Gasf-BP-based vibration signal pin fault detection method and system

Abstract

Disclosed are a GaSF-BP-based vibration signal pin fault detection method and system. The method of the present invention comprises: first, by means of a vibration sensor, collecting vibration data indicating a potential detachment of a power transmission tower pin. Then, employing a GaSF technique for raising the dimensionality of a one-dimensional vibration signal and converting same into a high-dimensional feature image. Then, by means of a BP neural network, performing recognition and classification on the generated GaSF image, and detecting a tower vibration signal in real time, so as to achieve the purpose of identifying pin detachment. The present invention integrates the accuracy of the GaSF method with respect to time series data feature characterizations and the image recognition rapidity of the BP neural network, thereby effectively improving detection effectiveness for pin detachment faults.

Description

GASF-BP-based vibration signal pin fault detection method and system TECHNICAL FIELD

[0001] The present application relates to the technical field of computer platform load balancing, in particular to a GASF-BP-based vibration signal pin fault detection method and system. BACKGROUND

[0002] In recent years, with the rapid development of the national economy, China's power industry has entered a new period of large-scale construction. As of the end of 2022, the total length of 110kV and above transmission line loops in the country reached 1.2 million kilometers. However, the rapid development of power construction has also brought a series of problems, especially in the operation of transmission lines. Factors such as harsh environment (hail, acid rain, strong wind), mechanical vibration and material aging have caused defects in power components, such as pin shedding, insulator self-explosion, stay tower stay wire fracture, and line damage caused by external forces. It has seriously affected the normal operation of the transmission line and caused great losses to the social economy.

[0003] Pins are important components that connect the conductor and the support (such as a pole or tower) of a transmission line. They play a role in fixing and supporting the conductor, ensuring the normal operation of the transmission line. However, pins are exposed to harsh environmental conditions for a long time, such as high temperature, humidity, and chemicals, which can cause metallic corrosion and weaken their structural strength, increasing the risk of shedding when working in external environments. In addition, as a key component of the transmission line, pins are affected by natural weather factors such as wind, ice and snow, lightning, etc., which can cause abnormal mechanical stress on the pins, leading to fracture or deformation. In addition, defects such as unqualified materials and improper processes may occur during the production and manufacturing of pins, which can easily cause faults.

[0004] Once a pin fails, it may gradually loosen or even fall off, and the structural rigidity of the transmission tower will also decrease, posing a serious threat to the safety of the tower. If this problem cannot be discovered and addressed in a timely manner, it may lead to instability or even collapse of the power system, causing significant impact on social and economic activities and public life, and in severe cases, it may lead to the tilting or collapse of the tower. In recent years, similar accidents have occurred from time to time, causing serious economic losses to the power grid. Therefore, to ensure the safety of the transmission tower, it is necessary to detect pin loosening in advance and maintain it in a timely manner to prevent accidents.

[0005] In the past few decades, power grid companies mainly use traditional manual inspection methods for power inspection. The inspection personnel need to hold instruments and climb the tower. The harsh geological environment not only affects the work efficiency of the inspection personnel, but also causes certain personal injury to the inspection personnel. Moreover, the inspection labor intensity is large, the inspection quality is affected by the subjective factors of the operating personnel, and the inspection range is greatly limited in the harsh environment. In recent years, the unmanned aerial vehicle is widely used in the power department due to its small size, low cost and good mobility, so as to reduce the work intensity of the operating personnel and improve the inspection range and efficiency. Some detection devices (such as laser sensors, visible light / infrared / ultraviolet image acquisition equipment, range finders and the like) are integrated into the unmanned aerial vehicle to detect the key components on the power system. However, it is difficult for the unmanned aerial vehicle to collect images for the small pin components of the transmission line, and the recognition accuracy is not high, so the method of deploying a vibration sensor is selected to achieve the purpose of accurately identifying the pin shedding. SUMMARY

[0006] In view of the above problems, the present application is proposed.

[0007] Therefore, the technical problem solved by the present application is that the existing cloud service platform computing method has the problems of large power consumption loss, load rate reward and punishment, high cost, and how to allocate task requests to each host to achieve load balancing optimization.

[0008] To solve the above technical problems, the present application provides the following technical scheme: a vibration signal pin fault detection method based on GASF-BP, comprising: collecting original vibration signals, and preprocessing the collected original vibration signals; performing coordinate system conversion processing on the time series data in the Cartesian coordinate system, converting the Cartesian coordinate system into the polar coordinate system, and understanding the relationship of the time series data with the angle of the polar coordinate system; using a back propagation neural network as an image recognition, training the BP neural network, and converting the vibration signals collected by the sensor into GASF images and inputting them into the trained neural network for recognition.

[0009] As a preferred scheme of the vibration signal pin fault detection method based on GASF-BP, wherein: the collecting of original vibration signals includes using a vibration exciter to generate low-frequency high-energy elastic waves in the power transmission tower, the elastic waves propagate in the rigid structure of the power transmission tower and cause the vibration of the whole or part of the tower, and the vibration signals containing the running state information of the pin are generated; a vibration sensor is installed at the other end of the power transmission tower, the sensor receives the generated vibration signals and converts the vibration signals into electrical signals, and the platform parameters of the cloud service end are collected.

[0010] As a preferred scheme of the GASF-BP-based vibration signal pin failure detection method, the pre-processing of the collected original vibration signal comprises filtering, noise reduction and sampling.

[0011] The filtering comprises that, for an input vibration signal x(n), after passing through a parameter-adjustable transverse filter, an output is y(n), and an LMS algorithm automatically adjusts parameters of the filter according to an error between the output signal y(n) of the filter and an expected signal d(n).

[0012] The input vibration signal is a vector X(n) = [x(n), x(n-1),..., x(n-L)] T , and an output of the LMS adaptive filter is y(n), which is expressed as:

[0013] wherein W(n) is a weight coefficient of the filter, L is an order of the filter, i represents a specific order of the filter, T represents transposition, w i represents the weight coefficient of the i-th order.

[0014] An error signal e(n) is defined as e(n) = d(n)-y(n) = d(n)-W T (n)X(n)

[0015] A cost function J(n) is defined by taking a minimum value of the mean square of the error signal e(n) as an optimal statistical criterion, and the cost function J(n) is expressed as: J(n) = E[e 2 (n)]

[0016] A change of the cost function J(n) with time represents a convergence speed and a convergence accuracy of the LMS filter, and when the cost function J(n) takes a minimum value, the weight coefficient of the LMS filter is an optimal value; by using a steepest descent algorithm, the LMS filter weight vector W(n) is adjusted along a steepest descent direction of a performance surface, a minimum point of the performance surface is searched, and an optimal weight vector W opt is solved, and the steepest descent direction is expressed as:

[0017] wherein, represents a partial derivative of the cost function to the weight vector.

[0018] Therefore, an update formula of the LMS adaptive filter weight vector is calculated as:

[0019] wherein μ is a step factor, and μ>0 is always a positive number, and W(n+1) is an updated weight vector.

[0020] As a preferred scheme of the vibration signal pin failure detection method based on GASF-BP, wherein the noise reduction includes a vibration signal s received by a vibration sensor and to be analyzed, expressed as:

[0021] wherein, s(t) represents a raw vibration signal received by the sensor, n(t) represents noise signal contained in the fault signal, and s(t) represents the vibration signal to be analyzed;

[0022] The vibration signal generated by the power transmission tower is discretely sampled to obtain N discrete signals f(n), n=0, 1, …, N-1, and the wavelet transform coefficient of f(t) is W f (j, k), expressed as:

[0023] In the actual collection and calculation of the vibration signal, the following equation is used for recursive calculation to obtain a recursive implementation method: S f (j+1, k) = S f (j, k) * h(j, k) W f (j+1, k) = S f (j, k) * g(j, k)

[0024] wherein, S f (j+1, k) represents an approximation part of the signal at scale j+1 and position k, k represents the position of the signal, j represents the scale, h and g represent low-pass filters and high-pass filters of the scale function φ(t) and the wavelet function ψ(t) respectively, S f (0, k) represents the original signal f(k), S f (j, k) represents the approximation coefficient at scale j, and W f (j, k) represents the wavelet coefficient, and the reconstruction formula of the wavelet transform is expressed as:

[0025] wherein, S f (j-1, k) represents an approximation signal at a finer scale j-1 and position k, represents the coefficient of the reconstruction filter.

[0026] As a preferred scheme of the vibration signal pin failure detection method based on GASF-BP, wherein the time series data in the Cartesian coordinate system is subjected to coordinate system conversion processing, including according to the vibration signal, n time series data points are grouped to form a 1*n-dimensional time series vector X={X1, X2, …, X n}, readjust the scale of X, scale all elements in vector X to [-1, 1], and the scaling formula is represented as:

[0027] Wherein, x i is the original element in x, is the scaled element, maxX represents the coefficient of the reconstruction filter, and minX represents the minimum value in vector X;

[0028] Convert the scaled acceleration time sequence into polar coordinates, and use the angle and radius of the polar coordinates to represent the value of acceleration and the corresponding timestamp respectively;

[0029] Encode the scaled acceleration value with the angle, the value of acceleration is between [0, π], and the angle cosine function is used for encoding, the acceleration value is mapped to the range of [0, 1], and then multiplied by π to obtain the corresponding angle value;

[0030] The timestamp is used as the radius of the polar coordinates, normalized to the range of [0, 1], and then multiplied by a suitable constant, the expression is as follows:

[0031] Wherein, t i is the corresponding timestamp of x i , M is the normalization factor, N is the constant factor of the normalized polar coordinate system span, is the radian representation of x i in the polar coordinate system, i.e., the angle cosine, r i is the corresponding radius, and θ i is the angle in the polar coordinates, is the set of scaled elements, is the element inverse cosine calculation;

[0032] Reconstruct the vibration time sequence by using GAF to obtain the sequence X, import GAF to obtain the two-dimensional time sequence diagram, i.e., Gram and Angle Field, the expression is as follows:

[0033] Wherein, A GASF (i,j) is represented as the Gram and Angle Field matrix, cos (θ i +θ j ) is represented as the matrix element expression, i and j are represented as the polar coordinate angle number.

[0034] As a preferred scheme of the vibration signal pin fault detection method based on GASF-BP, the back propagation neural network is used as image recognition, which includes network construction and feature extraction;

[0035] The network construction includes constructing a back propagation neural network model for image recognition, generally containing an input layer, a hidden layer and an output layer; the input layer receives a feature vector as input, the hidden layer contains a plurality of neurons, and the output layer represents the result of classification;

[0036] The feature extraction adopts a color histogram to describe the color feature of the image, a gray level co-occurrence matrix to extract the texture feature of the image, and an edge detection algorithm to extract the contour feature of the image.

[0037] The color feature color histogram divides the color space of the image into a plurality of regions, and counts the number or proportion of pixels in each region to obtain a histogram representing the color distribution of the image; the color histogram reflects the distribution of different colors in the image, and the description and comparison of the color feature.

[0038] The texture feature represents the texture structure and distribution of pixels in the image, and reflects the visual feature of homogeneity in the image; the gray level co-occurrence matrix calculates the relationship and frequency of gray values between pixels in the image, and obtains the texture feature by counting the co-occurrence matrix between pixel pairs.

[0039] The contour feature describes the boundary shape and contour information of an object or region, and uses a Canny edge detection algorithm to calculate the rate of change of pixel values in the image to detect boundary pixels.

[0040] As a preferred scheme of the vibration signal pin fault detection method based on GASF-BP, the training of the BP neural network includes updating the weights and biases in the network using a back propagation algorithm to minimize the error between the predicted results and the actual labels, so that the network can accurately distinguish whether the pin of the power transmission tower is detached.

[0041] The neural network algorithm simultaneously satisfies local convergence and has a gradient descent method, and adopts a stable L-M algorithm, which is represented as: x s+1 = x s -(H+αI) -1 G

[0042] Wherein, H is a matrix of multi-dimensional vectors, G is the first-order gradient of multi-dimensional vectors, x s is the function extreme value of the previous step, alpha is the step size, I is the unit matrix, and x s+1 is represented as the function value after iteration.

[0043] A further object of the present application is to provide a GASF-BP-based vibration signal pin failure detection system, which ensures the quality and reliability of the input data through the high-precision sensor of the original vibration acquisition module and the preprocessing step, provides noise-free vibration signals for subsequent processing, converts the Cartesian coordinate system to the polar coordinate system through the coordinate system conversion processing module, reveals the periodicity and trend of the time series data, enhances the interpretability of the data, and uses a well-trained neural network and GASF image input to analyze and accurately identify the vibration signal through the BP neural network identification module, thereby improving the intelligent diagnosis and predictive maintenance capabilities of the system, and the coordinated work of the three modules improves the efficiency and accuracy of vibration signal processing, and provides strong technical support for the condition monitoring and failure prediction of complex systems.

[0044] The application provides a GASF-BP-based vibration signal pin failure detection system, which comprises an original vibration acquisition module, a coordinate system conversion processing module and a BP neural network identification module.

[0045] The original vibration acquisition module acquires original vibration signals and pre-processes the acquired original vibration signals.

[0046] The coordinate system conversion processing module converts time series data in a Cartesian coordinate system, converts the Cartesian coordinate system into a polar coordinate system, and understands the relationship of the time series data by using the angle of the polar coordinate system.

[0047] The BP neural network identification module uses a back propagation neural network as an image recognition, trains the BP neural network, converts the vibration signals collected by the sensor into GASF images, and inputs the GASF images into the trained neural network for identification.

[0048] A computer device comprises a memory and a processor, the memory stores a computer program, and the processor implements the steps of the method described above when executing the computer program.

[0049] A computer-readable storage medium stores a computer program, and the computer program is executed by a processor to implement the steps of the method described above.

[0050] The GASF-BP-based vibration signal pin failure detection method provided by the application reduces the power consumption in the network, measures the load balancing between the cloud network and the server (host) using the fitness function, converts the load balancing problem into an optimization problem, and is suitable for processing resource optimization. The results of the particle swarm are used as the initial population of the genetic algorithm, and the application achieves better results in execution cost, load balancing and completion time. BRIEF DESCRIPTION OF DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0052] Fig. 1 is a schematic diagram of vibration sensor installation of a GASF-BP-based vibration signal pin fault detection method according to an embodiment of the present application.

[0053] Fig. 2 is a block diagram of image recognition process of a GASF-BP-based vibration signal pin fault detection method according to a second embodiment of the present application.

[0054] Fig. 3 is a block diagram of bookstore road vibration signal pin fault detection method based on deep learning of a GASF-BP-based vibration signal pin fault detection method according to the second embodiment of the present application. DETAILED DESCRIPTION

[0055] In order to make the above-mentioned objects, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without any creative effort should be within the scope of protection of the present application.

[0056] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the concept of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0057] Embodiment 1

[0058] As shown in Fig. 1 and Fig. 2, according to an embodiment of the present application, a GASF-BP-based vibration signal pin fault detection method is provided, which comprises:

[0059] S1: collecting original vibration signals and pre-processing the collected original vibration signals;

[0060] It should be noted that filtering, noise reduction and sampling are performed.

[0061] Further, the filtering is represented as:

[0062] It should be noted that the adaptive filter is used for processing, the coefficients of the adaptive filter can be automatically and continuously adapted to the vibration signal to obtain the best expected response, for the input vibration signal x(n), after passing through the parameter adjustable transverse filter, the output is y(n), the LMS algorithm automatically adjusts the parameters of the filter according to the error between the output signal y(n) of the filter and the expected signal d(n);

[0063] The input vibration signal is a vector X(n) = [x(n), x(n-1),..., x(n-L)] T The output of the LMS adaptive filter is y(n), which is represented as:

[0064] Wherein, W(n) is the weight coefficient of the filter, L is the order of the filter, i represents the specific order of the filter, T represents the transpose, and wi represents the weight coefficient of the i-th order;

[0065] The error signal e(n) is defined as e(n) = d(n) - y(n) = d(n) - W T (n)X(n)

[0066] The mean square minimum of the error signal e(n) is taken as the optimal statistical criterion, and the cost function J(n) expression is defined as: J(n) = E[e 2 (n)]

[0067] The change of the cost function J(n) with time represents the convergence speed and convergence accuracy of the LMS filter, when the cost function J(n) takes the minimum value, the weight coefficient of the LMS filter is the optimal value; using the steepest descent algorithm, the LMS filter weight vector W(n) is adjusted along the steepest descent direction of the performance surface, and the minimum point of the performance surface is searched, The filter weight vector w(n) is continuously adjusted and updated to approach the optimal weight vector W opt The optimal weight vector W opt The steepest descent direction is represented as:

[0068] Wherein, The partial derivative of the cost function to the weight vector is represented as;

[0069] Then the update formula of the LMS adaptive filter weight vector is calculated as:

[0070] Wherein, μ is a step factor, and μ>0 is always positive, The steepest descent direction is represented as, and W(n+1) is the updated weight vector;

[0071] Further, the noise reduction is represented as follows: the filtered signal is denoised by using the wavelet threshold denoising method, the filtered signal is wavelet transformed to decompose the signal into wavelet coefficients of different scales and frequencies, the wavelet coefficient amplitude of the vibration signal is large after wavelet decomposition, and the wavelet coefficient amplitude of the noise is small, the selection of the wavelet coefficient can be realized by setting the threshold, the coefficients smaller than the threshold are regarded as the wavelet coefficients of the noise, the part of the coefficients is removed, the remaining coefficients are regarded as the wavelet coefficients of the effective signal and are reserved, the reserved wavelet coefficients are inverse wavelet transformed to reconstruct the denoised signal, the denoising processing reduces the random noise in the signal and improves the clarity and readability of the signal.

[0072] The vibration signal s received by the vibration sensor for analysis is represented as follows:

[0073] Wherein, s(t) represents the original vibration signal received by the sensor, n(t) represents the noise signal contained in the fault signal, and s(t) represents the vibration signal for analysis.

[0074] The vibration signal generated by the transmission tower is discretely sampled to obtain N discrete signals f(n), n = 0, 1, …, N-1, and the wavelet transform coefficient is W f (j, k), represented as follows:

[0075] The double-scale equation is used for recursive calculation to obtain the recursive implementation method: S f (j+1, k) = S f (j, k) * h(j, k) W f (j+1, k) = S f (j, k) * g(j, k)

[0076] Wherein, S f (j+1, k) is represented as S f (0, k) represents the original signal f(k), S f (j, k) represents the approximation coefficient at the scale j, and W f (j, k) represents the wavelet coefficient, and the wavelet transform reconstruction formula is represented as follows:

[0077] Wherein, S f (j-1, k) represents the approximation signal at the fine scale j-1 and the position k, is represented as the coefficient of the reconstruction filter.

[0078] S2: Perform coordinate system transformation on the time series data in the Cartesian coordinate system, converting the Cartesian coordinate system into the polar coordinate system, and understand the relationship of the time series from the perspective of the polar coordinate system;

[0079] Furthermore, coordinate system transformation processing of time series data in the Cartesian coordinate system, including Gram and Angular Field (GAF), is an effective method for studying the relationship between time series data and timestamps. By converting time series data into matrix form, the closeness between data signals and time can be clearly revealed, showcasing the internal structure and complexity of the data. The core idea of ​​the GAF method is to combine the temporal and numerical information in time series data, reflecting the relationship between data through changes in angles, thereby more intuitively understanding the characteristics and patterns of time series data, deeply analyzing and mining its dynamic changes and evolution processes, and providing stronger support for data analysis and prediction.

[0080] Based on the vibration signal, n time series data points are used to form a 1*n-dimensional time series vector X = {X1, X2, ..., X...} n}, readjust the scale of X, scaling all elements of vector X to the range [-1, 1]. The scaling formula is expressed as:

[0081] Where, x i It is an element that was originally in x. These are scaled elements, where maxX represents the maximum value in vector X and minX represents the minimum value in vector X.

[0082] The scaled acceleration time series is converted into polar coordinates, and the angle and radius of the polar coordinates are used to represent the acceleration value and the corresponding timestamp, respectively.

[0083] Furthermore, transforming the Cartesian coordinate system into a polar coordinate system, it can be represented as:

[0084] The scaled acceleration values ​​are encoded using angles. The acceleration values ​​are between [0, π]. The angle cosine function is used to encode the acceleration values, mapping them to the range of [0, 1]. Then, the values ​​are multiplied by π to obtain the corresponding angle values.

[0085] Using the timestamp as the radius of the polar coordinates, normalizing it to a range between [0,1], and then multiplying it by a suitable constant, we get:

[0086] Among them, t i For x iCorresponding timestamp, M is a normalization factor, N is a constant factor for normalizing polar coordinate system span, is x i In polar coordinates, the radian representation is the angle cosine, r i is the corresponding radius, θ i is expressed as an angle in polar coordinates, is expressed as a set of scaled elements , is expressed as an element ,

[0087] Further, the relationship of the time series sequence is understood by the angle of the polar coordinate system, the vibration time series is reconstructed by using GAF, the sequence X is imported into GAF to obtain a two-dimensional time series graph, that is, Gram and Angle Field, which is expressed as:

[0088] where A GASF (i,j) is expressed as a Gram and Angle Field matrix, cos(θ i +θ j ) is expressed as a matrix element expression, i, j is expressed as a polar coordinate angle sequence number.

[0089] S3: As shown in FIG. 2, a back propagation neural network is used as image recognition, a GASF image is prepared, color features, texture features and contour features of the image are extracted after the neural network, the BP neural network is trained, and the vibration signal collected by the sensor is converted into a GASF image and input into the trained neural network for recognition.

[0090] Further, a back propagation neural network is used as image recognition.

[0091] It should be noted that the network is constructed, and features are extracted.

[0092] Further, the network construction includes constructing a back propagation neural network model for image recognition, which usually includes an input layer, a hidden layer and an output layer; the input layer receives a feature vector as input, the hidden layer includes multiple neurons, and the output layer represents the result of classification;

[0093] Further, the feature extraction includes using a color histogram to describe the color features of the image, using a gray level co-occurrence matrix to extract the texture features of the image, and using an edge detection algorithm to extract the contour features of the image.

[0094] Further, the color features include dividing the color space of the image into a plurality of regions by using a color histogram, counting the number or proportion of pixels in each region, and obtaining a histogram representing the color distribution of the image; the color histogram reflects the distribution of different colors in the image, and the description and comparison of color features.

[0095] Further, the texture feature includes the texture structure and the texture distribution between pixels in the image, reflects the visual feature of the homogeneity in the image, and the gray level co-occurrence matrix calculates the relationship and the appearance frequency of the gray value between pixels in the image.

[0096] Further, the contour feature includes the boundary shape and the contour information of the object or the region, and the Canny edge detection algorithm is used to calculate the change rate of the pixel value in the image to detect the boundary pixels.

[0097] Further, the BP neural network is trained, and the vibration signal collected by the sensor is converted into a GASF image and input into the trained neural network for identification, which is represented as:

[0098] It should be noted that the back propagation algorithm is used to update the weights and biases in the network.

[0099] The error between the prediction result and the actual label is minimized, so that the network can accurately distinguish whether the pin of the power tower is dropped.

[0100] The neural network algorithm satisfies the local convergence and has the gradient descent method, and the L-M algorithm with stable performance is used, which is represented as: x s+1 = x s -(H+αI) -1 G

[0101] Wherein, H is the matrix of the multi-dimensional vector, G is the first order gradient of the multi-dimensional vector, x s is the function extreme value of the last step, alpha is the step size, and I is the unit matrix.

[0102] The larger the alpha value is, the closer the algorithm is to the steepest descent method, and the smaller the alpha value is, the closer the algorithm is to the Gauss-Newton algorithm.

[0103] The performance of the trained neural network model is evaluated using the test data set, the test data is input into the network for prediction, and then compared with the actual label, the accuracy and recall rate evaluation indicators are calculated to measure the performance of the model.

[0104] The signal collected by the vibration sensor is converted into an image and input into the trained neural network model for real-time state recognition, and the prediction result can be obtained through feature extraction of the image and forward propagation of the neural network, so as to realize detection of whether the pin is abnormally dropped.

[0105] Embodiment 2

[0106] As shown in Figure 3, an embodiment of the present application provides a GASF-BP based vibration signal pin fault detection system, which comprises:

[0107] The original vibration collection module: through the high-precision sensor collection equipment, the module can ensure that the vibration signal obtained has high fidelity and reliability, and the preprocessing step further improves the data quality, removes possible noise interference, and lays a solid foundation for subsequent analysis and processing.

[0108] The coordinate system conversion processing module: converts the time series data in the Cartesian coordinate system into the polar coordinate system, not only simplifies the data structure, but also makes the periodicity and trend of the time series data more obvious, which helps to better understand the dynamic characteristics and periodic changes of the vibration signal, and provides a new perspective for signal analysis and feature extraction.

[0109] The BP neural network recognition module: using the strong learning ability of the back propagation neural network (BPNN), the module can effectively identify complex patterns and features from the vibration signal, and through the conversion of the vibration signal into the GASF image format and the combination of the image recognition ability of the neural network, the system can realize high-accuracy vibration signal classification and anomaly detection, thereby improving the intelligent diagnosis and predictive maintenance ability of the system.

[0110] Embodiment 3

[0111] One embodiment of the present application, which is different from the first two embodiments, is:

[0112] If the function is realized in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the present application or the part of the prior art that essentially contributes or the part of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device) to execute all or part of the steps of the method described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various program code storage media.

[0113] The logic and / or steps represented in flow diagrams or otherwise described herein, for example, can be considered as a sequence of executable instructions, and can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions, or a combination of the above. For the purposes of this specification, a "computer-readable medium" can be any apparatus that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device.

[0114] More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection having one or more wires (electrical connections), a portable computer diskette (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can also be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, for example, via optical scanning of the paper or other medium, then compiled, interpreted, or otherwise processed in a suitable manner, if necessary, and stored in a computer memory.

[0115] It should be understood that aspects of the application can be implemented in hardware, software, firmware or combinations thereof. In the above embodiments, the various steps or methods can be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, and in another embodiment, any of the following technologies, known in the art, or a combination thereof, can be used: discrete logic circuitry having logic gates for implementing logic functions on data signals, application specific integrated circuits having appropriate combinational logic gates, programmable gate arrays (PGA), field programmable gate arrays (FPGA), and the like.

[0116] Example 4

[0117] For an embodiment of the present application, a GASF-BP-based vibration signal pin fault detection method is provided. In order to verify the beneficial effects of the present application, economic benefit calculation and simulation experiments are used for scientific demonstration.

[0118] The method realizes the judgment of the working state of the transmission tower pin and the fault warning of the transmission line by collecting the vibration signal of the transmission tower pin and performing preprocessing, feature extraction, feature selection, and model establishment. The experimental results show that the method has the following characteristics:

[0119] 1) The accuracy of detection is improved: by combining the GASF dimensionality method, it can be more accurately judged whether the pin has abnormal shedding, avoiding the misjudgment and missed judgment in the traditional method, and improving the accuracy of detection.

[0120] 2) The detection efficiency is improved: this paper introduces a kind of vibration signal pin fault detection method based on GASF-BP, which can efficiently identify the working state that may cause the pin shedding of the transmission line, and can timely send early warning notice to the relevant staff, thereby effectively improving the detection efficiency of the pin shedding of the transmission line.

[0121] The effect of the pin shedding detection method and the detection method based on unmanned aerial vehicle image recognition is shown in Table 1.

[0122] Table 1 Accuracy of different methods

[0123] It should be noted that the above examples are only used to illustrate the technical solutions of the present application and are not limited. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, and they should be covered in the scope of the claims of the present application.

Claims

1. A GASF-BP based vibration signal pin failure detection method, characterized in that, The method comprises the following steps: Collecting original vibration signals, and preprocessing the collected original vibration signals; Converting the time series data in the Cartesian coordinate system into the polar coordinate system, and understanding the relationship of the time series data by the angle of the polar coordinate system; Using the back propagation neural network as image recognition, training the BP neural network, and converting the vibration signals collected by the sensor into GASF images and inputting them into the trained neural network for recognition.

2. A GASF-BP based method for detecting a failure of a vibratory signal pin as claimed in claim 1, wherein: The collecting of the original vibration signals comprises generating low-frequency high-energy elastic waves in the power transmission tower by using a vibration exciter, propagating the elastic waves in the rigid structure of the power transmission tower, causing the vibration of the whole or part of the tower, and generating vibration signals containing the pin running state information; installing a vibration sensor at the other end of the power transmission tower, receiving the generated vibration signals, converting the vibration signals into electrical signals, and collecting the platform parameters of the cloud server.

3. A GASF-BP based method for detecting a failure of a vibratory signal pin as claimed in claim 2, wherein: The preprocessing of the collected original vibration signals comprises filtering, noise reduction and sampling; The filtering comprises inputting the vibration signal x(n) through a parameter-adjustable transverse filter, and outputting y(n); the LMS algorithm automatically adjusts the parameters of the filter according to the error between the output signal y(n) of the filter and the expected signal d(n); The input vibration signal is a vector X(n) = [x(n), x(n-1),..., x(n-L)] T The output of the LMS adaptive filter is y(n), which is expressed as: wherein W(n) is a weight coefficient of the filter, L is the order of the filter, i is the specific order of the filter, T is a transpose, w i is the weight coefficient of the i-th order; The error signal e(n) is defined as e(n), and the expected signal d(n) is defined as d(n); e(n) = d(n) - y(n) = d(n) - W T (n) X(n) The mean square minimum value of the error signal e(n) is taken as the optimal statistical criterion, and the cost function J(n) is defined as: J(n) = E[e 2 (n)] The change of the cost function J(n) with time characterizes the convergence speed and convergence accuracy of the LMS filter, and the weight coefficient of the LMS filter is the optimal value when the cost function J(n) takes the minimum value; the steepest descent algorithm is used to adjust the weight vector W(n) of the LMS filter along the steepest descent direction of the performance surface, search the minimum point of the performance surface, and solve the optimal weight vector Wopt, and the steepest descent direction is represented as: wherein The partial derivative of the cost function with respect to the weight vector is represented as The update formula of the LMS adaptive filter weight vector is calculated as: wherein μ is a step size factor and μ > 0 is always positive, is the steepest descent direction, and W(n+1) is the updated weight vector.

4. A GASF-BP based method for detecting a failure of a vibratory signal pin as claimed in claim 3, characterized in that: The noise reduction comprises a vibration signal s received by the vibration sensor to be analyzed, denoted as: wherein The vibration signal to be analyzed is represented as n(t), the noise signal contained in the fault signal is represented as n(t), and the original vibration signal received by the sensor is represented as s(t); The vibration signals generated by the power transmission tower are discretely sampled to obtain N discrete signals f(n), n = 0, 1,..., N-1, and the wavelet transform coefficient is W f (j, k) is represented as: The double-scale equation is used for recursive calculation to obtain the recursive implementation method: S f (j+1,k) = S f (j,k)*h(j,k) W f (j+1,k) = S f (j,k)*g(j,k) where S f (j + 1, k) is the approximation part of the signal at scale j + 1 and position k, k is the signal position, j is the scale, h and g are the low pass filter and high pass filter of the scaling function φ(t) and the wavelet function ψ(t) respectively, S f (0, k) represents the original signal f(k), S f (j, k) represents the approximation coefficients at scale j, W f (j, k) represents the wavelet coefficients, and the reconstruction formula of the wavelet transform is represented as: where S f (j-1,k) denotes the approximation signal at the finer scale j-1 and position k, The coefficients of the reconstruction filter are represented as 5. A GASF-BP based method for fault detection of a vibrating signal pin as claimed in claim 4, wherein: The coordinate system transformation processing of the time series data in the Cartesian coordinate system includes forming a 1*n-dimensional time series vector X = {X1, X2, ..., X...} based on n time series data points given by the vibration signal. n }, readjust the scale of X, scaling all elements of vector X to the range [-1, 1]. The scaling formula is expressed as: wherein x i is an element originally in x, is the scaled element, maxX represents the maximum value in the vector X, and minX represents the minimum value in the vector X; The scaled acceleration time series is converted into polar coordinates, and the angle and radius of the polar coordinates are used to represent the value of the acceleration and the corresponding time stamp, respectively; The scaled acceleration value is encoded by the angle, the value of the acceleration is between [0, π], and the angle cosine function is used for encoding, so as to map the acceleration value to the range of [0, 1] and multiply it by π to obtain the corresponding angle value; The timestamp is used as the radius of the polar coordinate, normalized to the range [0, 1], multiplied by a constant, denoted as: where t i is x i the corresponding time stamp, M is a normalization factor, and N is a constant factor normalizing the span of the polar coordinate system, is x i In polar coordinates, the arc length is given by the angle cosine, r i is the corresponding radius, and θ i is expressed as an angle in polar coordinates, representing the scaled element a set of the plurality of images, Represented as an element The inverse cosine calculation is performed; The time series of vibration is reconstructed by GAF, and the sequence X is introduced into GAF to obtain a two-dimensional time series diagram, i.e. Gram and angle field, which is expressed as: where A GASF (i,j) represents the Gram and the angle field matrix, cos(θ i +θ j ) represents the matrix element expression, i represents, j represents the polar coordinate angle number.

6. A GASF-BP based method for detecting a failure of a vibratory signal pin as claimed in claim 5, wherein: The back propagation neural network is used as image recognition, which comprises network construction and feature extraction. The network construction comprises constructing a back propagation neural network model for image recognition, which comprises an input layer, a hidden layer and an output layer; the input layer receives a feature vector as input, the hidden layer comprises a plurality of neurons, and the output layer represents the classification result; The feature extraction uses a color histogram to describe the color features of the image, a gray level co-occurrence matrix to extract the texture features of the image, and an edge detection algorithm to extract the contour features of the image. The color feature includes a color histogram, which divides the color space of an image into several regions, counts the number or proportion of pixels in each region, and obtains a histogram representing the color distribution of the image; the color histogram reflects the distribution of different colors in the image, and the description and comparison of the color feature; The texture feature includes the texture structure and distribution between pixels in the image, reflecting the visual feature of homogeneity in the image; a gray level co-occurrence matrix calculates the relationship and frequency of gray values between pixels, and obtains the texture feature by counting the co-occurrence matrix between pixel pairs; The contour feature includes the boundary shape and contour information of an object or region, and a Canny edge detection algorithm is used to calculate the rate of change of pixel values in the image to detect boundary pixels.

7. A GASF-BP based method for detecting a failure of a vibratory signal pin as claimed in claim 6, wherein: The training of the BP neural network includes updating the weights and biases in the network using a backpropagation algorithm to minimize the error between the predicted results and the actual labels, so that the network can accurately distinguish whether the power tower pin has fallen off; The neural network algorithm satisfies local convergence and has a gradient descent method, and uses a stable L-M algorithm, which is represented as: x s+1 = x s - (H + aI) -1 G where H is a matrix of multi-dimensional vectors, G is a first order gradient of multi-dimensional vectors, x s is the function extremum of the previous step, a is the step size, I is the identity matrix, x s+1 is expressed as the function value after iteration.

8. A system for detecting a failure of a vibrating signal pin using a GASF-BP based method according to any one of claims 1 to 7, characterized in that: The method comprises an original vibration acquisition module, a coordinate system conversion processing module, and a BP neural network identification module. The original vibration acquisition module acquires original vibration signals and pre-processes the acquired original vibration signals. The coordinate system conversion processing module performs coordinate system conversion processing on time series data in a Cartesian coordinate system, converts the Cartesian coordinate system into a polar coordinate system, and understands the relationship of the time series data in terms of the angle of the polar coordinate system. The BP neural network identification module uses a backpropagation neural network as an image recognition method, trains the BP neural network, converts the vibration signals collected by the sensor into a GASF image, and inputs the GASF image into the trained neural network for identification. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. The processor executes the computer program to implement the steps of the method of any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 7.

Images