Method for diagnosing faults of rolling bearings of a top drive gearbox of a drilling rig based on machine vision

Abstract

This invention discloses a machine vision-based fault diagnosis method for rolling bearings in the top drive gearbox of drilling rigs, belonging to the field of large drilling rig fault diagnosis technology. The method includes the following steps: S1, acquiring dynamic frame images under typical fault conditions; S2, preprocessing each frame of the dynamic frame images; S3, plotting a curve showing the relationship between pixel coordinates of vibration monitoring points and time; S4, obtaining data samples of typical fault types; S5, establishing a rolling bearing fault diagnosis model; S6, optimizing the rolling bearing fault diagnosis model; and S7, diagnosing the bearing fault. Through the above method, this invention establishes a rolling bearing fault diagnosis model based on an artificial fish swarm optimization algorithm and a support vector machine. By inputting data samples into the model, fault diagnosis of rolling bearings in the top drive gearbox of drilling rigs can be performed with a high accuracy rate.

Description

Technical Field

[0001] This invention relates to the field of fault diagnosis technology for large drilling rigs, and specifically to a fault diagnosis method for rolling bearings in the top drive gearbox of a drilling rig based on machine vision. Background Technology

[0002] With the continuous progress of human society and the rapid development of the petroleum industry, oil drilling and production equipment is becoming increasingly automated and intelligent. The top drive, a new type of drilling operation drive device that emerged in the 1980s, is a key piece of equipment on offshore drilling platforms, and its operating status directly affects the efficiency and safety of drilling operations. The gear transmission system, composed of bearings, shafts, and housings, is an important component of the top drive unit. Rolling bearings effectively support the normal operation of the shaft and the components between it during rotation, ensuring the shaft is in its proper working position and thus improving rotational accuracy during operation. Therefore, a failure in a rolling bearing can trigger a large chain reaction, rendering the entire drilling rig unable to operate normally.

[0003] Currently, vibration monitoring is a common method for diagnosing rolling bearing faults. However, when collecting vibration signals under fault conditions, the signals are subject to on-site interference and significant noise, exhibiting nonlinear behavior and non-stationary characteristics. Traditional contact measurement methods, which acquire vibration signals by installing sensors, lack sufficient anti-interference capabilities. Furthermore, in large, high-speed machinery, locating sensor installation points or conducting multi-point measurements is labor-intensive and complex. Therefore, traditional fault diagnosis methods cannot meet the modern production needs of rolling bearings in drilling rig top drive gearboxes.

[0004] Based on this, the present invention designs a fault diagnosis method for rolling bearings of drilling rig top drive gearbox based on machine vision to solve the above problems. Summary of the Invention

[0005] In view of the above-mentioned shortcomings of the existing technology, the present invention provides a fault diagnosis method for rolling bearings of drilling rig top drive gearbox based on machine vision.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A machine vision-based fault diagnosis method for rolling bearings in the top drive gearbox of a drilling rig includes the following steps:

[0008] S1. Select a fixed position of the bearing as a vibration monitoring point and collect dynamic frame images under typical fault conditions;

[0009] S2. Use a bilateral filter to preprocess each frame of the dynamic frame image;

[0010] S3. Based on the camera's geometric imaging fundamentals, plot a graph showing the relationship between the pixel coordinates of the vibration monitoring points and time.

[0011] S4. Perform time-domain and frequency-domain analysis on the vibration signal to extract feature values, and use principal component analysis to extract feature parameters to obtain data samples of typical fault types.

[0012] S5. Establish a rolling bearing fault diagnosis model based on artificial fish swarm optimization algorithm and support vector machine;

[0013] S6. Input the data samples into the rolling bearing fault diagnosis model and optimize the rolling bearing fault diagnosis model:

[0014] S7. Perform bearing fault diagnosis based on the optimized rolling bearing fault diagnosis model.

[0015] Furthermore, in step S1, a fixed position of the bearing is selected as a vibration monitoring point. The outer ring of the rolling bearing is fixed, and a point on any inner ring of the bearing is selected as a vibration monitoring point. In step S1, typical fault states include normal bearing, inner ring fault, outer ring fault, and rolling element fault.

[0016] Furthermore, in step S2, the formula for the bilateral filter is expressed as follows:

[0017]

[0018] Where g(i,j) represents the output point; f(k,l) represents the input point; and w(i,j,k,l) ​​represents the value calculated using two Gaussian functions. The specific calculation process is as follows:

[0019] w = w s *w r

[0020]

[0021]

[0022] Among them, w s The template weights, w, are determined by the Euclidean distance of the pixel position. r The template weights are determined by the difference in pixel values; q(i,j) represents the coordinates of the other coefficients of the template window, and f(i,j) represents the pixel value of the image at point q(i,j); p(k,l) is the center coordinate point of the template window, and the corresponding pixel value is f(k,l); σ r is the standard deviation of the Gaussian function.

[0023] Furthermore, in step S3, the transformation relationship between the camera's geometric imaging basis, the world coordinate system, and the image coordinate system is as follows:

[0024]

[0025] Among them, [xy] T The coordinates of a pixel in the image coordinate system; [X x Y x Z x ] T f is the position of a point in the camera coordinate system; f / dx, f / dy, x0, and y0 are the camera's intrinsic parameters, and f is the camera's effective focal length. Represents the intrinsic parameter matrix; Represents the extrinsic parameter matrix; This represents the camera's projection matrix.

[0026] Furthermore, in step S4, the vibration signal is analyzed in the time domain and frequency domain to extract feature values, including eight time domain features such as maximum value, minimum value, peak-to-peak value, root mean square value, waveform index, impulse index, peak factor, and margin index, and four frequency domain features such as center frequency, mean square frequency, frequency variance, and frequency standard deviation. The 12 feature values ​​are used to extract feature parameters to form feature vectors, which are used as the input vector of the model.

[0027] Furthermore, in step S4, the main steps of principal component analysis are as follows:

[0028] A. Standardize the indicator data to obtain a standardized matrix;

[0029] B. Establish the covariance matrix based on standardized data;

[0030] C. Find the eigenvalues ​​of the covariance matrix, the principal component contribution rate, and the cumulative variance contribution rate to determine the number of principal components;

[0031] D. Establish the initial factor loading matrix and interpret the principal components;

[0032] E. Calculate the comprehensive scoring function.

[0033] Furthermore, in steps S5 and S6, the support vector machine introduces a Gaussian kernel function, and uses the artificial fish swarm optimization algorithm to optimize the penalty factor C and width parameter σ of the Gaussian kernel function. The data sample feature vectors are then used for training to obtain the optimal support vector machine diagnostic model.

[0034] Furthermore, the algorithm for Support Vector Machines is as follows:

[0035] Input: Training sample dataset T;

[0036] Output: Classification decision function;

[0037] Choose Gaussian kernel function k(x) i x jGiven a suitable parameter C, construct and solve the optimization problem:

[0038]

[0039]

[0040] The optimal solution α = (α1, α2, ..., α3) is obtained. N );

[0041] Choose one that satisfies 0 < α i <C component α i ,calculate:

[0042]

[0043] Construct the classification decision function:

[0044]

[0045] The Gaussian kernel function has the ability to extract local feature information and is also known as the radial basis function. The specific formula is as follows:

[0046]

[0047] Furthermore, the specific process for optimizing the penalty factor C and width parameter σ of the Gaussian kernel function using the artificial fish swarm optimization algorithm includes the following steps:

[0048] A. Initialization and parameter settings of artificial fish population: Let the initial population of the artificial fish population be W = {w1, w2, ..., w...} N}, where w k (k = 1, 2, ..., N) is the initial position vector of the kth artificial fish, and N is the number of individuals in the artificial fish group. The number of individuals in the artificial fish group is set to N = 100.

[0049] B. Calculate the fitness function (food concentration function) for each artificial fish, and record the maximum value and its corresponding position vector on the bulletin board;

[0050] C. Behavioral selection of artificial fish swarms: Each artificial fish engages in tail-chasing behavior. If successful, it performs swarming behavior; if unsuccessful, it performs foraging behavior. The position vector of the artificial fish changes, generating a new population.

[0051] Rear-end collision behavior:

[0052]

[0053] Among them, w k / next Indicates the next position; Rand represents a random number in the range (0, 1); step represents the step size of the artificial fish; wmax Indicates the state of maximum food concentration; n f Indicates the number of partners in the domain; z l This represents the food concentration value at a random location within the sensing range;

[0054] Group behavior:

[0055]

[0056] Among them, w c Indicates the position and state of the center of the fish school; z c This indicates the food concentration value at the center location;

[0057] Foraging behavior:

[0058]

[0059] Among them, w l This indicates the state of a random location within the fish's perception range;

[0060] D. Update the bulletin board; Each time the artificial fish's position vector changes, calculate the corresponding fitness function value, compare the maximum fitness function value with the maximum value before the update, and if the fitness function value is larger, update the content in the bulletin board;

[0061] E. Termination condition judgment: Determine whether the number of iterations of the artificial fish swarm has reached the maximum. If it has, output the current position vector and terminate the algorithm; otherwise, jump to the behavior selection of the artificial fish swarm.

[0062] Furthermore, in step S7, the input is the extracted feature vector, and the output is the correspondence between four types of rolling bearing fault states: where 0 is the normal state, 1 is the inner ring fault state, 2 is the outer ring fault state, and 3 is the rolling element fault state.

[0063] Beneficial effects

[0064] This invention addresses the shortcomings of traditional vibration measurement methods, such as insufficient anti-interference capability and complex operation. It proposes a machine vision-based fault diagnosis method for rolling bearings in drilling rig top drive gearboxes. The method employs machine vision to acquire dynamic frame images of the bearing under typical fault conditions. After image preprocessing, a time-domain signal diagram is plotted based on camera geometry. Time-domain and frequency-domain features are extracted as data samples. A rolling bearing fault diagnosis model is established based on the artificial fish swarm optimization algorithm and support vector machine. Inputting the data samples into the model enables fault diagnosis of rolling bearings in drilling rig top drive gearboxes. This method exhibits high fault diagnosis accuracy, good anti-interference capability, and convenient operation, making it of significant practical value for ensuring the safe operation of large drilling rigs. Attached Figure Description

[0065] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0066] Figure 1 This is a flowchart of the machine vision-based fault diagnosis method for rolling bearings in the top drive gearbox of a drilling rig, according to the present invention.

[0067] Figure 2 This is a flowchart illustrating the rolling bearing fault diagnosis model established based on the artificial fish swarm optimization algorithm and support vector machine in this invention.

[0068] Figure 3 This is a time-domain signal diagram of the vibration monitoring points of the rolling bearing under normal conditions.

[0069] Figure 4 This is a time-domain signal diagram of the vibration monitoring points of the rolling bearing when the outer ring fails.

[0070] Figure 5 The time-domain signal diagram of the vibration monitoring point of the rolling bearing when the inner ring fails;

[0071] Figure 6 This is a time-domain signal diagram of the vibration monitoring points of the rolling bearing when the rolling element fails.

[0072] Figure 7 The image shows the test results using this invention. Detailed Implementation

[0073] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0074] The present invention will be further described below with reference to embodiments.

[0075] Example 1

[0076] Please refer to the instruction manual appendix. Figure 1-7 A machine vision-based fault diagnosis method for rolling bearings in the top drive gearbox of a drilling rig includes the following steps:

[0077] S1. Select a fixed position of the bearing as a vibration monitoring point and collect dynamic frame images under typical fault conditions;

[0078] In step S1, a fixed position of the bearing is selected as the vibration monitoring point. The outer ring of the rolling bearing is fixed, and a point on any inner ring of the bearing is selected as the vibration monitoring point.

[0079] In step S1, dynamic frame images are acquired by using a high-speed industrial camera equipped with a flicker-free industrial light source; the high-speed industrial camera is fixed with a tripod, the high-speed industrial camera is aligned with the vibration monitoring point, and the dynamic frame images acquired by the high-speed industrial camera are transmitted to the computer for storage via a high-speed acquisition card.

[0080] In step S1, the position and intensity of the flicker-free industrial light source can be adjusted to improve the contrast and brightness of the images captured by the high-speed industrial camera.

[0081] In step S1, typical fault states include normal bearing, inner ring fault, outer ring fault, and rolling element fault.

[0082] S2. Use a bilateral filter to preprocess each frame of the dynamic frame image;

[0083] In step S2, the formula for the bilateral filter is as follows:

[0084]

[0085] Where i and j are the coordinates of the current convolved pixel, and k and l are the coordinates of the neighboring pixels; g(i, j) represents the output point; f(k, l) represents the input point; and w(i, j, k, l) represents the value calculated using two Gaussian functions, as detailed below:

[0086] w = w s *w r

[0087]

[0088]

[0089] Among them, w s The template weights, w, are determined by the Euclidean distance of the pixel position. r The template weights are determined by the difference in pixel values; q(i,j) represents the coordinates of the other coefficients of the template window, and f(i,j) represents the pixel value of the image at point q(i,j); p(k,l) is the center coordinate point of the template window, and the corresponding pixel value is f(k,l); σ r The standard deviation of the Gaussian function;

[0090] In step S2, each frame of the dynamic frame image is preprocessed using a bilateral filter, which can effectively remove noise while better preserving edge information.

[0091] S3. Based on the camera's geometric imaging fundamentals, plot a graph showing the relationship between the pixel coordinates of the vibration monitoring points and time.

[0092] In step S3, the transformation relationships between the camera's geometric imaging basis, the world coordinate system, and the image coordinate system are as follows:

[0093]

[0094] Among them, [xy] T The coordinates of a pixel in the image coordinate system; [X x Y x Z x ] T f is the position of a point in the camera coordinate system; f / dx, f / dy, x0, and y0 are the camera's intrinsic parameters, and f is the camera's effective focal length. Represents the intrinsic parameter matrix; Represents the extrinsic parameter matrix; The projection matrix of the camera is represented; Xx, Yx, and Zx represent the coordinates in the camera coordinate system; T is a mathematical operation (transpose); and A represents the rotation matrix.

[0095] a represents the translation matrix.

[0096] S4. Perform time-domain and frequency-domain analysis on the vibration signal to extract feature values, and use principal component analysis to extract feature parameters to obtain data samples of typical fault types.

[0097] In step S4, based on the curve of the relationship between the pixel coordinates of the vibration monitoring points and time drawn in step S3, a time-domain signal graph is obtained. The time-domain signal is converted into a frequency-domain signal through discrete Fourier transform. Feature values ​​are extracted from the time-domain and frequency-domain signals. Principal component analysis is used to fuse the feature parameters to obtain the fused state feature parameters.

[0098] In step S4, the vibration signal is analyzed in the time domain and frequency domain to extract feature values, including eight time domain features such as maximum value, minimum value, peak-to-peak value, root mean square value, waveform index, impulse index, peak factor, and margin index, and four frequency domain features such as center frequency, mean square frequency, frequency variance, and frequency standard deviation. The 12 feature values ​​are extracted using principal component analysis to form feature vectors, which are used as the input vectors of the model.

[0099] In step S4, the main steps of principal component analysis are as follows:

[0100] A. Standardize the indicator data to obtain a standardized matrix;

[0101] B. Establish the covariance matrix based on standardized data;

[0102] C. Find the eigenvalues ​​of the covariance matrix, the principal component contribution rate, and the cumulative variance contribution rate to determine the number of principal components;

[0103] D. Establish the initial factor loading matrix and interpret the principal components;

[0104] E. Calculate the comprehensive scoring function;

[0105] S5. Establish a rolling bearing fault diagnosis model based on artificial fish swarm optimization algorithm and support vector machine;

[0106] S6. Input the data samples into the rolling bearing fault diagnosis model and optimize the rolling bearing fault diagnosis model;

[0107] In steps S5 and S6, the support vector machine introduces a Gaussian kernel function, and uses the artificial fish swarm optimization algorithm to optimize the penalty factor C and width parameter σ of the Gaussian kernel function. The data sample feature vectors are then used for training to obtain the optimal support vector machine diagnostic model.

[0108] The algorithm for Support Vector Machine is as follows:

[0109] Input: Training sample dataset T;

[0110] Output: Classification decision function;

[0111] Choose Gaussian kernel function k(x) i x j Given a suitable parameter C, construct and solve the optimization problem:

[0112]

[0113]

[0114] The optimal solution α = (α1, α2, ..., α3) is obtained. N );

[0115] Choose one that satisfies 0 < α i <C component α i ,calculate:

[0116]

[0117] Construct the classification decision function:

[0118]

[0119] The Gaussian kernel function has the ability to extract local feature information and is also known as the radial basis function. The specific formula is as follows:

[0120]

[0121] Where xi is any point in space, xj is a center, α is the optimal solution, i = 1, 2, ..., N, N is the number of training samples, yi ∈ {-1, 1}, w and b are hyperplane parameters, α* and b* represent some components of α and b, and σ refers to the variance;

[0122] The specific process of optimizing the penalty factor C and width parameter σ of the Gaussian kernel function using the artificial fish swarm optimization algorithm includes the following steps:

[0123] A. Initialization and parameter settings of artificial fish population: Let the initial population of the artificial fish population be W = {w1, w2, ..., w...} N}, where w k (k = 1, 2, ..., N) is the initial position vector of the kth artificial fish, and N is the number of individuals in the artificial fish group. The number of individuals in the artificial fish group is set to N = 100.

[0124] B. Calculate the fitness function (food concentration function) for each artificial fish, and record the maximum value and its corresponding position vector on the bulletin board;

[0125] C. Behavioral selection of artificial fish swarms: Each artificial fish engages in tail-chasing behavior. If successful, it performs swarming behavior; if unsuccessful, it performs foraging behavior. The position vector of the artificial fish changes, generating a new population.

[0126] Rear-end collision behavior:

[0127]

[0128] Among them, w k / next Indicates the next position; Rand represents a random number in the range (0, 1); step represents the step size of the artificial fish; w max Indicates the state of maximum food concentration; n f Indicates the number of partners in the domain; z l The value represents the food concentration at a random location within the sensing range; δ represents the crowding factor; z max This indicates the highest food concentration value;

[0129] Group behavior:

[0130]

[0131] Among them, w c Indicates the position and state of the center of the fish school; z c This indicates the food concentration value at the center location;

[0132] Foraging behavior:

[0133]

[0134] Among them, w l This indicates the state of a random location within the fish's perception range;

[0135] D. Update the bulletin board; Each time the artificial fish's position vector changes, calculate the corresponding fitness function value, compare the maximum fitness function value with the maximum value before the update, and if the fitness function value is larger, update the content in the bulletin board;

[0136] E. Termination condition judgment: Determine whether the number of iterations of the artificial fish swarm has reached the maximum. If it has, output the current position vector and terminate the algorithm; otherwise, jump to the behavior selection of the artificial fish swarm.

[0137] S7. Perform bearing fault diagnosis based on the optimized rolling bearing fault diagnosis model.

[0138] In step S7, the input is the extracted feature vector, and the output is the correspondence between four types of rolling bearing fault states: where 0 is the normal state, 1 is the inner ring fault state, 2 is the outer ring fault state, and 3 is the rolling element fault state.

[0139] pass Figure 7 It can be seen that the method of the present invention has a high accuracy rate in fault diagnosis;

[0140] This invention addresses the shortcomings of traditional vibration measurement methods, such as insufficient anti-interference capability and complex operation. It proposes a machine vision-based fault diagnosis method for rolling bearings in drilling rig top drive gearboxes. The method employs machine vision to acquire dynamic frame images of the bearing under typical fault conditions. After image preprocessing, a time-domain signal diagram is plotted based on camera geometry. Time-domain and frequency-domain features are extracted as data samples. A rolling bearing fault diagnosis model is established based on the artificial fish swarm optimization algorithm and support vector machine. Inputting the data samples into the model enables fault diagnosis of rolling bearings in drilling rig top drive gearboxes. This method exhibits high fault diagnosis accuracy, good anti-interference capability, and convenient operation, making it of significant practical value for ensuring the safe operation of large drilling rigs.

[0141] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A fault diagnosis method for rolling bearings in the top drive gearbox of a drilling rig based on machine vision, characterized in that, Includes the following steps: S1. Select a fixed position of the bearing as a vibration monitoring point and collect dynamic frame images under typical fault conditions; S2. Use a bilateral filter to preprocess each frame of the dynamic frame image; S3. Based on the camera's geometric imaging fundamentals, plot a graph showing the relationship between the pixel coordinates of the vibration monitoring points and time. S4. Perform time-domain and frequency-domain analysis on the vibration signal to extract feature values, and use principal component analysis to extract feature parameters to obtain data samples of typical fault types. S5. Establish a rolling bearing fault diagnosis model based on artificial fish swarm optimization algorithm and support vector machine; S6. Input the data samples into the rolling bearing fault diagnosis model and optimize the rolling bearing fault diagnosis model; S7. Perform bearing fault diagnosis based on the optimized rolling bearing fault diagnosis model. In step S2, the formula for the bilateral filter is as follows: Where g(i,j) represents the output point; f(k,l) represents the input point; and w(i,j,k,l) ​​represents the value calculated using two Gaussian functions. The specific calculation process is as follows: w=w s *In r Among them, w s The template weights, w, are determined by the Euclidean distance of the pixel position. r The template weights are determined by the difference in pixel values; q(i,j) represents the coordinates of the other coefficients of the template window, and f(i,j) represents the pixel value of the image at point q(i,j); p(k,l) is the center coordinate point of the template window, and the corresponding pixel value is f(k,l); σ r The standard deviation of the Gaussian function; In step S3, the transformation relationships between the camera's geometric imaging basis, the world coordinate system, and the image coordinate system are as follows: Among them, [xy] T The coordinates of a pixel in the image coordinate system; [X x Y x Z x ] T f is the position of a point in the camera coordinate system; f / dx, f / dy, x0, and y0 are the camera's intrinsic parameters, and f is the camera's effective focal length. Represents the intrinsic parameter matrix; Represents the extrinsic parameter matrix; This represents the camera's projection matrix.

2. The fault diagnosis method for rolling bearings of drilling rig top drive gearbox based on machine vision according to claim 1, characterized in that, In step S1, a fixed position of the bearing is selected as the vibration monitoring point. The outer ring of the rolling bearing is fixed, and a point of any inner ring of the bearing is selected as the vibration monitoring point. In step S1, typical fault states include normal bearing, inner ring fault, outer ring fault and rolling element fault.

3. The fault diagnosis method for rolling bearings of drilling rig top drive gearbox based on machine vision according to claim 1, characterized in that, In step S4, the vibration signal is analyzed in the time domain and frequency domain to extract feature values, including eight time domain features such as maximum value, minimum value, peak-to-peak value, root mean square value, waveform index, impulse index, peak factor, and margin index, and four frequency domain features such as center frequency, mean square frequency, frequency variance, and frequency standard deviation. The 12 feature values ​​are extracted using principal component analysis to form feature vectors, which are used as the input vectors of the model.

4. The fault diagnosis method for rolling bearings of drilling rig top drive gearbox based on machine vision according to claim 1, characterized in that, In step S4, the steps of principal component analysis are as follows: A. Standardize the indicator data to obtain a standardized matrix; B. Establish the covariance matrix based on standardized data; C. Find the eigenvalues ​​of the covariance matrix, the principal component contribution rate, and the cumulative variance contribution rate to determine the number of principal components; D. Establish the initial factor loading matrix and interpret the principal components; E. Calculate the comprehensive scoring function.

5. The fault diagnosis method for rolling bearings of drilling rig top drive gearbox based on machine vision according to claim 1, characterized in that, In steps S5 and S6, the support vector machine introduces a Gaussian kernel function, and uses the artificial fish swarm optimization algorithm to optimize the penalty factor C and width parameter σ of the Gaussian kernel function. The data sample feature vectors are then used for training to obtain the optimal support vector machine diagnostic model.

6. The fault diagnosis method for rolling bearings of drilling rig top drive gearbox based on machine vision according to claim 5, characterized in that, The algorithm for Support Vector Machine is as follows: Input: Training sample dataset T; Output: Classification decision function; Choose Gaussian kernel function k(x) i x j Given a suitable parameter C, construct and solve the optimization problem: The optimal solution α = (α1, α2, ..., α3) is obtained. N ); Choose one that satisfies 0 < α i <C component α i ,calculate: Construct the classification decision function: The Gaussian kernel function has the ability to extract local feature information and is also known as the radial basis function. The specific formula is as follows:

7. The fault diagnosis method for rolling bearings of drilling rig top drive gearbox based on machine vision according to claim 6, characterized in that, The specific process for optimizing the penalty factor C and width parameter σ of the Gaussian kernel function using the artificial fish swarm optimization algorithm includes the following steps: A. Initialization and parameter settings of artificial fish population: Let the initial population of the artificial fish population be W = {w1, w2, ..., w...} N }, where w k Let N be the initial position vector of the kth artificial fish, k = 1, 2, ..., N, where N is the number of individuals in the artificial fish group. The number of individuals in the artificial fish group is set to N = 100. B. Calculate the fitness function of each artificial fish, and record the maximum value and its corresponding position vector on the bulletin board; C. Behavioral selection of artificial fish swarms: Each artificial fish engages in tail-chasing behavior. If successful, it performs swarming behavior; if unsuccessful, it performs foraging behavior. The position vector of the artificial fish changes, generating a new population. Rear-end collision behavior: Among them, w k / next Indicates the next position; Rand represents a random number in the range (0, 1); step represents the step size of the artificial fish; w max Indicates the state of maximum food concentration; n f Indicates the number of partners in the domain; z l This represents the food concentration value at a random location within the sensing range; Group behavior: Among them, w c Indicates the position and state of the center of the fish school; z c This indicates the food concentration value at the center location; Foraging behavior: Among them, w l This indicates the state of a random location within the fish's perception range; D. Update the bulletin board; Each time the artificial fish's position vector changes, calculate the corresponding fitness function value, compare the maximum fitness function value with the maximum value before the update, and if the fitness function value is larger, update the content in the bulletin board; E. Termination condition judgment: Determine whether the number of iterations of the artificial fish swarm has reached the maximum. If it has, output the current position vector and terminate the algorithm; otherwise, jump to the behavior selection of the artificial fish swarm.

8. The fault diagnosis method for rolling bearings of drilling rig top drive gearbox based on machine vision according to claim 1, characterized in that, In step S7, the input is the extracted feature vector, and the output is the correspondence between four types of rolling bearing fault states: 0 is the normal state, 1 is the inner ring fault state, 2 is the outer ring fault state, and 3 is the rolling element fault state.

Images