Instrument identification model training method based on digital twinning

Abstract

The invention relates to the technical field of digital twinning, in particular to an instrument identification model training method based on digital twinning. The method comprises the following steps: acquiring an industrial automation instrument identification image, and extracting an industrial instrument visible light image according to the industrial automation instrument identification image so as to obtain an industrial instrument visible light image; performing spectrum reconstruction according to the visible light image of the industrial instrument to obtain a multispectral image; performing instrument thermal imaging image extraction according to the industrial automation instrument identification image so as to obtain an instrument thermal imaging image; detecting the abnormal temperature of the instrument according to the thermal imaging image of the instrument so as to obtain abnormal temperature data of the instrument; performing instrument X-ray image extraction according to the industrial automation instrument identification image so as to obtain an instrument X-ray image; and welding joint abnormal area detection is carried out according to the instrument X-ray image, so that welding joint abnormal area data is obtained. The prediction accuracy of the model is improved based on the digital twinning technology.

Description

Technical Field

[0001] The present invention relates to the technical field of digital twins, and in particular to an instrument identification model training method based on digital twins. Background Art

[0002] With the rapid development of the Industrial Internet and the Internet of Things, the popularity of equipment and sensors has made it possible to obtain and process real-time data. The digital twin-based instrument identification model training method builds a digital twin of the instrument and uses its real-time data to interact with the model, thereby achieving accurate identification and prediction of the instrument status. This process not only improves the management efficiency of the instrument, but also provides data support for the intelligent decision-making of the equipment. Over time, the performance of the equipment will change due to wear, fatigue or other physical factors. Traditional digital twin models are usually built based on the performance data of new equipment in the initial training stage, but these data cannot reflect the actual performance of the equipment during long-term use. In the case of aging equipment, the prediction accuracy of the model decreases significantly, resulting in the inability to effectively identify the true status of the equipment. Summary of the invention

[0003] Based on this, it is necessary for the present invention to provide an instrument identification model training method based on digital twins to solve at least one of the above technical problems.

[0004] To achieve the above purpose, a digital twin-based instrument identification model training method includes the following steps:

[0005] Step S1: acquiring an industrial automation instrument recognition image, and extracting an industrial instrument visible light image according to the industrial automation instrument recognition image, thereby obtaining an industrial instrument visible light image; and performing spectrum reconstruction according to the industrial instrument visible light image, thereby obtaining a multispectral image;

[0006] Step S2: extracting the instrument thermal imaging image according to the industrial automation instrument recognition image, thereby obtaining the instrument thermal imaging image; performing abnormal temperature detection of the instrument according to the instrument thermal imaging image, thereby obtaining abnormal temperature data of the instrument;

[0007] Step S3: extracting an instrument X-ray image according to the industrial automation instrument recognition image, thereby obtaining an instrument X-ray image; detecting abnormal areas of welding joints according to the instrument X-ray image, thereby obtaining abnormal area data of welding joints;

[0008] Step S4: performing abnormal correlation analysis on the abnormal welding joint data according to the abnormal temperature data of the instrument, thereby obtaining the abnormal instrument temperature-welding joint data; performing material abnormal characteristic analysis on the abnormal instrument temperature-welding joint data according to the multispectral image, thereby obtaining the abnormal instrument material characteristic data;

[0009] Step S5: performing instrument identification according to the visible light image of the industrial instrument to obtain instrument identification data; constructing a digital twin instrument identification model according to the instrument identification data to obtain a digital twin instrument identification model; performing instrument material intelligent adaptation on the instrument identification data according to the digital twin instrument identification model and the abnormal characteristic data of the instrument material to obtain instrument material intelligent adaptation data;

[0010] Step S6: Optimize the digital twin instrument recognition model according to the instrument material intelligent adaptation data to obtain the digital twin instrument recognition optimization model, and upload it to the instrument recognition management platform to execute the instrument recognition model training task.

[0011] The present invention can obtain a multispectral image containing rich information by acquiring a visible light image of an industrial automation instrument and performing spectral reconstruction. This helps to more comprehensively analyze the surface features and material properties of the instrument, and provides a more accurate basis for subsequent identification and abnormality detection. Multispectral images can provide more information than single-band images, thereby improving the accuracy and reliability of instrument identification. Through thermal imaging image extraction and abnormal temperature detection, the temperature distribution of the instrument can be monitored in real time, abnormal conditions such as overheating can be discovered in time, and equipment failures can be prevented. Abnormal temperature data helps to warn potential safety hazards in advance and reduce the risk of accidents. X-ray image extraction and abnormal area detection of welding joints enable the discovery of hidden defects inside the instrument, which is crucial to ensuring product quality and equipment stability. Through abnormal area data of welding joints, stricter quality control can be implemented to improve the overall quality of the product. Correlation analysis of temperature data and abnormal welding joint data helps to reveal the potential connection between the two, providing more information for fault diagnosis. Analysis of multispectral images can help identify abnormal material characteristics and provide a basis for material selection and replacement. The instrument identification model constructed through digital twin technology can realize the rapid and accurate identification of instruments and improve management efficiency. The construction of the digital twin instrument identification model provides a basis for subsequent intelligent adaptation and optimization. The digital twin model is optimized according to the actual operation data, which improves the adaptability and accuracy of the model and enables it to better reflect the actual status of the equipment. The optimized model is uploaded to the management platform, which can realize real-time update and deployment of the model and ensure that the model is always in the latest state.

[0012] Optionally, step S1 specifically includes:

[0013] Step S11: acquiring an industrial automation instrument recognition image, and extracting an industrial instrument visible light image according to the industrial automation instrument recognition image, thereby obtaining an industrial instrument visible light image;

[0014] Step S12: extracting the spectral features of the industrial instrument according to the visible light image of the industrial instrument, thereby obtaining the spectral data of the industrial instrument;

[0015] Step S13: performing statistics on the near-infrared band range of the industrial instrument according to the spectrum data of the industrial instrument, thereby obtaining the near-infrared band range data of the industrial instrument;

[0016] Step S14: constructing a spectrum reconstruction model according to the spectrum data of the industrial instrument and the near-infrared band range data of the industrial instrument, thereby obtaining a spectrum reconstruction model;

[0017] Step S15: performing pixel-by-pixel spectral reflectance analysis on the visible light image of the industrial instrument according to the spectral reconstruction model, thereby obtaining pixel-by-pixel spectral reflectance data;

[0018] Step S16: reconstructing a multispectral image according to the pixel-by-pixel spectral reflectance data, thereby obtaining a multispectral image.

[0019] The present invention obtains industrial automation instrument recognition images as the first step of the whole process, laying the foundation for subsequent analysis. Through accurate image recognition, the accuracy and reliability of subsequent processing can be ensured. By extracting visible light images from the recognition images, the appearance characteristics of the instrument can be focused on, which is crucial for the identification and classification of the instrument. Extracting spectral features from visible light images can help identify the material composition and surface characteristics of the instrument. This process helps to establish basic data for instrument identification, thereby improving the depth of subsequent analysis. Spectral data provides support for understanding the optical properties of different materials and can help identify the material type of the instrument under various conditions. By performing statistics on the near-infrared band range of spectral data, specific spectral features and wavelength ranges can be identified, which is very important for material identification and quality control of the instrument. The obtained near-infrared band range data can provide important information for subsequent analysis and enhance the accuracy and reliability of the model. Combining spectral data and near-infrared band range data to construct a spectral reconstruction model can achieve a comprehensive understanding of the spectral characteristics of the instrument. This model provides a theoretical basis for subsequent pixel-by-pixel analysis. Through the scientifically constructed spectral reconstruction model, the error in the analysis process can be effectively reduced and the credibility of the results can be improved. By using the spectral reconstruction model to perform pixel-by-pixel spectral reflectance analysis on visible light images, the spectral characteristics of each pixel can be obtained, providing more detailed analysis results. This pixel-by-pixel analysis method can help identify small changes in materials, thereby improving the ability to monitor the status of the instrument. By reconstructing multispectral images with pixel-by-pixel reflectance data, images containing information from multiple bands can be generated, thereby providing a more comprehensive description of instrument characteristics. Multispectral images can provide rich information for subsequent instrument identification and characteristic analysis, helping to improve the accuracy and comprehensiveness of instrument identification.

[0020] Optionally, step S2 specifically includes:

[0021] Step S21: extracting the instrument thermal imaging image according to the industrial automation instrument recognition image, thereby obtaining the instrument thermal imaging image;

[0022] Step S22: performing pixel temperature mapping according to the instrument thermal imaging image, thereby obtaining instrument thermal imaging temperature data;

[0023] Step S23: drawing an instrument thermal imaging temperature spectrum diagram according to the instrument thermal imaging temperature data, thereby obtaining an instrument thermal imaging temperature spectrum diagram;

[0024] Step S24: performing temperature frequency statistics according to the instrument thermal imaging temperature spectrum diagram, thereby obtaining low-frequency instrument thermal imaging temperature data;

[0025] Step S25: Acquire rated temperature data of industrial automation equipment;

[0026] Step S26: performing abnormal temperature comparison on the low-frequency instrument thermal imaging temperature data according to the rated temperature data of the industrial automation instrument, thereby obtaining abnormal temperature data of the instrument.

[0027] The present invention can visually observe the temperature distribution of the instrument by extracting a thermal imaging image from an industrial automation instrument recognition image. This provides basic data for subsequent temperature analysis. Thermal imaging technology can capture the temperature changes of the instrument in real time, helping the operator to promptly discover the abnormal state of the equipment. By performing pixel temperature mapping on the thermal imaging image, the specific temperature value of each pixel can be obtained, thereby realizing accurate quantification of the thermal state of the instrument. This detailed temperature mapping can reveal the unevenness of the surface temperature of the instrument, help identify the existing hot spots or cold spots, and guide subsequent maintenance and detection. Drawing a temperature spectrum diagram based on the instrument thermal imaging temperature data can visualize the state of the instrument under different temperature conditions. This graphical representation helps to understand the thermal characteristics of the equipment. The temperature spectrum diagram can show the trend of temperature changes and provide data support for fault prediction and maintenance decisions. By performing temperature frequency statistics on the temperature spectrum diagram, the normal operating temperature range of the instrument and its deviation can be identified. This helps to understand the performance of the instrument under different working conditions. After identifying the low-frequency instrument thermal imaging temperature data, it can provide an important basis for subsequent analysis and help identify potential fault risks. Obtaining the rated temperature data of industrial automation instruments is an important basis for abnormal analysis. This data provides a standard reference for subsequent temperature comparisons. The rated temperature data helps to compare the actual temperature with the design specifications to ensure that the equipment operates within a safe range. By comparing the low-frequency instrument thermal imaging temperature data with the rated temperature data, the abnormal temperature state of the equipment can be effectively identified. This process is a key link in fault diagnosis and maintenance. The abnormal temperature data obtained from the instrument can timely warn of equipment failures, reduce the risk of equipment damage caused by abnormal temperature, and thus improve the reliability and safety of the equipment.

[0028] Optionally, step S3 specifically includes:

[0029] Step S31: extracting an instrument X-ray image according to the industrial automation instrument recognition image, thereby obtaining an instrument X-ray image;

[0030] Step S32: performing grayscale threshold image region segmentation according to the instrument X-ray image, thereby obtaining an X-ray grayscale threshold region image;

[0031] Step S33: performing welding joint crack region analysis on the X-ray grayscale threshold region image, thereby obtaining welding joint crack region data;

[0032] Step S34: performing welding joint pore area analysis on the X-ray grayscale threshold area image, thereby obtaining welding joint pore area data;

[0033] Step S35: Merge the number of abnormal areas of the weld joint according to the crack area data of the weld joint and the pore area data of the weld joint, so as to obtain the abnormal area data of the weld joint.

[0034] The present invention can obtain detailed information on the internal structure of the instrument by extracting X-ray images from the industrial automation instrument recognition image. This provides basic data for subsequent defect detection and welding quality assessment. X-ray technology can penetrate materials to visualize internal defects (such as cracks, pores, etc.), thereby achieving a comprehensive assessment of the quality of welded joints. By performing grayscale threshold segmentation on X-ray images, regions of interest (such as welded joints) can be extracted from complex images, which helps to improve the accuracy of subsequent analysis. Regional segmentation can effectively reduce the interference of background noise and other irrelevant information, making subsequent defect analysis more centralized and efficient. Performing welded joint crack region analysis on X-ray grayscale threshold region images can effectively identify crack defects in welded joints. This process is crucial to ensuring welding quality. By timely detecting cracks, structural failure caused by cracks can be prevented, thereby improving the overall reliability and safety of the equipment. Performing welded joint pore region analysis on X-ray grayscale threshold region images helps to detect pore defects in the welding process. These pores reduce the strength of welded joints. Through the detection of pores, data support can be provided for welding quality control to ensure that the product meets industry standards. Combining the crack area data and the pore area data of the welded joint can comprehensively evaluate the abnormal conditions of the welded joint. This provides an important basis for subsequent fault analysis and treatment. Through combined analysis, the overall quality status of the welded joint can be better understood, providing guidance for the formulation of maintenance and improvement measures.

[0035] Optionally, step S33 is specifically:

[0036] Step S331: setting the crack region grayscale threshold according to the X-ray grayscale threshold region image, thereby obtaining the crack region grayscale threshold;

[0037] Step S332: dividing the X-ray grayscale threshold region image into crack regions according to the crack region grayscale threshold, thereby obtaining a crack region image;

[0038] Step S333: classifying the crack types according to the crack region image, thereby obtaining a hot crack image and a cold crack image;

[0039] Step S334: performing thermal crack roughness calculation on the thermal crack image to obtain thermal crack roughness data;

[0040] Step S335: Calculating the cold crack smoothness of the cold crack image to obtain cold crack smoothness data;

[0041] Step S336: dividing the crack region image into weld joint crack regions according to the hot crack roughness data and the cold crack smoothness data, thereby obtaining hot crack region data and cold crack region data;

[0042] Step S337: performing high-frequency regional distribution statistics according to the hot crack regional data and the cold crack regional data, so as to obtain the hot crack high-frequency regional data and the cold crack high-frequency regional data;

[0043] Step S338: Merge the welding joint area according to the hot crack high-frequency area data and the cold crack high-frequency area data, so as to obtain the welding joint area data.

[0044] The present invention can accurately identify the crack area in the X-ray image by setting the grayscale threshold of the crack area. This helps to reduce misjudgment and missed judgment and improve the accuracy of crack detection. By dividing the crack area of ​​the X-ray image according to the set grayscale threshold, a clear crack area image can be obtained. This provides a basis for subsequent crack type classification and characteristic analysis. Through the crack area image, the crack type can be classified, such as hot cracks and cold cracks. This classification helps to understand the causes and characteristics of crack formation more deeply and provide guidance for subsequent processing. The roughness calculation of the hot crack image can obtain the surface characteristic data of the hot crack. These data help to evaluate the severity and impact of the crack. The smoothness calculation of the cold crack image can obtain the surface smoothness data of the cold crack. This helps to judge the expansion trend and potential risks of the crack. According to the characteristic data of the hot crack and the cold crack, the crack area image is further divided to obtain the hot crack area data and the cold crack area data. This helps to formulate maintenance and improvement measures in a targeted manner. By performing high-frequency regional distribution statistics on the hot crack and cold crack area data, high-frequency characteristic data of the crack area can be obtained. These data help to reveal the distribution law and characteristics of cracks. By merging the weld joint area based on the hot crack high-frequency area data and the cold crack high-frequency area data, comprehensive weld joint area data can be obtained. This helps to evaluate the quality status of the weld joint as a whole.

[0045] Optionally, step S34 is specifically:

[0046] Step S341: setting the grayscale threshold of the pore area on the X-ray grayscale threshold area image, thereby obtaining the grayscale threshold of the pore area;

[0047] Step S342: dividing the X-ray grayscale threshold region image into pore regions according to the pore region grayscale threshold, thereby obtaining a pore region image;

[0048] Step S343: performing pore position area statistics according to the pore area image, thereby obtaining high-frequency pore position area data and low-frequency pore position area data;

[0049] Step S344: performing statistics on the shapes and sizes of pores according to the pore area image, thereby obtaining large pore area data and small pore area data;

[0050] Step S345: performing a high-risk area intersection operation according to the high-frequency pore position area data and the large pore area data, thereby obtaining high-risk pore area data;

[0051] Step S346: performing a low-risk area intersection operation according to the low-frequency pore position area data and the small pore area data, thereby obtaining low-risk pore area data;

[0052] Step S347: Merge the weld joint porosity risk areas according to the high-risk porosity area data and the low-risk porosity area data, thereby obtaining the weld joint porosity area data.

[0053] The present invention can accurately identify the pore area in the X-ray image by setting the grayscale threshold of the pore area. This process helps to reduce misjudgment and ensure the accuracy of subsequent pore detection. By dividing the X-ray image into pore areas according to the set grayscale threshold, a clear pore area image can be obtained. This provides a basis for subsequent pore statistics and analysis. According to the pore area image, high-frequency and low-frequency statistics of the pore position can be performed to obtain detailed information on the pore distribution. This helps to identify dense areas of pores and their potential risks. By performing statistics on the shape and size of the pores, regional data of large pores and small pores can be obtained. These data help to evaluate the impact of pores on the strength of welded joints. According to the intersection operation of high-frequency pore position area data and large pore area data, high-risk pore areas in welded joints can be identified. This process helps to timely discover the key areas that cause structural failure. According to the intersection operation of low-frequency pore position area data and small pore area data, low-risk pore areas in welded joints can be identified. This helps to distinguish pores of different hazard levels. By combining high-risk porosity areas with low-risk porosity areas, comprehensive porosity area data of welded joints can be obtained. This helps to build a complete risk assessment model to support subsequent maintenance and treatment decisions.

[0054] Optionally, step S4 is specifically:

[0055] Step S41: Obtaining time distribution data of abnormal temperature of the instrument and time distribution data of abnormal welding joint according to time distribution statistics of abnormal temperature of the instrument and abnormal welding joint;

[0056] Step S42: Calculating the time distribution overlap data according to the abnormal temperature time distribution data of the instrument and the abnormal time distribution data of the welding joint, thereby obtaining the time distribution overlap data;

[0057] Step S43: extracting instrument temperature-welding joint abnormality features according to the time distribution overlapping data, thereby obtaining instrument temperature-welding joint abnormality data;

[0058] Step S44: calculating the material grain coarsening probability for the instrument temperature-welding joint abnormality data according to the multispectral image, thereby obtaining the material grain coarsening probability data;

[0059] Step S45: calculating the material phase change probability of the instrument temperature-welding joint abnormality data according to the multispectral image, thereby obtaining the material phase change probability data;

[0060] Step S46: Integrate the instrument material abnormality characteristics according to the material grain coarsening probability data and the material phase change probability data, so as to obtain the instrument material abnormality characteristic data.

[0061] The present invention can obtain the occurrence frequency and time pattern of two abnormal events by performing time distribution statistics on abnormal temperature data of the instrument and abnormal data of the welding joint. This provides a basis for understanding the occurrence law of abnormal events. The acquired time distribution data can help identify the potential association between temperature anomaly and welding joint anomaly, laying the foundation for subsequent analysis. By calculating the overlap of abnormal temperature time distribution data of the instrument and abnormal time distribution data of the welding joint, abnormal events occurring at the same time can be identified. This helps to reveal the relationship between temperature change and welding joint defects. The analysis of overlapping data can provide early warning of potential failures, helping engineers to take timely measures to reduce the risk of failure. According to the time distribution overlapping data, feature extraction of instrument temperature and welding joint anomaly can be performed to obtain richer abnormal data. This helps to comprehensively evaluate the operating status of the equipment. The extracted abnormal data can provide a basis for further analysis and help identify the impact of temperature anomaly on the quality of welding joints. The probability of grain coarsening is calculated for instrument temperature and welding joint abnormal data through multispectral images, and the grain change of the material under abnormal temperature conditions can be evaluated. This is crucial for the evaluation of material properties. The coarsening of grains affects the strength and toughness of the material. Timely identification of this change can help predict the performance degradation of the material. By calculating the probability of material phase change for abnormal data through multispectral images, the phase change behavior of materials under specific conditions can be evaluated. This is crucial to understanding the thermal stability and reliability of materials. Phase change probability data can provide a basis for material selection and design, ensuring the adaptability of materials under different working conditions. By integrating grain coarsening probability data and phase change probability data, comprehensive instrument material abnormality characteristic data can be obtained. This helps to comprehensively evaluate the performance changes and safety of materials. The integrated data can provide more accurate information for fault diagnosis and maintenance decisions, helping engineers better understand the performance of materials under abnormal working conditions.

[0062] Optionally, step S5 specifically includes:

[0063] Step S51: performing instrument identification according to the visible light image of the industrial instrument, thereby obtaining instrument identification data;

[0064] Step S52: constructing a digital twin instrument identification model according to the instrument identification data, thereby obtaining a digital twin instrument identification model;

[0065] Step S53: classifying the instruments according to the instrument identification data, thereby obtaining metal material instrument data and composite material instrument data;

[0066] Step S54: matching the metal material instrument data with abnormal metal materials according to the instrument material abnormal characteristic data, thereby obtaining abnormal metal material data;

[0067] Step S55: matching composite material instrument data with abnormal composite materials according to instrument material abnormal characteristic data, thereby obtaining abnormal composite material data;

[0068] Step S56: performing intelligent material adaptation on the metal material instrument and the composite material instrument respectively according to the abnormal metal material data and the abnormal composite material data, so as to obtain instrument material intelligent adaptation data.

[0069] The present invention can obtain basic information and characteristic data of the instrument by identifying the visible light image of the industrial instrument. This provides basic data for subsequent analysis and processing. By using image processing and computer vision technology, the accuracy of instrument identification can be improved and misidentification can be reduced. A digital twin instrument identification model is constructed according to the instrument identification data, and the instrument in the physical world can be combined with its virtual model. This model can reflect the state of the physical instrument in real time. The digital twin model can be used for simulation and prediction, which helps to optimize equipment management and maintenance decisions and improve efficiency. According to the classification of instrument identification data, the instrument can be divided into metal material instruments and composite material instruments. This helps to carry out targeted management and maintenance. Through classification, resources can be more effectively allocated and corresponding maintenance strategies can be formulated to improve management efficiency. Matching metal material instruments using instrument material abnormal characteristic data helps to identify metal materials with potential abnormalities. This provides a basis for subsequent fault diagnosis. By timely discovering abnormal metal materials, the risk of equipment failure caused by material problems can be reduced and the safety of the equipment can be improved. Matching composite material instruments according to instrument material abnormal characteristic data helps to identify abnormal composite materials. This also provides important information for fault diagnosis and maintenance. By detecting abnormalities in composite materials, the reliability of materials in specific working environments can be improved to ensure the normal operation of equipment. Intelligent adaptation of the instrument based on abnormal metal material data and abnormal composite material data can achieve dynamic adjustment and optimization of materials. This helps to improve the overall performance of the instrument. Through intelligent adaptation, the instrument can adjust material properties according to actual working conditions, thereby improving its adaptability and flexibility and optimizing overall operating efficiency.

[0070] Optionally, step S55 is specifically:

[0071] Step S551: Calculating thermal stability data of the composite material according to the instrument data of the composite material, thereby obtaining thermal stability data of the composite material;

[0072] Step S552: dividing the composite material instrument data into instrument thermal stability regions according to the composite material thermal stability data, thereby obtaining composite material instrument thermal stability region data;

[0073] Step S553: ​​extracting material phase change characteristics according to the instrument material abnormal characteristic data, thereby obtaining material phase change data;

[0074] Step S554: performing composite material instrument phase change analysis on composite material instrument thermal stability region data according to the material phase change data, thereby obtaining composite material instrument phase change data;

[0075] Step S555: clustering abnormal data of composite material instrument data according to composite material instrument phase change data, thereby obtaining abnormal composite material data.

[0076] The present invention can obtain the performance of composite materials at different temperatures by calculating the thermal stability of composite material instrument data. This provides a basis for understanding the stability of materials in a working environment. Thermal stability data helps to evaluate the performance of composite materials under high temperature and changing conditions, and provides a scientific basis for material selection and application. According to the thermal stability data, the thermal stability area of ​​the composite material instrument is divided, and the applicable range of the material under different temperature conditions can be clearly identified. This helps to optimize the use conditions of the material. By dividing the thermal stability area, the temperature range that causes the material failure can be identified in time, thereby improving the safety and reliability of the equipment. According to the abnormal characteristic data of the instrument material, the material phase change feature extraction is performed, and the phase change behavior data of the composite material under temperature change can be obtained. This is crucial for understanding the thermal response characteristics of the material. Through the extracted phase change data, a basis can be provided for the design and application of the material to ensure the performance of the material under specific conditions. According to the material phase change data, the thermal stability area of ​​the composite material instrument is analyzed for phase change, and the phase change characteristics of the material under different working conditions can be deeply understood. This helps to evaluate the performance and adaptability of the material. The phase change analysis can reveal the behavior of the material at high temperature or rapid temperature change, and provide support for the reliability evaluation of the material. According to the phase change data of the composite material instrument, abnormal data clustering is performed, and composite material data that behaves abnormally under specific conditions can be identified. This process helps detect potential problems. Through cluster analysis, abnormal material performance can be discovered in a timely manner, reducing the risk of equipment failure caused by material problems.

[0077] Optionally, step S54 is specifically:

[0078] Step S541: Calculate the grain size according to the metal material instrument data, thereby obtaining the metal material instrument grain size data;

[0079] Step S542: performing grain distribution statistics according to the metal material instrument grain size data, thereby obtaining the metal material instrument grain size distribution data;

[0080] Step S543: extracting material grain coarsening characteristics according to the instrument material abnormal characteristic data, thereby obtaining material grain coarsening data;

[0081] Step S544: performing a grain coarsening comparison on the metal material instrument grain size distribution data according to the material grain coarsening data, thereby obtaining the metal material instrument grain coarsening distribution data;

[0082] Step S545: clustering abnormal grain coarsening of the metal material instrument data according to the metal material instrument grain coarsening distribution data, thereby obtaining abnormal metal material data.

[0083] The present invention can obtain the grain characteristic data of the metal material by calculating the grain size of the metal material instrument data. This provides basic data for subsequent analysis. The grain size is an important factor affecting the mechanical properties of the metal material. By obtaining the grain size data, a basis can be provided for the performance evaluation of the material such as strength and toughness. According to the distribution statistics of the grain size data of the metal material instrument, the distribution of the grain size can be obtained. This helps to understand the microstructure and uniformity of the material. By analyzing the distribution of the grain size, guidance can be provided for the design and production of the material to ensure the performance of the material in a specific application. According to the material abnormal characteristic data, the grain coarsening characteristics are extracted to identify the situation where the grains are coarsened under specific conditions. This is crucial to understand the performance of the material under extreme conditions. Through the extracted coarsening characteristics, the stability and safety of the material under different working environments can be evaluated, thereby reducing the risk of failure. By comparing the material grain coarsening data with the grain size distribution data, the trend and distribution of grain coarsening can be identified. This helps to discover potential material defects. This comparison can help engineers discover abnormalities in the material structure in advance and reduce the risk of equipment failure caused by material problems. By performing abnormal cluster analysis based on the grain coarsening distribution data of metal material instruments, metal materials with abnormal coarsening can be identified. This provides an important basis for fault diagnosis. Through cluster analysis, potential problems of materials can be discovered in a timely manner, so that preventive measures can be taken to reduce the occurrence of equipment failures. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments thereof made with reference to the following drawings:

[0085] Figure 1 It is a schematic diagram of the steps of the instrument identification model training method based on digital twins of the present invention;

[0086] Figure 2 Detailed step flow diagram of step S1 in the present invention;

[0087] Figure 3 Detailed step flow diagram of step S2 in the present invention;

[0088] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0089] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0090] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0091] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0092] To achieve this, please refer to Figures 1 to 3 The present invention provides an instrument identification model training method based on digital twins, the method comprising the following steps:

[0093] Step S1: acquiring an industrial automation instrument recognition image, and extracting an industrial instrument visible light image according to the industrial automation instrument recognition image, thereby obtaining an industrial instrument visible light image; and performing spectrum reconstruction according to the industrial instrument visible light image, thereby obtaining a multispectral image;

[0094] In this embodiment, a high-resolution industrial camera is used to capture the recognition image of the industrial automation instrument, and the image resolution should reach 1920x1080 pixels to ensure clear details. Then, the visible light image of the instrument is extracted and the background noise is removed through image processing algorithms (such as edge detection and morphological processing). Then, a multi-spectral camera is used to reconstruct the spectrum and obtain the reflection spectrum data of different bands (such as 400nm to 700nm). These data will be used to generate high-quality multi-spectral images with a resolution of 640x480 pixels, which is convenient for subsequent material property analysis.

[0095] Step S2: extracting the instrument thermal imaging image according to the industrial automation instrument recognition image, thereby obtaining the instrument thermal imaging image; performing abnormal temperature detection of the instrument according to the instrument thermal imaging image, thereby obtaining abnormal temperature data of the instrument;

[0096] In this embodiment, an infrared thermal imager is used to obtain a thermal imaging image of the instrument, and the resolution of the thermal imaging is 320x240 pixels. The thermal imaging image is processed by image processing software to highlight the abnormal temperature area. Using the threshold segmentation method, the temperature threshold (such as more than 70°C) is set to detect the abnormal temperature area and generate the abnormal temperature data of the instrument. Finally, the abnormal temperature data is recorded in CSV format for subsequent analysis to ensure the traceability and easy processing of the data.

[0097] Step S3: extracting an instrument X-ray image according to the industrial automation instrument recognition image, thereby obtaining an instrument X-ray image; detecting abnormal areas of welding joints according to the instrument X-ray image, thereby obtaining abnormal area data of welding joints;

[0098] In this embodiment, an X-ray image of the instrument is obtained using an X-ray imaging device with a resolution of not less than 2048x1536 pixels to ensure that details are recognizable. Then, a machine learning model is applied to analyze the X-ray image, and a convolutional neural network (CNN) algorithm is used to identify abnormal areas of the weld joint. By training the model, a loss function is set to optimize the detection accuracy, and finally the data of the abnormal area of ​​the weld joint is output to ensure the accuracy and reliability of the identification.

[0099] Step S4: performing abnormal correlation analysis on the abnormal welding joint data according to the abnormal temperature data of the instrument, thereby obtaining the abnormal instrument temperature-welding joint data; performing material abnormal characteristic analysis on the abnormal instrument temperature-welding joint data according to the multispectral image, thereby obtaining the abnormal instrument material characteristic data;

[0100] In this embodiment, based on the abnormal temperature data of the instrument and the abnormal data of the welding joint, a multivariate statistical analysis method is used to perform abnormal correlation analysis, and the significance level (such as p<0.05) is set to detect the relationship between different abnormalities. Then, the material abnormality characteristics of the associated data are analyzed based on the multispectral image to extract the spectral characteristics of the material (such as reflectivity, absorptivity, etc.). Finally, the formed instrument material abnormality characteristics data will be used for subsequent intelligent adaptation analysis.

[0101] Step S5: performing instrument identification according to the visible light image of the industrial instrument to obtain instrument identification data; constructing a digital twin instrument identification model according to the instrument identification data to obtain a digital twin instrument identification model; performing instrument material intelligent adaptation on the instrument identification data according to the digital twin instrument identification model and the abnormal characteristic data of the instrument material to obtain instrument material intelligent adaptation data;

[0102] In this embodiment, computer vision algorithms are used to identify visible light images of industrial instruments to identify the model and specifications of the instruments. Then, based on the identification data, a digital twin model is constructed, and 3D modeling software (such as SolidWorks) is used to build the model. The model will incorporate material abnormality property data to achieve intelligent material adaptation. The intelligent adaptation system will use a rule engine to analyze material properties, adapt the optimal material, and ensure reliability under specific working conditions.

[0103] Step S6: Optimize the digital twin instrument recognition model according to the instrument material intelligent adaptation data to obtain the digital twin instrument recognition optimization model, and upload it to the instrument recognition management platform to execute the instrument recognition model training task.

[0104] In this embodiment, the digital twin model is optimized based on the instrument material intelligent adaptation data, and the model parameters are adjusted using an optimization algorithm (such as a genetic algorithm) to improve the accuracy and response speed of the model. The optimized digital twin instrument identification model will be uploaded to the instrument identification management platform, and the API interface will be used to implement real-time model updates and scheduling of training tasks. The platform will monitor model performance and retrain regularly to ensure the effectiveness and accuracy of the model.

[0105] Optionally, step S1 specifically includes:

[0106] Step S11: acquiring an industrial automation instrument recognition image, and extracting an industrial instrument visible light image according to the industrial automation instrument recognition image, thereby obtaining an industrial instrument visible light image;

[0107] In this embodiment, a high-resolution industrial camera (with a resolution of at least 1920x1080 pixels) is used to capture the recognition images of industrial automation instruments. These images should be taken under good lighting conditions to clearly show the details of the instrument. Next, the recognition images are processed using image processing software (such as OpenCV), and the visible light image of the instrument is extracted through an edge detection algorithm (such as the Canny algorithm) to remove background noise. Finally, the generated visible light image of the industrial instrument will be used for subsequent spectral analysis to ensure that the image quality is high enough to support accurate feature extraction.

[0108] Step S12: extracting the spectral features of the industrial instrument according to the visible light image of the industrial instrument, thereby obtaining the spectral data of the industrial instrument;

[0109] In this embodiment, based on the obtained visible light image of the industrial instrument, the spectral characteristics of the instrument are extracted using a spectral analysis algorithm. This process first requires converting the visible light image into spectral data, and processing the image using a standard spectral decomposition method (such as principal component analysis PCA). By sampling the spectral information of different regions in the image, key spectral features are extracted, including reflectivity, transmittance, etc. Ultimately, the generated industrial instrument spectral data should include reflectivity values ​​of multiple bands for subsequent analysis and modeling.

[0110] Step S13: performing statistics on the near-infrared band range of the industrial instrument according to the spectrum data of the industrial instrument, thereby obtaining the near-infrared band range data of the industrial instrument;

[0111] In this embodiment, the response characteristics of the instrument in the near-infrared spectrum (e.g., 700nm to 2500nm) are determined by performing near-infrared band statistics on the spectral data of the industrial instrument. The spectral data is analyzed using statistical analysis software (such as Matlab), and a threshold is set to identify effective spectral information within a specific wavelength range. A relationship diagram between wavelength and reflectivity is generated by a visualization tool, and the start and end points of the near-infrared band are marked. Ultimately, the obtained near-infrared band range data of the industrial instrument will provide the necessary parameters for the subsequent construction of the spectral reconstruction model.

[0112] Step S14: constructing a spectrum reconstruction model according to the spectrum data of the industrial instrument and the near-infrared band range data of the industrial instrument, thereby obtaining a spectrum reconstruction model;

[0113] In this embodiment, a spectral reconstruction model is constructed based on industrial instrument spectral data and near-infrared band range data. Machine learning techniques (such as random forests or support vector machines) are used for modeling, and spectral features extracted from multiple samples are used when training the model. The model training set should contain samples of various materials and different spectral conditions to improve the generalization ability of the model. Ultimately, the generated spectral reconstruction model can predict the corresponding near-infrared spectral features after the visible light image is input.

[0114] Step S15: performing pixel-by-pixel spectral reflectance analysis on the visible light image of the industrial instrument according to the spectral reconstruction model, thereby obtaining pixel-by-pixel spectral reflectance data;

[0115] In this embodiment, the spectral reconstruction model is used to perform pixel-by-pixel spectral reflectance analysis on the visible light image of the industrial instrument. The pixel-by-pixel analysis method is adopted, and the spectral reconstruction model is input for the RGB value of each pixel to calculate its corresponding spectral reflectance. This process is implemented using the Python programming language and its related libraries (such as NumPy and SciPy) to ensure that the spectral reflectance data of each pixel is accurately calculated and stored. Ultimately, the obtained pixel-by-pixel spectral reflectance data will lay the foundation for the reconstruction of multispectral images.

[0116] Step S16: reconstructing a multispectral image according to the pixel-by-pixel spectral reflectance data, thereby obtaining a multispectral image.

[0117] In this embodiment, a multispectral image is reconstructed based on pixel-by-pixel spectral reflectance data. Image processing software (such as ImageJ) is used to convert the pixel-by-pixel reflectance data into a multispectral image format to ensure that the spatial resolution of the image is consistent with the source visible light image. By normalizing the reflectance value of each pixel, it is ensured that the multispectral image finally generated can truly reflect the spectral characteristics of the material. The reconstructed multispectral image will be used for subsequent material property analysis and anomaly detection to ensure its high availability and reliability.

[0118] Optionally, step S2 specifically includes:

[0119] Step S21: extracting the instrument thermal imaging image according to the industrial automation instrument recognition image, thereby obtaining the instrument thermal imaging image;

[0120] In this embodiment, a high-definition recognition image of an industrial automation instrument is obtained. An industrial camera or similar high-resolution photographic equipment is used to capture images of the instrument under normal working conditions. These images should contain all the appearance features of the instrument to ensure that the different parts of the instrument can be accurately located and identified. These recognition images are then input into a thermal imaging camera for shooting. The thermal imaging camera has an infrared sensor that can detect and record the temperature distribution on the surface of an object. The camera should be set to an appropriate resolution and frame rate when shooting to ensure the clarity and detail of the thermal imaging image. The generated thermal imaging image should be saved in a standard image file format (such as TIFF or JPEG), and ensure that the image contains sufficient temperature gradients for subsequent processing.

[0121] Step S22: performing pixel temperature mapping according to the instrument thermal imaging image, thereby obtaining instrument thermal imaging temperature data;

[0122] In this embodiment, the acquired instrument thermal imaging image is processed using image processing software (such as MATLAB, LabVIEW or dedicated thermal imaging analysis software). The thermal imaging image is imported into the software, and the software uses the temperature data provided by the infrared sensor to assign a corresponding temperature value to each pixel. Through the color mapping function of the thermal imaging image, different temperature ranges can be mapped to different colors. Next, the temperature value of each pixel in the image is extracted and converted into a temperature data array. This temperature data array contains the temperature distribution of the entire instrument surface when shooting. These data are saved in numerical form for further analysis.

[0123] Step S23: drawing an instrument thermal imaging temperature spectrum diagram according to the instrument thermal imaging temperature data, thereby obtaining an instrument thermal imaging temperature spectrum diagram;

[0124] In the present embodiment, data analysis software (such as NumPy and Matplotlib libraries in Python) is used to perform spectrum analysis on temperature data. First, a one-dimensional or two-dimensional temperature matrix is ​​extracted from the temperature data obtained in step S22. Then, Fourier transform is performed on the temperature data to obtain its frequency domain representation. Fourier transform can convert the data in the time domain into the data in the frequency domain, thereby revealing the periodic changes in the temperature data. The spectrum diagram shows the temperature variation intensity under different frequencies. When drawing the spectrum diagram, use appropriate frequency range and resolution to clearly show the relationship between frequency and temperature variation. The spectrum diagram is saved as an image file for subsequent analysis.

[0125] Step S24: performing temperature frequency statistics according to the instrument thermal imaging temperature spectrum diagram, thereby obtaining low-frequency instrument thermal imaging temperature data;

[0126] In this embodiment, the frequency statistical analysis of the spectrum graph is performed. First, the range of the low-frequency part of the spectrum graph is determined. This usually involves selecting the low-frequency area in the spectrum graph, and the low frequency usually represents the part with slower temperature changes. The frequency statistical tool in the data analysis software is used to calculate the intensity of temperature changes in the low-frequency area. By counting the temperature values ​​of the low-frequency components, the low-frequency instrument thermal imaging temperature data is extracted. These data reflect the temperature fluctuations of the instrument on a long time scale and provide basic data for subsequent anomaly detection.

[0127] Step S25: Acquire rated temperature data of industrial automation equipment;

[0128] In this embodiment, the rated temperature data of the industrial automation instrument is obtained. This includes consulting the technical manual or equipment manual of the instrument and recording the rated operating temperature range of the instrument under normal operation. The rated temperature data includes the maximum allowable temperature, the minimum allowable temperature and the normal operating temperature range of the instrument. These data can also be obtained by contacting the instrument manufacturer or referring to industry standards. The rated temperature data is organized into a structured data format, such as a spreadsheet or database record, to facilitate subsequent comparison and analysis.

[0129] Step S26: performing abnormal temperature comparison on the low-frequency instrument thermal imaging temperature data according to the rated temperature data of the industrial automation instrument, thereby obtaining abnormal temperature data of the instrument.

[0130] In this embodiment, a data analysis tool is used to compare the low-frequency instrument thermal imaging temperature data with the rated temperature data. First, the low-frequency instrument thermal imaging temperature data is compared with the rated temperature range. The specific operation includes: setting a temperature threshold (for example, exceeding the upper or lower limit of the rated temperature range), and then checking whether there are temperature values ​​exceeding these thresholds in the low-frequency data. Through this comparison, abnormal temperatures of the instrument over a long time scale can be identified. These abnormal temperature values ​​are recorded and marked as abnormal temperature data for further analysis or processing. These data are of great significance for equipment maintenance and fault diagnosis.

[0131] Optionally, step S3 specifically includes:

[0132] Step S31: extracting an instrument X-ray image according to the industrial automation instrument recognition image, thereby obtaining an instrument X-ray image;

[0133] In this embodiment, an X-ray image of an industrial automation instrument is obtained. An X-ray device with high resolution and appropriate radiation intensity is selected to perform an X-ray scan on the instrument. During the scanning process, the position of the instrument should be ensured to be stable to avoid image blur. After the scan is completed, the X-ray image is acquired into a computer system through a digitizing device. The X-ray image is saved in the form of a digital image, usually in the DICOM (Digital Imaging and Communications in Medicine) format, which can retain all the details of the image and related metadata. The image data is imported into the image processing software to ensure that the image is clear and discernible, and to prepare for further image processing steps.

[0134] Step S32: performing grayscale threshold image region segmentation according to the instrument X-ray image, thereby obtaining an X-ray grayscale threshold region image;

[0135] In this embodiment, the acquired X-ray image is subjected to grayscale threshold segmentation. The X-ray image is loaded using image processing software (such as ImageJ or MATLAB). First, different areas in the image are distinguished by setting an appropriate grayscale threshold. The grayscale threshold is usually determined by the brightness value of the image, and the threshold needs to be adjusted according to the specific characteristics of the weld joint when selecting it. The threshold segmentation function in the image processing tool is used to extract areas in the image that are higher or lower than a specific grayscale value. The processed image is displayed as a binary image, in which the target area is displayed in white and the background area is displayed in black. The segmentation result is saved as an image file for subsequent regional analysis.

[0136] Step S33: performing welding joint crack region analysis on the X-ray grayscale threshold region image, thereby obtaining welding joint crack region data;

[0137] In this embodiment, crack region analysis is performed on the X-ray grayscale threshold region image. Connected region analysis is performed on the binary image using an image analysis tool (such as OpenCV or ImageJ). The crack region in the image is identified and marked by an algorithm, and the crack usually appears as an elongated and irregular dark area. Morphological processing techniques (such as corrosion and expansion) are applied to improve the significance of the crack and remove noise. The specific features of the crack region (such as area, length and shape) will be extracted and recorded. Crack region data are generated, which contain the coordinates, size and distribution information of the crack, providing basic data for subsequent anomaly detection.

[0138] Step S34: performing welding joint pore area analysis on the X-ray grayscale threshold area image, thereby obtaining welding joint pore area data;

[0139] In this embodiment, the X-ray grayscale threshold area image is subjected to pore area analysis. The pore area in the image is identified using a region detection algorithm (such as a region growing algorithm or a morphological operation) in the image processing software. Pores usually appear as small, circular or elliptical dark spots. By setting the minimum and maximum area size thresholds, noise can be filtered out and the actual pore area can be focused on. Feature information of the pore area, such as position, diameter, and number, is extracted, and pore area data is generated. This data is saved for subsequent anomaly detection and quality assessment.

[0140] Step S35: Merge the number of abnormal areas of the weld joint according to the crack area data of the weld joint and the pore area data of the weld joint, so as to obtain the abnormal area data of the weld joint.

[0141] In this embodiment, the crack area data and the pore area data are merged and analyzed. The crack area data and the pore area data are imported and organized into the same data set using a data processing tool (such as Excel or the Pandas library in Python). When merging, matching and deduplication processing are performed based on the spatial position and size of the region. For each detected abnormal area (whether it is a crack or a pore), its specific position and size on the weld joint are recorded. The merged results generate abnormal area data of the weld joint, which shows all the detected defective areas in the weld joint. Finally, the abnormal area data is summarized into a report for quality control and further engineering analysis.

[0142] Optionally, step S33 is specifically:

[0143] Step S331: setting the crack region grayscale threshold according to the X-ray grayscale threshold region image, thereby obtaining the crack region grayscale threshold;

[0144] In this embodiment, the grayscale threshold of the crack area is set for the X-ray grayscale threshold area image. The X-ray grayscale image is imported into image processing software (such as MATLAB, ImageJ). By observing the grayscale value distribution of the crack area in the image, the grayscale range of the crack area is determined. This process can be achieved through histogram analysis. Using the histogram tool of the software, the distribution of grayscale values ​​in the image is analyzed, and an appropriate grayscale threshold is selected, which should be able to effectively distinguish cracks from other areas. The set grayscale threshold should be able to accurately cover the grayscale range of the crack area, while excluding background noise and other non-crack areas. The determined grayscale threshold is recorded for subsequent crack area division.

[0145] Step S332: dividing the X-ray grayscale threshold region image into crack regions according to the crack region grayscale threshold, thereby obtaining a crack region image;

[0146] In this embodiment, the crack area grayscale threshold is applied to divide the X-ray grayscale threshold area image into crack areas. The grayscale value in the image is compared with the set threshold using the threshold segmentation tool in the image processing software. For pixels with grayscale values ​​higher than the set threshold, they are marked as crack areas, and pixels below the threshold are marked as background. This operation can be achieved through the binarization function in the image processing software. In the generated binary image, the crack area will be highlighted as white or other contrasting colors, and the background area will remain black. Save the processed crack area image for further crack type classification.

[0147] Step S333: classifying the crack types according to the crack region image, thereby obtaining a hot crack image and a cold crack image;

[0148] In this embodiment, the crack area image is classified into crack types. The crack area image is imported into the image analysis software and the crack classification algorithm is applied. The crack type classification can be based on the shape, width, texture and other characteristics of the crack. For example, hot cracks usually have irregular shapes and show a large width, while cold cracks are relatively slender and more uniform. The feature extraction function in the image analysis tool is used to identify and classify the crack types. The classification results are divided into hot crack images and cold crack images, and the respective image files are saved for subsequent roughness and smoothness calculations.

[0149] Step S334: performing thermal crack roughness calculation on the thermal crack image to obtain thermal crack roughness data;

[0150] In this embodiment, the roughness calculation is performed on the thermal crack image. The thermal crack image is imported using the surface roughness analysis tool in the image processing software. First, the image processing algorithm is applied to extract the surface profile of the crack. Then, the roughness parameters of the crack surface, such as RMS (root mean square roughness) or Ra (arithmetic mean roughness) are calculated by the algorithm. These parameters can be obtained by calculating the degree of fluctuation of the crack profile. The calculation results are saved as roughness data, and the roughness value of the thermal crack is recorded for subsequent crack area analysis and evaluation.

[0151] Step S335: Calculating the cold crack smoothness of the cold crack image to obtain cold crack smoothness data;

[0152] In this embodiment, the smoothness of the cold crack image is calculated. The cold crack image is imported into the smoothness analysis tool. The boundary of the crack area is extracted using the smoothness analysis function in the image processing software, and the smoothness of the boundary is measured. An edge detection algorithm, such as Canny edge detection, can be applied to determine the boundary of the crack, and then the smoothness of the boundary is calculated. Statistical indicators such as standard deviation or coefficient of variation are used to quantify the smoothness of the crack. The smoothness calculation results are saved as a data file, and the smoothness value of the cold crack is recorded for subsequent crack area analysis and comparison.

[0153] Step S336: dividing the crack region image into weld joint crack regions according to the hot crack roughness data and the cold crack smoothness data, thereby obtaining hot crack region data and cold crack region data;

[0154] In this embodiment, the hot crack roughness data and the cold crack smoothness data are combined to further divide the crack region image. The roughness and smoothness data of the hot crack and the cold crack are imported using a data analysis tool. The crack region image is divided into hot crack regions and cold crack regions according to the set roughness and smoothness thresholds. These thresholds are applied to classify the crack region image to obtain detailed hot crack region data and cold crack region data. The image files and data records of the division results are saved for further abnormality detection and analysis.

[0155] Step S337: performing high-frequency regional distribution statistics according to the hot crack regional data and the cold crack regional data, so as to obtain the hot crack high-frequency regional data and the cold crack high-frequency regional data;

[0156] In this embodiment, the high-frequency regional distribution statistics of the hot crack region data and the cold crack region data are performed. The hot crack region data and the cold crack region data are imported into the data analysis software. The high-frequency component analysis of the crack region is performed using a spectrum analysis tool. The high-frequency components in the crack region are extracted by Fourier transform or wavelet transform, and these components usually reflect the slight changes and textures of the cracks. The distribution of the high-frequency region is counted, and the high-frequency region data of the hot crack and the cold crack are recorded. The statistical results are saved for further analysis of the overall quality of the welded joint.

[0157] Step S338: Merge the welding joint area according to the hot crack high-frequency area data and the cold crack high-frequency area data, so as to obtain the welding joint area data.

[0158] In this embodiment, the high-frequency area data of hot cracks is merged with the high-frequency area data of cold cracks. The high-frequency area data of hot cracks and cold cracks are integrated according to the actual area of ​​the weld joint using the data merging tool. Combine the data of the two to create a comprehensive weld joint area data map to display all high-frequency abnormal areas. The merging process needs to consider the overlap of the area and the consistency of the data to ensure that the merged data accurately reflects the overall state of the weld joint. The merged weld joint area data is saved for welding quality assessment and maintenance decision-making.

[0159] Optionally, step S34 is specifically:

[0160] Step S341: setting the grayscale threshold of the pore area on the X-ray grayscale threshold area image, thereby obtaining the grayscale threshold of the pore area;

[0161] In this embodiment, the grayscale threshold of the pore area is set to extract the pore area from the X-ray grayscale threshold area image. First, load the X-ray image into the image processing software (such as MATLAB or ImageJ). Use the grayscale histogram function of the software to analyze the grayscale value distribution of the pore area in the image. Pores usually appear as lower grayscale values ​​in the image. By adjusting the threshold slider, observe the changes in the pore area in the image, and select a grayscale threshold that can effectively distinguish the pores from the background area. Record the selected grayscale threshold for subsequent pore area division.

[0162] Step S342: dividing the X-ray grayscale threshold region image into pore regions according to the pore region grayscale threshold, thereby obtaining a pore region image;

[0163] In this embodiment, the grayscale threshold of the pore area is applied to divide the X-ray grayscale threshold area image into pore areas. The image is imported into the image processing software, and the grayscale value in the image is compared with the set pore threshold using the threshold segmentation function. Pixels above the set threshold are marked and regarded as pore areas, and pixels below the threshold are regarded as background. The generated image will be displayed as a binary image, in which the pore area is highlighted in white and the background area is black. The processed pore area image is saved for subsequent regional statistics and shape analysis.

[0164] Step S343: performing pore position area statistics according to the pore area image, thereby obtaining high-frequency pore position area data and low-frequency pore position area data;

[0165] In this embodiment, the position area statistics are performed on the pore area image. The position coordinates of the pores are extracted using an image analysis tool (such as ImageJ or OpenCV in Python). Fourier transform or other frequency domain analysis methods are performed to identify the high-frequency and low-frequency pore position areas in the image. High-frequency pores are usually shown as dense pore distributions, while low-frequency pores are shown as dispersed pore distributions. According to the frequency domain analysis results, the pore positions are divided into high-frequency areas and low-frequency areas. The high-frequency pore position data and the low-frequency pore position data are saved for subsequent dangerous area analysis.

[0166] Step S344: performing statistics on the shapes and sizes of pores according to the pore area image, thereby obtaining large pore area data and small pore area data;

[0167] In this embodiment, a statistical analysis of the shape and size of the pores is performed. The pore area image is imported into the image processing software, and the pores are measured using the regional analysis tool. Morphological analysis methods (such as connected domain analysis) are applied to identify the shape and size of the pores. A threshold value for the pore size is set, and the pores are divided into "large pores" and "small pores". Large pores generally refer to pores whose diameter or area exceeds a set value, while small pores are smaller than this value. The regional data of each type of pore is recorded, and these data are saved as files for subsequent hazardous area assessments.

[0168] Step S345: performing a high-risk area intersection operation according to the high-frequency pore position area data and the large pore area data, thereby obtaining high-risk pore area data;

[0169] In this embodiment, the high-frequency pore position area data and the large pore area data are intersected. Using data processing tools, the area data of both are imported into the same data set. Through the intersection calculation, the overlapping parts of the high-frequency pore area and the large pore area are extracted. These overlapping areas are marked as high-risk pore areas because they have both high-frequency distribution and large size characteristics. Using the logical operation function of the image processing software, an image of the high-risk pore area is generated, and the high-risk pore area data is saved.

[0170] Step S346: performing a low-risk area intersection operation according to the low-frequency pore position area data and the small pore area data, thereby obtaining low-risk pore area data;

[0171] In this embodiment, the low-frequency pore location area data and the small pore area data are intersected. The data of both are imported into the data processing tool, and a logical operation is performed to determine the overlapping area. The intersection of the low-frequency pore area and the small pore area is marked as a low-risk pore area, because these pores are sparsely distributed and small in size, but still need attention. Generate a low-risk pore area image and save the low-risk pore area data.

[0172] Step S347: Merge the weld joint porosity risk areas according to the high-risk porosity area data and the low-risk porosity area data, thereby obtaining the weld joint porosity area data.

[0173] In this embodiment, the high-risk porosity area data is merged with the low-risk porosity area data. The data merging tool is used to integrate the regional data of high-risk porosity and low-risk porosity into a comprehensive data set. The high-risk and low-risk porosity areas are marked separately through image processing software, and these areas are merged to generate complete porosity area data of the weld joint. Ensure that the merged data accurately reflects the potential dangers of all porosity. Save the final weld joint porosity area data and generate a report for quality assessment and maintenance decision-making.

[0174] Optionally, step S4 is specifically:

[0175] Step S41: Obtaining time distribution data of abnormal temperature of the instrument and time distribution data of abnormal welding joint according to time distribution statistics of abnormal temperature of the instrument and abnormal welding joint;

[0176] In this embodiment, the abnormal temperature data of the instrument and the abnormal data of the welding joint are collected. These data record the time point when the abnormal event occurs. Use data processing software (such as Excel or Pandas library in Python) to arrange the data in chronological order and perform time distribution statistics. The specific steps include: 1) importing the data into the software, 2) grouping the time data, such as statistics by hour, day or week, 3) calculating the frequency of abnormal events in each time period, 4) generating a time distribution diagram to show the time distribution of abnormal temperature of the instrument and abnormal welding joint. Save the generated time distribution data for subsequent overlap calculation and feature extraction.

[0177] Step S42: Calculating the time distribution overlap data according to the abnormal temperature time distribution data of the instrument and the abnormal time distribution data of the welding joint, thereby obtaining the time distribution overlap data;

[0178] In this embodiment, the time distribution overlapping data calculation is performed. Using the time distribution data processing tool, the abnormal temperature time distribution data of the instrument is compared and analyzed with the abnormal time distribution data of the weld joint. The two data sets are aligned according to the same time granularity (for example, every hour or every day). Then, the overlapping parts of the time points in the two data sets are found by intersection calculation. These overlapping parts represent instrument abnormalities and weld joint abnormal events that occur simultaneously in the same time period. Generate charts and statistical results of overlapping data, and record the overlapping frequency in each time period for further analysis.

[0179] Step S43: extracting instrument temperature-welding joint abnormality features according to the time distribution overlapping data, thereby obtaining instrument temperature-welding joint abnormality data;

[0180] In this embodiment, feature extraction is performed on the time distribution overlapping data. The overlapping data is imported into data analysis software (such as MATLAB or Python). By analyzing the time distribution overlapping data, the relevant features between the instrument temperature anomaly and the welding joint anomaly are extracted. Including: 1) calculating the number of overlapping events in each time period, 2) analyzing the temperature range of overlapping events and the type of welding joint anomaly, 3) extracting statistical features, such as the average temperature and maximum temperature of overlapping events and their correlation with the anomaly type. Generate an instrument temperature-welding joint anomaly data report, record the extracted features, and use them for subsequent data integration and analysis.

[0181] Step S44: calculating the material grain coarsening probability for the instrument temperature-welding joint abnormality data according to the multispectral image, thereby obtaining the material grain coarsening probability data;

[0182] In this embodiment, multispectral images are used to calculate the probability of material grain coarsening for instrument temperature-welding joint abnormal data. The multispectral image is imported into the image processing software, and the grain information of the material is extracted using spectral analysis technology. Image segmentation technology is applied to separate the grain area in the image from the background. The size of the grains is counted by image analysis tools (such as MATLAB or ImageJ), and the distribution of the grains and the probability of coarsening are calculated. The calculation of the probability of coarsening is based on the statistical data of the grain size distribution, and a probability model is used to evaluate the degree of grain coarsening under different temperature conditions. The calculation results are saved as material grain coarsening probability data for subsequent analysis.

[0183] Step S45: calculating the material phase change probability of the instrument temperature-welding joint abnormality data according to the multispectral image, thereby obtaining the material phase change probability data;

[0184] In this embodiment, the probability of material phase change is calculated. The multispectral image is imported into the image processing software, and the spectral information is used to analyze the phase change phenomenon of the material. The appropriate spectral band is selected to analyze the changes in the spectral characteristics of the material. Through image processing and spectral analysis technology, the phase change characteristics of the material under different temperature conditions are detected. The probability of phase change of each material is calculated using a phase change probability model (such as a statistical model based on eigenvalues). The phase change probability under different abnormal conditions is counted, and a material phase change probability data report is generated. The results are saved for analyzing the heat treatment effect and structural changes of the material.

[0185] Step S46: Integrate the instrument material abnormality characteristics according to the material grain coarsening probability data and the material phase change probability data, so as to obtain the instrument material abnormality characteristic data.

[0186] In this embodiment, the material grain coarsening probability data and the material phase change probability data are integrated. Use the data integration tool to import the data of both into the data analysis software. First, match the grain coarsening probability data and phase change probability data for each time period. Then, through statistical analysis, identify the characteristic patterns of material anomalies, such as the degree of correlation between grain coarsening and phase change within certain temperature ranges. Generate a comprehensive report to record detailed data on the abnormal characteristics of the instrument material, including the combined effects of grain coarsening and phase change. Save the integrated data for instrument failure prediction and maintenance decisions.

[0187] Optionally, step S5 specifically includes:

[0188] Step S51: performing instrument identification according to the visible light image of the industrial instrument, thereby obtaining instrument identification data;

[0189] In this embodiment, industrial instruments are identified by visible light images. Obtain visible light images of industrial instruments. These images can be taken by a high-resolution camera to ensure that the images are clear and detailed. Import the images into image processing software (such as OpenCV or MATLAB). Use image segmentation and feature extraction algorithms (such as edge detection and contour recognition) to identify the instruments in the image. Extract the characteristic information of the instrument, such as shape, size, identifier, etc. Generate instrument identification data, which includes the identification results of each instrument and its characteristic description. Organize the identification data into a structured format for subsequent processing and analysis.

[0190] Step S52: constructing a digital twin instrument identification model according to the instrument identification data, thereby obtaining a digital twin instrument identification model;

[0191] In this embodiment, a digital twin instrument recognition model is constructed using instrument recognition data. The instrument recognition data is organized into a training set, which contains image features and classification labels for each instrument. Using a modeling tool for a digital twin model (such as TensorFlow or PyTorch), these data are input into the model for training. Select a suitable algorithm, such as a convolutional neural network (CNN), for feature learning and model training. Through repeated training and verification, the model parameters are optimized to improve recognition accuracy. After the training is completed, the obtained digital twin instrument recognition model is saved, and the model can accurately recognize and classify instruments in new images.

[0192] Step S53: classifying the instruments according to the instrument identification data, thereby obtaining metal material instrument data and composite material instrument data;

[0193] In this embodiment, the instruments are classified to obtain metal material instrument data and composite material instrument data. The instrument identification data is used in combination with material information (such as material labels or attributes) to classify the identified instruments. The identification data is imported into data processing software (such as Excel or Python), and the instruments are divided into two categories according to the material characteristics: metal material instruments and composite material instruments. The data is classified using a classification algorithm (such as a decision tree or a support vector machine) to generate two types of data sets. The metal material instrument data and the composite material instrument data are saved to facilitate subsequent abnormal matching and intelligent adaptation operations.

[0194] Step S54: matching the metal material instrument data with abnormal metal materials according to the instrument material abnormal characteristic data, thereby obtaining abnormal metal material data;

[0195] In this embodiment, the abnormal characteristic data of the instrument material is used to perform abnormal matching on the metal material instrument data. Collect metal material instrument data and material abnormal characteristic data. Import these data into a data analysis tool (such as MATLAB or Python) to perform abnormal matching operations. Identify abnormal features in the metal material instrument data by setting a threshold or using anomaly detection algorithms (such as isolation forest or cluster analysis). Generate a data set containing abnormal metal material information, including abnormal detection results and related features. Save abnormal metal material data to prepare for subsequent intelligent adaptation.

[0196] Step S55: matching composite material instrument data with abnormal composite materials according to instrument material abnormal characteristic data, thereby obtaining abnormal composite material data;

[0197] In this embodiment, the abnormal characteristic data of the instrument material is used to perform abnormal matching on the composite material instrument data. The composite material instrument data and the material abnormal characteristic data are obtained. The data is imported into a data analysis tool (such as Python's Scikit-learn or MATLAB) to perform abnormal matching. Anomaly detection algorithms (such as density peak clustering or principal component analysis) are used to identify abnormal features in the composite material instrument data. Abnormal composite material data is generated, including a feature description and matching results for each abnormal item. The abnormal composite material data is saved to provide a basis for further intelligent adaptation of materials.

[0198] Step S56: performing intelligent material adaptation on the metal material instrument and the composite material instrument respectively according to the abnormal metal material data and the abnormal composite material data, so as to obtain instrument material intelligent adaptation data.

[0199] In this embodiment, abnormal metal material data and abnormal composite material data are used to perform intelligent material adaptation on the instrument. Integrate abnormal data and import adaptation algorithm tools (such as rule-based matching systems or optimization algorithms). Adapt metal material instruments and composite material instruments separately. For metal material instruments, use abnormal metal data to adjust and optimize material properties; for composite material instruments, adjust corresponding material parameters based on abnormal composite material data. Apply the adaptation algorithm to generate the final material adaptation results, including recommended material adjustment plans and adaptation suggestions. Save instrument material intelligent adaptation data for subsequent maintenance, optimization and quality control.

[0200] Optionally, step S55 is specifically:

[0201] Step S551: Calculating thermal stability data of the composite material according to the instrument data of the composite material, thereby obtaining thermal stability data of the composite material;

[0202] In the present embodiment, the thermal stability data of the composite material instrument is calculated to obtain the data of the composite material instrument, including the performance of the material at different temperatures. These data can be obtained from material testing experiments, such as high temperature furnace testing. These data are imported into data analysis tools (such as NumPy libraries in MATLAB or Python). The data is subjected to thermal stability analysis, including calculation of thermal stability indicators, such as weight loss, thermal expansion coefficient and its rate of change of the material at different temperatures. Detailed thermal stability data is obtained using thermal analysis techniques, such as differential scanning calorimeter (DSC) or thermogravimetric analysis (TGA). After the calculation is completed, the thermal stability data of the composite material is saved for subsequent analysis.

[0203] Step S552: dividing the composite material instrument data into instrument thermal stability regions according to the composite material thermal stability data, thereby obtaining composite material instrument thermal stability region data;

[0204] In this embodiment, the thermal stability data of the composite material is applied to the composite material instrument data for thermal stability region division. The thermal stability data is imported into a data processing software (such as MATLAB or Python). Different regions in the composite material instrument data are divided according to the temperature range and stability index of the thermal stability data. The instrument data is divided into multiple thermal stability regions, such as a high stability region, a medium stability region, and a low stability region using a segmented processing method. By setting a thermal stability threshold, the stability of each region is marked according to the performance of the material in these regions. Generate thermal stability region data, and record the temperature range and stability level of each region to provide a basis for further analysis.

[0205] Step S553: ​​extracting material phase change characteristics according to the instrument material abnormal characteristic data, thereby obtaining material phase change data;

[0206] In this embodiment, phase change characteristics of the material are extracted. Material grain coarsening probability data and material phase change probability data are collected. These data are imported into a data analysis tool (such as the Pandas library in MATLAB or Python). Using a phase change analysis algorithm, the phase change characteristics of the material under different temperature conditions are extracted, such as the phase change temperature point of the material, the phase change type, and the rate of the phase change process. The sample is tested using a differential scanning calorimeter (DSC) to record the thermal effect data of the phase change. The phase change data of each material is calculated, including the starting temperature, the end temperature of the phase change, and its behavior during the phase change process. The phase change data is saved to provide information for subsequent phase change analysis.

[0207] Step S554: performing composite material instrument phase change analysis on composite material instrument thermal stability region data according to the material phase change data, thereby obtaining composite material instrument phase change data;

[0208] In this embodiment, the material phase change data is applied to the thermal stability region data of the composite material instrument for phase change analysis. First, the material phase change data is imported into the analysis software. The thermal stability region data of the composite material instrument is partitioned and analyzed, and the phase change characteristic data of each thermal stability region is evaluated. For example, by comparing the phase change temperature points of the material in different regions with the temperature range of the thermal stability region, it is determined which regions have undergone phase change. Using data analysis tools (such as the SciPy library in MATLAB or Python), the phase change behavior of different thermal stability regions is analyzed, and phase change data, including the regional distribution, intensity and type of phase change, is generated. The phase change data of the composite material instrument is saved to prepare for the subsequent processing of abnormal data.

[0209] Step S555: clustering abnormal data of composite material instrument data according to composite material instrument phase change data, thereby obtaining abnormal composite material data.

[0210] In this embodiment, abnormal cluster analysis of composite material instrument data is performed. Composite material instrument phase change data is obtained. These data are imported into a cluster analysis tool (such as the Scikit-learn library in Python). Select a suitable clustering algorithm (such as K-means clustering or DBSCAN) to cluster the phase change data and identify abnormal data points. Set clustering parameters, such as distance threshold or number of clusters, and perform cluster analysis. Through the clustering results, abnormal composite material regions are identified, which show characteristics different from normal phase change behavior. Abnormal composite material data is generated, including a feature description and clustering results for each abnormal point. Abnormal composite material data is saved to provide a basis for further analysis and processing of the material.

[0211] Optionally, step S54 is specifically:

[0212] Step S541: Calculate the grain size according to the metal material instrument data, thereby obtaining the metal material instrument grain size data;

[0213] In this embodiment, the grain size is calculated for the metal material instrument data. The data of the metal material instrument is obtained, including microscope images or scanned images of material samples. These images are imported using image processing software (such as ImageJ or MATLAB). Image analysis techniques (such as threshold segmentation and edge detection) are applied to identify grain boundaries. The grain size is measured using a particle analysis tool to calculate the diameter or area of ​​each grain. All measurement results are counted to generate the grain size data of the metal material instrument. The data includes the size value of each grain and its distribution characteristics, which are stored in a structured format for subsequent statistical analysis and processing.

[0214] Step S542: performing grain distribution statistics according to the metal material instrument grain size data, thereby obtaining the metal material instrument grain size distribution data;

[0215] In this embodiment, the grain size data of the metal material instrument is distributed and counted. The grain size data is imported into a statistical analysis tool (such as the Pandas library in Python or MATLAB). The data is statistically analyzed to calculate the frequency distribution, mean, median and standard deviation of the grain size. A grain size distribution graph (such as a histogram or a kernel density estimation graph) is drawn to show the distribution of grains in different size ranges. The distribution data is analyzed to identify the main concentration range of the grain size and its changing trend. The grain size distribution data of the metal material instrument is generated to provide a basis for the subsequent grain coarsening analysis.

[0216] Step S543: extracting material grain coarsening characteristics according to the instrument material abnormal characteristic data, thereby obtaining material grain coarsening data;

[0217] In this embodiment, the material grain coarsening characteristics are extracted. First, the material grain coarsening probability data is obtained. The data is imported into a data analysis tool (such as the SciPy library in MATLAB or Python). The coarsening feature extraction algorithm (such as statistical analysis or image processing technology) is used to identify the change trend of the grain size, especially the part where the grain size increases. The degree of grain coarsening is calculated, and relevant features are extracted, such as the proportion of coarsening, the rate of size growth, etc. The material grain coarsening data is generated, which includes the coarsening characteristics and degree of each material, and saved as structured data for further analysis.

[0218] Step S544: performing a grain coarsening comparison on the metal material instrument grain size distribution data according to the material grain coarsening data, thereby obtaining the metal material instrument grain coarsening distribution data;

[0219] In this embodiment, the material grain coarsening data is applied to the grain size distribution data of the metal material instrument for comparison. First, the grain size distribution data and grain coarsening data are obtained. These data are imported into a data analysis tool (such as MATLAB or Python). The coarsening data is compared with the size distribution data, and the grain area with increased size is identified by setting a coarsening threshold. The comparison results are calculated to generate the grain coarsening distribution data of the metal material instrument. These data include the distribution characteristics and frequency of coarsened grains, providing spatial distribution information of grain coarsening.

[0220] Step S545: clustering abnormal grain coarsening of the metal material instrument data according to the metal material instrument grain coarsening distribution data, thereby obtaining abnormal metal material data.

[0221] In this embodiment, the grain coarsening distribution data is used to perform abnormal clustering on the metal material instrument data. Obtain the grain coarsening distribution data generated in step S544. Import the data into a clustering analysis tool (such as the Scikit-learn library in Python). Select a suitable clustering algorithm (such as K-means clustering or DBSCAN) to perform cluster analysis on the grain coarsening data. Set clustering parameters to identify abnormal grain areas that are significantly different from the normal distribution. Generate abnormal metal material data, including feature descriptions and clustering results for each abnormal area. Save these data as a basis for identifying abnormal materials to provide support for subsequent quality control and improvement measures.

[0222] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0223] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A digital twin-based instrument identification model training method, characterized in that: The following steps are involved: Step S1: acquiring an industrial automation instrument recognition image, and extracting an industrial instrument visible light image according to the industrial automation instrument recognition image, thereby obtaining an industrial instrument visible light image; and performing spectrum reconstruction according to the industrial instrument visible light image, thereby obtaining a multispectral image; Step S2: extracting the instrument thermal imaging image according to the industrial automation instrument recognition image, thereby obtaining the instrument thermal imaging image; performing abnormal temperature detection of the instrument according to the instrument thermal imaging image, thereby obtaining abnormal temperature data of the instrument; Step S3: extracting an instrument X-ray image according to the industrial automation instrument recognition image, thereby obtaining an instrument X-ray image; detecting abnormal areas of welding joints according to the instrument X-ray image, thereby obtaining abnormal area data of welding joints; Step S4: performing abnormal correlation analysis on the abnormal welding joint data according to the abnormal temperature data of the instrument, thereby obtaining the abnormal instrument temperature-welding joint data; performing material abnormal characteristic analysis on the abnormal instrument temperature-welding joint data according to the multispectral image, thereby obtaining the abnormal instrument material characteristic data; Step S5: performing instrument identification according to the visible light image of the industrial instrument, thereby obtaining instrument identification data; A digital twin instrument identification model is constructed based on the instrument identification data to obtain the digital twin instrument identification model; the instrument identification data is intelligently adapted to the instrument material based on the digital twin instrument identification model and the abnormal characteristic data of the instrument material to obtain the instrument material intelligent adaptation data; Step S6: Optimize the digital twin instrument recognition model according to the instrument material intelligent adaptation data to obtain the digital twin instrument recognition optimization model, and upload it to the instrument recognition management platform to execute the instrument recognition model training task.

2. The instrument identification model training method based on digital twin according to claim 1 is characterized in that: Step S1 is specifically as follows: Step S11: acquiring an industrial automation instrument recognition image, and extracting an industrial instrument visible light image according to the industrial automation instrument recognition image, thereby obtaining an industrial instrument visible light image; Step S12: extracting the spectral features of the industrial instrument according to the visible light image of the industrial instrument, thereby obtaining the spectral data of the industrial instrument; Step S13: performing statistics on the near-infrared band range of the industrial instrument according to the spectrum data of the industrial instrument, thereby obtaining the near-infrared band range data of the industrial instrument; Step S14: constructing a spectrum reconstruction model according to the spectrum data of the industrial instrument and the near-infrared band range data of the industrial instrument, thereby obtaining a spectrum reconstruction model; Step S15: performing pixel-by-pixel spectral reflectance analysis on the visible light image of the industrial instrument according to the spectral reconstruction model, thereby obtaining pixel-by-pixel spectral reflectance data; Step S16: reconstructing a multispectral image according to the pixel-by-pixel spectral reflectance data, thereby obtaining a multispectral image.

3. The instrument identification model training method based on digital twin according to claim 1 is characterized in that: Step S2 is specifically as follows: Step S21: extracting the instrument thermal imaging image according to the industrial automation instrument recognition image, thereby obtaining the instrument thermal imaging image; Step S22: performing pixel temperature mapping according to the instrument thermal imaging image, thereby obtaining instrument thermal imaging temperature data; Step S23: drawing an instrument thermal imaging temperature spectrum diagram according to the instrument thermal imaging temperature data, thereby obtaining an instrument thermal imaging temperature spectrum diagram; Step S24: performing temperature frequency statistics according to the instrument thermal imaging temperature spectrum diagram, thereby obtaining low-frequency instrument thermal imaging temperature data; Step S25: Acquire rated temperature data of industrial automation equipment; Step S26: performing abnormal temperature comparison on the low-frequency instrument thermal imaging temperature data according to the rated temperature data of the industrial automation instrument, thereby obtaining abnormal temperature data of the instrument.

4. The instrument identification model training method based on digital twin according to claim 1 is characterized in that: Step S3 is specifically as follows: Step S31: extracting an instrument X-ray image according to the industrial automation instrument recognition image, thereby obtaining an instrument X-ray image; Step S32: performing grayscale threshold image region segmentation according to the instrument X-ray image, thereby obtaining an X-ray grayscale threshold region image; Step S33: performing welding joint crack region analysis on the X-ray grayscale threshold region image, thereby obtaining welding joint crack region data; Step S34: performing welding joint pore area analysis on the X-ray grayscale threshold area image, thereby obtaining welding joint pore area data; Step S35: Merge the number of abnormal areas of the weld joint according to the crack area data of the weld joint and the pore area data of the weld joint, so as to obtain the abnormal area data of the weld joint.

5. The instrument identification model training method based on digital twin according to claim 4 is characterized in that: Step S33 is specifically as follows: Step S331: setting the crack region grayscale threshold according to the X-ray grayscale threshold region image, thereby obtaining the crack region grayscale threshold; Step S332: dividing the X-ray grayscale threshold region image into crack regions according to the crack region grayscale threshold, thereby obtaining a crack region image; Step S333: classifying the crack types according to the crack region image, thereby obtaining a hot crack image and a cold crack image; Step S334: performing thermal crack roughness calculation on the thermal crack image to obtain thermal crack roughness data; Step S335: Calculating the cold crack smoothness of the cold crack image to obtain cold crack smoothness data; Step S336: dividing the crack region image into weld joint crack regions according to the hot crack roughness data and the cold crack smoothness data, thereby obtaining hot crack region data and cold crack region data; Step S337: performing high-frequency regional distribution statistics according to the hot crack regional data and the cold crack regional data, so as to obtain the hot crack high-frequency regional data and the cold crack high-frequency regional data; Step S338: Merge the welding joint area according to the hot crack high-frequency area data and the cold crack high-frequency area data, so as to obtain the welding joint area data.

6. The instrument identification model training method based on digital twin according to claim 4 is characterized in that: Step S34 is specifically as follows: Step S341: setting the grayscale threshold of the pore area on the X-ray grayscale threshold area image, thereby obtaining the grayscale threshold of the pore area; Step S342: dividing the X-ray grayscale threshold region image into pore regions according to the pore region grayscale threshold, thereby obtaining a pore region image; Step S343: performing pore position area statistics according to the pore area image, thereby obtaining high-frequency pore position area data and low-frequency pore position area data; Step S344: performing statistics on the shapes and sizes of pores according to the pore area image, thereby obtaining large pore area data and small pore area data; Step S345: performing a high-risk area intersection operation according to the high-frequency pore position area data and the large pore area data, thereby obtaining high-risk pore area data; Step S346: performing a low-risk area intersection operation according to the low-frequency pore position area data and the small pore area data, thereby obtaining low-risk pore area data; Step S347: Merge the weld joint porosity risk areas according to the high-risk porosity area data and the low-risk porosity area data, thereby obtaining the weld joint porosity area data.

7. The instrument identification model training method based on digital twin according to claim 1 is characterized in that: Step S4 is specifically as follows: Step S41: Obtaining time distribution data of abnormal temperature of the instrument and time distribution data of abnormal welding joint according to time distribution statistics of abnormal temperature of the instrument and abnormal welding joint; Step S42: Calculating the time distribution overlap data according to the abnormal temperature time distribution data of the instrument and the abnormal time distribution data of the welding joint, thereby obtaining the time distribution overlap data; Step S43: extracting instrument temperature-welding joint abnormality features according to the time distribution overlapping data, thereby obtaining instrument temperature-welding joint abnormality data; Step S44: calculating the material grain coarsening probability for the instrument temperature-welding joint abnormality data according to the multispectral image, thereby obtaining the material grain coarsening probability data; Step S45: calculating the material phase change probability of the instrument temperature-welding joint abnormality data according to the multispectral image, thereby obtaining the material phase change probability data; Step S46: Integrate the instrument material abnormality characteristics according to the material grain coarsening probability data and the material phase change probability data, so as to obtain the instrument material abnormality characteristic data.

8. The instrument identification model training method based on digital twin according to claim 1 is characterized in that: Step S5 is specifically as follows: Step S51: performing instrument identification according to the visible light image of the industrial instrument, thereby obtaining instrument identification data; Step S52: constructing a digital twin instrument identification model according to the instrument identification data, thereby obtaining a digital twin instrument identification model; Step S53: classifying the instruments according to the instrument identification data, thereby obtaining metal material instrument data and composite material instrument data; Step S54: matching the metal material instrument data with abnormal metal materials according to the instrument material abnormal characteristic data, thereby obtaining abnormal metal material data; Step S55: matching composite material instrument data with abnormal composite materials according to instrument material abnormal characteristic data, thereby obtaining abnormal composite material data; Step S56: performing intelligent material adaptation on the metal material instrument and the composite material instrument respectively according to the abnormal metal material data and the abnormal composite material data, so as to obtain instrument material intelligent adaptation data.

9. The digital twin-based instrument identification model training method according to claim 8, characterized in that: Step S55 is specifically as follows: Step S551: Calculating thermal stability data of the composite material according to the instrument data of the composite material, thereby obtaining thermal stability data of the composite material; Step S552: dividing the composite material instrument data into instrument thermal stability regions according to the composite material thermal stability data, thereby obtaining composite material instrument thermal stability region data; Step S553: ​​extracting material phase change characteristics according to the instrument material abnormal characteristic data, thereby obtaining material phase change data; Step S554: performing composite material instrument phase change analysis on composite material instrument thermal stability region data according to the material phase change data, thereby obtaining composite material instrument phase change data; Step S555: clustering abnormal data of composite material instrument data according to composite material instrument phase change data, thereby obtaining abnormal composite material data.

10. The digital twin-based instrument identification model training method according to claim 8, characterized in that: Step S54 is specifically as follows: Step S541: Calculate the grain size according to the metal material instrument data, thereby obtaining the metal material instrument grain size data; Step S542: performing grain distribution statistics according to the metal material instrument grain size data, thereby obtaining the metal material instrument grain size distribution data; Step S543: extracting material grain coarsening characteristics according to the instrument material abnormal characteristic data, thereby obtaining material grain coarsening data; Step S544: performing a grain coarsening comparison on the metal material instrument grain size distribution data according to the material grain coarsening data, thereby obtaining the metal material instrument grain coarsening distribution data; Step S545: clustering abnormal grain coarsening of the metal material instrument data according to the metal material instrument grain coarsening distribution data, thereby obtaining abnormal metal material data.

Images