Track circuit fault diagnosis method and system

Abstract

The invention discloses a track circuit fault diagnosis method and system. The method comprises the steps of obtaining first data of a target track circuit, and performing first preprocessing on the first data; the first preprocessing at least comprises performing first expansion operation on the first data; taking the first data after the first preprocessing as a sample set, and training a first diagnosis model; and verifying the trained first diagnosis model, and performing target track circuit fault diagnosis according to the verified first diagnosis model. Through automatic data acquisition, preprocessing, model training and verification processes, the fault of the track circuit can be quickly diagnosed, the diagnosis time is greatly shortened, and the diagnosis efficiency is improved. Through automatic fault diagnosis, the fault can be found and positioned in time, and equipment damage and shutdown time caused by the fault are avoided, so that the maintenance cost is reduced.

Description

Technical Field

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

[0002] With the rapid development of high-speed railway construction, driving safety and operation efficiency issues are becoming more and more important. Track circuits are one of the core devices of railway signal systems and are widely used in train operation control and dispatching. Their main function is to detect the occupancy status of track sections in real time to ensure the safety and efficiency of train operation. Due to the long-term exposure of track circuits to complex environments (such as bad weather, vibration, electromagnetic interference, etc.) and factors such as equipment aging, faults often occur. Common faults include track circuit disconnection, insulation failure, equipment short circuit, power supply abnormality, etc. These faults may cause signal errors, train operation delays and even serious safety accidents.

[0003] Traditional track circuit fault diagnosis methods mainly rely on manual inspections and empirical judgments, which are not only inefficient, but also have the risk of misjudgment and missed judgments. With the development of artificial intelligence, big data and Internet of Things technologies, track circuit fault diagnosis methods based on artificial intelligence technology have gradually become a research hotspot. This method collects operating parameters such as current, voltage, frequency, etc. of the track circuit, combined with machine learning or deep learning models, to quickly identify the fault type and locate the fault. However, there are still some limitations, such as the small amount of fault data leading to data imbalance that affects model training, and the data feature extraction is too single to mine deeper information in the data, which affects the accuracy of model recognition. Summary of the invention

[0004] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0005] In view of the above existing problems, the present invention is proposed.

[0006] Therefore, the present invention provides a track circuit fault diagnosis method and system, which can solve the problems mentioned in the background technology.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In a first aspect, the present invention provides a track circuit fault diagnosis method, comprising:

[0009] Acquire first data of a target track circuit, and perform first preprocessing on the first data;

[0010] The first preprocessing at least includes performing a first expansion operation on the first data;

[0011] Using the first data after the first preprocessing as a sample set to train a first diagnostic model;

[0012] The trained first diagnostic model is verified, and target track circuit fault diagnosis is performed according to the verified first diagnostic model.

[0013] As a preferred solution of the track circuit fault diagnosis method of the present invention, the first diagnostic model includes:

[0014] The first diagnostic model is any model that takes as input the first data of the target track circuit and outputs the probability distribution of different fault categories of the track circuit or can directly or indirectly obtain relevant parameters of the probability distribution of different fault categories of the track circuit.

[0015] As a preferred solution of the track circuit fault diagnosis method of the present invention, the first preprocessing includes:

[0016] performing a first expansion operation on the first data;

[0017] The first expansion operation is performed by a preset generative adversarial network;

[0018] The generative adversarial network includes several logical combinations of generators and discriminators.

[0019] As a preferred solution of the track circuit fault diagnosis method of the present invention, the several logical combinations include:

[0020] Determine the logic combination method and combination target;

[0021] Determine the input and output of the generator and the discriminator according to the logical combination method;

[0022] According to the input and output of the generator and the discriminator, a generative adversarial network is established in combination with the logical combination method and the combination target.

[0023] As a preferred solution of the track circuit fault diagnosis method described in the present invention, the first diagnostic model also includes an input layer, a one-dimensional convolutional layer, an LSTM layer, a multi-head attention mechanism layer, a fully connected layer and an output layer.

[0024] As a preferred solution of the track circuit fault diagnosis method described in the present invention, the one-dimensional convolution layer includes two one-dimensional convolution layers, the convolution kernel of the one-dimensional convolution layer slides from the starting time step of the track circuit data, multiplies and adds the local track circuit voltage data, generates a new eigenvalue at each time step, and extracts the local correlation and short-term pattern in the track circuit data.

[0025] As a preferred scheme of the track circuit fault diagnosis method described in the present invention, wherein: the first data of the target track circuit at least includes the power output voltage, the power output current, the sending end cable side voltage, the sending end cable side current, the receiving end cable side main rail voltage, the receiving end cable side small rail voltage, the receiving end equipment side main rail voltage, the receiving end equipment side small rail voltage, the receiving entrance main rail voltage, the receiving entrance small rail voltage, the main rail relay and the small rail relay.

[0026] In a second aspect, the present invention provides a track circuit fault diagnosis system, comprising:

[0027] A data acquisition and processing module, used for acquiring first data of a target track circuit and performing first preprocessing on the first data;

[0028] The first preprocessing at least includes performing a first expansion operation on the first data;

[0029] A model training model, used to train a first diagnostic model using the first data after the first preprocessing as a sample set;

[0030] The diagnostic module is used to verify the first diagnostic model after training and perform target track circuit fault diagnosis based on the verified first diagnostic model.

[0031] In a third aspect, the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned method when executing the computer program.

[0032] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the method described above when executed by a processor.

[0033] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention proposes a track circuit fault diagnosis method and system, which obtains the first data of the target track circuit and performs a first preprocessing on the first data; the first preprocessing at least includes a first expansion operation on the first data; the first data after the first preprocessing is used as a sample set to train a first diagnostic model; the trained first diagnostic model is verified, and the target track circuit fault diagnosis is performed according to the verified first diagnostic model. Through the automated data acquisition, preprocessing, model training and verification process, the present invention can quickly diagnose the fault of the track circuit, greatly shorten the diagnosis time, and improve the diagnosis efficiency. The use of deep learning algorithms for model training can learn the characteristics and laws of track circuit faults, thereby improving the accuracy of diagnosis. Through automated fault diagnosis, the present invention can timely detect and locate faults, avoid equipment damage and downtime caused by faults, and thus reduce maintenance costs.

[0034] Specifically, using generative adversarial networks (GANs) to generate diverse sample data and expand data sets can significantly improve the performance of the model in small sample or unbalanced data scenarios, solve the problem of insufficient track circuit fault data, avoid overfitting of the model, and improve the robustness and generalization ability of the model; optimizing the ConvLSTM deep learning network model, using one-dimensional convolution to extract local features of track circuit data, the LSTM layer to capture long-term dependencies in track circuit data, and the multi-head attention mechanism to extract global features from multiple angles. The combination of the three enhances the feature extraction capability of the model, and can fully and comprehensively mine and learn hidden features in the data; at the same time, introducing residual connections to solve the gradient vanishing problem caused by the increase in network depth, improve training stability, and further improve the accuracy of fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0036] Figure 1 A method flow chart of a track circuit fault diagnosis method and system provided by one embodiment of the present invention;

[0037] Figure 2 A detailed flow chart of a track circuit fault diagnosis method and system provided by one embodiment of the present invention;

[0038] Figure 3A schematic diagram of a generative adversarial network (GAN) structure of a track circuit fault diagnosis method and system provided by an embodiment of the present invention;

[0039] Figure 4 A schematic diagram of the network structure of a ConvLSTM deep learning model of a track circuit fault diagnosis method and system provided by an embodiment of the present invention;

[0040] Figure 5 A schematic diagram of extracting feature information using a one-dimensional convolutional layer of a track circuit fault diagnosis method and system provided by an embodiment of the present invention;

[0041] Figure 6 A schematic diagram of an LSTM hidden unit in an LSTM layer of a track circuit fault diagnosis method and system provided by an embodiment of the present invention;

[0042] Figure 7 A schematic diagram of a multi-head attention mechanism layer of a track circuit fault diagnosis method and system provided for one embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0044] Example 1

[0045] Reference Figure 1-Figure 7 , which is the first embodiment of the present invention, provides a track circuit fault diagnosis method and system, including:

[0046] In the existing related technologies, there are some problems, such as the scarcity and imbalance of track circuit fault data, which often leads to poor training results of fault diagnosis models.

[0047] The present application provides a method that can effectively solve the above-mentioned problems. Next, how to implement the track circuit fault diagnosis method will be described in detail in combination with multiple embodiments.

[0048] Figure 1 A method flow chart of a track circuit fault diagnosis method and system is shown, including:

[0049] S101, obtaining first data of a target track circuit, and performing first preprocessing on the first data;

[0050] In an embodiment of the present application, the first data of the target track circuit includes at least the power output voltage, the power output current, the voltage on the sending cable side, the current on the sending cable side, the main rail voltage on the receiving cable side, the small rail voltage on the receiving cable side, the main rail voltage on the receiving device side, the small rail voltage on the receiving device side, the main rail voltage on the receiving input, the small rail voltage on the receiving input, the main rail voltage on the receiving input, the main rail relay and the small rail relay.

[0051] In an optional embodiment, the first data of the target track circuit can be acquired in real time through sensors, monitoring equipment or other data acquisition means to ensure the accuracy and timeliness of the data. These data cover the key parameters of the track circuit operation and are crucial for subsequent diagnostic analysis.

[0052] In the embodiment of the present application, the first data of the target track circuit is collected by a centralized signal monitoring system in a railway machine room;

[0053] In an optional embodiment, after acquiring these data, a first preprocessing operation is immediately performed. This step is intended to improve data quality and lay a solid foundation for subsequent model training.

[0054] Exemplarily, the track circuit data is collected by the centralized signal monitoring system in the railway machine room, and the track circuit data is expressed as: i =[x i1 ,x i2 ,...,x ij ] T , where X i represents a single track circuit data, each x represents a feature quantity, j is the number of feature quantities, and the collected track circuit data is normalized so that the data is distributed in the range of [0,1]. The data normalization formula is as follows:

[0055]

[0056] Among them, x' ij represents the normalized value of the jth characteristic value of the i-th track circuit data, x ij represents the original value of the jth feature value of the i-th track circuit data, x j_max and x j_min They respectively represent the maximum and minimum values ​​of the j-th feature quantity in all track circuit data.

[0057] It should be noted that the above first preprocessing operation is only an example and does not constitute a limitation of the present invention. In practical applications, the first preprocessing operation may also include one or more data preprocessing means such as data cleaning, data smoothing, data filtering, feature selection, feature extraction, etc., to further improve data quality and model training effect.

[0058] In the embodiment of the present application, 12 characteristic quantities are selected, including power output voltage, power output current, sending cable side voltage, sending cable side current, receiving cable side main rail voltage, receiving cable side small rail voltage, receiving device side main rail voltage, receiving device side small rail voltage, receiving input main rail voltage, receiving input small rail voltage, main rail relay, and small rail relay.

[0059] It should be noted that the above-mentioned characteristic quantities are selected as the key inputs of the diagnostic model. These characteristic quantities can fully reflect the operating status of the track circuit and are crucial for the diagnosis and location of faults. By deeply analyzing these characteristic quantities, the diagnostic model can capture potential fault signals in the track circuit, thereby achieving fast and accurate fault diagnosis.

[0060] It should also be noted that since track circuit fault data is usually scarce and unbalanced, this will bring great challenges to the training of fault diagnosis models. In order to solve this problem, the present invention adopts data expansion technology to generate more diverse sample data by transforming, combining or generating operations on the original data, thereby expanding the scale of the data set and improving the generalization ability and robustness of the model.

[0061] In the embodiment of the present application, the first preprocessing at least includes performing a first expansion operation on the first data;

[0062] It should be noted that in addition to basic data processing such as cleaning and denoising, the first preprocessing also emphasizes the first expansion operation. This is because in practical applications, track circuit fault data is often scarce and has imbalance problems. By presetting data expansion operations, sample diversity can be effectively increased, and the challenges brought by data imbalance can be alleviated, thereby improving the training effect and generalization ability of the model.

[0063] In an optional embodiment, the first expansion operation can be implemented by presetting a generative adversarial network GAN. The generative adversarial network GAN consists of a generator and a discriminator, which are mutually adversarial and trained collaboratively. The generator is responsible for generating realistic false data to deceive the discriminator; while the discriminator is dedicated to distinguishing real data from false data. Through continuous adversarial training, the generator can gradually learn the distribution characteristics of real data, thereby generating high-quality false data.

[0064] In an optional embodiment, the first expansion operation can also be implemented by other data enhancement techniques, such as random noise addition, data smoothing, feature transformation, etc. These techniques can be used alone or in combination with the generative adversarial network GAN to further enhance the effect and diversity of data expansion.

[0065] In an optional embodiment, the first expansion operation can also generate more diverse fault data by simulating track circuit fault scenarios. For example, different types of faults can be simulated by adjusting track circuit parameters or introducing external interference, thereby collecting more fault data samples. These data samples can not only increase the diversity of the training set, but also help improve the model's ability to identify unknown fault types.

[0066] In an embodiment of the present application, the first expansion operation is implemented by presetting a generative adversarial network GAN.

[0067] Specifically, the first preprocessing includes:

[0068] Performing a first expansion operation on the first data;

[0069] The first expansion operation is performed by a preset generative adversarial network;

[0070] The generative adversarial network consists of several logical combinations of generators and discriminators.

[0071] In the embodiment of the present application, several logical combinations include:

[0072] Determine the logical combination method and combination target;

[0073] Determine the input and output of the generator and discriminator according to the logical combination method;

[0074] According to the input and output of the generator and discriminator, a generative adversarial network is established by combining the logical combination method and the combination goal.

[0075] For example, the generative adversarial network (GAN) is used to enhance the track circuit fault data and expand the data set; Figure 3 As shown in the figure, the GAN network structure consists of a generator G and a discriminator D. The input of the generator G is a noise vector z, and the output is synthesized track circuit data. The input of the discriminator D is the normalized real data or the synthesized data output by the generator, and the output is a scalar value, which represents the probability that the input data is real data. The generator G and the discriminator D are trained adversarially, and G and D are trained alternately and iteratively until the Nash equilibrium is reached. After the training is completed, the generator G is used to generate a large amount of synthesized track circuit data as a supplement to the original training data to expand the data set.

[0076] It should be noted that through the above method, a variety of track circuit fault data samples can be generated, effectively alleviating the problem of data scarcity and imbalance. These expanded data samples will be used as part of the training set to train the first diagnosis model, thereby improving the model's diagnostic performance and generalization ability.

[0077] S102, using the first data after the first preprocessing as a sample set to train a first diagnostic model;

[0078] In the embodiment of the present application, the first diagnostic model includes:

[0079] The first diagnostic model is any model that takes as input the first data of the target track circuit and outputs the probability distribution of different fault categories of the track circuit or can directly or indirectly obtain relevant parameters of the probability distribution of different fault categories of the track circuit.

[0080] In an optional embodiment, the relevant parameters of the probability distribution of different fault categories of the track circuit may be directly or indirectly obtained, including but not limited to the confidence of the fault category, the score of the fault category, the probability density function value of the fault category, etc. These relevant parameters can directly or indirectly reflect the possibility of different fault categories of the track circuit, thereby helping the diagnosis system to more accurately identify the fault type of the track circuit.

[0081] In an optional embodiment, the first diagnostic model can be constructed by a deep learning framework, such as TensorFlow or PyTorch. These frameworks provide a wealth of neural network components and optimization algorithms, which can support the construction and training of complex models.

[0082] In another optional embodiment, when constructing the first diagnostic model, a suitable neural network architecture may be selected, such as a convolutional neural network (CNN), a recurrent neural network (RNN) or its variants, such as a long short-term memory network (LSTM) and a gated recurrent unit (GRU). These network architectures are capable of learning the spatial and temporal characteristics of data, and have advantages in analyzing track circuit fault data.

[0083] In another optional embodiment, in order to further improve the performance of the model, a transfer learning method can be used. Transfer learning allows us to use a model that has been trained in a related field or task and adapt it to a new task or data set through fine-tuning. This can greatly shorten the time of model training and may improve the accuracy of the model.

[0084] In an optional embodiment, the first diagnostic model can also integrate multiple diagnostic algorithms, such as support vector machine (SVM), random forest, etc., to form an integrated learning model, and further improve the accuracy and stability of fault diagnosis by integrating the prediction results of multiple models.

[0085] In an optional embodiment, the first diagnostic model may also use model fusion technology to combine multiple models with different architectures and parameters, utilize their respective advantages, achieve complementary advantages, and thus improve overall performance.

[0086] In an embodiment of the present application, the first diagnostic model ConvLSTM deep learning network model is trained, and 80% of the expanded data set (i.e., the first data after the first preprocessing) is used as a training data set for training the ConvLSTM deep learning network model, and the remaining 20% ​​is used as a test data set for testing and verifying the ConvLSTM deep learning network model.

[0087] In an embodiment of the present application, the first diagnostic model also includes an input layer, a one-dimensional convolutional layer, an LSTM layer, a multi-head attention mechanism layer, a fully connected layer and an output layer.

[0088] In an embodiment of the present application, the one-dimensional convolution layer includes two one-dimensional convolution layers. The convolution kernel of the one-dimensional convolution layer slides from the starting time step of the track circuit data, multiplies and adds the local track circuit voltage data, generates a new eigenvalue at each time step, and extracts the local correlation and short-term pattern in the track circuit data.

[0089] In the embodiment of the present application, a ConvLSTM deep learning network model is constructed:

[0090] The constructed ConvLSTM deep learning network model consists of a one-dimensional convolutional layer, an LSTM layer, a multi-head attention mechanism layer, a fully connected layer, and an output layer connected in sequence, as shown in Figure 2. Figure 4 As shown;

[0091] Among them, the one-dimensional convolution layer sets two layers of one-dimensional convolution, the convolution kernel sizes of the first and second layers are 3 and 5 respectively, and the activation function replaces the original Relu activation function with LeakyRelu; Figure 5 As shown in the figure, the convolution kernel starts to slide from the starting time step of the track circuit data, multiplies and adds the local track circuit voltage data, generates new eigenvalues ​​at each time step, extracts local correlation and short-term patterns in the track circuit data, and improves the model's ability to understand the characteristics of the track circuit data. The specific formula of the one-dimensional convolution layer is as follows:

[0092]

[0093] in, represents the i-th convolution kernel of the l-th layer, represents the jth local input of the l-1th layer, It represents the output features of the track circuit data after convolution operation and activation function. The operator * represents the convolution operation. is the i-th bias of the l-th layer, N represents the convolution calculation area, and f is the activation function.

[0094] In the embodiment of the present application, the number of LSTM layers is set to 2, the number of hidden units in each LSTM layer is 128, the output from the one-dimensional convolution layer is received, and the long-term dependencies in the time series data are captured through the memory unit and the gating mechanism, such as Figure 6 As shown, the formula for a single LSTM hidden unit is as follows:

[0095] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0096] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0097]

[0098] o t =σ(W o ·[h t-1 ,x t ]+b o )

[0099] h t =o t tanh(C t )

[0100] Among them, h t-1 is the output value of the LSTM hidden unit at time t-1, x t is the input value of the LSTM hidden unit at time t; W i , W f , W o , W C and b i , b f , b o , b C Respectively represent the weight matrix and bias of the input gate, forget gate, output gate, and candidate memory unit; i t , f t , o t They are the output states of the input gate, forget gate, and output gate at time t respectively; is the candidate memory unit at time t, C tis the memory unit updated at time t; σ is the Sigmoid activation function, which is used for the forget gate, input gate and output gate, and outputs a value between 0 and 1, indicating the degree of information forgetting or the degree of information passing; tanh(x) is the hyperbolic tangent function, which is used to generate candidate memory units and outputs a value between -1 and 1, indicating the new state of the information.

[0101] In an optional embodiment, in order to avoid the vanishing gradient due to the increase of network depth, a residual connection is introduced in the LSTM layer, and the input of the LSTM layer is directly added to the output to enhance the information flow and training stability. The specific calculation formula of the residual connection is as follows:

[0102] H C =LSTM(Y)

[0103] Z c =H C +Linear(Y)

[0104] Where Y represents the output value of the track circuit data after passing through the one-dimensional convolution layer, that is, the input of the LSTM layer; H C is the output value after the input data passes through the LSTM layer; Linear() is a linear mapping function used to adjust the dimension of Y and H C Consistent, ensuring that the residual connection can be directly added; Z c Represents the output value after the LSTM layer introduces residual connection.

[0105] In the embodiments of the present application, Figure 7 As shown, the multi-head attention mechanism layer sets the number of attention heads to 8 and inputs the data Z c The input is given to 8 attention heads, parameterized to different feature subspaces through linear projection, and each attention head is processed in parallel. Then the results of each head's attention calculation are spliced ​​and projected to obtain the output matrix to extract the global feature information of the track circuit data. Through different attention heads, different features of the track circuit data are paid attention to from multiple angles, which improves the diversity of feature expression and global modeling capabilities. Each attention head in the multi-head attention mechanism has three very important matrix parameters Q, K, and V, and the calculation formula is as follows:

[0106] Q i =ZW i Q , K i =ZW i K , V i =ZW i V

[0108] Among them, Q i , Ki 、V i are the query matrix, key matrix, and value matrix in the i-th attention head, respectively, and W i Q , W i K , W i V are the projection matrices of the query, key, and value in the i-th attention head, respectively, and Z represents the same input matrix;

[0109] In the embodiment of the present application, the calculation formula of each attention head is as follows:

[0110]

[0111] Among them, head i represents the result of the self-attention calculation of the i-th attention head in its subspace, Attention represents the attention weight calculation function, softmax represents the normalization function, and d k is the dimension of the key, is the scaling factor;

[0112] In the embodiment of the present application, the calculation formula of the multi-head attention mechanism is as follows:

[0113] O=MultiHead(Q,K,V)=Concat(head 1 ,head 2 ,...,head n )W O

[0114] Among them, O represents the output value after the multi-head attention mechanism, MultiHead represents the multi-head attention function, Q, K, V represent all query matrices, key matrices and value matrices in the multi-head attention mechanism, Concat represents the concatenation of the outputs of all attention heads, and W O Represents the output weight matrix.

[0115] In an embodiment of the present application, the fully connected layer maps and integrates the extracted track circuit fault data features; in the output layer, the Softmax activation function is used to convert the final output of the fully connected layer into a probability distribution of different track circuit fault categories.

[0116] It should be noted that using the first data after the first preprocessing as a sample set and training the first diagnostic model can make full use of the preprocessed and enhanced track circuit data and improve the training effect of the model. Through deep learning network models, such as ConvLSTM, it is possible to automatically learn complex features in the data and accurately classify track circuit faults. In addition, by using a one-dimensional convolutional layer to extract local features, an LSTM layer to capture long-term dependencies, and a multi-head attention mechanism layer to extract global features, the characteristics of the track circuit data can be fully understood, thereby improving the accuracy and stability of fault diagnosis. This training method not only improves the generalization ability of the model, but also provides strong support for subsequent fault diagnosis.

[0117] S103, verifying the trained first diagnostic model, and performing target track circuit fault diagnosis according to the verified first diagnostic model.

[0118] In an optional embodiment, the first diagnostic model after verification training can be performed by cross-validation, holdout method or other verification methods. Cross-validation is to divide the data set into k parts, take k-1 parts as training data in turn, and the remaining part as test data, perform k training and testing, and finally obtain the average value of the k test results. The holdout rule is to directly divide the data set into a training set and a test set, train the model with the training set, and then test the performance of the model with the test set. These methods can objectively evaluate the performance of the model and ensure that the model is stable and reliable in practical applications.

[0119] In an optional embodiment, when verifying the first diagnostic model, multiple evaluation indicators can be used to comprehensively measure the performance of the model, such as accuracy, precision, recall, F1 score, etc. These evaluation indicators can reflect the accuracy of the model in classifying track circuit faults from different perspectives, and help users select the optimal model parameters and architecture.

[0120] In an optional embodiment, the target track circuit can be diagnosed for faults based on the verified first diagnostic model. The real-time data of the target track circuit is input into the trained first diagnostic model, and the model automatically extracts data features and classifies faults. The accuracy and reliability of the model can be verified by comparing the fault category output by the model with the actual fault condition.

[0121] In an optional embodiment, the first diagnostic model can also be integrated with other diagnostic methods or systems to form a more complete track circuit fault diagnosis system. For example, the first diagnostic model can be combined with an expert system to further optimize and interpret the results output by the model using the knowledge and experience of the expert system. Alternatively, the first diagnostic model can be integrated with other machine learning algorithms to improve the accuracy and stability of fault diagnosis by integrating the prediction results of multiple models.

[0122] In summary, the present invention proposes a track circuit fault diagnosis method, which obtains first data of a target track circuit and performs a first preprocessing on the first data; the first preprocessing at least includes a first expansion operation on the first data; the first data after the first preprocessing is used as a sample set to train a first diagnostic model; the trained first diagnostic model is verified, and the target track circuit fault diagnosis is performed according to the verified first diagnostic model. Through the automated data acquisition, preprocessing, model training and verification process, the present invention can quickly diagnose the fault of the track circuit, greatly shorten the diagnosis time, and improve the diagnosis efficiency. The use of a deep learning algorithm for model training can learn the characteristics and laws of track circuit faults, thereby improving the accuracy of the diagnosis. Through automated fault diagnosis, the present invention can promptly detect and locate faults, avoid equipment damage and downtime caused by faults, and thus reduce maintenance costs.

[0123] Specifically, using generative adversarial networks (GANs) to generate diverse sample data and expand data sets can significantly improve the performance of the model in small sample or unbalanced data scenarios, solve the problem of insufficient track circuit fault data, avoid overfitting of the model, and improve the robustness and generalization ability of the model; optimizing the ConvLSTM deep learning network model, using one-dimensional convolution to extract local features of track circuit data, the LSTM layer to capture long-term dependencies in track circuit data, and the multi-head attention mechanism to extract global features from multiple angles. The combination of the three enhances the feature extraction capability of the model, and can fully and comprehensively mine and learn hidden features in the data; at the same time, introducing residual connections to solve the gradient vanishing problem caused by the increase in network depth, improve training stability, and further improve the accuracy of fault diagnosis.

[0124] Example 2

[0125] In a preferred embodiment, the model is trained and tested using a training data set and a test data set, and the fault diagnosis performance of the model is evaluated according to four evaluation indicators: accuracy, precision, recall, and F1 score. The model is optimized by adjusting the model optimization parameters, such as the convolution kernel size, the number of LSTM hidden units and layers, the number of attention heads, and the learning rate and batch size of the model multiple times.

[0126] Specifically, the initial parameter learning rate of the training model is set to 0.001, the batch size is 32, and the cross entropy loss function is used to handle multi-classification problems to guide model optimization; the Adam optimizer is used to adjust all trainable parameters in the network according to the loss obtained by the cross entropy loss function, and the learning rate of each parameter of the model is adaptively adjusted; the fault diagnosis performance of the model is evaluated by the accuracy, precision, recall and F1 score evaluation indicators to achieve the optimal effect of the model. The calculation formula is as follows:

[0127]

[0128]

[0129] Among them, TP represents the number of samples correctly predicted by the model as positive, TN represents the number of samples correctly predicted by the model as negative, FP represents the number of samples incorrectly predicted by the model as positive (actually negative), and FN represents the number of samples incorrectly predicted by the model as negative (actually positive).

[0130] It should be noted that during the model training process, through continuous iterative optimization, the model's loss function value is gradually reduced, and the fault diagnosis performance is gradually improved. When the performance indicators of the model on the test data set meet the preset requirements, the model training can be considered complete. At this time, the trained model can be used to perform fault diagnosis on the new target track circuit data. In practical applications, only the real-time data of the target track circuit needs to be input into the model, and the model can quickly output the fault diagnosis results, providing strong support for the maintenance and management of the track circuit.

[0131] Example 3

[0132] This embodiment also provides a track circuit fault diagnosis system, including:

[0133] A data acquisition and processing module, used for acquiring first data of a target track circuit and performing first preprocessing on the first data;

[0134] The first preprocessing at least includes performing a first expansion operation on the first data;

[0135] A model training model, used to train a first diagnostic model using the first data after the first preprocessing as a sample set;

[0136] The diagnostic module is used to verify the first diagnostic model after training and perform target track circuit fault diagnosis based on the verified first diagnostic model.

[0137] The above-mentioned unit modules may be embedded in or independent of a processor in a computer device in the form of hardware, or may be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to the above-mentioned modules.

[0138] This embodiment also provides a computer device, which can be a terminal. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a track circuit fault diagnosis method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse.

[0139] This embodiment further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0140] Acquire first data of a target track circuit, and perform first preprocessing on the first data;

[0141] The first preprocessing at least includes performing a first expansion operation on the first data;

[0142] Using the first data after the first preprocessing as a sample set, training a first diagnostic model;

[0143] The trained first diagnostic model is verified, and target track circuit fault diagnosis is performed according to the verified first diagnostic model.

[0144] 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 preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0145] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code. The scheme in the embodiments of the present application can be implemented in various computer languages, for example, object-oriented programming language Java and literal scripting language JavaScript, etc.

[0146] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0147] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0148] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0149] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0150] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A track circuit fault diagnosis method, characterized in that: include: Acquire first data of a target track circuit, and perform first preprocessing on the first data; The first preprocessing at least includes performing a first expansion operation on the first data; Using the first data after the first preprocessing as a sample set to train a first diagnostic model; The trained first diagnostic model is verified, and target track circuit fault diagnosis is performed according to the verified first diagnostic model.

2. The track circuit fault diagnosis method according to claim 1, characterized in that: The first diagnostic model comprises: The first diagnostic model is any model that takes as input the first data of the target track circuit and outputs the probability distribution of different fault categories of the track circuit or can directly or indirectly obtain relevant parameters of the probability distribution of different fault categories of the track circuit.

3. The track circuit fault diagnosis method according to claim 2, characterized in that: The first preprocessing comprises: performing a first expansion operation on the first data; The first expansion operation is performed by a preset generative adversarial network; The generative adversarial network includes several logical combinations of generators and discriminators.

4. The track circuit fault diagnosis method according to claim 3, characterized in that: The several logical combinations include: Determine the logic combination method and combination target; Determine the input and output of the generator and the discriminator according to the logical combination method; According to the input and output of the generator and the discriminator, a generative adversarial network is established in combination with the logical combination method and the combination target.

5. The track circuit fault diagnosis method according to claim 4, characterized in that: The first diagnostic model also includes an input layer, a one-dimensional convolutional layer, an LSTM layer, a multi-head attention mechanism layer, a fully connected layer and an output layer.

6. The track circuit fault diagnosis method according to claim 5, characterized in that: The one-dimensional convolution layer includes two one-dimensional convolution layers. The convolution kernel of the one-dimensional convolution layer slides from the starting time step of the track circuit data, multiplies and adds the local track circuit voltage data, generates a new eigenvalue at each time step, and extracts the local correlation and short-term pattern in the track circuit data.

7. The track circuit fault diagnosis method according to claim 6, characterized in that: The first data of the target track circuit at least includes the power output voltage, the power output current, the sending end cable side voltage, the sending end cable side current, the receiving end cable side main rail voltage, the receiving end cable side small rail voltage, the receiving end equipment side main rail voltage, the receiving end equipment side small rail voltage, the receiving entrance main rail voltage, the receiving entrance small rail voltage, the main rail relay and the small rail relay.

8. A track circuit fault diagnosis system, characterized in that: include: A data acquisition and processing module, used for acquiring first data of a target track circuit and performing first preprocessing on the first data; The first preprocessing at least includes performing a first expansion operation on the first data; A model training model, used to train a first diagnostic model using the first data after the first preprocessing as a sample set; The diagnostic module is used to verify the first diagnostic model after training and perform target track circuit fault diagnosis based on the verified first diagnostic model.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

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

Images