CN118822958A - A method for evaluating the health status of electromagnetic coils based on image features - Google Patents

Abstract

The present invention relates to the field of image data processing technology, and in particular to a method for evaluating the health status of an electromagnetic coil based on image features. The method comprises the following steps: performing a series of processing and analysis on the electromagnetic coil image, including super-resolution reconstruction, feature enhancement, deformation analysis, and color change association, so as to obtain the deformation data and color change data of the coil. The two types of data are integrated and analyzed to determine the variation factors of the coil, and a virtual model is used to simulate the development process of the coil. Finally, the actual coil image is matched with the degree of development according to the simulated data, the remaining service life of the electromagnetic coil is estimated, and the health status is evaluated; the present invention realizes a more comprehensive and efficient method for evaluating the health status of an electromagnetic coil.

Description

A method for evaluating the health status of electromagnetic coils based on image features

Technical Field

The present invention relates to the technical field of image data processing, and in particular to a method for evaluating the health status of an electromagnetic coil based on image features.

Background Art

Initially, the health status assessment method of electromagnetic coils mainly relied on manual visual inspection of the coil appearance to find obvious damage or deformation problems. However, this method is highly subjective, difficult to find hidden internal problems, and has low assessment accuracy. With the development of image analysis technology, people began to use simple image processing techniques, such as grayscale histograms and edge detection, to conduct preliminary analysis of the visual features of the coil surface, but they were unable to capture complex surface defect features, resulting in low accuracy. Subsequently, it entered the feature extraction and pattern recognition stage, using richer texture features, shape features, etc. to extract coil surface information, and combined with supervised learning algorithms for pattern recognition. However, this method requires complex feature engineering and a large amount of labeled data, and its generalization performance is limited. In recent years, deep learning has been widely used in this field. Using deep learning models such as convolutional neural networks, features can be automatically extracted from original images and classified and identified. However, this method also requires a large amount of training data, and the generalization ability of the model is limited, and it is difficult to explain the internal decision-making mechanism.

Summary of the invention

Based on this, it is necessary to provide an electromagnetic coil health status assessment method based on image features to solve at least one of the above technical problems.

To achieve the above purpose, a method for evaluating the health status of an electromagnetic coil based on image features comprises the following steps:

Step S1: acquiring an electromagnetic coil image; performing super-resolution reconstruction on the electromagnetic coil image to generate a reconstructed electromagnetic coil image; performing visual feature optimization on the reconstructed electromagnetic coil image to obtain an optimized coil image;

Step S2: acquiring a standard coil image; performing covering object recognition on the optimized coil image data to generate surface attachment data; performing deformation analysis on the optimized coil image according to the standard coil image and the surface attachment data to generate coil deformation data;

Step S3: performing regional light reflection analysis on the optimized coil image based on the coil deformation data to generate coil color change data;

Step S4: based on the coil deformation data and the coil color change data, reversely deduce and trace the coil variation data to obtain the coil variation factor; perform variation field simulation on the coil variation factor to generate simulated coil development data;

Step S5: Perform health status assessment on the electromagnetic coil image according to the simulated coil development data to execute the electromagnetic coil health status assessment method based on image features.

The present invention can greatly improve the quality and details of electromagnetic coil images through super-resolution reconstruction and visual feature optimization, laying a good foundation for subsequent analysis. This is crucial for evaluation methods based on image features. Covering recognition and deformation analysis can effectively capture degradation features such as coil surface attachments and geometric deformations, providing a key basis for accurate diagnosis of coil status. Light reflection analysis and inverse simulation help to explore the internal causes of coil changes, which is conducive to an in-depth understanding of the degradation mechanism of the coil and provide a basis for preventive maintenance. Integrating the above steps into a complete evaluation process can make full use of the advantages of image analysis, evaluate the health status of the coil more comprehensively and systematically, and provide reliable support for subsequent maintenance decisions. Therefore, the present invention can comprehensively and dynamically grasp the actual use status of the electromagnetic coil through a series of systematic image processing and analysis technologies, greatly improving the accuracy and reliability of coil health assessment. This is of great significance for improving the safety and use efficiency of electromagnetic equipment.

The benefit of the present invention lies in that high-quality and reliable coil image data can be obtained by collecting coil images through a digital camera and using processing methods such as super-resolution reconstruction, edge sharpening, and feature enhancement, laying a good foundation for subsequent analysis. By performing feature comparison analysis on the standard coil image and the optimized coil image, the surface attachments of the coil can be identified and the deformation can be analyzed, so as to fully grasp the actual use status of the coil. By analyzing the light reflection characteristics of the coil image and combining the deformation data, the color change information of the coil can be obtained, which provides an important basis for accurately evaluating the state change of the coil. By integrating and analyzing the coil deformation and color change data, the main factors causing the coil variation can be determined, and the variation field simulation can be performed based on the virtual coil model to simulate the future development trend of the coil. By comparing and matching the simulated data with the actual coil image, the remaining service life of the coil can be estimated, and its health status can be evaluated in combination with the preset threshold, providing a decision-making basis for the maintenance and replacement of the coil. Therefore, the present invention can comprehensively and dynamically grasp the actual use status of the electromagnetic coil through a series of systematic image processing and analysis technologies, greatly improving the accuracy and reliability of the coil health assessment. This is of great significance for improving the safety and use efficiency of electromagnetic equipment.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of a process flow of a method for evaluating the health status of an electromagnetic coil based on image features;

FIG2 is a schematic diagram of a detailed implementation process of step S2 in FIG1 ;

FIG3 is a schematic diagram of a detailed implementation process of step S3 in FIG1 ;

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

The technical method of the present invention is described clearly and completely below 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.

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.

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.

To achieve the above purpose, please refer to Figures 1 to 3, a method for evaluating the health status of an electromagnetic coil based on image features includes the following steps:

Step S1: acquiring an electromagnetic coil image; performing super-resolution reconstruction on the electromagnetic coil image to generate a reconstructed electromagnetic coil image; performing visual feature optimization on the reconstructed electromagnetic coil image to obtain an optimized coil image;

Step S2: acquiring a standard coil image; performing covering object recognition on the optimized coil image data to generate surface attachment data; performing deformation analysis on the optimized coil image according to the standard coil image and the surface attachment data to generate coil deformation data;

Step S3: performing regional light reflection analysis on the optimized coil image based on the coil deformation data to generate coil color change data;

Step S4: based on the coil deformation data and the coil color change data, reversely deduce and trace the coil variation data to obtain the coil variation factor; perform variation field simulation on the coil variation factor to generate simulated coil development data;

Step S5: Perform health status assessment on the electromagnetic coil image according to the simulated coil development data to execute the electromagnetic coil health status assessment method based on image features.

The present invention can greatly improve the quality and details of the electromagnetic coil image through super-resolution reconstruction and visual feature optimization, laying a good foundation for subsequent analysis. This is crucial for the evaluation method based on image features. Covering recognition and deformation analysis can effectively capture the surface attachments and geometric deformation features of the coil, providing a key basis for accurate diagnosis of the coil status. Light reflection analysis and inverse simulation help to explore the internal causes of coil changes, which is conducive to a deep understanding of the degradation mechanism of the coil and provide a basis for preventive maintenance. Integrating the above steps into a complete evaluation process can make full use of the advantages of image analysis, evaluate the health status of the coil more comprehensively and systematically, and provide reliable support for subsequent maintenance decisions. Therefore, the present invention can comprehensively and dynamically grasp the actual use status of the electromagnetic coil through a series of systematic image processing and analysis technologies, greatly improving the accuracy and reliability of coil health assessment. This is of great significance for improving the safety and use efficiency of electromagnetic equipment.

In the embodiment of the present invention, referring to FIG. 1 , a schematic diagram of a step flow of a method for evaluating the health status of an electromagnetic coil based on image features of the present invention is shown. In this example, the method for evaluating the health status of an electromagnetic coil based on image features includes the following steps:

Step S1: acquiring an electromagnetic coil image; performing super-resolution reconstruction on the electromagnetic coil image to generate a reconstructed electromagnetic coil image; performing visual feature optimization on the reconstructed electromagnetic coil image to obtain an optimized coil image;

In the embodiment of the present invention, a 5-megapixel digital camera is used to shoot the electromagnetic coil at a 45-degree angle and save it in a 24-bit color depth, uncompressed RGB image format, and the image size is 2560×1920 pixels. A super-resolution reconstruction method based on deep learning is adopted, and the original 2560×1920 pixel electromagnetic coil image is input. After being processed by 3 layers of convolutional neural networks and 2 layers of deconvolution layers, a reconstructed electromagnetic coil image with a resolution of 5120×3840 pixels is generated. The network training uses the MS-COCO data set, the loss function is PSNR loss, the optimizer is Adam, and the number of training epochs is 50. Unsharp Masking edge sharpening is performed on the reconstructed image to improve the image details, the sharpening factor is set to 1.2, and the radius is 1.5 pixels. Then the histogram equalization technology is used to enhance the contrast of the image, and the edge and texture features of the coil are highlighted by dynamically adjusting the pixel gray value distribution. Finally, the non-local mean denoising algorithm is used to eliminate image noise, the filter intensity coefficient is 0.6, and the search window is 21×21 pixels.

Step S2: acquiring a standard coil image; performing covering object recognition on the optimized coil image data to generate surface attachment data; performing deformation analysis on the optimized coil image according to the standard coil image and the surface attachment data to generate coil deformation data;

In the embodiment of the present invention, a high-definition RGB image of 2560x1920 pixels is obtained from the standard coil object using the same equipment and parameters as the standard coil reference image. The material, shape, and size features of the coil in the image are kept completely consistent, without any surface attachments. A semantic segmentation network based on deep learning is used, and the optimized 5120x3840 pixel coil image is input, and the U-Net network structure of encoding and decoding is trained to identify dust, oil stains, and corroded surface attachments in the image, generate a mask map, and annotate the distribution of surface attachments at the pixel level. The network training uses a self-built electromagnetic coil attachment data set, the loss function is cross entropy loss, the optimizer is SGD, and the number of training epochs is 100. First, the image registration technology based on SIFT feature points is used to accurately align the optimized coil image with the standard coil image. Then the optical flow field between the two images is calculated to obtain the displacement vector of each pixel point, and the coil deformation is quantitatively analyzed. Finally, a coil deformation heat map is generated based on the displacement vector data to intuitively reflect the degree of local deformation of the coil.

Step S3: performing regional light reflection analysis on the optimized coil image based on the coil deformation data to generate coil color change data;

In the embodiment of the present invention, the optimized coil image is taken in 16-bit RAW format, and the reflected light imaging data with sufficient dynamic range is obtained by adjusting the exposure time and gain. The image is radiated and corrected using a calibrated reflectivity reference plate to eliminate the influence of uneven light source intensity factors and generate accurate initial reflected light data. The image size is 5120x3840 pixels, and the reflected light brightness value is quantized to a grayscale of 0-65535. Based on the Beckmann-Spizzichino reflection model, combined with the coil surface roughness parameter, the initial reflected light data is subjected to microscopic optical analysis. By simulating the normal vector randomly distributed on the micro surface, the specular reflection and diffuse reflection components of each point are calculated to generate reflected light data reflecting the microscopic optical characteristics. The model parameters are set to: surface RMS roughness 0.02um, relative dielectric constant 5.0, and incident angle 45 degrees. The entire analysis process is optimized by parallel computing, and the processing speed reaches 1 second per square meter. The coil deformation heat map is matched one by one with the optimized coil image, and the correlation coefficient matrix between the deformation amount and the color difference value is calculated by statistical analysis. A mathematical model describing the relationship between deformation and color change is obtained, which provides a basis for subsequent pre-compensation. During the analysis, the RANSAC algorithm is used to fit the linear model, and the cross-validation method is used to evaluate the model performance. The R^2 goodness of fit reaches above 0.85. The microscopic reflected light data is mapped through the deformation-color change association model to estimate the color difference at each position. This color difference is then superimposed on the initial reflected light data to correct the color distortion caused by the coil deformation and generate the final coil color change data. The entire pre-compensation process is accelerated by GPU parallelism, and the processing time per frame is less than 20 milliseconds.

Step S4: based on the coil deformation data and the coil color change data, reversely deduce and trace the coil variation data to obtain the coil variation factor; perform variation field simulation on the coil variation factor to generate simulated coil development data;

In the embodiment of the present invention, the optimized coil image is taken in 16-bit RAW format, and the reflected light imaging data with sufficient dynamic range is obtained by adjusting the exposure time and gain. The image is radiated and corrected using a calibrated reflectivity reference plate to eliminate the influence of uneven light source intensity factors and generate accurate initial reflected light data. The image size is 5120x3840 pixels, and the reflected light brightness value is quantized to a grayscale of 0-65535. Based on the Beckmann-Spizzichino reflection model, combined with the coil surface roughness parameter, the initial reflected light data is subjected to microscopic optical analysis. By simulating the normal vector randomly distributed on the micro surface, the specular reflection and diffuse reflection components of each point are calculated to generate reflected light data reflecting the microscopic optical characteristics. The model parameters are set to: surface RMS roughness 0.02um, relative dielectric constant 5.0, and incident angle 45 degrees. The entire analysis process is optimized by parallel computing, and the processing speed reaches 1 second per square meter. The coil deformation heat map is matched one by one with the optimized coil image, and the correlation coefficient matrix between the deformation amount and the color difference value is calculated by statistical analysis. A mathematical model describing the relationship between deformation and color change is obtained, which provides a basis for subsequent pre-compensation. During the analysis process, the RANSAC algorithm is used to fit the linear model, and the cross-validation method is used to evaluate the model performance to ensure that the goodness of fit reaches above 0.85. The microscopic reflected light data is mapped through the deformation-color change association model to estimate the color difference at each position. This color difference is then superimposed on the initial reflected light data to correct the color distortion caused by the coil deformation and generate the final coil color change data. The entire pre-compensation process is accelerated by GPU parallelism, and the processing time per frame is less than 20 milliseconds.

Step S5: Perform health status assessment on the electromagnetic coil image according to the simulated coil development data to execute the electromagnetic coil health status assessment method based on image features.

In an embodiment of the present invention, the simulated coil development data is feature matched with the actually collected electromagnetic coil image. A deep convolutional neural network model is used to extract the key features of the geometric shape and material texture in the electromagnetic coil image, and compare and analyze it with the corresponding features in the simulated data. The mapping function between the actual development degree of the coil and the simulated data is fitted by the least squares method, and the development degree index of each electromagnetic coil sample is given. The network model parameters are set as: convolution kernel size 3x3, number of channels 32-64-128, number of nodes in the fully connected layer 512-256. The Adam optimizer is used, and the loss function is MSE. Based on the electromagnetic coil development degree data, a coil life prediction model is established through accelerated aging tests. The Weibull distribution is used to fit the coil failure time, and the parameters are set as: shape parameter 3.2, scale parameter 12000 hours. Substituting the development degree data into the model, the estimated life of each electromagnetic coil sample can be calculated. In order to improve the prediction accuracy, while considering the influence of workload and environmental factors on life, a comprehensive prediction model is constructed using the multivariate linear regression method. The electromagnetic coil life threshold is set to 20,000 hours, and coils below this threshold are judged to be in abnormal state. The estimated life data is compared with the threshold to give the health status assessment results of each coil sample. For abnormal coils, the deviation between their image features and simulation data can be further analyzed to obtain more detailed fault diagnosis results. For example, severe local deformation can cause insulation damage, and color deviation reflects the decline of the magnetic properties of the core. The entire status assessment process is fully automatic, and the results are presented in the form of a report.

Preferably, step S1 comprises the following steps:

Step S11: capturing an electromagnetic coil image by a digital camera;

Step S12: super-resolution reconstruction is performed on the electromagnetic coil image to obtain a high-resolution image;

Step S13: sharpening the edges of the high-resolution image to generate a reconstructed electromagnetic coil image;

Step S14: performing image feature enhancement on the reconstructed electromagnetic coil image to generate an enhanced coil image;

Step S15: performing histogram equalization processing on the enhanced coil image to obtain an optimized coil image.

The benefit of the present invention lies in that by acquiring the electromagnetic coil image through a digital camera, it is possible to ensure high resolution and good quality of the original image data, providing a reliable basis for subsequent image processing and analysis. Through super-resolution reconstruction and edge sharpening processing, the image details can be enhanced, making the coil structure and surface features more clearly visible, which is conducive to subsequent feature extraction and analysis. Image feature enhancement and histogram equalization can optimize the image contrast and brightness distribution, highlight the key visual features of the coil, and further improve the reliability of the analysis. It can effectively improve the image quality and usability, lay a good foundation for the subsequent covering identification, deformation analysis, and light reflection analysis steps, and ensure the accuracy and effectiveness of the entire evaluation process. It is crucial for the electromagnetic coil health status assessment method based on image features. By optimizing the quality of acquiring and processing image data, the reliability and accuracy of the entire evaluation process can be greatly improved, providing support for more accurate coil status diagnosis.

In an embodiment of the present invention, a 12-megapixel CMOS digital camera is used to capture high-definition close-up images of electromagnetic coils under constant lighting conditions. The camera parameters are set as: ISO 200, aperture f/8, shutter speed 1/125s. The image acquisition distance is maintained at about 30cm to ensure that the coil structure details are fully captured. During acquisition, an LED fill light is used to eliminate ambient light interference and obtain a clear original coil image. The original coil image is input into a super-resolution reconstruction network based on deep learning. The network adopts an ESPCN structure, including 3 convolutional layers and 1 subpixel convolutional layer. The convolution kernel size is set to 3x3, and the number of channels is 64-32-3. During training, the MSE loss function, Adam optimizer, and learning rate are 0.001. The trained network can increase the input image resolution by 4 times and output a high-definition coil image of 16 million pixels. The high-resolution coil image is input into the Unsharp Masking algorithm for edge sharpening. The algorithm parameters are set as: radius is 1.5 pixels and intensity is 0.6. This process can highlight the key contour features of the coil winding and insulation layer, eliminate blur, and generate a clear reconstructed electromagnetic coil image. The Adaptive Histogram Equalization (CLAHE) algorithm is used to perform local contrast enhancement on the reconstructed image. The image block size is 8x8 pixels, and each sub-block is individually histogram equalized and reorganized using bilinear interpolation. This method can highlight the texture features of the coil surface, improve the overall visual quality and information content of the image, and lay the foundation for subsequent analysis. The enhanced coil image is input into the global histogram equalization algorithm for grayscale adjustment. By stretching the image grayscale histogram, the overall contrast is improved, making the image brightness distribution more uniform.

Preferably, step S2 comprises the following steps:

Step S21: Obtaining a standard coil image by viewing a log;

Step S22: performing feature layer overlay comparison on the annotated coil image and the optimized coil image data to generate feature image difference data;

Step S23: performing covering object recognition on the optimized coil image according to the characteristic image difference data to generate surface attachment data;

Step S24: performing deformation analysis on the optimized coil image based on the surface attachment data to generate coil deformation data.

The benefit of the present invention lies in that by viewing the log to obtain the standard coil image, a reliable benchmark is provided for the subsequent comparative analysis, ensuring the accuracy and pertinence of the subsequent analysis. By superimposing and comparing the feature layers, the subtle difference features between the standard coil image and the optimized coil image can be deeply explored, providing more accurate input data for covering identification and deformation analysis. Based on the image difference feature data, the surface attachments on the optimized coil image can be more accurately identified, providing reliable basic data for subsequent deformation analysis. By making full use of the surface attachment data, the geometric deformation features in the optimized coil image can be more accurately analyzed, providing an important basis for accurately diagnosing the coil status. By giving full play to the advantages of the reference standard image and image difference analysis, the key data reflecting the coil degradation characteristics can be more accurately extracted, laying a solid foundation for the subsequent health status assessment. This is crucial to improving the accuracy and reliability of the entire assessment process.

As an example of the present invention, referring to FIG. 2 , in this example, step S2 includes:

Step S21: Obtaining a standard coil image by viewing a log;

In the embodiment of the present invention, standard image samples of electromagnetic coils at the time of initial mass production are extracted from the quality management information system. The system automatically records multi-angle high-definition coil images collected at the initial stage of production and testing of each batch of products, and classifies and archives them. Through the system query function, standard electromagnetic coil images that are highly representative and have no obvious defects when the coil products are just put into mass production can be searched.

Step S22: performing feature layer overlay comparison on the annotated coil image and the optimized coil image data to generate feature image difference data;

In an embodiment of the present invention, the standard coil image and the optimized coil image obtained from the quality management system are uploaded to the image processing software. The key features of the two sets of images, such as coil shape, wire diameter, and number of turns, are aligned and superimposed one by one. The system automatically calculates the pixel difference between the two layers and generates a comparison chart reflecting the feature difference. In the comparison chart, the area with darker and darker colors indicates the degree of feature difference between the two images at that location. Further analysis is performed on the areas with larger differences to find the deviation of the appearance features between the optimized coil image and the standard sample.

Step S23: performing covering object recognition on the optimized coil image according to the characteristic image difference data to generate surface attachment data;

In the embodiment of the present invention, a regional segmentation algorithm is used to perform fine cutting on the feature difference map. The boundary lines of the difference area are identified, and the image is cut into multiple sub-image blocks along these boundary lines. Each sub-image block contains the difference information of a non-repeating area, which can be used for subsequent targeted analysis. During the cutting process, the coordinate position and size attribute data of each sub-image are recorded to provide a positioning basis for surface texture analysis. For the difference cutting image, the texture analysis algorithm based on the gray level co-occurrence matrix is used to extract its surface texture features. The quantitative indicators of each sub-image block in terms of contrast, roughness, and regularity are automatically calculated to form a complete surface texture data. These data are used to describe the material characteristics of the coil surface, such as whether there are impurity particles and whether the surface is flat and smooth. By comparing the differences between the standard coil image and the optimized coil image in these texture indicators, the optimization effect can be further analyzed to find the existing surface attachment problems. Based on the coil surface texture data, an algorithm based on texture gradient analysis is used to quantitatively evaluate its fluency. The algorithm automatically calculates the smoothness of the texture change of each difference cutting image block in the local area, forming a quantitative data reflecting the overall fluency. Areas with high fluency indicate that the texture transition of the coil surface is natural, without obvious mutations or discontinuities; while areas with low fluency have surface attachments. By comparing the difference in fluency index between the standard coil image and the optimized coil image, the problem areas with poor optimization effect can be further identified. The optimized coil image is analyzed using the abnormal area detection algorithm combined with the texture fluency data. Identify areas with abnormally low fluency and mark them on the image with borders to identify them as problem areas with surface attachments. At the same time, the system will record the location coordinates and quantitative data of the area size of these attachments and generate a detailed surface attachment report. This report can be used to guide subsequent image optimization work, helping engineers quickly locate problematic areas and take targeted cleaning or corrective measures.

Step S24: performing deformation analysis on the optimized coil image based on the surface attachment data to generate coil deformation data.

In the embodiment of the present invention, a three-dimensional reconstruction algorithm based on machine vision is used to construct a three-dimensional model of the coil using optimized coil image data. The algorithm automatically extracts feature points in the coil image, calculates the three-dimensional coordinates of each feature point based on the geometric relationship between the multi-view images, and assembles a complete three-dimensional model. The model will contain the contour shape, winding direction, and end structure of the coil, providing accurate geometric reference for subsequent surface contamination analysis and structural deformation detection. The surface attachment data is associated with the three-dimensional coil model for analysis. First, according to the position coordinate information of the attachment, these contaminated areas are accurately located on the three-dimensional model. Then, the model surface is locally repaired using a three-dimensional modeling tool to simulate the clean state of the coil after removing these contaminants. Through this simulation method based on measured data, a three-dimensional model of a clean coil close to reality can be obtained, providing a reliable reference for the next step of structural analysis. A three-dimensional reconstruction algorithm is used to construct a three-dimensional model of the coil using standard coil image data. The model will contain the contour shape, winding direction, and end structure information of the standard coil as a reference object for subsequent structural deformation analysis. The three-dimensional model of the clean coil is compared and analyzed with the step standard coil model. First, the registration tool of the 3D modeling software is used to align the two models in terms of position and orientation attributes. Then, the finite element analysis method is used to calculate the geometric deformation of the clean model in each area relative to the standard model. These deformation data include displacement, deformation, and stress indicators, which are used to reflect the changes that will occur in the coil structure during the optimization process. Through comparative analysis, the optimization effect is further evaluated to understand whether the coil structure is affected by surface contamination, which provides an important basis for subsequent coil performance tuning.

Preferably, step S23 includes the following steps:

Step S231: performing difference image cutting on the optimized coil image according to the characteristic image difference data to generate a difference cutting image;

Step S232: performing surface texture analysis on the difference cut image to obtain coil surface texture data;

Step S233: performing a fluency analysis on the coil surface texture data to generate texture fluency data;

Step S234: Based on the texture fluency data, the optimized coil image is subjected to cover identification to generate surface attachment data.

The benefit of the present invention is that by performing differential image cutting on the optimized coil image, the area where there is a significant difference between the standard coil image and the optimized coil image can be accurately located, providing accurate data input for subsequent surface texture analysis, which is crucial to improving the accuracy of analysis. A detailed surface texture analysis of the differential cut image can extract more subtle and complex texture features on the coil surface, which is more comprehensive than the traditional simple texture analysis. This provides a richer and more accurate data basis for the subsequent fluency analysis. By performing a fluency analysis on the surface texture data of the coil, the state of the surface texture can be more accurately evaluated. This analysis method based on texture fluency is more reasonable than the traditional texture analysis method. Using texture fluency data as the basis for identifying surface attachments can significantly improve the accuracy of covering identification, provide more reliable input data for subsequent health status assessment, and have a strong application value. The advantages of image difference analysis and surface texture feature extraction are fully utilized, and fluency analysis is carried out on this basis, and finally a more accurate surface attachment identification is achieved.

In the embodiment of the present invention, a regional segmentation algorithm is used to perform fine cutting on the feature difference map. The boundary lines of the difference area are identified, and the image is cut into multiple sub-image blocks along these boundary lines. Each sub-image block contains the difference information of a non-repeating area, which can be used for subsequent targeted analysis. During the cutting process, the coordinate position and size attribute data of each sub-image are recorded to provide a positioning basis for surface texture analysis. For the difference cutting image, the texture analysis algorithm based on the gray level co-occurrence matrix is used to extract its surface texture features. The quantitative indicators of each sub-image block in terms of contrast, roughness, and regularity are automatically calculated to form a complete surface texture data. These data are used to describe the material characteristics of the coil surface, such as whether there are impurity particles and whether the surface is flat and smooth. By comparing the differences between the standard coil image and the optimized coil image in these texture indicators, the optimization effect can be further analyzed to find the existing surface attachment problems. Based on the coil surface texture data, an algorithm based on texture gradient analysis is used to quantitatively evaluate its fluency. The algorithm automatically calculates the smoothness of the texture change of each difference cutting image block in the local area, forming a quantitative data reflecting the overall fluency. Areas with high fluency indicate that the texture transition of the coil surface is natural, without obvious mutations or discontinuities; while areas with low fluency have surface attachments. By comparing the difference in fluency index between the standard coil image and the optimized coil image, the problem areas with poor optimization effect can be further identified. The optimized coil image is analyzed using the abnormal area detection algorithm combined with the texture fluency data. Identify areas with abnormally low fluency and mark them on the image with borders to identify them as problem areas with surface attachments. At the same time, the system will record the location coordinates and quantitative data of the area size of these attachments and generate a detailed surface attachment report. This report can be used to guide subsequent image optimization work, helping engineers quickly locate problematic areas and take targeted cleaning or corrective measures.

Preferably, step S24 includes the following steps:

Step S241: performing three-dimensional reconstruction on the optimized coil image to generate a three-dimensional model of the electromagnetic coil;

Step S242: Analyze the surface contamination of the electromagnetic coil three-dimensional model based on the surface attachment data to obtain a simulated electromagnetic coil cleaning model;

Step S243: reconstructing the standard coil image in three dimensions to generate a standard electromagnetic coil model;

Step S244: using the standard electromagnetic coil model to perform structural deformation analysis on the simulated electromagnetic coil cleaning model to generate coil deformation data.

The invention is beneficial in that by performing three-dimensional reconstruction on the optimized coil image, a three-dimensional model of the electromagnetic coil can be generated, providing visual data input for subsequent surface contamination analysis and structural deformation analysis, which is of great significance for more intuitive analysis of the coil state. Based on the surface attachment data, the surface contamination analysis of the three-dimensional model can be performed to generate a simulated electromagnetic coil cleaning model. This step provides an ideal benchmark reference for subsequent structural deformation analysis, which helps to more accurately identify the deformation of the coil during actual use. By performing three-dimensional reconstruction on the standard coil image, a standard electromagnetic coil model is generated. This provides a reliable comparison benchmark for subsequent structural deformation analysis, which helps to more accurately evaluate the difference between the optimized coil and the standard coil. Using the standard electromagnetic coil model, the structural deformation analysis of the simulated electromagnetic coil cleaning model can generate coil deformation data. These data provide key inputs for the final health status assessment, which helps to diagnose the actual use status of the coil more comprehensively and accurately. The advantages of three-dimensional model reconstruction and comparative analysis are fully utilized, and a reliable data basis is provided for structural deformation analysis by establishing a simulated cleaning model and a standard comparison model. This not only improves the accuracy of the entire evaluation process, but also lays a solid foundation for the comprehensive diagnosis of the health status of the coil.

In the embodiment of the present invention, a three-dimensional reconstruction algorithm based on machine vision is used to construct a three-dimensional model of the coil using optimized coil image data. The algorithm automatically extracts feature points in the coil image, calculates the three-dimensional coordinates of each feature point based on the geometric relationship between the multi-view images, and assembles a complete three-dimensional model. The model will contain the contour shape, winding direction, and end structure of the coil, providing accurate geometric reference for subsequent surface contamination analysis and structural deformation detection. The surface attachment data is associated with the three-dimensional coil model for analysis. First, according to the position coordinate information of the attachment, these contaminated areas are accurately located on the three-dimensional model. Then, the model surface is locally repaired using a three-dimensional modeling tool to simulate the clean state of the coil after removing these contaminants. Through this simulation method based on measured data, a three-dimensional model of a clean coil close to reality can be obtained, providing a reliable reference for the next step of structural analysis. A three-dimensional reconstruction algorithm is used to construct a three-dimensional model of the coil using standard coil image data. The model will contain the contour shape, winding direction, and end structure information of the standard coil as a reference object for subsequent structural deformation analysis. The three-dimensional model of the clean coil is compared and analyzed with the step standard coil model. First, the registration tool of the 3D modeling software is used to align the two models in terms of position and orientation attributes. Then, the finite element analysis method is used to calculate the geometric deformation of the clean model in each area relative to the standard model. These deformation data include displacement, deformation, and stress indicators, which are used to reflect the changes that will occur in the coil structure during the optimization process. Through comparative analysis, the optimization effect is further evaluated to understand whether the coil structure is affected by surface contamination, which provides an important basis for subsequent coil performance tuning.

Preferably, step S3 comprises the following steps:

Step S31: capturing regional light reflection of the optimized coil image to generate initial reflected light data;

Step S32: performing microscopic light reflection analysis on the initial reflected light data to generate microscopic reflected light data;

Step S33: performing deformation-color change correlation analysis on the optimized coil image based on the coil deformation data to obtain deformation-color change correlation data;

Step S34: using the microscopic reflected light data and the deformation-color change correlation data to perform pre-compensation color change analysis on the initial reflected light data to generate coil color change data.

The benefit of the present invention is that by capturing the regional light reflection of the optimized coil image, the initial reflected light data can be obtained. These data provide an important basis for the subsequent microscopic light reflection analysis and deformation-color change correlation analysis. Based on the initial reflected light data, further microscopic light reflection analysis can be carried out to extract more detailed and complex reflected light features. This in-depth analysis helps to better understand the actual state of the coil surface. By performing deformation-color change correlation analysis on the coil deformation data and the optimized coil image, the intrinsic connection between the two can be found. This provides an important basis for the subsequent pre-compensation color change analysis. Using microscopic reflected light data and deformation-color change correlation data, pre-compensation color change analysis of the initial reflected light data can generate more accurate coil color change data. This provides a more reliable input for the final health status assessment. The light reflection characteristics, deformation characteristics and the relationship between the two are fully explored, and through in-depth analysis, accurate pre-compensation of coil color change is finally achieved. This series of innovative analysis methods not only improves the accuracy of the entire evaluation process, but also reflects the innovation and advancement of this method in optical feature analysis.

As an example of the present invention, referring to FIG3 , in this example, step S3 includes:

Step S31: capturing regional light reflection of the optimized coil image to generate initial reflected light data;

In an embodiment of the present invention, a light reflection capture technology based on machine vision is adopted, and a high-precision digital camera is used to shoot the surface of the optimized coil under multiple angles and multiple lighting conditions. The camera sensor will finely collect the light reflection characteristic data of each area on the surface of the coil, including the reflected light intensity and the reflected light angle parameters. To ensure the accuracy and reliability of the data, the camera position and the shooting conditions of the light source direction need to be strictly controlled and calibrated. At the same time, the collected raw image data needs to be denoised and corrected preprocessed to eliminate the interference of external environmental factors. The processed reflected light data is saved in a matrix form, covering the overall information of the coil surface, providing a basic basis for subsequent surface state analysis.

Step S32: performing microscopic light reflection analysis on the initial reflected light data to generate microscopic reflected light data;

In an embodiment of the present invention, a microscopic light reflection analysis algorithm based on machine learning is used to conduct an in-depth analysis of the initial reflected light data. First, the reflected light data is spatially resolved with high resolution to identify the microscopic texture features of different areas on the coil surface. Then, based on the pre-trained reflected light model, the microscopic reflection parameters of each area are calculated, including surface roughness, reflectivity, and normal angle indicators. By analyzing these microscopic reflection characteristics, the material state and contamination degree information of the coil surface can be inferred. To improve the analysis accuracy, the algorithm will combine the 3D geometric model data of the coil to perform angle correction and lighting effect compensation on the reflected light data.

Step S33: performing deformation-color change correlation analysis on the optimized coil image based on the coil deformation data to obtain deformation-color change correlation data;

In an embodiment of the present invention, a deformation-color change correlation analysis model based on deep learning is adopted, and correlation analysis is performed using optimized coil image data and corresponding coil geometric deformation data. The model first performs high-precision image segmentation on the input coil image to accurately extract various areas on the coil surface. Then, the microscopic reflected light data of these areas are deeply correlated with the corresponding deformation parameters (such as strain and deformation angle). Through training with a large amount of sample data, the deformation-color change coupling law of different areas on the coil surface is established. When analyzing a new coil image, the geometric deformation state of the area is quickly predicted based on the reflected light information of the surface area. The analysis results are saved in the form of a highly compressed tensor, which contains the deformation-color change mapping relationship of each area on the coil surface.

Step S34: using the microscopic reflected light data and the deformation-color change correlation data to perform pre-compensation color change analysis on the initial reflected light data to generate coil color change data.

In the embodiment of the present invention, full use is made of microscopic reflected light data and deformation-color change correlation data. The algorithm first maps the initial reflected light data to each microscopic area on the coil surface, and corrects the original data according to the corresponding microscopic reflection characteristics, and uses the deformation-color change correlation model to predict the geometric deformation state of each area, and accordingly performs secondary compensation on the reflected light data.

Preferably, step S34 includes the following steps:

Step S341: performing deformation influence analysis on the initial reflected light data based on the deformation-color change correlation data to generate a deteriorated optical feature;

Step S342: using the deteriorated optical feature to perform deteriorated light stripping on the initial reflected light data to obtain clean reflected light data;

Step S343: performing light reflection analysis on the standard coil image to generate standard light reflection data;

Step S344: Perform color change analysis on the microscopic reflected light data and the clean reflected light data based on the standard light reflection data to generate coil color change data.

The benefit of the present invention is that by analyzing the deformation-color change correlation data, the deterioration optical characteristics, that is, the influence of the coil deformation on the reflected light data, can be identified. This lays the foundation for the subsequent deterioration light stripping. Using the deterioration optical characteristics, the initial reflected light data is deteriorated light stripped to obtain clean reflected light data. This step ensures that only the optical characteristics of the coil itself are considered in the subsequent analysis without being disturbed by the deformation. The standard coil image is subjected to light reflection analysis to generate standard light reflection data. This provides a reliable comparison benchmark for the subsequent color change analysis and helps to more accurately evaluate the actual state of the coil. Based on the standard light reflection data, the clean reflected light data and the microscopic reflected light data are subjected to color change analysis, and finally the coil color change data is generated. These data provide key inputs for health status assessment, reflecting the innovation of the method in optical feature analysis. By separating the deformation effect and establishing a standard benchmark, the accurate analysis of the coil color change is finally achieved. This not only improves the reliability of the entire evaluation process, but also further enriches the coil health diagnosis method based on image features, providing more accurate data support for practical applications.

In the embodiment of the present invention, a deformation-color change association model based on deep learning is adopted. The model establishes a complete set of deformation-color change association data by accurately mapping and modeling the geometric deformation parameters of the coil surface with the corresponding optical property changes. The obtained initial reflected light data is accurately matched with the deformation parameters to determine the specific deformation state of each surface area. Then, based on the trained mapping relationship, the algorithm can predict the optical property changes of these areas, including reflectivity and color shift indicators. By analyzing the entire surface area, a complete set of deteriorated optical feature data is generated, which describes the various optical effects caused by the geometric deformation of the coil surface in the form of a high-dimensional tensor. Deteriorated optical feature data, accurate optical correction of the original reflected light data. First, the geometric deformation parameters in the deteriorated optical features are accurately matched with the corresponding reflected light data. Then, based on the trained mapping relationship, the optical property changes of each area, including reflectivity and color shift indicators, are predicted. According to these change information, the original reflected light data is accurately mathematically corrected and filtered to remove various optical distortions and color shift effects caused by geometric deformation. A detailed light reflection analysis is performed on the standard coil image, that is, the coil sample under ideal conditions. First, the standard coil image is segmented and the region is identified to determine the geometric features of each surface area. Then, based on the optical model, the ideal reflected light characteristics of each area under standard conditions are calculated, including reflectivity and color indicators. These data constitute the standard light reflection data, which describes the optical characteristic benchmark of the coil without any deformation. The standard light reflection data is used to perform in-depth color analysis and comparison of the clean reflected light data and the original microscopic reflected light data. First, the clean reflected light data and the microscopic reflected light data are accurately matched and aligned with the standard light reflection data. Then, based on color theory, the color difference indicators between the two are calculated, including hue, saturation, and brightness dimensions.

Preferably, step S4 comprises the following steps:

Step S41: integrating the variation results of the coil deformation data and the coil color change data to obtain the coil variation data;

Step S42: reversely tracing the coil variation data to obtain the coil variation factors;

Step S43: constructing a virtual coil for the electromagnetic coil image to obtain a virtual coil model;

Step S44: performing variation field simulation on the virtual coil model based on the coil variation factor to generate simulated coil development data.

The benefit of the present invention lies in that more comprehensive coil change data is obtained by integrating coil deformation data and coil color change data. This provides more abundant input information for subsequent reverse deduction and tracing. Based on the integrated coil change data, the key factors that cause the coil to change can be found by the reverse deduction method. This is of great significance for accurately diagnosing the health status of the coil and preventing potential problems. By constructing a virtual coil for the electromagnetic coil image, a digital model that can be simulated and analyzed is obtained. This provides a basis for subsequent variation field simulation. Using the determined coil variation factors, the virtual coil model is simulated for the variation field to generate simulated coil development data. These data not only reflect the future development trend of the coil, but also provide strong support for the formulation of targeted maintenance strategies. In-depth integration and analysis of coil change information is achieved, the root causes of the changes are found, and a virtual model is constructed based on this for trend prediction. This not only improves the accuracy and reliability of coil health status assessment, but also provides an important basis for formulating preventive and maintenance measures, and has high value in practical applications.

In an embodiment of the present invention, the coil deformation data and the coil color change data are fully connected and integrated. Based on a highly coordinated and consistent data model, the two types of change indicators of different dimensions are precisely matched and mapped to generate a complete coil change data set. This data set not only contains detailed information on the geometric deformation of the coil surface, but also integrates the corresponding changes in optical properties, such as reflectivity and color. The coil change data is used to determine the root causes of these changes by reverse deduction. First, a coil variation analysis model based on machine learning is established, which can capture the various laws and correlations contained in the coil change data. Then, by reverse deduction, starting from the coil change data, the root causes of these changes, such as material degradation, magnetic field interference, and mechanical stress, are determined. For the original electromagnetic coil image obtained, advanced computer vision technology is used for three-dimensional reconstruction and modeling to generate a virtual model that highly restores the actual coil. First, the coil image is carefully segmented and feature extracted to determine the geometric size, material properties, and winding structure information of the coil. Then, based on this information, a finite element-based modeling method is used to construct a high-fidelity three-dimensional virtual coil model. This model can not only accurately describe the physical form of the coil, but also simulate its internal electromagnetic field distribution and flux link parameters. The coil variation factors are imported into the virtual coil model for in-depth simulation analysis and prediction. First, according to the specific characteristics of the variation factors, such as the degree of material degradation and changes in magnetic field strength, the relevant parameters of the virtual coil model are modified and adjusted in a targeted manner. Then, based on advanced electromagnetic field calculation methods, the effects of these variation factors on the electromagnetic characteristics and flux distribution indicators of the coil are simulated. Through a comprehensive analysis of this series of simulation results, a set of highly reliable simulated coil development data is finally output, including performance attenuation curves and insulation layer degradation trends, which provide reliable basic data for subsequent life prediction and fault analysis.

Preferably, step S42 includes the following steps:

Step S421: Analyze the change trend of the coil change data to obtain the coil change mode;

Step S422: performing variation mechanism driving processing on the coil variation pattern to generate a variation mechanism model;

Step S423: using the variation mechanism model to perform parameter inversion on the coil variation data to generate the maximum contribution parameter;

Step S424: using the variation mechanism model to perform sensitivity analysis on the coil variation data to generate a maximum sensitivity parameter;

Step S425: jointly analyze the maximum contribution parameter and the maximum sensitivity parameter to obtain the coil variation factor.

The benefit of the present invention is that by performing trend analysis on the coil change data, the typical pattern of coil change can be identified. This lays the foundation for the subsequent exploration of the variation mechanism. Based on the variation pattern, a mathematical model capable of describing the coil variation process is established by adopting the variation mechanism driven method. This provides a theoretical basis for parameter inversion and sensitivity analysis. The variation mechanism model is used to perform parameter inversion on the coil change data to find the key parameters that contribute the most to the coil change. This helps to accurately diagnose the health status of the coil. By performing sensitivity analysis on the variation mechanism model, the parameters that are most sensitive to coil changes are found. This helps to predict the potential failure points of the coil and provide a basis for formulating maintenance strategies. By jointly analyzing the maximum contribution parameters and the maximum sensitivity parameters, the key factors that cause the coil to change are finally determined. This provides an important basis for subsequent health status assessment and fault diagnosis. By analyzing the coil change pattern and constructing the variation mechanism model, the key parameters that cause the coil change are deeply excavated, which not only improves the accuracy of the health status assessment, but also provides an important basis for preventive maintenance. This method based on the combination of data-driven and physical mechanisms reflects the scientificity and advancement of the evaluation method.

In an embodiment of the present invention, in-depth statistical analysis and trend mining are performed on the coil change data. Time series analysis and pattern recognition methods are used to identify typical patterns of changes in various performance indicators of the coil (such as geometric dimensions and electromagnetic characteristics) over time. For example, some parameters show a linear attenuation trend, while other parameters show a step-like change. By modeling and fitting a large amount of historical data, various change laws and patterns of the coil during use can be accurately captured. These change patterns provide an important basis for subsequent fault diagnosis and life prediction. Further in-depth exploration of the root mechanism that leads to these change patterns. First, a series of variation mechanism models based on physical-chemical-mechanical principles are established. These models can accurately describe the influence of material degradation, magnetic field interference, and thermal stress factors on coil performance. Then, the variation pattern is matched and driven with these physical mechanism models, and through repeated iterative optimization, a set of variation mechanism models that are highly consistent with the actual observed data are finally obtained. The variation mechanism model is used to conduct in-depth analysis of the coil change data. Specifically, the parameter inversion method is used to adjust the values of various parameters in the variation mechanism model to fit the observed coil change data to the greatest extent. Through this process, it is possible to determine which parameters have the greatest contribution to the change in coil performance. For example, the change in material stiffness coefficient is the main factor causing the change in geometric dimensions, while the increase in winding resistance is the key to the deterioration of electromagnetic characteristics. The variation mechanism model is used to perform sensitivity analysis on the coil variation data. Specifically, for each parameter in the variation mechanism model, the degree of influence of the parameter on the coil performance is analyzed one by one. For example, through perturbation analysis, it is possible to determine how a slight change in the material thermal expansion coefficient will cause a change in the coil size; it is possible to calculate how much the increase in the number of turns around the coil will affect the coil inductance. Through this sensitivity analysis, it is possible to determine which parameters have the greatest sensitivity to the change in coil performance, and these maximum sensitivity parameters provide a key basis for subsequent fault diagnosis and life prediction. The maximum contribution parameters and maximum sensitivity parameters are deeply compared and analyzed and comprehensively processed to determine the key factors that cause the change in coil performance. For example, if it is found that the change in material stiffness coefficient is both the main factor causing the change in geometric dimensions and has significant sensitivity to the change in electromagnetic characteristics, then this parameter can be identified as the most critical coil variation factor. By jointly analyzing multiple parameters, it is possible to more accurately determine which factors are the root causes of the deterioration of the overall performance of the coil. This analysis result provides reliable basic data for subsequent fault diagnosis and life prediction.

Preferably, step S5 comprises the following steps:

Step S51: matching the development degree of the electromagnetic coil image according to the simulated coil development data to obtain the electromagnetic coil development degree data;

Step S52: estimating the life of the electromagnetic coil development degree data to obtain estimated electromagnetic coil life data;

Step S53: performing a health status assessment on the estimated electromagnetic coil life data based on a preset electromagnetic coil life threshold, so as to execute an electromagnetic coil health status assessment method based on image features.

The benefit of the present invention is that by comparing and matching the simulated coil development data with the actual electromagnetic coil image, the development degree of the current coil can be accurately evaluated. This lays the foundation for subsequent life prediction. Based on the coil development degree data, the life prediction model is used for analysis to obtain the estimated remaining life of the electromagnetic coil. This provides a quantitative basis for judging the health status of the coil. Combined with the preset life threshold, the estimated coil life data is analyzed to obtain the health status of the current coil. This evaluation method based on image features can more objectively and comprehensively reflect the actual condition of the coil. The results of the health status evaluation can provide a basis for the preventive maintenance of the coil. For example, for a coil that is about to reach the life threshold, a more timely and effective maintenance plan can be formulated to extend its service life. The simulation data and variation mechanism analysis accumulated in the early stage are fully utilized to make the coil health status evaluation more accurate and reliable. This helps to reduce the uncertainty in the evaluation process and improve the overall applicability of the method. The complete process from coil development degree evaluation to life prediction and then to health status judgment is realized, which provides strong support for the electromagnetic coil health status evaluation method based on image features. This not only improves the accuracy and operability of the evaluation, but also lays the foundation for subsequent preventive maintenance work.

In an embodiment of the present invention, a coil variation mechanism model is used to simulate and predict the development trend of the coil in the future. These simulation data include the changes of various performance indicators of the coil (such as geometric dimensions and electromagnetic characteristics) over time. Next, these simulation data are compared and matched with the actual collected coil image. The specific method is to extract several characteristic parameters (such as coil radius and line width) in the coil image and compare them with the simulation data one by one. Through this process, the degree of deviation between the development degree of the actual coil and the simulation result can be determined, thereby obtaining a set of electromagnetic coil development degree data. The electromagnetic coil development degree data is used to predict and evaluate the remaining service life of the coil. Specifically, according to the change trend of various performance indicators of the coil, combined with the variation mechanism model, it is predicted which performance deterioration process the coil will experience in the future, and the remaining service life of the coil is estimated accordingly. For example, if the algorithm finds that the coil winding resistance is increasing at a faster rate, it can be predicted that the electromagnetic performance of the coil will drop significantly in the future, thereby calculating the remaining service life of the coil. The estimated electromagnetic coil life data is compared with the pre-set life threshold to determine the health status of the coil. For example, if the remaining service life of the coil is predicted to be less than 1,000 hours, the health status of the coil will be assessed as unhealthy. Through this assessment process, the health status of the coil is determined, providing a basis for subsequent maintenance. At the same time, the health status assessment results are correlated with the electromagnetic coil development degree data to explore the health status characteristics corresponding to different development degrees. This provides important support for the electromagnetic coil health assessment method based on image features.

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.

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 (10)

1. An electromagnetic coil health state assessment method based on image features is characterized by comprising the following steps:

Step S1: acquiring an electromagnetic coil image; performing super-resolution reconstruction on the electromagnetic coil image to generate a reconstructed electromagnetic coil image; visual characteristic optimization is carried out on the reconstructed electromagnetic coil image, and an optimized coil image is obtained;

step S2: acquiring a standard coil image; performing covering identification on the optimized coil image data to generate surface layer attachment data; performing deformation analysis on the optimized coil image according to the standard coil image and the surface layer attachment data to generate coil deformation data;

Step S3: performing regional light reflection analysis on the optimized coil image based on the coil deformation data to generate coil deformation data;

Step S4: performing reverse deduction and tracing on the coil change data based on the coil deformation data and the coil color data to obtain a coil variation factor; performing variation field simulation on coil variation factors to generate simulated coil development data;

Step S5: and carrying out health state assessment on the electromagnetic coil image according to the simulated coil development data so as to execute an electromagnetic coil health state assessment method based on image characteristics.

2. The method for evaluating the health of an electromagnetic coil based on image features as recited in claim 1, wherein the step S1 includes the steps of:

Step S11: acquiring an electromagnetic coil image through a digital camera;

Step S12: performing super-resolution reconstruction on the electromagnetic coil image to obtain a high-resolution image;

Step S13: performing edge sharpening on the high-resolution image to generate a reconstructed electromagnetic coil image;

step S14: image characteristic enhancement is carried out on the reconstructed electromagnetic coil image, and an enhanced coil image is generated;

step S15: and carrying out histogram equalization processing on the enhanced coil image to obtain an optimized coil image.

3. The method for evaluating the health of an electromagnetic coil based on image features as recited in claim 1, wherein the step S2 includes the steps of:

Step S21: obtaining a standard coil image by checking a log;

step S22: performing feature map layer stacking comparison on the labeling coil image and the optimizing coil image data to generate feature image difference data;

step S23: performing covering identification on the optimized coil image according to the characteristic image difference data to generate surface layer attachment data;

Step S24: and performing deformation analysis on the optimized coil image based on the surface layer attachment data to generate coil deformation data.

4. The electromagnetic coil health state assessment method based on image features as set forth in claim 3, wherein step S23 includes the steps of:

Step S231: performing differential image cutting on the optimized coil image according to the characteristic image differential data to generate a differential cutting image;

step S232: performing surface texture analysis on the difference cut image to obtain coil surface texture data;

step S233: fluency analysis is carried out on the texture data of the coil surface layer, and texture fluency data is generated;

Step S234: and performing covering identification on the optimized coil image based on the texture fluency data to generate surface layer attachment data.

5. The electromagnetic coil health state assessment method based on image features as set forth in claim 3, wherein step S24 includes the steps of:

step S241: performing three-dimensional reconstruction on the optimized coil image to generate an electromagnetic coil three-dimensional model;

step S242: performing surface pollution analysis on the electromagnetic coil three-dimensional model based on surface attachment data to obtain a simulated electromagnetic coil cleaning model;

step S243: three-dimensional reconstruction is carried out on the standard coil image, and a standard electromagnetic coil model is generated;

Step S244: and carrying out structural deformation analysis on the simulation electromagnetic coil cleaning model by using the standard electromagnetic coil model to generate coil deformation data.

6. The method for evaluating the health of an electromagnetic coil based on image features as recited in claim 1, wherein the step S3 includes the steps of:

Step S31: performing regional light reflection capturing on the optimized coil image to generate initial reflected light data;

Step S32: performing microscopic light reflection analysis on the initial reflected light data to generate microscopic reflected light data;

step S33: performing deformation-color change association analysis on the optimized coil image based on the coil deformation data to obtain deformation-color change association data;

step S34: and performing precompensation color change analysis on the initial reflected light data by utilizing the microcosmic reflected light data and the deformation-color change associated data to generate coil color change data.

7. The method of image feature based electromagnetic coil health assessment of claim 6, wherein step S34 includes the steps of:

step S341: performing deformation influence analysis on the initial reflected light data based on deformation-color change associated data to generate metamorphic optical characteristics;

step S342: carrying out metamorphic light stripping on the initial reflected light data by utilizing metamorphic optical characteristics to obtain clean reflected light data;

step S343: performing light reflection analysis on the standard coil image to generate standard light reflection data;

step S344: and performing color change analysis on the microscopic reflected light data and the clean reflected light data based on the standard light reflection data to generate coil color change data.

8. The method of evaluating the health of an electromagnetic coil based on image features as recited in claim 1, wherein step S4 includes the steps of:

Step S41: performing variation result integration on the coil deformation data and the coil color data to obtain coil variation data;

Step S42: performing reverse deduction and tracing on the coil variation data to obtain coil variation factors;

Step S43: virtual coil construction is carried out on the electromagnetic coil image, and a virtual coil model is obtained;

step S44: and performing variation field simulation on the virtual coil model based on the coil variation factors to generate simulated coil development data.

9. The method of image feature based electromagnetic coil health assessment of claim 8, wherein step S4 comprises the steps of:

step S421: performing change trend analysis on the coil change data to obtain a coil change mode;

step S422: performing variation mechanism driving treatment on the coil variation mode to generate a variation mechanism model;

Step S423: performing parameter inversion on coil change data by using a variation mechanism model to generate a maximum contribution parameter;

step S424: performing sensitivity analysis on coil change data by using a variation mechanism model to generate a maximum sensitivity parameter;

Step S425: and carrying out joint analysis on the maximum contribution parameter and the maximum sensitivity parameter to obtain the coil variation factor.

10. The method for evaluating the health of an electromagnetic coil based on image features as recited in claim 1, wherein the step S5 includes the steps of:

Step S51: matching the development degree of the electromagnetic coil image according to the development data of the simulation coil to obtain development degree data of the electromagnetic coil;

step S52: carrying out service life prediction on the development degree data of the electromagnetic coil to obtain predicted service life data of the electromagnetic coil;

Step S53: and carrying out health state assessment on the estimated electromagnetic coil life data based on a preset electromagnetic coil life threshold value so as to execute an electromagnetic coil health state assessment method based on image characteristics.