Interlocking subsystem real-time fault prediction method and system based on gray scale prediction

Abstract

The invention discloses an interlocking subsystem real-time fault prediction method and system based on gray scale prediction, and relates to the technical field of rail transit, and the method comprises the steps: mapping real-time operation data into target analysis data of each interlocking device, carrying out the hierarchical division of the target analysis data, and carrying out the hierarchical division of the target analysis data; the health degree evaluation score of the interlocking subsystem is calculated based on the hierarchical data set in combination with historical data, and the complex operation state of the interlocking subsystem is quantified; according to the health degree evaluation score and the equipment health state of each level, calculating to obtain a real-time health degree score of the interlocking subsystem, constructing a gray scale prediction sequence, taking the gray scale prediction sequence as an input of a gray scale model, outputting a gray scale sequence value, and improving pertinence and accuracy of prediction; and the health degree prediction value of the interlocking subsystem at the next moment is calculated based on the gray sequence value, so that the fault prediction efficiency of the interlocking subsystem is remarkably improved, the fault processing efficiency of the interlocking subsystem is further improved, and safe, stable and reliable operation of the interlocking subsystem is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of rail transit, and particularly to a real-time fault prediction method and system for an interlocking subsystem based on grey prediction. Background Art

[0002] With the rapid development of the railway industry, the interlocking system, as one of the crucial systems in the field of rail transit, is an important link to ensure the safe operation of railway trains. Due to the complex structure and harsh working environment of the interlocking system, and the fact that the interlocking system operates day and night without interruption every day, it is inevitable for failures to occur. Therefore, the interlocking system is a high-frequency area for failures. At present, the process of fault location in the interlocking system is complex. Relying on on-site maintenance personnel to infer the fault point based on the phenomena and combined with experience, it is highly dependent on the professional qualities and work experience of the staff. There are large errors in human factors during the troubleshooting process. With the increase in train operation speed and traffic density, relying solely on the maintenance of maintenance personnel is extremely inefficient, and the invested human and time costs are also relatively large; after a failure occurs, the method of manually searching for the fault point requires high requirements for maintenance personnel, takes a certain amount of time to troubleshoot, and may not be able to completely solve the fault. If the fault cannot be excluded in a timely and accurate manner, it will endanger the safety of train operation, affect the operation of rail transit, and cause relatively large losses.

[0003] The patent "An Intelligent Operation and Maintenance Method and System for a Railway Station Interlocking System", publication number: CN114625837A, publication date: June 14, 2022, discloses: constructing a fault operation and maintenance knowledge graph for a railway station interlocking system, analyzing the knowledge related to the faults of the interlocking system, summarizing the schema layer of the knowledge graph, and constructing the data layer in a top-down manner; making a decision on the operation and maintenance plan for the optical cable transmission system based on the knowledge graph, mapping the fault description into the fault knowledge graph to obtain a fault subgraph, further searching in the knowledge graph, matching the fault description with the existing knowledge in the knowledge graph, and outputting the operation and maintenance plan; based on the on-site fault data, comparing the fault phenomenon description with the existing fault data, and using Bayes' formula to obtain the most probable fault and output it; the system includes: a fault data preprocessing module, a knowledge graph storage module, and a result output and display module. However, this invention does not fully consider the dynamic change of the fault probability. When using Bayes' formula to calculate the probability, it is necessary to assume that the probability relationship between the fault phenomenon, fault cause, and fault location conforms to a certain prior distribution. It is difficult to ensure the accuracy of the calculation results and the applicability is poor in the face of various complex fault factors in the actual interlocking system. Summary of the Invention

[0004] The present invention addresses the problem that the fault detection of the interlocking subsystem overly relies on manual troubleshooting, resulting in low efficiency in system fault handling. A real-time fault prediction method and system for the interlocking subsystem based on grey prediction are proposed. By mapping real-time operation data into target analysis data for each interlocking device, hierarchical partitioning is performed on the target analysis data. Based on the hierarchical data set combined with historical data, the health evaluation score of the interlocking subsystem is calculated, quantifying the complex operation state of the interlocking subsystem. According to the health evaluation score and the device health states at each level, the real-time health score of the interlocking subsystem is calculated, and a grey prediction sequence is constructed and used as the input of the grey model to output grey sequence values, fully considering the importance differences of devices at different levels in the entire interlocking subsystem, improving the pertinence and accuracy of prediction. Based on the grey sequence values, the health prediction value of the interlocking subsystem at the next moment is calculated, significantly improving the fault prediction efficiency of the interlocking subsystem, and further improving the fault handling efficiency of the interlocking subsystem to ensure the safe, stable, and reliable operation of the interlocking subsystem.

[0005] To solve the above technical problems, according to the first aspect provided by the embodiments of the present invention, a real-time fault prediction method for an interlocking subsystem based on grey prediction is provided, including the following steps: S1. Obtain the real-time operation data of the interlocking subsystem and map the real-time operation data into target analysis data for each interlocking device; S2. Based on the interlocking architecture and combined with device attributes, perform hierarchical partitioning on the target analysis data to obtain a hierarchical data set; S3. Based on the hierarchical data set combined with historical data, calculate the health evaluation score of the interlocking subsystem; S4. Calculate the real-time health score of the interlocking subsystem according to the health evaluation score and the device health states at each level; S5. Based on the real-time health score and the historical health score, construct a grey prediction sequence and use it as the input of the grey model to output grey sequence values; S6. Based on the grey sequence values, calculate the health prediction value of the interlocking subsystem at the next moment.

[0006] In this solution, by mapping the acquired real-time operation data into the target analysis data of each interlocking device, key information directly related to the device status can be accurately extracted, irrelevant data interference can be eliminated, and an accurate and targeted data basis can be provided for subsequent analysis; by hierarchically organizing the data, the data can be closely linked to the physical structure of the interlocking subsystem. By comprehensively analyzing the characteristics of these different levels, the patterns and rules of fault occurrence can be understood more comprehensively; by calculating the health evaluation score of the interlocking subsystem, the complex operation status of the interlocking subsystem can be quantified, so as to intuitively understand the current operation status of the interlocking subsystem; by calculating the real-time health score, the importance differences of different-level devices in the entire interlocking subsystem are fully considered, making the health score more in line with the impact of each part on the whole in the actual operation of the interlocking subsystem. This helps to focus on the health change trends of key devices during fault prediction, improving the pertinence and accuracy of prediction; by constructing a gray prediction sequence and inputting it into the gray model to output the gray sequence value, leveraging the advantages of the gray prediction method, the potential change rules in the data can be mined, and the future health status of the system can be predicted prospectively, significantly improving the fault prediction efficiency of the interlocking subsystem; by calculating the predicted health value of the interlocking subsystem at the next moment based on the gray sequence value, the change trend of the system health can be accurately grasped, enabling the operation and maintenance personnel to clearly understand the future short-term health trend of the system. Thus, a maintenance plan can be formulated more scientifically, resources can be allocated reasonably, the probability of faults can be effectively reduced, and the stable and reliable operation of the interlocking subsystem can be ensured.

[0007] Preferably, the S1 includes the following sub-steps: Collect the real-time operation data of the interlocking subsystem based on the communication protocol of the interlocking subsystem, and encapsulate the real-time operation data into a number of data packets; Analyze the data packets, extract the key data associated with each interlocking device, and use the key data as the target analysis data of the interlocking device.

[0008] Preferably, the S2 includes the following sub-steps: According to the interlocking architecture, perform the first-level division on the target analysis data to obtain a number of first data sets corresponding to the interlocking subsystems; According to the device types within each interlocking subsystem, perform the second-level division on the first data set to obtain a second data set; according to the data types, perform the third-level division on the second data set to obtain a third data set; Establish a data mapping table, associate and store the data in each data set based on the data key-value pairs, and complete the hierarchical division of the target analysis data.

[0009] In the solution, through three - layer data partitioning, a hierarchical data architecture is constructed, which can closely link the data with the physical structure of the interlocking subsystem. Different - level data partitioning helps to extract fault - related features. Each level focuses on data features from different perspectives. The first level focuses on the macroscopic features at the module level, the second level focuses on the differential features of equipment types, and the third level focuses on the detailed features of specific operating states. By comprehensively analyzing these features at different levels, the mode and law of fault occurrence can be understood more comprehensively. When a fault occurs, the operation and maintenance personnel can gradually narrow down the troubleshooting scope according to the modules, equipment types, and operating states that the fault may involve.

[0010] Preferably, S3 includes the following sub - steps: Initialize the first health - degree evaluation scores of each interlocking device, and determine the minimum failure rate of the data layer within the historical operation period according to the historical data set; Determine the dynamic parameters of the health - degree evaluation model according to the initialized first health - degree evaluation scores and the minimum failure rate; Generate the health - degree calculation coefficients of the interlocking devices based on the importance degrees of the partitioning elements of each data layer, and calculate the first health - degree evaluation scores of the devices at each level according to the health - degree calculation coefficients and the corresponding data sets; Calculate the current failure rate of the data layer according to the first health - degree evaluation scores, and update the dynamic parameters by combining the health - degree scores and the current failure rate; Calculate the health - degree evaluation scores based on the updated dynamic parameters.

[0011] In this solution, by querying the fault alarm and early - warning information in the historical operation period to calculate the minimum failure rate, the historical fault conditions of the interlocking subsystem are fully considered, so that the calculation of the health - degree score is not only based on the current state, but also can comprehensively consider past operation experience; by initializing the first health - degree evaluation scores of the interlocking devices at each level and substituting them into the model together with the minimum failure rate to determine the dynamic parameters, a basis is provided for subsequent health - degree calculation. This basic parameter can reflect the health tendency of the interlocking devices in the initial state and also consider the possibility of historical faults; by calculating the health - degree evaluation scores, the complex operating state of the interlocking subsystem can be quantified, and the health degree of the system is represented by a specific value, enabling the operation and maintenance personnel to intuitively understand the current operating condition of the system.

[0012] Preferably, the generating the health - degree calculation coefficients of the interlocking devices based on the importance degrees of the partitioning elements of each data layer, and calculating the first health - degree evaluation scores of the devices at each level according to the health - degree calculation coefficients and the corresponding data sets includes: based on the importance degrees of the partitioning elements of each data layer, marking the importance degrees of each partitioning element through gradient marking; Establish an importance judgment matrix according to the importance level identifier; Extract the maximum eigenvector from the importance judgment matrix, and calculate the health degree calculation coefficient of the corresponding interlocking device through the maximum eigenvector; Calculate the first health degree evaluation score of each layer of devices according to the health degree calculation coefficient of the interlocking device and the corresponding data set.

[0013] Preferably, the S4 includes the following sub-steps: Based on the health status of the devices corresponding to each data layer, identify the health degree of each interlocking device according to the health status through gradient labeling; Establish a health degree judgment matrix according to the health degree identifier; Extract the maximum eigenvector from the health degree judgment matrix, and calculate the health status weight coefficient of the corresponding interlocking device through the maximum eigenvector; Calculate the real-time health degree score of each layer of devices in the interlocking subsystem based on the health degree evaluation score and the health status weight coefficient.

[0014] Preferably, the formula for calculating the real-time health degree score is as follows: In the formula, Y t is the real-time health degree score at time t, k represents the total number of layers, ω i represents the health status weight coefficient of the i-th layer, represents the health degree evaluation score of the i-th layer at time t.

[0015] In this solution, the final health degree score is calculated through the health status weight coefficients of the interlocking devices at each layer, taking into account the importance differences of devices at different layers in the entire interlocking subsystem. Based on the health degree weight allocation method, the health degree score can more accurately reflect the overall health status of the system.

[0016] Preferably, the S5 includes the following sub-steps: Generate a first gray prediction sequence based on the historical health degree scores of the interlocking devices at each layer, accumulate the first gray prediction sequences of the interlocking devices at the same layer, and construct a first gray value queue; Establish a gray model according to the fusion of the first gray value queue and the first gray prediction sequence, and determine the gray parameter of the gray model based on the historical health degree score; Generate a second gray prediction sequence based on the real-time health degree scores of the interlocking devices at each layer, accumulate the second gray prediction sequences of the interlocking devices at the same layer, and construct a second gray value queue; Based on the grayscale parameter, substitute the second grayscale value queue into the grayscale gradient equation of the grayscale model to obtain grayscale sequence values.

[0017] Preferably, S6 includes the following sub-steps: Based on the subtraction of the grayscale sequence value at the current level from the grayscale sequence value at the previous level, obtain the health prediction value of the interlocking subsystem at the next moment.

[0018] In this solution, by constructing a grayscale prediction sequence according to the real-time health score and the historical health score, the health status of the interlocking subsystem can be dynamically evaluated, problems that may occur in the system can be pre-warned in advance, and it helps to improve the reliability and availability of the interlocking system. Among them, by using the grayscale prediction method to calculate the health prediction value at the next moment by combining the grayscale sequence value and the historical grayscale prediction value, the prediction model can be adjusted according to the actual operation data, so that the grayscale model can continuously optimize the prediction result, achieve a more accurate prediction of the health change trend of the interlocking subsystem, and thus better ensure the stable operation of the interlocking subsystem.

[0019] According to another aspect provided by the embodiments of the present invention, a real-time fault prediction system for an interlocking subsystem based on grayscale prediction is provided, including: A data acquisition module, configured to acquire target analysis data of interlocking devices; A data processing module, configured to perform hierarchical division on the target analysis data to obtain a hierarchical data set; A data analysis module, configured to calculate the health score of the interlocking subsystem according to the hierarchical data set; A data prediction module, configured to calculate the health score to obtain the health prediction value of the interlocking subsystem at the next moment.

[0020] The beneficial effects of the present invention: 1. By constructing a grayscale prediction sequence according to the real-time health score and the historical health score, the health status of the interlocking system can be dynamically evaluated. It no longer simply judges whether a fault occurs, but can pre-warn problems that may occur in the system in advance, improving the reliability and availability of the interlocking system; 2. By calculating the final health score through the health state weight coefficients of interlocking devices at each level, the importance differences of different-level devices in the entire interlocking subsystem are considered. Based on the health degree weight distribution method, the health score can more accurately reflect the overall health status of the system. At the same time, through reasonable weight distribution, the role of these key parts can be highlighted when calculating the health score, thereby avoiding the potential fault risks of key devices being masked by small problems of some non-key devices, and thus optimizing the reliability assessment of the fault prediction of the entire interlocking subsystem; 3. By constructing a grayscale prediction sequence based on real-time health scores and historical health scores, the health status of the interlocking subsystem can be dynamically evaluated, and possible problems in the system can be warned in advance, which helps to improve the reliability and availability of the interlocking system. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Other features, objects and advantages of the present invention will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings. The drawings are only for the purpose of illustrating preferred embodiments and are not to be considered as limiting the present invention. Also, the same reference symbols are used throughout the drawings to represent the same parts.

[0022] Figure 1 The present invention is a flowchart of a method for real-time fault prediction of an interlocking subsystem based on grayscale prediction according to an embodiment of the present invention.

[0023] Figure 2 A schematic diagram of a hierarchical framework of interlocking subsystem data according to an embodiment of the present invention.

[0024] Figure 3 The figure is a flow chart of calculating a health evaluation score according to an embodiment of the present invention.

[0025] Figure 4 The present invention is a schematic diagram of a real-time fault prediction system module for an interlocking subsystem based on grayscale prediction according to an embodiment of the present invention.

[0026] Figure 5 The present invention is a schematic diagram of an interlocking fault prediction analysis process according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific implementation method described herein is only an optimal embodiment of the present invention, which is only used to explain the present invention and does not limit the scope of protection of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0028] Example 1: Figure 1 As shown, a real-time fault prediction method for an interlocking subsystem based on grayscale prediction includes steps S1-S6, wherein: S1. Acquire the real-time operation data of the interlocking subsystem and map the real-time operation data into target analysis data of each interlocking device.

[0029] Specifically, S1 includes the following sub-steps: Collecting real-time operation data of the interlocking subsystem based on the communication protocol of the interlocking subsystem, and encapsulating the real-time operation data into a plurality of data packets; Parse the data packet, extract the key data associated with each interlocking device, and use the key data as the target analysis data of the interlocking device.

[0030] Specifically, the real-time operation data at least includes network communication status data, the operation status of the interlocking subsystem, IO acquisition data, alarm data, operation data; log messages received by the interlocking subsystem, etc. Establish a communication connection with the interlocking subsystem through the communication protocol of the interlocking subsystem, obtain various operation data of the interlocking subsystem based on the communication relationship, and map the collected data into analyzable data in the corresponding format of each interlocking device through the configuration file, that is, the target analysis data.

[0031] Among them, when communicating with the interlocking subsystem for the first time, all the operation data of the interlocking subsystem is collected. Subsequently, only the data whose status has changed is collected. The collection of non-first communication data can be achieved by configuring a data status monitoring mechanism in the configuration file. For example, define a data status record table to record the previous status of each data. When communicating, compare the currently obtained data status with the recorded status, and only the data whose status has changed is collected.

[0032] Furthermore, the configuration file at least includes data mapping rules, which include establishing a target analysis data table for establishing the correspondence between the operation data of the interlocking subsystem and the relevant parameters of each interlocking device in the target score data format, so as to map the collected raw data into the target analysis data in the corresponding format of each interlocking device. For example, a certain byte in the raw data represents the temperature of the interlocking slave computer, and in the data mapping table of the configuration file, the position of this byte will be associated with the field representing the temperature of the interlocking slave computer in the target analysis data. The configuration file also includes defining relevant protocol parameters according to the communication protocol of the interlocking subsystem, such as protocol type, port number, data transmission format, etc. These parameters are used to establish a communication connection with the interlocking subsystem to ensure that the operation data can be correctly obtained.

[0033] In this embodiment, the communication protocol stipulates important information such as the data transmission format, transmission rate, and transmission content. Data collection is carried out through the communication protocol to avoid data loss or incorrect collection. For example, some devices in the interlocking subsystem may send status information at specific time intervals and data formats. According to the communication protocol, these information can be accurately obtained, providing a reliable data basis for subsequent analysis. The real-time operation data is encapsulated into several data packets to facilitate the storage, transmission, and processing of data in the system. In addition, the encapsulation of data packets is also beneficial to ensuring the integrity of data in a complex network environment or storage system, preventing data from being damaged or lost during transmission.

[0034] S2. Hierarchically divide the target analysis data based on the interlocking architecture in combination with device attributes to obtain a hierarchical data set.

[0035] Specifically, S2 includes the following sub-steps: Perform a first-level division on the target analysis data according to the interlocking architecture to obtain a first data set corresponding to several interlocking subsystems; Perform a second-level division on the first data set according to the device types within each interlocking subsystem to obtain a second data set; perform a third-level division on the second data set according to the data types to obtain a third data set; Establish a data mapping table, and associate and store the data in each data set based on data key-value pairs to complete the hierarchical division of the target analysis data.

[0036] Furthermore, as Figure 2 shown, the first data set can be divided according to the organizational structure of the interlocking subsystem. For example, take the board card module, drive and acquisition loop, CBI-A / B machine, communication module, and auxiliary module in the interlocking subsystem as the first data set division benchmark, and classify the collected data according to these modules once to obtain the first-level data set, that is, the first data set; then divide the first data set according to the device types in each module. For example, cpsu boards, com boards, VIB boards, vob boards, etc. in the board card module, and classify the first data set according to these board card types a second time to obtain the second data set at the second level; finally, classify the second data set a third time according to the specific operating status of each device through the corresponding data types, such as alarm status data, operation attribute data, analog quantity, etc., to obtain the third data set at the third level.

[0037] As an implementation manner, as Figure 2 shown, hierarchically process the operation data of the board card module under the interlocking subsystem. First, extract and aggregate the status data, alarm data, etc. of cpsu in different interface protocols of the subsystem to form the three-level device cpsu board data; similarly, aggregate the relevant data types such as alarm and status of other boards such as VIB and VOB into other three-level boards; then form the second-level board card module from these three-level board card data sets; finally, form the first-level interlocking subsystem data set from the second-level board card module, drive and acquisition loop module, etc. In this embodiment, the interlocking structure is different from the actual interlocking architecture. In this embodiment, the hierarchy tends to be a combination of the device hierarchy and the function hierarchy. For example, according to the hierarchical levels from the interlocking lower computer to the device board card to the drive and acquisition IO, as Figure 2As shown, the board card is just the board card, excluding the driving and acquisition components on the board card. Instead, the driving and acquisition components and the board card are regarded as the same level. The purpose is that after calculating the fault prediction value, the interlocking device can be quickly located according to the input data of the prediction model, the fault point range can be restricted, the response and handling efficiency of the fault can be improved, and it helps to comprehensively and clearly view the status of specific devices through the target analysis data mapped above, so that the operation and maintenance personnel can make corresponding operation and maintenance measures in time to further ensure the safe and stable operation of the interlocking subsystem.

[0038] In this embodiment, the data is hierarchically organized by referring to the interlocking architecture and combining functions. After classifying the subsystem data according to the device type, data sets are classified by combining different data types. When a fault occurs, the operation and maintenance personnel can gradually narrow down the troubleshooting scope according to the modules, device types, and operating states that the fault may involve. For example, if the fault manifests as communication anomalies, clues can first be found in the data related to the communication module in the first data set, then the specific communication device board card can be determined in the second data set, and finally the data types related to the operating state of the communication device (such as alarm data) can be viewed through the third data set, so as to accurately locate the cause of the fault.

[0039] At the same time, the data division at different levels helps to extract features related to faults. Each level focuses on data features from different perspectives. The first level focuses on the macroscopic features at the module level, the second level focuses on the differential features of device types, and the third level focuses on the detailed features of specific operating states. By comprehensively analyzing these features at different levels, the patterns and rules of fault occurrence can be understood more comprehensively. For example, for the fault analysis of the board card module, it can first be determined from the first level whether there are abnormalities in the overall board card module, then it can be analyzed from the second level which specific board card has problems, and finally the data changes under the specific operating state of the board card can be viewed from the third level, so as to provide rich feature information for fault prediction and diagnosis. In addition, the hierarchical division helps with the update, maintenance, and storage of data; for newly added data, it can be added according to the established hierarchical and category rules to ensure data consistency and integrity. At the same time, in terms of data storage, a reasonable hierarchical division can optimize the storage structure, improve storage efficiency, and reduce data redundancy.

[0040] S3. Calculate the health evaluation score of the interlocking subsystem based on the hierarchical data set and historical data.

[0041] Specifically, S3 includes the following sub-steps: Initialize the first health evaluation score of each interlocking device, and determine the minimum failure rate of the data layer during the historical operation period according to the historical data set; Determine the dynamic parameters of the health evaluation model according to the initialized first health evaluation score and the minimum failure rate; Generate a health calculation coefficient for the interlocking device based on the importance degree of the division elements of each data layer, and calculate the first health evaluation score of each level of equipment according to the health calculation coefficient and the corresponding data set; Calculate the current failure rate of the data layer according to the first health evaluation score, and update the dynamic parameter by combining the health score and the current failure rate; Calculate the health evaluation score based on the updated dynamic parameter.

[0042] Specifically, the step of generating a health calculation coefficient for the interlocking device based on the importance degree of the division elements of each data layer, and calculating the first health evaluation score of each level of equipment according to the health calculation coefficient and the corresponding data set includes: Based on the importance degree of the division elements of each data layer, identify the importance degree of each division element through gradient annotation; Establish an importance judgment matrix according to the importance degree identification; Extract the maximum eigenvector in the importance judgment matrix, and calculate the health calculation coefficient of the corresponding interlocking device through the maximum eigenvector; Calculate the first health evaluation score of each level of equipment according to the health calculation coefficient of the interlocking device and the corresponding data set.

[0043] S4. Calculate the real-time health score of the interlocking subsystem according to the health evaluation score and the health status of each level of equipment.

[0044] Specifically, the S4 includes the following sub-steps: Based on the health status of the equipment corresponding to each data layer, identify the health degree of each interlocking device according to the health status through gradient annotation; Establish a health judgment matrix according to the health degree identification; Extract the maximum eigenvector in the health judgment matrix, and calculate the health status weight coefficient of the corresponding interlocking device through the maximum eigenvector; Calculate the real-time health score of each level of equipment in the interlocking subsystem based on the health evaluation score and the health status weight coefficient.

[0045] Specifically, the calculation formula of the first health evaluation score is as follows: In the formula, represents the first health evaluation score of each level of equipment, k is the number of levels, k = 1, 2, 3; α k,j is the health calculation coefficient of the jth device at the kth level; represents the data set of the jth device in the kth level.

[0046] Further, the health degree calculation coefficient configuration method can be to compare each pair of elements at the same level. The elements can be the functions of the devices, the types of the devices, etc. According to the importance degree, the comparison results are marked by the 1-9 gradient marking method to form a judgment matrix, that is, the importance judgment matrix. The meanings of the gradient markings are shown in Table 1. Table 1 Meanings of Health Degree Gradient Markings After determining the importance judgment matrix information of each level of devices through gradient marking, such as the importance judgment matrix data of each device in the board card module shown in Table 2. Table 2 Importance Judgment Matrix Data of Board Card Module Board card module COM communication board CPSU board Core processor board VIB board VOB board COM communication board 1 1 3 2 2 CPSU board 1 1 3 2 2 Core processor board 1 / 3 1 / 3 1 1 / 2 1 / 2 VIB board 1 / 2 1 / 2 2 1 1 VOB board 1 / 2 1 / 2 2 1 1 Extract the maximum eigenvector from the above matrix data, normalize it, and obtain the corresponding health degree calculation coefficients of each device in the board card module, as shown in Table 3. Table 3 Health Degree Calculation Coefficients of Board Card Module COM communication board CPSU board Core processor board VIB board VOB board Calculation parameter 0.1087 0.1087 0.3587 0.212 0.2119 In this embodiment, as the interlocking subsystem operates and the devices are updated, the importance among the devices may change. The method for determining the health degree calculation coefficients in this application can re-perform pairwise comparison and calculation according to the actual situation, dynamically adjust the coefficients, so as to adapt to the changes in the interlocking subsystem, ensure that the calculation of the health degree score can always accurately reflect the actual health status of each interlocking device, and further provide a more reliable basis for fault prediction.

[0047] Further, the calculation formula of the health degree evaluation model is as follows: In the formula, represents the health degree score of the i-th layer of the interlocking subsystem at time t, i ∈ k, that is, the layer when dividing the target analysis data; κ and ε are the dynamic parameters of the model, representing the health state of the interlocking subsystem. The higher the health degree score of the devices in the interlocking subsystem, the smaller the dynamic parameters; represents the failure rate of the i-th layer of the interlocking subsystem at time t.

[0048] In this embodiment, the historical operation period can be dynamically set, and the first health degree evaluation scores of each layer of interlocking devices are initialized, which can also represent the health deduction scores of each layer of interlocking devices; query the historical data of each layer of related devices according to a certain historical operation period, and calculate a minimum failure rate based on the historical fault alarm data or early warning information data, etc. during this period; substitute the initialized health degree score and the minimum failure rate into the health degree evaluation model, that is, formula (2), to obtain the initialized dynamic parameters of the model.

[0049] Further, according to the data set collected and partitioned at time t and the corresponding health calculation coefficient, substitute them into formula (1) to calculate the real-time first health evaluation score, and then divide it by the health evaluation threshold to obtain the failure rate; substitute the real-time first health evaluation score, the failure rate, and the model dynamic parameters at time t-1 into formula (2) to calculate the real-time health evaluation score of the interlocking device.

[0050] As an implementation manner, as Figure 3 shown, when initializing the model dynamic parameters, the minimum failure rate within 180 days 1 / 180*240 and the first health evaluation score 100 are used, and the common failure rate within 6 months (the failure rate actually occurred within 6 months, with a score of 80) are substituted into formula (2) to obtain the corresponding dynamic parameters; after receiving the real-time data and hierarchically partitioning, calculate the health deduction value of the interlocking device according to formula (1), obtain the health score based on formula (2), that is, the first health evaluation score, and then substitute the score value and the failure rate into formula (2) to update the dynamic parameters of formula (2). Among them, the common failure rate within 6 months is selected, that is, the failure rate in the 6 months before the current moment, and the first health evaluation score of each current level is calculated through formula (1); when running for the first time, if it is less than 6 months, it is the current failure rate, and the corresponding first health evaluation score at this time is 80. The time length t can be dynamically configured. In this embodiment, the empirical value of 5 minutes is adopted. It should be noted that the longer the time length, the worse the calculation accuracy, and it cannot effectively reflect the state of the interlocking device.

[0051] In some alternative implementation manners, not only can the health evaluation score of the entire interlocking subsystem be calculated, but also the health evaluation scores can be calculated separately for each level of equipment, so as to grasp the system state from both the overall and local perspectives. For the overall system, the score can reflect whether the system is running stably; for each level of equipment, such as board card modules, drive and acquisition circuits, etc., the score can help locate the specific parts that may have problems. For example, if the health evaluation score of the overall system is acceptable, but the score of a certain level of equipment (such as the communication module) is low, the operation of this module can be focused on.

[0052] In this embodiment, the minimum failure rate is calculated by querying the fault alarms and early warning information of the historical operation cycle, fully considering the historical fault conditions of the interlock subsystem, enabling the fault prediction model to learn the past fault patterns and frequencies. When predicting future faults, the probability of faults occurring in similar historical situations can be referred to, thereby improving the accuracy of prediction. By initializing the first health evaluation scores of interlock devices at each level and substituting them into the model together with the minimum failure rate to determine the dynamic parameters, a basis is provided for subsequent health calculation. This basic parameter can reflect the health tendency of interlock devices in the initial state and also consider the possibility of historical faults. By calculating the real-time health evaluation scores of interlock devices based on the real-time first health evaluation scores, failure rates, and model dynamic parameters, the state changes of the interlock subsystem at different times can be captured, adapting to the dynamic characteristics of the system and improving the adaptability of the fault prediction system. Among them, by calculating the health evaluation scores, the operation and maintenance personnel can intuitively understand the current operation status of the system. For example, the health evaluation scores can be set between 0 and 100. A score above 80 indicates that the system is in good condition, 40 - 80 is in a sub-healthy state, and below 40 may mean that there are serious fault risks in the system. This quantitative method avoids judging the system state solely based on experience or vague feelings, making the evaluation results more objective and accurate.

[0053] Specifically, the calculation formula for the real-time health score is as follows: In the formula, Y t is the real-time health score at time t, k represents the total number of levels, ω i represents the health state weight coefficient of the i-th level, represents the health evaluation score of the i-th level at time t.

[0054] In this embodiment, the configuration of the health state weight coefficient can use the configuration method of the above-mentioned health calculation coefficient. By comparing elements pairwise at the same level, according to the importance degree, they are compared and marked by the 1 - 9 gradient marking method, and then a health judgment matrix is established based on the health degree marking. For example, according to the execution time of each module in the second level of the interlock subsystem, the control relationship between modules, etc., the importance degree is marked for the maximum authority or the control and being controlled relationship, as shown in Table 4. Table 4 Judgment matrix information of the second-level modules Interlock health status Board card module Drive and acquisition loop CBI-A / B machine Communication module Auxiliary module Board card module 1 2 2 1 1 / 2 Drive and acquisition loop 1 / 2 1 1 1 1 / 3 CBI-A / B machine 1 / 2 1 1 1 1 / 3 Communication module 1 1 1 1 1 / 2 Auxiliary module 3 3 3 2 1 Taking the maximum eigenvector of the above matrix and normalizing it is the health state weight coefficient, as shown in Table 5. Table 5 Health state weight coefficient Board card module Drive and acquisition loop CBI-A / B machine Communication module Auxiliary module Weight 0.1956 0.2609 0.2609 0.1956 0.087

[0055] In this embodiment, the final health score is calculated through the health state weight coefficients of the interlock devices at each level, taking into account the importance differences of devices at different levels in the entire interlock subsystem. Based on the health weight allocation method, the health score can more accurately reflect the overall health status of the system. At the same time, through reasonable weight allocation, the role of these key parts can be highlighted when calculating the health score, thereby avoiding the potential failure risks of key devices being masked by minor problems of some non-key devices, and thus optimizing the reliability assessment of the fault prediction for the entire interlock subsystem.

[0056] S5. Construct a gray prediction sequence based on the real-time health score and the historical health score, and use it as the input of the gray model to output the gray sequence value.

[0057] Specifically, the S5 includes the following sub-steps: Generate a first gray prediction sequence based on the historical health scores of the interlock devices at each level, accumulate the first gray prediction sequences of the interlock devices at the same level, and construct a first gray value queue; Establish a gray model by fusing the first gray value queue and the first gray prediction sequence, and determine the gray parameter of the gray model based on the historical health score; Generate a second gray prediction sequence based on the real-time health scores of the interlock devices at each level, accumulate the second gray prediction sequences of the interlock devices at the same level, and construct a second gray value queue; Based on the gray parameter, substitute the second gray value queue into the gray gradient equation of the gray model to obtain the gray sequence value.

[0058] S6. Calculate the health prediction value of the interlock subsystem at the next moment based on the gray sequence value.

[0059] Specifically, the S6 includes the following sub-steps: Subtract the gray sequence value of the current level from the gray sequence value of the previous level to obtain the health prediction value of the interlock subsystem at the next moment.

[0060] Specifically, the first gray prediction sequence first obtains the historical health score from the historical data of the interlock subsystem devices, constructs the first gray value according to the historical health scores of the interlock devices at the same level, and is denoted as Y (0) , Y (0) =(Y (0) (1), Y (0) (2), Y (0) (3), …, Y 0 (n))(4) A sequence representing the historical health scores of n elements, i.e., n interlocking devices at the same level; among them, the historical health score can also be expressed as the historical health prediction value of the interlocking subsystem. Therefore, the first gray prediction sequence can represent the original prediction sequence of the interlocking subsystem; Accumulate the first gray values Y at each level (0) to obtain the first gray value queue of the interlocking subsystem, denoted as Z (1) , Z (1) =(Z (1) (2), Z (1) (3), …, Z (1) (n))(5) According to the first gray values Y at each level (0) and the first gray value queue Z (1) fuse to construct a gray model, and the calculation formula is as follows: Y (0) (k)+aZ (1) (k)=b(7) where a and b are the gray parameters of the gray model respectively. The gray parameters can be calculated according to the historical health scores. Specifically: construct the matrix of the first gray value Y (0) and the matrix of the first gray value queue Z (1) : Then the values of a and b can be expressed as: (a, b) T =(B T B) -1 B T Y(9) Substitute the historical health scores into formula (9) to obtain the values of a and b. Among them, the length of the first gray value Y (0) is at least greater than or equal to 2.

[0061] Furthermore, construct a second gray prediction sequence according to the real-time health scores, denoted as Y (1) . Substitute the second gray prediction sequences Y (1) (k) at different levels into the gradient equation corresponding to the gray model, as follows: Furthermore, calculate the gray sequence value according to formula (10): In the formula, e is the natural constant, and Y (1) (k + 1) represents the gray sequence value of the interlocking device at the current level; The health score at the next moment is obtained by subtracting the gray - scale sequence value at the current moment from the previous gray - scale sequence value. The calculation formula is as follows: Y (1) =Y (1) (k + 1)-Y (1) (k); In the formula, Y (1) represents the predicted value of the health of a single interlocking device, and Y (1) (k) represents the gray - scale sequence value of the interlocking device at the previous level of the current level.

[0062] It can be understood that the construction of gray - scale values is mainly for predicting the health score at the next moment. Gray - scale prediction mainly generates and processes the original data to find the law of system changes, generates a data sequence with strong regularity, and then establishes a corresponding differential equation model to predict the future development trend of things. It constructs a gray - scale prediction model with a series of numerical values reflecting the characteristics of the prediction object observed at equal time intervals to predict the characteristic quantity at a certain future moment or the time to reach a certain characteristic quantity. Since the operation data of the interlocking subsystem often has uncertainty and incompleteness, therefore, using the gray - scale prediction method, which has relatively flexible requirements for data, to conduct fault prediction on the interlocking subsystem can well fit the change trend of the data. Based on the internal dynamic change law of the system, by generating and accumulating the original data, the potential law of the data can be mined. For example, the state change of signal equipment may be affected by factors such as equipment aging and small environmental fluctuations in the short term. Gray - scale prediction can, according to the recent operation data (such as signal transmission delay time, signal strength change, etc.), accurately predict whether the equipment is likely to fail in the short term, which helps the operation and maintenance personnel take measures in advance, such as arranging equipment inspections or repairs, thereby reducing the probability of failure.

[0063] In addition, the gray - scale prediction method can, to a certain extent, filter out various interference factors existing in the actual operation environment of the interlocking subsystem, such as electromagnetic interference, mechanical vibration, etc. Gray - scale prediction mainly focuses on the overall change trend of the data rather than the precise fitting of each data point. For example, when electromagnetic interference causes some communication parameters of the communication equipment in the interlocking subsystem to have instantaneous abnormal fluctuations, the gray - scale prediction model will not produce large deviations due to these local abnormal data and can still make a relatively reasonable fault prediction based on the overall trend of the data.

[0064] In this embodiment, by constructing a gray prediction sequence based on the real-time health score and the historical health score, the health status of the interlocking subsystem can be dynamically evaluated, problems that may occur in the system can be warned in advance, and it helps to improve the reliability and availability of the interlocking system. Among them, by using the gray prediction method, the health prediction value at the next moment can be calculated by combining the gray sequence value and the historical gray prediction value, and the prediction model can be adjusted according to the actual operation data, so that the gray model can continuously optimize the prediction result, realize more accurate prediction of the health change trend of the interlocking subsystem, and thus better ensure the stable operation of the interlocking subsystem.

[0065] Embodiment 2, as Figure 4 shown, the fault prediction system of the interlocking subsystem based on gray prediction includes: A data acquisition module, used to acquire the target analysis data of the interlocking device; A data processing module, used to perform hierarchical division on the target analysis data to obtain a hierarchical data set; A data analysis module, used to calculate the health score of the interlocking subsystem according to the hierarchical data set; A data prediction module, used to calculate the health score to obtain the health prediction value of the interlocking subsystem at the next moment.

[0066] In some examples, as Figure 5 shown, the first health evaluation score of each level corresponding to each module in the interlocking subsystem is the mean value of the first health evaluation scores of each device, that is the mean value of. For example, when a certain cpsu board in the interlocking system fails, the specific fault data processing process is as follows: The subsystem collects the fault information, encapsulates the fault information collected according to the communication protocol and then sends it to the fault prediction system of the interlocking subsystem in this application;

[0067] The fault prediction system receives the message data of the interlocking subsystem through the data acquisition module, unpacks the packet, parses and assigns values according to the above configuration file and aggregates them to the alarm and status of each cpsu board at the board level, and assigns the status and alarm to the fault and alarm; Based on the data processing module, combined with the information of each cpsu board of the board type in the historical data table, comprehensively give the fault and alarm of the cpsu board, the corresponding status deduction value, the alarm deduction value; at the same time, count the failure rate of other boards of the cpsu board, the evaluation value of other boards, etc., calculate the alarm situation, and update the failure rate of the board module;

[0068] Based on the data analysis module, through formula (1), after accumulating the status value deduction scores of each cpsu board and taking the mean value, the evaluation value of the three-level cpsu board is obtained, that is, the first health evaluation score;

[0069] Combine the evaluation value and the evaluation values of each board card type with the evaluation parameters (health calculation coefficients) in formula (1) to calculate the deduction points of the board card module at the second level;

[0070] Substitute the failure rate of the board card module into formula (2) to obtain the evaluation score (health evaluation score) at the board card level. Divide it by the deduction points of the board card module to get a new evaluation score. Combine the minimum failure rate and the new evaluation score and substitute them into formula (2) to update the dynamic parameters of the health evaluation model;

[0071] Calculate the new health evaluation score according to the new dynamic parameters. Combine the health status weight coefficient and calculate the interlocking health score at this moment through formula (3). At the same time, record the data in the database, including the deduction points and health evaluation scores at each level;

[0072] Based on the data prediction module, sort the health scores and historical health scores to establish the original prediction sequence Y (0) and accumulate the original sequence Y (0) in sequence to construct the gray value sequence Z (1) and construct the gray model, that is, formula (7);

[0073] Determine the gray parameters of the gray model based on the historical health scores. Calculate the predicted gray sequence values based on the gray prediction sequence (the second gray prediction sequence) of the health scores and the gray parameters;

[0074] Subtract the previous gray sequence value from the predicted gray sequence value to obtain the predicted value of the health score at the next moment.

[0075] In this embodiment, by the mutual cooperation of each module to execute the steps of the interlocking subsystem fault prediction method, it is realized to predict faults by analyzing the change trend of the operation data of the interlocking subsystem, issue early warnings before the faults are fully formed, so as to timely remind the operation and maintenance personnel to conduct inspections and maintenance, avoid the occurrence of faults, and improve the reliability and safety of the interlocking subsystem.

[0076] The above specific implementation manners are the preferred implementation manners of the present invention, and do not limit the specific implementation scope of the present invention. The scope of the present invention includes but is not limited to this specific implementation manner. All equivalent changes made according to the shape, structure, and method of the present invention are within the protection scope of the present invention.

Claims

1. A real-time fault prediction method for interlocking subsystem based on grayscale prediction, characterized in that: The steps include: S1. Obtain the real-time operation data of the interlocking subsystem and map the real-time operation data into target analysis data of each interlocking device; S2. Based on the interlocking architecture and equipment attributes, hierarchical division is performed on the target analysis data to obtain a hierarchical data set; S3, calculate the health evaluation score of the interlocking subsystem based on the hierarchical data set combined with historical data; S4. Calculate the real-time health score of the interlocking subsystem based on the health evaluation score and the health status of the equipment at each level; S5. Construct a grayscale prediction sequence based on the real-time health score and the historical health score, and use it as the input of the grayscale model to output the grayscale sequence value; S6. Calculate the predicted health value of the interlocking subsystem at the next moment based on the grayscale sequence value.

2. The interlocking subsystem real-time fault prediction method based on grayscale prediction according to claim 1 is characterized in that: The S1 comprises the following sub-steps: Collecting real-time operation data of the interlocking subsystem based on the communication protocol of the interlocking subsystem, and encapsulating the real-time operation data into a plurality of data packets; The data packet is parsed to extract key data associated with each interlocking device, and the key data is used as target analysis data of the interlocking device.

3. The interlocking subsystem real-time fault prediction method based on grayscale prediction according to claim 1 is characterized in that: The S2 comprises the following sub-steps: According to the interlocking architecture, the target analysis data is divided into a first layer to obtain a first data set corresponding to a plurality of interlocking subsystems; Performing a second-level division on the first data set according to the device type inside each interlocking subsystem to obtain a second data set; Performing a third-level division on the second data set according to data types to obtain a third data set; A data mapping table is established to associate and store the data in each data set based on data key-value pairs, thereby completing the hierarchical division of the target analysis data.

4. The interlocking subsystem real-time fault prediction method based on grayscale prediction according to claim 3 is characterized in that: The S3 comprises the following sub-steps: Initializing a first health evaluation score of each interlocking device, and determining a minimum failure rate of the data layer in a historical operation cycle based on a historical data set; Determining dynamic parameters of a health evaluation model according to the initialized first health evaluation score and the minimum failure rate; Generate a health calculation coefficient of the interlocking device based on the importance of the divided elements of each data layer, and calculate a first health evaluation score of each level of equipment according to the health calculation coefficient and the corresponding data set; Calculate the current failure rate of the data layer according to the first health evaluation score, and update the dynamic parameter by combining the health score with the current failure rate; A health evaluation score is calculated based on the updated dynamic parameters.

5. The interlocking subsystem real-time fault prediction method based on grayscale prediction according to claim 4 is characterized in that: The step of generating a health calculation coefficient of the interlocking device based on the importance of the divided elements of each data layer, and calculating a first health evaluation score of each level of equipment according to the health calculation coefficient and the corresponding data set includes: Based on the importance of the division elements of each data layer, the importance of each division element is marked through gradient annotation; Establishing an importance judgment matrix according to the importance identification; Extracting the maximum eigenvector in the importance judgment matrix, and calculating the health calculation coefficient of the corresponding interlocking device through the maximum eigenvector; The first health evaluation score of each level of equipment is calculated based on the health calculation coefficient of the interlocking device and the corresponding data set.

6. The interlocking subsystem real-time fault prediction method based on grayscale prediction according to claim 5 is characterized in that: The S4 comprises the following sub-steps: Based on the equipment health status corresponding to each data layer, the health level of each interlocking device is marked according to the equipment health status through gradient labeling; Establishing a health degree judgment matrix according to the health degree identifier; Extracting the maximum eigenvector in the health judgment matrix, and calculating the health status weight coefficient of the corresponding interlocking device through the maximum eigenvector; The real-time health scores of the devices at each level of the interlocking subsystem are calculated based on the health evaluation score and the health status weight coefficient.

7. The interlocking subsystem real-time fault prediction method based on grayscale prediction according to claim 6 is characterized in that: The real-time health score calculation formula is as follows: Where Y t is the real-time health score at time t, k represents the total number of levels, ω i represents the health status weight coefficient of the i-th level, Represents the health evaluation score of the i-th level at time t.

8. The interlocking subsystem real-time fault prediction method based on grayscale prediction according to claim 7 is characterized in that: The S5 comprises the following sub-steps: Generate a first grayscale prediction sequence based on the historical health scores of interlocking devices at each level, accumulate the first grayscale prediction sequences of interlocking devices at the same level, and construct a first grayscale value queue; Establishing a grayscale model according to the fusion of the first grayscale value queue and the first grayscale prediction sequence, and determining grayscale parameters of the grayscale model based on the historical health score; Generate a second grayscale prediction sequence based on the real-time health scores of interlocking devices at each level, accumulate the second grayscale prediction sequences of interlocking devices at the same level, and construct a second grayscale value queue; Based on the grayscale parameters, the second grayscale value queue is substituted into the grayscale gradient equation of the grayscale model to obtain a grayscale sequence value.

9. The interlocking subsystem real-time fault prediction method based on grayscale prediction according to claim 8 is characterized in that: The S6 comprises the following sub-steps: Based on the subtraction of the grayscale sequence value of the current level from the grayscale sequence value of the previous level, the predicted health value of the interlocking subsystem at the next moment is obtained.

10. An interlocking subsystem fault prediction system based on grayscale prediction, applicable to the interlocking subsystem real-time fault prediction method based on grayscale prediction as described in any one of claims 1 to 9, characterized in that: include: Data acquisition module, used to obtain target analysis data of interlocking equipment; A data processing module is used to hierarchically divide the target analysis data and obtain hierarchical data sets; A data analysis module for calculating the health scores of interlocking subsystems based on hierarchical data sets; The data prediction module is used to calculate the health score and obtain the health prediction value of the interlocking subsystem at the next moment.

Images