[0083] Optionally, the reconstruction of the preliminary arc behavior pattern gradually fine-tunes the input parameters in the multi-physics field coupling simulation model, verifies the degree of matching between the simulation results and the actual abnormal phenomenon after each adjustment, obtains the influence of different parameter changes on the degree of abnormality, refines the simulation of the abnormal behavior pattern, accurately simulates the actually observed abnormalities, and generates refined abnormal behavior pattern simulation results, including:

[0084] Using the preliminary reconstruction of the arc behavior pattern, the input parameters in the multi-physics field coupling simulation model are gradually fine-tuned, and one or more key parameters are adjusted each time to obtain an adjusted parameter combination, wherein the parameters include current intensity, voltage level, and material properties; according to the adjusted parameter combination, the simulation model is run to generate new simulation results, and a variety of quantitative indicators are used to compare and analyze the new simulation results with the actual detected abnormal phenomena and evaluate the degree of matching to obtain the matching evaluation results after each adjustment; based on the matching evaluation results after each adjustment, the parameter changes that affect the abnormal phenomena are recorded and analyzed to obtain simulation results under different parameter settings, and a detailed analysis report on the impact of parameter changes on the abnormalities is obtained by comparing the simulation results under different parameter settings; according to the detailed analysis report, the input parameters in the simulation model are fine-tuned and verified until the simulation model accurately reproduces the actual observed abnormal phenomena and generates accurate abnormal behavior pattern simulation results;

[0085] Optionally, based on the matching evaluation result after each adjustment, the parameter changes that affect the abnormal phenomenon are recorded and analyzed to obtain simulation results under different parameter settings, and by comparing the simulation results under different parameter settings, a detailed analysis report on the impact of parameter changes on the abnormality is obtained, including: based on the matching evaluation result after each adjustment, a detailed record of key parameter changes that significantly affect the abnormal phenomenon during the simulation process is recorded, and each fine-tuning and the results corresponding to each fine-tuning are accurately saved using the key parameters to obtain a detailed parameter change record, wherein the key parameters include current intensity, voltage level, and material properties; according to the detailed parameter change record, the simulation results generated after each parameter adjustment are sorted and summarized to form a complete data set, and a summary of simulation results under different parameter settings is obtained; and statistical analysis is applied and data mining technology, comparing the simulation result summaries under the different parameter settings, identifying the impact of parameter changes on the degree of abnormality, and based on the simulation result summaries under the different parameter settings, establishing a mathematical model or a machine learning model to quantify the degree of impact of each parameter on the abnormality, and obtaining a parameter impact model; using the parameter impact model to perform sensitivity analysis, generate sensitivity analysis results, and display the sensitivity analysis results in the form of charts to present the impact trends of different parameter changes on abnormal phenomena, and obtain visual charts; based on the visual charts and the sensitivity analysis results, obtaining a detailed analysis report on the impact of parameter changes on abnormal phenomena, the detailed analysis report describes the specific impact of parameter changes on abnormal phenomena and ranks the importance of each key parameter, and points out the degree of impact of each key parameter on the abnormality.

[0086] In this step, based on the preliminary reconstruction results of the arc behavior pattern, the input parameters in the multi-physics field coupling simulation model are gradually fine-tuned, and one or more key parameters are adjusted each time to generate a new parameter combination. These parameters are the key factors affecting the arc behavior. By fine-tuning them, the abnormal phenomena actually observed can be simulated more accurately. The adjusted parameter combination is used to run the simulation model to generate new simulation results, and a variety of quantitative indicators are used to compare and analyze the new simulation results with the actual detected abnormal phenomena in detail to evaluate the degree of match between the two. The matching evaluation results after each adjustment are recorded for subsequent analysis and verification to ensure that the simulation results can accurately reproduce the actual abnormal phenomena. After each adjustment, the changes in key parameters that significantly affect the abnormal phenomena are recorded and analyzed, and each fine-tuning and its corresponding results are saved to form a detailed record of parameter changes. The simulation results generated after each parameter adjustment are sorted and summarized to establish a complete data set containing simulation results under different parameter settings. Statistical analysis and data mining techniques are applied to compare the simulation results under different parameter settings, identify the impact of parameter changes on the degree of abnormality, and establish a mathematical model or machine learning model to quantify the impact of each parameter on the abnormality. Finally, a visual chart is generated through sensitivity analysis to show the impact trend of different parameter changes on abnormal phenomena, and a detailed analysis report is formed to describe the specific impact of parameter changes on abnormal phenomena, and the importance of each key parameter is ranked to indicate the degree of their impact on the abnormality.

[0087] In the embodiment of the present application, based on the preliminary arc behavior pattern reconstruction results, the technicians gradually fine-tune the input parameters in the multi-physics field coupling simulation model, adjusting one or more key parameters each time to generate a new parameter combination. Then, these parameter combinations are used to run the simulation model to generate new simulation results, and a variety of quantitative indicators are used to compare and analyze the new simulation results with the actual detected abnormal phenomena to evaluate the matching degree after each adjustment. Next, the key parameter changes that significantly affect the abnormal phenomenon after each adjustment are recorded and analyzed, and each fine-tuning and its corresponding results are saved to form a detailed parameter change record. All simulation results are sorted and summarized to establish a complete data set, and statistical analysis and data mining techniques are applied to compare the simulation results under different parameter settings to identify the impact of parameter changes on the degree of abnormality. Finally, a mathematical model or machine learning model is established to quantify the degree of influence of each parameter on the abnormality, perform sensitivity analysis, and display the influence trend of different parameter changes on the abnormal phenomenon in the form of a chart, generate a detailed analysis report, and guide subsequent design and manufacturing process adjustments.

[0088] In a power equipment manufacturing plant, technicians gradually fine-tuned the input parameters in the multi-physics coupling simulation model based on the preliminary reconstruction results of the arc behavior pattern. Each time one or more key parameters were adjusted, a new parameter combination was generated, and the simulation model was run to generate new simulation results. They used a variety of quantitative indicators to compare and analyze the new simulation results with the actual detected abnormal phenomena to evaluate the degree of match after each adjustment. The technicians recorded the changes in key parameters that significantly affected the abnormal phenomena after each adjustment, and saved each fine-tuning and its corresponding results to form a detailed record of parameter changes. Subsequently, they sorted and summarized all simulation results and established a data set containing simulation results under different parameter settings. Statistical analysis and data mining techniques were applied to compare the simulation results under different parameter settings, identify the impact of parameter changes on the degree of abnormality, and establish mathematical models or machine learning models to quantify the degree of influence of each parameter on the abnormality. Finally, the technicians conducted a sensitivity analysis and generated a visual chart to show the trend of the impact of different parameter changes on abnormal phenomena and formed a detailed analysis report. The report pointed out that current intensity and material thermal conductivity were the main factors leading to the anomaly, and put forward improvement suggestions: select materials with higher thermal conductivity and strictly control current intensity, thereby significantly improving the product's anti-ablation uniformity and overall quality.

[0089] Optionally, step 103 performs abnormality detection on the evaluation result, identifies abnormal areas with uneven anti-ablation performance, and constructs an autoencoder or variational autoencoder unsupervised learning model based on the abnormal area, learns normal patterns from data collected under normal operating conditions to accurately identify abnormal areas and generate abnormal identification results, including: using multiple quantitative indicators in the evaluation result to perform abnormality detection on the difference between the actual surface image of the static arc contact workpiece and the theoretical reference image, applying a local outlier factor or an isolation forest algorithm to identify abnormal areas indicating uneven anti-ablation performance, and obtain preliminary abnormality detection results; based on the preliminary abnormality detection results, constructs an autoencoder or variational autoencoder unsupervised learning model, learns normal patterns from data collected under normal operating conditions, and generates a trained abnormality identification model; inputs the preliminary abnormality detection result into the trained abnormality identification model, further analyzes and processes it through the trained abnormality identification model, accurately identifies the abnormal area, and uses the learning ability of the model to capture the existing minor abnormalities to generate an abnormality identification result. Optionally, the method utilizes multiple quantitative indicators in the evaluation result to perform anomaly detection on the difference between the actual surface image of the static arc contact workpiece and the theoretical reference image, applies a local outlier factor or an isolation forest algorithm to identify abnormal areas indicating uneven anti-ablation performance, and obtains preliminary anomaly detection results, including: utilizing the actual surface image of the static arc contact workpiece to perform a detailed comparative analysis between the actual surface image and a series of theoretical reference images covering a wide range of working parameter combinations to obtain an evaluation result; performing quantitative evaluation processing on the difference between the actual surface image and the theoretical reference image according to the multiple quantitative indicators in the evaluation result to generate a matching degree evaluation result, wherein the multiple quantitative indicators include a structural similarity index and a peak signal-to-noise ratio; based on the matching degree evaluation result, applying a local outlier factor or an isolation forest algorithm to perform anomaly detection processing on the difference between the actual surface image and the theoretical reference image to identify abnormal areas indicating uneven anti-ablation performance; and using the output of the LOF or isolation forest algorithm to mark the abnormal areas to generate preliminary anomaly detection results.

[0090] In this step, the difference between the actual surface image of the static arc contact workpiece and the theoretical reference image is detected by using multiple quantitative indicators in the evaluation results, and the abnormal area indicating the presence of uneven anti-ablation performance is identified by applying the local outlier factor or the isolation forest algorithm to obtain preliminary abnormality detection results. Based on these preliminary results, the autoencoder or variational autoencoder is selected and configured as the unsupervised learning model, and the model architecture and parameter settings are determined according to the needs of abnormality detection to generate a model structure suitable for the application environment. By preprocessing a large amount of actual surface image data of the static arc contact workpiece collected under normal operating conditions, a data set for training is constructed, and a detailed network architecture is designed. The AE or VAE model is trained using the preprocessed data set, and the model parameters are gradually adjusted by optimizing the loss function until convergence to generate a trained abnormality recognition model. The trained model is evaluated by an independent validation set to ensure its reconstruction effect on normal samples and the detection accuracy of known abnormal samples, and the model structure or training strategy is adjusted according to the evaluation results, and finally an optimized model that effectively distinguishes normal and abnormal situations is generated. The preliminary anomaly detection results are input into the trained anomaly recognition model for further analysis and processing to accurately identify the abnormal areas, and the model's learning ability is used to capture the tiny anomalies that exist to generate the final anomaly recognition results.

[0091] In the embodiment of the present application, based on the multiple quantitative indicators in the evaluation results, the difference between the actual surface image of the static arc contact workpiece and the theoretical reference image is detected for anomaly, and the local outlier factor or the isolation forest algorithm is used to identify the abnormal area indicating the uneven anti-ablation performance, and obtain the preliminary anomaly detection result. Then, based on the preliminary anomaly detection results, the autoencoder or variational autoencoder is selected and configured as the unsupervised learning model, and the model architecture and parameter settings are determined according to the anomaly detection requirements to generate a model structure suitable for the application environment. Then, a large amount of actual surface image data of the static arc contact workpiece collected under normal operating conditions is preprocessed to generate a data set prepared for training, and a detailed network architecture is designed. The AE or VAE model is trained using the preprocessed data set, and the loss function is optimized to gradually adjust the model parameters until convergence to generate a trained anomaly recognition model. The trained model is evaluated by an independent validation set to ensure its reconstruction effect on normal samples and the detection accuracy of known abnormal samples, and the model structure or training strategy is adjusted according to the evaluation results, and finally an optimized model that effectively distinguishes normal and abnormal situations is generated. Finally, the preliminary anomaly detection results are input into the trained anomaly recognition model for further analysis and processing to accurately identify the abnormal areas, and the model’s learning ability is used to capture the tiny anomalies that exist, generating the final anomaly recognition results.

[0092] In a power equipment manufacturing plant, technicians first used multiple quantitative indicators in the evaluation results to perform anomaly detection on the difference between the actual surface image of the static arc contact workpiece and the theoretical reference image, and applied local outlier factors or isolation forest algorithms to identify abnormal areas indicating uneven anti-ablation performance, and obtained preliminary anomaly detection results. Based on these preliminary results, they selected variational autoencoders as unsupervised learning models, determined the model architecture and parameter settings according to the needs of anomaly detection, and generated a model structure suitable for the application environment. Next, the technicians preprocessed a large number of actual surface image data of static arc contact workpieces collected under normal operating conditions, including image cropping, normalization and enhancement, generated a data set ready for training, and designed the network architecture of VAE in detail. The VAE model was trained using the preprocessed data set, and the model parameters were gradually adjusted by optimizing the loss function until convergence, generating a trained anomaly recognition model. The trained VAE model was evaluated through an independent validation set to ensure its reconstruction effect on normal samples and detection accuracy of known abnormal samples, and the model structure or training strategy was adjusted according to the evaluation results, and finally an optimized model that effectively distinguishes normal and abnormal situations was generated. Finally, the technicians input the preliminary anomaly detection results into the trained VAE model for further analysis and processing, accurately identifying the abnormal area, and using the model's learning ability to capture the tiny anomalies that exist, generating the final anomaly identification results. This process not only improves the accuracy of anomaly detection, but also provides a scientific basis for subsequent design and manufacturing processes, significantly improving product quality and reliability.

[0093] Optionally, in step 102, the actual surface image of the static arc contact workpiece is obtained, and the actual surface image is compared and analyzed according to the theoretical reference image, multiple quantitative indicators are used, and a convolutional neural network is introduced to extract high-level image features, and a scale-invariant feature transformation or a key point matching algorithm of accelerated robust features is applied to perform image registration, and an evaluation result of the degree of matching between the actual surface image and the theoretical reference image is generated, including:

[0094] The actual surface of the static arc contact workpiece is photographed by using a high-resolution imaging device to obtain high-quality surface image data and obtain the actual surface image of the static arc contact workpiece; the actual surface image is compared and analyzed according to the theoretical reference image to identify the similarities and differences between the theoretical reference image and the actual surface image, and a plurality of quantitative indicators are used to evaluate the similarity between the actual surface image and the theoretical reference image to obtain a preliminary result of the degree of matching between the two; based on deep learning technology, a convolutional neural network is introduced to perform high-level feature extraction on the actual surface image to generate a high-level feature representation of the image, and the convolutional neural network refers to automatically learning and extracting complex patterns and features in the image; scale-invariant feature transformation or accelerated robust feature algorithm is used to perform key point matching on the high-level feature representation of the image to obtain the result of key point matching; the evaluation result of the degree of matching between the actual surface image and the theoretical reference image is generated by combining the plurality of quantitative indicators, the high-level feature representation of the image and the result of the key point matching.

[0095] In this step, the actual surface of the static arc contact workpiece is photographed by a high-resolution imaging device to obtain high-quality surface image data and obtain the actual surface image of the static arc contact workpiece. According to a series of theoretical reference images covering a wide range of working parameter combinations, the actual surface image is compared and analyzed to identify the similarities and differences between the theoretical reference image and the actual surface image, and a variety of quantitative indicators are used to evaluate the similarity between the actual surface image and the theoretical reference image, and the preliminary results of the matching degree between the two are obtained. Based on deep learning technology, a convolutional neural network (CNN) is introduced to extract advanced features from the actual surface image and generate an advanced feature representation of the image. CNN can automatically learn and extract complex patterns and features in the image and capture subtle structural changes. The scale-invariant feature transform (SIFT) or accelerated robust feature (SURF) algorithm is used to match the key points of the advanced feature representation of the image to ensure the accuracy of image registration and obtain the key point matching results. Combining the results of multiple quantitative indicators, advanced feature representation of the image and key point matching, the evaluation results of the matching degree between the actual surface image and the theoretical reference image are finally generated, providing a solid foundation for subsequent anomaly detection.

[0096] In the embodiment of the present application, a high-resolution imaging device is used to shoot the actual surface of the static arc contact workpiece, obtain high-quality surface image data, and form the actual surface image of the static arc contact workpiece. Then, according to a series of theoretical reference images covering a wide range of working parameter combinations, a detailed comparative analysis of the actual surface image is performed to identify the similarities and differences between the two, and a variety of quantitative indicators are used to evaluate the similarity between the actual surface image and the theoretical reference image, and a preliminary matching degree result is obtained. Next, based on deep learning technology, a convolutional neural network is introduced to extract advanced features from the actual surface image, automatically generate an advanced feature representation of the image, and capture complex structure and texture information. Then, a scale-invariant feature transformation or an accelerated robust feature algorithm is applied to perform key point matching on the advanced feature representation of the image to ensure the accuracy of image registration and obtain the result of key point matching. Finally, the matching degree between the actual surface image and the theoretical reference image is comprehensively evaluated by combining the results of multiple quantitative indicators, advanced feature representation of the image, and key point matching, and the final evaluation result is generated, providing detailed and accurate basic data for subsequent abnormality detection.

[0097] In an electric equipment manufacturing plant, technicians first used a high-resolution industrial camera to photograph the actual surface of the static arc contact workpiece, obtained high-quality surface image data, and formed the actual surface image of the static arc contact workpiece. Then, they conducted a detailed comparative analysis of the actual surface image based on a series of theoretical reference images covering a wide range of working parameter combinations, identified the similarities and differences between the two, and used multiple quantitative indicators to evaluate the similarity between the actual surface image and the theoretical reference image, and obtained preliminary matching results. In order to further improve the evaluation accuracy, the technicians introduced a convolutional neural network based on deep learning technology to extract advanced features from the actual surface image, automatically generated an advanced feature representation of the image, and captured complex structure and texture information. Subsequently, the scale-invariant feature transformation algorithm was applied to perform key point matching on the advanced feature representation of the image, ensuring the accuracy of image registration and obtaining the key point matching results. Finally, combining the results of multiple quantitative indicators, advanced feature representation of the image, and key point matching, the technicians comprehensively evaluated the matching degree between the actual surface image and the theoretical reference image and generated the final evaluation results.

[0098] This process not only improves the accuracy and reliability of image comparison and analysis, but also provides a solid scientific basis for subsequent anomaly detection, significantly improving the product's anti-ablation uniformity and overall quality.

[0099] More specifically, the embodiment of the present application also provides a calculation formula for a structural similarity index, which is used to calculate the similarity between the actual surface image of the static arc contact workpiece and the theoretical reference image, and obtain a preliminary result of the matching degree between the two;

[0100] Among them, the specific calculation formula of similarity iSSIM(x,y) is as follows:

[0101] iSSIM(x,y)=α·SSIM(x,y)+(1-α)·(w1·LGI(x,y)+w2·PCI(x,y))

[0102] Among them, iSSIM(x,y) represents the similarity between the actual surface image x and the theoretical reference image y, α is the balancing factor that controls the weight of iSSIM and other features; LGI(x,y) is the local gradient consistency, which measures the similarity of the two images in terms of edge green and texture; PCI(x,y) is the phase consistency index, which measures the similarity of the two images in phase information; w1 and w2 are the weights of the local gradient consistency and phase consistency indexes, which ensure that the importance of different features is properly reflected.

[0103] The structural similarity index is a method to measure the similarity between two images and is widely used in image quality assessment. The traditional SSIM takes into account brightness, contrast, and structural information, but in order to more accurately capture the subtle differences between images, especially the changes in texture-rich areas or near edges, the researchers introduced an improved structural similarity index. This improved version adds local gradients and phase consistency as additional features, allowing iSSIM to more accurately reflect the human eye's perception of image changes.

[0104] The purpose of the overall formula is to improve the ability to capture subtle differences between images, especially for those information that traditional SSIM may not be able to fully express, such as edge details and textures. By combining brightness, contrast, and structural information with local gradients and phase consistency, iSSIM can provide a more comprehensive similarity evaluation metric that is consistent with the human visual system. This is especially important in applications that require high-precision image comparison, such as medical imaging, industrial inspection, etc.

[0105] The following is a brief introduction to the design reasons of each sub-item of the formula:

[0106] iSSIM(x,y)=α·SSIM(x,y)+(1-α)·(w1·LGI(x,y)+w2·PCI(x,y))

[0107] SSIM(x,y): It provides a comprehensive assessment of image brightness, contrast, and structural information. These factors are crucial to the human visual system because they directly affect our perception of image quality. However, SSIM may be insufficient when dealing with complex textures and edges, so additional features are needed to supplement its limitations.

[0108] LGI(x,y): To enhance the sensitivity to changes in image boundaries. Local gradient information reflects the edge strength in an image, which is very important for detecting significant changes in an image. By adding LGI, iSSIM is able to provide more refined similarity assessments in areas with edges and rich textures, thus making up for the shortcomings of SSIM in this regard.

[0109] PCI(x,y): Based on phase consistency, it is highly sensitive to the consistency and stability of the local structure of the image. PCI helps to capture local patterns and texture features in the image, especially in complex backgrounds. By introducing PCI, iSSIM can better identify small structural changes in the image, further improving the accuracy of the evaluation.

[0110] α, w1, w2: To balance the importance of different components. These weight parameters allow the influence of each component to be adjusted according to the needs of specific application scenarios. For example, in some applications, edge information may be more critical than structural information. In this case, the role of LGl or PCI can be emphasized by adjusting the weights. The choice of weights is usually based on experimental verification to ensure the best overall performance.

[0111] The following is a brief introduction to how to obtain the parameters of the formula:

[0112] Weight parameters (α, w1, w2): α is usually determined by cross-validation or grid search techniques. In a specific application scenario, technicians will prepare a set of standard image pairs and use different α values ​​for training and testing, and select the α value that optimizes the model performance. Weights such as w1 and w2 can also be determined by similar methods. For example, optimization algorithms such as genetic algorithms or particle swarm optimization (PSO) can be used to automatically search for the optimal weight combination so that the results of iSSIM are closest to expert annotations or the actual situation. In addition, a reasonable initial weight range can be pre-set based on domain knowledge, and then fine-tuned through experiments.

[0113] SSIM(x,y): Directly calculated using existing open source libraries. These libraries provide efficient implementations that can quickly get the SSIM value between two images. For each pair of images, just call the corresponding function to get the result.

[0114] LGI(x,y): Calculating local gradient information involves calculating the gradient of the image. Gradient information can be extracted by calculating the first-order derivative of the image (such as the Sobel operator or the Canny edge detector), and then statistical analysis (such as mean, variance, etc.) of these gradient values ​​is performed to finally form the LGI. This process can also be simplified by using functions provided by existing libraries (such as OpenCV).

[0115] PCI(x,y): The calculation of phase consistency index is relatively complex because it involves frequency domain analysis. You can use the skimage.feature.phase_congruency function in Python to calculate PCI. These toolkits have implemented efficient phase consistency calculation methods that can be directly used in practical applications.

[0116] In order to more accurately capture the subtle differences between images and ensure that the relative importance of different features in similarity assessment is reasonably reflected, the system needs to dynamically adjust the balance between brightness, contrast, structural information and local gradient and phase consistency. By introducing the balance factor α, it can be ensured that under different image conditions, the importance of each feature can be automatically adjusted according to the actual situation. Specifically, the calculation formula of the balance factor α is as follows:

[0117]

[0118] Among them, α is the balance factor, which controls the relative importance between traditional SSIM and other features (LGI and PCI); β is the adjustment parameter, which determines the steepness of the function. A larger β will make the function change more drastically, while a smaller β will make the change more gentle; σ x and σ y They represent the standard deviation of the actual surface image x and the theoretical reference image y, respectively, reflecting the contrast of the image; γ represents the threshold, which is used to set when to start paying attention to the local gradient consistency (LGI) and phase consistency index (PCI). x +σ y When it is greater than γ, the α value decreases, increasing the emphasis on LGI and PCI;

[0119] In the calculation formula of the balance factor α, the purpose of the design of this formula is to achieve adaptive weight adjustment, to ensure that the relative importance between brightness, contrast, structural information and local gradient and phase consistency under different image conditions can change dynamically according to the actual situation. The design of the balance factor aims to automatically adjust the importance of each feature according to the complexity of the image. By considering the standard deviation of the image to reflect its complexity or degree of change, this method can rely more on local gradient and phase consistency in areas with rich edges or complex textures, while paying more attention to the traditional structural similarity index in smooth areas. This method uses a nonlinear mapping function to ensure that the weight coefficient is always within a valid range and provides a smooth transition effect, making the weight adjustment more natural and smooth.

[0120] In order to more accurately capture the subtle differences between images, especially in the edges and texture-rich areas, the system needs to quantify the importance of local gradient information. By introducing the weight coefficient w1, it can ensure that the relative importance of local gradients in similarity evaluation is reasonably reflected. Specifically, the calculation formula of the local gradient consistency weight w1 is as follows:

[0121]

[0122] Among them, w1 represents the local gradient consistency weight, which reflects the importance of edge information in the image; G x (i,j) and G y (i, j) are the first-order derivatives (gradients) of the image at point (i, j) along the x and y directions, respectively, which are used to measure the edge strength of the image; ∈ represents a small constant (e.g. 10 -6 ) to avoid the situation where the denominator is zero and ensure the numerical stability of the formula;

[0123] In the calculation formula of the local gradient consistency weight w1, the purpose of this formula is to measure the ratio of the sum of the absolute values ​​of the gradients of all pixels in the image to the maximum possible sum of the gradients, so as to highlight significant edge and structural changes, while preventing the situation where the gradient is too large or too small from causing unreasonable impact on the overall evaluation results. Through normalization, it is ensured that the weight value is always within a reasonable range and can adapt to the differences in gradient distribution between different images.

[0124] Among them, in order to more accurately capture the subtle differences between images, especially in areas with complex textures and rich local structural changes, the system needs to quantify the importance of phase consistency information. By introducing the phase consistency index weight w2, it can ensure that the relative importance of phase consistency features is reasonably reflected in the similarity evaluation. Specifically, the calculation formula of the phase consistency index weight w2 is as follows:

[0125]

[0126] Where w2 represents the phase consistency index weight, which emphasizes the contribution of phase consistency in areas with high structural complexity; PC(i,j) represents the phase consistency value at the image point (i,j), which is used to measure the phase consistency of the image at that point and reflects the structural details of the image; ∈ represents a small constant (e.g. 10 -6 ) is used to avoid the situation where the denominator is zero and ensure the numerical stability of the formula.

[0127] In the calculation formula of the phase consistency index weight w2, the purpose of this formula is to measure the ratio of the sum of the phase consistency values ​​of all pixels in the image to the sum of the maximum possible phase consistency. This weight can highlight the local patterns and structural features in the image, while preventing the phase consistency from being too strong or too weak from causing unreasonable effects on the overall evaluation results. Normalization ensures that the weight value is always within a reasonable range and adapts to the differences in phase consistency distribution between different images; the introduction of a minimum value avoids division by zero errors and ensures numerical stability. Therefore, w2 can dynamically adjust the importance of phase consistency information, provide a more consistent and accurate similarity assessment, and is suitable for high-precision image comparison analysis under complex backgrounds.

[0128] Figure 2 A schematic diagram of a structural diagram of a detection and processing system for abnormal image of a static arc contact workpiece anti-ablation uniformity sample is provided for an embodiment of the present application. Figure 2 As shown, the system includes:

[0129] A modeling module 21 is used to model the physical behavior of the static arc contact workpiece under different arc conditions to obtain a multi-physics field coupling simulation model, and generate a theoretical reference image based on the multi-physics field coupling simulation model, wherein the theoretical reference image refers to an image of the expected physical behavior of the static arc contact workpiece under specific working conditions;

[0130] An acquisition module 22 is used to acquire an actual surface image of the static arc contact workpiece, compare and analyze the actual surface image according to the theoretical reference image, use multiple quantitative indicators, introduce a convolutional neural network to extract high-level image features, apply a scale-invariant feature transformation or a key point matching algorithm of an accelerated robust feature to perform image registration, and generate an evaluation result of the degree of matching between the actual surface image and the theoretical reference image, wherein the multiple quantitative indicators include a structural similarity index and a peak signal-to-noise ratio;

[0131] A detection module 23 is used to perform abnormality detection on the evaluation results, identify abnormal areas with uneven anti-ablation performance, and construct an autoencoder or variational autoencoder unsupervised learning model based on the abnormal areas, learn normal patterns from data collected under normal operating conditions, so as to accurately identify abnormal areas and generate abnormal identification results;

[0132] The reconstruction module 24 is used to reconstruct the arc behavior pattern that causes the abnormal phenomenon based on the abnormality identification result by using the reverse engineering principle combined with the particle swarm optimization algorithm, adjust the input parameters through the inversion of the multi-physical field coupling simulation model until the abnormality is reproduced, obtain preliminary cause diagnosis information, apply the causal reasoning framework to explore the causal relationship between variables, and generate a detailed cause diagnosis report, so as to make targeted adjustments to the design or manufacturing process parameters of the static arc contact workpiece through the detailed cause diagnosis report.

[0133] Figure 2 The ablation resistance uniformity sample abnormal image detection and processing system of a static arc contact workpiece can be performed Figure 1 The implementation principle and technical effect of the method for detecting abnormal image of anti-ablation uniformity sample of static arc contact workpiece described in the embodiment shown are not repeated here. The specific way in which each module and unit performs the operation of the abnormal image detection and processing system for anti-ablation uniformity sample of static arc contact workpiece in the above embodiment has been described in detail in the embodiment of the method, and will not be elaborated here.

[0134] In one possible design, Figure 2 The anti-ablation uniformity sample abnormal image detection and processing system of a static arc contact workpiece in the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0135] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0136] The processing component 32 is used to: model the physical behavior of the static arc contact workpiece under different arc conditions to obtain a multi-physics field coupling simulation model, and generate a theoretical reference image based on the multi-physics field coupling simulation model, wherein the theoretical reference image refers to an image of the expected physical behavior of the static arc contact workpiece under specific working conditions; obtain an actual surface image of the static arc contact workpiece, and compare and analyze the actual surface image according to the theoretical reference image, using multiple quantitative indicators and introducing a convolutional neural network to extract high-level image features, applying a scale-invariant feature transformation or a key point matching algorithm of an accelerated robust feature to perform image registration, and generating an evaluation result of the degree of matching between the actual surface image and the theoretical reference image, wherein the multiple quantitative indicators include a structural similarity index and a peak value signal-to-noise ratio; perform anomaly detection on the evaluation result, identify abnormal areas with uneven anti-ablation performance, and construct an autoencoder or variational autoencoder unsupervised learning model based on the abnormal area, learn the normal mode from the data collected under normal operating conditions to accurately identify the abnormal area, and generate an abnormal identification result; according to the abnormal identification result, use the reverse engineering principle in combination with the particle swarm optimization algorithm to reconstruct the arc behavior pattern that causes the abnormal phenomenon, adjust the input parameters through the inversion of the multi-physical field coupling simulation model until the anomaly is reproduced, obtain preliminary cause diagnosis information, apply the causal reasoning framework to explore the causal relationship between variables, and generate a detailed cause diagnosis report, so as to make targeted adjustments to the design or manufacturing process parameters of the static arc contact workpiece through the detailed cause diagnosis report.

[0137] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0138] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0139] Of course, the computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0140] The input / output interface provides an interface between the processing component and the peripheral interface module, which may be an output device, an input device, etc.

[0141] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0142] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0143] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The embodiment shown is a method for detecting abnormal images of a static arc contact workpiece anti-ablation uniformity sample.

[0144] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0145] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0146] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0147] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for detecting abnormal images of samples of static arc contact workpiece anti-ablation uniformity, characterized in that: include: Modeling the physical behavior of the static arc contact workpiece under different arc conditions to obtain a multi-physics field coupling simulation model, and generating a theoretical reference image based on the multi-physics field coupling simulation model, wherein the theoretical reference image refers to an image of the expected physical behavior of the static arc contact workpiece under specific working conditions; Acquire an actual surface image of the static arc contact workpiece, perform comparative analysis on the actual surface image according to the theoretical reference image, adopt multiple quantitative indices, introduce a convolutional neural network to extract high-level image features, apply a scale-invariant feature transformation or a key point matching algorithm of an accelerated robust feature to perform image registration, and generate an evaluation result of the degree of matching between the actual surface image and the theoretical reference image, wherein the multiple quantitative indices include a structural similarity index and a peak signal-to-noise ratio; Performing anomaly detection on the evaluation results to identify abnormal areas with uneven anti-ablation performance, and constructing an autoencoder or variational autoencoder unsupervised learning model based on the abnormal areas to learn normal patterns from data collected under normal operating conditions to accurately identify abnormal areas and generate abnormal identification results; According to the abnormality identification result, the arc behavior pattern that causes the abnormal phenomenon is reconstructed by utilizing the reverse engineering principle combined with the particle swarm optimization algorithm. The input parameters are adjusted by inversion through the multi-physical field coupling simulation model until the abnormality is reproduced, thereby obtaining preliminary causal diagnosis information. The causal relationship between variables is explored using a causal reasoning framework, and a detailed causal diagnosis report is generated, so that the design or manufacturing process parameters of the static arc contact workpiece can be adjusted in a targeted manner through the detailed causal diagnosis report.

2. The method according to claim 1, characterized in that According to the abnormality identification result, the reverse engineering principle is combined with the particle swarm optimization algorithm to reconstruct the arc behavior mode that causes the abnormal phenomenon, and the input parameters are adjusted by the multi-physics field coupling simulation model until the abnormality is reproduced to obtain preliminary cause diagnosis information, and the causal relationship between variables is explored by applying the causal reasoning framework to generate a detailed cause diagnosis report, including: By using the reverse engineering principle combined with the particle swarm optimization algorithm, based on the non-uniform anti-ablation performance area determined by the abnormality identification result, the key parameters affecting the arc behavior in the multi-physics field coupling simulation model are analyzed, a series of key parameters are set, and the optimal parameter combination is automatically searched through the particle swarm optimization algorithm, and the optimal parameter combination is adjusted. The multi-physics field coupling simulation model reappears the abnormal phenomenon actually detected, and a preliminary reconstruction of the arc behavior pattern is obtained; According to the preliminary reconstruction of the arc behavior pattern, the input parameters in the multi-physics field coupling simulation model are gradually fine-tuned, and after each adjustment, the matching degree between the simulation result and the actual abnormal phenomenon is verified to obtain the influence of different parameter changes on the abnormal degree, and the simulation of the abnormal behavior pattern is refined to accurately simulate the actually observed abnormality and generate a refined abnormal behavior pattern simulation result; Based on the detailed abnormal behavior pattern simulation results, a causal reasoning framework is applied to conduct a systematic causal relationship analysis of all involved variables to determine the factors that directly or indirectly cause the abnormality, and quantify the causal effects between the factors to obtain a detailed cause diagnosis report. All the involved variables include key parameters of arc behavior, material properties, simulation model parameters and other potential influencing factors.

3. The method according to claim 2, characterized in that According to the preliminary reconstruction of the arc behavior pattern, the input parameters in the multi-physics field coupling simulation model are gradually fine-tuned, and after each adjustment, the matching degree between the simulation result and the actual abnormal phenomenon is verified to obtain the influence of different parameter changes on the abnormal degree, and the simulation of the abnormal behavior pattern is refined to accurately simulate the actually observed abnormality and generate a refined abnormal behavior pattern simulation result, including: Using the preliminary arc behavior pattern reconstruction, gradually fine-tuning the input parameters in the multi-physics field coupling simulation model, adjusting one or more key parameters each time to obtain an adjusted parameter combination, wherein the parameters include current intensity, voltage level, and material properties; According to the adjusted parameter combination, the simulation model is run to generate new simulation results, and a plurality of quantitative indicators are used to compare and analyze the new simulation results with the abnormal phenomena actually detected and evaluate the matching degree, so as to obtain the matching evaluation results after each adjustment; Based on the matching evaluation results after each adjustment, the parameter changes that affect the abnormal phenomenon are recorded and analyzed to obtain simulation results under different parameter settings, and a detailed analysis report on the impact of the parameter changes on the abnormality is obtained by comparing the simulation results under the different parameter settings; According to the detailed analysis report, the input parameters in the simulation model are fine-tuned and verified until the simulation model accurately reproduces the abnormal phenomena actually observed and generates accurate abnormal behavior pattern simulation results.

4. The method according to claim 3, characterized in that Based on the matching evaluation results after each adjustment, the parameter changes that affect the abnormal phenomenon are recorded and analyzed to obtain simulation results under different parameter settings. By comparing the simulation results under different parameter settings, a detailed analysis report on the impact of parameter changes on the abnormality is obtained, including: Based on the matching evaluation results after each adjustment, the key parameter changes that significantly affect the abnormal phenomenon during the simulation process are recorded in detail, and each fine-tuning and the results corresponding to each fine-tuning are accurately saved using the key parameters to obtain a detailed parameter change record, wherein the key parameters include current intensity, voltage level, and material properties; According to the detailed parameter change records, the simulation results generated after each parameter adjustment are sorted and summarized to form a complete data set, and a summary of simulation results under different parameter settings is obtained; Apply statistical analysis and data mining techniques to compare the simulation results under the different parameter settings, identify the impact of parameter changes on the degree of abnormality, and establish a mathematical model or a machine learning model based on the simulation results under the different parameter settings to quantify the impact of each parameter on the abnormality, and obtain a parameter impact model; Perform sensitivity analysis using the parameter impact model to generate sensitivity analysis results, present the sensitivity analysis results in the form of charts, present the impact trend of different parameter changes on abnormal phenomena, and obtain visual charts; According to the visualization chart and the sensitivity analysis results, a detailed analysis report of the impact of parameter changes on abnormalities is obtained, which describes the specific impact of parameter changes on abnormal phenomena and ranks the importance of each key parameter, and points out the degree of impact of each key parameter on the abnormality.

5. The method according to claim 1, characterized in that The abnormality detection is performed on the evaluation results to identify abnormal areas with uneven anti-ablation performance, and an autoencoder or a variational autoencoder unsupervised learning model is constructed based on the abnormal areas to learn normal patterns from data collected under normal operating conditions to accurately identify abnormal areas and generate abnormality identification results, including: Using multiple quantitative indicators in the evaluation results, anomaly detection is performed on the difference between the actual surface image of the static arc contact workpiece and the theoretical reference image, and a local outlier factor or an isolation forest algorithm is applied to identify abnormal areas indicating the presence of uneven anti-ablation performance, thereby obtaining preliminary anomaly detection results; Based on the preliminary anomaly detection results, construct an autoencoder or variational autoencoder unsupervised learning model to learn normal patterns from data collected under normal operating conditions and generate a trained anomaly recognition model; The preliminary anomaly detection result is input into the trained anomaly recognition model, and further analyzed and processed by the trained anomaly recognition model to accurately identify the abnormal area, and use the learning ability of the model to capture the existing minor anomalies to generate anomaly recognition results.

6. The method according to claim 5, characterized in that The method utilizes the multiple quantitative indicators in the evaluation results to perform anomaly detection on the difference between the actual surface image of the static arc contact workpiece and the theoretical reference image, applies a local outlier factor or an isolation forest algorithm, identifies an abnormal area indicating the presence of uneven anti-ablation performance, and obtains a preliminary anomaly detection result, including: Using the actual surface image of the static arc contact workpiece, a detailed comparative analysis is performed between the actual surface image and a series of theoretical reference images covering a wide range of working parameter combinations to obtain an evaluation result; According to multiple quantitative indicators in the evaluation result, a quantitative evaluation process is performed on the difference between the actual surface image and the theoretical reference image to generate a matching degree evaluation result, wherein the multiple quantitative indicators include a structural similarity index and a peak signal-to-noise ratio; Based on the matching degree evaluation result, applying a local outlier factor or an isolation forest algorithm, performing anomaly detection processing on the difference between the actual surface image and the theoretical reference image, and identifying an abnormal area indicating the presence of non-uniform anti-ablation performance; The output of the LOF or Isolation Forest algorithm is used to mark the abnormal area and generate a preliminary abnormality detection result.

7. The method according to claim 1, characterized in that The actual surface image of the static arc contact workpiece is obtained, and the actual surface image is compared and analyzed according to the theoretical reference image, multiple quantitative indicators are used, and a convolutional neural network is introduced to extract high-level image features, and a scale-invariant feature transformation or a key point matching algorithm of an accelerated robust feature is applied to perform image registration, and an evaluation result of the degree of matching between the actual surface image and the theoretical reference image is generated, including: Using high-resolution imaging equipment, the actual surface of the static arc contact workpiece is photographed to obtain high-quality surface image data and obtain the actual surface image of the static arc contact workpiece; According to the theoretical reference image, the actual surface image is compared and analyzed to identify the similarities and differences between the theoretical reference image and the actual surface image, and a plurality of quantitative indicators are used to evaluate the similarity between the actual surface image and the theoretical reference image to obtain a preliminary result of the matching degree between the two; Based on deep learning technology, a convolutional neural network is introduced to extract high-level features from the actual surface image to generate high-level feature representation of the image, wherein the convolutional neural network automatically learns and extracts complex patterns and features in the image; Applying a scale-invariant feature transform or an accelerated robust feature algorithm to perform key point matching on the high-level feature representation of the image to obtain a key point matching result; The plurality of quantitative indicators, the high-level feature representation of the image and the key point matching results are combined to generate an evaluation result of the degree of matching between the actual surface image and the theoretical reference image.

8. A static arc contact workpiece anti-ablation uniformity sample abnormal image detection system, characterized in that: include: A modeling module is used to model the physical behavior of the static arc contact workpiece under different arc conditions to obtain a multi-physics field coupling simulation model, and generate a theoretical reference image based on the multi-physics field coupling simulation model, wherein the theoretical reference image refers to an image of the expected physical behavior of the static arc contact workpiece under specific working conditions; an acquisition module, for acquiring an actual surface image of the static arc contact workpiece, performing comparative analysis on the actual surface image according to the theoretical reference image, using a variety of quantitative indices, introducing a convolutional neural network to extract high-level image features, applying a scale-invariant feature transformation or a key point matching algorithm of an accelerated robust feature to perform image registration, and generating an evaluation result of the degree of matching between the actual surface image and the theoretical reference image, wherein the multiple quantitative indices include a structural similarity index and a peak signal-to-noise ratio; A detection module, configured to perform anomaly detection on the evaluation results, identify abnormal areas with uneven anti-ablation performance, and construct an autoencoder or variational autoencoder unsupervised learning model based on the abnormal areas, learn normal patterns from data collected under normal operating conditions, so as to accurately identify abnormal areas and generate abnormal identification results; A reconstruction module is used to reconstruct the arc behavior pattern that causes the abnormal phenomenon based on the abnormality identification result by using the reverse engineering principle combined with the particle swarm optimization algorithm, adjust the input parameters through the multi-physical field coupling simulation model inversion until the abnormality is reproduced, obtain preliminary cause diagnosis information, apply the causal reasoning framework to explore the causal relationship between variables, and generate a detailed cause diagnosis report, so as to make targeted adjustments to the design or manufacturing process parameters of the static arc contact workpiece through the detailed cause diagnosis report.

9. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a method for detecting abnormal images of static arc contact workpiece anti-ablation uniformity samples as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a computer, the abnormal image detection method of the static arc contact workpiece anti-ablation uniformity sample according to any one of claims 1 to 7 is implemented.

Images