Track fault diagnosis method and system

Abstract

The invention discloses an orbit fault diagnosis method and system, and the method comprises the steps: calculating a sample entropy, an energy entropy, a power spectrum and a permutation entropy of an optimal solution # imgabs0 #, selecting an optimal entropy from the sample entropy, the energy entropy, the power spectrum and the permutation entropy according to a preset optimal entropy weight selection strategy, and extracting typical features according to the optimal entropy; an improved CNN-BiLSTM-SA neural network is constructed, and sequence space features are extracted according to the improved CNN-BiLSTM-SA neural network; hyper-parameters of the CNN-BiLSTM-SA neural network are optimized and improved according to a Cauchy-Euclidean clustering particle swarm optimization algorithm, and an orbit fault diagnosis model is obtained; the optimal solution # imgabs1 # of the steel rail vibration signals, the typical features and the sequence space features are fused to obtain fused features, the fused features are input into a rail fault diagnosis model, and the rail fault diagnosis model outputs and obtains a rail fault diagnosis result. The accuracy and efficiency of track fault category diagnosis can be improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of rail transit fault diagnosis, and in particular relates to a rail fault diagnosis method and system. Background Art

[0002] In high-speed railway systems, the health of the track is crucial to ensure the safety, comfort and efficiency of train operation. With the increase in railway operating mileage, the external environmental impact and wheel-rail friction on the track are becoming increasingly severe, resulting in increasingly prominent track disease problems. These diseases include abnormal fasteners, track plate damage, rail corrugation, track unevenness, CA mortar problems, support layer or base plate damage, and main structure diseases caused by changes in the lower roadbed structure. If not detected and repaired in time, it will seriously affect the service life and operational efficiency of the track. At present, track fault diagnosis has the following defects: Lack of real-time and accuracy: High-speed railways run at high speeds, and existing fault diagnosis technologies are unable to collect fault data in real time and obtain accurate diagnosis results. Summary of the invention

[0003] The present invention provides a rail fault diagnosis method and system, which are used to solve the technical problem that the existing fault diagnosis technology cannot collect fault data in real time and obtain accurate diagnosis results.

[0004] In a first aspect, the present invention provides a track fault diagnosis method, comprising: The rail vibration signal is obtained based on the Beer morphology adaptive filtering denoising strategy; The rail vibration signal is decomposed according to the improved scale-space guided variational mode decomposition to obtain the optimal solution of the rail vibration signal. ; Calculate the optimal solution The sample entropy, energy entropy, power spectrum and permutation entropy of the embodiment of the present invention are obtained, and the optimal entropy is selected from the sample entropy, energy entropy, power spectrum and permutation entropy according to a preset optimal entropy weight selection strategy, and typical features are extracted according to the optimal entropy; Constructing an improved CNN-BiLSTM-SA neural network, and extracting sequence space features according to the improved CNN-BiLSTM-SA neural network; Optimizing the hyperparameters of the improved CNN-BiLSTM-SA neural network according to the Cauchy-Euclidean clustering particle swarm optimization algorithm to obtain a track fault diagnosis model; The optimal solution of the rail vibration signal , the typical features and the sequence space features are fused to obtain fused features, and the fused features are input into the track fault diagnosis model, and the track fault diagnosis model outputs a track fault diagnosis result.

[0005] In a second aspect, the present invention provides a track fault diagnosis system, comprising: An acquisition module, configured to acquire rail vibration signals based on a Beer morphology adaptive filtering and denoising strategy; A decomposition module is configured to decompose the rail vibration signal according to the improved scale-space guided variational mode decomposition to obtain an optimal solution of the rail vibration signal. ; Select the module and configure it to calculate the optimal solution The sample entropy, energy entropy, power spectrum and permutation entropy of the embodiment of the present invention are obtained, and the optimal entropy is selected from the sample entropy, energy entropy, power spectrum and permutation entropy according to a preset optimal entropy weight selection strategy, and typical features are extracted according to the optimal entropy; An extraction module, configured to construct an improved CNN-BiLSTM-SA neural network, and extract sequence space features according to the improved CNN-BiLSTM-SA neural network; an optimization module configured to optimize hyperparameters of the improved CNN-BiLSTM-SA neural network according to a Cauchy-Euclidean clustering particle swarm optimization algorithm to obtain a track fault diagnosis model; An output module is configured to output the optimal solution of the rail vibration signal , the typical features and the sequence space features are fused to obtain fused features, and the fused features are input into the track fault diagnosis model, and the track fault diagnosis model outputs a track fault diagnosis result.

[0006] In a third aspect, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can perform the steps of the track fault diagnosis method of any embodiment of the present invention.

[0007] In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor executes the steps of the track fault diagnosis method of any embodiment of the present invention.

[0008] The rail fault diagnosis method and system of the present application designs a vibration signal acquisition system of Bilateral Morphological Adaptive Filtering Denoising (BMAFD); proposes an improved scale-space guided variational mode decomposition (MSVMD) to decompose the rail vibration signal after denoising; uses the best entropy weight selection strategy (BEWM) to select the most appropriate entropy and calculate the decomposed The optimal entropy of the CNN-BiLSTM-SA network model is proposed to extract sequence space features and capture their front-to-back dependencies. The self-attention mechanism is introduced to solve information overload and increase the attention to key information. The Cauchy-Euclidean distance Particle Swarm Optimization Algorithm (CEPSO) is proposed to optimize the hyperparameters of the CNN-BiLSTM-SA neural network, and the optimal entropy loss function is proposed. The fusion features are input into the improved model for track fault diagnosis, which can improve the accuracy and efficiency of track fault category diagnosis and achieve accurate and efficient diagnosis of track fault categories. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 A flow chart of a track fault diagnosis method provided by one embodiment of the present invention; Figure 2 A structural block diagram of a track fault diagnosis system provided by an embodiment of the present invention; Figure 3 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0011] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0012] See also Figure 1 , which shows a flow chart of a track fault diagnosis method of the present application.

[0013] like Figure 1 As shown, the track fault diagnosis method specifically includes the following steps: Step S101, obtaining rail vibration signals based on the Beer morphology adaptive filtering and denoising strategy.

[0014] In this step, the vibration sensor is improved, and a vibration signal acquisition system with Beer morphology adaptive filtering and denoising is designed to collect rail vibration signals. Specifically, the vibration signal acquisition system with Beer morphology adaptive filtering and denoising uses 8-channel PCB three-axis digital acceleration vibration sensors to collect rail vibration signals, and performs signal amplification, Beer adaptive morphology filtering and denoising, bias adjustment, single-ended to differential conversion and other signal preprocessing on the signal through the signal conditioning module; in order to weaken noise, eliminate scattered points, fill breakpoints with burrs, and improve the signal-to-noise ratio, the expression of the Beer morphology adaptive filtering and denoising strategy is: , , , In the formula, is the signal obtained after Beer morphological filtering and denoising, is the weighting factor for opening and closing operations, is the minimum value of the opening and closing operation, is the original signal of rail vibration, is the weighting factor for the closed-open operation, is the maximum value of the closed-open operation, is the nth structural element sequence in the Beer morphological filter, is the n+1th structural element sequence in the Beer morphological filter, For opening and closing operations , For opening and closing operations , for right The opening and closing operation, is the original signal bandwidth, is the circumference of a circle, is the filter order, is the sampling frequency, is the carrier frequency.

[0015] Step S102, decomposing the rail vibration signal according to the improved scale-space guided variational mode decomposition to obtain the optimal solution of the rail vibration signal .

[0016] In this step, the collected rail vibration signal is Fourier transformed, and then convolution is performed using the Gaussian kernel function to obtain the scale space set and space length, which is expressed as: , , , In the formula, is the rail vibration signal after Fourier transformation, is the signal length, is the discrete time signal of the rail vibration signal, is a discrete time point, is an imaginary unit, is the frequency index, is the scale space set of rail vibration signals, is the local minimum position of each scale layer, is the number of local minima in each scale layer, is the number of local minima at each scale layer in the previous circular convolution, is the spatial length of the rail vibration signal, is the number of current circular convolutions; The corresponding threshold is calculated by using the coupled scale space constraint rail vibration signal decomposition strategy to convert the original scale space curve It is divided into the first category with the smallest intra-class variance and the second category with the largest inter-class variance. The frequency band between the local minima represented by two scale space curves exceeding a certain length (corresponding to the threshold T) is considered to be a valid frequency band, and the frequency band boundary corresponding to the threshold of the scale space curve is obtained based on this, and the number of resonant frequency bands k is further calculated to guide the variational mode decomposition. The expression is: , , In the formula, is the probability weighted sum of the two types of variance, is the probability of the (t-1)th point in the i-th scale space curve, is the probability of the first scale-space curve, is the tensor product operation, is the i-th heterogeneous scale space curve at point t, is the variance change of the (t-1)th point in the i-th scale space curve, is the mean change of the t-1th point of the i-th scale space curve, is the probability change of the tth point in the i-th scale space curve, is the between-class variance, is the class mean, is the probability of the tth point in the i-th scale space curve, is the probability of the first type of scale space curve at point t in the scale space curve set, is the intra-class mean of the t+1th point in the scale space curve of the first class, is the mean value of the i-th scale space curve, is the intra-class mean of the t-th point in the scale space curve of the first class, is the mean value of the kth IMF signal in the i-th scale space, is the kth IMF signal in the i-th scale space, is the number of categories, n=1, 2, is the number of resonant frequency bands, In order to find the k value that makes the subsequent objective function reach the maximum value, is the probability of the i-th scale-space curve, is the variance of the i-th category, is the probability of the i-th scale space curve at point t in the scale space curve set, is the class mean; The rail vibration signal is subjected to variational mode decomposition, and the single-sided spectrum of the analytical signal corresponding to each IMF signal is obtained through Hilbert transform. The expression is: , In the formula, is the impulse function, is an imaginary unit, For time, is the circumference of a circle, is the decomposed IMF signal; Through the complex exponential function with the respective estimated center frequency as parameter Multiply them together to adjust the frequency of each IMF to the baseband. The expression is: , In the formula, is the center frequency corresponding to each IMF signal after decomposition; The bandwidth of each IMF signal is estimated using square regularization and Gaussian smoothed offset signals, and the expression is: , In the formula, is the bandwidth of the IMF signal, for, is the set of IMF components after decomposition, is the set of center frequencies corresponding to each IMF component after decomposition, is the partial derivative with respect to t, is the original rail vibration signal, is the decomposed IMF signal, is the IMF component, is the original input signal; Given a discriminant accuracy ,like , the iteration is stopped, and the number of IMF signals k is obtained by decomposing the rail vibration signal according to the coupled scale space constraint, and the central frequency strategy is improved by selecting the heterogeneous scale space. , , and The variational mode decomposition of , and finally the optimal solution and center frequency of each IMF signal are obtained. The expression of the strategy of selecting the center frequency in the heterogeneous scale space is: , In the formula, is the kth mode function at frequency in the n+th iteration The value at is the frequency domain representation of the original signal f(t), is the modal function of modal number i at the frequency in the n+th iteration The value at is the modal function with modal number i at the frequency in the nth iteration The value at is the representation of the signal in the time domain after being processed by the bandpass filter, is the Lagrangian operator corresponding to the constraint condition in the nth iteration at frequency The value at is the penalty term coefficient, is the frequency variable, is the center frequency of the kth mode function in the nth iteration; , In the formula, is the indicated center frequency of the kth mode, For In the enclosed area and Perform double integration, is the time variable used for integration, is the representation of the signal in the time-frequency domain after being processed by the bandpass filter, For The single integral of is the frequency domain representation of the signal after being processed by the bandpass filter, For The single integral of ; , In the formula, is the frequency in the n+1th iteration The Lagrangian operator at , is the frequency in the nth iteration The Lagrangian operator at , is the frequency domain representation of the signal after being processed by the bandpass filter, is the Fourier transform of the original signal f(t), is the number of IMF signals, is the kth mode function in the n+1th iteration at frequency The value at , In the formula, is the penalty term coefficient of the kth mode, is the sampling frequency, is the signal frequency after being processed by the bandpass filter.

[0017] Step S103, calculate the optimal solution The sample entropy, energy entropy, power spectrum and permutation entropy are obtained, and the best entropy is selected from the sample entropy, energy entropy, power spectrum and permutation entropy according to a preset best entropy weight selection strategy, and typical features are extracted according to the best entropy.

[0018] In this step, the object of typical feature extraction is the rail vibration signal. Constructing the initial evaluation index matrix , i=1,2,…,m, j=1,2,…,n, that is, there are m evaluation samples in total, and each sample has n evaluation indicators; Standardize the indicators to eliminate the dimensional differences between different indicators. The expression in which the value is proportional to the evaluation indicator is: , The expression that the value is inversely proportional to the evaluation index is: , In the formula, is the indicator feature weight, is the value of the jth feature of the ith sample, is the minimum value of all features in the i-th sample, is the maximum value of all features in the i-th sample; Calculate the weight matrix based on the Newell coefficient-Heller matrix method , the expression is: , In the formula, is the Newell coefficient, is the weight of the first evaluation index of the mth evaluation sample under the weighting of n-1 Vigneault coefficient, is the weight of the first evaluation index of the mth evaluation sample under the weighting of the 2-dimensional coefficient, is the weight of the nth evaluation index of the mth evaluation sample under the weighting of the n-1 Vigneault coefficient, is the weight of the nth evaluation index of the mth evaluation sample under the weighting of the second Vigneault coefficient; Calculate the information entropy of the jth evaluation index and the weight of the jth evaluation index , the expression is: , , In the formula, is the mth evaluation sample, is the indicator feature weight of the nth evaluation indicator of the mth evaluation sample.

[0019] Step S104, constructing an improved CNN-BiLSTM-SA neural network, and extracting sequence space features according to the improved CNN-BiLSTM-SA neural network.

[0020] In this step, sequence space features refer to multi-level feature representations formed by feature extraction at different levels (local features of CNN, long-term dependencies of BiLSTM, and global relationships of SA). The extraction object is the rail vibration signal. Construct a convolutional neural network (CNN). CNN has a powerful feature extraction function. Its network composition includes input layer, output layer, activation layer, fully connected layer, pooling layer and convolution layer. The convolution layer is used to extract features. Its operation expression is: , In the formula, For input, is the feature output after convolution of the i-th convolution kernel of the first layer, and Represent the weight and bias of the i-th convolution kernel in the first layer respectively; The activation layer adds a nonlinear Relu activation function to enable the model to have cross-layer nonlinear mapping learning capabilities. Its operation expression is: , In the formula, is the input of the activation layer after the i-th convolution kernel of the first layer is processed. The output after adding the nonlinear ReLU activation function to the feature output after convolution of the i-th convolution kernel of the first layer; Use the maximum pooling layer to downsample the data after the convolution layer to extract the main features, and then integrate them to form new features to avoid overfitting. The operation expression is: , in, is the new feature formed in the maximum pooling layer after the i-th convolution kernel of the first layer is processed. is the pooling kernel width, represents the step length, is the characteristic length; The bidirectional long short-term memory network (BiLSTM) is embedded in the maximum pooling layer and the fully connected layer of the CNN. Through three gating structures, the functions of controlling the retention of previous unit information, preserving the input information, and determining the unit output are realized. BiLSTM can fully obtain the time series data information and improve the model dependency, specifically: , , , , , , , in, yes function, is the hyperbolic tangent function, is the input signal, Represents the weight of the input signal and the memory unit at the current moment, Represents the weight of the hidden layer and the memory unit value at the current moment, The weights of the input signal and the input gate, is the weight from the hidden state to the input gate, is the weight from the memory cell state to the input gate, is the weight of the input signal to the forget gate; is the bias of the memory unit at the current moment, is the input gate bias, Forget gate bias, Output gate bias, are the network output bias respectively; is the temporary memory unit value, is the memory cell value at the current moment; represents the value of the input gate, represents the value of the forget gate, and is the output result of the current output gate and hidden layer, is the network output; BiLSTM consists of two LSTMs moving in opposite directions. The BiLSTM update process is as follows: , , , in, and represents a forward and backward update of LSTM, and is the update result, and are the forward and reverse weight values ​​of the output and hidden layers, Bias for the output gate; In order to improve the robustness of the entire network and the ability to characterize data features, a fully connected layer is added before the output layer, specifically: , in, for The activation function, the hidden layer weights and biases of the m+1th layer are , ; The Self-Attention module is introduced to convert the N sequences of a single sample into three vectors of length d through adaptive linear mapping, and then connect them in parallel to form a matrix Q, which is the query matrix representing the information that needs to be paid attention to, M is the key matrix representing the information available for query, and V is the value matrix representing the information or features corresponding to the key, so as to solve the problem of information overload and improve the attention of key information. The specific method is as follows: , Use the Self-learning Variable Squeeze and Excitation Network (SVSEN) to improve the feature channels that are useful for the current task. Specifically: , , In the formula, is the compressed feature, is the total number of channels, is the original feature, is element-wise multiplication, is the total number of features, is the compression weight, is the original feature, is the compression ratio, is the self-learning weight, is the tensor product operation, To stimulate the calibrated features, is the characteristic after excitation, is the learning rate of the ith feature, is the excitation weight of the jth channel.

[0021] Step S105, optimizing the hyperparameters of the improved CNN-BiLSTM-SA neural network according to the Cauchy-Euclidean clustering particle swarm optimization algorithm to obtain a track fault diagnosis model.

[0022] In this step, the particle swarm is initialized; According to the Euclidean distance nonlinear dynamic inertia weight competition adjustment strategy, the Euclidean distance between all particles in the population is calculated to obtain the Euclidean distance matrix , the expression is: , In the formula, is the inertia weight in nonlinear dynamics The Euclidean distance between the nth particle in group A and the nth particle in group B; , In the formula, To perform a product operation on the number of iterations m, is the tensor product operation, is the competitive factor, is the maximum number of iterations set, is the velocity of the d-dimensional i-th particle at the m-1th iteration, is the Euclidean distance of all particles in the population after the mth iteration, is the Euclidean distance of all particles in the population after the m-1th iteration, is the partial derivative operator symbol, is the velocity of the d-dimensional i-th particle at the m-th iteration, is the velocity of the d-dimensional i-th particle at the m+1-th iteration; , In the formula, is the inertia weight adjusted for the m-1th iteration, is the nonlinear dynamic correlation variable of the particle Euclidean distance at the mth iteration; Based on the random trajectory correction strategy of the Cauchy-Lorenz coefficient distribution, each particle is dynamically corrected with the global optimal particle. And when the number of iterations or accuracy requirements are met, output , the expression is: , , , In the formula, is the random trajectory correction result of the Cauchy-Lorenz coefficient distribution, is the particle position of the mth iteration before correction, is the Lorenz coefficient , is the probability density function of the Cauchy-Lorenz coefficient distribution, is the global optimal particle position, is the scale parameter of the mth iteration, is the number of iterations, is the initial scale parameter, is the time constant.

[0023] Step S106: The optimal solution of the rail vibration signal , the typical features and the sequence space features are fused to obtain fused features, and the fused features are input into the track fault diagnosis model, and the track fault diagnosis model outputs a track fault diagnosis result.

[0024] In this step, the optimal solution of the rail vibration signal is , the typical features and sequence space features are fused and normalized, the optimal entropy loss function is calculated, and the improved network model is input for training and optimization for subsequent fault classification and diagnosis in the network, specifically: , In the formula, is the result of feature fusion, is the normalization factor, is the weight coefficient of the i-th feature layer, is the total number of feature layers, is the fusion feature of the i-th feature layer, is the mean of the i-th feature layer, is the standard deviation of the i-th feature layer; , In the formula, is the weight factor of the predicted probability of belonging to category B in the a-th sample, is the weight factor of the predicted probability that the a-th sample does not belong to category B, is the predicted probability of belonging to category B in the a-th sample, for The expected value of is the regularization parameter, is the weight factor in regularization, is the weight factor in regularization.

[0025] In summary, the method of the present application obtains the rail vibration signal based on the Beer morphology adaptive filtering denoising strategy; decomposes the rail vibration signal according to the improved scale space guided variational mode decomposition to obtain the optimal solution of the rail vibration signal ; Calculate the optimal solution The sample entropy, energy entropy, power spectrum and permutation entropy are obtained, and the optimal entropy is selected from the sample entropy, energy entropy, power spectrum and permutation entropy according to the preset optimal entropy weight selection strategy, and typical features are extracted according to the optimal entropy; an improved CNN-BiLSTM-SA neural network is constructed, and sequence space features are extracted according to the improved CNN-BiLSTM-SA neural network; the hyperparameters of the improved CNN-BiLSTM-SA neural network are optimized according to the Cauchy-Euclidean clustering particle swarm optimization algorithm to obtain a track fault diagnosis model; the optimal solution of the rail vibration signal is converted into , typical features and sequence space features are fused to obtain fused features, and the fused features are input into the track fault diagnosis model. The track fault diagnosis model outputs the track fault diagnosis results, which can improve the accuracy and efficiency of track fault category diagnosis.

[0026] See also Figure 2 , which shows a structural block diagram of a track fault diagnosis system of the present application.

[0027] like Figure 2 As shown, the track fault diagnosis system 200 includes an acquisition module 210 , a decomposition module 220 , a selection module 230 , an extraction module 240 , an optimization module 250 and an output module 260 .

[0028] The acquisition module 210 is configured to acquire the rail vibration signal based on the Beer morphology adaptive filtering denoising strategy; the decomposition module 220 is configured to decompose the rail vibration signal according to the variational mode decomposition guided by the improved scale space to obtain the optimal solution of the rail vibration signal. ; Select module 230, configured to calculate the optimal solution The sample entropy, energy entropy, power spectrum and permutation entropy of the track fault diagnosis model are obtained by extracting the optimal entropy from the sample entropy, energy entropy, power spectrum and permutation entropy according to the preset optimal entropy weight selection strategy, and extracting typical features according to the optimal entropy; an extraction module 240 is configured to construct an improved CNN-BiLSTM-SA neural network, and extract sequence space features according to the improved CNN-BiLSTM-SA neural network; an optimization module 250 is configured to optimize the hyperparameters of the improved CNN-BiLSTM-SA neural network according to the Cauchy-Euclidean clustering particle swarm optimization algorithm to obtain a track fault diagnosis model; an output module 260 is configured to output the optimal solution of the rail vibration signal to the track fault diagnosis model; , the typical features and the sequence space features are fused to obtain fused features, and the fused features are input into the track fault diagnosis model, and the track fault diagnosis model outputs a track fault diagnosis result.

[0029] It should be understood that Figure 2 Modules and references documented in Figure 1 Therefore, the operations and features described above for the method and the corresponding technical effects are also applicable to Figure 2 The modules in it will not be described in detail here.

[0030] In some other embodiments, the embodiments of the present invention further provide a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor is caused to execute the rail fault diagnosis method in any of the above method embodiments; As an implementation mode, the computer-readable storage medium of the present invention stores computer-executable instructions, and the computer-executable instructions are configured as follows: The rail vibration signal is obtained based on the Beer morphology adaptive filtering denoising strategy; The rail vibration signal is decomposed according to the improved scale-space guided variational mode decomposition to obtain the optimal solution of the rail vibration signal. ; Calculate the optimal solution The sample entropy, energy entropy, power spectrum and permutation entropy of the embodiment of the present invention are obtained, and the optimal entropy is selected from the sample entropy, energy entropy, power spectrum and permutation entropy according to a preset optimal entropy weight selection strategy, and typical features are extracted according to the optimal entropy; Constructing an improved CNN-BiLSTM-SA neural network, and extracting sequence space features according to the improved CNN-BiLSTM-SA neural network; Optimizing the hyperparameters of the improved CNN-BiLSTM-SA neural network according to the Cauchy-Euclidean clustering particle swarm optimization algorithm to obtain a track fault diagnosis model; The optimal solution of the rail vibration signal , the typical features and the sequence space features are fused to obtain fused features, and the fused features are input into the track fault diagnosis model, and the track fault diagnosis model outputs a track fault diagnosis result.

[0031] The computer-readable storage medium may include a program storage area and a data storage area, wherein the program storage area may store an operating system and an application required by at least one function; the data storage area may store data created according to the use of the track fault diagnosis system, etc. In addition, the computer-readable storage medium may include a high-speed random access memory, and may also include a memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the computer-readable storage medium may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the track fault diagnosis system via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0032] Figure 3 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Figure 3 As shown, the device includes: a processor 310 and a memory 320. The electronic device may also include: an input device 330 and an output device 340. The processor 310, the memory 320, the input device 330 and the output device 340 may be connected via a bus or other means. Figure 3 The example of the bus connection is taken. The memory 320 is the computer-readable storage medium mentioned above. The processor 310 executes various functional applications and data processing of the server by running the non-volatile software programs, instructions and modules stored in the memory 320, that is, the track fault diagnosis method of the above method embodiment is realized. The input device 330 can receive input digital or character information, and generate key signal input related to user settings and function control of the track fault diagnosis system. The output device 340 may include display devices such as display screens.

[0033] The electronic device can execute the method provided by the embodiment of the present invention, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not described in detail in this embodiment, please refer to the method provided by the embodiment of the present invention.

[0034] As an implementation mode, the electronic device is applied to a rail fault diagnosis system and is used for a client, and includes: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can: The rail vibration signal is obtained based on the Beer morphology adaptive filtering denoising strategy; The rail vibration signal is decomposed according to the improved scale-space guided variational mode decomposition to obtain the optimal solution of the rail vibration signal. ; Calculate the optimal solution The sample entropy, energy entropy, power spectrum and permutation entropy of the embodiment of the present invention are obtained, and the optimal entropy is selected from the sample entropy, energy entropy, power spectrum and permutation entropy according to a preset optimal entropy weight selection strategy, and typical features are extracted according to the optimal entropy; Constructing an improved CNN-BiLSTM-SA neural network, and extracting sequence space features according to the improved CNN-BiLSTM-SA neural network; Optimizing the hyperparameters of the improved CNN-BiLSTM-SA neural network according to the Cauchy-Euclidean clustering particle swarm optimization algorithm to obtain a track fault diagnosis model; The optimal solution of the rail vibration signal , the typical features and the sequence space features are fused to obtain fused features, and the fused features are input into the track fault diagnosis model, and the track fault diagnosis model outputs a track fault diagnosis result.

[0035] Through the description of the above implementation modes, those skilled in the art can clearly understand that each implementation mode can be implemented by means of software plus a necessary general hardware platform, or of course by hardware. Based on such an understanding, the above technical solution can essentially or in other words be embodied in the form of a software product that contributes to the prior art. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiment.

[0036] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A track fault diagnosis method, characterized in that: include: The rail vibration signal is obtained based on the Beer morphology adaptive filtering denoising strategy; The rail vibration signal is decomposed according to the improved scale-space guided variational mode decomposition to obtain the optimal solution of the rail vibration signal. ; Calculate the optimal solution The sample entropy, energy entropy, power spectrum and permutation entropy of the embodiment of the present invention are obtained, and the optimal entropy is selected from the sample entropy, energy entropy, power spectrum and permutation entropy according to a preset optimal entropy weight selection strategy, and typical features are extracted according to the optimal entropy; Constructing an improved CNN-BiLSTM-SA neural network, and extracting sequence space features according to the improved CNN-BiLSTM-SA neural network; Optimizing the hyperparameters of the improved CNN-BiLSTM-SA neural network according to the Cauchy-Euclidean clustering particle swarm optimization algorithm to obtain a track fault diagnosis model; The optimal solution of the rail vibration signal , the typical features and the sequence space features are fused to obtain fused features, and the fused features are input into the track fault diagnosis model, and the track fault diagnosis model outputs a track fault diagnosis result.

2. A rail fault diagnosis method according to claim 1, characterized in that: in, The expression of the Beer morphology adaptive filtering denoising strategy is: ， ， ， In the formula, is the signal obtained after Beer morphological filtering and denoising, is the weighting factor for opening and closing operations, is the minimum value of the opening and closing operation, is the original signal of rail vibration, is the weighting factor for the closing and opening operation, is the maximum value of the closed-open operation, is the nth structural element sequence in the Beer morphological filter, is the n+1th structural element sequence in the Beer morphological filter, For opening and closing operations , For opening and closing operations , for right The opening and closing operation, is the original signal bandwidth, is the circumference of a circle, is the filter order, is the sampling frequency, is the carrier frequency.

3. A rail fault diagnosis method according to claim 1, characterized in that: The rail vibration signal is decomposed according to the improved scale-space guided variational mode decomposition to obtain the optimal solution of the rail vibration signal. include: The collected rail vibration signal is subjected to Fourier transform, and then convolution is performed using the Gaussian kernel function to obtain the scale space set and space length, which are expressed as follows: ， ， ， In the formula, is the rail vibration signal after Fourier transformation, is the signal length, is the discrete time signal of the rail vibration signal, is a discrete time point, is an imaginary unit, is the frequency index, is the scale space set of rail vibration signals, is the local minimum position of each scale layer, is the number of local minima in each scale layer, is the number of local minima at each scale layer in the previous circular convolution, is the spatial length of the rail vibration signal, is the number of current circular convolutions; The corresponding threshold is calculated by using the coupled scale space constraint rail vibration signal decomposition strategy to convert the original scale space curve It is divided into the first category with the smallest intra-class variance and the second category with the largest inter-class variance. The expression is: ， ， In the formula, is the probability weighted sum of the two types of variance, is the probability of the (t-1)th point in the i-th scale space curve, is the probability of the first scale-space curve, is the tensor product operation, is the i-th heterogeneous scale space curve at point t, is the variance change of the (t-1)th point in the i-th scale space curve, is the mean change of the t-1th point of the i-th scale space curve, is the probability change of the tth point in the i-th scale space curve, is the between-class variance, is the class mean, is the probability of the tth point in the i-th scale space curve, is the probability of the first type of scale space curve at point t in the scale space curve set, is the intra-class mean of the t+1th point in the scale space curve of the first class, is the mean value of the i-th scale space curve, is the intra-class mean of the t-th point in the scale space curve of the first class, is the mean value of the kth IMF signal in the i-th scale space, is the kth IMF signal in the i-th scale space, is the number of categories, n=1, 2, is the number of resonant frequency bands, In order to find the k value that makes the subsequent objective function reach the maximum value, is the probability of the i-th scale-space curve, is the variance of the i-th category, is the probability of the i-th scale space curve at point t in the scale space curve set, is the class mean; The rail vibration signal is subjected to variational mode decomposition, and the single-sided spectrum of the analytical signal corresponding to each IMF signal is obtained through Hilbert transform. The expression is: ， In the formula, is the impulse function, is an imaginary unit, For time, is the circumference of a circle, is the decomposed IMF signal; Through the complex exponential function with the respective estimated center frequency as parameter Multiply them together to adjust the frequency of each IMF to the baseband. The expression is: ， In the formula, is the center frequency corresponding to each IMF signal after decomposition; The bandwidth of each IMF signal is estimated using square regularization and Gaussian smoothed offset signals, and the expression is: ， In the formula, is the bandwidth of the IMF signal, for, is the set of IMF components after decomposition, is the set of center frequencies corresponding to each IMF component after decomposition, is the partial derivative with respect to t, is the original rail vibration signal, is the decomposed IMF signal, is the IMF component, is the original input signal; Given a discriminant accuracy ,like , the iteration is stopped, and the number of IMF signals k is obtained by decomposing the rail vibration signal according to the coupled scale space constraint, and the central frequency strategy is improved by selecting the heterogeneous scale space. , , and The variational mode decomposition of , and finally the optimal solution and center frequency of each IMF signal are obtained. The expression of the strategy of selecting the center frequency in the heterogeneous scale space is: ， In the formula, is the kth mode function at frequency in the n+th iteration The value at is the frequency domain representation of the original signal f(t), is the modal function of modal number i at the frequency in the n+th iteration The value at is the modal function with modal number i at the frequency in the nth iteration The value at is the representation of the signal in the time domain after being processed by the bandpass filter, is the Lagrangian operator corresponding to the constraint condition in the nth iteration at frequency The value at is the penalty term coefficient, is the frequency variable, is the center frequency of the kth mode function in the nth iteration; ， In the formula, is the indicated center frequency of the kth mode, For In the enclosed area and Perform double integration, is the time variable used for integration, is the representation of the signal in the time-frequency domain after being processed by the bandpass filter, For The single integral of is the frequency domain representation of the signal after being processed by the bandpass filter, For The single integral of ; ， In the formula, is the frequency in the n+1th iteration The Lagrangian operator at , is the frequency in the nth iteration The Lagrangian operator at , is the frequency domain representation of the signal after being processed by the bandpass filter, is the Fourier transform of the original signal f(t), is the number of IMF signals, is the kth mode function in the n+1th iteration at frequency The value at ， In the formula, is the penalty term coefficient of the kth mode, is the sampling frequency, is the frequency of the signal after being processed by the bandpass filter.

4. A rail fault diagnosis method according to claim 1, characterized in that: The selecting the best entropy from sample entropy, energy entropy, power spectrum and permutation entropy according to the preset best entropy weight selection strategy includes: Constructing the initial evaluation index matrix , i=1,2,…,m, j=1,2,…,n, that is, there are m evaluation samples in total, and each sample has n evaluation indicators; Standardize the indicators to eliminate the dimensional differences between different indicators. The expression in which the value is proportional to the evaluation indicator is: ， The expression that the value is inversely proportional to the evaluation index is: ， In the formula, is the indicator feature weight, is the value of the jth feature of the ith sample, is the minimum value of all features in the i-th sample, is the maximum value of all features in the i-th sample; Calculate the weight matrix based on the Newell coefficient-Heller matrix method , the expression is: ， In the formula, is the Newell coefficient, is the weight of the first evaluation index of the mth evaluation sample under the weighting of n-1 Vigneault coefficient, is the weight of the first evaluation index of the mth evaluation sample under the weighting of the 2-dimensional coefficient, is the weight of the nth evaluation index of the mth evaluation sample under the weighting of the n-1 Vigneault coefficient, is the weight of the nth evaluation index of the mth evaluation sample under the weighting of the second Vigneault coefficient; Calculate the information entropy of the jth evaluation index and the weight of the jth evaluation index , the expression is: ， ， In the formula, is the mth evaluation sample, is the indicator feature weight of the nth evaluation indicator of the mth evaluation sample.

5. A rail fault diagnosis method according to claim 1, characterized in that: The improved CNN-BiLSTM-SA neural network includes a self-learning compression and excitation network, which is expressed as: ， ， In the formula, is the compressed feature, is the total number of channels, is the original feature, is element-wise multiplication, is the total number of features, is the compression weight, is the original feature, is the compression ratio, is the self-learning weight, is the tensor product operation, To stimulate the calibrated features, is the characteristic after excitation, is the learning rate of the ith feature, is the excitation weight of the jth channel.

6. A rail fault diagnosis method according to claim 1, characterized in that: The hyper parameters of the improved CNN-BiLSTM-SA neural network are optimized according to the Cauchy-Euclidean clustering particle swarm optimization algorithm, including: Initialize particle swarm; According to the Euclidean distance nonlinear dynamic inertia weight competition adjustment strategy, the Euclidean distance between all particles in the population is calculated to obtain the Euclidean distance matrix , the expression is: ， In the formula, is the inertia weight in nonlinear dynamics The Euclidean distance between the nth particle in group A and the nth particle in group B; ， In the formula, To perform a product operation on the number of iterations m, is the tensor product operation, is the competitive factor, is the maximum number of iterations set, is the velocity of the d-dimensional i-th particle at the m-1th iteration, is the Euclidean distance of all particles in the population after the mth iteration, is the Euclidean distance of all particles in the population after the m-1th iteration, is the partial derivative operator symbol, is the velocity of the d-dimensional i-th particle at the m-th iteration, is the velocity of the d-dimensional i-th particle at the m+1-th iteration; ， In the formula, is the inertia weight adjusted for the m-1th iteration, is the nonlinear dynamic correlation variable of the particle Euclidean distance at the mth iteration; Based on the random trajectory correction strategy of the Cauchy-Lorenz coefficient distribution, each particle is dynamically corrected with the global optimal particle. And when the number of iterations or accuracy requirements are met, output , the expression is: ， ， ， In the formula, is the random trajectory correction result of the Cauchy-Lorenz coefficient distribution, is the particle position of the mth iteration before correction, is the Lorenz coefficient , is the probability density function of the Cauchy-Lorenz coefficient distribution, is the global optimal particle position, is the scale parameter of the mth iteration, is the number of iterations, is the initial scale parameter, is the time constant.

7. A track fault diagnosis system, characterized in that: include: An acquisition module, configured to acquire rail vibration signals based on a Beer morphology adaptive filtering and denoising strategy; A decomposition module is configured to decompose the rail vibration signal according to the improved scale-space guided variational mode decomposition to obtain an optimal solution of the rail vibration signal. ; Select the module and configure it to calculate the optimal solution The sample entropy, energy entropy, power spectrum and permutation entropy of the embodiment of the present invention are obtained, and the optimal entropy is selected from the sample entropy, energy entropy, power spectrum and permutation entropy according to a preset optimal entropy weight selection strategy, and typical features are extracted according to the optimal entropy; An extraction module, configured to construct an improved CNN-BiLSTM-SA neural network, and extract sequence space features according to the improved CNN-BiLSTM-SA neural network; an optimization module configured to optimize hyperparameters of the improved CNN-BiLSTM-SA neural network according to a Cauchy-Euclidean clustering particle swarm optimization algorithm to obtain a track fault diagnosis model; An output module is configured to output the optimal solution of the rail vibration signal , the typical features and the sequence space features are fused to obtain fused features, and the fused features are input into the track fault diagnosis model, and the track fault diagnosis model outputs a track fault diagnosis result.

8. An electronic device, characterized in that: include: At least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images