Aluminum ash processing electrical system insulation aging analysis method and system based on image analysis

Abstract

The invention relates to the technical field of image processing, in particular to an aluminum ash processing electrical system insulation aging analysis method and system based on image analysis, and the method comprises the steps: carrying out the pore and crack feature extraction of a target image through a feature analysis model, and obtaining the pore space gradient and crack skeleton topology of an aluminum ash electrical system; according to the pore space gradient of the aluminum ash electrical system, a pore communication network is constructed through a pore analysis model, pore characterization data of the pore communication network is calculated, and branch density analysis is performed according to the crack skeleton topology to obtain crack characterization data; and calling daily average fluoride concentration data in the aluminum ash, measuring basic electrical parameters of the target part, and performing electrical system insulation aging evaluation through an aging evaluation model in combination with pore characterization data and crack characterization data to obtain an aging evaluation result. Therefore, the insulation aging of the aluminum ash processing electrical system can be analyzed and evaluated more accurately under the corrosion influence of fluoride in the aluminum ash.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, and in particular to an aluminum dust treatment electrical system insulation aging analysis method and system based on image analysis. BACKGROUND

[0002] The aluminum dust contains a large amount of fluoride, and the fluoride ions in the fluoride can penetrate and diffuse through the pores of the insulating material, react with active groups such as hydroxyl groups in the epoxy resin, cause changes in the pore structure of the insulating material to form penetration channels, and cause serious corrosion to the electrical system, thereby accelerating the aging of the insulating material in the electrical system. If the replacement of the insulating aging parts is not timely, it will cause performance failure of the electrical system. The traditional aluminum dust treatment electrical system insulation aging analysis mainly relies on the disassembly analysis after insulation breakdown in the maintenance process after the electrical system fails, which has a lag and cannot be preventive maintenance. Some existing technical teams evaluate the insulation aging state by measuring macroscopic electrical parameters such as insulation resistance, but these methods have low accuracy due to the failure to fully consider the synergistic effect of fluoride penetration on insulation material aging, and cannot effectively reflect the actual impact of corrosion in the aluminum dust treatment environment on the insulation performance of the parts. Therefore, how to more accurately and effectively analyze and evaluate the insulation aging state of the aluminum dust treatment electrical system has become a problem to be solved. SUMMARY

[0003] The present application provides an aluminum dust treatment electrical system insulation aging analysis method and system based on image analysis to solve the problem of inaccurate insulation aging evaluation caused by corrosion of fluoride penetration in the aluminum dust treatment electrical system.

[0004] To achieve the above-mentioned purpose, on the one hand, the present application provides an aluminum dust treatment electrical system insulation aging analysis method based on image analysis, which comprises:

[0005] Collecting images of a preset sampling point of an aluminum dust treatment electrical system and preprocessing to obtain an enhanced image as a target image, and extracting pore and crack features of the target image through a feature analysis model to obtain a pore space gradient and a crack skeleton topology of the aluminum dust electrical system.

[0006] According to the pore space gradient of the aluminum dust electrical system, a pore analysis model is used to construct a pore connected network and calculate pore representation data of the pore connected network, and branch density analysis is performed according to the crack skeleton topology to obtain crack representation data.

[0007] Retrieving daily average fluoride concentration data in the aluminum dust and measuring basic electrical parameters of the target parts, and performing electrical system insulation aging evaluation according to the daily average fluoride concentration, the basic electrical parameters of the target parts, the pore representation data and the crack representation data through an aging evaluation model to obtain an aging evaluation result, and implementing a preset maintenance measure according to the aging evaluation result.

[0008] Furthermore, the method for extracting pore and crack features from the target image using a feature analysis model to obtain the pore spatial gradient and crack skeleton topology of the aluminum gray electrical system includes:

[0009] The target image is binarized to obtain an initial binary image. The initial binary image is then processed by morphological opening to separate the pore region and the crack region, resulting in a pore binary image and a crack binary image.

[0010] The equivalent diameter of each pore in the binary image of pores is calculated by the distance transformation method, and the spatial gradient of the pores is obtained by performing gradient calculation on the pore boundary using the Sobel operator.

[0011] The crack skeleton is extracted iteratively from the binary image of the crack by skeletonizing pixels and crack branch points are detected. The crack branch length is read from the crack branch points and the angle between adjacent branches is calculated as the crack branch angle. The crack skeleton topology is constructed based on the crack branch length and crack branch angle.

[0012] Furthermore, the method of calculating the equivalent diameter of each pore in the binary pore image using the distance transform method and obtaining the pore spatial gradient by calculating the gradient of the pore boundary using the Sobel operator includes:

[0013] The equivalent diameter of a pore is obtained by obtaining the distance from each pixel inside a pore to the nearest pore boundary in the binary image of the pore and extracting the maximum distance value:

[0014] .

[0015] in It is the first The maximum distance from a pixel inside a pore to the nearest pore boundary. It is the first The equivalent diameter of each pore.

[0016] The boundary pixels of the aperture are convolved using the Sobel operator to obtain the boundary pixel value. directional gradient components and The directional gradient component, based on each boundary pixel point in directional gradient components and The directional gradient component yields the spatial gradient amplitude of the pore. for:

[0017] .

[0018] in It is the first Each pore boundary pixel exist directional gradient component, is the th pore boundary pixel point In directional gradient component, is the total number of boundary pixel points of the pore, is the th pore pore space gradient amplitude.

[0019] Further, the pore representation data includes pore connectivity and permeation path tortuosity, and the method for constructing a pore connectivity network and calculating pore representation data of the pore connectivity network according to the pore space gradient of the aluminum ash electrical system through a pore analysis model includes:

[0020] According to the pore binary image, the centroid coordinates of all pores are extracted, and the centroid coordinates are taken as nodes to construct a triangular network through triangulation.

[0021] According to the connectivity determination rule, the edges in the triangular network are screened to obtain a pore connectivity network, and the connectivity determination rule is specifically:

[0022] When the difference between the equivalent diameters of two adjacent pores is less than a preset equivalent diameter difference threshold and the distance between the centroids of the two pores is less than a preset centroid distance threshold, the edge connecting the centroids of the two pores is retained, otherwise the edge connecting the centroids of the two pores is deleted.

[0023] The pore connectivity is calculated according to the pore connectivity network and is:

[0024] .

[0025] Wherein is the actual number of connected edges in the pore connectivity network, is the maximum number of edges of a complete connected graph, is the number of pores, is the th pore pore space gradient amplitude, is the maximum value in the pore space gradient amplitude, is the pore connectivity.

[0026] The shortest path length and the straight-line distance between any two nodes in the pore connectivity network are calculated by a shortest path algorithm, and the permeation path tortuosity is calculated according to the shortest path length and the straight-line distance between all nodes and is:

[0027] .

[0028] Wherein is the total number of node pairs, is the node and the shortest path length between nodes , is the straight-line distance between nodes and nodes , is the equivalent diameter of the th pore, is the equivalent diameter of the th pore, is the tortuosity of the permeation path.

[0029] Further, the method for extracting the crack skeleton and detecting the crack branch point by skeletonizing the pixel points iteratively, reading the crack branch length according to the crack branch point, and calculating the included angle between adjacent branches as the crack branch angle, and constructing the crack skeleton topology according to the crack branch length and the crack branch angle comprises:

[0030] Alternately performing the following two iterations on the crack binary image until no changes are obtained to obtain a single-pixel-wide crack centerline as the crack skeleton:

[0031] Iteration 1: deleting the pixel point that satisfies the following conditions: the number of 1s in the 4-connected neighborhood is between 2 and 6, the number of 0s to 1s in the 8-connected neighborhood is 1, and the 4-neighborhood is not all 1s.

[0032] Iteration 2: deleting the pixel point that satisfies the conditions of iteration 1 and the number of 1s in the 8-neighborhood is between 2 and 6.

[0033] Checking the 8-neighborhood connectivity of each pixel point in the crack skeleton, and marking the pixel point as a crack branch point when the number of connected pixel points in the 8-neighborhood of the pixel point is greater than or equal to 3.

[0034] Starting from the crack branch point, tracing the skeleton pixels along each branch until the branch endpoint to obtain a branch path, counting the number of pixels on the branch path and multiplying by the pixel resolution to obtain the crack branch length.

[0035] Forming a branch vector according to the crack branch point and its branch endpoint, and calculating the crack branch angle according to adjacent branch vectors .

[0036] .

[0037] wherein is the th branch vector, is the th branch vector, is the crack branch angle formed by the branch and the branch .

[0038] The crack branch length and the crack branch angle are stored as a topological structure as a crack skeleton topology.

[0039] Further, the crack characterization data includes crack branch density and crack fractal dimension, and the method of obtaining crack characterization data according to crack skeleton topology branch density analysis includes:

[0040] The crack branch density is obtained according to the number of branch points and the crack branch length as follows:

[0041] .

[0042] Wherein is the total number of branch points, is the crack branch length corresponding to the th crack branch point, is the crack branch density.

[0043] The crack skeleton topology is covered with square grids with side length The number of grids containing crack pixels is counted This process is repeated for different grid scales, and the crack fractal dimension is obtained by linear fitting of and by the least square method.

[0044] The specific fitting formula is:

[0045] .

[0046] Wherein is the number of grid scales, is the side length of the square grid of the th grid scale, is the number of grids containing crack pixels under the th grid scale, is the mean value of , is the mean value of , is the crack fractal dimension.

[0047] Further, the method of obtaining the aging evaluation result by the aging evaluation model according to the daily average fluoride concentration, the basic electrical parameter of the target part, the pore characterization data and the crack characterization data includes:

[0048] The daily average fluoride concentration data in aluminum ash is retrieved from aluminum ash production data. This daily average fluoride concentration data, along with pore connectivity and permeation path tortuosity from pore characterization data, and crack branch density and crack fractal dimension from crack characterization data, are loaded into a component life correction model to obtain an aging acceleration index. The life correction model is a mapping relationship between the daily average fluoride concentration data, pore connectivity, permeation path tortuosity, crack branch density, and crack fractal dimension historical data, established using a neural network algorithm and the aging acceleration index.

[0049] The neural network employs a four-layer structure. The input layer has five nodes, and the LeakyReLU activation function is used in both hidden layers to enhance the interaction between porosity and crack characterization data. The output layer has one node, and the Sigmoid activation function is used to output the aging acceleration index. The aging acceleration index quantifies the degree of aging acceleration caused by the corrosive effect of fluorides in aluminum ash on insulating materials. The training set aging acceleration index is calculated by recording the ratio of the actual lifespan lead to the theoretical replacement lifespan of components from the measurement time to the final aging replacement under different daily average fluoride concentrations, porosity, permeation path tortuosity, crack branch density, and crack fractal dimension.

[0050] The basic electrical parameters of the target component are measured, including insulation resistance and dielectric loss. The theoretical remaining time for aging replacement of the target component is obtained by comparing the insulation resistance and dielectric loss with a preset component insulation aging index comparison table. The theoretical remaining time for aging replacement is then corrected based on the aging acceleration index to obtain a corrected value for the remaining time for aging replacement. for:

[0051] .

[0052] in This is the theoretical value of the remaining time before replacement due to aging. It is the aging acceleration index. This is the correction value for the remaining time of aging replacement.

[0053] The remaining time for aging replacement is used as the aging assessment result, and preset maintenance measures are implemented according to the aging assessment result.

[0054] Furthermore, the method of extracting the centroid coordinates of all pores from the binary pore image and constructing a triangular network using the centroid coordinates as nodes through triangulation includes:

[0055] Connectivity labeling is performed on the binary image of pores to obtain individual pores, and the geometric centroid coordinates are calculated based on the moments of the individual pores.

[0056] The centroid coordinates of the pores are used as nodes to call the Delaunay function of the SciPy library to generate an initial connection network. The initial connection network is then subjected to invalid triangle removal according to the side length constraint to obtain a triangular network. The side length constraint is set as follows: when any side length is greater than three times the average pore diameter, it is marked as an invalid triangle and removed.

[0057] Based on the same inventive concept, the present invention provides an insulation aging analysis system for aluminum ash processing electrical systems based on image analysis. The system includes a first analysis module, a second analysis module, and a correction module connected in sequence.

[0058] The first analysis module is used to acquire images of preset sampling points of the aluminum ash processing electrical system and preprocess them to obtain enhanced images as target images. The target images are then used to extract pore and crack features through a feature analysis model to obtain the pore spatial gradient and crack skeleton topology of the aluminum ash electrical system.

[0059] The second analysis module is used to construct a pore connectivity network based on the pore space gradient of the aluminum ash electrical system through a pore analysis model and calculate the pore characterization data of the pore connectivity network. It also performs branch density analysis based on the crack skeleton topology to obtain crack characterization data.

[0060] The correction module is used to retrieve the daily average fluoride concentration data in aluminum ash and measure the basic electrical parameters of the target components. Based on the daily average fluoride concentration, the basic electrical parameters of the target components, the porosity characterization data, and the crack characterization data, the aging assessment results are obtained by performing an aging assessment model on the electrical system insulation aging. Pre-set maintenance measures are then implemented according to the aging assessment results.

[0061] Compared with the prior art, the beneficial effects of this invention are: it realizes the extraction of pore space gradient and crack skeleton topology data features through feature analysis model, realizes the quantification of corrosion degree characterization data through pore analysis and crack analysis, and further combines the daily average fluoride concentration data to correct insulation aging through lifetime correction model, thereby achieving a more accurate insulation aging assessment of aluminum ash treatment electrical system as a whole. Attached Figure Description

[0062] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0063] Figure 1 This is a flowchart of the insulation aging analysis method for aluminum ash treatment electrical system based on image analysis according to Embodiment 1 of the present invention;

[0064] Figure 2 This is a schematic diagram of the module composition of the image analysis-based electrical system insulation aging analysis system for aluminum ash processing according to Embodiment 2 of the present invention. Detailed Implementation

[0065] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0066] Example 1: As Figure 1 As shown, this embodiment provides an image analysis-based method for insulation aging analysis of electrical systems processing aluminum ash. The method includes:

[0067] S1. Acquire images of preset sampling points in the aluminum ash processing electrical system and preprocess them to obtain an enhanced image as the target image. Extract pore and crack features from the target image using a feature analysis model to obtain the pore spatial gradient and crack skeleton topology of the aluminum ash electrical system. The preset sampling points are set on the surface of the target components, and key components of the electrical system are selected as the target components. The pore spatial gradient and crack skeleton topology reflect the microstructure of the electrical equipment's insulation layer under aluminum ash corrosion. The preprocessing includes noise reduction of the sampling point images using a filtering algorithm and enhancement of local contrast of pores and cracks through histogram equalization to obtain the target image.

[0068] S2. Based on the pore space gradient of the aluminum ash electrical system, a pore connectivity network is constructed using a pore analysis model, and the pore characterization data of the pore connectivity network is calculated. Crack characterization data is obtained by performing branch density analysis based on the crack skeleton topology. The pore characterization data affects the conductivity of the electrical insulation and can measure the degree of chemical corrosion. The crack characterization data is related to the degree of insulation aging and can measure the degree of mechanical degradation.

[0069] S3. Retrieve the daily average fluoride concentration data in aluminum ash and measure the basic electrical parameters of the target components. Based on the daily average fluoride concentration, the basic electrical parameters of the target components, the porosity characterization data, and the crack characterization data, conduct an aging assessment of the electrical system insulation through an aging assessment model to obtain the aging assessment results. Implement the preset maintenance measures according to the aging assessment results.

[0070] It should be noted that the method for extracting pore and crack features from the target image using a feature analysis model to obtain the pore spatial gradient and crack skeleton topology of the aluminum gray electrical system includes:

[0071] The target image is binarized to obtain an initial binary image. The initial binary image is then processed by morphological opening to separate the pore region and the crack region, resulting in a pore binary image and a crack binary image.

[0072] The morphological opening operation includes performing morphological erosion followed by morphological dilation. For example, a threshold of 128 is set to binarize the target image to obtain an initial binary image. Morphological erosion is then performed on the initial binary image to shrink the pore and crack regions and remove white noise. In the image, pixels around the pores shrink, and the pores are eroded away when they approach the edge. Then, morphological dilation is performed on the eroded image to expand the white areas and extend the boundaries of the pores and cracks. The opening operation effectively removes small noise and irregular pore regions from the image, while smoothing the boundaries of the pores and cracks, making the pore and crack regions clearer and more prominent. This effectively separates the pore and crack regions, forming a pore binary image and a crack binary image.

[0073] The equivalent diameter of each pore in the binary image of pores is calculated by the distance transformation method, and the spatial gradient of the pores is obtained by performing gradient calculation on the pore boundary using the Sobel operator.

[0074] The crack skeleton is extracted iteratively from the binary image of the crack by skeletonizing pixels and crack branch points are detected. The crack branch length is read from the crack branch points and the angle between adjacent branches is calculated as the crack branch angle. The crack skeleton topology is constructed based on the crack branch length and crack branch angle.

[0075] It should be noted that the method of calculating the equivalent diameter of each pore in the binary pore image using the distance transform method and obtaining the pore spatial gradient by calculating the gradient of the pore boundary using the Sobel operator includes:

[0076] The equivalent diameter of a pore is obtained by obtaining the distance from each pixel inside a pore to the nearest pore boundary in the binary image of the pore and extracting the maximum distance value:

[0077] .

[0078] in It is the first The maximum distance from a pixel inside a pore to the nearest pore boundary. It is the first The equivalent diameter of each pore.

[0079] For example: the maximum distance from a pixel in the first aperture to the aperture boundary is Then the equivalent diameter of the pore is .

[0080] The boundary pixels of the aperture are convolved using the Sobel operator to obtain the boundary pixel value. directional gradient components and The directional gradient component, based on each boundary pixel point in directional gradient components and The directional gradient component yields the spatial gradient amplitude of the pore. for:

[0081] .

[0082] in It is the first Each pore boundary pixel exist Oriented gradient components, It is the first Each pore boundary pixel exist Oriented gradient components, It is the total number of boundary pixels of the aperture. It is the first The pore space gradient amplitude of each pore.

[0083] For example: Suppose the first aperture has three boundary pixels. The boundary pixels of the aperture are convolved using the Sobel operator to obtain the boundary pixels at... directional gradient components and The directional gradient components are respectively , , Then the pore space gradient magnitude of the pore .

[0084] It should be noted that the pore characterization data includes pore connectivity and permeation path tortuosity. The method for constructing a pore connectivity network based on the pore spatial gradient of the aluminum ash electrical system through a pore analysis model and calculating the pore characterization data of the pore connectivity network includes:

[0085] The centroid coordinates of all pores are extracted from the binary image of the pores, and the centroid coordinates are used as nodes to construct a triangular network through triangulation.

[0086] The edges in the triangular network are filtered according to the connectivity determination rule to obtain the porous connected network. The specific connectivity determination rule is as follows:

[0087] When the difference in equivalent diameter between two adjacent pores is less than a preset equivalent diameter difference threshold and the distance between the centroids of the two pores is less than a preset centroid distance threshold, the edge connecting the centroids of the two pores is retained; otherwise, the edge connecting the centroids of the two pores is deleted.

[0088] The preset equivalent diameter difference threshold is half of the average equivalent diameter of all pores, and the preset centroid distance threshold is twice the average equivalent diameter of all pores. Through experiments, it has been verified that this rule, by limiting the diameter ratio and relative distance, can both eliminate unreasonable pseudo-connectivity caused by excessive diameter differences or excessive spacing, and retain effective connectivity paths that may actually allow fluoride ion penetration.

[0089] Pore ​​connectivity is calculated based on the pore connectivity network. for:

[0090] .

[0091] in It is the actual number of connected edges in a porous network. It is the maximum number of edges in a fully connected graph. It is the number of pores. It is the first The amplitude of the pore space gradient of each pore. It is the maximum value of the gradient amplitude in the pore space. It refers to pore connectivity.

[0092] The formula for calculating the maximum number of edges in a fully connected graph is: For example: Suppose there are 4 pores in a porous connected network forming 5 connecting edges. The spatial gradient magnitudes of the 4 pores are 4.37, 5.21, 3.86, and 4.62, respectively. Then the maximum value of the spatial gradient magnitude is 5.21, and the maximum number of edges in the fully connected graph is... Then the pore connectivity can be obtained. .

[0093] The closer the pore connectivity is to 0, the more the fluoride is adsorbed on the surface and the less the aging effect is. The closer the pore connectivity is to 1, the greater the penetration of the fluoride and the faster the aging.

[0094] The shortest path length and straight-line distance between any two nodes in the pore connectivity network are calculated using the shortest path algorithm. Based on the shortest path lengths and straight-line distances between all nodes, the tortuosity of the permeation path is then calculated. for:

[0095] .

[0096] in It is the total number of node pairs. It is a node and nodes The shortest path length between them. It is a node and nodes The straight-line distance between them It is the first The equivalent diameter of each pore. It is the first The equivalent diameter of each pore. It refers to the tortuosity of the penetration path.

[0097] For example: Pore 1 is connected to pores 2 and 3; pore 2 is connected to pores 3 and 4; and pore 3 is connected to pore 4. The centroid coordinates of these four pores are as follows: , , , Based on these centroid coordinates, we can calculate the straight-line distance between any two nodes using Euclidean distance. For example, the straight-line distance between pore 1 and pore 2 is... The straight-line distance between pore 2 and pore 3 is Similarly, we can obtain , , , The shortest path length between any two nodes in a porous connected network is calculated using Dijkstra's algorithm. For example, the shortest path length from pore 1 to pore 2 is... The shortest path from pore 1 to pore 3 is The shortest path from pore 1 to pore 4 is Since pore 2 is directly connected to pores 3 and 4, and pore 3 is directly connected to pore 4, the shortest path and the straight-line distance are the same. , , Since the total number of node pairs is 6, assuming the equivalent diameters of the pores are respectively , , , The tortuosity of the penetration path can then be determined. .

[0098] The smaller the tortuosity of the penetration path, the higher the corrosion efficiency of fluoride, and the faster the insulation of the parts will age due to corrosion. The larger the tortuosity of the penetration path, the slower the aging.

[0099] The shortest path algorithm can calculate the shortest path length between any two nodes in a porous connected network using Dijkstra's algorithm and calculate the straight-line distance between any two nodes in a porous connected network using Euclidean distance algorithm.

[0100] Because fluoride ions preferentially diffuse through uniform pore size regions in the corrosive environment of aluminum ash, they diffuse through the diameter difference coefficient. The introduction of this method makes the contribution of the tortuosity of the permeation path between nodes with large differences in the equivalent pore diameter to the overall tortuosity of the permeation path less than the contribution of the tortuosity of the permeation path between nodes with small differences in the equivalent pore diameter, thus making the overall tortuosity of the permeation path more accurate.

[0101] It should be noted that the method for iteratively extracting the crack skeleton from the binary image of the crack by skeletonizing pixels and detecting crack branch points, reading the crack branch length based on the crack branch points and calculating the angle between adjacent branches as the crack branch angle, and constructing the crack skeleton topology based on the crack branch length and crack branch angle includes:

[0102] Define the foreground pixel as 1 and the background pixel as 0, and alternately perform the following two iterations on the binary image of the crack until no change occurs, obtaining a single-pixel-wide crack centerline as the crack skeleton:

[0103] Iteration 1: Delete pixels whose number of 1s in a 4-connected neighborhood is between 2 and 6, whose number of transitions from 0 to 1 in an 8-connected neighborhood is 1, and whose 4-neighborhood is not all 1s.

[0104] Iteration 2: Delete pixels that satisfy the conditions of Iteration 1 and whose number of 1s in their 8-neighborhood is between 2 and 6.

[0105] For each pixel in the crack skeleton, check its 8-neighbor connectivity. If the number of connected pixels in the 8-neighborhood of a pixel is greater than or equal to 3, mark the pixel as a crack branch point.

[0106] Starting from the crack branch point, trace the skeleton pixels along each branch until the branch endpoint to obtain the branch path. Count the number of pixels on the branch path and multiply by the pixel resolution to obtain the crack branch length.

[0107] For example: Suppose that there are 3 pixels in the crack skeleton image marked as crack branch points, with the number of pixels on the branch path being 10 pixels, 20 pixels, and 15 pixels respectively, and the resolution being 1 micrometer / pixel. Then the crack branch length of branch point 1 is 10 micrometers, the crack branch length of branch point 2 is 20 micrometers, and the crack branch length of branch point 3 is 15 micrometers.

[0108] The crack branching angle is calculated by forming a branch vector based on the crack branching point and its branch endpoints, and then calculating the branching angle based on the adjacent branch vectors. for:

[0109] .

[0110] in It is the first Branch vectors It is the first Branch vectors It is a branch and branches The angle of the branching crack.

[0111] Store the crack branch length and crack branch angle as a topological structure as the crack skeleton topology.

[0112] For example, taking the first two branch points mentioned above as examples, assume the direction vector of branch 1 is... The direction vector of branch 2 Then the crack branching angles of branch 1 and branch 2 are: .

[0113] The crack skeleton topology is formed based on all crack branch lengths and crack branch angles, and the data can be stored and represented as follows: ,in The point set contains all crack branch points. It is an edge set containing all crack branch angles and crack branch lengths.

[0114] It should be noted that the crack characterization data includes crack branch density and crack fractal dimension, and the method for obtaining crack characterization data by performing branch density analysis based on crack skeleton topology includes:

[0115] The crack branch density is obtained based on the number of branch points and the crack branch length:

[0116] .

[0117] in It is the total number of branch points. It is the first Crack branch length corresponding to each crack branch point It is the crack branch density.

[0118] For example, taking the above three branching points as an example, when there are three crack branching points with lengths of 10, 20, and 15 respectively, the crack branching density is: .

[0119] Crack branch density reflects the number of potential channels for fluoride ion penetration. The higher the crack branch density, the greater the impact of fluoride chemical corrosion on the crack. The lower the crack branch density, the greater the impact of mechanical stress and other factors on the crack.

[0120] The crack skeleton topology is represented by a side length of... A square grid at a grid scale is used to count the number of grid cells containing crack pixels. Repeat this process at different grid scales, and apply the least squares method to... and The crack fractal dimension was obtained by linear fitting.

[0121] The specific fitting formula is as follows:

[0122] .

[0123] in It is the number of grid scales. It is the first The side length of a square grid at each grid size. It is the first The number of grids containing crack pixels at each grid scale. yes The mean, yes The mean, It is the crack fractal dimension.

[0124] The phrase "according to different grid scales" refers to... Repeat the process with different values, and the rules for selecting the values ​​are as follows: The condition must satisfy a value greater than or equal to 2 pixels and less than or equal to the length of the shorter side of the image. For example, the number of crack pixels contained in a square grid covered with grid sizes of 2, 4, and 8 are 100, 50, and 20 respectively. , , , , , The fractal dimension of the crack can be obtained. .

[0125] The crack fractal dimension reveals the propagation pattern of cracks and is sensitive to changes in the morphology of microcracks. It can reflect early aging that cannot be detected by traditional electrical parameters. The larger the crack fractal dimension, the more severe the crack propagation, the more severe the corrosion, and the higher the degree of aging.

[0126] It should be noted that the method for obtaining aging assessment results by retrieving daily average fluoride concentration data in aluminum ash, measuring the basic electrical parameters of the target components, and then using an aging assessment model to conduct an aging assessment of the electrical system insulation based on the daily average fluoride concentration, the basic electrical parameters of the target components, porosity characterization data, and crack characterization data includes:

[0127] The daily average fluoride concentration data in aluminum ash is retrieved from aluminum ash production data. This daily average fluoride concentration data, along with pore connectivity and permeation path tortuosity from pore characterization data, and crack branch density and crack fractal dimension from crack characterization data, are loaded into a component life correction model to obtain an aging acceleration index. The life correction model is a mapping relationship between the daily average fluoride concentration data, pore connectivity, permeation path tortuosity, crack branch density, and crack fractal dimension historical data, established using a neural network algorithm and the aging acceleration index.

[0128] The neural network employs a four-layer structure. The input layer has five nodes, and the LeakyReLU activation function is used in both hidden layers to enhance the interaction between porosity and crack characterization data. The output layer has one node, and the Sigmoid activation function is used to output the aging acceleration index. The aging acceleration index quantifies the degree of aging acceleration caused by the corrosive effect of fluorides in aluminum ash on insulating materials. The training set aging acceleration index is calculated by recording the ratio of the actual lifespan lead to the theoretical replacement lifespan of components from the measurement time to the final aging replacement under different daily average fluoride concentrations, porosity, permeation path tortuosity, crack branch density, and crack fractal dimension.

[0129] The basic electrical parameters of the target component are measured, including insulation resistance and dielectric loss. The theoretical remaining time for aging replacement of the target component is obtained by comparing the insulation resistance and dielectric loss with a preset component insulation aging index comparison table. The theoretical remaining time for aging replacement is then corrected based on the aging acceleration index to obtain a corrected value for the remaining time for aging replacement. for:

[0130] .

[0131] in This is the theoretical value of the remaining time before replacement due to aging. It is the aging acceleration index. This is the correction value for the remaining time of aging replacement.

[0132] For example, when the daily average fluoride concentration in aluminum ash is 0.5 mol / m³, the pore connectivity is 0.72, the tortuosity of the penetration path is 0.92, the crack branch density is 0.07, and the crack fractal dimension is 1.16, and the aging acceleration index obtained by inputting it into the lifetime correction model is 0.15, this means that the corrosion effect of fluoride has accelerated the aging process of the insulation material by 15%. If the theoretical value of the remaining aging replacement time for the target component is obtained by comparing the current insulation resistance and dielectric loss with the preset component insulation aging index comparison table, then the correction value for the remaining aging replacement time is 5000 hours. Hour.

[0133] The remaining time for aging replacement is used as the aging assessment result, and preset maintenance measures are implemented according to the aging assessment result.

[0134] The method commonly used in the industry is to compare the insulation resistance and dielectric loss with a preset reference table of component insulation aging indicators to obtain the theoretical value of the remaining time for replacement of the target component. The preset reference table of component insulation aging indicators is a database of theoretical values ​​of the remaining time for replacement of the component determined by accelerated aging tests under standard operating conditions. It is mainly obtained by recording the cumulative time when each parameter reaches the critical threshold as the theoretical total life, and constructing a correspondence between the electrical parameter combination of different aging stages and the remaining life.

[0135] The more severe the fluoride permeation effect, the greater the aging acceleration index, and the smaller the remaining time correction value for aging replacement, requiring earlier maintenance and replacement of parts.

[0136] It should be noted that the method of extracting the centroid coordinates of all pores from the binary pore image and constructing a triangular network using the centroid coordinates as nodes through triangulation includes:

[0137] Connectivity labeling is performed on the binary image of pores to obtain individual pores, and the geometric centroid coordinates are calculated based on the moments of the individual pores.

[0138] The centroid coordinates of the pores are used as nodes to call the Delaunay function of the SciPy library to generate an initial connection network. The initial connection network is then subjected to invalid triangle removal according to the side length constraint to obtain a triangular network. The side length constraint is set as follows: when any side length is greater than three times the average pore diameter, it is marked as an invalid triangle and removed.

[0139] Example 2: Based on the same inventive concept, such as Figure 2 As shown, this embodiment provides an image analysis-based insulation aging analysis system for aluminum ash processing electrical systems. The system includes a first analysis module, a second analysis module, and a correction module connected in sequence.

[0140] The first analysis module is used to acquire images of preset sampling points of the aluminum ash processing electrical system and preprocess them to obtain enhanced images as target images. The target images are then used to extract pore and crack features through a feature analysis model to obtain the pore spatial gradient and crack skeleton topology of the aluminum ash electrical system.

[0141] The second analysis module is used to construct a pore connectivity network based on the pore space gradient of the aluminum ash electrical system through a pore analysis model and calculate the pore characterization data of the pore connectivity network. It also performs branch density analysis based on the crack skeleton topology to obtain crack characterization data.

[0142] The correction module is used to retrieve the daily average fluoride concentration data in aluminum ash and measure the basic electrical parameters of the target components. Based on the daily average fluoride concentration, the basic electrical parameters of the target components, the porosity characterization data, and the crack characterization data, the aging assessment results are obtained by performing an aging assessment model on the electrical system insulation aging. Pre-set maintenance measures are then implemented according to the aging assessment results.

[0143] It should be noted that the specific methods by which each module performs operations in the system described in the above embodiments have been described in detail in Embodiment 1 of the method, and will not be elaborated here.

[0144] Finally, it should be noted that although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for analyzing the aging of the insulation of an electrical system of an aluminum dust handling system based on image analysis, characterized in that, The method comprises: Collecting an image of a preset sampling point of an aluminum ash treatment electrical system and performing preprocessing to obtain an enhanced image as a target image, performing feature analysis on the target image to extract pore and crack features to obtain a pore space gradient and a crack skeleton topology of the aluminum ash treatment electrical system; According to the pore space gradient of the aluminum ash treatment electrical system, a pore connectivity network is constructed by a pore analysis model, and pore representation data of the pore connectivity network is calculated, and branch density analysis is performed according to the crack skeleton topology to obtain crack representation data; The daily average fluoride concentration data in the aluminum ash are called, and the basic electrical parameters of the target component are measured, the daily average fluoride concentration, the basic electrical parameters of the target component, the pore representation data and the crack representation data are input into an aging evaluation model to perform electrical system insulation aging evaluation to obtain an aging evaluation result, and a preset maintenance measure is implemented according to the aging evaluation result.

2. The image analysis-based electrical system insulation aging analysis method for aluminum dust treatment according to claim 1, characterized by, The method comprises: The target image is binarized to obtain an initial binary image, and the initial binary image is processed by a morphological opening operation to separate pore regions and crack regions to obtain a pore binary image and a crack binary image; The equivalent diameter of each pore in the pore binary image is calculated by a distance transformation method, and the pore space gradient is calculated by performing gradient calculation on the pore boundary by a Sobel operator; The crack skeleton is extracted by iteratively extracting the pixel points of the crack binary image, and the crack branch points are detected, the crack branch length is read according to the crack branch points, the included angle between adjacent branches is calculated as the crack branch angle, and the crack skeleton topology is constructed according to the crack branch length and the crack branch angle.

3. The image analysis-based electrical system insulation aging analysis method for aluminum dust treatment according to claim 2, characterized by, The method comprises: The equivalent diameter of each pore in the pore binary image is calculated by a distance transformation method, and the pore space gradient is calculated by performing gradient calculation on the pore boundary by a Sobel operator; ； wherein is the maximum distance value from the first pore interior pixel to the nearest pore boundary, is the equivalent diameter of the first pore. The boundary pixel points of the pores are subjected to convolution operation by a Sobel operator to obtain the directional gradient component of each boundary pixel point in the x direction and the y direction. The directional gradient component of each boundary pixel point in the x direction and the y direction is obtained. The directional gradient component of each boundary pixel point in the x direction and the y direction is obtained. The directional gradient component of each boundary pixel point in the x direction and the y direction is obtained. The directional gradient component of each boundary pixel point in the x direction and the y direction is obtained. The pore spatial gradient amplitude is: ； wherein is the number of the pore boundary pixel point is the number of the pore boundary pixel point is the gradient component in the is the gradient component in the is the number of the pore boundary pixel point is the number of the pore boundary pixel point is the gradient component in the is the gradient component in the is the total number of the boundary pixel points of the pore is the number of the pore is the pore spatial gradient magnitude of the pore 4. The image analysis-based electrical system insulation aging analysis method for aluminum dust treatment according to claim 3, characterized by, The maximum distance value of each pixel point inside the pore in the pore binary image to the nearest pore boundary is obtained, and the equivalent diameter of the pore is extracted as: The pore representation data comprises a pore connectivity degree and a permeation path tortuosity, and the method comprises: The centroid coordinates of all pores are extracted from the pore binary image, and the centroid coordinates are taken as nodes to construct a triangular network by triangulation; The edges in the triangular network are screened to obtain a pore connectivity network according to a connectivity determination rule, and the connectivity determination rule is as follows: The porosity connectivity is calculated according to the pore connectivity network is: ； wherein is the actual number of connected edges in the pore connectivity network, is the maximum number of edges in a fully connected graph, is the number of pores, is the pore space gradient amplitude of the th pore, is the maximum value in the pore space gradient amplitudes, is the pore connectivity; The shortest path algorithm is used to calculate the shortest path length and the straight line distance between any two nodes in the pore connectivity network, and the tortuosity of the penetration path is calculated according to the shortest path length and the straight line distance between all nodes is: ； wherein is the total number of node pairs, is the shortest path length between node and node , is the straight-line distance between node and node , is the equivalent diameter of the th pore, is the equivalent diameter of the th pore, is the tortuosity of the permeation path.

5. The image analysis-based electrical system insulation aging analysis method for aluminum dust treatment according to claim 2, characterized by, When the difference between the equivalent diameters of two adjacent pores is less than a preset equivalent diameter difference threshold and the distance between the two pore centroids is less than a preset centroid distance threshold, the edge connecting the two pore centroids is retained, otherwise the edge connecting the two pore centroids is deleted; The crack skeleton is extracted by iteratively extracting the pixel points of the crack binary image, and the crack branch points are detected, the crack branch length is read according to the crack branch points, the included angle between adjacent branches is calculated as the crack branch angle, and the crack skeleton topology is constructed according to the crack branch length and the crack branch angle. Alternately perform the following two iterations on the crack binary image until no change is obtained to get the single-pixel wide crack centerline as the crack skeleton: Iteration 1: delete the pixel points whose number of 1 in 4-connected neighborhood is between 2 and 6, and whose number of 0 to 1 in 8-connected neighborhood is 1, and whose 4-neighborhood is not all 1; Iteration 2: delete the pixel points which satisfy the condition of iteration 1 and whose number of 1 in 8-connected neighborhood is between 2 and 6; Check the 8-neighborhood connectivity of each pixel point in the crack skeleton, and mark the pixel point as a crack branch point when the number of connected pixel points in the 8-neighborhood of the pixel point is greater than or equal to 3; From the crack branch point, trace the skeleton pixels along each branch until the branch endpoint to obtain the branch path, count the number of pixels on the branch path and multiply by the pixel resolution to obtain the crack branch length; A branch vector is formed according to the crack branching point and its branching endpoint, and a crack branching angle is calculated according to adjacent branch vectors is: ； wherein is the first branch vector, is the first branch vector, is the branch and the branch crack branch angle formed by Store the crack branch length and crack branch angle as the topological structure as the crack skeleton topology.

6. The image analysis-based electrical system insulation aging analysis method for aluminum dust treatment according to claim 5, characterized by, The crack representation data includes crack branch density and crack fractal dimension, and the method for branch density analysis according to the crack skeleton topology to obtain the crack representation data includes: According to the number of branch points and the crack branch length, the crack branch density is obtained as: ； wherein is the total number of branch points, is the length of the crack branch corresponding to the th crack branch point, is the crack branch density; The crack skeleton topology is covered with square grids of mesh size The number of grids containing crack pixels is counted This process is repeated for different mesh sizes, and the crack fractal dimension is obtained by linear fitting of and using the least squares method.

7. The image analysis-based electrical system insulation aging analysis method for aluminum dust treatment according to claim 1, characterized by, The method for obtaining the aging assessment result by the aging assessment model according to the daily average fluoride concentration in the aluminum ash, the basic electrical parameters of the target component, the pore representation data, and the crack representation data includes: The method for obtaining the aging acceleration index by loading the daily average fluoride concentration data, the pore connectivity in the pore representation data, the permeation path tortuosity, the crack branch density in the crack representation data, and the crack fractal dimension into the component life correction model includes: the life correction model is a mapping relationship between the aging acceleration index and the daily average fluoride concentration data, the pore connectivity, the permeation path tortuosity, the crack branch density, and the crack fractal dimension historical data through a neural network algorithm; The basic electrical parameters of the target component are measured, including insulation resistance and dielectric loss. The theoretical value of the remaining life of the target component is obtained by comparing the insulation resistance and the dielectric loss with a preset comparison table of insulation aging indicators of the component. The remaining life of the target component is corrected according to the aging acceleration index to obtain a corrected value of the remaining life of the target component is: ； wherein is a theoretical value of the remaining duration of the replacement of the aging, is an aging acceleration index, is a correction value of the remaining duration of the replacement of the aging; The aging replacement remaining duration correction value is taken as the aging assessment result, and a preset maintenance measure is implemented according to the aging assessment result.

8. The image analysis-based electrical system insulation aging analysis method for aluminum dust treatment according to claim 4, characterized by, The method for extracting the centroid coordinates of all pores from the pore binary image and constructing a triangular network by triangulation using the centroid coordinates as nodes includes: Perform connected component labeling on the pore binary image to obtain independent pores, and calculate the geometric centroid coordinates of the independent pores according to the independent pores; Call the Delaunay function of the SciPy library to generate an initial connection network using the centroid coordinates of the pores as nodes, and remove invalid triangles from the initial connection network according to the edge length constraint to obtain a triangular network, wherein the edge length constraint is set as: when any edge length is greater than three times the average pore diameter, it is marked as an invalid triangle and removed.

9. An image analysis based electrical system insulation ageing analysis system for electrical systems with aluminium ash handling, for performing the method according to any one of claims 1-8, characterised in that, The system includes a first analysis module, a second analysis module, and a correction module connected in sequence; The first analysis module is used to collect images of a predetermined sampling point of an aluminum ash treatment electrical system and perform preprocessing to obtain an enhanced image as a target image, and perform pore and crack feature extraction on the target image through a feature analysis model to obtain a pore spatial gradient and a crack skeleton topology of the aluminum ash electrical system; The system includes a first analysis module, a second analysis module, and a correction module connected in sequence; The second analysis module is configured to construct a pore connectivity network and calculate pore characterization data of the pore connectivity network according to a pore analysis model based on a pore space gradient of the aluminum ash electrical system, and to obtain crack characterization data by branch density analysis based on a crack skeleton topology; The correction module is configured to call daily average fluoride concentration data in the aluminum ash and measure basic electrical parameters of the target component, perform electrical system insulation aging assessment by an aging assessment model based on the daily average fluoride concentration, the basic electrical parameters of the target component, the pore characterization data and the crack characterization data to obtain an aging assessment result, and implement a preset maintenance measure according to the aging assessment result.

Images