Track circuit fault diagnosis method, device, equipment, medium and product

Abstract

The invention belongs to the technical field of track circuit fault diagnosis, and particularly relates to a track circuit fault diagnosis method, device and equipment, a medium and a product. Comprising the following steps: extracting a plurality of fault type voltage curves and a normal voltage curve based on a main rail historical voltage curve; segmenting the fault type voltage curve and the normal voltage curve, and extracting statistical characteristics of each segment; performing correlation analysis and feature screening on the statistical features, and performing dimensionality reduction by adopting a principal component analysis method to obtain a feature data set; constructing a fault classification model, and inputting the feature data set for training; and extracting feature data from the real-time fault voltage data of the track circuit, and inputting the feature data into the trained fault classification model for fault diagnosis. According to the invention, the accuracy, reliability and intelligent level of track circuit fault diagnosis are effectively improved.

Description

Technical Field

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

[0002] Track circuits are installed in stations and along the railway lines. They are in an outdoor environment for a long time and are easily disturbed by the external environment, which greatly increases the possibility of equipment failure. Its working conditions are harsh and changeable, and the operation and maintenance work is complicated, which has caused many difficulties in the diagnosis and maintenance of track circuit faults. At present, the maintenance of track circuit equipment mainly relies on regular inspections by on-site maintenance personnel. However, the large number of track circuit equipment makes the workload of maintenance personnel extremely heavy; considering the complex structure of the equipment itself, relying solely on the on-site maintenance experience of maintenance personnel to diagnose track circuit faults is not only inefficient, but also difficult to guarantee reliability. Therefore, how to efficiently realize the intelligent diagnosis of track circuit equipment faults is crucial to improving train operation efficiency and driving safety. Summary of the invention

[0003] In view of the shortcomings of the prior art, the present invention provides a track circuit fault diagnosis method, comprising: Based on the historical output voltage curve of the main rail, several fault type voltage curves and a normal voltage curve are extracted; The fault type voltage curve and the normal voltage curve are segmented, and statistical features are extracted for each segment; Perform correlation analysis and feature screening on statistical features, and use principal component analysis to reduce the dimension to obtain feature data sets; Build a fault classification model and input feature data sets for training; Feature data is extracted from the real-time fault voltage data of the track circuit, and the feature data is input into the trained fault classification model for fault diagnosis.

[0004] Furthermore, the fault types include: slow train, poor branch characteristics, poor compensation capacitor, loose plugs, broken rail fault, poor tuning unit, and poor attenuation disk.

[0005] Furthermore, the statistical characteristics include: maximum value, minimum value, total value, mean value, median, standard deviation, skewness, and kurtosis.

[0006] Furthermore, the correlation analysis includes: performing correlation analysis using the Pearson correlation coefficient.

[0007] Furthermore, the fault classification model is constructed, including: an LSTM layer and a fully connected layer.

[0008] Furthermore, the input feature data set is trained, including: using a softmax function to convert the output, using a cross entropy function as a loss function and a stochastic gradient descent as an optimizer to evaluate network performance, and updating network parameters according to the loss function result through a back propagation algorithm.

[0009] The present invention also provides a track circuit fault diagnosis device, comprising: Curve data extraction module: used to extract several fault type voltage curves and a normal voltage curve based on the main rail historical output voltage curve; Feature extraction module: used to segment the fault type voltage curve and the normal voltage curve, and extract statistical features for each segment; Feature analysis module: used to perform correlation analysis and feature screening on statistical features, and use principal component analysis to reduce the dimension to obtain feature data sets; Model building module: used to build a fault classification model and input feature data sets for training; Fault diagnosis module: used to extract feature data from the real-time fault voltage data of the track circuit and input the feature data into the trained fault classification model for fault diagnosis.

[0010] The present invention provides a computer device, which includes a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor executes the steps of the track circuit fault diagnosis method.

[0011] The present invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the track circuit fault diagnosis method.

[0012] The present invention provides a computer program product, including a computer program, wherein the computer program is executed by a processor to implement the track circuit fault diagnosis method. Compared with the prior art, the present invention has the following advantages: 1. Through reasonable data collection and analysis strategies, accurately capture the voltage change characteristics when the track circuit fails, obtain multiple types of voltage curve data, and provide an effective basis for fault diagnosis; 2. Use feature extraction and dimensionality reduction technology to deeply explore the inherent laws of data, build high-quality feature data sets, and greatly improve data processing efficiency and accuracy; 3. With the help of a carefully constructed fault classification model, the LSTM layer and the fully connected layer work together, combined with the softmax function, cross entropy function, optimizer and back propagation algorithm, accurate identification and classification of track circuit fault types can be achieved, effectively improving the accuracy, reliability and intelligence level of track circuit fault diagnosis.

[0013] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. 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 paying creative work.

[0015] Figure 1 A flow chart of a track circuit fault diagnosis method provided in an embodiment of the present invention is shown; Figure 2 A schematic diagram showing a normal type of track circuit voltage curve of a track circuit fault diagnosis method provided in an embodiment of the present invention; Figure 3 A schematic diagram of a track circuit voltage curve of a slow-moving train in a track circuit fault diagnosis method provided in an embodiment of the present invention is shown; Figure 4 A schematic diagram showing a voltage curve of a track circuit with poor shunt characteristics in a track circuit fault diagnosis method provided in an embodiment of the present invention; Figure 5 A schematic diagram showing a voltage curve of a track circuit with poor compensation capacitance in a track circuit fault diagnosis method provided in an embodiment of the present invention; Figure 6 A schematic diagram of a track circuit voltage curve of a loosened pin in a track circuit fault diagnosis method provided in an embodiment of the present invention is shown; Figure 7 A schematic diagram of a track circuit voltage curve of a broken rail fault in a track circuit fault diagnosis method provided in an embodiment of the present invention is shown; Figure 8 A schematic diagram showing a track circuit voltage curve of a bad tuning unit in a track circuit fault diagnosis method provided in an embodiment of the present invention is shown; Fig. 9 A schematic diagram showing a voltage curve of a track circuit with a bad attenuation disk in a track circuit fault diagnosis method provided in an embodiment of the present invention is shown; Fig.10 A schematic diagram of a fault classification model of a track circuit fault diagnosis method provided in an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0016] 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.

[0017] See also Figure 1 As shown, an application embodiment of the present invention provides a track circuit fault diagnosis method, comprising: S1: Based on the historical output voltage curve of the main rail, several fault type voltage curves and a normal voltage curve are extracted.

[0018] During the operation monitoring of the track circuit, when a fault occurs, the voltage changes of each track circuit usually tend to stabilize within two minutes. Based on this feature, voltage data is collected once per second during the two-minute period and used as the data basis for experiments and analysis. By fixing the starting time of the voltage curve, the time when the train passes the track can be determined based on the time node, that is, the connection point of the curve before and after the train passes. Since there is a significant difference in the voltage curve before and after the train passes, this difference can be directly related to different fault situations.

[0019] It should be noted that the main rail output voltage presents different numerical characteristics under different working conditions of the track circuit. When the track circuit is in the adjustment state, that is, when the track is idle and the equipment is intact and there is no train occupying it, the main rail output voltage has a specific stable value; and when the track circuit enters the shunt state, that is, when the train occupies the track and causes the circuit to be shunt, the main rail output voltage will change significantly, and its value is different from that in the adjustment state. In the overall curve record of the main rail output voltage, the proportion of the adjustment state and the shunt state is not fixed, and the curve contains parts of both states at the same time.

[0020] During the operation of the track circuit, the track circuit switches between the adjustment state and the shunt state due to the entry and exit of the train. At this time, the voltage change is plotted as a curve. During the passage of a train, it includes the adjustment state before the train enters the track (the first curve part), the shunt state after the train enters the track, and the adjustment state after the train leaves (the second curve part). These two curves are adjacent, and through them, we can analyze how the main rail output voltage changes when the train enters and exits the track. This change is very critical for judging whether there is a fault in the track circuit and the type of fault.

[0021] Therefore, after selecting the division according to the branch status, the last 20 seconds of the previous curve and the first 100 seconds of the next curve can clearly show the change trend of the main track output voltage before and after the train passes, and clearly know whether it is rising or falling, so as to effectively distinguish different fault types. With the help of the microcomputer monitoring system, track circuit voltage curve data including seven fault types and one normal type can be obtained.

[0022] Seven fault types and one normal type (see Figure 2 ) track circuit voltage curve: See also Figure 3 As shown, the train is running slowly: the step-up curve, the main rail output voltage rises step-by-step from the shunt state to the adjustment state; See also Figure 4 As shown, the shunt characteristics are poor: the main rail output voltage drops and rises slowly before and after the shunt; See also Figure 5 As shown, the compensation capacitor is bad: the rail voltage is in the adjustment state, and the main rail output voltage drops; See also Figure 6 As shown, the plug is loose: the track circuit is in the adjustment state, the main rail output voltage fluctuates, and the section is red; See also Figure 7 As shown, rail break fault: the main rail output voltage drops to 0mv; See also Figure 8 As shown, the tuning unit is bad: the main rail output voltage fluctuates after the vehicle passes, and then drops by more than 100mv after stabilization; See also Fig. 9 As shown, the attenuation disk is bad: after the train clears the section, the main rail output voltage rises very little; S2: Divide the fault type voltage curve and the normal voltage curve into segments, and extract statistical features for each segment.

[0023] The relevant static properties of the on-site track circuit, such as the track circuit number, time and the original data of the main rail output voltage, are stored in the form of a data table to provide a data basis for subsequent processing.

[0024] Since each voltage curve lasts 120 seconds, each main rail output voltage curve is divided into 4 sections by cutting every 30 data. For the overall curve and each segmented curve, the eight statistical indicators of maximum value, minimum value, total value, average value, median (sorted first), standard deviation, skewness, and kurtosis are calculated respectively, and a total of 40 ((4 segmented curves + 1 overall curve) × 8 statistical indicators) statistical features are accumulated. Among them, the maximum and minimum values ​​reflect the voltage range, the total value, average value and median reflect the overall status of each group of voltage data, the standard deviation measures the discrete trend, the kurtosis describes the morphological difference with the normal distribution, and the skewness measures the skew direction and degree of data distribution. The original data is mapped to the [0-1] interval, and the maximum value of the original data is normalized to 1.

[0025] S3: Perform correlation analysis and feature screening on statistical features, and use principal component analysis to reduce the dimension to obtain the feature data set.

[0026] For the standardized main rail output voltage data set, the Pearson correlation coefficient was first used to perform correlation analysis on the eight basic statistical indicator variables. It was found that the sum, mean, and median were highly correlated, so the two characteristic variables, the sum and median, were removed, and the mean was retained to reflect the overall situation. The remaining six statistical indicators formed a total of 30 ((4 segmented curves + 1 overall curve) × 6 statistical indicators) characteristic values. Then the 30 characteristic values ​​of these six statistical indicators were analyzed again for correlation. The results showed that the minimum, maximum, and mean values ​​were highly correlated, while the standard deviation, skewness, and kurtosis were not highly correlated.

[0027] It should be noted that when the correlation coefficient is between 0.8-1.0, the correlation is extremely strong; when the correlation coefficient is between 0.6-0.8, the correlation is relatively strong; when the correlation coefficient is between 0.4-0.6, the correlation is not strong; and when the correlation coefficient is between 0-0.2, there is no correlation.

[0028] For the features with strong correlation, principal component analysis is used to reduce the dimension, and finally obtain the feature data set. The principal component is a new variable formed by linear combination of the original variables. The core of principal component analysis lies in "dimensionality reduction", that is, integrating multiple original indicators into a few independent comprehensive indicators. In this process, most of the information of the original variables can be retained, effectively avoiding a large amount of information loss.

[0029] First, construct an original matrix of n rows and p columns, which consists of relevant sample data; centralize the sample data to make the data easier to analyze and calculate during the processing; calculate the covariance matrix; calculate the eigenvalues ​​and corresponding eigenvectors of the covariance matrix, where the eigenvector is a non-zero vector that describes the structure of the data set; diagonalize the covariance matrix to obtain the matrix P, whose column vector group is the n standard orthogonal eigenvectors of the covariance matrix, and take the first k rows to form the matrix P. In this process, there are three main bases for determining the number of principal components: first, the eigenvalue is greater than 1; second, the cumulative variance contribution reaches an acceptable level (such as more than 80%); third, based on the steep slope information of the gravel map, the number of principal components is determined when an obvious turning point is observed.

[0030] In actual operation, SPSSAU software is used to perform automatic principal component analysis, and the software will output information such as eigenvalues, variance ratios, and scree plots. Based on this information and the basis for determining the number of principal components mentioned above, it is determined that when two principal components are selected, the cumulative variance contribution rate is the largest.

[0031] The data set contains the standard deviation, kurtosis, skewness and two principal components of the entire data and each segmented data. These features can be used for subsequent model training and other related work, which helps to improve processing efficiency and accuracy and more accurately reflect the characteristics and laws of the main rail output voltage data.

[0032] S4: Build a fault classification model and input feature dataset for training.

[0033] The fault classification model is mainly composed of an LSTM layer (long short-term memory network layer) and a fully connected layer (FC). Its purpose is to build a complex nonlinear model between the input and output to achieve effective classification of track circuit faults.

[0034] See also Fig.10 As shown in the figure, in the fully connected layer, the softmax function is used to process the output of the last neuron in the LSTM_1 layer, thereby converting it into a probability distribution corresponding to the seven types of track circuit faults. This is because there are a total of seven types of track circuit faults, so the number of neurons set in the fully connected layer is 7.

[0035] The cross entropy function is used as the loss function in the LSTM network to evaluate the performance of the entire network. The specific formula is:

[0036] In this formula, is the label value (1 for positive class and 0 for negative class), is the predicted probability value (the value range is between (0,1)).

[0037] This function reflects the difference between the real sample label and the predicted probability. By calculating the difference, we can measure the degree of deviation between the network prediction result and the actual situation, and then evaluate the quality of the network performance.

[0038] In the model involved in this article, the optimizer uses stochastic gradient descent (SGD). In the early stage of model training, the initial learning rate is set to 0.1, and as the number of iterations increases, the learning rate will gradually decrease. This dynamic adjustment of the learning rate helps the model converge better during training and improves the accuracy and stability of the model.

[0039] During the operation of the LSTM network, after using the cross entropy function as the loss function to evaluate the network performance, the network parameters will be updated through the back propagation algorithm based on the results calculated by the loss function, so that the network can continuously optimize its own prediction capabilities and thus more accurately classify track circuit faults.

[0040] S5: Extract feature data from the real-time fault voltage data of the track circuit, and input the feature data into the trained fault classification model for fault diagnosis.

[0041] An application embodiment of the present invention provides a track circuit fault diagnosis device, comprising: Curve data extraction module: used to extract several fault type voltage curves and a normal voltage curve based on the main rail historical output voltage curve; Feature extraction module: used to segment the fault type voltage curve and the normal voltage curve, and extract statistical features for each segment; Feature analysis module: used to perform correlation analysis and feature screening on statistical features, and use principal component analysis to reduce the dimension to obtain feature data sets; Model building module: used to build a fault classification model and input feature data sets for training; Fault diagnosis module: used to extract feature data from the real-time fault voltage data of the track circuit and input the feature data into the trained fault classification model for fault diagnosis.

[0042] An application embodiment of the present invention provides a computer device, which includes a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor executes the steps of the track circuit fault diagnosis method.

[0043] An application embodiment of the present invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of a track circuit fault diagnosis method.

[0044] An embodiment of the present invention further provides a computer program product corresponding to the track circuit fault diagnosis method provided by the aforementioned embodiment. The computer program product includes a computer program, and the computer program is executed by a processor to implement the track circuit fault diagnosis method provided by the aforementioned embodiment.

[0045] The above description and the accompanying drawings fully illustrate the embodiments of the present invention so that those skilled in the art can practice them. Other embodiments may include structural and other changes. The embodiments represent only possible variations. Unless explicitly required, individual components and functions are optional, and the order of operations may vary. Parts and features of some embodiments may be included in or replace parts and features of other embodiments. The embodiments of the present invention are not limited to the structures described above and shown in the accompanying drawings, and various modifications and changes may be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A track circuit fault diagnosis method, characterized in that: include: Based on the historical output voltage curve of the main rail, several fault type voltage curves and a normal voltage curve are extracted; The fault type voltage curve and the normal voltage curve are segmented, and statistical features are extracted for each segment; Perform correlation analysis and feature screening on statistical features, and use principal component analysis to reduce the dimension to obtain feature data sets; Build a fault classification model and input feature data sets for training; Feature data is extracted from the real-time fault voltage data of the track circuit, and the feature data is input into the trained fault classification model for fault diagnosis.

2. The track circuit fault diagnosis method according to claim 1, characterized in that: The fault types include: slow train, poor branch characteristics, poor compensation capacitor, loose plugs, broken rail faults, poor tuning units, and poor attenuation disks.

3. The track circuit fault diagnosis method according to claim 1, characterized in that: The statistical characteristics include: maximum value, minimum value, total value, mean value, median, standard deviation, skewness, and kurtosis.

4. The track circuit fault diagnosis method according to claim 1, characterized in that: The correlation analysis includes: performing correlation analysis using the Pearson correlation coefficient.

5. The track circuit fault diagnosis method according to claim 1, characterized in that: The fault classification model is constructed, including: an LSTM layer and a fully connected layer.

6. The track circuit fault diagnosis method according to claim 1, characterized in that: The input feature data set is trained, including: using the softmax function to convert the output, using the cross entropy function as the loss function and the stochastic gradient descent as the optimizer to evaluate the network performance, and updating the network parameters according to the loss function result through the back propagation algorithm.

7. A track circuit fault diagnosis device, characterized in that: include: Curve data extraction module: used to extract several fault type voltage curves and a normal voltage curve based on the main rail historical output voltage curve; Feature extraction module: used to segment the fault type voltage curve and the normal voltage curve, and extract statistical features for each segment; Feature analysis module: used to perform correlation analysis and feature screening on statistical features, and use principal component analysis to reduce the dimension to obtain feature data sets; Model building module: used to build a fault classification model and input feature data sets for training; Fault diagnosis module: used to extract feature data from the real-time fault voltage data of the track circuit and input the feature data into the trained fault classification model for fault diagnosis.

8. A computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the computer program is executed by a processor, the processor executes the steps of the track circuit fault diagnosis method described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the processor executes the steps of the track circuit fault diagnosis method described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that The computer program is executed by a processor to implement the track circuit fault diagnosis method according to any one of claims 1 to 6.

Images