Modeling method and system of steel-cored aluminum strand fire cause identification model and fire cause identification method

Abstract

The invention relates to a modeling method and system for a steel-cored aluminum strand fire cause identification model and a fire cause identification method, and the modeling method for the steel-cored aluminum strand fire cause identification model comprises the following steps: obtaining SEM images of a plurality of steel-cored aluminum strand samples, and carrying out the cutting and graying processing of the SEM images of the steel-cored aluminum strand samples, a corresponding SEM gray level image is obtained; performing binarization processing on each SEM gray scale image to obtain a corresponding binarization image; extracting trace features of partial areas of cracks and holes in the binarized image, and performing screening to obtain correlation features; constructing a training sample set comprising the binary images of the steel-cored aluminum strand samples, the corresponding correlation characteristics and the fire cause labels of the corresponding steel-cored aluminum strand samples, and establishing a steel-cored aluminum strand fire cause identification model; training the steel-cored aluminum strand fire cause identification model by using the training sample set; and a trained steel-cored aluminum strand fire cause identification model is obtained.

Description

Technical Field

[0001] The invention relates to a modeling method and system of a steel core aluminum stranded wire fire cause identification model and a fire cause identification method, and belongs to the technical field of fire cause identification. Background Art

[0002] As the infrastructure of modern society, the safety of the power system is of vital importance. Fire is one of the main factors affecting the safe operation of the power system, especially the analysis of the damage to the steel core aluminum stranded wire after the fire, which is directly related to the accurate judgment of the cause of the fault and the subsequent preventive measures. Steel core aluminum stranded wire is widely used in overhead transmission lines due to its high strength and good conductivity. However, after a fire occurs, the surface of the steel core aluminum stranded wire will leave unique traces, which are the key clues to determine the cause of the fire.

[0003] Traditional post-fire trace analysis methods mainly rely on macroscopic observation and empirical judgment, which are subjective and uncertain. With the development of science and technology, microscopic analysis techniques such as scanning electron microscopy (SEM) are gradually applied to the analysis of surface traces of materials. SEM can provide high-resolution surface morphology images, which helps to identify and analyze the microstructure and trace characteristics of the material surface; SEM is often used in the field of materials to observe the microscopic morphology and structure of different materials such as metals and ceramics. Samples of different materials have different conductivity and cleanliness requirements, which leads to differences in SEM sample preparation.

[0004] Although SEM technology has made significant progress in the field of material micro-analysis, there are still some technical difficulties in the diagnosis of power equipment faults, especially in the identification of traces of steel-core aluminum stranded wire after fire. For example, the SEM image analysis of steel-core aluminum stranded wire in the existing technology mostly focuses on the identification of simple features such as surface cracks and holes, lacks a systematic and automated analysis method, and is difficult to accurately extract key information on the cause of the fire from the complex micro-morphology.

[0005] In addition, the identification of the cause of fire often requires comprehensive consideration of multiple trace features. The existing technology lacks effective feature extraction and comprehensive analysis methods, making it difficult to quickly and accurately determine the cause of the fire. In addition, the cause of the fire cannot be effectively identified based on the macroscopic traces of the conductor. The SEM image investigation part of GBT 16840.6-2012 "Technical Identification Methods for Physical Evidence of Electrical Fire Trace" is mostly about copper conductors, and does not consider the steel-core aluminum stranded wire in overhead lines. As a result, the existing technology has certain difficulties in identifying steel-core aluminum stranded wire fires, and it is difficult to distinguish between the scenes of "flame causing line fault" and "line fault discharge causing fire"; more sophisticated microscopic trace feature analysis and more advanced data processing technology are needed.

[0006] The Chinese patent application with application publication number CN116342950A discloses "A method for metallographic image recognition of wire melt marks based on semantic segmentation". It uses a neural network to perform semantic segmentation of images and combines a classifier to classify different types of melt marks, but lacks effective feature extraction, resulting in inaccurate judgments. Summary of the invention

[0007] In order to solve the above problems existing in the prior art, the present invention proposes a modeling method and system for identifying the cause of fire of steel core aluminum stranded wire and a method for identifying the cause of fire.

[0008] The technical solution of the present invention is as follows:

[0009] In one aspect, the present invention provides a method for modeling a fire cause identification model for a steel core aluminum stranded wire, comprising the following steps:

[0010] Acquire SEM images of several steel core aluminum stranded wire samples, and perform cropping and grayscale processing on the SEM images of each steel core aluminum stranded wire sample to obtain a corresponding SEM grayscale image;

[0011] Binarization is performed on each SEM grayscale image to obtain a corresponding binary image; trace features of cracks and holes in the binary image are extracted, and correlation features are screened for the extracted trace features to obtain correlation features;

[0012] Construct a training sample set including binary images of each steel core aluminum stranded wire sample, corresponding correlation features and fire cause labels of the corresponding steel core aluminum stranded wire samples, and establish a steel core aluminum stranded wire fire cause identification model based on machine learning algorithm;

[0013] The fire cause identification model of steel core aluminum stranded wire is trained by using the training sample set; and a trained fire cause identification model of steel core aluminum stranded wire is obtained.

[0014] Preferably, the binarization process adopts an iterative segmentation threshold method, and the specific steps include:

[0015] The initial estimation threshold is set based on the maximum grayscale value and the minimum grayscale value in the SEM grayscale image. The specific formula is as follows:

[0016]

[0017] Among them, N0 is the initial estimation threshold; Z max is the maximum gray value; Z min is the minimum gray value;

[0018] Set the kth iteration threshold to N k , and N kThe initial value of is N0; the image is divided into foreground and background based on the iterative threshold, and the average grayscale value of the foreground and background is calculated. The specific formulas are as follows:

[0019]

[0020] Among them, Z F is the average gray value of the foreground; Z B is the average gray value of the background; Z(i, j) is the gray value of the pixel at coordinate (i, j); P(i, j) is the weight coefficient of coordinate (i, j), taking the probability of the gray value of coordinate (i, j);

[0021] Calculate the new threshold, the specific formula is as follows:

[0022]

[0023] Set the iteration end condition to |N k+1 -N k |<ε, ε is the preset parameter; if the condition is not met, the new threshold N k+1 Substitute into the formula for calculating the average gray value of the foreground and background; if the conditions are met, output N k is the optimal segmentation threshold;

[0024] The SEM grayscale image is binarized based on the output optimal segmentation threshold.

[0025] Preferably, the step of extracting trace features of crack and hole regions in the binary image is specifically as follows:

[0026] Identify cracks and holes, and remove small black areas with grayscale 0 through morphological corrosion operations;

[0027] The grayscale of the image is inverted, 0 grayscale is converted to 255 grayscale, 255 grayscale is converted to 0 grayscale, and the white area of ​​grayscale 255 represents the concave part in the SEM image;

[0028] Use the regionprops function to identify the white connected area, read the parameter BoundingBox, which is the diagonal coordinates of the external matrix of the connected area and the length and width of the rectangle, and obtain the minimum external matrix corresponding to the crack and hole parts;

[0029] Construct the following conditional probability density formula:

[0030] P(i,j,d,θ)={[(x,y),(x+a,x+b)|f(x,y)]=i; f(x+a,y+b)=j}

[0031] Where P(i,j,d,θ) represents the probability of a group of image pixels with a relative distance of d having grayscale values ​​i and j respectively in the direction of θ; (x, y) is the pixel with grayscale value i, (x+a, y+b) is the pixel with grayscale value j; d is the preset distance; θ is the preset direction;

[0032] The maximum gray level of the image is G, and the gray values ​​i and j are arranged and combined to obtain the numerical combination of all P(i, j, d, θ) to form a co-occurrence gray matrix;

[0033] Trace features are extracted based on the minimum circumscribed matrix and the co-occurrence grayscale matrix.

[0034] Preferably, the trace features include: contrast, correlation, energy, homogeneity, entropy in the initial state, the number of holes, the mean of the hole area, the variance of the hole area, the mean of the minimum circumscribed matrix, the mean of the minimum circumscribed matrix perimeter, the mean of the hole / circumscribed matrix ratio, the variance of the minimum circumscribed matrix, the variance of the minimum circumscribed matrix perimeter, the variance of the hole / circumscribed matrix ratio, the entropy of the processed binary image, the mean of the center point X, the mean of the center point Y, the variance of the center point X, and the variance of the center point Y.

[0035] Preferably, the correlation feature screening is specifically:

[0036] The correlation between trace features and labels is analyzed by Pearson correlation coefficient. xy , the specific formula is as follows:

[0037]

[0038] Where n is the sample size; x i is any feature of the trace features of the i-th observation point; y i is the corresponding label data; is the sample mean corresponding to any item in the trace feature; is the sample mean of the corresponding label data;

[0039] Calculate the correlation coefficient of each trace feature, set the correlation coefficient threshold, and screen out the trace features with correlation coefficients greater than the correlation coefficient threshold as correlation features.

[0040] On the other hand, the present invention also provides a method for identifying traces of steel-core aluminum stranded wire after fire by SEM images, comprising the following steps:

[0041] Collect a single strand of aluminum-steel core stranded wire from the burning side of the fire, pre-treat the aluminum-steel core stranded wire, and obtain an aluminum-steel core stranded wire sample;

[0042] Obtaining a SEM image of the steel core aluminum stranded wire sample by scanning electron microscopy;

[0043] The SEM image is identified using a trained steel core aluminum stranded wire fire cause identification model obtained by the modeling method of the steel core aluminum stranded wire fire cause identification model described in any embodiment of the present invention to obtain the cause of the fire.

[0044] Preferably, the method for pre-treating the steel core aluminum stranded wire is specifically as follows:

[0045] The steel core aluminum stranded wire is adjusted to the set size and impurities on the steel core aluminum stranded wire are removed by ultrasonic cleaning.

[0046] In another aspect, the present invention further provides a modeling system for a fire cause identification model of a steel core aluminum stranded wire, comprising:

[0047] An image processing module is used to obtain SEM images of several steel core aluminum stranded wire samples, and to perform cropping and grayscale processing on the SEM images of each steel core aluminum stranded wire sample to obtain a corresponding SEM grayscale image;

[0048] The feature extraction module performs binarization processing on each SEM grayscale image to obtain a corresponding binary image; extracts trace features of cracks and holes in the binary image, and performs correlation feature screening on the extracted trace features to obtain correlation features;

[0049] The training sample and model building module constructs a training sample set including the binary images of each steel core aluminum stranded wire sample, the corresponding correlation features, and the fire cause labels of the corresponding steel core aluminum stranded wire samples, and establishes a steel core aluminum stranded wire fire cause identification model based on a machine learning algorithm;

[0050] The model training module uses the training sample set to train the steel core aluminum stranded wire fire cause identification model; and obtains the trained steel core aluminum stranded wire fire cause identification model.

[0051] Preferably, the feature extraction module includes a binarization processing unit, which uses an iterative segmentation threshold method to perform binarization processing on each SEM grayscale image, and the specific steps include:

[0052] The initial estimation threshold is set based on the maximum grayscale value and the minimum grayscale value in the SEM grayscale image. The specific formula is as follows:

[0053]

[0054] Among them, N0 is the initial estimation threshold; Z max is the maximum gray value; Z min is the minimum gray value;

[0055] Set the kth iteration threshold to N k , and Nk The initial value of is N0; the image is divided into foreground and background based on the iterative threshold, and the average grayscale value of the foreground and background is calculated. The specific formulas are as follows:

[0056]

[0057] Among them, Z F is the average gray value of the foreground; Z B is the average gray value of the background; Z(i, j) is the gray value of the pixel at coordinate (i, j); P(i, j) is the weight coefficient of coordinate (i, j), taking the probability of the gray value of coordinate (i, j);

[0058] Calculate the new threshold, the specific formula is as follows:

[0059]

[0060] Set the iteration end condition to |N k+1 -N k |<ε, ε is the preset parameter; if the condition is not met, the new threshold N k+1 Substitute into the formula for calculating the average gray value of the foreground and background; if the conditions are met, output N k is the optimal segmentation threshold;

[0061] The SEM grayscale image is binarized based on the output optimal segmentation threshold.

[0062] Preferably, the feature extraction module includes a trace feature extraction unit, which is used to extract trace features of cracks and holes in the binary image. The specific steps are:

[0063] Identify cracks and holes, and remove small black areas with grayscale 0 through morphological corrosion operations;

[0064] The grayscale of the image is inverted, 0 grayscale is converted to 255 grayscale, 255 grayscale is converted to 0 grayscale, and the white area of ​​grayscale 255 represents the concave part in the SEM image;

[0065] Use the regionprops function to identify the white connected area, read the parameter BoundingBox, which is the diagonal coordinates of the external matrix of the connected area and the length and width of the rectangle, and obtain the minimum external matrix corresponding to the crack and hole parts;

[0066] Construct the following conditional probability density formula:

[0067] P(i,j,d,θ)={[(x,y),(x+a,x+b)|f(x,y)]=i; f(x+a,y+b)=j}

[0068] Where P(i,j,d,θ) represents the probability of a group of image pixels with a relative distance of d having grayscale values ​​i and j respectively in the direction of θ; (x, y) is the pixel with grayscale value i, (x+a, y+b) is the pixel with grayscale value j; d is the preset distance; θ is the preset direction;

[0069] The maximum gray level of the image is G, and the gray values ​​i and j are arranged and combined to obtain the numerical combination of all P(i, j, d, θ) to form a co-occurrence gray matrix;

[0070] Trace features are extracted based on the minimum circumscribed matrix and the co-occurrence grayscale matrix.

[0071] Preferably, the trace features include: contrast, correlation, energy, homogeneity, entropy in the initial state, the number of holes, the mean of the hole area, the variance of the hole area, the mean of the minimum circumscribed matrix, the mean of the minimum circumscribed matrix perimeter, the mean of the hole / circumscribed matrix ratio, the variance of the minimum circumscribed matrix, the variance of the minimum circumscribed matrix perimeter, the variance of the hole / circumscribed matrix ratio, the entropy of the processed binary image, the mean of the center point X, the mean of the center point Y, the variance of the center point X, and the variance of the center point Y.

[0072] Preferably, the feature extraction module includes a feature screening unit for performing correlation feature screening on the extracted trace features, and the specific steps are:

[0073] The correlation between trace features and labels is analyzed by Pearson correlation coefficient. xy , the specific formula is as follows:

[0074]

[0075] Where n is the sample size; x i is any feature of the trace features of the i-th observation point; y i is the corresponding label data; is the sample mean corresponding to any item in the trace feature; is the sample mean of the corresponding label data;

[0076] Calculate the correlation coefficient of each trace feature, set the correlation coefficient threshold, and screen out the trace features with correlation coefficients greater than the correlation coefficient threshold as correlation features.

[0077] On the other hand, the present invention further provides an electronic device having a computer program stored thereon, wherein when the computer program is executed by a processor, the method for modeling a steel core aluminum stranded wire fire cause identification model as described in any embodiment of the present invention is implemented.

[0078] On the other hand, the present invention also provides a computer-readable storage medium for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement a modeling method for a steel core aluminum stranded wire fire cause identification model as described in any embodiment of the present invention.

[0079] The present invention has the following beneficial effects:

[0080] 1. The present invention extracts fire trace features by graying and binarizing the SEM images of steel core aluminum stranded wire samples with different fire causes, and improves feature correlation through correlation screening; then the model is trained through images, correlation features and labels of fire causes, and a fire cause classification model with good recognition accuracy can be obtained.

[0081] 2. The present invention adopts an iterative segmentation threshold method to binarize the image, gradually assigning each pixel in the image to the foreground or background category to obtain a stable binary image, which can better adapt to the local characteristics of SEM grayscale images with complex backgrounds or uneven lighting.

[0082] 3. The present invention performs corrosion processing on the image, thereby improving the ability to recognize the trace features of the wire, and uses the minimum circumscribed matrix and the co-occurrence grayscale matrix to extract the trace features, which can extract feature information with rich types and multiple levels.

[0083] 4. The present invention extracts multiple features from the processed spectrum, and performs feature dimension reduction through correlation coefficients to screen out high-correlation features, which helps to improve the accuracy and efficiency of the identification model.

[0084] 5. The wire SEM sample preparation of the present invention is processed by ultrasonic cleaning, but does not use surface polishing technology, so that the sample with wire traces can minimize the residue of fire ash, and at the same time, the traces on the wire surface that need to be observed will not be damaged due to polishing. The wire is bonded to the sample holder with conductive glue for observation, and the sample is not wrapped with conductive glue. The use of conductive glue is reduced according to the good conductivity of the wire itself, and the sample preparation cost is reduced. Through the SEM sample preparation process, the wire trace sample is clean, conductive, and has no dust on the surface, which meets the requirements for cleanliness and conductivity when using SEM for microscopic analysis of wire traces, while saving sample preparation costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 This is a schematic diagram of a method flow chart of Embodiment 1 of the present invention;

[0086] Figure 2 It is a schematic diagram of iterative binarization segmentation of the present invention;

[0087] Figure 3This is a schematic diagram of the method flow of Embodiment 2 of the present invention;

[0088] Figure 4 This is a schematic diagram of placing the SEM wire sample of the present invention;

[0089] Figure 5 It is a schematic diagram of a single-strand conductor of the present invention;

[0090] Figure 6 A typical SEM image of the wire trace of the present invention;

[0091] Figure 7 This is a schematic diagram of the system framework of Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0092] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0093] It should be understood that the step numbers used in this document are only for convenience of description and are not intended to limit the order in which the steps are executed.

[0094] It should be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0095] The terms “include” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0096] The term "and / or" means and includes any and all possible combinations of one or more of the associated listed items.

[0097] Embodiment 1:

[0098] See also Figure 1 This embodiment provides a method for modeling a fire cause identification model for steel core aluminum stranded wire, comprising the following steps:

[0099] S10, obtaining SEM images of several steel core aluminum stranded wire samples, and cropping and graying the SEM images of each steel core aluminum stranded wire sample to obtain a corresponding SEM grayscale image;

[0100] S20, performing binarization processing on each SEM grayscale image to obtain a corresponding binarized image; extracting trace features of cracks and holes in the binarized image, and performing correlation feature screening on the extracted trace features to obtain correlation features;

[0101] S30, constructing a training sample set including a binary image of each steel core aluminum stranded wire sample, a corresponding correlation feature, and a fire cause label of the corresponding steel core aluminum stranded wire sample (in this embodiment, the fire cause is divided into flame-induced line fault and line fault discharge), and establishing a steel core aluminum stranded wire fire cause identification model based on a machine learning algorithm;

[0102] S40, using the training sample set to train the steel core aluminum stranded wire fire cause identification model; obtaining the trained steel core aluminum stranded wire fire cause identification model.

[0103] In this embodiment, in order to distinguish the fire causes "line fault caused by flame" from "line fault discharge", it is specifically implemented as follows: when the map label is identified as "discharge trace", it is classified as "line fault discharge", and when the map label is identified as "flame burning trace", it is classified as "line fault caused by flame".

[0104] As a preferred implementation of this embodiment, the binarization process adopts an iterative segmentation threshold method. Assume that there is a grayscale image to be processed, the grayscale value of the pixel at the coordinate (i, j) on the image is Z(i, j), P(i, j) is the weight coefficient of the coordinate (i, j), and P(i, j) takes the probability of the grayscale of the point (i, j). The iterative threshold selection is close to the idea of ​​approximation. The iterative process flow chart is as follows: Figure 2 shown.

[0105] The specific iterative threshold selection process is as follows:

[0106] STEP1: Set the initial estimation threshold N0 according to the maximum Zmax and minimum Zmin of the gray value in the image;

[0107]

[0108] Among them, N0 is the initial estimation threshold; Z max is the maximum gray value; Z min is the minimum gray value;

[0109] STEP2: According to the threshold N k (represents the threshold of the kth iteration, N k The initial value is N0. The image is divided into foreground and background. The average grayscale values ​​of the foreground and background are Z and F and Z B ;

[0110]

[0111] Among them, Z F is the average gray value of the foreground; Z B is the average gray value of the background; Z(i, j) is the gray value of the pixel at coordinate (i, j); P(i, j) is the weight coefficient of coordinate (i, j), taking the probability of the gray value of coordinate (i, j);

[0112] STEP3: Find the new threshold N K+1 ;

[0113]

[0114] STEP4: Set the iteration end condition to N k Approximately equal to N k+1 , that is |N k+1 -N k |<ε, if the end condition is not met, the new threshold is brought into STEP2 and steps STEP2 to 3 are repeated;

[0115] STEP5: Iteration ends N k That is, the optimal segmentation threshold is selected to complete the binarization.

[0116] As a preferred implementation of this embodiment, a feature method is used to extract features from a binary image. Among them, the trace features of cracks and holes in the binary image can have 31 dimensions, mainly including: contrast, correlation, energy, homogeneity (0, 45, 90, 135), entropy in the initial state, number of holes, mean of hole area, variance of hole area, mean of minimum circumscribed matrix, mean of minimum circumscribed matrix perimeter, mean of hole / circumscribed matrix ratio, variance of minimum circumscribed matrix, variance of minimum circumscribed matrix perimeter, variance of hole / circumscribed matrix ratio, entropy of the processed binary image, mean of center point X, mean of center point Y, variance of center point X, variance of center point Y.

[0117] Parameters such as contrast, correlation, energy, homogeneity (0, 45, 90, 135), entropy in the initial state, and entropy of the processed binary image are all derived from the gray-level co-occurrence matrix (GLCM). GLCM represents texture with conditional probability density Pi,j,d,θ)[68,69]. Pi,j,d,θ) represents the probability of a bunch of image pixels with a relative distance d in the direction of θ having gray values ​​i and j respectively. The calculation formula is as follows:

[0118] P(i, j, d, θ)

[0119] ={[(x,y),(x+a,x+b)|f(x,y)]

[0120] =i; f(x+a, y+b)=j}

[0121] In the formula, (x, y) is the pixel with gray value i, (x+a, x+b) is the pixel with gray value j. In this project, d=1 is taken to represent adjacent pixels; θ considers four typical angles: horizontal (0°), diagonal (45°), vertical (90°), and anti-diagonal (135°).

[0122] In the grayscale image, assume that the grayscale value of the pixel (x, y) is i, and the grayscale value of the pixel (x+a, y+b) at a certain angle and a certain distance from it is j. Let (x, y) move in the image coordinate system to obtain the values ​​of each (x, y). The maximum grayscale level of the grayscale image is G. The combinations of i and j are statistically obtained to obtain all Pi, j, d, θ) values ​​to form a co-occurrence grayscale matrix W.

[0123] W = [P (i, j, d, θ)] G×G

[0124] The number of holes, the mean of the hole area, the variance of the hole area, the mean of the minimum circumscribed matrix, the mean of the minimum circumscribed matrix perimeter, the mean of the hole / circumscribed matrix ratio, the variance of the minimum circumscribed matrix perimeter, the variance of the hole / circumscribed matrix ratio, the mean of the center point X, the mean of the center point Y, the variance of the center point X, the variance of the center point Y and other parameters are obtained using the matlab tool regionprops function, and the coordinates of each white connected area BoundingBox are returned:

[0125] BoundingBox = [B1, B2, ... B m ]

[0126] Bm represents all external matrices, and m represents the total number of external matrices. Return parameter Area to obtain the pixel area of ​​each connected region.

[0127] Furthermore, this embodiment provides a correlation feature screening method. In this embodiment, there are 31 dimensions of trace features. The Pearson correlation coefficient is used to analyze the correlation between features and labels. 0-0.2 is very weak or almost no correlation; 0.2-0.4 is weak correlation; 0.4-0.6 is medium correlation; 0.6-0.8 is strong correlation; 0.8-1.0 is very strong correlation. Highly correlated features with a correlation coefficient greater than or equal to 0.3 are selected.

[0128] The correlation calculation formula is:

[0129]

[0130] Where n is the sample size; x iis any feature of the trace features of the i-th observation point; y is the corresponding label data; is the sample mean corresponding to any item in the trace feature; is the sample mean of the corresponding label data;

[0131] In this embodiment, a total of 19 features are selected, among which the key identification features are, from high to low, respectively: mean ratio of hole to external matrix area (0.75), number of holes (0.47), energy (0.4, 0.44, 0.45, 0.45), mean of minimum external matrix perimeter (0.39), mean of hole area (0.38), mean of minimum external matrix area (0.31), and homogeneity (0.3, 0.31).

[0132] As an implementation method, this embodiment uses the SVM model as the benchmark of the steel core aluminum stranded wire fire cause identification model, and divides the training sample set into a training set and a test set in a ratio of 70%:30%;

[0133] The samples in the training set are input into the steel core aluminum stranded wire fire cause identification model, wherein the binary image of the steel core aluminum stranded wire sample and the corresponding correlation feature are used as input features, and the fire cause label is used as the label value. The model is iteratively trained based on the error between the predicted value output by the steel core aluminum stranded wire fire cause identification model and the label value;

[0134] The test set test is input into the steel-core aluminum stranded wire fire cause identification model to evaluate whether the model accuracy reaches the preset threshold. If so, the training is completed.

[0135] Embodiment 2:

[0136] See Figure 3 This embodiment provides a method for identifying traces of steel-core aluminum stranded wire after fire by SEM images, comprising the following steps:

[0137] A10, collecting single-strand steel-core aluminum stranded wires on the burning side of fires caused by different fire causes (in this embodiment, the fire causes are classified into two types: flame-induced line faults and line fault discharges), pre-treating the steel-core aluminum stranded wires to obtain steel-core aluminum stranded wire samples;

[0138] A20. Obtain the SEM image of the steel core aluminum stranded wire sample by scanning electron microscopy;

[0139] A30. Use the trained steel core aluminum stranded wire fire cause identification model obtained by the modeling method of the steel core aluminum stranded wire fire cause identification model described in any embodiment of the present invention to identify the SEM image to obtain the cause of the fire.

[0140] As a preferred implementation of this embodiment, in step A10, the steel core aluminum stranded wire is pre-processed to obtain a steel core aluminum stranded wire sample as follows:

[0141] Step 1 Sampling: Cut 30-50mm long single-strand steel-core aluminum stranded wire on the burning side for sample preparation. Figure 4 As shown;

[0142] Step 2 Cutting: Use a cutting machine to adjust the sample to the appropriate size. The sample length is 10mm and the height cannot exceed 10mm.

[0143] Step 3: Cleaning: Use an ultrasonic dust removal instrument to clean the sample surface and remove the ash attached to the steel core aluminum stranded wire by smoke;

[0144] Step 4 Observation: Aluminum steel core aluminum stranded wire is a conductive material and can be directly observed. Use conductive glue to glue the sample to the sample holder. The sample is placed as follows: Figure 5 As shown, appropriate parameters were set to obtain SEM images. The acceleration voltage was 20 kV, the distance between the electron microscope lens and the sample surface was 7.5 mm, the magnification was 100 to 500 times, and the observation step was 200.

[0145] Typical SEM images are shown in Figure 6 . The surface of a normal conductor is relatively smooth, without cracks, holes or melt marks. As the medium-voltage line expands due to the high temperature of the forest flame, the microscopic image of its surface traces shows an increase in cracks and irregular depressions as the heating time increases. Irregular pits appear under the action of long-term flames, which is mainly caused by the aluminum strands melting under high temperature and then cooling and solidifying. Under the action of discharge, the medium-voltage line has fewer cracks in the microscopic image sample, and there are local small holes and local cracks, accompanied by pits caused by local deeper arcs. This is because the local discharge arc action time is short, but the temperature is as high as several thousand degrees, which can cause local burn pits in the conductor, and it is not easy for large areas of the conductor surface to melt, so there are fewer cracks.

[0146] It can identify the cause of medium voltage conductor traces, which are caused by discharge or flame, with an identification accuracy of 85%-90%.

[0147] Embodiment three:

[0148] See also Figure 6 This embodiment provides a SEM spectrum identification system for traces of steel-core aluminum stranded wire after fire, including:

[0149] An image processing module is used to obtain SEM images of several steel core aluminum stranded wire samples, and to perform cropping and grayscale processing on the SEM images of each steel core aluminum stranded wire sample to obtain a corresponding SEM grayscale image;

[0150] The feature extraction module performs binarization processing on each SEM grayscale image to obtain a corresponding binary image; extracts trace features of cracks and holes in the binary image, and performs correlation feature screening on the extracted trace features to obtain correlation features;

[0151] The training sample and model building module constructs a training sample set including the binary images of each steel core aluminum stranded wire sample, the corresponding correlation features, and the fire cause labels of the corresponding steel core aluminum stranded wire samples, and establishes a steel core aluminum stranded wire fire cause identification model based on a machine learning algorithm;

[0152] The model training module uses the training sample set to train the steel core aluminum stranded wire fire cause identification model; and obtains the trained steel core aluminum stranded wire fire cause identification model.

[0153] As a preferred implementation of this embodiment, the feature extraction module includes a binarization processing unit, and the binarization processing unit uses an iterative segmentation threshold method to perform binarization processing on each SEM grayscale image. The specific steps include:

[0154] The initial estimation threshold is set based on the maximum grayscale value and the minimum grayscale value in the SEM grayscale image. The specific formula is as follows:

[0155]

[0156] Among them, N0 is the initial estimation threshold; Z max is the maximum gray value; Z min is the minimum gray value;

[0157] Set the kth iteration threshold to N k , and N k The initial value of is N0; the image is divided into foreground and background based on the iterative threshold, and the average grayscale value of the foreground and background is calculated. The specific formulas are as follows:

[0158]

[0159] Among them, Z F is the average gray value of the foreground; Z B is the average gray value of the background; Z(i, j) is the gray value of the pixel at coordinate (i, j); P(i, j) is the weight coefficient of coordinate (i, j), taking the probability of the gray value of coordinate (i, j);

[0160] Calculate the new threshold, the specific formula is as follows:

[0161]

[0162] Set the iteration end condition to |N k+1 -N k|<ε, ε is the preset parameter; if the condition is not met, the new threshold N k+1 Substitute into the formula for calculating the average gray value of the foreground and background; if the conditions are met, output N k is the optimal segmentation threshold;

[0163] The SEM grayscale image is binarized based on the output optimal segmentation threshold.

[0164] As a preferred implementation of this embodiment, the feature extraction module includes a trace feature extraction unit, which is used to extract trace features of cracks and holes in the binary image. The specific steps are:

[0165] Identify cracks and holes, and remove small black areas with grayscale 0 through morphological corrosion operations;

[0166] The grayscale of the image is inverted, 0 grayscale is converted to 255 grayscale, 255 grayscale is converted to 0 grayscale, and the white area of ​​grayscale 255 represents the concave part in the SEM image;

[0167] Use the regionprops function to identify the white connected area, read the parameter BoundingBox, which is the diagonal coordinates of the external matrix of the connected area and the length and width of the rectangle, and obtain the minimum external matrix corresponding to the crack and hole parts;

[0168] Construct the following conditional probability density formula:

[0169] P(i,j,d,θ)={[(x,y),(x+a,x+b)|f(x,y)]=i; f(x+a,y+b)=j}

[0170] Where P(i,j,d,θ) represents the probability of a group of image pixels with a relative distance of d having grayscale values ​​i and j respectively in the direction of θ; (x, y) is the pixel with grayscale value i, (x+a, y+b) is the pixel with grayscale value j; d is the preset distance; θ is the preset direction;

[0171] The maximum gray level of the image is G, and the gray values ​​i and j are arranged and combined to obtain the numerical combination of all P(i, j, d, θ) to form a co-occurrence gray matrix;

[0172] Trace features are extracted based on the minimum circumscribed matrix and the co-occurrence grayscale matrix.

[0173] As a preferred implementation manner of this embodiment, the trace features include: contrast, correlation, energy, homogeneity, entropy in the initial state, the number of holes, the mean of the hole area, the variance of the hole area, the mean of the minimum circumscribed matrix, the mean of the minimum circumscribed matrix perimeter, the mean of the hole / circumscribed matrix ratio, the variance of the minimum circumscribed matrix, the variance of the minimum circumscribed matrix perimeter, the variance of the hole / circumscribed matrix ratio, the entropy of the processed binary image, the mean of the center point X, the mean of the center point Y, the variance of the center point X, and the variance of the center point Y.

[0174] As a preferred implementation of this embodiment, the feature extraction module includes a feature screening unit for performing correlation feature screening on the extracted trace features, and the specific steps are:

[0175] The correlation between trace features and labels is analyzed by Pearson correlation coefficient. xy , the specific formula is as follows:

[0176]

[0177] Where n is the sample size; x i is any feature of the trace features of the i-th observation point; y i is the corresponding label data; is the sample mean corresponding to any item in the trace feature; is the sample mean of the corresponding label data;

[0178] The correlation coefficient of each trace feature is calculated, a correlation coefficient threshold is set, and trace features with correlation coefficients greater than the correlation coefficient threshold are screened out as correlation features.

[0179] Embodiment 4:

[0180] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, a modeling method for a steel core aluminum stranded wire fire cause identification model as described in any embodiment of the present invention is implemented.

[0181] Embodiment five:

[0182] The present embodiment provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements a modeling method for a steel core aluminum stranded wire fire cause identification model as described in any embodiment of the present invention.

[0183] In the embodiments of the present application, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, c can be single or multiple.

[0184] Those of ordinary skill in the art will appreciate that the various units and algorithm steps described in the embodiments disclosed herein can be implemented in a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0185] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0186] In several embodiments provided in the present application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory; hereinafter referred to as: ROM), random access memory (Random Access Memory; hereinafter referred to as: RAM), disk or optical disk, and other media that can store program codes.

[0187] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A modeling method for identifying the cause of fire in steel core aluminum stranded wire, characterized in that: The following steps are involved: Acquire SEM images of several steel core aluminum stranded wire samples, and perform cropping and grayscale processing on the SEM images of each steel core aluminum stranded wire sample to obtain a corresponding SEM grayscale image; Binarization is performed on each SEM grayscale image to obtain a corresponding binary image; trace features of cracks and holes in the binary image are extracted, and correlation features are screened for the extracted trace features to obtain correlation features; Construct a training sample set including binary images of each steel core aluminum stranded wire sample, corresponding correlation features and fire cause labels of the corresponding steel core aluminum stranded wire samples, and establish a steel core aluminum stranded wire fire cause identification model based on machine learning algorithm; The fire cause identification model of steel core aluminum stranded wire is trained by using the training sample set; and a trained fire cause identification model of steel core aluminum stranded wire is obtained.

2. The modeling method of a steel core aluminum stranded wire fire cause identification model according to claim 1, characterized in that: The binarization process adopts an iterative segmentation threshold method, and the specific steps include: The initial estimation threshold is set based on the maximum grayscale value and the minimum grayscale value in the SEM grayscale image. The specific formula is as follows: Among them, N0 is the initial estimation threshold; Z max is the maximum gray value; Z min is the minimum gray value; Set the kth iteration threshold to N k , and N k The initial value of is N0; the image is divided into foreground and background based on the iterative threshold, and the average grayscale value of the foreground and background is calculated. The specific formulas are as follows: Among them, Z F is the average gray value of the foreground; Z B is the average gray value of the background; Z(i, j) is the gray value of the pixel at coordinate (i, j); P(i, j) is the weight coefficient of coordinate (i, j), taking the probability of the gray value of coordinate (i, j); Calculate the new threshold, the specific formula is as follows: Set the iteration end condition to |N k+1 -N k |<ε, ε is the preset parameter; if the condition is not met, the new threshold N k+1 Substitute into the formula for calculating the average gray value of the foreground and background; if the conditions are met, output N k is the optimal segmentation threshold; The SEM grayscale image is binarized based on the output optimal segmentation threshold.

3. The modeling method of the steel core aluminum stranded wire fire cause identification model according to claim 1 is characterized in that: The step of extracting the trace features of the crack and hole areas in the binary image is specifically as follows: Identify cracks and holes, and remove small black areas with grayscale 0 through morphological corrosion operations; The grayscale of the image is inverted, 0 grayscale is converted to 255 grayscale, 255 grayscale is converted to 0 grayscale, and the white area of ​​grayscale 255 represents the concave part in the SEM image; Use the regionprops function to identify the white connected area, read the parameter BoundingBox, which is the diagonal coordinates and the length and width of the rectangle of the white connected area's external matrix, and obtain the minimum external matrix corresponding to the crack and hole parts; Construct the following conditional probability density formula: P(i,j,d,θ)={[(x,y),(x+a,x+b)|f(x,y)]=i; f(x+a,y+b)=j} Where P(i,j,d,θ) represents the probability of a group of image pixels with a relative distance of d having grayscale values ​​i and j respectively in the direction of θ; (x, y) is the pixel with grayscale value i, (x+a, y+b) is the pixel with grayscale value j; d is the preset distance; θ is the preset direction; The maximum gray level of the image is G, and the gray values ​​i and j are arranged and combined to obtain the numerical combination of all P(i, j, d, θ) to form a co-occurrence gray matrix; Trace features are extracted based on the minimum circumscribed matrix and the co-occurrence grayscale matrix.

4. The modeling method of a steel core aluminum stranded wire fire cause identification model according to claim 3 is characterized by: The trace features include: contrast, correlation, energy, homogeneity, entropy in the initial state, the number of holes, the mean of the hole area, the variance of the hole area, the mean of the minimum circumscribed matrix, the mean of the minimum circumscribed matrix perimeter, the mean of the hole / circumscribed matrix ratio, the variance of the minimum circumscribed matrix, the variance of the minimum circumscribed matrix perimeter, the variance of the hole / circumscribed matrix ratio, the entropy of the processed binary image, the mean of the center point X, the mean of the center point Y, the variance of the center point X, and the variance of the center point Y.

5. The modeling method of a steel core aluminum stranded wire fire cause identification model according to claim 1, characterized in that: The correlation feature screening is specifically as follows: The correlation between trace features and labels is analyzed by Pearson correlation coefficient. xy , the specific formula is as follows: Where n is the sample size; x i is any feature of the trace features of the i-th observation point; y i is the corresponding label data; is the sample mean corresponding to any item in the trace feature; is the sample mean of the corresponding label data; The correlation coefficient of each trace feature is calculated, a correlation coefficient threshold is set, and trace features with correlation coefficients greater than the correlation coefficient threshold are screened out as correlation features.

6. A method for identifying traces of steel-core aluminum stranded wire after fire by SEM images, characterized in that: The following steps are involved: Collect a single strand of aluminum-steel core stranded wire from the burning side of the fire, pre-treat the aluminum-steel core stranded wire, and obtain an aluminum-steel core stranded wire sample; Obtaining a SEM image of the steel core aluminum stranded wire sample by scanning electron microscopy; The trained steel core aluminum stranded wire fire cause identification model obtained by the modeling method of the steel core aluminum stranded wire fire cause identification model as claimed in any one of claims 1 to 5 is used to identify the SEM image to obtain the cause of the fire.

7. The method for identifying traces of steel-core aluminum stranded wire after fire according to claim 6, characterized in that: The method for pre-treating the steel core aluminum stranded wire is specifically as follows: The steel core aluminum stranded wire is adjusted to the set size and impurities on the steel core aluminum stranded wire are removed by ultrasonic cleaning.

8. A modeling system for identifying the cause of fire in steel core aluminum stranded wire, characterized in that: include: An image processing module is used to obtain SEM images of several steel core aluminum stranded wire samples, and to perform cropping and grayscale processing on the SEM images of each steel core aluminum stranded wire sample to obtain a corresponding SEM grayscale image; The feature extraction module performs binarization processing on each SEM grayscale image to obtain a corresponding binary image; extracts trace features of cracks and holes in the binary image, and performs correlation feature screening on the extracted trace features to obtain correlation features; The training sample and model building module constructs a training sample set including the binary images of each steel core aluminum stranded wire sample, the corresponding correlation features, and the fire cause labels of the corresponding steel core aluminum stranded wire samples, and establishes a steel core aluminum stranded wire fire cause identification model based on a machine learning algorithm; The model training module uses the training sample set to train the steel core aluminum stranded wire fire cause identification model; and obtains the trained steel core aluminum stranded wire fire cause identification model.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method for modeling a steel core aluminum stranded wire fire cause identification model as described in any one of claims 1 to 5 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for modeling a steel core aluminum stranded wire fire cause identification model as described in any one of claims 1 to 5 is implemented.

Images