Nondestructive detection method for missing nuts and bolts of transmission tower foundation bolts based on INGO-SVM

Through the support vector machine model based on the improved Northern Goshawk optimization algorithm optimization algorithm optimization problem, the problem of nut missing in electromagnetic non-destructive detection in complex environments is solved, and the accuracy of the loss of bolts and nuts of the transmission tower is achieved, which improves the detection accuracy and resistance Interference capability.

CN117407775BActive Publication Date: 2024-12-06CHINA THREE GORGES UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202311189067.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-09-15
Publication Date
2024-12-06
Estimated Expiration
2043-09-15

AI Technical Summary

Technical Problem

The existing electromagnetic induction non-destructive detection methods cannot effectively detect the missing bolts and nuts of power transmission towers in complex environments, especially when ferromagnetic substances are disturbed greatly, it is impossible to accurately analyze the missing type of nuts.

Method used

Using the Support Vector Machine (SVM) classification detection method (INGO-SVM) based on improved Northern Goshawk optimization algorithm (INGO) optimization, data were obtained through electromagnetic non-destructive detection experiments, and using the Northern Goshawk optimization algorithm, Cubic chaos mapping and small hole imaging Reverse learning strategy optimizes the model to improve the accuracy of detection.

Benefits of technology

Improve the accuracy of the evaluation of the missing grade of the anchor bolt and nut of the transmission tower, reduce the impact of ferromagnetic interference, improve the accuracy of electromagnetic induction non-destructive detection, and can effectively map the screw diameter, number of nuts, thickness of the pad plate and magnetic field strength. The nonlinear relationship provides a feasible detection method.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117407775B_ABST
    Figure CN117407775B_ABST
Patent Text Reader

Abstract

The nondestructive detection method for the missing of the anchor bolts and nuts of the transmission tower based on INGO-SVM is characterized by comprising the following steps: S1. Performing electromagnetic nondestructive testing experiments to obtain a data set for evaluating the missing defects of the anchor bolts and nuts of the transmission tower; S2. Using the Northern Goshawk optimization algorithm to process the data; S3. Using Cubic chaotic mapping and pinhole imaging reverse learning strategy to optimize the Northern Goshawk optimization algorithm to obtain an improved Northern Goshawk optimization algorithm; S4. Using the improved Northern Goshawk optimization algorithm to optimize the support vector machine to obtain an improved Northern Goshawk optimization algorithm to optimize the INGO-SVM model; S5. Importing the specific parameters of the anchor bolts of the transmission tower to be inspected into the INGO-SVM model to evaluate the missing level of the anchor bolts and nuts of the tower. The method uses the high-dimensional mapping of SVM to map the nonlinearly related input and output quantities in the low dimension to the high dimension for classification, thereby improving the accuracy of the missing level evaluation of the anchor bolts and nuts of the transmission tower.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of safety detection of power transmission towers, and in particular to an INGO-SVM-based nondestructive detection method for missing anchor bolts and nuts of power transmission towers. Background Art

[0002] Transmission towers are important carriers of electric energy transmission, and ensuring their safe operation is a key link in ensuring safe transmission of electric energy. The anchor bolts used to fix the foundation and tower legs of the transmission tower are affected by various uncertain factors such as construction workers during the construction process, which may lead to the possibility of missing nuts. This defect makes the connection between the tower and the foundation loose, directly affecting the bearing capacity of the transmission tower, and leaving huge potential risks for the safe operation of the entire transmission line. Therefore, the defect of missing nuts in the anchor bolts of lossless transmission towers is of great significance to ensuring the safe transmission of electric energy.

[0003] In the non-destructive testing of ferromagnetic materials wrapped in concrete, electromagnetic induction non-destructive testing is currently mainly used at home and abroad. Among them, Yu Jin et al. introduced an electromagnetic non-destructive testing method for steel bar positioning and burial depth based on the magnetization effect of steel bars. This method detects the position and burial depth of a single steel bar without the interference of ferromagnetic materials through the reasonable design of the position of the Hall sensor. However, this method is not suitable for electromagnetic non-destructive testing in complex environments. Due to the mutual interference of magnetic field intensity signals, the signals will show a nonlinear relationship, and it is impossible to achieve the ideal detection effect through simple signal processing; Ran Chongwei et al. proposed a method based on magnetic coupling resonant eddy current detection for the detection of the position of a single steel bar in concrete. This method uses the detection characteristics of magnetic coupling resonant eddy current, which has greatly improved the detection distance of electromagnetic non-destructive testing. Similarly, when facing the detection object with large interference of magnetic field signal such as missing anchor bolts and nuts, it is still impossible to directly analyze the missing type of nuts through the detection results. In order to solve the above problems, the present invention proposes a support vector machine (SVM) classification detection method (INGO-SVM) based on improved Northern Goshawk Optimization (INGO) algorithm optimization, aiming at the electromagnetic nondestructive detection problem of missing nuts on transmission tower anchor bolts. Summary of the invention

[0004] The technical problem to be solved by the present invention is: to solve the problems existing in the above-mentioned background technology, and to provide a nondestructive detection method for missing nuts and bolts of transmission tower anchor bolts based on INGO-SVM. This method uses the high-dimensional mapping of SVM to map the nonlinearly related input and output quantities in low dimensions to high dimensions for classification, thereby improving the accuracy of the evaluation of the missing level of nuts and bolts of transmission tower anchor bolts.

[0005] In order to realize the above-mentioned technical features, the purpose of the present invention is achieved as follows:

[0006] The nondestructive detection method for missing anchor bolts and nuts of power transmission tower based on INGO-SVM is characterized by comprising the following steps:

[0007] S1. Conduct electromagnetic nondestructive testing experiments to obtain a data set for evaluating the missing defects of the anchor bolts and nuts of transmission towers;

[0008] S2. Process the data using the Northern Goshawk optimization algorithm;

[0009] S3. The optimization algorithm of northern goshawk is optimized by using Cubic chaotic mapping and pinhole imaging reverse learning strategy to obtain an improved northern goshawk optimization algorithm;

[0010] S4. Optimizing the support vector machine using the improved northern goshawk optimization algorithm to obtain an improved northern goshawk optimization algorithm to optimize the support INGO-SVM model;

[0011] S5. Import the specific parameters of the transmission tower anchor bolts to be inspected into the INGO-SVM model to evaluate the level of missing nuts of the tower anchor bolts.

[0012] In S1, an electromagnetic nondestructive testing experiment for missing nuts in the anchor bolts of the transmission tower was carried out. By changing the screw diameter, protective layer thickness, pad thickness and number of nuts of the anchor bolts in the concrete protective cap, the magnetic field strength signal under the change of four factors was obtained. At the same time, the defect of missing nuts was divided into no missing nuts, one missing nut and two missing nuts, which were recorded as three levels of 0, 1 and 2. The above experimental results were combined with the evaluation criteria, and the normalized data was used as the analysis sample of the algorithm model.

[0013] In S2, the northern goshawk population is initialized, and the fitness function is set. The equation of the fitness function is set as follows:

[0014]

[0015] In formula (1), W represents the transmission tower anchor bolt detection dataset; Si represents the actual value of the magnetic field intensity of the sample; Represents the training value of the magnetic field strength of the sample.

[0016] In S2, the optimization algorithm processing of the northern goshawk is divided into two steps, namely, exploration and identification of target prey and pursuit and escape of target prey. In the exploration and identification of target prey stage, the mathematical model of this stage is as follows:

[0017]

[0018]

[0019]

[0020] In formulas (2), (3), and (4): Pi represents the position of the target prey pursued by the i-th goshawk; represents the objective function value at the corresponding position; k is a random natural number in the range [1, N] and is not i; represents the new state of the i-th goshawk, where represents its new state in the jth dimension, F i new,p1 is the objective function value corresponding to its latest state; r takes a random number in the range of [0,1], and I takes 1 or 2, both of which are used to generate random numbers for the random behavior of the northern goshawk optimization algorithm during search and update;

[0021] In the stage of pursuit and escape of the target prey, assuming that the hunting radius is R, the mathematical model of this stage is as follows:

[0022]

[0023]

[0024]

[0025] In formula (5), (6) and (7): represents the new state of the i-th goshawk after the second stage update, where represents its new state in the jth dimension, F i new,p2 is the objective function value corresponding to its latest state; t represents the current number of iterations, and T represents the maximum number of iterations.

[0026] In S3, Cubic chaos mapping is added to the Northern Goshawk optimization algorithm, and its mathematical model is as follows:

[0027] x n+1= ρx n (1-x n 2 ) (8)

[0028] In formula (8): xn∈(0,1), ρ represents the control parameter, which is used to control the chaotic properties and nonlinearity;

[0029] After a certain number of iterations, the fitness value will remain fixed, and the pinhole imaging reverse learning strategy is introduced to optimize the population. In the northern goshawk optimization algorithm, it is assumed that the position change interval of each goshawk is [ai,bi], and the horizontal coordinate of the flame with a height of h is xi,j. Through pinhole imaging, the flame will produce a reflection with a height of H at the position of the horizontal coordinate xi,j*. According to the triangle similarity principle, it can be obtained:

[0030]

[0031] Let m = h / H represent the scaling factor, then the expression of xi,j* is:

[0032]

[0033] From equations (9) and (10), we can see that as the scaling factor m decreases, the difference between the positive solution and the inverse solution becomes larger. When m = 1, the equation degenerates into a general reverse learning strategy.

[0034] In S4, through S2 and S3, the population of northern goshawks is updated, the fitness value of each northern goshawk and the corresponding optimal position are calculated, and the algorithm is run to the maximum number of iterations to stop updating, and the minimum fitness value and the corresponding optimal position are assigned to the penalty parameter c and kernel function parameter g in the support vector machine; the goal of the support vector machine is to find an optimal hyperplane to distinguish two or more categories, and the hyperplane optimization problem model is as follows:

[0035]

[0036] st y i (ωx i +b)≥1-ε i (11)

[0037] In formula (11), ω is the normal vector of the hyperplane; b is the offset distance of the hyperplane; ε is the introduced slack variable used to relax the interval distance limit of the function; min represents the point closest to the classification line; c is the introduced penalty factor used to weigh the trade-off between the function classification accuracy and the function interval distance limit;

[0038] Another important part of the support vector machine is the kernel function and its kernel function parameters. The radial basis is selected as the kernel function, and its expression is as follows:

[0039]

[0040] In formula (12), x1 is the RBF center; g is the kernel function parameter, which indicates the coverage of the used restriction function; K() indicates the kernel function; K(x1,x2) indicates the maximum position from the center point x1;

[0041] The optimization problem of support vector machine is transformed into optimizing its penalty factor c and the kernel function parameter g of radial basis.

[0042] Furthermore, the optimal c, g parameters and the minimum fitness value obtained by iteration were used to construct an SVM evaluation model, and the experimental sample data in S1 and the improved evaluation algorithm were combined to obtain the evaluation model of the support vector machine optimized by the trained improved northern goshawk optimization algorithm;

[0043] The data collected by S1 is input into the support vector machine and used as the initial condition of the fitness formula. The initial fitness value is calculated by formula (1), and then the value is imported into the northern goshawk optimization algorithm to solve the penalty factor c of the support vector machine and the radial basis kernel function parameter g. The fitness value of each northern goshawk in each iteration and the corresponding c and g parameters are iteratively updated according to formulas (2) to (10). When the algorithm iterates to the maximum number of iterations, the operation is terminated, and the global optimal position and the minimum fitness value are output to obtain the optimal c and g values. Finally, the optimal c and g values ​​are applied to the support vector machine to obtain the INGO-SVM model.

[0044] In S5, the target transmission tower is inspected using the same electromagnetic nondestructive testing experimental equipment as that used for the training sample detection, and the corresponding magnetic field strength signal is obtained. The signal value and the screw diameter, protective layer thickness, and pad thickness corresponding to the anchor bolt are input into the INGO-SVM evaluation model for evaluation. The evaluation model calculates and displays the defect level of the anchor bolt nut missing.

[0045] The present invention has the following beneficial effects:

[0046] 1. The present invention aims at electromagnetic nondestructive testing under the interference of ferromagnetic materials. By adopting a large amount of experimental data to perform sample training of the algorithm, the influence of ferromagnetic interference during a single measurement is reduced, and the accuracy of nondestructive testing using the electromagnetic induction principle in the face of more ferromagnetic interference is improved.

[0047] 2. The INGO-SVM model proposed in the present invention can well map the nonlinear relationship between the screw diameter, the number of nuts, the pad thickness, the protective layer thickness and the magnetic field strength in the transmission tower anchor bolts, and analyze the number of nuts on the anchor bolts according to the screw diameter, pad thickness, protective layer thickness and magnetic field strength in actual detection, providing a feasible method for the detection of missing nuts on the transmission tower anchor bolts. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a flowchart of the nondestructive testing process of the present invention.

[0049] Figure 2 Schematic diagram of pinhole imaging. DETAILED DESCRIPTION

[0050] This section is a specific application example of a nondestructive detection method for missing nuts and bolts of a transmission tower anchor bolt based on INGO-SVM provided by the present invention. The purpose is to explain the specific application and operation of the present invention, make the technical solution and effect clearer and easier to understand, and give examples to further explain the present invention in detail. Therefore, the specific implementation described here is only used to explain the present invention and is not used to limit the present invention.

[0051] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments:

[0052] Step 1: Conduct electromagnetic nondestructive testing experiments to obtain a dataset for evaluating the missing defects of the anchor bolts and nuts of transmission towers.

[0053] An electromagnetic nondestructive testing experiment was conducted to detect the missing defects of the anchor bolts and nuts of the transmission tower. By changing the four factors that affect the detection results of the anchor bolts in the concrete protective layer, namely, changing the screw diameter, protective layer thickness, pad thickness and the number of nuts, the magnetic field intensity data under the corresponding influencing factors were obtained. Figure 2 As shown in the figure, the screw diameters d used in the detection platform are respectively selected from M48, M42, M36 and M30; the setting range of the pad thickness s is 4-12mm, with an interval of 4mm; the variation range of the protective layer thickness w is 30-60mm, with an interval of 6mm; at the same time, there are three types of nuts on the screw, 0, 1, and 2, corresponding to no nut missing, one nut missing, and two nuts missing, respectively, recorded as three defect levels of 0, 1, and 2; for 4 screw diameters, 3 pad thicknesses, and 6 protective layer thicknesses, a classification detection experiment of the number of nuts was carried out, and a total of 216 sets of data samples were collected; and the above detection experimental data were combined with the evaluation criteria, and the normalized data were used as training samples for algorithm model evaluation.

[0054] Step 2: The Northern Goshawk Optimization Algorithm (NGO) processes the data.

[0055] The NGO algorithm performs data preprocessing in two main steps: exploring and identifying the target prey and chasing and escaping the target prey. In the exploration and identification stage, the northern goshawk will randomly select a prey and attack it quickly. In this stage, the algorithm needs to perform a global search to find the optimal area where the prey exists. The mathematical model of this stage is as follows:

[0056]

[0057]

[0058]

[0059] In formulas (2), (3), and (4): Pi represents the position of the target prey pursued by the i-th goshawk; represents the objective function value at the corresponding position; k is a random natural number in the range [1, N] and is not i; represents the new state of the i-th goshawk, where represents its new state in the jth dimension, F i new,p1 is the objective function value corresponding to its latest state; r takes a random number in the range of [0,1], and I takes 1 or 2, which are used to generate random numbers for NGO random behavior during search and update.

[0060] When the northern goshawk attacks its prey, the algorithm will enter the stage of pursuit and escape of the target prey. In this stage, the prey will try to escape. At this time, the northern goshawk will chase the target prey. Because the goshawk is extremely fast, it can capture the target prey under any circumstances. Therefore, assuming that the hunting radius is R, the mathematical model of this stage is as follows:

[0061]

[0062]

[0063]

[0064] In formula (5), (6) and (7): represents the new state of the i-th goshawk after the second stage update, where represents its new state in the jth dimension, F i new,p2 is the objective function value corresponding to its latest state; t represents the current number of iterations, and T represents the maximum number of iterations.

[0065] Step 3: Improve the Northern Goshawk Optimization Algorithm (NGO) to obtain the Improved Northern Goshawk Optimization Algorithm (INGO).

[0066] In the NGO algorithm, increasing the diversity of population initialization can effectively expand the algorithm's search range, allowing the algorithm to find potential optimal solutions as much as possible and analyze the optimization effect, and improve the algorithm's convergence speed. In this step, the Cubic chaotic map is added, and its mathematical model is as follows:

[0067] x n+1 =ρx n (1-x n 2 ) (8)

[0068] In formula (8), xn∈(0,1), ρ represents the control parameter, which is used to control the chaotic properties and the degree of nonlinearity. Here ρ is 2.52.

[0069] After a certain number of iterations, the fitness value will remain fixed. In order to optimize the population of goshawks, a pinhole imaging reverse learning strategy is introduced to optimize the population and prevent the algorithm from falling into the local optimum. This strategy is a reverse learning strategy based on the principle of imaging of the derivative phenomenon of light. It dynamically adjusts the number of reverse solutions through a scaling factor to prevent the algorithm from falling into the local optimum. Figure 2 shown.

[0070] In the northern goshawk optimization algorithm, it is assumed that the position change interval of each goshawk is [ai,bi], and the horizontal coordinate of the flame with a height of h is xi,j. Through pinhole imaging, the flame will produce a reflection with a height of H at the position of the horizontal coordinate xi,j*. According to the triangle similarity principle, it can be obtained:

[0071]

[0072] Let m = h / H represent the scaling factor, then the expression of xi,j* is:

[0073]

[0074] It can be seen from equations (9) and (10) that as the scaling factor m decreases, the difference between the positive solution and the inverse solution becomes larger. When m = 1, the equation degenerates into a general reverse learning strategy. It can be seen that the solution obtained by the pinhole imaging reverse learning strategy is a more complete solution.

[0075] Step 4: Use the improved Northern Goshawk Optimization Algorithm (INGO) to optimize the support vector machine (SVM) to obtain the improved Northern Goshawk Optimization Algorithm optimized support vector machine model (INGO-SVM).

[0076] SVM is a common supervised learning algorithm, mainly used for classification of a small number of training sample data sets. The goal of SVM is to find an optimal hyperplane to distinguish two or more categories. The hyperplane optimization problem model is as follows:

[0077]

[0078] st y i (ωx i +b)≥1-ε i (11)

[0079] In formula (11), ω is the normal vector of the hyperplane; b is the offset distance of the hyperplane; ε is the introduced slack variable used to relax the interval distance limit of the function; min represents the point closest to the classification line; c is the introduced penalty factor used to weigh the trade-off between the function classification accuracy and the function interval distance limit.

[0080] Another important part of SVM is the kernel function and its kernel function parameters. The radial basis function (RBF) is selected as the kernel function. RBF can reflect the data into a relatively high dimension and can better complete the processing purpose when facing nonlinear data. Its expression is as follows:

[0081]

[0082] In formula (12), x1 is the RBF center; g is the kernel function parameter, which indicates the coverage of the used restriction function; K() indicates the kernel function; K(x1,x2) indicates the maximum position from the center point x1. Based on this, the optimization problem of the support vector machine is transformed into optimizing its penalty factor c and RBF kernel function parameter g.

[0083] Furthermore, the model is trained by inputting the 216 sets of data collected in step 1 into the SVM and using them as the initial conditions of the fitness formula. The initial fitness value is calculated by formula (1), and then the value is imported into the INGO algorithm to solve the penalty factor c and RBF kernel function parameter g of the SVM. The fitness value of each northern goshawk in each iteration and the corresponding c and g parameters are iteratively updated according to formulas (2) to (10). When the algorithm iterates to the maximum number of iterations, the operation ends, and the global optimal position and the minimum fitness value are output to obtain the optimal c and g values. Finally, the optimal c and g values ​​are applied to the SVM to obtain the INGO-SVM model.

[0084] Step 5: Import the specific parameters of the transmission tower anchor bolts to be inspected into the INGO-SVM model to evaluate the level of missing nuts of the tower anchor bolts.

[0085] According to the construction drawings of the tower to be inspected, the specific parameters of the anchor bolt screw diameter, protective layer thickness, and pad thickness of the tower to be inspected are obtained. The magnetic field strength data under the corresponding influencing factors are collected and brought into the INGO-SVM model. The types of nuts on the screw are analyzed by formulas (11) to (12), and the nut missing level of the tower anchor bolt is given. It is judged whether the nut status on the transmission tower anchor bolt screw is no nut missing, one nut missing, or two nuts missing.

Claims

1. A nondestructive detection method for missing anchor bolts and nuts of transmission towers based on INGO-SVM, characterized in that: The following steps are involved: S1. Conduct electromagnetic nondestructive testing experiments to obtain a dataset for evaluating the missing defects of the anchor bolts and nuts of transmission towers; S2. The data were processed using the Northern Goshawk optimization algorithm; S3. The optimization algorithm of northern goshawk is optimized by using Cubic chaotic mapping and pinhole imaging reverse learning strategy, and the improved northern goshawk optimization algorithm is obtained; S4. Use the improved northern goshawk optimization algorithm to optimize the support vector machine, and obtain the improved northern goshawk optimization algorithm to optimize the support INGO-SVM model; S5. Import the specific parameters of the transmission tower anchor bolts to be inspected into the INGO-SVM model to evaluate the level of missing nuts of the tower anchor bolts; In S1, an electromagnetic nondestructive testing experiment of the missing nut defect of the transmission tower anchor bolt is carried out. By changing the screw diameter, protective layer thickness, pad thickness and number of nuts of the anchor bolt in the concrete protective cap, the magnetic field strength signal under the change of four factors is obtained. At the same time, the defect of missing nuts is divided into no missing nuts, missing one nut and missing two nuts, which are recorded as three levels of 0, 1 and 2. The above experimental results are combined with the evaluation criteria, and the normalized data is used as the analysis sample of the algorithm model; In S2, the northern goshawk population is initialized, and the fitness function is set. The equation of the fitness function is set as follows: ; (1) In formula (1), W represents the transmission tower anchor bolt detection dataset; S i Represents the actual value of the magnetic field strength of the sample; represents the training value of magnetic field strength of the sample; In S3, Cubic chaos mapping is added to the Northern Goshawk optimization algorithm, and its mathematical model is as follows: ; (8) In formula (8): xn∈(0,1), ρ represents the control parameters, which are used to control the chaotic properties and the degree of nonlinearity; After a certain number of iterations, the fitness value will remain fixed, and the pinhole imaging reverse learning strategy is introduced to optimize the population; in the northern goshawk optimization algorithm, it is assumed that the position change interval of each goshawk is [ a i , b i ], the horizontal coordinate of the flame with height h is x i,j , through pinhole imaging, the flame will be at x i,j * The position produces a height of H The reflection of , through the triangle similarity principle, we can get: ;(9) Let m = h / H represent the scaling factor, then x i,j * The expression of is: ;(10) From equations (9) and (10), we can see that as the scaling factor m decreases, the difference between the positive solution and the inverse solution becomes larger. When m = 1, the equation degenerates into a general reverse learning strategy. In S4, through S2 and S3, the population of northern goshawks is updated, the fitness value of each northern goshawk and the corresponding optimal position are calculated, and the algorithm is run to the maximum number of iterations to stop updating, and the minimum fitness value and the corresponding optimal position are assigned to the penalty parameter c and kernel function parameter g in the support vector machine; the goal of the support vector machine is to find an optimal hyperplane to distinguish two or more categories, and the hyperplane optimization problem model is as follows: ;(11) In formula (11), ω is the normal vector of the hyperplane; b is the offset distance of the hyperplane; ε is the introduced slack variable used to relax the interval distance limit of the function; min represents the point closest to the classification line; c is the introduced penalty factor used to weigh the trade-off between the function classification accuracy and the function interval distance limit; Another important part of the support vector machine is the kernel function and its kernel function parameters. The radial basis is selected as the kernel function, and its expression is as follows: ;(12) In formula (12), x1 is the RBF center; g is the kernel function parameter, which indicates the coverage of the used restriction function; K() indicates the kernel function; K(x1,x2) indicates the maximum position from the center point x1; The optimization problem of support vector machine is transformed into optimizing its penalty factor c and the kernel function parameter g of radial basis; In S5, the target transmission tower anchor bolts are tested using the same electromagnetic nondestructive testing experimental equipment as that used for the training sample testing, and the corresponding magnetic field strength signal is obtained. The signal value and the screw diameter, protective layer thickness, and pad thickness corresponding to the anchor bolt are input into the INGO-SVM evaluation model for evaluation. The evaluation model calculates and displays the defect level of the anchor bolt nut missing.

2. The INGO-SVM-based nondestructive detection method for missing foundation bolts and nuts of power transmission towers according to claim 1 is characterized in that: In S2, the optimization algorithm processing of the northern goshawk is divided into two steps, namely, exploration and identification of target prey and pursuit and escape of target prey. In the exploration and identification of target prey stage, the mathematical model of this stage is as follows: ;(2) ;(3) ; (4) In formulas (2), (3), and (4): Pi represents the position of the target prey pursued by the i-th goshawk; represents the objective function value at the corresponding position; k is a random natural number in the range [1, N] and is not i; represents the new state of the i-th goshawk, where represents its new state in the j-th dimension, is the objective function value corresponding to its latest state; r takes a random number in the range of [0,1], and I takes 1 or 2, both of which are used to generate random numbers for the random behavior of the northern goshawk optimization algorithm during search and update; In the stage of pursuit and escape of the target prey, assuming that the hunting radius is R, the mathematical model of this stage is as follows: ;(5) ;(6) ; (7) In formula (5), (6) and (7): represents the new state of the i-th goshawk after the second stage update, where represents its new state in the j-th dimension, is the objective function value corresponding to its latest state; t represents the current number of iterations, and T represents the maximum number of iterations.

3. The INGO-SVM-based nondestructive detection method for missing foundation bolts and nuts of power transmission towers according to claim 1 is characterized in that: The optimal c, g parameters and minimum fitness value obtained by iteration are used to build the SVM evaluation model, and the experimental sample data in S1 and the improved evaluation algorithm are combined to obtain the evaluation model of the support vector machine optimized by the improved northern goshawk optimization algorithm after training; The data collected by S1 is input into the support vector machine and used as the initial condition of the fitness formula. The initial fitness value is calculated by formula (1), and then the value is imported into the northern goshawk optimization algorithm to solve the penalty factor c of the support vector machine and the radial basis kernel function parameter g. According to formulas (2) to (10), the fitness value of each northern goshawk in each iteration and the corresponding c and g parameters are iteratively updated. When the algorithm iterates to the maximum number of iterations, the operation is terminated, and the global optimal position and the minimum fitness value are output to obtain the optimal c and g values. Finally, the optimal c and g values ​​are applied to the support vector machine to obtain the INGO-SVM model.