Railway signal system fault early warning method, system, equipment and medium

Abstract

The invention relates to industrial railway diagnosis and prediction, and discloses a railway signal system fault early warning method, which comprises the following steps: acquiring historical operation data of a railway signal system, performing data cleaning, constructing a feature data set of the cleaned data in a plurality of historical time sequence intervals, selecting a preset number of historical time sequence intervals as target intervals, and performing fault early warning on the target intervals. Generating an interval label according to the feature data set included in each target interval, screening the feature data sets in the target intervals to obtain a training sample, obtaining a time sequence training model established based on a bidirectional long-short term neural network, training the time sequence training model by using the training sample to obtain a time sequence analysis model, and obtaining a time sequence analysis result; and acquiring system operation data of the railway signal system, analyzing the system operation data by using the time sequence analysis model, and performing fault early warning according to a result. According to the method, the analysis error of the operation data of the railway signal system can be reduced, and the accuracy of railway signal fault hidden danger analysis and the timeliness of fault early warning are improved.

Description

Technical Field

[0001] The present invention relates to the field of industrial railway diagnosis and prediction, and in particular to a railway signal system fault early warning method, system, equipment and medium. Background Art

[0002] At present, the industrial railway overhaul has not yet established a condition assessment technology system, and the overhaul is still mainly planned. The operating conditions of each line are different, and the maintenance methods are different. The planned maintenance method will inevitably lead to a certain degree of advance or lag, affecting the maintenance efficiency and equipment safety. At the same time, the line equipment condition assessment and maintenance management have not considered the entire life cycle of design, production, transportation, laying, use, maintenance, and off-road disposal. The correlation between each link is not strong, and the track structure design and later maintenance methods are rarely considered from the perspective of the entire life cycle. It is difficult to achieve the lowest cost and the highest cost performance in the entire life cycle.

[0003] Traditional "planned maintenance" is actually a blanket inspection. There is no scientific basis for maintenance, which will cause a lot of waste of manpower and material resources. Excessive and repeated maintenance, and the equipment being opened and assembled many times, also increase the risk of human failure. Frequent on-site line operations also pose a hidden danger to personal safety.

[0004] Establish a maintenance and repair model based on condition-based repair, establish an equipment condition assessment technology system based on full life cycle management, improve equipment quality, and enhance maintenance and repair efficiency and benefits. To achieve the above goals, the following three concepts are proposed.

[0005] "Condition-based maintenance" means real-time monitoring of the status of each device in the railway signal system, analyzing the health status of the equipment, making health predictions and preventive maintenance warnings for possible hidden faults, and then providing maintenance personnel with scientific maintenance plans, transforming the existing fixed-period equipment inspections into rationalized and intelligent on-demand inspections.

[0006] "Pre-emptive repair" means that the system monitors the health status of on-site equipment in real time, predicts possible faults and provides preventive maintenance warnings, and performs maintenance before equipment failure occurs, eliminating the failure in the bud and ensuring the normal operation of the equipment. This avoids production stoppages caused by equipment failures and ensures railway transportation safety and uninterrupted continuous production.

[0007] "Fault diagnosis" means that in the event of a device failure, the system can quickly perform online intelligent fault diagnosis, analyze the cause of the equipment failure, give recommended maintenance measures, and assist maintenance personnel to quickly repair the equipment and restore normal operation.

[0008] Currently, industrial railways are transforming from the traditional planned maintenance mode to condition-based maintenance, advance maintenance, and fault diagnosis modes. The existing technology has problems such as large errors in the analysis of railway signal system operation data, low accuracy in the analysis of railway signal fault hazards, and insufficient timely fault warning. Summary of the invention

[0009] The present invention provides a railway signal system fault warning method, system, electronic equipment and computer-readable storage medium, the main purpose of which is to solve the problems of large errors in railway signal system operation data analysis, low accuracy in railway signal fault hidden danger analysis and insufficient timeliness of fault warning.

[0010] To achieve the above object, the present invention provides a railway signal system fault early warning method, comprising:

[0011] Acquire historical operation data of a railway signal system, and perform data cleaning on the historical operation data to obtain cleaned data;

[0012] Constructing a feature data set of the cleaned data within a plurality of preset historical time series intervals;

[0013] Select a preset number of historical time series intervals as target intervals, and generate interval labels for each target interval based on the feature data set contained in each target interval;

[0014] Screening the feature data set within the target interval according to the interval label to obtain a training sample;

[0015] Obtaining a time series training model based on a bidirectional long-short term neural network, and using the training samples to train the time series training model to obtain a time series analysis model;

[0016] The system operation data of the railway signal system in the target time sequence interval is obtained, the system operation data is analyzed using the time sequence analysis model, and a fault warning is issued according to the analysis result.

[0017] In one embodiment of the present invention, the step of performing data cleaning on the historical operation data to obtain cleaned data includes:

[0018] Filling missing values ​​of the historical operation data using a linear interpolation method to obtain preliminary optimized data;

[0019] Calculating the first-order difference of each data in the preliminary optimized data;

[0020] Select target data and adjacent data one by one from the preliminary optimized data, and calculate the dislocation inner product corresponding to the target data according to the first-order differences corresponding to the target data and the adjacent data;

[0021] The data in the preliminary optimized data whose misaligned inner product is less than or equal to a preset threshold are collected as cleaned data.

[0022] In one embodiment of the present invention, constructing a feature data set of the cleaned data within a plurality of preset historical time series intervals includes:

[0023] Sorting the cleaned data, and counting the five-number features in the sorted cleaned data;

[0024] Calculating sample entropy and permutation entropy of the cleaned data;

[0025] The five-number features, the sample entropy and the permutation entropy are aggregated into a feature data set of the cleaned data.

[0026] In one embodiment of the present invention, generating an interval label for each target interval according to a feature data set included in each target interval includes:

[0027] Calculate the feature values ​​corresponding to the feature data sets contained in each target interval one by one;

[0028] Assigning the interval label of the target interval corresponding to the characteristic value less than or equal to the first preset threshold as a fault label;

[0029] Assigning an interval label of a target interval where the characteristic value is between a first preset threshold and a second preset threshold as a degraded label;

[0030] The interval label of the target interval corresponding to the characteristic value greater than the second preset threshold is allocated as a normal label.

[0031] In one embodiment of the present invention, the step of screening the feature data set in the target interval according to the interval label to obtain a training sample includes:

[0032] Perform equidistant sampling on the feature data set within the target interval whose interval label is a normal label to obtain a sampled data set;

[0033] The feature data set and the sampling data set in the target interval whose interval labels are the fault labels and the degradation labels are collected into training samples.

[0034] In one embodiment of the present invention, the step of training the time series training model using the training samples to obtain a time series analysis model includes:

[0035] Obtaining sample labels of the training samples;

[0036] Performing time series analysis on the training samples using the time series training model to obtain prediction labels for the training samples;

[0037] Calculating the loss value of the time series training model according to the sample label and the predicted label using a regularization term;

[0038] The parameters of the time series training model are optimized according to the loss value to obtain a time series analysis model.

[0039] In one embodiment of the present invention, the analyzing the system operation data by using the timing analysis model and providing a fault warning according to the analysis result includes:

[0040] Analyzing the system operation data using the timing analysis model to obtain prediction data;

[0041] Comparing the predicted data with a first preset threshold and a second preset threshold;

[0042] If the predicted data is less than or equal to a first preset threshold, the predicted data is determined to be fault data;

[0043] If the predicted data is between a preset first threshold and a preset second threshold, the predicted data is determined to be degraded data;

[0044] When the predicted data is fault data or degradation data, a fault warning is performed.

[0045] In order to solve the above problems, the present invention also provides a railway signal system fault warning system, the system comprising:

[0046] A data cleaning module is used to obtain historical operation data of the railway signal system, and perform data cleaning on the historical operation data to obtain cleaned data;

[0047] A feature data set construction module, used to construct a feature data set of the cleaned data within a plurality of preset historical time series intervals;

[0048] A label generation module is used to select a preset number of historical time series intervals as target intervals, and generate interval labels for each target interval based on the feature data set contained in each target interval;

[0049] A feature data set screening module is used to screen the feature data set in the target interval according to the interval label to obtain a training sample;

[0050] A model training module is used to obtain a time series training model based on a bidirectional long-term and short-term neural network, and train the time series training model using the training samples to obtain a time series analysis model;

[0051] The data analysis module is used to obtain the system operation data of the railway signal system in the target timing interval, analyze the system operation data using the timing analysis model, and issue a fault warning based on the analysis result.

[0052] In order to solve the above problem, the present invention further provides an electronic device, the electronic device comprising:

[0053] at least one processor; and,

[0054] a memory communicatively connected to the at least one processor; wherein,

[0055] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can perform the steps of the above-mentioned railway signal system fault warning method.

[0056] In order to solve the above problem, the present invention also provides a computer-readable storage medium, in which at least one computer program is stored. The at least one computer program is executed by a processor in an electronic device to implement the steps of the above-mentioned railway signal system fault warning method.

[0057] The present invention performs data cleaning on historical operation data to ensure the quality and integrity of the data, improve the accuracy and reliability of the historical operation data analysis capability and the establishment of a time series analysis model; constructs a feature data set of the cleaned data in multiple preset historical time series intervals, which helps to identify the central trend, dispersion and outliers of the cleaned data, and provides comprehensive data statistical features; generates interval labels based on the feature data set in each target interval, which can effectively count a representative and diverse training sample set, and can also help distinguish the feature patterns and trends of data in different time periods, which helps to more accurately capture the dynamic changes of data; the time series analysis model built by the bidirectional long short-term memory neural network combines the processing of forward and backward information, effectively predicts the trend changes of railway signals and the abnormal conditions of the railway signal system, and performs fault warning, providing the ability of continuous optimization for system operation. Therefore, the present invention proposes a railway signal system fault warning method, system, electronic device and computer-readable storage medium, the main purpose of which is to solve the problems of large errors in the analysis of railway signal system operation data, low accuracy in the analysis of railway signal fault hazards and insufficient timeliness of fault warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 A schematic diagram of a flow chart of a railway signal system fault warning method provided by an embodiment of the present invention;

[0059] Figure 2A functional module diagram of a railway signal system fault warning system provided by one embodiment of the present invention;

[0060] Figure 3 A schematic diagram of the structure of an electronic device for implementing a railway signal system fault warning method provided in an embodiment of the present invention.

[0061] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0062] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0063] The embodiment of the present application provides a railway signal system fault warning method. The execution subject of the railway signal system fault warning method includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiment of the present application. In other words, the railway signal system fault warning method can be executed by software or hardware installed in a terminal device or a server device.

[0064] Reference Figure 1 FIG. 1 is a flow chart of a method for early warning of a railway signal system failure provided by an embodiment of the present invention. In this embodiment, the method for early warning of a railway signal system failure includes:

[0065] S1. Acquire historical operation data of a railway signal system, and perform data cleaning on the historical operation data to obtain cleaned data.

[0066] In an embodiment of the present invention, the historical operation data of the railway signal system includes data such as date, time interval, voltage value, current value, etc.; the historical operation data of the railway signal system can be obtained by configuring a data collector, connecting the required data source, and extracting the historical operation data of the railway signal system through data extraction logic written in a scripting language.

[0067] In an embodiment of the present invention, the data cleaning of the historical operation data to obtain the cleaned data includes: using linear interpolation to fill missing values ​​in the historical operation data to obtain preliminary optimized data; calculating the first-order difference of each data in the preliminary optimized data; selecting target data and adjacent data one by one from the preliminary optimized data, and calculating the misaligned inner product corresponding to the target data based on the first-order difference corresponding to the target data and the adjacent data; and aggregating the data in the preliminary optimized data whose misaligned inner product is less than or equal to a preset threshold as the cleaned data.

[0068] Specifically, the missing values ​​of the historical operation data are filled using the following formula:

[0069]

[0070] in, represents the missing values ​​in the historical operation data, w represents the horizontal coordinate of the missing values ​​in the historical operation data, w0 and w1 represent the horizontal coordinates of the values ​​adjacent to the missing values ​​in the historical operation data, and v0 and v1 represent the vertical coordinates of the values ​​adjacent to the missing values ​​in the historical operation data.

[0071] The present invention utilizes a linear interpolation method to fill missing values ​​in the historical operation data, and can generate a continuous function or curve on limited data points, thereby realizing analysis and prediction of data change trends.

[0072] In the embodiment of the present invention, the formula Δp x+c =p x+c+1 -p x+c Calculate the first-order difference of the data at the x+cth moment in the preliminary optimization data, where p x+c It represents the value of the initial optimization data at the x+cth moment, p x+c+1 represents the value of the initial optimization data at the x+c+1th moment, Δp x+c It represents the first-order difference of the preliminary optimized data at the x+cth moment.

[0073] Specifically, the target data and adjacent data are selected one by one from the preliminary optimization data, wherein the target data is the data at the x+cth moment, and the adjacent data is the data at the x+c+1th moment. The offset inner product corresponding to the target data is calculated based on the first-order difference corresponding to the target data and the adjacent data, specifically by the formula Δp 2 x+c =Δp x+c *Δp x+c+1 The inner product Δp of the initial optimized data at the x+c moment is calculated by offset calculation 2 x+c , where Δp x+c It represents the first-order difference of the initial optimization data at the x+cth moment, Δp x+c+1 It represents the first-order difference of the initial optimization data at the time x+c+1, Δp 2 x+c It represents the inner product of the initial optimized data at the x+cth moment.

[0074] In the embodiment of the present invention, the data whose inner product of the initial optimization data is less than or equal to the preset threshold value are collected as the cleaning data. Specifically, the threshold value is first set to Then the inner product Δp calculated by preliminary optimization data at time x+c is 2 x+c With threshold In contrast, it is necessary to determine whether the preliminary optimization data at time x+c is retained in the historical operation data for time series analysis. If the inner product Then delete the initial optimization data at time x+c, if The preliminary optimized data at the x+cth moment is retained for time series analysis, and finally the retained preliminary optimized data is aggregated to obtain the cleaned data.

[0075] The present invention processes outliers through first-order differences, which can disperse the sudden fluctuations caused by outliers to adjacent time points, thereby better showing the long-term trend and periodic changes of the data, and helping to improve the robustness of the model and the prediction accuracy.

[0076] Furthermore, by obtaining the historical operating data of the railway signal system, the data is cleaned. By cleaning the data, we can delete or correct erroneous, inconsistent or duplicate data items, fill in missing values, and process outliers, so as to make the data set more complete, reduce interference in the model training process, and improve the accuracy of the results.

[0077] S2. Constructing a feature data set of the cleaned data within a plurality of preset historical time series intervals.

[0078] In an embodiment of the present invention, constructing a feature data set of the cleaned data within a plurality of preset historical time series intervals includes: sorting the cleaned data, and counting the five-number features in the sorted cleaned data; calculating the sample entropy and the permutation entropy of the cleaned data; and aggregating the five-number features, the sample entropy, and the permutation entropy into a feature data set of the cleaned data.

[0079] In detail, the characteristic data set of the cleaned data can be constructed by the five-number summary method, which is a method for describing the degree of data dispersion. The five-number features include the minimum value, the first quartile, the median, the third quartile and the maximum value. These features are conducive to quickly understanding the central position, dispersion degree and potential outliers of the data, providing an important reference for the preliminary analysis of the data.

[0080] In the embodiment of the present invention, the sample entropy and permutation entropy of the cleaned data are obtained by calculation, wherein the sample entropy is calculated as follows: a time series consisting of N data is obtained from the cleaned data as {f(n)}=f(1), f(2), f(3), ..., f(N), and the time series consisting of N data is reconstructed into an m-dimensional vector according to the following formula:

[0081] F m (i)={f(i),f(i+1),…,f(i+m-1)},1≤i≤N-m+1

[0082] Among them, Fm (i) represents the m consecutive values ​​of f starting from the i-th point;

[0083] Definition F m (i) to F m (j) Vector distance:

[0084]

[0085] Among them, F m (i) represents the m consecutive values ​​of f starting from the i-th point, F m (j) represents the m consecutive values ​​of f starting from the jth point, d[F m (i),F m (j)] represents F m (i) to F m (j) The distance of the vector, f(i+k) represents the data at the i+kth moment, and f(j+k) represents the data at the j+kth moment;

[0086] Statistics F m (i) to F m The number of j (1≤j≤Nm,j≠i) whose distance (j) is less than or equal to r, and is recorded as B i ; For 1≤i≤Nm, Defined as:

[0087]

[0088] Among them, B i It means F m (i) to F m (j) is the number of j whose distance is less than or equal to r, N represents the number of data obtained from the cleaned data, m represents the dimension of the vector reconstructed from the time series composed of N data, and r represents the similarity tolerance;

[0089] By analyzing the entire time series, we can get:

[0090]

[0091] Among them, N represents the number of data obtained from the cleaned data, m represents the dimension of the vector reconstructed from the time series composed of N data, and B (m) (r) represents F m (i) and F m (j) The probability that two sequences match m points under similarity tolerance r;

[0092] Increase the dimension to m+1, count F m+1 (i) to F m+1(j)(1≤j≤Nm,j≠i) The number of distances less than or equal to r, denoted as A i ; Defined as:

[0093]

[0094] Among them, A i It means F m+1 (i) and F m+1 (j)(1≤j≤Nm,j≠i) is the number of distances less than or equal to r, where N represents the number of data obtained from the cleaned data, m represents the dimension of the vector reconstructed from the time series composed of N data, and r represents the similarity tolerance;

[0095] Analyzing the subsequences of the entire time series, we get:

[0096]

[0097] Among them, N represents the number of data obtained from the cleaned data, m represents the dimension of the vector reconstructed from the time series composed of N data, and A m (r) represents F m+1 (i) and F m+1 (j) The probability that two sequences match m+1 points;

[0098] The final sample entropy is defined as:

[0099]

[0100] Among them, A m (r) represents F m+1 (i) and F m+1 (j) The probability that two sequences match m+1 points, B (m) (r) represents F m (i) and F m (j) The probability that two sequences match m points under the similarity tolerance r. SampEn(m,r,N) represents the sample entropy of the cleaned data.

[0101] The calculation method of permutation entropy is as follows: select a time series {a(t), t=1,2,…,M} with a length of M from the cleaned data, reconstruct the time series {a(t), t=1,2,…,M} in phase space to obtain the following matrix:

[0102]

[0103] Where h represents the embedding dimension, λ represents the delay time, T = M-(h-1)λ, each row in the matrix Z is a reconstruction component, and there are T reconstruction components in total;

[0104] Sort each row of the matrix Z in ascending order:

[0105] a(e+(j1-1)λ≤a(e+(j2-1)λ≤…≤a(e+(j h -1)λ

[0106] Among them, j1,…,j h , represents the index of the column where each element is located, and λ represents the delay time;

[0107] For each row of the matrix, the reconstruction sequence can obtain the corresponding symbol sequence, denoted as S(g):

[0108] S(g)=(j1,j2,…,j h )

[0109] Here, g = {1,…,T} and T ≤ h!, there are at most h! symbol sequences in the h-dimensional phase space, j1,…,j h , represents the index of the column where each element is located;

[0110] Calculate the probability of each sequence appearing as p1,…,p e , define the permutation entropy of the time series a(t) as:

[0111] H p (h)=-∑p e ln(p e )

[0112] The permutation entropy of the time series a(t) is normalized:

[0113]

[0114] Among them, H p (h) represents the complexity of the cleaned data. The smaller it is, the stronger the regularity of the cleaned data is. h represents the embedding dimension.

[0115] Sample entropy is a time series complexity measurement method. The calculation of this method does not depend on the data length and has better consistency, that is, the changes in parameters m and r have the same impact on sample entropy. The permutation entropy algorithm belongs to the algorithm used to detect mutations in dynamics. It has an amplifying effect on small changes in signals and can detect the mutation moment of the signal, which is conducive to health prediction and preventive maintenance warnings for possible fault hazards.

[0116] Constructing a feature data set of the cleaned data within multiple preset historical time series intervals helps to provide a comprehensive data summary and statistical characteristics by describing the central trend, dispersion and identification of outliers of the data. This method can better display the distribution of data, discover anomalies and trends, and thus support more accurate data analysis, decision making and formulation of problem-solving strategies.

[0117] S3. Select a preset number of historical time series intervals as target intervals, and generate an interval label for each target interval based on the feature data set contained in each target interval.

[0118] In an embodiment of the present invention, selecting a preset number of historical time series intervals as the target interval specifically means using a preset sliding window as the target interval. For example, selecting a 10-day sliding window as the target interval means that the ten days from July 3 to July 13 are the target interval.

[0119] In an embodiment of the present invention, the generation of an interval label for each target interval based on the feature data set contained in each target interval includes: calculating the feature values ​​corresponding to the feature data set contained in each target interval one by one; assigning the interval label of the target interval corresponding to the feature value less than or equal to the first preset threshold as a fault label; assigning the interval label of the target interval with a feature value between the first preset threshold and the second preset threshold as a degraded label; and assigning the interval label of the target interval corresponding to the feature value greater than the second preset threshold as a normal label.

[0120] Generating interval labels for each target interval based on the feature data set contained in each target interval can help identify and describe the characteristics and trends of each interval. These labels can be used to distinguish different data patterns or behaviors, which is conducive to more accurate data classification, anomaly detection and model prediction.

[0121] S4. Screening the feature data set within the target interval according to the interval label to obtain training samples.

[0122] In an embodiment of the present invention, the feature data set in the target interval is screened according to the interval label to obtain a training sample, including: equidistant sampling of the feature data set in the target interval whose interval label is a normal label to obtain a sampling data set; and aggregating the feature data set and the sampling data set in the target interval whose interval label is a fault label and a degradation label into a training sample.

[0123] By screening the feature data set within the target interval according to the interval label, training samples that meet specific conditions or standards can be effectively extracted, which is conducive to ensuring the quality and representativeness of the training data, thereby improving the accuracy and generalization ability of the model, reducing the risk of overfitting, and improving the model's prediction ability for unknown data.

[0124] S5. Obtain a time series training model based on a bidirectional long-short term neural network, and use the training samples to train the time series training model to obtain a time series analysis model.

[0125] In an embodiment of the present invention, the time series training model based on the bidirectional long short-term neural network is specifically composed of two LSTMs, one LSTM processes the input sequence in chronological order, and the other LSTM processes the input sequence in reverse chronological order.

[0126] In an embodiment of the present invention, the method of using the training samples to train the timing training model to obtain a timing analysis model includes: obtaining sample labels of the training samples; performing timing analysis on the training samples using the timing training model to obtain predicted labels of the training samples; calculating the loss value of the timing training model based on the sample labels and the predicted labels using a regularization term; and optimizing parameters of the timing training model based on the loss values ​​to obtain a timing analysis model.

[0127] In the embodiment of the present invention, the loss value of the time series training model is calculated according to the sample label and the predicted label using the regularization term, wherein the loss value is calculated using the L2 regularization term, and the formula is as follows:

[0128]

[0129] Among them, C0 represents the original loss function, represents the regularization term, q represents the sample size of the training set, and λ1 represents the regularization term coefficient.

[0130] L2 regularization adds a penalty term to the loss function based on the sum of squares of the model weights to control the complexity and weight size of the model, thereby effectively preventing overfitting, improving the generalization ability and stability of the model, and ensuring the model's good adaptability to new data.

[0131] In detail, the time series training model is trained using the training samples, and the training samples can be specifically divided into a training set, a validation set, and a real test set. In the time series training model, the ratio of the training set, the validation set, and the real test set can be divided into 6:2:2. This ratio of training data can achieve the best model training effect and reduce the model error without destroying the original model structure. During the training process, the last training result of each layer can be used as the initial iteration parameter for the next time, reducing the number of times the model is retrained and improving the efficiency of the model.

[0132] The time series training model built based on the bidirectional long short-term memory neural network can effectively capture the long-term dependencies and complex time patterns in time series data, and improve the model's prediction accuracy and generalization ability for sequence data. The bidirectional long short-term memory neural network combines the processing of forward and backward information, so that the model has better generalization ability when facing new and unseen data, and can more accurately predict and analyze future sequence trends and patterns.

[0133] S6. Obtain system operation data of the railway signal system in a target timing interval, analyze the system operation data using the timing analysis model, and issue a fault warning based on the analysis result.

[0134] In an embodiment of the present invention, the system operation data of the railway signal system in the target timing interval is obtained specifically including date, time, voltage value, and current value.

[0135] In an embodiment of the present invention, the method of analyzing the system operation data using the timing analysis model and performing a fault warning based on the analysis results includes: analyzing the system operation data using the timing analysis model to obtain prediction data; comparing the prediction data with a first preset threshold and a second preset threshold; if the prediction data is less than or equal to the first preset threshold, determining that the prediction data is fault data; if the prediction data is between the preset first threshold and the preset second threshold, determining that the prediction data is degraded data; and performing a fault warning when the prediction data is fault data or degraded data.

[0136] Using the time series analysis model to analyze the continuous time series of system operation data can effectively predict trend changes, detect abnormal situations, and analyze the predicted data in real time. When fault data or degraded data is detected, fault warnings can be issued, providing timely warning capabilities for system operation.

[0137] like Figure 2 , is a functional module diagram of a railway signal system fault warning system provided by an embodiment of the present invention. According to the functions implemented, the railway signal system fault warning system 400 may include a data cleaning module 401, a feature data set construction module 402, a label generation module 403, a feature data set screening module 404, a model training module 405 and a data analysis module 406. The module described in the present invention may also be referred to as a unit, which refers to a series of computer program segments that can be executed by a processor of the railway signal system fault warning system and can complete fixed functions, which are stored in the memory of the railway signal system fault warning system.

[0138] In this embodiment, the functions of each module / unit are as follows:

[0139] A data cleaning module is used to obtain historical operation data of the railway signal system, and perform data cleaning on the historical operation data to obtain cleaned data;

[0140] A feature data set construction module, used to construct a feature data set of the cleaned data within a plurality of preset historical time series intervals;

[0141] A label generation module is used to select a preset number of historical time series intervals as target intervals, and generate interval labels for each target interval based on the feature data set contained in each target interval;

[0142] A feature data set screening module is used to screen the feature data set in the target interval according to the interval label to obtain a training sample;

[0143] A model training module is used to obtain a time series training model based on a bidirectional long-term and short-term neural network, and train the time series training model using the training samples to obtain a time series analysis model;

[0144] The data analysis module is used to obtain the system operation data of the railway signal system in the target timing interval, analyze the system operation data using the timing analysis model, and issue a fault warning based on the analysis result.

[0145] In detail, each module described in a railway signal system fault warning system 400 described in an embodiment of the present invention adopts the same technical means as a railway signal system fault warning method described in the accompanying drawings when used, and can produce the same technical effects, which will not be repeated here.

[0146] like Figure 3 FIG. 1 is a schematic diagram of the structure of an electronic device for implementing a railway signal system fault warning method provided by an embodiment of the present invention.

[0147] The electronic device 1000 may include a processor 1001, a memory 1002, a communication bus 1003 and a communication interface 1004, and may also include a computer program stored in the memory 1002 and executable on the processor 1001, such as a railway signal system fault warning model generation program.

[0148] Among them, the processor 1001 is the control core (Control Unit) of the electronic device, which uses various interfaces and lines to connect the various components of the entire electronic device, and executes various functions of the electronic device and processes data by running or executing programs or modules stored in the memory 1002, and calling data stored in the memory 1002.

[0149] The memory 1002 includes at least one type of readable storage medium. In some embodiments, the memory 1002 may be an internal storage unit of the electronic device, such as a mobile hard disk of the electronic device. The memory 1002 may be used to store not only application software installed in the electronic device but also various data.

[0150] The communication bus 1003 is configured to implement connection and communication between the memory 1002 and at least one processor 1001, etc.

[0151] The communication interface 1004 is used for communication between the electronic device and other devices, including a network interface and a user interface.

[0152] The figure only shows an electronic device with components. Those skilled in the art will understand that the structure shown in the figure does not constitute a limitation on the electronic device 1000, and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0153] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

[0154] The railway signal system fault warning system model generation program stored in the memory 1002 in the electronic device 1000 is a combination of multiple instructions. When running in the processor 1001, the steps in the above-mentioned railway signal system fault warning method can be implemented.

[0155] Specifically, the specific implementation system of the processor 1001 for the above instructions can refer to the description of the relevant steps in the corresponding embodiment of the accompanying drawings, which will not be repeated here.

[0156] The present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor of an electronic device, it can implement the steps in the above-mentioned railway signal system fault warning method.

[0157] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0158] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any attached figure mark in the claims should not be regarded as limiting the claims involved.

[0159] In addition, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the system claim can also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.

[0160] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution 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 solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A railway signal system fault early warning method, characterized in that: The method comprises: Acquire historical operation data of a railway signal system, and perform data cleaning on the historical operation data to obtain cleaned data; Constructing a feature data set of the cleaned data within a plurality of preset historical time series intervals; Select a preset number of historical time series intervals as target intervals, and generate interval labels for each target interval based on the feature data set contained in each target interval; Screening the feature data set within the target interval according to the interval label to obtain a training sample; Obtaining a time series training model based on a bidirectional long-short term neural network, and using the training samples to train the time series training model to obtain a time series analysis model; The system operation data of the railway signal system in the target time sequence interval is obtained, the system operation data is analyzed using the time sequence analysis model, and a fault warning is issued according to the analysis result.

2. A railway signal system fault early warning method according to claim 1, characterized in that: The step of performing data cleaning on the historical operation data to obtain cleaned data includes: Filling missing values ​​of the historical operation data using a linear interpolation method to obtain preliminary optimized data; Calculating the first-order difference of each data in the preliminary optimized data; Select target data and adjacent data one by one from the preliminary optimized data, and calculate the dislocation inner product corresponding to the target data according to the first-order differences corresponding to the target data and the adjacent data; The data in the preliminary optimized data whose misaligned inner product is less than or equal to a preset threshold are collected as cleaned data.

3. A railway signal system fault early warning method according to claim 1, characterized in that: The step of constructing a feature data set of the cleaned data within a plurality of preset historical time series intervals includes: Sorting the cleaned data, and counting the five-number features in the sorted cleaned data; Calculating sample entropy and permutation entropy of the cleaned data; The five-number features, the sample entropy and the permutation entropy are aggregated into a feature data set of the cleaned data.

4. A railway signal system fault early warning method according to claim 1, characterized in that: The step of generating an interval label for each target interval according to the feature data set contained in each target interval includes: Calculate the feature values ​​corresponding to the feature data sets contained in each target interval one by one; Assigning the interval label of the target interval corresponding to the characteristic value less than or equal to the first preset threshold as a fault label; Assigning an interval label of a target interval where the characteristic value is between a first preset threshold and a second preset threshold as a degraded label; The interval label of the target interval corresponding to the characteristic value greater than the second preset threshold is allocated as a normal label.

5. A railway signal system fault early warning method according to claim 1, characterized in that: The step of screening the feature data set within the target interval according to the interval label to obtain a training sample includes: Perform equidistant sampling on the feature data set within the target interval whose interval label is a normal label to obtain a sampled data set; The feature data set and the sampling data set in the target interval whose interval labels are the fault labels and the degradation labels are collected into training samples.

6. A railway signal system fault early warning method according to claim 1, characterized in that: The step of training the time series training model using the training samples to obtain a time series analysis model includes: Obtaining sample labels of the training samples; Performing time series analysis on the training samples using the time series training model to obtain prediction labels for the training samples; Calculating the loss value of the time series training model according to the sample label and the predicted label using a regularization term; The parameters of the time series training model are optimized according to the loss value to obtain a time series analysis model.

7. A railway signal system fault early warning method according to claim 1, characterized in that: The using of the timing analysis model to analyze the system operation data and providing fault warning according to the analysis result includes: Analyzing the system operation data using the timing analysis model to obtain prediction data; Comparing the predicted data with a first preset threshold and a second preset threshold; If the predicted data is less than or equal to a first preset threshold, the predicted data is determined to be fault data; If the predicted data is between a preset first threshold and a preset second threshold, the predicted data is determined to be degraded data; When the predicted data is fault data or degradation data, a fault warning is performed.

8. A railway signal system fault warning system, characterized in that: The system comprises: A data cleaning module is used to obtain historical operation data of the railway signal system, and perform data cleaning on the historical operation data to obtain cleaned data; A feature data set construction module, used to construct a feature data set of the cleaned data within a plurality of preset historical time series intervals; A label generation module is used to select a preset number of historical time series intervals as target intervals, and generate interval labels for each target interval based on the feature data set contained in each target interval; A feature data set screening module is used to screen the feature data set in the target interval according to the interval label to obtain a training sample; A model training module is used to obtain a time series training model based on a bidirectional long-term and short-term neural network, and train the time series training model using the training samples to obtain a time series analysis model; The data analysis module is used to obtain the system operation data of the railway signal system in the target timing interval, analyze the system operation data using the timing analysis model, and issue a fault warning based on the analysis result.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute a railway signal system fault warning method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, a railway signal system fault warning method as described in any one of claims 1 to 7 is implemented.

Images