Method and system for evaluating bolt health state based on cable bridge bolt maintenance data

Abstract

The invention relates to the technical field of data analysis, in particular to a method and system for evaluating the health state of a bolt based on cable bridge bolt overhaul data, and the method comprises the steps: obtaining a bolt overhaul picture and position data; bolt corrosion characteristic parameters and displacement characteristic parameters are extracted; preprocessing the historical maintenance pictures, performing clustering analysis to establish a bolt replacement feature association rule, and screening according to a preset threshold to obtain a strong association rule; evaluating a bolt state based on a strong association rule; and establishing health monitoring indexes and weights, inputting the health monitoring indexes and weights into a TOPSIS model, calculating an Euclidean distance in combination with optimal and worst parameters, and judging whether early warning is performed or not according to the state relative proximity. According to the invention, subjectivity and uncertainty of traditional manual detection are avoided, and consistency of evaluation standards can be maintained especially when a large amount of bolt maintenance data is processed. Meanwhile, association rules of multiple variables such as corrosion characteristics and displacement characteristics for bolt failure are mined.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and in particular to a method and system for evaluating the health status of cable bridge bolts based on maintenance data of the bolts. Background Art

[0002] The two common bridge structures for cable bridges across rivers or across the sea are cable-stayed bridges and suspension bridges. The principle of a cable-stayed bridge is that the cable towers support the main beams of the bridge deck through cables, while the suspension bridge uses cables to connect two cable towers, and the cables hang down multiple spaced hangers to suspend the main beams of the bridge deck. The cables and hangers of a suspension bridge are mainly connected by cable clamp bolts, while the cables, main beams of the bridge deck, and lock towers of a cable-stayed bridge use high-strength fixed bolts. At the beginning of bridge construction, bolts are generally purchased from the same batch or adjacent batches of the same manufacturer, and undergo strict quality spot checks and inspections, so it can be assumed that they have roughly the same service life. As the years of use increase, the initial bolts will become loose and rust and fail. These two types of failures need to be discovered and replaced through inspections. Generally speaking, a 1,500-meter cable-stayed bridge across the river requires 120,000 cable bolts (excluding high-strength bolts in the steel truss splicing structure, which number is approximately 800,000). The bridge maintenance unit will perform maintenance on the bolts every year. The earliest was manual high-altitude inspection and investigation. Currently, drone inspections have also been used. However, facing the huge number of bolt maintenance volumes, the maintenance workload is also extremely large. Therefore, under the appropriate service life, the cost of collective replacement is less than the cost of replacing a single maintenance that finds an abnormality. The appropriate service life is related to the environment of the bridge, such as the bolt process material, force load size, environmental corrosion, environmental wind load, etc., so it is difficult to predict the appropriate service life. Sometimes, for safety reasons, although most of the bolts are still under safe use conditions, they are replaced as a whole in advance when they reach the conservative service life, wasting a lot of costs. Therefore, it is necessary to evaluate the true health status of the bolts through daily maintenance data. Summary of the invention

[0003] (1) Technical issues to be solved

[0004] The object of the present invention is to provide a method and system for evaluating the health status of cable bridge bolts based on the maintenance data of the bolts, so as to solve the problem of bolt evaluation and replacement.

[0005] (2) Technical solution

[0006] To achieve the above objectives, on the one hand, the present invention provides a method for evaluating the health status of a bolt based on the maintenance data of a cable bridge bolt, the method comprising:

[0007] Obtain a cable bridge bolt maintenance picture and corresponding cable bridge bolt position data, wherein the cable bridge bolt position data includes a cable number and a serial number of the bolt corresponding to the cable number; extract bolt detection data through the cable bridge bolt maintenance picture, including a first bolt corrosion characteristic parameter and a first bolt displacement characteristic parameter, wherein the first bolt corrosion characteristic parameter is characterized by bolt surface texture data, and the first bolt displacement characteristic parameter is characterized by a bolt axial displacement parameter and a bolt structure deformation parameter;

[0008] Preprocess the cable bridge bolt maintenance history image to obtain the first maintenance data, perform cluster analysis on the second bolt corrosion characteristic parameter and the second bolt displacement characteristic parameter of the replaced bolt in the first maintenance data to obtain the second maintenance data; establish a bolt replacement feature association rule through the second maintenance data, and obtain a bolt replacement feature strong association rule based on a preset feature association threshold, and evaluate the first bolt corrosion characteristic parameter and the first bolt displacement characteristic parameter based on the bolt replacement feature strong association rule to obtain a bolt status evaluation result;

[0009] The bolt health monitoring index and the bolt health monitoring weight are established based on the bolt corrosion characteristic parameters and the bolt displacement characteristic parameters through the bolt state assessment result; the bolt health monitoring index and the bolt health monitoring weight are input into the TOPSIS assessment model, and the Euclidean distance parameters between the bolt detection data and the optimal parameters of the bolt state and the worst parameters of the bolt state are calculated in combination with the preset optimal parameters of the bolt state and the worst parameters of the bolt state; the relative proximity of the bolt state is calculated according to the Euclidean distance parameters, and a warning signal is issued when the relative proximity of the bolt state is greater than a preset bolt state proximity threshold.

[0010] Further, the bolt detection data extracted through the cable bridge bolt maintenance picture includes a first bolt corrosion characteristic parameter and a first bolt displacement characteristic parameter, the first bolt corrosion characteristic parameter is characterized by bolt surface texture data, and the first bolt displacement characteristic parameter is characterized by a bolt axial displacement parameter and a bolt structure deformation parameter. The method includes:

[0011] The cable bridge bolt inspection picture is subjected to image preprocessing to obtain a first preprocessed image, wherein the image preprocessing includes image size calibration processing, image illumination correction processing and regional white balance processing; the bolt area in the first preprocessed image is extracted by the Otsu adaptive threshold segmentation method to obtain the gray value matrix G of the bolt area s ; Extract the edge contour line of the bolt area by Laplace edge detection operator, and obtain the coordinates of the bolt center positioning point and the bolt radius size by Hough circle transform;

[0012] The gray value matrix G of the bolt area sPerform wavelet decomposition to obtain multiple subband coefficient matrices including the low-frequency approximate coefficient matrix C a , horizontal detail coefficient matrix C h , vertical detail coefficient matrix C v and the diagonal detail coefficient matrix C d ; Calculate the texture feature vector T of the bolt area f for:

[0013]

[0014] Where N p is the total number of pixels in the bolt area; according to the texture feature vector T f and a preset corrosion texture vector threshold to calculate the first bolt corrosion characteristic parameter, including a bolt corrosion area ratio parameter and a bolt corrosion depth ratio parameter;

[0015] A polar coordinate system is established with the bolt center positioning point as the origin, and the edge contour line of the bolt area is discretized into a preset N e sampling points, record the polar coordinates of each sampling point [r i ,θ i ]; Construct the bolt edge profile feature vector E f for:

[0016]

[0017] where r 0 is the nominal radius of the bolt, θ 0 is the nominal angle of the sampling point, and are radial unit vectors and tangential unit vectors respectively; calculate the first bolt displacement characteristic parameter according to the bolt edge profile characteristic vector, including the bolt axial displacement parameter and the bolt structure deformation parameter; and normalize the first bolt corrosion characteristic parameter and the first bolt displacement characteristic parameter.

[0018] Furthermore, the method of preprocessing the cable bridge bolt maintenance history image to obtain first maintenance data, clustering the second bolt corrosion characteristic parameter and the second bolt displacement characteristic parameter of the replaced bolt in the first maintenance data to obtain second maintenance data; establishing bolt replacement feature association rules through the second maintenance data, and obtaining bolt replacement feature strong association rules according to a preset feature association threshold includes:

[0019] The cable bridge bolt maintenance history picture is subjected to image preprocessing to obtain a second preprocessed image, wherein the image preprocessing includes image grayscale processing, Gaussian filter noise reduction processing and grayscale histogram equalization processing; a bolt area grayscale value matrix X of the second preprocessed image is obtained. g, the bolt area gray value matrix X is transformed into g Map to high-dimensional reproducing Hilbert space and calculate the kernel matrix K h for:

[0020]

[0021] where σ 1 is the preset Gaussian kernel function bandwidth parameter; the kernel matrix K h Perform eigenvalue decomposition to obtain the eigenvalue λ i and its corresponding eigenvector v i , select the eigenvectors corresponding to the preset first m maximum eigenvalues, project the second preprocessed image into the low-dimensional subspace spanned by the m eigenvectors to obtain the first maintenance data X 1 for:

[0022] X 1 =X g V m ;

[0023] Where V m is a projection matrix composed of m eigenvectors; the first maintenance data X is clustered by DBSCAN density clustering algorithm 1 Perform cluster analysis based on the first maintenance data X 1 Calculate the Euclidean distance matrix D between samples e , for the first inspection data X 1 Each sample point p in i Calculate its ∈ d The number of sample points in the neighborhood N ∈ (p i ), where ∈ d is the preset bolt feature cluster radius parameter; when N ∈ (p i ) is greater than the preset bolt feature density threshold M p When the sample point p i Mark as core point, traverse all core sample points in turn, and compare each core sample point and its Euclidean distance less than ∈ d The sample points of the same bolt feature cluster are taken as the bolt feature cluster, and the bolt feature cluster is taken as the second maintenance data X 2 ;

[0024] The second maintenance data X 2 Discretization processing obtains the bolt feature item set I f , the bolt feature item set I is calculated by Apriori algorithm f The support S u for:

[0025]

[0026] Where A is the bolt feature item set I f subset, N(A) is the number of samples containing subset A, N t is the total number of samples; calculate the bolt feature item set I f The confidence level C f for:

[0027]

[0028] Where A and B are mutually exclusive subsets of bolt feature items; according to the preset feature association first threshold θ s and feature association second threshold θ c Filter to meet S u >θ s And C f >θ c The association rule of is used as the strong association rule of the bolt replacement feature;

[0029] The first threshold value θ of the feature association of the bolt feature item set s for:

[0030]

[0031] Where n f is the number of bolt failure samples in the historical cable bridge bolt maintenance, N t is the total number of samples in the historical cable bridge bolt maintenance; the first threshold value θ of the feature association of the bolt feature item set s for:

[0032]

[0033] Where n c is the number of samples in which bolt failure occurred during two consecutive inspections in the historical cable bridge bolt inspections, n f is the number of bolt failure samples in the historical cable bridge bolt maintenance.

[0034] Furthermore, the method of evaluating the first bolt corrosion characteristic parameter and the first bolt displacement characteristic parameter according to the bolt replacement characteristic strong association rule to obtain a bolt state evaluation result includes:

[0035] The first bolt corrosion characteristic parameter is defined as the bolt corrosion area ratio r a and bolt corrosion depth ratio r d , the first bolt displacement characteristic parameter is defined as the bolt axial displacement d a and bolt structure deformation d s ; According to the strong association rule of the bolt replacement characteristics, the critical parameter vector of bolt failure is obtained in is the critical value of bolt corrosion area ratio, is the critical value of bolt corrosion depth ratio, is the critical value of the bolt axial displacement, is the critical value of bolt structure deformation;

[0036] Calculate the characteristic parameter vector X of the current detected bolt f =[r a ,r d ,d a ,d s ] T and the bolt failure critical parameter vector θ f The Mahalanobis distance D m :

[0037]

[0038] Where Σ is the covariance matrix of the eigenvalue vector; calculate the eigenvalue vector X f The failure probability P f :

[0039]

[0040] where α f is the preset failure probability coefficient; according to the failure probability P f Get the bolt status assessment results.

[0041] Further, the bolt health monitoring index and bolt health monitoring weight based on the bolt corrosion characteristic parameter and the bolt displacement characteristic parameter are established through the bolt state evaluation result; the bolt health monitoring index and the bolt health monitoring weight are input into the TOPSIS evaluation model, and the Euclidean distance parameter between the bolt detection data and the optimal bolt state parameter and the worst bolt state parameter is calculated in combination with the preset optimal bolt state parameter and the worst bolt state parameter; the relative proximity of the bolt state is calculated according to the Euclidean distance parameter, and the method of issuing an early warning signal when the relative proximity of the bolt state is greater than the preset bolt state proximity threshold comprises:

[0042] The bolt state evaluation results are used to construct a bolt state evaluation matrix M according to the number of bolts. a , the bolt condition assessment matrix M a Contains a feature evaluation vector for each bolt, wherein the feature evaluation vector includes a first bolt corrosion feature parameter and a first bolt displacement feature parameter of the bolt; and calculates the bolt state evaluation matrix M by using the Euclidean norm. a The square sum of each characteristic parameter in each column is obtained, and each characteristic parameter is divided by the corresponding square sum to obtain the bolt state normalization matrix Mb ; Normalize the bolt state matrix M b Each column characteristic parameter of is multiplied by the bolt health monitoring weight to obtain the bolt state weighted matrix M c ;

[0043] Through the second inspection data X 2 The failure parameter vector V of the TOPSIS evaluation model is calculated by using the second bolt corrosion characteristic parameter and the second bolt displacement characteristic parameter of the bolt failure sample in the bolt maintenance of the historical cable bridge f , through the second maintenance data X 2 The health parameter vector V of the TOPSIS evaluation model is obtained by using the second bolt corrosion characteristic parameter and the second bolt displacement characteristic parameter of the bolt non-failure sample in the historical cable bridge bolt maintenance. h ; Calculate the bolt state weighted matrix M c The feature evaluation vector L for each bolt in i′ With the failure parameter vector V f and the health parameter vector V h The state distances are D i′f and D i′h for:

[0044]

[0045] Where n f is the feature evaluation vector L i′ The number of elements in L i′ (j) is the j′th characteristic parameter of the i′th bolt, V f (j) is the failure parameter vector V f The j′th characteristic parameter of h (j) is the health parameter vector V h The j′th characteristic parameter of the bolt is calculated; the health status score S of the i′th bolt is calculated. s′ for:

[0046]

[0047] Get the adjacent preset number N of the cable numbers corresponding to the bolt numbers s Health status score sequence of bolts where j ′ The value range is 1 to N s ; Through the second inspection data X 2 The health status score sequence of bolt failure samples in the bolt maintenance of historical cable bridges is used to calculate the mean health status score U s and standard deviation D s , get the bolt status warning threshold T sfor:

[0048] T s =U s +B s D s ;

[0049] Among them B s is a preset status scoring coefficient; when the health status score S i′ Greater than the bolt status warning threshold T s and is greater than the health status score sequence When the arithmetic mean of

[0050] Based on the same inventive concept, on the other hand, the present invention also provides a system for evaluating the health status of bolts based on cable bridge bolt inspection data, the system comprising:

[0051] A maintenance data acquisition module is used to obtain a cable bridge bolt maintenance picture and corresponding cable bridge bolt position data, wherein the cable bridge bolt position data includes a cable number and a serial number of the bolt corresponding to the cable number; bolt detection data extracted from the cable bridge bolt maintenance picture includes a first bolt corrosion characteristic parameter and a first bolt displacement characteristic parameter, wherein the first bolt corrosion characteristic parameter is characterized by bolt surface texture data, and the first bolt displacement characteristic parameter is characterized by a bolt axial displacement parameter and a bolt structure deformation parameter;

[0052] A historical data training module is used to pre-process the cable bridge bolt maintenance history pictures to obtain first maintenance data, and to cluster the second bolt corrosion characteristic parameters and the second bolt displacement characteristic parameters of the replaced bolts in the first maintenance data to obtain second maintenance data; to establish bolt replacement feature association rules through the second maintenance data, and to obtain bolt replacement feature strong association rules based on the preset feature association threshold, and to evaluate the first bolt corrosion characteristic parameters and the first bolt displacement characteristic parameters based on the bolt replacement feature strong association rules to obtain bolt status evaluation results;

[0053] The health scoring module is used to establish a bolt health monitoring index and a bolt health monitoring weight based on the bolt corrosion characteristic parameter and the bolt displacement characteristic parameter according to the bolt state evaluation result; input the bolt health monitoring index and the bolt health monitoring weight into the TOPSIS evaluation model, and calculate the Euclidean distance parameter between the bolt detection data and the optimal bolt state parameter and the worst bolt state parameter in combination with the preset optimal bolt state parameter and the worst bolt state parameter; calculate the relative proximity of the bolt state according to the Euclidean distance parameter, and issue a warning signal when the relative proximity of the bolt state is greater than a preset bolt state proximity threshold.

[0054] Furthermore, the system further comprises:

[0055] The maintenance data feature extraction module is used to perform image preprocessing on the cable bridge bolt maintenance picture to obtain a first preprocessed image, wherein the image preprocessing includes image size calibration processing, image illumination correction processing and regional white balance processing; the bolt area in the first preprocessed image is extracted by the Otsu adaptive threshold segmentation method to obtain the gray value matrix G of the bolt area s ; Extract the edge contour line of the bolt area by Laplace edge detection operator, and obtain the coordinates of the bolt center positioning point and the bolt radius size by Hough circle transform;

[0056] The gray value matrix G of the bolt area s Perform wavelet decomposition to obtain multiple subband coefficient matrices including the low-frequency approximate coefficient matrix C a , horizontal detail coefficient matrix C h , vertical detail coefficient matrix C v and the diagonal detail coefficient matrix C d ; Calculate the texture feature vector T of the bolt area f for:

[0057]

[0058] Where N p is the total number of pixels in the bolt area; according to the texture feature vector T f and a preset corrosion texture vector threshold to calculate the first bolt corrosion characteristic parameter, including a bolt corrosion area ratio parameter and a bolt corrosion depth ratio parameter;

[0059] A polar coordinate system is established with the bolt center positioning point as the origin, and the edge contour line of the bolt area is discretized into a preset N e sampling points, record the polar coordinates of each sampling point [r i ,θ i ]; Construct the bolt edge profile feature vector E f for:

[0060]

[0061] where r 0 is the nominal radius of the bolt, θ 0 is the nominal angle of the sampling point, and are radial unit vectors and tangential unit vectors respectively; calculate the first bolt displacement characteristic parameter according to the bolt edge profile characteristic vector, including the bolt axial displacement parameter and the bolt structure deformation parameter; and normalize the first bolt corrosion characteristic parameter and the first bolt displacement characteristic parameter.

[0062] Furthermore, the system further comprises:

[0063] The historical data feature mapping module is used to perform image preprocessing on the cable bridge bolt maintenance history image to obtain a second preprocessed image, wherein the image preprocessing includes image grayscale processing, Gaussian filter noise reduction processing and grayscale histogram equalization processing; obtain the bolt area grayscale value matrix X of the second preprocessed image g , the bolt area gray value matrix X is transformed into g Map to high-dimensional reproducing Hilbert space and calculate the kernel matrix K h for:

[0064]

[0065] where σ 1 is the preset Gaussian kernel function bandwidth parameter; the kernel matrix K h Perform eigenvalue decomposition to obtain the eigenvalue λ i and its corresponding eigenvector v i , select the eigenvectors corresponding to the preset first m maximum eigenvalues, project the second preprocessed image into the low-dimensional subspace spanned by the m eigenvectors to obtain the first maintenance data X 1 for:

[0066] X 1 =X g V m ;

[0067] Where V m is a projection matrix composed of m eigenvectors; the first maintenance data X is clustered by DBSCAN density clustering algorithm 1 Perform cluster analysis based on the first maintenance data X 1 Calculate the Euclidean distance matrix D between samples e , for the first inspection data X 1 Each sample point p in i Calculate its ∈ d The number of sample points in the neighborhood N ∈ (p i ), where ∈ d is the preset bolt feature cluster radius parameter; when N ∈ (p i ) is greater than the preset bolt feature density threshold M p When the sample point p i Mark as core point, traverse all core sample points in turn, and compare each core sample point and its Euclidean distance less than ∈ d The sample points of the same bolt feature cluster are taken as the bolt feature cluster, and the bolt feature cluster is taken as the second maintenance data X2 ;

[0068] The second maintenance data X 2 Discretization processing obtains the bolt feature item set I f , the bolt feature item set I is calculated by Apriori algorithm f The support S u for:

[0069]

[0070] Where A is the bolt feature item set I f subset, N(A) is the number of samples containing subset A, N t is the total number of samples; calculate the bolt feature item set I f The confidence level C f for:

[0071]

[0072] Where A and B are mutually exclusive subsets of bolt feature items; according to the preset feature association first threshold θ s and feature association second threshold θ c Filter to meet S u >θ s And C f >θ c The association rule of is used as the strong association rule of the bolt replacement feature;

[0073] The first threshold value θ of the feature association of the bolt feature item set s for:

[0074]

[0075] Where n f is the number of bolt failure samples in the historical cable bridge bolt maintenance, N t is the total number of samples in the historical cable bridge bolt maintenance; the first threshold value θ of the feature association of the bolt feature item set s for:

[0076]

[0077] Where n c is the number of samples in which bolt failure occurred during two consecutive inspections in the historical cable bridge bolt inspections, n f is the number of bolt failure samples in the historical cable bridge bolt maintenance.

[0078] Furthermore, the system further comprises:

[0079] A feature matching evaluation module is used to define the first bolt corrosion feature parameter as a bolt corrosion area ratio r a and bolt corrosion depth ratio r d , the first bolt displacement characteristic parameter is defined as the bolt axial displacement d a and bolt structure deformation d s ; According to the strong association rule of the bolt replacement characteristics, the critical parameter vector of bolt failure is obtained in is the critical value of bolt corrosion area ratio, is the critical value of bolt corrosion depth ratio, is the critical value of the bolt axial displacement, is the critical value of bolt structure deformation;

[0080] Calculate the characteristic parameter vector X of the current detected bolt f =[r a ,r d ,d a ,d s ] T and the bolt failure critical parameter vector θ f The Mahalanobis distance D m :

[0081]

[0082] Where Σ is the covariance matrix of the eigenvalue vector; calculate the eigenvalue vector X f The failure probability P f :

[0083]

[0084] where α f is the preset failure probability coefficient; according to the failure probability P f Get the bolt status assessment results.

[0085] Furthermore, the system further comprises:

[0086] The bolt replacement warning module is used to construct a bolt status evaluation matrix M based on the bolt status evaluation results according to the number of bolts. a , the bolt condition assessment matrix M a Contains a feature evaluation vector for each bolt, wherein the feature evaluation vector includes a first bolt corrosion feature parameter and a first bolt displacement feature parameter of the bolt; and calculates the bolt state evaluation matrix M by using the Euclidean norm. a The square sum of each characteristic parameter in each column is obtained, and each characteristic parameter is divided by the corresponding square sum to obtain the bolt state normalization matrix M b ; Normalize the bolt state matrix Mb Each column characteristic parameter of is multiplied by the bolt health monitoring weight to obtain the bolt state weighted matrix M c ;

[0087] Through the second inspection data X 2 The failure parameter vector V of the TOPSIS evaluation model is calculated by using the second bolt corrosion characteristic parameter and the second bolt displacement characteristic parameter of the bolt failure sample in the bolt maintenance of the historical cable bridge f , through the second maintenance data X 2 The health parameter vector V of the TOPSIS evaluation model is obtained by using the second bolt corrosion characteristic parameter and the second bolt displacement characteristic parameter of the bolt non-failure sample in the historical cable bridge bolt maintenance. h ; Calculate the bolt state weighted matrix M c The feature evaluation vector L for each bolt in i′ With the failure parameter vector V f and the health parameter vector V h The state distances are D i′f and D i′h for:

[0088]

[0089] Where n f is the feature evaluation vector L i′ The number of elements in L i′ (j) is the j′th characteristic parameter of the i′th bolt, V f (j) is the failure parameter vector V f The jth ′ characteristic parameters; V h (j) is the health parameter vector V h The jth ′ characteristic parameters; calculate the i-th ′ The health status score of each bolt is S s′ for:

[0090]

[0091] Get the adjacent preset number N of the cable numbers corresponding to the bolt numbers s Health status score sequence of bolts where j ′ The value range is 1 to N s ; Through the second inspection data X 2 The health status score sequence of bolt failure samples in the bolt maintenance of historical cable bridges is used to calculate the mean health status score U s and standard deviation D s , get the bolt status warning threshold Ts for:

[0092] T s =U s +B s D s ;

[0093] Among them B s is a preset status scoring coefficient; when the health status score S i′ Greater than the bolt status warning threshold T s and is greater than the health status score sequence When the arithmetic mean of

[0094] (3) Beneficial effects

[0095] Compared with the prior art, the invention has the beneficial effect of avoiding the subjectivity and uncertainty of traditional manual detection, especially when processing a large amount of bolt maintenance data, it can maintain the consistency of the evaluation standard. At the same time, it mines the association rules of multiple variables such as corrosion characteristics and displacement characteristics for bolt failure. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] Figure 1 This is a block diagram of a method for evaluating the health status of bolts based on cable bridge bolt maintenance data according to Embodiment 1 of the present invention;

[0097] Figure 2 This is a module block diagram of a system for evaluating the health status of bolts based on cable bridge bolt maintenance data according to embodiment 2 of the present invention. DETAILED DESCRIPTION

[0098] 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.

[0099] Before giving examples, it is necessary to explain the application scenarios of the present invention. For a 2,000-meter cable-stayed bridge, the tower height is 300 meters, the distance between the towers is 1,600 meters, and the tension of a single cable is close to 20,000 tons. All tensions are fastened through bolt connection points. Nearly 150,000 bolts need to be inspected, and the inspection is to determine whether they are rusted or loose by the appearance of the bolts. Whether it is manual inspection or inspection by means of drones, it is necessary to take photos and mark the appearance of the bolts. The failure severity is analyzed by the obtained bolt photos, and it can be replaced after further judgment. In the process of analysis, it is necessary to further evaluate the health status of the bolts based on the overall maintenance data of the bolts. If the health status of most of the bolts is not good, then the bolts can be replaced as a whole to reduce the cost of inspection, replacement and maintenance. However, in actual judgment, such as processing the bolt maintenance pictures, rust textures and bolt displacement or deformation will be obtained. The difficulty in accurately judging whether replacement is necessary is that the rust texture and bolt displacement or deformation may occur alone or simultaneously. If the judgment is based solely on the rust texture, it may be that only the surface of the threaded stud is rusted, while the internal thread contact is well tightened and has not yet reached the point of replacement. Therefore, it is necessary to comprehensively analyze historical maintenance data, comprehensively judge the characteristic parameters of various aspects, and reasonably plan the weight of each parameter.

[0100] Example 1: Figure 1 As shown, this embodiment provides a method for evaluating the health status of bolts based on cable bridge bolt inspection data, the method comprising:

[0101] Obtain a cable bridge bolt maintenance picture and corresponding cable bridge bolt position data, wherein the cable bridge bolt position data includes a cable number and a serial number of the bolt corresponding to the cable number; extract bolt detection data through the cable bridge bolt maintenance picture, including a first bolt corrosion characteristic parameter and a first bolt displacement characteristic parameter, wherein the first bolt corrosion characteristic parameter is characterized by bolt surface texture data, and the first bolt displacement characteristic parameter is characterized by a bolt axial displacement parameter and a bolt structure deformation parameter;

[0102] Preprocess the cable bridge bolt maintenance history image to obtain the first maintenance data, perform cluster analysis on the second bolt corrosion characteristic parameter and the second bolt displacement characteristic parameter of the replaced bolt in the first maintenance data to obtain the second maintenance data; establish a bolt replacement feature association rule through the second maintenance data, and obtain a bolt replacement feature strong association rule based on a preset feature association threshold, and evaluate the first bolt corrosion characteristic parameter and the first bolt displacement characteristic parameter based on the bolt replacement feature strong association rule to obtain a bolt status evaluation result;

[0103] The bolt health monitoring index and the bolt health monitoring weight are established based on the bolt corrosion characteristic parameters and the bolt displacement characteristic parameters through the bolt state assessment result; the bolt health monitoring index and the bolt health monitoring weight are input into the TOPSIS assessment model, and the Euclidean distance parameters between the bolt detection data and the optimal parameters of the bolt state and the worst parameters of the bolt state are calculated in combination with the preset optimal parameters of the bolt state and the worst parameters of the bolt state; the relative proximity of the bolt state is calculated according to the Euclidean distance parameters, and a warning signal is issued when the relative proximity of the bolt state is greater than a preset bolt state proximity threshold.

[0104] For example, the operation and maintenance department of a Yangtze River Bridge in a certain city carried out routine maintenance on the cable-stayed system of the bridge in January 2024. The bridge has a total of 128 cables, each of which is installed with an average of 1,200 high-strength bolts, totaling about 154,000 bolts that need to be inspected. The maintenance personnel used a DJI M300 RTK industrial drone equipped with a 60x zoom lens and a Zenmuse H20T camera to take high-definition photos of each bolt (resolution 4000×3000 pixels). While the drone takes pictures, it automatically records the detailed location information of each bolt, such as "C045-089" for the 89th bolt on the No. 45 cable, and "C045-090" for the 90th bolt on the same cable. The photos and location information of each bolt are uploaded to the bridge health monitoring database in real time to form a complete bolt maintenance data set.

[0105] Then, two types of key test data were collected for each bolt photo. The first type was the bolt corrosion characteristic parameters: by analyzing the surface texture characteristics, it was found that the surface of the bolt numbered "C045-089" showed irregular brown-red rust patches, covering an area of ​​25% of the entire visible surface. The rust depth measured by 3D profile scanning reached 30% of the surface treatment layer, and some areas had eroded the bolt body. In contrast, the adjacent "C045-090" bolt only had a few rust spots on the threaded part, with an area of ​​only 5% of the surface and a depth of no more than 10% of the surface treatment layer. The second type was the bolt displacement characteristic parameters: measured by a high-precision displacement sensor, the "C045-089" bolt was displaced 2.8 mm in the axial direction relative to the installation reference plane, the bolt head was elliptical, and the structural deformation reached 3% of the original size; while the axial displacement of the "C045-090" bolt was only 0.3 mm, the bolt head remained intact and round, and there was no obvious structural deformation.

[0106] Subsequently, all the historical data of bolt maintenance of the bridge from 2019 to 2023 were retrieved, totaling 770,000 detailed records. Among them, 12,500 bolt samples that had been replaced were found. These samples can be roughly divided into three categories: 2,100 were replaced simply due to severe corrosion, 1,500 were replaced simply due to excessive displacement, and 8,900 were replaced due to both severe corrosion and obvious displacement. These failed samples were analyzed using a density clustering algorithm, and four typical failure mode combinations were identified: Type A (severe corrosion + slight displacement), Type B (moderate corrosion + moderate displacement), Type C (slight corrosion + severe displacement), and Type D (severe corrosion + severe displacement). Each failure mode has its characteristic parameter range. For example, the bolts in Type D generally show a corrosion area of ​​more than 20%, a corrosion depth of more than 25%, and an axial displacement of more than 2.5 mm or a structural deformation of more than 2.5%.

[0107] Based on the cluster analysis results, it was found that when the surface corrosion area of ​​the bolt exceeded 20% and the depth exceeded 25%, its support was 0.71 (8,900 / 12,500) and the confidence reached 0.75, indicating that in this case, the bolt has a 75% probability of needing to be replaced; when the axial displacement of the bolt exceeded 2.5 mm or the structural deformation exceeded 2.5%, its support was 0.75 (9,400 / 12,500) and the confidence reached 0.80, indicating that in this case, the bolt has an 80% probability of needing to be replaced. When setting the feature association threshold, the historical data distribution was considered: the support threshold was set to 0.70 (derived from 12,500 / 77,000, indicating the proportion of failed samples in the total samples), and the confidence threshold was set to 0.75 (derived from 9,400 / 12,500, indicating the proportion of continuous failed samples in the failed samples). After screening, 8 strong association rules were finally determined as the evaluation criteria for bolt replacement.

[0108] Subsequently, a bolt health assessment system was established through the TOPSIS model, and two types of monitoring indicators and their weights were set: the weight of the corrosion characteristic indicator was 60% (of which the area accounted for 35% and the depth accounted for 25%), and the weight of the displacement characteristic indicator was 40% (of which the axial displacement accounted for 25% and the structural deformation accounted for 15%). The optimal parameter combination of the bolt state (corrosion area less than 5%, depth less than 10%, axial displacement less than 0.3 mm, structural deformation less than 1%) and the worst parameter combination (corrosion area more than 30%, depth more than 35%, axial displacement more than 3 mm, structural deformation more than 4%) were pre-set. The evaluation results of the bolt numbered "C045-089" showed that the Euclidean distance between the detection data of the bolt and the optimal parameter was 0.51, and the Euclidean distance with the worst parameter was 0.28, and the calculated relative closeness of the state was 0.65. Since the proximity exceeded the preset bolt status proximity threshold of 0.60 and was significantly different from the average proximity of 0.35 of the 10 adjacent bolts on the same cable, an early warning signal was automatically triggered, and the bolt was marked as "high risk" in the maintenance recommendation report, recommending that it be replaced as a priority during the next maintenance.

[0109] Further, the bolt detection data extracted through the cable bridge bolt maintenance picture includes a first bolt corrosion characteristic parameter and a first bolt displacement characteristic parameter, the first bolt corrosion characteristic parameter is characterized by bolt surface texture data, and the first bolt displacement characteristic parameter is characterized by a bolt axial displacement parameter and a bolt structure deformation parameter. The method includes:

[0110] The cable bridge bolt inspection picture is subjected to image preprocessing to obtain a first preprocessed image, wherein the image preprocessing includes image size calibration processing, image illumination correction processing and regional white balance processing; the bolt area in the first preprocessed image is extracted by the Otsu adaptive threshold segmentation method to obtain the gray value matrix G of the bolt area s ; Extract the edge contour line of the bolt area by Laplace edge detection operator, and obtain the coordinates of the bolt center positioning point and the bolt radius size by Hough circle transform;

[0111] The gray value matrix G of the bolt area s Perform wavelet decomposition to obtain multiple subband coefficient matrices including the low-frequency approximate coefficient matrix C a , horizontal detail coefficient matrix C h , vertical detail coefficient matrix C v and the diagonal detail coefficient matrix C d ; Calculate the texture feature vector T of the bolt area f for:

[0112]

[0113] Where Np is the total number of pixels in the bolt area; according to the texture feature vector T f and a preset corrosion texture vector threshold to calculate the first bolt corrosion characteristic parameter, including a bolt corrosion area ratio parameter and a bolt corrosion depth ratio parameter;

[0114] A polar coordinate system is established with the bolt center positioning point as the origin, and the edge contour line of the bolt area is discretized into a preset N e sampling points, record the polar coordinates of each sampling point [r i ,θ i ]; Construct the bolt edge profile feature vector E f for:

[0115]

[0116] where r 0 is the nominal radius of the bolt, θ 0 is the nominal angle of the sampling point, and are radial unit vectors and tangential unit vectors respectively; calculate the first bolt displacement characteristic parameter according to the bolt edge profile characteristic vector, including the bolt axial displacement parameter and the bolt structure deformation parameter; and normalize the first bolt corrosion characteristic parameter and the first bolt displacement characteristic parameter.

[0117] For example, for the inspection image of the bolt No. "C045-089" mentioned above, standardization preprocessing was first performed. The original 4000×3000 pixel image was uniformly scaled to 1000×750 pixels, and the uneven lighting area was compensated to make the overall brightness of the image balanced. In particular, the regional white balance technology was used for the shadow area on the edge of the bolt to make the outline of the bolt clearer.

[0118] The image was then divided into the bolt area and the background area using the Otsu adaptive threshold segmentation method. The grayscale value of the "C045-089" bolt area showed a clear bimodal distribution, with one peak corresponding to the intact bolt surface (grayscale value of about 210) and the other peak corresponding to the rusted area (grayscale value of about 120). The edge contour of the bolt was extracted using the Laplace operator, and the positions where the grayscale value change rate of the edge pixels exceeded 50% were marked as contour points. The circular contour of the bolt was identified using the Hough circle transform, and the bolt center coordinates (500,375) were determined, with a nominal radius of 50 pixels.

[0119] Wavelet decomposition was performed on the extracted bolt area, and four sub-bands were obtained, namely, low-frequency approximation coefficient, horizontal detail coefficient, vertical detail coefficient, and diagonal detail coefficient. The total number of regional pixels of bolt No. "C045-089" is 7854. The calculated texture feature vector shows that the mean value of low-frequency coefficient is 0.85 (intact area), the mean value of horizontal detail coefficient is 0.45 (rust boundary), the mean value of vertical detail coefficient is 0.42 (rust texture), and the mean value of diagonal detail coefficient is 0.38 (rust depth). Comparing these feature values ​​with the preset corrosion standard, the corrosion area ratio is 25% and the corrosion depth ratio is 30%. In the bolt edge profile analysis, the circumference is evenly divided into 72 sampling points, and the radial and tangential deviations of each sampling point relative to the nominal circle are recorded. The calculation results show that the displacement of the bolt in the axial direction is 2.8 mm and the structural deformation is 3%. Finally, these feature parameters are normalized and the numerical range is uniformly mapped to between 0 and 1, which is convenient for subsequent comprehensive evaluation.

[0120] Furthermore, the method of preprocessing the cable bridge bolt maintenance history image to obtain first maintenance data, clustering the second bolt corrosion characteristic parameter and the second bolt displacement characteristic parameter of the replaced bolt in the first maintenance data to obtain second maintenance data; establishing bolt replacement feature association rules through the second maintenance data, and obtaining bolt replacement feature strong association rules according to a preset feature association threshold includes:

[0121] The cable bridge bolt maintenance history picture is subjected to image preprocessing to obtain a second preprocessed image, wherein the image preprocessing includes image grayscale processing, Gaussian filter noise reduction processing and grayscale histogram equalization processing; a bolt area grayscale value matrix X of the second preprocessed image is obtained. g , the bolt area gray value matrix X is transformed into g Map to high-dimensional reproducing Hilbert space and calculate the kernel matrix K h for:

[0122]

[0123] where σ 1 is the preset Gaussian kernel function bandwidth parameter; the kernel matrix K h Perform eigenvalue decomposition to obtain the eigenvalue λ i and its corresponding eigenvector v i , select the eigenvectors corresponding to the preset first m maximum eigenvalues, project the second preprocessed image into the low-dimensional subspace spanned by the m eigenvectors to obtain the first maintenance data X 1 for:

[0124] X 1 =X g Vm ;

[0125] Where V m is a projection matrix composed of m eigenvectors; the first maintenance data X is clustered by DBSCAN density clustering algorithm 1 Perform cluster analysis based on the first maintenance data X 1 Calculate the Euclidean distance matrix D between samples e , for the first inspection data X 1 Each sample point p in i Calculate its ∈ d The number of sample points in the neighborhood N ∈ (p i ), where ∈ d is the preset bolt feature cluster radius parameter; when N ∈ (p i ) is greater than the preset bolt feature density threshold M p When the sample point p i Mark as core point, traverse all core sample points in turn, and compare each core sample point and its Euclidean distance less than ∈ d The sample points of the same bolt feature cluster are taken as the bolt feature cluster, and the bolt feature cluster is taken as the second maintenance data X 2 ;

[0126] The second maintenance data X 2 Discretization processing obtains the bolt feature item set I f , the bolt feature item set I is calculated by Apriori algorithm f The support S u for:

[0127]

[0128] Where A is the bolt feature item set I f subset, N(A) is the number of samples containing subset A, N t is the total number of samples; calculate the bolt feature item set I f The confidence level C f for:

[0129]

[0130] Where A and B are mutually exclusive subsets of bolt feature items; according to the preset feature association first threshold θ s and feature association second threshold θ c Filter to meet S u >θ s And C f >θ c The association rule of is used as the strong association rule of the bolt replacement feature;

[0131] The first threshold value θ of the feature association of the bolt feature item set s for:

[0132]

[0133] Where n f is the number of bolt failure samples in the historical cable bridge bolt maintenance, N t is the total number of samples in the historical cable bridge bolt maintenance; the first threshold value θ of the feature association of the bolt feature item set s for:

[0134]

[0135] Where n c is the number of samples in which bolt failure occurred during two consecutive inspections in the historical cable bridge bolt inspections, n d is the number of bolt failure samples in the historical cable bridge bolt maintenance.

[0136] For example, an in-depth analysis was conducted on 770,000 historical photos of bolt maintenance on the aforementioned bridge between 2019 and 2023. Each image was first converted into a 256-level grayscale image. Each image was denoised using a Gaussian filter, with the filter window size set to 5×5 pixels and a standard deviation of 1.5. The image contrast was enhanced through grayscale histogram equalization, making the subtle texture features on the bolt surface more clearly visible.

[0137] Specifically, taking a batch of 1,000 inspection photos of replaced bolts taken in 2022 as an example, after extracting the grayscale value matrix of the bolt area in each photo, the Gaussian kernel function is used for feature mapping. The bandwidth parameter of the Gaussian kernel function is set to 2.0, and the processed kernel matrix contains the main characteristic information of the surface state of the bolt. The kernel matrix is ​​decomposed by eigenvalue, and the top 5 eigenvalues ​​with the largest contribution rate and their corresponding eigenvectors are selected. These 5 eigenvalues ​​are 6.8, 5.2, 4.1, 3.3 and 2.7, respectively, which cumulatively explain 85% of the variability of the original data. The original image data is projected into the low-dimensional subspace formed by these 5 eigenvectors, which achieves data dimensionality reduction while retaining the key characteristic information of the surface state of the bolt.

[0138] The DBSCAN density clustering algorithm is used to analyze the data after dimensionality reduction. The bolt feature clustering radius parameter is set to 0.15 (this value is determined by cross-validation of 100 typical samples), and the density threshold is set to 8 sample points. In the clustering process, the Euclidean distance matrix between sample points is first calculated. Taking the "C045-089" bolt as an example, a total of 12 similar samples are found within its neighborhood range of 0.15, which exceeds the preset density threshold of 8, so it is marked as a core point. In this way, all sample points are traversed, and the data of 1000 pictures are finally divided into 4 feature clustering clusters, representing different bolt degradation modes. The feature data obtained by clustering are discretized to establish item sets containing features such as rust degree and displacement. The Apriori algorithm is used to analyze the association between feature item sets, and the support is calculated (taking the batch data of 2022 as an example, 45 samples have severe rust and displacement at the same time, and the support is 45 / 1000=0.045). Similarly, the confidence level is calculated (45 of the 60 samples with severe corrosion were displaced at the same time, and the confidence level is 45 / 60=0.75). Based on historical data statistics, the first threshold of feature association is set to 0.04 (derived from the proportion of 50 failed samples in the batch of 1,000 samples), and the second threshold of feature association is set to 0.70 (derived from the proportion of 35 consecutive failures in the 50 failed samples). Finally, the rules that meet the dual threshold conditions are selected as the basis for bolt replacement.

[0139] Furthermore, the method of evaluating the first bolt corrosion characteristic parameter and the first bolt displacement characteristic parameter according to the bolt replacement characteristic strong association rule to obtain a bolt state evaluation result includes:

[0140] The first bolt corrosion characteristic parameter is defined as the bolt corrosion area ratio r a and bolt corrosion depth ratio r d , the first bolt displacement characteristic parameter is defined as the bolt axial displacement d a and bolt structure deformation d s ; According to the strong association rule of the bolt replacement characteristics, the critical parameter vector of bolt failure is obtained in is the critical value of bolt corrosion area ratio, is the critical value of bolt corrosion depth ratio, is the critical value of the bolt axial displacement, is the critical value of bolt structure deformation;

[0141] Calculate the characteristic parameter vector X of the current detected bolt f =[r a ,r d ,d a ,d s ]T and the bolt failure critical parameter vector θ f The Mahalanobis distance D m :

[0142]

[0143] Where Σ is the covariance matrix of the eigenvalue vector; calculate the eigenvalue vector X f The failure probability P f :

[0144]

[0145] where α f is the preset failure probability coefficient; according to the failure probability P f Get the bolt status assessment results.

[0146] For example, the test results of the aforementioned "C045-089" bolts are analyzed in depth based on the evaluation criteria established above. First, the four key characteristic parameters of the bolt are clearly defined: bolt corrosion area ratio (current measured value 25%), bolt corrosion depth ratio (current measured value 30%), bolt axial displacement (current measured value 2.8 mm) and bolt structural deformation (current measured value 3%). Referring to the characteristic distribution law of failed samples in historical data, the critical parameter vector of bolt failure is determined: the critical value of the corrosion area ratio is 20% (based on the statistical mean of 2,100 simple corrosion failure samples), the critical value of the corrosion depth ratio is 25% (based on laboratory corrosion test results), the critical value of the axial displacement is 2.5 mm (based on the statistical mean of 1,500 simple displacement failure samples), and the critical value of the structural deformation is 2.5% (based on material mechanics calculation results).

[0147] When calculating the Mahalanobis distance for the inspection data of bolt No. "C045-089", the covariance matrix of the characteristic parameters is first constructed. By analyzing 12,500 failure samples in the historical data, the correlation between the characteristic parameters is obtained: the correlation coefficient between the corrosion area and the depth is 0.85 (indicating that the two are highly correlated), the correlation coefficient between the axial displacement and the structural deformation is 0.72 (indicating that there is a significant correlation), and the correlation coefficient between the corrosion feature and the displacement feature is low (both less than 0.3). Based on this covariance matrix, the Mahalanobis distance between the currently detected characteristic parameter vector [25%, 30%, 2.8mm, 3%] and the failure critical parameter vector [20%, 25%, 2.5mm, 2.5%] is calculated, and the distance value is 1.8.

[0148] Then, the failure probability coefficient was set to 0.5 (this value was determined by survival analysis of historical failure samples), and the Mahalanobis distance was substituted into the failure probability calculation formula. The calculation results for bolt No. "C045-089" showed that the failure probability of this bolt reached 0.78, which was significantly higher than the average failure probability of 0.25 of other bolts on the same cable. Further analysis found that the main reason for such a high failure probability was that all characteristic parameters of the bolt exceeded the corresponding critical values: the corrosion area exceeded 25% (exceeded the critical value by 5 percentage points), the corrosion depth exceeded 20% (exceeded the critical value by 5 percentage points), the axial displacement exceeded 0.3 mm (exceeded the critical value by 0.3 mm), and the structural deformation exceeded 0.5% (exceeded the critical value by 0.5 percentage points). Based on the failure probability assessment results, bolt No. "C045-089" was marked as "needing priority replacement" level.

[0149] Further, the bolt health monitoring index and bolt health monitoring weight based on the bolt corrosion characteristic parameter and the bolt displacement characteristic parameter are established through the bolt state evaluation result; the bolt health monitoring index and the bolt health monitoring weight are input into the TOPSIS evaluation model, and the Euclidean distance parameter between the bolt detection data and the optimal bolt state parameter and the worst bolt state parameter is calculated in combination with the preset optimal bolt state parameter and the worst bolt state parameter; the relative proximity of the bolt state is calculated according to the Euclidean distance parameter, and the method of issuing an early warning signal when the relative proximity of the bolt state is greater than the preset bolt state proximity threshold comprises:

[0150] The bolt state evaluation results are used to construct a bolt state evaluation matrix M according to the number of bolts. a , the bolt condition assessment matrix M a Contains a feature evaluation vector for each bolt, wherein the feature evaluation vector includes a first bolt corrosion feature parameter and a first bolt displacement feature parameter of the bolt; and calculates the bolt state evaluation matrix M by using the Euclidean norm. a The square sum of each characteristic parameter in each column is obtained, and each characteristic parameter is divided by the corresponding square sum to obtain the bolt state normalization matrix M b ; Normalize the bolt state matrix M b Each column characteristic parameter of is multiplied by the bolt health monitoring weight to obtain the bolt state weighted matrix M c ;

[0151] Through the second inspection data X 2 The failure parameter vector V of the TOPSIS evaluation model is calculated by using the second bolt corrosion characteristic parameter and the second bolt displacement characteristic parameter of the bolt failure sample in the bolt maintenance of the historical cable bridge f , through the second maintenance data X 2The health parameter vector V of the TOPSIS evaluation model is obtained by using the second bolt corrosion characteristic parameter and the second bolt displacement characteristic parameter of the bolt non-failure sample in the historical cable bridge bolt maintenance. h ; Calculate the bolt state weighted matrix M c The feature evaluation vector L for each bolt in i′ With the failure parameter vector V f and the health parameter vector V h The state distances are D i′f and D i′h for:

[0152]

[0153]

[0154] Where n f is the feature evaluation vector L i′ The number of elements in L i′ (j) is the j′th characteristic parameter of the i′th bolt, V f (j) is the failure parameter vector V f The jth ′ characteristic parameters; V h (j) is the health parameter vector V h The jth ′ characteristic parameters; calculate the i-th ′ The health status score of each bolt is S i′ for:

[0155]

[0156] Get the adjacent preset number N of the cable numbers corresponding to the bolt numbers s Health status score sequence of bolts where j ′ The value range is 1 to N s ; Through the second inspection data X 2 The health status score sequence of bolt failure samples in the bolt maintenance of historical cable bridges is used to calculate the mean health status score U s and standard deviation D s , get the bolt status warning threshold T s for:

[0157] T s =U s +B s D s ;

[0158] Among them B s is a preset status scoring coefficient; when the health status score S i′Greater than the bolt status warning threshold T s and is greater than the health status score sequence When the arithmetic mean of

[0159] Exemplarily, based on the aforementioned evaluation results of bolt No. "C045-089", a bolt status evaluation matrix is ​​constructed. The matrix contains the status data of 20 adjacent bolts (from C045-080 to C045-099), and each bolt records four characteristic parameters: corrosion area ratio, corrosion depth ratio, axial displacement, and structural deformation. The evaluation matrix is ​​normalized by the Euclidean norm, and the sum of squares of the characteristic parameters in each column is calculated: the corrosion area ratio is 158.3, the corrosion depth ratio is 142.7, the axial displacement is 35.6, and the structural deformation is 28.9. Each characteristic parameter is normalized by dividing it by the corresponding sum of squares, and then multiplied by the corresponding health monitoring weights (corrosion area 35%, corrosion depth 25%, axial displacement 25%, structural deformation 15%).

[0160] 12,500 failed bolt samples replaced between 2019 and 2023 were selected from the historical database, and their characteristic parameters were extracted to calculate the failure parameter vector [0.85, 0.82, 0.78, 0.75] of the TOPSIS model. At the same time, 650,000 bolt samples in good condition were selected from 770,000 records, and the health parameter vector [0.15, 0.18, 0.22, 0.25] was calculated. The characteristic evaluation vector [0.78, 0.76, 0.73, 0.71] of the "C045-089" bolt was calculated to have a state distance from the failure parameter vector and the health parameter vector, respectively, and the distance from the failure state was 0.12, and the distance from the health state was 0.65, and the health state score was calculated to be 0.84.

[0161] Comparative analysis found that in the health status score sequence of 12 adjacent bolts (C045-084 to C045-095) on the same cable, the scores of other bolts were between 0.2 and 0.4, while the score of 0.84 for bolt C045-089 significantly deviated from the group. By analyzing the failure samples in the historical data, the mean of the health status score was 0.75 and the standard deviation was 0.15. The status score coefficient was set to 1.5 (this value was determined by statistical analysis of thousands of failure cases), and the bolt status warning threshold was calculated to be 0.82. Since the health status score of bolt C045-089 was 0.84, which exceeded the warning threshold of 0.82 and was significantly higher than the arithmetic mean of 0.3 of the adjacent bolts, the warning signal was triggered, and it was recommended to replace the bolt first during the next maintenance.

[0162] Embodiment 2: Based on the same inventive concept, Figure 2As shown, this embodiment also provides a system for evaluating the health status of bolts based on cable bridge bolt inspection data, the system comprising:

[0163] A maintenance data acquisition module is used to obtain a cable bridge bolt maintenance picture and corresponding cable bridge bolt position data, wherein the cable bridge bolt position data includes a cable number and a serial number of the bolt corresponding to the cable number; bolt detection data extracted from the cable bridge bolt maintenance picture includes a first bolt corrosion characteristic parameter and a first bolt displacement characteristic parameter, wherein the first bolt corrosion characteristic parameter is characterized by bolt surface texture data, and the first bolt displacement characteristic parameter is characterized by a bolt axial displacement parameter and a bolt structure deformation parameter;

[0164] A historical data training module is used to pre-process the cable bridge bolt maintenance history pictures to obtain first maintenance data, and to cluster the second bolt corrosion characteristic parameters and the second bolt displacement characteristic parameters of the replaced bolts in the first maintenance data to obtain second maintenance data; to establish bolt replacement feature association rules through the second maintenance data, and to obtain bolt replacement feature strong association rules based on the preset feature association threshold, and to evaluate the first bolt corrosion characteristic parameters and the first bolt displacement characteristic parameters based on the bolt replacement feature strong association rules to obtain bolt status evaluation results;

[0165] The health scoring module is used to establish a bolt health monitoring index and a bolt health monitoring weight based on the bolt corrosion characteristic parameter and the bolt displacement characteristic parameter according to the bolt state evaluation result; input the bolt health monitoring index and the bolt health monitoring weight into the TOPSIS evaluation model, and calculate the Euclidean distance parameter between the bolt detection data and the optimal bolt state parameter and the worst bolt state parameter in combination with the preset optimal bolt state parameter and the worst bolt state parameter; calculate the relative proximity of the bolt state according to the Euclidean distance parameter, and issue a warning signal when the relative proximity of the bolt state is greater than a preset bolt state proximity threshold.

[0166] Furthermore, the system further comprises:

[0167] The maintenance data feature extraction module is used to perform image preprocessing on the cable bridge bolt maintenance picture to obtain a first preprocessed image, wherein the image preprocessing includes image size calibration processing, image illumination correction processing and regional white balance processing; the bolt area in the first preprocessed image is extracted by the Otsu adaptive threshold segmentation method to obtain the gray value matrix G of the bolt area s ; Extract the edge contour line of the bolt area by Laplace edge detection operator, and obtain the coordinates of the bolt center positioning point and the bolt radius size by Hough circle transform;

[0168] The gray value matrix G of the bolt areas Perform wavelet decomposition to obtain multiple subband coefficient matrices including the low-frequency approximate coefficient matrix C a , horizontal detail coefficient matrix C h , vertical detail coefficient matrix C v and the diagonal detail coefficient matrix C d ; Calculate the texture feature vector T of the bolt area f for:

[0169]

[0170] Where N p is the total number of pixels in the bolt area; according to the texture feature vector T f and a preset corrosion texture vector threshold to calculate the first bolt corrosion characteristic parameter, including a bolt corrosion area ratio parameter and a bolt corrosion depth ratio parameter;

[0171] A polar coordinate system is established with the bolt center positioning point as the origin, and the edge contour line of the bolt area is discretized into a preset N e sampling points, record the polar coordinates of each sampling point [r i ,θ i ]; Construct the bolt edge profile feature vector E d for:

[0172]

[0173] where r 0 is the nominal radius of the bolt, θ 0 is the nominal angle of the sampling point, and are radial unit vectors and tangential unit vectors respectively; calculate the first bolt displacement characteristic parameter according to the bolt edge profile characteristic vector, including the bolt axial displacement parameter and the bolt structure deformation parameter; and normalize the first bolt corrosion characteristic parameter and the first bolt displacement characteristic parameter.

[0174] Furthermore, the system further comprises:

[0175] The historical data feature mapping module is used to perform image preprocessing on the cable bridge bolt maintenance history image to obtain a second preprocessed image, wherein the image preprocessing includes image grayscale processing, Gaussian filter noise reduction processing and grayscale histogram equalization processing; obtain the bolt area grayscale value matrix X of the second preprocessed image g , the bolt area gray value matrix X is transformed into g Map to high-dimensional reproducing Hilbert space and calculate the kernel matrix K h for:

[0176]

[0177] where σ 1 is the preset Gaussian kernel function bandwidth parameter; the kernel matrix K h Perform eigenvalue decomposition to obtain the eigenvalue λ i and its corresponding eigenvector v i , select the eigenvectors corresponding to the preset first m maximum eigenvalues, project the second preprocessed image into the low-dimensional subspace spanned by the m eigenvectors to obtain the first maintenance data X 1 for:

[0178] X 1 =X g V m ;

[0179] Where V m is a projection matrix composed of m eigenvectors; the first maintenance data X is clustered by DBSCAN density clustering algorithm 1 Perform cluster analysis based on the first maintenance data X 1 Calculate the Euclidean distance matrix D between samples e , for the first inspection data X 1 Each sample point p in i Calculate its ∈ d The number of sample points in the neighborhood N ∈ (p i ), where ∈ d is the preset bolt feature cluster radius parameter; when N ∈ (p i ) is greater than the preset bolt feature density threshold M p When the sample point p i Mark as core point, traverse all core sample points in turn, and compare each core sample point and its Euclidean distance less than ∈ d The sample points of the same bolt feature cluster are taken as the bolt feature cluster, and the bolt feature cluster is taken as the second maintenance data X 2 ;

[0180] The second maintenance data X 2 Discretization processing obtains the bolt feature item set I f , the bolt feature item set I is calculated by Apriori algorithm f The support S u for:

[0181]

[0182] Where A is the bolt feature item set I f subset, N(A) is the number of samples containing subset A, N tis the total number of samples; calculate the bolt feature item set I f The confidence level C f for:

[0183]

[0184] Where A and B are mutually exclusive subsets of bolt feature items; according to the preset feature association first threshold θ s and feature association second threshold θ c Filter to meet S u >θ s And C f >θ c The association rule of is used as the strong association rule of the bolt replacement feature;

[0185] The first threshold value θ of the feature association of the bolt feature item set s for:

[0186]

[0187] Where n f is the number of bolt failure samples in the historical cable bridge bolt maintenance, N t is the total number of samples in the historical cable bridge bolt maintenance; the first threshold value θ of the feature association of the bolt feature item set s for:

[0188]

[0189] Where n c is the number of samples in which bolt failure occurred during two consecutive inspections in the historical cable bridge bolt inspections, n f is the number of bolt failure samples in the historical cable bridge bolt maintenance.

[0190] Furthermore, the system further comprises:

[0191] A feature matching evaluation module is used to define the first bolt corrosion feature parameter as a bolt corrosion area ratio r a and bolt corrosion depth ratio r d , the first bolt displacement characteristic parameter is defined as the bolt axial displacement d a and bolt structure deformation d s ; According to the strong association rule of the bolt replacement characteristics, the critical parameter vector of bolt failure is obtained in is the critical value of bolt corrosion area ratio, is the critical value of bolt corrosion depth ratio, is the critical value of the bolt axial displacement, is the critical value of bolt structure deformation;

[0192] Calculate the characteristic parameter vector X of the current detected bolt f =[r a ,r d ,d a ,d s ] T and the bolt failure critical parameter vector θ f The Mahalanobis distance D m :

[0193]

[0194] Where Σ is the covariance matrix of the eigenvalue vector; calculate the eigenvalue vector X f The failure probability P f :

[0195]

[0196] where α f is the preset failure probability coefficient; according to the failure probability P f Get the bolt status assessment results.

[0197] Furthermore, the system further comprises:

[0198] The bolt replacement warning module is used to construct a bolt status evaluation matrix M based on the bolt status evaluation results according to the number of bolts. a , the bolt condition assessment matrix M a Contains a feature evaluation vector for each bolt, wherein the feature evaluation vector includes a first bolt corrosion feature parameter and a first bolt displacement feature parameter of the bolt; and calculates the bolt state evaluation matrix M by using the Euclidean norm. a The square sum of each characteristic parameter in each column is obtained, and each characteristic parameter is divided by the corresponding square sum to obtain the bolt state normalization matrix M b ; Normalize the bolt state matrix M b Each column characteristic parameter of is multiplied by the bolt health monitoring weight to obtain the bolt state weighted matrix M c ;

[0199] Through the second inspection data X 2 The failure parameter vector V of the TOPSIS evaluation model is calculated by using the second bolt corrosion characteristic parameter and the second bolt displacement characteristic parameter of the bolt failure sample in the bolt maintenance of the historical cable bridge f , through the second maintenance data X 2 The health parameter vector V of the TOPSIS evaluation model is obtained by using the second bolt corrosion characteristic parameter and the second bolt displacement characteristic parameter of the bolt non-failure sample in the historical cable bridge bolt maintenance. h; Calculate the bolt state weighted matrix M c The feature evaluation vector L for each bolt in i′ With the failure parameter vector V f and the health parameter vector V h The state distances are D i′f and D i′h for:

[0200]

[0201] Where n f is the feature evaluation vector L i′ The number of elements in L i′ (j) is the j′th characteristic parameter of the i′th bolt, V f (j) is the failure parameter vector V f The jth ′ characteristic parameters; V h (j) is the health parameter vector V h The jth ′ characteristic parameters; calculate the i-th ′ The health status score of each bolt is S i′ for:

[0202]

[0203] Get the adjacent preset number N of the cable numbers corresponding to the bolt numbers s Health status score sequence of bolts where j ′ The value range is 1 to N s ; Through the second inspection data X 2 The health status score sequence of bolt failure samples in the bolt maintenance of historical cable bridges is used to calculate the mean health status score U s and standard deviation D s , get the bolt status warning threshold T s for:

[0204] T s =U s +B s D s ;

[0205] Among them B s is a preset status scoring coefficient; when the health status score S i′ Greater than the bolt status warning threshold T s and is greater than the health status score sequence When the arithmetic mean of

[0206] It should be noted that, regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0207] Finally, it should be noted that: Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for evaluating the health status of bolts based on cable bridge bolt maintenance data, characterized in that: The method comprises: Obtain a cable bridge bolt maintenance picture and corresponding cable bridge bolt position data, wherein the cable bridge bolt position data includes a cable number and a serial number of the bolt corresponding to the cable number; extract bolt detection data through the cable bridge bolt maintenance picture, including a first bolt corrosion characteristic parameter and a first bolt displacement characteristic parameter, wherein the first bolt corrosion characteristic parameter is characterized by bolt surface texture data, and the first bolt displacement characteristic parameter is characterized by a bolt axial displacement parameter and a bolt structure deformation parameter; Preprocess the cable bridge bolt maintenance history image to obtain the first maintenance data, perform cluster analysis on the second bolt corrosion characteristic parameter and the second bolt displacement characteristic parameter of the replaced bolt in the first maintenance data to obtain the second maintenance data; establish a bolt replacement feature association rule through the second maintenance data, and obtain a bolt replacement feature strong association rule based on a preset feature association threshold, and evaluate the first bolt corrosion characteristic parameter and the first bolt displacement characteristic parameter based on the bolt replacement feature strong association rule to obtain a bolt status evaluation result; The bolt health monitoring index and the bolt health monitoring weight are established based on the bolt corrosion characteristic parameters and the bolt displacement characteristic parameters through the bolt state assessment result; the bolt health monitoring index and the bolt health monitoring weight are input into the TOPSIS assessment model, and the Euclidean distance parameters between the bolt detection data and the optimal parameters of the bolt state and the worst parameters of the bolt state are calculated in combination with the preset optimal parameters of the bolt state and the worst parameters of the bolt state; the relative proximity of the bolt state is calculated according to the Euclidean distance parameters, and a warning signal is issued when the relative proximity of the bolt state is greater than a preset bolt state proximity threshold.

2. The method for evaluating the health status of bolts based on cable bridge bolt maintenance data according to claim 1 is characterized in that: The method of extracting bolt detection data from the cable bridge bolt maintenance picture includes a first bolt corrosion characteristic parameter and a first bolt displacement characteristic parameter, wherein the first bolt corrosion characteristic parameter is characterized by bolt surface texture data, and the first bolt displacement characteristic parameter is characterized by a bolt axial displacement parameter and a bolt structure deformation parameter, including: The cable bridge bolt inspection picture is subjected to image preprocessing to obtain a first preprocessed image, wherein the image preprocessing includes image size calibration processing, image illumination correction processing and regional white balance processing; the bolt area in the first preprocessed image is extracted by the Otsu adaptive threshold segmentation method to obtain the gray value matrix G of the bolt area s ; Extract the edge contour line of the bolt area by Laplace edge detection operator, and obtain the coordinates of the bolt center positioning point and the bolt radius size by Hough circle transform; The gray value matrix G of the bolt area s Perform wavelet decomposition to obtain multiple subband coefficient matrices including the low-frequency approximate coefficient matrix C a , horizontal detail coefficient matrix C h , vertical detail coefficient matrix C v and the diagonal detail coefficient matrix C d ; Calculate the texture feature vector T of the bolt area f for: Where N p is the total number of pixels in the bolt area; according to the texture feature vector T f and a preset corrosion texture vector threshold to calculate the first bolt corrosion characteristic parameter, including a bolt corrosion area ratio parameter and a bolt corrosion depth ratio parameter; A polar coordinate system is established with the bolt center positioning point as the origin, and the edge contour line of the bolt area is discretized into a preset N e sampling points, record the polar coordinates of each sampling point [r i ,θ i ]; Construct the bolt edge profile feature vector E f for: Where r0 is the nominal radius of the bolt, θ0 is the nominal angle of the sampling point, and are radial unit vectors and tangential unit vectors respectively; calculate the first bolt displacement characteristic parameter according to the bolt edge profile characteristic vector, including the bolt axial displacement parameter and the bolt structure deformation parameter; and normalize the first bolt corrosion characteristic parameter and the first bolt displacement characteristic parameter.

3. The method for evaluating the health status of bolts based on cable bridge bolt maintenance data according to claim 2 is characterized in that: The method of preprocessing the cable bridge bolt maintenance history image to obtain the first maintenance data, clustering the second bolt corrosion characteristic parameter and the second bolt displacement characteristic parameter of the replaced bolt in the first maintenance data to obtain the second maintenance data; establishing the bolt replacement feature association rule through the second maintenance data, and screening according to the preset feature association threshold to obtain the bolt replacement feature strong association rule includes: The cable bridge bolt maintenance history picture is subjected to image preprocessing to obtain a second preprocessed image, wherein the image preprocessing includes image grayscale processing, Gaussian filter noise reduction processing and grayscale histogram equalization processing; a bolt area grayscale value matrix X of the second preprocessed image is obtained. g , the bolt area gray value matrix X is transformed into g Map to high-dimensional reproducing Hilbert space and calculate the kernel matrix K h for: in σ 1 is the preset Gaussian kernel function bandwidth parameter; the kernel matrix K h Perform eigenvalue decomposition to obtain the eigenvalue λ i and its corresponding eigenvector v i , select the eigenvectors corresponding to the preset first m maximum eigenvalues, project the second preprocessed image to the low-dimensional subspace spanned by the m eigenvectors to obtain the first maintenance data X1 as follows: X1=X g V m ; Where V m is a projection matrix composed of m eigenvectors; cluster analysis is performed on the first maintenance data X1 using the DBSCAN density clustering algorithm, and the Euclidean distance matrix D between samples in the first maintenance data X1 is calculated based on the first maintenance data X1 e , for each sample point p in the first maintenance data X1 i Calculate its ∈ d The number of sample points in the neighborhood N ∈ (p i ), where ∈ d is the preset bolt feature cluster radius parameter; when N ∈ (p i ) is greater than the preset bolt feature density threshold M p When the sample point p i Mark as core point, traverse all core sample points in turn, and compare each core sample point and its Euclidean distance less than ∈ d The sample points of are taken as the same bolt feature cluster, and the bolt feature cluster is taken as the second maintenance data X2; The second maintenance data X2 is discretized to obtain the bolt feature item set I f , the bolt feature item set I is calculated by Apriori algorithm f The support S u for: Where A is the bolt feature item set I f subset, N(A) is the number of samples containing subset A, N t is the total sample size; calculate the bolt feature item set I f The confidence level C f for: Where A and B are mutually exclusive subsets of bolt feature items; according to the preset feature association first threshold θ s and feature association second threshold θ c Filter to meet S u >θ s And C f >θ c The association rule of is used as the strong association rule of the bolt replacement feature; The first threshold value θ of the feature association of the bolt feature item set s for: Where n f is the number of bolt failure samples in the historical cable bridge bolt maintenance, N t is the total number of samples in the historical cable bridge bolt maintenance; the first threshold value θ of the feature association of the bolt feature item set s for: Where n c is the number of samples in which bolt failure occurred during two consecutive inspections in the historical cable bridge bolt inspections, n f is the number of bolt failure samples in the historical cable bridge bolt maintenance.

4. The method for evaluating the health status of bolts based on cable bridge bolt maintenance data according to claim 3 is characterized in that: The method for evaluating the first bolt corrosion characteristic parameter and the first bolt displacement characteristic parameter according to the bolt replacement characteristic strong association rule to obtain the bolt state evaluation result includes: The first bolt corrosion characteristic parameter is defined as the bolt corrosion area ratio r a and bolt corrosion depth ratio r d , the first bolt displacement characteristic parameter is defined as the bolt axial displacement d a and bolt structure deformation d s ; According to the strong association rule of the bolt replacement characteristics, the critical parameter vector of bolt failure is obtained in is the critical value of bolt corrosion area ratio, is the critical value of bolt corrosion depth ratio, is the critical value of the bolt axial displacement, is the critical value of bolt structure deformation; Calculate the characteristic parameter vector X of the current detected bolt f =[r a , r d , d a , d s ] T and the bolt failure critical parameter vector θ f The Mahalanobis distance D m : Where Σ is the covariance matrix of the eigenvalue vector; calculate the eigenvalue vector X f The failure probability P f : where α f is the preset failure probability coefficient; according to the failure probability P f Get the bolt status assessment results.

5. The method for evaluating the health status of bolts based on cable bridge bolt maintenance data according to claim 4 is characterized in that: The method of establishing a bolt health monitoring index and a bolt health monitoring weight based on the bolt corrosion characteristic parameter and the bolt displacement characteristic parameter through the bolt state evaluation result; inputting the bolt health monitoring index and the bolt health monitoring weight into the TOPSIS evaluation model, and calculating the Euclidean distance parameter between the bolt detection data and the optimal bolt state parameter and the worst bolt state parameter in combination with the preset optimal bolt state parameter and the worst bolt state parameter; calculating the relative proximity of the bolt state according to the Euclidean distance parameter, and issuing an early warning signal when the relative proximity of the bolt state is greater than a preset bolt state proximity threshold comprises: The bolt state evaluation results are used to construct a bolt state evaluation matrix M according to the number of bolts. a , the bolt condition assessment matrix M a Contains a feature evaluation vector for each bolt, wherein the feature evaluation vector includes a first bolt corrosion feature parameter and a first bolt displacement feature parameter of the bolt; and calculates the bolt state evaluation matrix M by using the Euclidean norm. a The square sum of each characteristic parameter in each column is obtained, and each characteristic parameter is divided by the corresponding square sum to obtain the bolt state normalization matrix M b ; Normalize the bolt state matrix M b Each column characteristic parameter of is multiplied by the bolt health monitoring weight to obtain the bolt state weighted matrix M c ; The failure parameter vector V of the TOPSIS evaluation model is calculated by using the second bolt corrosion characteristic parameter and the second bolt displacement characteristic parameter of the bolt failure sample in the historical cable bridge bolt maintenance in the second maintenance data X2. f The health parameter vector V of the TOPSIS evaluation model is obtained by using the second bolt corrosion characteristic parameter and the second bolt displacement characteristic parameter of the bolt non-failure sample in the historical cable bridge bolt maintenance in the second maintenance data X2. h ; Calculate the bolt state weighted matrix M c The feature evaluation vector L for each bolt in i′ With the failure parameter vector V f and the health parameter vector V h The state distances are D i′f and D i′h for: Where n f is the feature evaluation vector L i′ The number of elements in L i′ (j) is the j′th characteristic parameter of the i′th bolt, V f (j) is the failure parameter vector V f The j′th characteristic parameter of h (j) is the health parameter vector V h The j′th characteristic parameter of the bolt is calculated; the health status score S of the i′th bolt is calculated. i′ for: Get the adjacent preset number N of the cable numbers corresponding to the bolt numbers s Health status score sequence of bolts The value of j′ ranges from 1 to N. s ; Calculate the mean health status score U through the health status score sequence of bolt failure samples in the historical cable bridge bolt maintenance in the second maintenance data X2 s and standard deviation D s , get the bolt status warning threshold T s for: T s =U s +B s D s ; Among them B s is a preset status scoring coefficient; when the health status score S i′ Greater than the bolt status warning threshold T s and is greater than the health status score sequence When the arithmetic mean of 6. A system for evaluating the health status of bolts based on cable bridge bolt inspection data, characterized in that: The system comprises: A maintenance data acquisition module is used to obtain a cable bridge bolt maintenance picture and corresponding cable bridge bolt position data, wherein the cable bridge bolt position data includes a cable number and a serial number of the bolt corresponding to the cable number; bolt detection data extracted from the cable bridge bolt maintenance picture includes a first bolt corrosion characteristic parameter and a first bolt displacement characteristic parameter, wherein the first bolt corrosion characteristic parameter is characterized by bolt surface texture data, and the first bolt displacement characteristic parameter is characterized by a bolt axial displacement parameter and a bolt structure deformation parameter; A historical data training module is used to pre-process the cable bridge bolt maintenance history pictures to obtain first maintenance data, and to cluster the second bolt corrosion characteristic parameters and the second bolt displacement characteristic parameters of the replaced bolts in the first maintenance data to obtain second maintenance data; to establish bolt replacement feature association rules through the second maintenance data, and to obtain bolt replacement feature strong association rules based on the preset feature association threshold, and to evaluate the first bolt corrosion characteristic parameters and the first bolt displacement characteristic parameters based on the bolt replacement feature strong association rules to obtain bolt status evaluation results; The health scoring module is used to establish a bolt health monitoring index and a bolt health monitoring weight based on the bolt corrosion characteristic parameter and the bolt displacement characteristic parameter according to the bolt state evaluation result; input the bolt health monitoring index and the bolt health monitoring weight into the TOPSIS evaluation model, and calculate the Euclidean distance parameter between the bolt detection data and the optimal bolt state parameter and the worst bolt state parameter in combination with the preset optimal bolt state parameter and the worst bolt state parameter; calculate the relative proximity of the bolt state according to the Euclidean distance parameter, and issue a warning signal when the relative proximity of the bolt state is greater than a preset bolt state proximity threshold.

7. The system for evaluating the health status of bolts based on cable bridge bolt inspection data according to claim 6, characterized in that: The system further comprises: The maintenance data feature extraction module is used to perform image preprocessing on the cable bridge bolt maintenance picture to obtain a first preprocessed image, wherein the image preprocessing includes image size calibration processing, image illumination correction processing and regional white balance processing; the bolt area in the first preprocessed image is extracted by the Otsu adaptive threshold segmentation method to obtain the gray value matrix G of the bolt area s ; Extract the edge contour line of the bolt area by Laplace edge detection operator, and obtain the coordinates of the bolt center positioning point and the bolt radius size by Hough circle transform; The gray value matrix G of the bolt area s Perform wavelet decomposition to obtain multiple subband coefficient matrices including the low-frequency approximate coefficient matrix C a , horizontal detail coefficient matrix C h , vertical detail coefficient matrix C v and the diagonal detail coefficient matrix C d ; Calculate the texture feature vector T of the bolt area f for: Where N p is the total number of pixels in the bolt area; according to the texture feature vector T f and a preset corrosion texture vector threshold to calculate the first bolt corrosion characteristic parameter, including a bolt corrosion area ratio parameter and a bolt corrosion depth ratio parameter; A polar coordinate system is established with the bolt center positioning point as the origin, and the edge contour line of the bolt area is discretized into a preset N e sampling points, record the polar coordinates of each sampling point [r i ,θ i ]; Construct the bolt edge profile feature vector E f for: Where r0 is the nominal radius of the bolt, θ0 is the nominal angle of the sampling point, and are radial unit vectors and tangential unit vectors respectively; calculate the first bolt displacement characteristic parameter according to the bolt edge profile characteristic vector, including the bolt axial displacement parameter and the bolt structure deformation parameter; and normalize the first bolt corrosion characteristic parameter and the first bolt displacement characteristic parameter.

8. The system for evaluating the health status of bolts based on cable bridge bolt maintenance data according to claim 7, characterized in that: The system further comprises: The historical data feature mapping module is used to perform image preprocessing on the cable bridge bolt maintenance history image to obtain a second preprocessed image, wherein the image preprocessing includes image grayscale processing, Gaussian filter noise reduction processing and grayscale histogram equalization processing; obtain the bolt area grayscale value matrix X of the second preprocessed image g , the bolt area gray value matrix X is transformed into g Map to high-dimensional reproducing Hilbert space and calculate the kernel matrix K h for: Where σ1 is the preset Gaussian kernel function bandwidth parameter; the kernel matrix K h Perform eigenvalue decomposition to obtain the eigenvalue λ i and its corresponding eigenvector v i , select the eigenvectors corresponding to the preset first m maximum eigenvalues, project the second preprocessed image to the low-dimensional subspace spanned by the m eigenvectors to obtain the first maintenance data X1 as follows: X1=X g V m ; Where V m is a projection matrix composed of m eigenvectors; cluster analysis is performed on the first maintenance data X1 using the DBSCAN density clustering algorithm, and the Euclidean distance matrix D between samples in the first maintenance data X1 is calculated based on the first maintenance data X1 e , for each sample point p in the first maintenance data X1 i Calculate its ∈ d The number of sample points in the neighborhood N ∈ (p i ), where ∈ d is the preset bolt feature cluster radius parameter; when N ∈ (p i ) is greater than the preset bolt feature density threshold M p When the sample point p i Mark as core point, traverse all core sample points in turn, and compare each core sample point and its Euclidean distance less than ∈ d The sample points of are taken as the same bolt feature cluster, and the bolt feature cluster is taken as the second maintenance data X2; The second maintenance data X2 is discretized to obtain the bolt feature item set I f , the bolt feature item set I is calculated by Apriori algorithm f The support S u for: Where A is the bolt feature item set I f subset, N(A) is the number of samples containing subset A, N t is the total sample size; calculate the bolt feature item set I f The confidence level C f for: Where A and B are mutually exclusive subsets of bolt feature items; according to the preset feature association first threshold θ s and feature association second threshold θ c Filter to meet S u >θ s And C f >θ c The association rule of is used as the strong association rule of the bolt replacement feature; The first threshold value θ of the feature association of the bolt feature item set s for: Where n f is the number of bolt failure samples in the historical cable bridge bolt maintenance, N t is the total number of samples in the historical cable bridge bolt maintenance; the first threshold value θ of the feature association of the bolt feature item set s for: Where n c is the number of samples in which bolt failure occurred during two consecutive inspections in the historical cable bridge bolt inspections, n f is the number of bolt failure samples in the historical cable bridge bolt maintenance.

9. The system for evaluating the health status of bolts based on cable bridge bolt inspection data according to claim 8, characterized in that: The system further comprises: A feature matching evaluation module is used to define the first bolt corrosion feature parameter as a bolt corrosion area ratio r a and bolt corrosion depth ratio r d , the first bolt displacement characteristic parameter is defined as the bolt axial displacement d a and bolt structure deformation d s ; According to the strong association rule of the bolt replacement characteristics, the critical parameter vector of bolt failure is obtained in is the critical value of bolt corrosion area ratio, is the critical value of bolt corrosion depth ratio, is the critical value of the bolt axial displacement, is the critical value of bolt structure deformation; Calculate the characteristic parameter vector X of the current detected bolt f =[r a , r d , d a , d s ] T and the bolt failure critical parameter vector θ f The Mahalanobis distance D m : Where Σ is the covariance matrix of the eigenvalue vector; calculate the eigenvalue vector X f The failure probability P f : where α f is the preset failure probability coefficient; according to the failure probability P f Get the bolt status assessment results.

10. The system for evaluating the health status of bolts based on cable bridge bolt inspection data according to claim 9, characterized in that: The system further comprises: The bolt replacement warning module is used to construct a bolt status evaluation matrix M based on the bolt status evaluation results according to the number of bolts. a , the bolt condition assessment matrix M a Contains a feature evaluation vector for each bolt, wherein the feature evaluation vector includes a first bolt corrosion feature parameter and a first bolt displacement feature parameter of the bolt; and calculates the bolt state evaluation matrix M by using the Euclidean norm. a The square sum of each characteristic parameter in each column is obtained, and each characteristic parameter is divided by the corresponding square sum to obtain the bolt state normalization matrix M b ; Normalize the bolt state matrix M b Each column characteristic parameter of is multiplied by the bolt health monitoring weight to obtain the bolt state weighted matrix M c ; The failure parameter vector V of the TOPSIS evaluation model is calculated by using the second bolt corrosion characteristic parameter and the second bolt displacement characteristic parameter of the bolt failure sample in the historical cable bridge bolt maintenance in the second maintenance data X2. f The health parameter vector V of the TOPSIS evaluation model is obtained by using the second bolt corrosion characteristic parameter and the second bolt displacement characteristic parameter of the bolt non-failure sample in the historical cable bridge bolt maintenance in the second maintenance data X2. h ; Calculate the bolt state weighted matrix M c The feature evaluation vector L for each bolt in i′ With the failure parameter vector V f and the health parameter vector V h The state distances are D i′f and D i′h for: Where n f is the feature evaluation vector L i′ The number of elements in L i′ (j) is the j′th characteristic parameter of the i′th bolt, V f (j) is the failure parameter vector V f The j′th characteristic parameter of h (j) is the health parameter vector V h The j′th characteristic parameter of the bolt is calculated; the health status score S of the i′th bolt is calculated. i′ for: Get the adjacent preset number N of the cable numbers corresponding to the bolt numbers s Health status score sequence of bolts The value of j′ ranges from 1 to N. s ; Calculate the mean health status score U through the health status score sequence of bolt failure samples in the historical cable bridge bolt maintenance in the second maintenance data X2 s and standard deviation D s , get the bolt status warning threshold T s for: T s =U s +B s D s ; Among them B s is a preset status scoring coefficient; when the health status score S i′ Greater than the bolt status warning threshold T s and is greater than the health status score sequence When the arithmetic mean of

Images