Non-contact detection method for loose bolts on top cover of hydraulic turbine in hydropower station

Through video monitoring and sparse optical flow algorithm, the vibration frequency and frequency entropy of the top cover bolt of the turbine are monitored in real time, solving the problem of loose detection in the existing technology that is not real-time and susceptible to noise, and achieving efficient and accurate bolt loose detection and position.

CN118710725BActive Publication Date: 2024-11-12NORTHWEST ENGINEERING CORPORATION LIMITED
View PDF 9 Cites 0 Cited by

Patent Information

Application Number
CN202411195087.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-29
Publication Date
2024-11-12
Estimated Expiration
2044-08-29

AI Technical Summary

Technical Problem

The existing turbine top cover bolt loose detection methods rely on manual inspection or expensive ultrasonic sensors, which cannot achieve real-time monitoring and are susceptible to industrial frequency noise, resulting in poor detection results and safety hazards.

Method used

Video monitoring and sparse optical flow algorithm are used to detect the vibration state of the top cover bolt. By calculating the vibration frequency and frequency entropy of the bolt, a frequency matrix and thermal map are constructed to realize contactless real-time monitoring and positioning of the loose bolts.

Benefits of technology

Real-time monitoring and accurate positioning of the loose bolts of the turbine roof cover are realized, reducing the monitoring cost and the impact of power frequency noise, and improving detection efficiency and accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118710725B_ABST
    Figure CN118710725B_ABST
Patent Text Reader

Abstract

The present invention discloses a non-contact detection method for loose bolts of a top cover of a hydropower station turbine, and the specific steps are as follows: S1, video monitoring of the vibration state of the top cover bolts to obtain a video image; S2, processing the video image to obtain optical flow information; S3, based on the relationship between the structural optical flow information and the structural vibration response information, obtaining multiple bolt vibration response information; S4, according to the multiple bolt vibration response information, obtaining the bolt vibration frequency, and constructing a top cover bolt frequency matrix; S5, obtaining the frequency entropy of the frequency matrix, and making a bolt loosening judgment; S6, constructing a circular heat map of different bolt frequencies of the top cover, and locating the position of the loose bolts. The present invention discloses a non-contact detection method for loose bolts of a top cover of a hydropower station turbine, and the vibration signals of bolts at different positions are calculated in real time by an image processing method, and then the positioning of the top cover bolts and real-time monitoring of the state are realized by the change of the bolt vibration signals.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of turbine bolt loosening detection, and relates to a non-contact detection method for loosening of turbine top cover bolts of a hydropower station. Background Art

[0002] As an electromechanical device that converts water energy into electrical energy, the turbine occupies the most core position in the hydropower station. The turbine top cover is used to install the water guide bearing, and together with the main shaft seal, it isolates the waterwheel chamber and the runner chamber to prevent water from backflowing into the waterwheel chamber. At present, the turbine top cover is mainly connected to the cylinder body by bolts. However, as the turbine runs, the vibration during the operation of the unit will induce the top cover bolts to loosen. In severe cases, the water in the turbine cylinder will overflow from the top cover, causing major accidents such as flooding the plant, endangering the lives and property of hydropower station workers and people downstream.

[0003] At present, the main method for detecting the looseness of the bolts on the turbine top cover is manual inspection, that is, within a specific maintenance cycle, the operation and maintenance personnel check the looseness of the top cover bolts by means of hammering, visual inspection, etc. However, for the turbine top cover, the number of fixing bolts usually reaches about 150. The manual method alone is time-consuming and labor-intensive. More importantly, this regular maintenance and overhaul method cannot grasp the real-time status of the bolts, and the detection effect depends on the subjective will of the operation and maintenance personnel, which makes the operation process of the turbine have certain risks. In recent years, with the development of new technologies, some real-time detection methods for the looseness of bolts have been produced, such as ultrasonic sensors, which detect the looseness of bolts by real-time monitoring the elongation of the bolts or the change in the elongation of the bolts. However, a single ultrasonic sensor is expensive, and the use of a large number of sensors makes the test system complicated and cannot be used in large quantities. In addition, the ultrasonic sensor needs to be in direct contact with the structure, which will be affected by the on-site power frequency noise and cannot accurately judge the monitoring results. Summary of the invention

[0004] The purpose of the present invention is to provide a non-contact detection method for loosening of top cover bolts of a turbine in a hydropower station, calculate the vibration signals of bolts at different positions in real time through an image processing method, and then realize real-time monitoring of the status of top cover bolts through changes in the bolt vibration signals.

[0005] The technical solution adopted by the present invention is a non-contact detection method for loosening of top cover bolts of a hydropower station turbine, which specifically comprises the following steps:

[0006] S1, performing video monitoring on the vibration state of the top cover bolts to obtain video images;

[0007] S2, using a sparse optical flow algorithm to process the video image obtained in S1 to obtain the optical flow information of the bolt structure;

[0008] S3, obtaining vibration response information of multiple bolts based on the relationship between the structured optical flow information and the structural vibration response information;

[0009] S4, obtaining multiple bolt vibration frequencies according to the multiple bolt vibration response information, and constructing a top cover bolt frequency matrix;

[0010] S5, obtaining the frequency entropy of the frequency matrix through the frequency matrix of the top cover bolts, and judging whether there is looseness in the bolts by comparing the magnitude of the frequency entropy before and after the bolts vibrate in real time;

[0011] S6, construct a circular heat map of different bolt frequencies of the top cover to locate the locations of loose bolts.

[0012] The present invention is also characterized in that:

[0013] S1 is specifically:

[0014] By installing a camera above the turbine top cover, the vibration status of the top cover bolts is monitored via video and video images are obtained.

[0015] S2 is specifically:

[0016] The sparse optical flow algorithm is used to calculate the optical flow information of the bolt. The specific formula is:

[0017] (1);

[0018] Where: and are the brightness gradients of the bolt vibration video image pixels in the horizontal and vertical directions, is the derivative of the image pixel brightness with respect to time t, u and v are the optical flows of image pixels in the horizontal and vertical directions respectively;

[0019] Since the optical flows u and v in formula (1) are unknown quantities and cannot be solved directly, we assume that the target pixel P(P x ,P y ) and its neighbors W(W x ,W y ) The k pixels in the same image have similar motions, that is, there are k equations satisfying formula (1); thus, formula (1) can be expressed as:

[0020] (2);

[0021] In the formula, p1 , p 2 , , p k The target pixel P Pixels in the neighborhood W, k Neighborhood K The number of pixels inside, and Neighborhood pixels p k Brightness gradients in horizontal and vertical directions, Neighborhood pixels p k The derivative of brightness with respect to time t, u and v are the optical flows of image pixels in the horizontal and vertical directions respectively;

[0022] definition:

[0023] , (3);

[0024] In the formula, A is all pixels in the neighborhood W (a total of k ) in the horizontal and vertical directions, b is all pixels in the neighborhood W (a total of k ) Brightness with respect to time t The matrix composed of the derivatives of ;

[0025] At this time, formula (2) can be expressed as:

[0026] (4);

[0027] In the formula, A is all pixels in the neighborhood W (a total of k ) in the horizontal and vertical directions, b is all pixels in the neighborhood W (a total of k ) Brightness with respect to time t The matrix composed of the derivatives of ; u and v are the optical flows of image pixels in the horizontal and vertical directions respectively;

[0028] There are k equations in formula (4) and two unknown quantities u and v , the least square method is used to solve it, and the optical flow formula of the target pixel in the horizontal and vertical directions can be expressed as:

[0029] (5);

[0030] In the formula, A T is the transposed matrix of matrix A, T is the transposed symbol, u and v are the optical flows of image pixels in the horizontal and vertical directions, respectively. b is all pixels in the neighborhood W (a total of k ) Brightness with respect to time t The matrix composed of the derivatives of ;

[0031] From formula (5), we can see that the accuracy of the target pixel optical flow calculation depends on the target gradient matrix , when the matrix When the reversibility is strong, the feature point corresponding to the pixel is easy to track with high accuracy. The target pixel screening method is used to calculate the directional gradients of all pixels on the bolt structure in the horizontal and vertical directions, namely:

[0032] (6);

[0033] (7);

[0034] In formula (6), is the directional gradient of the image pixel in the horizontal direction, I Represents the brightness value of the pixels in the image. represents the convolution operation, is the gradient operator in the horizontal direction;

[0035] In formula (7), is the directional gradient of the image pixel in the vertical direction, I Represents the brightness value of the pixels in the image. represents the convolution operation, is the gradient operator in the vertical direction;

[0036] in, for:

[0037] (8);

[0038] In the formula, is the gradient operator in the horizontal direction, which is combined with the image brightness matrix I By performing convolution operation, the brightness gradient of the image pixel brightness in the horizontal direction can be obtained;

[0039] for:

[0040] (9);

[0041] In the formula, is the gradient operator in the vertical direction, which is combined with the image brightness matrix I By performing convolution operation, the brightness gradient of the image pixel brightness in the vertical direction can be obtained;

[0042] According to the magnitude of the directional gradient of different pixels in the horizontal and vertical directions, the gradient amplitude is calculated, which is specifically expressed as:

[0043] (10);

[0044] In the formula, is the gradient magnitude of the image pixel, is the directional gradient of the image pixel in the horizontal direction, is the directional gradient of the image pixel in the vertical direction;

[0045] The gradient amplitudes of all pixels on each bolt are calculated, and the five pixels with the largest gradient amplitudes are selected as target pixels to obtain the optical flow of the target pixels. Combined with formula (5), the optical flow values ​​of the five target pixels are averaged to obtain the optical flow information of the bolt structure, which is specifically expressed as:

[0046] (11);

[0047] Where: u(t) and v(t) are the optical flows of the bolt in the horizontal and vertical directions at time t, u i (t) and v i (t) Respectively indicate the number i The optical flow of the pixel at time t.

[0048] S3 is specifically: According to the optical flow of the bolt at different times, combined with the corresponding relationship between the optical flow and the vibration response, the vibration response information of the bolt structure at time t can be expressed as:

[0049] (12);

[0050] Where: and are the vibration response information of the bolt structure, is the time interval between two adjacent frames, k The camera uses frequency, u(t) and v(t) are the optical flows of the bolt in the horizontal and vertical directions at time t, respectively.

[0051] S4 is specifically as follows: considering that after the turbine top cover bolts are loosened, their vibration characteristics will change, that is, the frequency of the bolts will change, firstly, the turbine top cover bolts are numbered as 1, 2, ..., n;

[0052] Secondly, according to the vibration response information of different bolts, the vibration frequency of different bolts can be obtained by Fourier transform, which is expressed as:

[0053] (13);

[0054] Where: S n For the number n The bolt vibration response information, f n For the number n The vibration frequency of the bolt;

[0055] Finally, the frequency matrix of the top cover bolts is constructed by calculating the structure of different bolt vibration frequencies, which can be expressed as:

[0056] (14).

[0057] S5 is specifically as follows: the constructed frequency matrix is ​​a diagonal matrix, and its diagonal elements, that is, the singular values ​​of the matrix are the vibration frequencies of bolts at different positions. Through the frequency matrix, the probability of the matrix singular value distribution can be obtained, which is recorded as:

[0058] (15);

[0059] Where: E q For the number q The probability of the bolt singular value distribution, f q For the number q The frequency of the bolts;

[0060] According to the probability of the bolt singular value distribution, the frequency entropy H of the bolt vibration can be calculated as:

[0061] (16);

[0062] In the formula, E q For the number q The probability of the singular value distribution of the bolt, H is the frequency entropy of the bolt vibration;

[0063] Among them, taking the initial vibration frequency entropy when the top cover bolts are not loose as a benchmark, by comparing the size relationship between the initial frequency entropy and the current frequency entropy in real time, it is judged whether the top cover bolts are loose, and an alarm is issued for abnormal looseness of the bolts.

[0064] S6 is specifically as follows: when an abnormal alarm of bolt loosening occurs, in order to facilitate the location of the loose bolt, the method of constructing a circular heat map of bolt frequency is used to locate the loose bolt, and the frequency arrangement is arranged according to the actual position of the bolt; since the frequency of the bolt will be abnormal after the bolt is loose, that is, it will be different from the vibration characteristics of the surrounding bolts, the position of the loose bolt is identified through the circular heat map of frequency;

[0065] In addition, when the bolt structure is just loosened, it has little effect on its low-order frequency, but mainly affects its high-order frequency characteristics. Therefore, based on the camera video image acquisition frame rate k, the maximum frequency that can be collected is k / 2, that is, the first j-order frequencies of the turbine top cover bolts can be calculated, which are recorded as:

[0066] (17);

[0067] Where: f j is the highest order frequency that can be calculated, which satisfies f j <k / 2;

[0068] Finally, by drawing the frequency heat map of the first j-order top cover bolts and comprehensively considering the distribution size of the first j-order frequency heat map, the position of the loose bolts can be located, thus completing the bolt loosening detection and loose bolt positioning.

[0069] Compared with the prior art, the present invention has the following beneficial effects:

[0070] (1) The non-contact detection method for loose top cover bolts provided by the present invention is not easily affected by the power frequency environment in the factory in practical application, thereby improving the monitoring efficiency and reducing the monitoring cost.

[0071] (2) In terms of efficiency, the non-contact detection method for loose top cover bolts provided by the present invention is based on video images and can realize simultaneous monitoring of the status of top cover bolts, thus having high measurement efficiency.

[0072] (3) In terms of convenience, the non-contact detection method for loose top cover bolts disclosed in the present invention proposes a bolt structure vibration measurement strategy based on an optical flow algorithm, which has the advantage of not requiring prior structural marking and can achieve rapid measurement of the vibration response of the bolt structure.

[0073] (4) In terms of accuracy, the non-contact detection method for loose top cover bolts provided by the present invention proposes a bolt loosening detection method based on frequency entropy and a loose bolt positioning method based on frequency circular heat map, which can achieve accurate monitoring of bolt status and accurate positioning of loose bolts. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1It is a schematic flow chart of the non-contact detection method for loosening of top cover bolts of the present invention;

[0075] Figure 2 It is a schematic diagram of the relationship between the target pixel and the neighboring pixel in the non-contact detection method for loosening of the top cover bolts of the present invention;

[0076] Figure 3 It is a schematic diagram of the distribution of the turbine top cover bolts of the present invention;

[0077] Figure 4 It is a circular thermal diagram of the frequency of the top cover bolts of the present invention. DETAILED DESCRIPTION

[0078] The present invention will be further explained below in conjunction with the accompanying drawings and specific embodiments.

[0079] Example 1

[0080] The present invention provides a non-contact detection method for loosening of top cover bolts of a hydropower station turbine, such as Figure 1 As shown, the specific steps are as follows:

[0081] S1, by installing a camera above the turbine top cover, the vibration status of the top cover bolts is monitored and video images are obtained.

[0082] S2, using an optical flow algorithm to process the obtained video image and calculate the optical flow information of the bolt structure;

[0083] S3, calculating the bolt vibration response information based on the relationship between the structural optical flow information and the structural vibration response information;

[0084] S4, calculating the vibration frequencies of different bolts according to the vibration response information of different bolts, and constructing the frequency matrix of the top cover bolts;

[0085] S5, based on the frequency matrix of the top cover bolts, the frequency entropy of the frequency matrix is ​​calculated, and by comparing the magnitude of the frequency entropy before and after the bolts vibrate in real time, it is determined whether there is any looseness in the bolts;

[0086] S6, construct a circular heat map of different bolt frequencies of the top cover to locate the locations of loose bolts.

[0087] Example 2

[0088] On the basis of Example 1, Figure 1-2 As shown, S2 is specifically as follows: considering the large number of bolts on the top cover, the time-consuming and labor-intensive use of the labeled video processing technology, and the large number of image pixels and the low calculation efficiency of the dense optical flow algorithm, the sparse optical flow algorithm is used to calculate the optical flow information of the bolts;

[0089] Based on the constant brightness and small motion assumptions of the sparse optical flow algorithm, the optical flow formula is:

[0090] (1);

[0091] Where: and are the brightness gradients of the bolt vibration video image pixels in the horizontal and vertical directions, is the derivative of the image pixel brightness with respect to time t, u and v are the optical flows of image pixels in the horizontal and vertical directions respectively;

[0092] Considering the optical flow in formula (1) u and v is an unknown quantity and cannot be solved directly. Therefore, in the sparse optical flow algorithm, the pixel and its neighborhood are considered to have similar motions. P(P x ,P y ) and its neighbors W(W x ,W y ) The k pixels in the same image have similar motions, that is, there are k equations satisfying formula (1); thus, formula (1) can be expressed as:

[0093] (2);

[0094] In the formula, p 1 , p 2 , , p k The target pixel P Pixels in the neighborhood W, k Neighborhood K The number of pixels inside, and Neighborhood pixels p k Brightness gradients in horizontal and vertical directions, Neighborhood pixels p k The derivative of brightness with respect to time t, u and v are the optical flows of image pixels in the horizontal and vertical directions respectively;

[0095] Among them, the definition is:

[0096] , (3);

[0097] In the formula, A is all pixels in the neighborhood W (a total of k ) in the horizontal and vertical directions, b is all pixels in the neighborhood W (a total of k ) Brightness with respect to time t The matrix composed of the derivatives of ;

[0098] At this time, formula (2) can be expressed as:

[0099] (4);

[0100] In the formula, A is all pixels in the neighborhood W (a total of k ) in the horizontal and vertical directions, b is all pixels in the neighborhood W (a total of k ) Brightness with respect to time t The matrix composed of the derivatives of ; u and v are the optical flows of image pixels in the horizontal and vertical directions, respectively.

[0101] Among them, considering that formula (4) has k equations, two unknown quantities u and v , so the least squares method is used to solve it, and the optical flow of the target pixel in the horizontal and vertical directions can be calculated as:

[0102] (5);

[0103] In the formula, A T is the transposed matrix of matrix A, T is the transposed symbol, u and v are the optical flows of image pixels in the horizontal and vertical directions, respectively. b is all pixels in the neighborhood W (a total of k ) Brightness with respect to time t The matrix composed of the derivatives of ;

[0104] From formula (5), we can see that the accuracy of the target pixel optical flow calculation depends on the target gradient matrix , when the matrix When the reversibility is strong, the feature point corresponding to the pixel is easy to track with high accuracy;

[0105] The target pixel screening method is used to calculate the gradients of all pixels on the bolt structure in the horizontal and vertical directions, namely:

[0106] (6);

[0107] (7);

[0108] In formula (6), is the directional gradient of the image pixel in the horizontal direction, I represents the brightness value of the pixel in the image, represents the convolution operation, is the gradient operator in the horizontal direction;

[0109] In formula (7), is the directional gradient of the image pixel in the vertical direction, I represents the brightness value of the pixel in the image, represents the convolution operation, is the gradient operator in the vertical direction;

[0110] in, for:

[0111] (8);

[0112] In the formula, is the gradient operator in the horizontal direction, which is combined with the image brightness matrix I By performing convolution operation, the brightness gradient of the image pixel brightness in the horizontal direction can be obtained;

[0113] for:

[0114] (9);

[0115] In the formula, is the gradient operator in the vertical direction, which is combined with the image brightness matrix I By performing convolution operation, the brightness gradient of the image pixel brightness in the vertical direction can be obtained;

[0116] According to the gradient size of different pixels in the horizontal and vertical directions, the gradient amplitude is calculated and recorded as:

[0117] (10);

[0118] Finally, the target pixel is selected by calculating the gradient amplitude of any pixel of the turbine top cover bolt. Then, the optical flow of the target pixel is calculated by the optical flow algorithm to obtain the optical flow information of the bolt structure.

[0119] Example 3

[0120] On the basis of Example 2, Figure 1-4As shown, S3 is specifically as follows: In order to improve the accuracy of the measurement of the optical flow information of the bolt structure, the gradient amplitudes of all pixels on each bolt are calculated, and the five pixels with the largest gradient amplitudes are selected as target pixels to obtain the optical flow of the target pixels, and the optical flow values ​​of the five target pixels are averaged to obtain the optical flow information of the bolt structure, which is specifically expressed as:

[0121] (11);

[0122] Where: u(t) and v(t) are the optical flows of the bolt in the horizontal and vertical directions at time t, u i (t) and v i (t) Respectively indicate the number i The optical flow of the pixel at time t.

[0123] On this basis, according to the optical flow of the bolt at different times and the corresponding relationship between the optical flow and the vibration response, the vibration response information of the bolt structure at time t can be expressed as:

[0124] (12);

[0125] Where: and are the vibration response information of the bolt structure, is the time interval between two adjacent frames, k The camera uses frequency, u(t) and v(t) are the optical flows of the bolt in the horizontal and vertical directions at time t, respectively.

[0126] S4 is specifically as follows: considering that after the turbine top cover bolts are loosened, their vibration characteristics will change, that is, the frequency of the bolts will change, firstly, the turbine top cover bolts are numbered as 1, 2, ..., n;

[0127] Secondly, according to the vibration response information of different bolts, the vibration frequency of different bolts can be obtained by Fourier transform, which is expressed as:

[0128] (13);

[0129] Where: S n For the number n The bolt vibration response information, f n For the number n The vibration frequency of the bolt;

[0130] Finally, the frequency matrix of the top cover bolts is constructed by calculating the structure of different bolt vibration frequencies, which can be expressed as:

[0131] (14).

[0132] Example 4

[0133] On the basis of Example 3, S5 is specifically as follows: wherein the constructed frequency matrix is ​​a diagonal matrix, and its diagonal elements, that is, the singular values ​​of the matrix are the vibration frequencies of the bolts at different positions. Through the frequency matrix, the probability of the matrix singular value distribution can be obtained, which is recorded as:

[0134] (15);

[0135] Where: E q For the number q The probability of the bolt singular value distribution, f q For the number q The frequency of the bolts;

[0136] According to the probability of the bolt singular value distribution, the frequency entropy H of the bolt vibration can be calculated as:

[0137] (16);

[0138] In the formula, E q For the number q The probability of the singular value distribution of the bolt, H is the frequency entropy of the bolt vibration;

[0139] Among them, taking the initial vibration frequency entropy when the top cover bolts are not loose as a benchmark, by comparing the size relationship between the initial frequency entropy and the current frequency entropy in real time, it is judged whether the top cover bolts are loose, and an alarm is issued for abnormal looseness of the bolts.

[0140] S6 is specifically as follows: when an abnormal alarm of bolt loosening occurs, in order to facilitate the location of the loose bolt, the location of the loose bolt is realized by constructing a circular heat map of bolt frequency, and the frequency arrangement is arranged according to the actual position of the bolt;

[0141] In the present invention, it is considered that the frequency of a loose bolt may be abnormal, that is, it may be different from the vibration characteristics of surrounding bolts. Therefore, the position of the loose bolt can be identified by bolts 1, bolts 2, ..., bolt n in the frequency circular thermal diagram.

[0142] In addition, when the bolt structure is just loosened, it has little effect on its low-order frequency, and mainly affects its high-order frequency characteristics. Therefore, based on the camera video image acquisition frame rate k, the maximum frequency that can be collected is k / 2, that is, the first j-order frequencies of the turbine top cover bolts can be calculated, which are recorded as:

[0143] (17);

[0144] Where: f j is the highest order frequency that can be calculated, which satisfies f j <k / 2;

[0145] Finally, by drawing the frequency heat map of the first j-order top cover bolts and integrating the distribution size of the first j-order frequency heat map, the loose bolt detection and loose bolt positioning are completed.

Claims

1. A non-contact detection method for loose bolts on a turbine cover of a hydropower station, characterized in that: The specific steps include: S1, performing video monitoring on the vibration state of the top cover bolts to obtain video images; S2, using a sparse optical flow algorithm to process the video image obtained in S1 to obtain the optical flow information of the bolt structure; S3, obtaining vibration response information of multiple bolts based on the relationship between the structured optical flow information and the structural vibration response information; S4, obtaining multiple bolt vibration frequencies according to the multiple bolt vibration response information, and constructing a top cover bolt frequency matrix; S5, obtaining the frequency entropy of the frequency matrix through the frequency matrix of the top cover bolts, and judging whether there is looseness in the bolts by comparing the magnitude of the frequency entropy before and after the bolts vibrate in real time; S6, constructing a circular heat map of different bolt frequencies of the top cover, locating the positions of loose bolts, and completing the bolt loosening detection; The S2 is specifically: The sparse optical flow algorithm is used to calculate the optical flow information of the bolt. The specific formula is: (1); Where: and are the brightness gradients of the bolt vibration video image pixels in the horizontal and vertical directions, is the derivative of the image pixel brightness with respect to time t, u and v are the optical flows of image pixels in the horizontal and vertical directions respectively; Since the optical flows u and v in formula (1) are unknown quantities and cannot be solved directly, we assume that the target pixel P (P x , P y ) and its neighborhood W (W x , W y ) have similar motions, that is, there are k equations satisfying formula (1); Therefore, formula (1) can be expressed as: (2); In the formula, p 1, p 2, , p k for Target Pixels P Pixels in the neighborhood W, k Neighborhood K The number of pixels inside, and Neighborhood pixels p k Brightness gradients in horizontal and vertical directions, Neighborhood pixels p k The derivative of brightness with respect to time t, u and v are the optical flows of image pixels in the horizontal and vertical directions respectively; definition: , (3); In the formula, A is all pixels in the neighborhood W (a total of k ) in the horizontal and vertical directions, b is all pixels in the neighborhood W (a total of k ) Brightness with respect to time t The matrix composed of the derivatives of ; At this time, formula (2) can be expressed as: (4); In the formula, A is all pixels in the neighborhood W (a total of k ) in the horizontal and vertical directions, b is all pixels in the neighborhood W (a total of k ) Brightness with respect to time t The matrix composed of the derivatives of ; u and v are the optical flows of image pixels in the horizontal and vertical directions respectively; There are k equations in formula (4) and two unknown quantities u and v , the least square method is used to solve it, and the optical flow formula of the target pixel in the horizontal and vertical directions can be expressed as: (5); In the formula, A T is the transposed matrix of matrix A, T is the transposed symbol, u and v are the optical flows of image pixels in the horizontal and vertical directions, respectively. b is all pixels in the neighborhood W (a total of k ) Brightness with respect to time t The matrix composed of the derivatives of ; From formula (5), we can see that the accuracy of the target pixel optical flow calculation depends on the target gradient matrix , when the matrix When the reversibility is strong, the feature point corresponding to the pixel is easy to track with high accuracy; (6); (7); In formula (6), is the directional gradient of the image pixel in the horizontal direction, I Represents the brightness value of the pixels in the image. represents the convolution operation, is the gradient operator in the horizontal direction; In formula (7), is the directional gradient of the image pixel in the vertical direction, I Represents the brightness value of the pixels in the image. represents the convolution operation, is the gradient operator in the vertical direction; in, for: (8); In the formula, is the gradient operator in the horizontal direction, which is combined with the image brightness matrix I By performing convolution operation, the brightness gradient of the image pixel brightness in the horizontal direction can be obtained; for: (9); In the formula, is the gradient operator in the vertical direction, which is combined with the image brightness matrix I By performing convolution operation, the brightness gradient of the image pixel brightness in the vertical direction can be obtained; According to the magnitude of the directional gradient of different pixels in the horizontal and vertical directions, the gradient amplitude is calculated, which is specifically expressed as: (10); In the formula, is the gradient magnitude of the image pixel, is the directional gradient of the image pixel in the horizontal direction, is the directional gradient of the image pixel in the vertical direction; The gradient amplitudes of all pixels on each bolt are calculated, and the five pixels with the largest gradient amplitudes are selected as target pixels to obtain the optical flow of the target pixels. The optical flow values ​​of the five target pixels are averaged to obtain the optical flow information of the bolt structure, which is specifically expressed as: (11); Where: u(t) and v(t) are the optical flows of the bolt in the horizontal and vertical directions at time t, u i (t) and v i (t) Respectively indicate the number i The optical flow of the pixel at time t; S3 is specifically: According to the optical flow of the bolt at different times, combined with the corresponding relationship between the optical flow and the vibration response, the vibration response information of the bolt structure at time t can be expressed as: (12); Where: and are the vibration response information of the bolt structure, is the time interval between two adjacent frames, k The camera uses frequency, u(t) and v(t) are the optical flows of the bolt in the horizontal and vertical directions at time t respectively; S4 is specifically as follows: considering that after the turbine top cover bolts are loosened, their vibration characteristics will change, that is, the frequency of the bolts will change. First, the turbine top cover bolts are numbered 1, 2, ..., n; secondly, according to the vibration response information of different bolts, the vibration frequencies of different bolts can be obtained by Fourier transform, which is recorded as: (13); Where: S n For the number n The bolt vibration response information, f n For the number n The vibration frequency of the bolt; Finally, the frequency matrix of the top cover bolts is constructed by calculating the structure of different bolt vibration frequencies, which can be expressed as: (14); S5 is specifically as follows: the constructed frequency matrix is ​​a diagonal matrix, and its diagonal elements, that is, the singular values ​​of the matrix are the vibration frequencies of bolts at different positions. Through the frequency matrix, the probability of the matrix singular value distribution can be obtained, which is recorded as: (15); Where: E q For the number q The probability of the bolt singular value distribution, f q For the number q The frequency of the bolts; According to the probability of the bolt singular value distribution, the frequency entropy H of the bolt vibration can be calculated as: (16); In the formula, E q For the number q The probability of the singular value distribution of the bolt, H is the frequency entropy of the bolt vibration; Among them, taking the initial vibration frequency entropy when the top cover bolts are not loose as a benchmark, by comparing the size relationship between the initial frequency entropy and the current frequency entropy in real time, it is judged whether the top cover bolts are loose, and an alarm is issued for abnormal looseness of the bolts.

2. The non-contact detection method for loose bolts on a turbine cover of a hydropower station according to claim 1 is characterized in that: The S1 is specifically: By installing a camera above the turbine top cover, the vibration status of the top cover bolts is monitored via video and video images are obtained.

3. The non-contact detection method for loose bolts on a turbine cover of a hydropower station according to claim 1 is characterized in that: Specifically, S6 includes: when an abnormal alarm of bolt loosening occurs, in order to facilitate the location of the loose bolt, the location of the loose bolt is realized by constructing a circular heat map of bolt frequency, and the frequency arrangement is arranged according to the actual position of the bolt; since the frequency of the bolt will be abnormal after the bolt is loose, that is, it will be different from the vibration characteristics of the surrounding bolts, the position of the loose bolt is identified through the circular heat map of frequency; In addition, when the bolt structure is just loosened, it has little effect on its low-order frequency, but mainly affects its high-order frequency characteristics. Therefore, based on the camera video image acquisition frame rate k, the maximum frequency that can be collected is k / 2, that is, the first j-order frequencies of the turbine top cover bolts can be calculated, which are recorded as: (17); Where: f j is the highest order frequency that can be calculated, which satisfies f j <k / 2; Finally, by drawing the frequency heat map of the first j-order top cover bolts and integrating the distribution size of the first j-order frequency heat map, the loose bolt detection and loose bolt positioning are completed.

Citation Information

Patent Citations

  • Method for measuring galloping of transmission conductor on the basis of video difference analysis

    CN102155933A

  • Device and method for online monitoring looseness of bolts on transmission tower

    CN107976307A

  • Method for judging degree of looseness of bolt based on vibration mode information

    CN110044566A

  • Hydropower station electromechanical equipment fault early warning method based on quadruped robot

    CN117589404A

  • Pumped storage power station water turbine top cover bolt looseness detection method based on video analysis

    CN118470609A