A knowledge base construction method for railway collaborative operation and maintenance

Abstract

This invention relates to the field of railway operation and maintenance technology, specifically a method for constructing a knowledge base for collaborative railway operation and maintenance. The method includes: performing knowledge association searches according to railway collaborative operation and maintenance instructions, configuring multiple event nodes; extracting dependencies between any two event nodes, determining hierarchical labels and conflict types between event nodes, and setting mapping relationships; identifying the mapping path corresponding to the event node change when the number of nodes mapped to any event node changes, and determining the association between the mapping path and the number of node changes; when there are conflict types in the association relationships, calculating the propagation range of event node updates based on local and global association characteristics; when there are no conflict types, determining the mapping path that satisfies the minimum node constraint, and configuring the propagation range of event node updates based on the timeliness differences of event nodes. This improves the efficiency and adaptability of railway operation and maintenance.

Description

Technical Field

[0001] This invention relates to the field of railway operation and maintenance technology, specifically a method for constructing a knowledge base for collaborative railway operation and maintenance. Background Technology

[0002] With the expansion of the railway network, the number of collaborative operation and maintenance instructions has surged, and these instructions involve multi-dimensional information such as equipment location, fault handling, and cross-departmental collaboration. Existing operation and maintenance knowledge bases often use simple keyword matching to parse railway collaborative operation and maintenance instructions, or only search by category or tag, failing to clearly define event boundaries and scope. This results in missing attributes in event node configurations, easily leading to operation and maintenance safety incidents.

[0003] For example, Chinese Patent Publication No. CN117271767A discloses a method for establishing an operation and maintenance knowledge base based on multi-agent intelligence. The method includes the following steps: acquiring AI knowledge reserve data; performing database classification processing on the AI ​​knowledge reserve data to generate a classification database; performing data channel initialization processing on the classification database via a local area network to generate an initial transmission channel; performing node setting processing on the classification database according to the local area network to generate a transmission monitoring master node; acquiring user input data; performing semantic analysis processing on the user input data based on natural language processing technology to generate user semantic parsing data; and performing dimensional transformation processing on the user semantic parsing data according to a dimensional transformation algorithm to generate user semantic transformation data.

[0004] For example, Chinese Patent Publication No. CN117194459A discloses a method, system, device, and storage medium for updating an operations and maintenance (O&M) knowledge base based on O&M events. The method includes the following steps: acquiring O&M event data; performing relevance filtering on the O&M event data to obtain a word vector training set; updating a historical O&M knowledge base model based on the word vector training set to obtain a first O&M knowledge base model; inputting newly received O&M events into the first O&M knowledge base model to obtain a first knowledge base entry and inputting the new O&M events into a preset O&M knowledge base to obtain a second knowledge base entry; and updating the preset O&M knowledge base based on the first and second knowledge base entries. This method can improve the efficiency of data maintenance.

[0005] Existing technologies manage data channels through decision trees and use semantic similarity for dimensional transformation to achieve integrated processing of operation and maintenance knowledge; they also filter time data through relevance, use time data to form a set of operation and maintenance events, and finally complete the knowledge base update process. However, these processing methods are biased towards semantic filtering and emphasize the direct relevance of data processing, rather than determining the boundary range corresponding to each operation and maintenance event. This leads to missing data update dimensions, resulting in attribute missing problems when reusing railway collaborative operation and maintenance instructions, and reducing the adaptation effect of operation and maintenance knowledge base data configuration. Summary of the Invention

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a knowledge base construction method for railway collaborative operation and maintenance, including: S1, obtaining railway collaborative operation and maintenance instructions, performing knowledge association search according to the railway collaborative operation and maintenance instructions, and configuring multiple event nodes.

[0007] S2: Extract the dependency relationship between any two event nodes, determine the hierarchical labels and conflict types between each event node, and set the mapping relationship between each event node.

[0008] S3, when the number of nodes mapped to any event node changes, identify the mapping path corresponding to the event node change and determine the relationship between the mapping path and the number of nodes changed.

[0009] S4. When there is a conflict between the mapping path and the number of node changes, extract the local and global association features of each event node change, and use the local and global association features to estimate the propagation range of each event node update.

[0010] S5. When there is no conflict type in the relationship between the mapping path and the number of node changes, obtain the fault label of the event node, perform reverse calculation based on the fault label, determine the mapping path that satisfies the minimum node constraint, and configure the propagation range when the event node is updated based on the time difference of each event node in the mapping path.

[0011] The beneficial effects of this invention are as follows: First, this invention transforms railway collaborative operation and maintenance instructions in natural language form into structured tag sequences and complete event nodes through a process of named entity recognition → equipment / fault / maintenance operation tag extraction → tag sequence combination → direct / indirect pointing relationship verification → context supplementation; ensuring that each event node contains core information such as equipment location, fault characteristics, operation and maintenance operations, and responsible entity, laying a precise data foundation for subsequent knowledge base association search and dynamic updates.

[0012] Second, this invention establishes a structured mapping relationship between event nodes by classifying dependencies (causality / temporality / coordination / constraint) → configuring hierarchical labels → constructing conflict datasets → classifying conflict types → combining hierarchical sequences of conflict-free scenarios. This determines the data mapping under conflict and non-conflict conditions and performs differentiated storage simultaneously to avoid data dependency chaos caused by data conflicts, which could lead to data operation and maintenance accidents.

[0013] Third, this invention tracks the changes in the number of nodes before and after data updates by generating node change records → setting mapping paths → quantifying association relationships (number of node changes / path complexity / path association degree), establishing a quantitative association between changes and mapping paths, ensuring quantitative processing of paths before and after data changes, providing data association support for each collaborative instruction issuance, ensuring that paths can be quickly adapted after node changes, avoiding the failure of association logic in the knowledge base, and supporting the dynamic iteration of the knowledge base as railway operation and maintenance progresses.

[0014] Fourth, this invention estimates the propagation range of event nodes in conflict scenarios through a method of local / global correlation feature extraction (hierarchical label data / number of node changes / number of path branches / related departments) → delay calculation based on feature distribution ratio → collaborative evolution (affected node identification → path distance-propagation delay calculation → convergence condition extraction). This ensures that the propagation range covers cross-departmental nodes and security constraint nodes necessary for conflict resolution, and that the propagation range remains stable through convergence conditions, supporting the setting of the knowledge base update range in conflict scenarios. Finally, based on fault label field splitting → minimum node constraint determination → intersection of mapping path sets → timeliness difference setting → multi-path batch processing, it plans event nodes in conflict-free scenarios; avoiding conflict situations in conflict-free scenarios, thereby improving the efficiency and speed of knowledge base updates. Attached Figure Description

[0015] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0016] Figure 1 This is a flowchart illustrating a knowledge base construction method for railway collaborative operation and maintenance.

[0017] Figure 2 This is a flowchart illustrating step S1 of a knowledge base construction method for railway collaborative operation and maintenance.

[0018] Figure 3 This is a flowchart illustrating step S2 of a knowledge base construction method for railway collaborative operation and maintenance.

[0019] Figure 4 This is a flowchart illustrating step S3 of a knowledge base construction method for railway collaborative operation and maintenance.

[0020] Figure 5This is a flowchart illustrating step S4 of a knowledge base construction method for railway collaborative operation and maintenance.

[0021] Figure 6 This is a flowchart illustrating step S5 of a knowledge base construction method for railway collaborative operation and maintenance. Detailed Implementation

[0022] The embodiments of the present invention are described in detail below. The embodiments described below are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. Where specific techniques or conditions are not specified in the embodiments, they shall be performed in accordance with the techniques or conditions described in the literature in the art or in accordance with the product manual.

[0023] See Figure 1 A knowledge base construction method for railway collaborative operation and maintenance includes: S1, obtaining railway collaborative operation and maintenance instructions, performing knowledge association search according to the railway collaborative operation and maintenance instructions, and configuring multiple event nodes.

[0024] S2: Extract the dependency relationship between any two event nodes, determine the hierarchical labels and conflict types between each event node, and set the mapping relationship between each event node.

[0025] S3, when the number of nodes mapped to any event node changes, identify the mapping path corresponding to the event node change and determine the relationship between the mapping path and the number of nodes changed.

[0026] S4. When there is a conflict between the mapping path and the number of node changes, extract the local and global association features of each event node change, and use the local and global association features to estimate the propagation range of each event node update.

[0027] S5. When there is no conflict type in the relationship between the mapping path and the number of node changes, obtain the fault label of the event node, perform reverse calculation based on the fault label, determine the mapping path that satisfies the minimum node constraint, and configure the propagation range when the event node is updated based on the time difference of each event node in the mapping path.

[0028] Each event node represents a fault operation and maintenance event in a collaborative operation and maintenance scenario. When performing a correlation search, the event node can be set with descriptive information based on the device, fault, and operation and maintenance related descriptions.

[0029] For example, when the system receives a docking instruction from a department and a linkage instruction from the dispatch center, it searches the database for named entities that are text-related to the request, and finds data related to the device ID, type, location, and lifecycle stage; data related to the fault, such as fault code, severity, and scope of impact; and data directly representing the operation and maintenance process, such as operation ID, required skills, and standard working hours. This data is set as multiple event nodes in a context-related manner to interpret the data nodes set under the current collaborative operation and maintenance.

[0030] like Figure 2 As shown, the implementation of step S1 includes: S11, analyzing the named entities of railway collaborative operation and maintenance instructions, where the named entities represent specific text, keywords, and other descriptive terms; identifying the equipment tags, fault tags, and operation and maintenance tags corresponding to the railway collaborative operation and maintenance instructions.

[0031] When acquiring equipment tags, fault tags, and maintenance operation tags, each tag is acquired in a different way. Equipment tags will search for named entities containing keywords such as equipment name, model, and serial number to obtain the basic information of each device, and then combine this basic information to form the equipment tag. Fault tags will identify named entities that represent the specific time and conditions of the fault, such as overheating, abnormal noise, and power failure, and combine this data to form the fault tag. As for maintenance operation tags, they will view the explicit or implicit operation requirements in the instructions and use those operation requirements as the maintenance operation tags.

[0032] S12 combines the equipment tags, fault tags, and maintenance operation tags corresponding to railway collaborative operation and maintenance instructions into a tag sequence for association search. Each tag sequence will be used as a processing method for context retrieval and prompt word input, thereby completing the event perception of fault operation and maintenance. In railway fault operation and maintenance scenarios, the same fault tag has different impacts on train operation safety and operation and maintenance collaboration under different equipment tags. At this time, based on the fault orientation of equipment tags, fault tags, and maintenance operation tags in the corresponding scenarios, relevant event nodes will be configured for the current association search, thereby perceiving the collaborative association triggered by each collaborative operation instruction.

[0033] S13, perform fault pointing verification on each tag in the tag sequence, and obtain the basic event information after tag combination.

[0034] When performing fault location verification, the events when the tags are combined are distinguished by establishing the relationship between the tags, thereby labeling the basic information of the corresponding events.

[0035] The implementation of step S13 includes: based on the input device tag, fault tag and operation tag, performing direct pointing relationship and indirect pointing relationship verification in sequence, and defining the tag corresponding to each pointing relationship after verification.

[0036] The relevant tags are combined into a tag set; the boundaries and scope of the event are determined by the tag set, which serves as the basic information of the event in the output.

[0037] Direct pointing verification verifies the association between devices and faults, devices and maintenance operations, and faults and maintenance operations. For example, if label A is a device label, label B is a fault label, and label C is a maintenance operation label, then A and B have a direct pointing relationship if the location of A matches the location of B's ​​occurrence, and the state of A is compatible with the fault type of B. If the operation type of C is the standard processing flow of B, and the time window of C includes the occurrence time of B, then C and B have a processing pointing relationship. If C is a processing flow for A, and the timestamps of the two correspond, then A and C have a direct pointing relationship. As for fault pointing verification, it describes whether there is a pointing relationship between devices, faults, and maintenance operations in fault maintenance scenarios.

[0038] Indirect pointing is a method for verifying causal relationships and time series. Causal pointing means that if any label X causes label Y, and label Y causes label Z, then there is an indirect pointing relationship between X and Z. In this case, labels X, Y, and Z can be any of the following: device label, fault label, or maintenance operation. As for time series verification, it involves checking whether these corresponding data are within the same time window to find related labels within the same time period; or checking whether the timestamps before and after the causal relationship conflict. For example, if label X is a relative cause but is later than label Z, it indicates that there is a problem with the timing of the indirect association, and it is necessary to point out that there is no contradictory indirect pointing relationship. After these fault and maintenance-related data are identified as belonging to one-to-one, one-to-many, many-to-one, and many-to-many label sets, the relevant data are combined into basic event information.

[0039] Then, the boundaries and scope of the event are identified based on the number and description of the tags in the tag set. For example, the boundaries are essentially composed of the progress of the event type (fault type / operation type), the event status (in progress / completed / cancelled), and related tags (fault tag ID / major equipment tag ID / critical operation tag ID); the scope describes the time range (start time, end time) and spatial range (location of occurrence, affected area), as well as other data lists statistically analyzed through tags, thereby explaining the basic information of the current event.

[0040] S14, supplement the basic information of the event with contextual information, and regard the supplemented data as the output event node.

[0041] Supplementing contextual information involves examining the contextual information when the railway collaborative operation and maintenance instruction was issued, such as the standard processing flow corresponding to the instruction and other instructions in the same location as the instruction, to supplement the information of the current event node. Then, the event nodes are described in the form of a set of tags, such as event nodes with collaborative operation as the main combination, event nodes with equipment as the main combination, event nodes with fault as the main combination, and event nodes with specific operation and maintenance operation as the main combination.

[0042] The method of information supplementation is to use natural language processing algorithms to extract keywords and high-frequency words from the text to complete the information filling, which will not be elaborated on here.

[0043] In one embodiment of the present invention, such as Figure 3 As shown, the implementation of step S2 includes: S21, clarifying the dependency type between each event node. The dependency type is represented as causal dependency, temporal dependency, collaborative dependency, and constraint dependency. Causal dependency means that the previous node is the triggering cause of the next node; without the former, the latter is meaningless. Temporal dependency means that nodes must be executed in the order of railway operation and maintenance process and cannot be reversed. Collaborative dependency means that the next node needs the collaborative result of the previous node to support it, such as the relationship of collaboration → operation and maintenance, which indicates the nodes involved in the collaborative operation at this time. Constraint dependency means that the safety specifications and setting standards in its execution procedure must be satisfied before the relevant operation can be executed.

[0044] S22, based on the dependency relationship type, configure the hierarchical labels of each event node in sequence, and integrate the hierarchical labels and dependencies into a structured mapping relationship; when extracting dependencies in step S2, determine the hierarchical labels of the event nodes based on the event node division method, such as configuring the hierarchical labels in sequence according to the main combination of device, fault, collaborative operation and specific operation and maintenance operation in its event basic information, forming a hierarchy such as device → fault → collaborative operation → operation and maintenance operation. At the same time, considering constraint dependencies, the relevant hierarchy of constraints is introduced after device and fault, and finally realizes the hierarchy of device → fault → constraint → collaborative operation → operation and maintenance operation, and configure its hierarchical label ID. Then, according to the mutual relationship of each event node in the hierarchical label, complete the mapping relationship setting of each event node.

[0045] In addition, when integrating hierarchical labels and dependencies, it is also necessary to identify dependency conflicts under multi-department and multi-constraint collaborative work, clarify the mapping relationship under conflict scenarios, and then extract the changes of event nodes when receiving railway collaborative operation and maintenance instructions.

[0046] When integrating hierarchical labels and dependencies in step S22, the implementation method also includes: S221, determining whether there is a dependency conflict when the current event node is in collaborative operation and maintenance. If there is a dependency conflict, integrating the hierarchical labels with the multiple hierarchical labels corresponding to the dependency conflict to form a conflict dataset.

[0047] Dependency conflict determination will be based on the following conditions: 1. Incorrect hierarchical label order, such as the hierarchical labels of some event nodes being triggered prematurely, resulting in the execution of specific operation and maintenance operations before the collaborative operation is executed; 2. Constraint conflict, that is, the constraint requirements of different event nodes conflict; 3. Resource or permission conflict, that is, when issuing railway collaborative operation and maintenance instructions, some operation and maintenance operations cannot be executed, or the actual content of the operation and maintenance operations does not conform to the collaborative operation but is not labeled.

[0048] If conflicts exist, these conflicting data need to be combined and mapped according to the scenario at the time of the conflict to assist in updating the data in the knowledge base.

[0049] S222: Conflict types are divided based on the operation and maintenance operations corresponding to the conflict dataset, and the divided data is regarded as the output mapping relationship.

[0050] The description of maintenance operations in the conflict dataset will represent the specific scenarios of railway maintenance operations. According to the scenarios, the data will be mapped and output separately. For example, different forms of maintenance operations such as outdoor maintenance operations, multi-department collaborative operations, and resource allocation operations will be used as the main mapping part to illustrate the mapping relationship under the data conflict scenario.

[0051] S223, when there are no dependency conflicts, group the hierarchical tags in hierarchical order and mark them as mapping relationships.

[0052] When there are no conflicts, the corresponding data will be fixed with a mapping relationship. This fixed mapping relationship is also the main basis for viewing data updates later, in order to determine the event nodes involved in each instruction update, to determine the control status of each railway collaborative operation and maintenance instruction, and finally to complete the data change processing.

[0053] In one embodiment of the present invention, the configuration of the mapping relationship has been completed in step S2. The event nodes under the mapping relationship can form a mapping path. Step S3 is used to solidify the mapping path, quantify the association between the occurrence of conflicts and data changes, so as to determine the update of the data corresponding to the events.

[0054] like Figure 4 As shown, the implementation of step S3 includes: S31, determining the mapping relationship between event nodes before and after data update, and generating node change records based on the number of event nodes corresponding to each mapping relationship.

[0055] When updating data, first check whether it is conflicting data or normal mapped data. Then check the event nodes of data changes such as adding nodes, deleting nodes, and modifying nodes. Record the changed node ID, hierarchical label, constraint conditions, and reason for change. Then, record the data of instruction execution according to its mapping relationship.

[0056] S32, view the triggering reason of the node change record, and set the mapping path based on the number of node changes recorded in the node change record.

[0057] S33 quantifies the association relationship based on the number of node changes, path complexity, and path correlation for each mapping path, and synchronizes the quantified values ​​to each mapping path.

[0058] In the above processing, the mapping path records the traces of node changes. For example, in conflict handling scenarios, it records conflicting data and quantifies the relationship between conflict type and path conflict based on the number of node changes, path complexity, and path relevance. Path complexity is determined by the product of the number of dependency types and branches of the mapping path in the event of a conflict. As for path relevance, it is set based on the matching degree between the mapping path and railway rules. The matching degree of railway rules is mainly reflected in compliance with traffic safety, personal safety, process compliance, and multi-department confirmation compliance, to reflect the command compliance of railway collaborative operation and maintenance instructions. This part is mainly set by checking whether the changed event nodes conform to the normal operation and maintenance processing specifications, and then setting the matching degree. When multiple event nodes are involved, the path relevance is set by weighted average. The weight is based on the conflict type of the corresponding event node and the relevant weight value is retrieved from the database. The quantified correlation will then be based on the product of the number of node changes, path complexity, and path correlation. The calculated data will be normalized and used as the quantified value of the corresponding mapping path. The larger the value, the more obvious the change of the event node. This will enable rapid tracking of the update status under railway collaborative operation and maintenance instructions, thereby maintaining the update processing in railway operation and maintenance scenarios.

[0059] Meanwhile, if the current mapping path is not a conflict resolution scenario, but only a path maintenance or monitoring of explicit changes, such as a newly added constraint event node being a blizzard, or the termination of some railway collaborative operation and maintenance instructions, resulting in the deletion of relevant event nodes for collaborative operations, the specific updated data will be recorded and its data will be synchronously mapped and updated. The quantification method of the association in non-conflict scenarios is consistent with that in conflict resolution scenarios. After the update is completed, a data description file needs to be configured to record the update details to achieve data management.

[0060] In step S3, the association quantification method tends to identify the severity of node updates and quantify and sort multiple mapping paths according to the number of updates, so as to realize node update processing under different mapping relationships.

[0061] In one embodiment of the present invention, such as Figure 5 As shown, the implementation of step S4 includes: S41, based on the mapping relationship between each event node in the mapping path, using the data corresponding to the hierarchical label as the local association feature, and using the number of node changes, the number of path branches, and the associated departments corresponding to the mapping path as the global association features.

[0062] Local correlation features will specifically point to fault labels, operation and maintenance labels, equipment labels, and their collaborative operations and constraints to illustrate the distribution of each event node in a specific implementation scenario. Global correlation features will describe the conflicting departments, the number of nodes, and path branches involved to explain the impact on a global scale.

[0063] S42. Based on the distribution ratio of local and global correlation features under different conflict types, determine the delay time when executing railway collaborative operation and maintenance instructions.

[0064] The delay time consists of two parts: a basic delay based on historical data and knowledge base, and a real-time calculated value based on the current conflict type. For each event node, the distribution ratio of its local correlation features in different conflict types is calculated first, and then the distribution ratio of the global correlation features of the entire mapping path in different conflict types is calculated.

[0065] The base latency will be directly obtained from the database and consists of three parts: the average processing time under similar scenarios, the average working time of each operation, and the adjustment value under different environmental conditions. It is set by weighted summation. The weighted data are, in order, the average processing time, the average working time, and the geometric mean of the average processing time and the average working time. The weight will be set based on the ratio of the normalized value of each part to the sum, that is, the ratio of the normalized value of the average processing time, the average working time, and the geometric mean to 1, thereby setting the weight and completing the base latency setting.

[0066] As for the real-time calculated value based on the current conflict type, this value can be calculated in the following way.

[0067] ;in, This represents the real-time calculated value based on the current conflict type. C1 represents the set of all conflict types, and c represents the index number of the conflict type, with a value range from 1 to C1. This represents the number of event nodes, and i represents the index of the corresponding event node on the currently processed mapping path, with a value range from 1 to N. The reference delay for conflict type c is obtained by looking at the average delay for the corresponding conflict type. This reference delay and the base delay represent different scenarios. The base delay represents the baseline time estimate for a specific collaborative operation and maintenance task, while the reference delay describes the average delay for a specific conflict type. This represents the proportion of the local association features of event node i in conflict type c. This indicates the proportion of global correlation features in conflict type c; , This represents the weight of local and global correlation features in conflict type c; the weight is based on the ratio of the proportion of conflict type c to the sum of all conflict types. This real-time calculated value is added to the base delay to obtain the delay time for executing railway collaborative operation and maintenance instructions.

[0068] S43. Based on the acquired delay time, perform collaborative evolution on each event node to determine the propagation range under collaborative evolution. When the propagation range generated by any event node reaches convergence, output the propagation range corresponding to the current event node to determine the specific situation of data update when the knowledge base changes.

[0069] When performing cooperative evolution in step S43, the implementation method includes: S431, identifying other nodes affected by the current event node to obtain the set of affected nodes.

[0070] S432, for each affected node and the current event node, calculate the propagation delay of each affected node.

[0071] First, analyze the impact range of the current delayed event node on other nodes in the mapping path. Based on the dependencies between nodes, resource dependencies, and time window overlap, determine the set of nodes that need to receive delay information and mark them as affected nodes.

[0072] For each affected node, the propagation delay is calculated based on its path distance to the original delayed node and its dependencies. The path distance can be a logical distance (such as the number of intermediate nodes) or a temporal distance (such as the scheduling interval). The propagation delay decreases with increasing distance, and the degree of decrease is controlled by a decay coefficient determined by domain experience.

[0073] The affected nodes will be extracted directly through dependencies, including subsequent nodes directly affected by the current event node, other nodes using the same resources as the current event node, and adjacent nodes that overlap with the current event node in time. These adjacent nodes indicate nodes that are adjacent in the hierarchical relationship. These event nodes that can cause delays due to the operation delay of the current node will be marked to obtain the set of affected nodes.

[0074] The propagation delay will be calculated based on an exponential decay formula, and the propagation delay value is expressed as follows.

[0075] ;in, This represents the propagation delay of node j, where node j represents an event node affected by node i. This represents the delay time of node i. Represents an exponential constant. This represents the path distance between nodes i and j. Its value can be the number of intermediate nodes or its scheduling interval to quantify the interval between different event nodes. If both logical distance and time distance are introduced, their normalized average value should be selected as the input path distance. The attenuation coefficient can be obtained by solving the propagation delay based on the current conflict type. It reflects the rate at which the delay of different event nodes decays with their logical or temporal distance. The influence weight is set based on the ratio of the sum of the delay times of nodes i and j to the total delay time, or based on the dependency relationship between nodes i and j, and the weight value is retrieved from the database.

[0076] S433: Based on the propagation delay of each co-evolution, extract the convergence condition corresponding to the propagation delay by the difference between the propagation delays of two adjacent co-evolutions, and output the part of the propagation delay that is less than the convergence threshold to obtain the propagation range of the corresponding event node.

[0077] After structuring the delay information and the adjustment data of collaborative evolution, it is directed to the affected nodes. The information includes the delay time, the scope of affected nodes, and the propagation delay. By examining the propagation delay of the same event node after multiple iterations, it is determined whether the affected event nodes have reached convergence. For example, the convergence threshold is set to 0.1 hours to correspond to the time delay generated by each specific operation and maintenance operation under the collaborative command, and whether these time delays can reach stability, thereby minimizing the overall impact of the fault on railway operation, completing the data update under collaborative data processing, and improving the response efficiency and quality of similar events.

[0078] In one embodiment of the present invention, in the processing method of step S5, compared with the conflict type processing of step S4, this step is biased towards viewing the mapping path of each fault, finding the data that meets the requirements through the minimum node constraint, and updating it according to the data, thereby obtaining the propagation range that different faults can generate.

[0079] like Figure 6As shown, the implementation of step S5 includes: S51, for the minimum node constraint of any mapping path, splitting the fault labels of each event node on the mapping path into multiple fields, and determining the set of fields corresponding to the mapping path when the data is updated, such as device ID, fault level, response time and other fields directly marked on the fault label.

[0080] The minimum node constraint mentioned above represents the required node corresponding to the fault label, and the mapping path corresponding to its required node is regarded as the mapping path of the current scenario.

[0081] For example, the path selected under the minimum node constraint should at least follow the pattern of device → fault → collaborative operation → maintenance operation, clearly defining the faulty device, fault content, collaborative operation topic, and specific operation content. The path should be mapped to form a sub-process for fault handling, and the necessary field set should be selected by splitting the sub-process according to the fields it contains.

[0082] For example, for event nodes corresponding to equipment, fields such as equipment ID, kilometer marker, and equipment type can be extracted; for event nodes corresponding to faults, fields such as fault ID, fault level, and fault type can be extracted; for event nodes corresponding to collaborative operations, fields such as responsible department ID, responsible person, and response time can be extracted; and for event nodes corresponding to specific operation and maintenance operations, fields such as operation process ID, operation time, and safety standard ID can be extracted.

[0083] Each field is filtered in relation to the scenario corresponding to the current event node. Pairwise comparisons are performed continuously to output the data with the least common intersection. The part of this data that conforms to the pattern of device → fault → collaborative operation → maintenance operation is output as the mapping path identified at this time. The mapping path identified at this time represents the minimum node constraint. Its reading order and related data can be viewed to complete the data update of the node.

[0084] The minimum node constraint requires different nodes in different scenarios. For example, in a typical scenario, it requires four nodes: device → fault → collaborative operation → maintenance operation. If there is no collaborative operation, only three nodes are required: device → fault → maintenance operation. In scenarios that depend on constraints, five nodes are required: device → fault → constraint → collaborative operation → maintenance operation, in order to complete the processing of related event nodes.

[0085] When the fault label is split into multiple fields in step S51, the implementation method also includes: determining whether the currently identified data allows the minimum node constraint processing of multiple mapping paths to be executed in a single execution. If so, the timeliness information of each mapping path during the update is viewed, the reading order and time range of the field set are set, and the timeliness information will further reveal the response time, standard time limit, actual update time and other data generated on its mapping path, and the data update order and time range are obtained according to its timestamp.

[0086] If not, it means that the minimum node constraint is not met, field splitting is missing, or there is a conflict when checking the relevant data using fault labels. This indicates that there are still hidden conflicts in the current data, and the corresponding data needs to be processed in a way that targets the conflict type.

[0087] S52, based on the reading order and content of the field set, performs set intersection on the mapping paths and sets the time difference of the mapping paths; the set intersection is based on the field set, extracts the nodes contained in all mapping paths, ensures that the minimum node constraint can be focused when updating multiple paths, and updates the relevant data to the nodes of their intersection to complete the unified data update processing; as for the time difference, based on its response time, standard time limit, actual update time and other data, it can further explain the difference in its delay time, and mark the relevant nodes on each mapping path according to its difference.

[0088] S53. Based on the timeliness differences of the mapping paths, plan the update paths for each event node, and consider the update paths as the propagation range of the output. When planning the update paths, prioritize overlapping nodes and then non-overlapping nodes. Sort the relevant event nodes according to the timeliness differences from smallest to largest, and update the event nodes to the corresponding time periods based on the identified time ranges. Then, collect the trajectory and order of the data updates, and use these updated event nodes as the propagation range of the output to determine the processing flow of the operation and maintenance system under normal fault tag binding, so as to play a role in reuse and enlightenment for subsequent similar faults.

[0089] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention, which are still covered within the protection scope of the present invention.

Claims

1. A method for constructing a knowledge base for collaborative railway operation and maintenance, characterized in that, include: S1: Obtain railway collaborative operation and maintenance instructions, perform knowledge association search according to railway collaborative operation and maintenance instructions, and configure multiple event nodes; S2, extract the dependency relationship between any two event nodes, determine the hierarchical labels and conflict types between each event node, and set the mapping relationship between each event node; S3, when the number of nodes mapped to any event node changes, identify the mapping path corresponding to the event node change and determine the relationship between the mapping path and the number of nodes changed; S4. When there is a conflict between the mapping path and the number of node changes, extract the local and global association features of each event node change, and use the local and global association features to estimate the propagation range of each event node update. S5. When there is no conflict type in the relationship between the mapping path and the number of node changes, obtain the fault label of the event node, perform reverse calculation based on the fault label, determine the mapping path that satisfies the minimum node constraint, and configure the propagation range when the event node is updated based on the time difference of each event node in the mapping path. The implementation of step S4 includes: S41, based on the mapping relationship between each event node in the mapping path, using the data corresponding to the hierarchical label as the local association feature, and using the number of node changes, the number of path branches, and the associated departments corresponding to the mapping path as the global association feature; S42, based on the distribution ratio of the local association feature and the global association feature under different conflict types, determining the delay time when the railway collaborative operation and maintenance instruction is executed; S43, based on the obtained delay time, performing collaborative evolution on each event node, determining the propagation range under collaborative evolution, and outputting the propagation range corresponding to the current event node when the propagation range generated by any event node reaches convergence; The minimum node constraint mentioned above represents the required node corresponding to the fault label, and the mapping path corresponding to its required node is regarded as the mapping path of the current scenario.

2. The knowledge base construction method for railway collaborative operation and maintenance according to claim 1, characterized in that, The implementation methods for step S1 include: S11, analyze the named entities of railway collaborative operation and maintenance instructions. Named entities represent specific text, keywords, and descriptive terms; identify the equipment tags, fault tags, and operation and maintenance tags corresponding to railway collaborative operation and maintenance instructions; S12 combines the equipment tags, fault tags, and operation tags corresponding to railway collaborative operation and maintenance instructions into a tag sequence for association search; S13, perform fault pointing verification on each tag in the tag sequence, and obtain the basic event information after tag combination; S14, supplement the basic information of the event with contextual information, and regard the supplemented data as the output event node.

3. The knowledge base construction method for railway collaborative operation and maintenance according to claim 2, characterized in that, The implementation methods of step S13 include: Based on the input device tags, fault tags, and operation tags, the direct and indirect pointing relationships are verified in sequence, and the tags corresponding to each pointing relationship are defined after verification. The relevant tags are combined into a tag set; the boundaries and scope of the event are determined by the tag set, which serves as the basic information of the event in the output.

4. The method for constructing a knowledge base for railway collaborative operation and maintenance according to claim 1, characterized in that, Step S2 can be implemented in the following ways: S21, specify the dependency type between each event node, which is represented as causal dependency, temporal dependency, cooperative dependency and constraint dependency; S22, based on the dependency relationship type, configure the hierarchical labels of each event node in sequence, and integrate the hierarchical labels and dependencies into a structured mapping relationship.

5. The knowledge base construction method for railway collaborative operation and maintenance according to claim 4, characterized in that, When integrating hierarchical tags and dependencies in step S22, the implementation also includes: S221, determine whether there is a dependency conflict during collaborative operation and maintenance of the current event node. If there is a dependency conflict, integrate the hierarchical labels corresponding to the dependency conflict to form a conflict dataset. S222, divide the conflict types according to the operation and maintenance operations corresponding to the conflict dataset, and regard the divided data as the output mapping relationship; S223, when there are no dependency conflicts, group the hierarchical tags in hierarchical order and mark them as mapping relationships.

6. The method for constructing a knowledge base for railway collaborative operation and maintenance according to claim 1, characterized in that, Step S3 can be implemented in the following ways: S31, determine the mapping relationship between event nodes before and after data update, and generate node change records based on the number of event nodes corresponding to each mapping relationship; S32, view the triggering reason of the node change record, and set the mapping path based on the number of node changes recorded in the node change record; S33 quantifies the association relationship based on the number of node changes, path complexity, and path correlation for each mapping path, and synchronizes the quantified values ​​to each mapping path.

7. The method for constructing a knowledge base for railway collaborative operation and maintenance according to claim 6, characterized in that, When performing cooperative evolution in step S43, the implementation methods include: S431, Identify other nodes affected by the current event node to obtain the set of affected nodes; S432, For each affected node and the current event node, calculate the propagation delay of each affected node; S433: Based on the propagation delay of each co-evolution, extract the convergence condition corresponding to the propagation delay by the difference between the propagation delays of two adjacent co-evolutions, and output the part of the propagation delay that is less than the convergence threshold to obtain the propagation range of the corresponding event node.

8. A method for constructing a knowledge base for railway collaborative operation and maintenance according to claim 7, characterized in that, When the fault label is split into multiple fields in step S51, the implementation method also includes: Determine whether the currently identified data allows for the minimum node constraint processing of multiple mapping paths in a single execution. If so, check the timeliness information of each mapping path during the update and set the reading order and time range of the field set.

9. A method for constructing a knowledge base for railway collaborative operation and maintenance according to claim 1, characterized in that, Step S5 can be implemented in the following ways: The implementation of step S5 includes: S51, for the minimum node constraint of any mapping path, splitting the fault label of each event node on the mapping path into multiple fields, and determining the set of fields corresponding to the mapping path when the data is updated; S52, based on the reading order and content of the field set, perform set intersection on the mapping path and set the time difference of the mapping path; S53, based on the timeliness differences of the mapping path, plan the update path of each event node, and regard the update path as the propagation range of the output.

Images