Railway fault tracing method and system based on weighted path sorting algorithm

Abstract

The invention relates to the technical field of railway fault traceability, in particular to a railway fault traceability method and system based on a weighted path sorting algorithm, and the method comprises the steps: constructing a traceability knowledge graph according to historical fault data, equipment operation data and maintenance records of a railway system; and then a weighted path sorting algorithm is carried out on the traceability knowledge graph by combining a Bayesian network model, a random walk algorithm and an attention mechanism model, so that efficient and accurate railway safety fault traceability is realized, the safety and reliability of railway transportation are remarkably improved, and the method has a wide application prospect and a remarkable practical value.

Description

Technical Field

[0001] The present application relates to the technical field of railway fault tracing, and in particular to a railway fault tracing method and system based on a weighted path sorting algorithm. Background Art

[0002] The traceability knowledge graph is a structured data model used to describe, record and track the source of information or data and its change process. Its main goal is to help users ensure key tasks such as operational safety, maintenance efficiency and accident tracking, so that users can trace the source and understand the entire process of data generation, transmission, change and use.

[0003] Railway transportation is one of the important modes of transportation in modern society, and its safety is directly related to public safety and economic operation. However, the railway system is complex, involving multiple subsystems and a large number of equipment. Once a failure occurs, quickly and accurately determining the root cause of the failure and handling it is an important part of ensuring railway transportation safety.

[0004] Existing railway safety fault tracing methods mainly rely on manual experience and traditional data analysis techniques, and have problems such as low diagnostic efficiency and insufficient accuracy. Summary of the invention

[0005] In order to at least to some extent overcome the problem that the railway safety fault tracing method in the related art mainly relies on manual experience and traditional data analysis technology, and has low diagnostic efficiency and insufficient accuracy, the present application provides a railway fault tracing method and system based on a weighted path sorting algorithm.

[0006] The scheme of this application is as follows:

[0007] According to a first aspect of an embodiment of the present application, a railway fault tracing method based on a weighted path sorting algorithm is provided, comprising:

[0008] Obtain the historical fault data, equipment operation data and maintenance records of the railway system to construct triples;

[0009] Store triple information in a graph database and build a traceability knowledge graph;

[0010] Constructing a subgraph of the traceability knowledge graph as a fault causal chain, and extracting a Bayesian network model corresponding to the fault causal chain;

[0011] Weighting the fault causal chain according to the Bayesian network model, the random walk algorithm, and the attention mechanism model;

[0012] Input the traceability knowledge graph into the weighted path sorting algorithm, execute the weighted path sorting algorithm according to the weighted fault causal chain, output the score of each path, and sort the paths according to the scores of the paths;

[0013] The highest ranked path or paths are used as the tracing results of the faulty node.

[0014] Preferably, after acquiring the historical fault data, equipment operation data and maintenance records of the railway system, the method further comprises:

[0015] Clean the historical fault data, equipment operation data and maintenance records of the railway system to remove noise and abnormal data;

[0016] Perform entity extraction on the processed data and count the frequency of occurrence of each entity;

[0017] The occurrence probability of each entity is calculated as the prior probability according to the total amount and the occurrence frequency of each entity, and the prior probability is used as the attribute value of the entity.

[0018] Preferably, triple construction is performed, including:

[0019] Extracting entity information and relationship information from the processed data;

[0020] Perform deduplication and synonym mapping on entity information with the same name attributes;

[0021] Knowledge fusion is performed on entity information and relationship information that represent the same meaning.

[0022] Preferably, the method further comprises:

[0023] The traceability knowledge graph is stored and managed through the graph database, supporting the query and update of the traceability knowledge graph.

[0024] Preferably, constructing a subgraph of the traceability knowledge graph as a fault causal chain and extracting a Bayesian network model corresponding to the fault causal chain includes:

[0025] Extracting a subgraph of the traceability knowledge graph as a fault causal chain through Cypher query;

[0026] Construct the Bayesian network structure parameter distribution and learn the Bayesian network structure together with samples;

[0027] A Bayesian network model corresponding to the traceability knowledge graph is constructed through the fault causal chain.

[0028] Preferably, weighting the fault causal chain according to the Bayesian network model, the random walk algorithm and the attention mechanism model includes:

[0029] By using the Bayesian network model, the posterior probability of the fault node in the fault causal chain is calculated as the event importance of the fault node;

[0030] Determine the characteristic values ​​of the faulty nodes in the path through the random walk algorithm;

[0031] The attention mechanism model is used to assign corresponding attention mechanism weights to the fault nodes in the path according to their importance levels.

[0032] The fault causal chain is weighted according to the event importance of the fault node, the node feature value and the attention mechanism weight to obtain the weighted importance of the fault node.

[0033] Preferably, the traceability knowledge graph is input into a weighted path sorting algorithm, and the weighted path sorting algorithm is executed according to the weighted fault causal chain, including:

[0034] Starting from the starting node, execute the graph traversal algorithm to explore all paths in the traceability knowledge graph;

[0035] During the traversal process, all paths from the start node to the target node are generated and collected into a path set;

[0036] According to the predefined evaluation indicators and sorting rules, each path is evaluated and its weight or score is calculated;

[0037] Sort the paths according to their weights or scores;

[0038] Output the sorting results in a list or in a ranking order.

[0039] Preferably, after cleaning the historical fault data, equipment operation data and maintenance records of the railway system to remove noise and abnormal data, the method further comprises:

[0040] The processed data is segmented through the BiLSTM neural network model to generate a fault entity corpus in the railway safety field;

[0041] Annotating the railway safety field fault entity corpus according to predefined entity types, obtaining annotation information, and converting it into BIO format to generate railway safety fault training data;

[0042] The BERT model was used to process the training data of railway safety faults in Siping City and generate word vectors for railway safety faults.

[0043] Process the fault word vectors in the railway safety field through a bidirectional neural network BiLSTM model to extract deep semantic information;

[0044] Processing the deep semantic information through a CRF model to obtain a maximum probability annotation sequence, and outputting a railway safety field fault entity according to the maximum probability annotation sequence;

[0045] Extracting features of the railway safety field fault entity based on the railway safety fault word vector to obtain a feature vector;

[0046] Performing a convolution operation on the feature vector through a convolutional neural network to obtain a convolution result;

[0047] The maximum eigenvalue in the convolution result is extracted through a maximum pooling operation, and the fault entity relationship in the railway safety field is output by a convolutional neural network.

[0048] According to a second aspect of an embodiment of the present application, a railway fault tracing system based on a weighted path sorting algorithm is provided, comprising:

[0049] The acquisition module is used to obtain the historical fault data, equipment operation data and maintenance records of the railway system for triple construction;

[0050] The knowledge graph module is used to store triple information in the graph database and build a traceability knowledge graph;

[0051] A Bayesian network module, used to construct a subgraph of the traceability knowledge graph as a fault causal chain, and extract a Bayesian network model corresponding to the fault causal chain;

[0052] A weighting module, used for weighting the fault causal chain according to the Bayesian network model, the random walk algorithm and the attention mechanism model;

[0053] The sorting module is used to input the traceability knowledge graph into the weighted path sorting algorithm, execute the weighted path sorting algorithm according to the weighted fault causal chain, output the score of each path, and sort the paths according to the scores of the paths;

[0054] The tracing module is used to use the highest ranked path or paths as the tracing result of the faulty node.

[0055] The technical solution provided by this application may have the following beneficial effects:

[0056] This technical solution first constructs a traceability knowledge graph based on the historical fault data, equipment operation data and maintenance records of the railway system, and then combines the Bayesian network model, random walk algorithm and attention mechanism model to perform a weighted path sorting algorithm on the traceability knowledge graph to achieve efficient and accurate railway safety fault tracing, which significantly improves the safety and reliability of railway transportation, and has broad application prospects and significant practical value.

[0057] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0059] Figure 1 It is a flowchart of a railway fault tracing method based on a weighted path sorting algorithm provided by an embodiment of the present application;

[0060] Figure 2 A schematic diagram of data preprocessing of a railway intrinsic safety fault tracing method provided by an embodiment of the present application;

[0061] Figure 3 A schematic diagram of a triplet constructed for a railway intrinsic safety fault tracing method provided in one embodiment of the present application;

[0062] Figure 4 A schematic diagram of constructing a knowledge graph in a railway intrinsic safety fault tracing method provided in one embodiment of the present application;

[0063] Figure 5 A schematic diagram of a fault cause-effect chain in a railway intrinsic safety fault tracing method provided by an embodiment of the present application;

[0064] Figure 6 It is a structural schematic diagram of a railway fault tracing system based on a weighted path sorting algorithm provided by an embodiment of the present application.

[0065] Figure numerals: acquisition module-21; knowledge graph module-22; Bayesian network module-23; weighting module-24; sorting module-25; tracing module-26. DETAILED DESCRIPTION

[0066] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0067] Embodiment 1

[0068] Figure 1 is a flow chart of a railway fault tracing method based on a weighted path sorting algorithm provided by an embodiment of the present application, with reference to Figure 1, a railway fault tracing method based on weighted path sorting algorithm, comprising:

[0069] S11: Obtain historical fault data, equipment operation data and maintenance records of the railway system to construct triples;

[0070] S12: Store the triple information in the graph database and construct a traceability knowledge graph;

[0071] S13: Construct a subgraph of the traceability knowledge graph as the fault causal chain and extract the Bayesian network model corresponding to the fault causal chain;

[0072] S14: Weighting the fault causal chain based on the Bayesian network model, random walk algorithm and attention mechanism model;

[0073] S15: Input the traceability knowledge graph into the weighted path sorting algorithm, execute the weighted path sorting algorithm according to the weighted fault causal chain, output the score of each path, and sort the paths according to the scores of the paths;

[0074] S16: Use the highest ranked path or paths as the tracing result of the faulty node.

[0075] It should be noted that after obtaining the historical fault data, equipment operation data and maintenance records of the railway system, the method further includes:

[0076] Clean the historical fault data, equipment operation data and maintenance records of the railway system to remove noise and abnormal data;

[0077] Perform entity extraction on the processed data and count the frequency of occurrence of each entity;

[0078] The occurrence probability of each entity is calculated as the prior probability according to the total amount and the occurrence frequency of each entity, and the prior probability is used as the attribute value of the entity.

[0079] In this technical solution, after collecting historical fault data, equipment operation data, maintenance records and other data of the railway system, the data is cleaned and preprocessed to remove noise and outliers to ensure data quality. Figure 2 shown.

[0080] In specific practice, after cleaning the historical fault data, equipment operation data and maintenance records of the railway system to remove noise and abnormal data, the method further includes:

[0081] The processed data is segmented through the BiLSTM neural network model to generate a fault entity corpus in the railway safety field;

[0082] Annotating the railway safety field fault entity corpus according to predefined entity types, obtaining annotation information, and converting it into BIO format to generate railway safety fault training data;

[0083] The BERT model was used to process the training data of railway safety faults in Siping City and generate word vectors for railway safety faults.

[0084] Process the fault word vectors in the railway safety field through a bidirectional neural network BiLSTM model to extract deep semantic information;

[0085] Processing the deep semantic information through a CRF model to obtain a maximum probability annotation sequence, and outputting a railway safety field fault entity according to the maximum probability annotation sequence;

[0086] Extracting features of the railway safety field fault entity based on the railway safety fault word vector to obtain a feature vector;

[0087] Performing a convolution operation on the feature vector through a convolutional neural network to obtain a convolution result;

[0088] The maximum eigenvalue in the convolution result is extracted through a maximum pooling operation, and the fault entity relationship in the railway safety field is output by a convolutional neural network.

[0089] Finally, the frequency of each entity in the overall data is counted; the probability of each entity is calculated as the prior probability based on the frequency corresponding to the total amount, and the prior probability is used as the attribute value of the entity.

[0090] It should be noted that the construction of triples includes:

[0091] Extracting entity information and relationship information from the processed data;

[0092] Perform deduplication and synonym mapping on entity information with the same name attributes;

[0093] Knowledge fusion is performed on entity information and relationship information that represent the same meaning.

[0094] For the processed text data, triples are constructed according to the entity information and relationship information of the fault event, and the entity information and relationship information with the same meaning are fused through expert experience. In the embodiment of the present invention, the constructed triples are as follows: Figure 3 shown.

[0095] The extracted entity words and corresponding verbs are stored as entity triples in the traceability knowledge graph according to preset semantic relationships, wherein the fault nodes include nodes such as fault type, fault phenomenon, and fault cause, and the various relationships between the nodes include various relationships such as belonging to relationships and causing relationships.

[0096] According to the literature, network data query and expert experience, a synonym mapping word library is constructed. If various entities such as "no indication of turnout reverse position", "no indication of turnout", "no indication of turnout positioning", "no indication of turnout idling", "no indication of turnout loss", "no indication of turnout reverse position to positioning", "no indication of turnout positioning to empty position idling", "no indication of turnout fixed reverse position", "no indication of turnout card gap", "no indication of reverse position", "no indication of positioning" and so on appear in the entity, they are all replaced with "no indication of turnout" entity;

[0097] Entities with the same name attributes are deduplicated, and when multiple entity words have the same original meaning but different names, the multiple entity words are synonym mapped, the multiple entity words are normalized, and the normalized multiple entity words are merged.

[0098] Based on the preprocessed entity information and key information, a traceability knowledge graph in the field of railway safety failures is constructed.

[0099] It should be noted that the method also includes:

[0100] The traceability knowledge graph is stored and managed through the graph database, supporting the query and update of the traceability knowledge graph.

[0101] In the embodiment of the present invention, Neo4j graph database is selected as the platform to construct the traceability knowledge graph of the embodiment. Figure 4 shown.

[0102] It should be noted that the subgraph of the traceability knowledge graph is constructed as the fault causal chain, and the Bayesian network model corresponding to the fault causal chain is extracted, including:

[0103] Extract subgraphs of the traceability knowledge graph as fault causal chains through Cypher queries;

[0104] Construct the Bayesian network structure parameter distribution and learn the Bayesian network structure together with samples;

[0105] A Bayesian network model corresponding to the traceability knowledge graph is constructed through the fault causal chain.

[0106] In an embodiment of the present invention, the fault causal chain is extracted as follows: Figure 5 shown.

[0107] 1) According to the acquired fault tracing report information, determine the specific information of the fault node, the fault type and the fault cause; for example, in the tracing report information, obtain the keyword "switch has no indication" through word segmentation.

[0108] 2) Extract the subgraph structure (fault causal chain) through Cypher query on the traceability knowledge graph, that is, all the paths that lead to the fault node, directly construct the Bayesian network structure parameter distribution, and then combine with samples to jointly learn the Bayesian network structure.

[0109] 3) Map the node set of each fault causal chain in the traceability knowledge graph to the node set in the Bayesian network. For example, in the example, the first one is [“low temperature”, “shrinkage of point rail”, “displacement of indicating rod”, “no indication of turnout”], the second one is [“low temperature”, “smaller gap”, “no indication of turnout”], and the third one is [“screw fracture”, “no indication of turnout”]. Map the node sets of the above fault causal chains to the node sets of the corresponding Bayesian network;

[0110] 4) determining the conditional probability information of each node in the Bayesian network based on the fault cause node attributes of the fault causal chain in the railway intrinsic safety fault tracing knowledge graph, and taking each node attribute - probability as a priori probability;

[0111] 5) Mapping the relationship between the fault cause nodes of the fault causal chain in the railway intrinsic safety fault tracing knowledge graph into a directed edge set of the Bayesian network, such as the first item in the example [low temperature causes the point rail to shrink, the point rail shrinkage causes the displacement of the indicating rod, the displacement of the indicating rod causes the turnout to have no representation], the second item [low temperature causes the gap to become smaller, the gap becomes smaller, this path causes the turnout to have no representation], and the third item [screw fracture causes the turnout to have no representation]. The above-mentioned edge relationships are mapped to the Bayesian network, and the former points to the latter;

[0112] 6) Constructing a Bayesian network model corresponding to the fault causal chain according to the determined node set in the Bayesian network, conditional probability information of each node in the Bayesian network, and the directed edge set of the Bayesian network.

[0113] 7) Use historical data to learn the parameters of the Bayesian network and determine the conditional probability distribution of each node in the fault causal chain.

[0114] It should be noted that the fault causal chain is weighted according to the Bayesian network model, random walk algorithm and attention mechanism model, including:

[0115] Through the Bayesian network model, the posterior probability of the fault node in the fault causal chain is calculated as the event importance of the fault node;

[0116] Determine the characteristic values ​​of the faulty nodes in the path through the random walk algorithm;

[0117] The attention mechanism model is used to assign corresponding attention mechanism weights to the fault nodes in the path according to their importance levels.

[0118] The fault causal chain is weighted according to the event importance of the fault node, the node feature value and the attention mechanism weight to obtain the weighted importance of the fault node.

[0119] 1) According to the Bayesian network model, the posterior probability of each fault node in the fault causal chain obtained by inference is used as the event importance of the node, denoted as p i ;

[0120] 2) According to the fault causal chain, the obtained fault causal chain is processed by the word2vec algorithm, and the nodes and edges in the path sequence of the fault causal chain are represented as vectors. For example, for the path sequence ["low temperature", "point rail shrinkage", "displacement of the representation rod", "no representation of the turnout"], its corresponding word vector representation is generated;

[0121] 3) Use the attention mechanism model to assign higher weights to key nodes in the path sequence, such as "representing rod displacement", and record it as w i ;

[0122] 4) The characteristic value of the calculated path is generated by recording the probability value in the process of random walk. The random walk algorithm is used to select the starting node: a starting node is selected from the graph as the starting point of the walk. A position is randomly selected (recorded as time i ) and at the next moment (time i+1 ) passes through the next node, and the probability time i+1 =time i + propability, or starting from the next node at this moment, the probability is time i+1 =time i -propability, move according to the movement rules, move from the current node to the next node, and repeat this step until the specified termination condition is reached, record the walking path or sampling results: record the passed nodes in each step. Get the eigenvalue of the path, and then generate the corresponding eigenvalue set. Each node eigenvalue is recorded as q i .

[0123] 5) The obtained node feature value, attention mechanism weight and event importance are used to weight the fault causal chain, and the weighted node importance is Among them, a, b, and c are weighted factors of the three. In this example, the values ​​of a, b, and c are 0.3, 0.2, and 0.5 respectively. According to the importance of the event, additional weight adjustments are made to the nodes in the path, and the weighted path sorting algorithm is performed on the queried fault causal chain.

[0124] It should be noted that the traceability knowledge graph is input into the weighted path sorting algorithm, and the weighted path sorting algorithm is executed according to the weighted fault causal chain, including:

[0125] Starting from the starting node, execute the graph traversal algorithm to explore all paths in the traceability knowledge graph;

[0126] During the traversal process, all paths from the start node to the target node are generated and collected into a path set;

[0127] According to the predefined evaluation indicators and sorting rules, each path is evaluated and its weight or score is calculated;

[0128] Sort the paths according to their weights or scores;

[0129] Output the sorting results in a list or in a ranking order.

[0130] The weighted path sorting algorithm aims to find the best path or ordered path combination to meet specific requirements and optimization goals. The input of the algorithm includes a graph structure in a knowledge graph or graph database, as well as a start node and a target node. The output is a sorted path set, which contains all paths from the start node to the target node, and is sorted according to certain sorting rules. In an embodiment of the present invention:

[0131] 1) Graph traversal: Starting from the starting node, use a suitable graph traversal algorithm to explore all possible paths.

[0132] 2) Construct a path set. During the traversal process, all paths from the start node to the target node are generated and collected into a path set.

[0133] 3) Path evaluation: Each path is evaluated and its weight or score is calculated according to predefined evaluation indicators and sorting rules. Evaluation indicators can include path length, weight of association relationship, similarity of node attributes, etc.

[0134] 4) Path sorting: sorting the paths according to their weights or scores to determine their importance or relevance. Generally, paths with higher scores are considered more important or relevant.

[0135] 5) Output the results. The sorted path results are output, usually in the form of a list or in the form of importance sorting.

[0136] It should be noted that, in the embodiment of the present invention, the source tracing analysis includes:

[0137] 1) Obtain path scores and sort paths according to the weighted path sorting algorithm to determine the importance and relevance of the paths;

[0138] 2) Based on the results of path sorting, the top three paths with the highest probability are selected as the tracing results. For example, the score report of the first path ["low temperature", "shrinkage of the point rail", "displacement of the indicator rod", "no indicator on the turnout"] is "the probability that this path leads to fault tracing is 91.09%", the score report of the second path ["screw fracture", "no indicator on the turnout"] is "the probability that this path leads to fault tracing is 78.34%", and the score report of the third path ["low temperature", "smaller gap", "no indicator on the turnout"] is "the probability that this path leads to fault tracing is 69.52%"

[0139] This technical solution first constructs a traceability knowledge graph based on the historical fault data, equipment operation data and maintenance records of the railway system, and then combines the Bayesian network model, random walk algorithm and attention mechanism model to perform a weighted path sorting algorithm on the traceability knowledge graph to achieve efficient and accurate railway safety fault tracing, which significantly improves the safety and reliability of railway transportation, and has broad application prospects and significant practical value.

[0140] Embodiment 2

[0141] A railway fault tracing system based on weighted path sorting algorithm, referring to Figure 6 ,include:

[0142] An acquisition module 21 is used to acquire historical fault data, equipment operation data and maintenance records of the railway system for triple construction;

[0143] The knowledge graph module 22 is used to store triple information in a graph database and construct a traceability knowledge graph;

[0144] The Bayesian network module 23 is used to construct a subgraph of the traceability knowledge graph as a fault causal chain and extract the Bayesian network model corresponding to the fault causal chain;

[0145] A weighting module 24, used for weighting the fault causal chain according to the Bayesian network model, the random walk algorithm and the attention mechanism model;

[0146] The sorting module 25 is used to input the traceability knowledge graph into the weighted path sorting algorithm, execute the weighted path sorting algorithm according to the weighted fault causal chain, output the score of each path, and sort the paths according to the scores of the paths;

[0147] The source tracing module 26 is used to use the highest ranked path or paths as the source tracing result of the faulty node.

[0148] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0149] It should be noted that, in the description of this application, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of this application, unless otherwise specified, the meaning of "plurality" refers to at least two.

[0150] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0151] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0152] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0153] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0154] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0155] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0156] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A railway fault tracing method based on a weighted path sorting algorithm, characterized in that: include: Obtain the historical fault data, equipment operation data and maintenance records of the railway system to construct triples; Store triple information in a graph database and build a traceability knowledge graph; Constructing a subgraph of the traceability knowledge graph as a fault causal chain, and extracting a Bayesian network model corresponding to the fault causal chain; Weighting the fault causal chain according to the Bayesian network model, the random walk algorithm, and the attention mechanism model; Input the traceability knowledge graph into the weighted path sorting algorithm, execute the weighted path sorting algorithm according to the weighted fault causal chain, output the score of each path, and sort the paths according to the scores of the paths; The highest ranked path or paths are used as the tracing results of the faulty node.

2. The method according to claim 1, characterized in that After obtaining the historical fault data, equipment operation data and maintenance records of the railway system, the method further includes: Clean the historical fault data, equipment operation data and maintenance records of the railway system to remove noise and abnormal data; Perform entity extraction on the processed data and count the frequency of occurrence of each entity; The occurrence probability of each entity is calculated as the prior probability according to the total amount and the occurrence frequency of each entity, and the prior probability is used as the attribute value of the entity.

3. The method according to claim 2, characterized in that Perform triple construction, including: Extracting entity information and relationship information from the processed data; Perform deduplication and synonym mapping on entity information with the same name attributes; Knowledge fusion is performed on entity information and relationship information that represent the same meaning.

4. The method according to claim 1, characterized in that: The method further comprises: The traceability knowledge graph is stored and managed through the graph database, supporting the query and update of the traceability knowledge graph.

5. The method according to claim 1, characterized in that Constructing a subgraph of the traceability knowledge graph as a fault causal chain, and extracting a Bayesian network model corresponding to the fault causal chain, including: Extracting a subgraph of the traceability knowledge graph as a fault causal chain through Cypher query; Construct the Bayesian network structure parameter distribution and learn the Bayesian network structure together with samples; A Bayesian network model corresponding to the traceability knowledge graph is constructed through the fault causal chain.

6. The method according to claim 1, characterized in that The fault causal chain is weighted according to the Bayesian network model, the random walk algorithm and the attention mechanism model, including: By using the Bayesian network model, the posterior probability of the fault node in the fault causal chain is calculated as the event importance of the fault node; Determine the characteristic values ​​of the faulty nodes in the path through the random walk algorithm; The attention mechanism model is used to assign corresponding attention mechanism weights to the fault nodes in the path according to their importance levels. The fault causal chain is weighted according to the event importance of the fault node, the node feature value and the attention mechanism weight to obtain the weighted importance of the fault node.

7. The method according to claim 1, characterized in that Input the traceability knowledge graph into the weighted path sorting algorithm, and execute the weighted path sorting algorithm according to the weighted fault causal chain, including: Starting from the starting node, execute the graph traversal algorithm to explore all paths in the traceability knowledge graph; During the traversal process, all paths from the start node to the target node are generated and collected into a path set; According to the predefined evaluation indicators and sorting rules, each path is evaluated and its weight or score is calculated; Sort the paths according to their weights or scores; Output the sorting results in a list or in a ranking order.

8. The method according to claim 2, characterized in that: After cleaning the historical fault data, equipment operation data and maintenance records of the railway system to remove noise and abnormal data, the method further includes: The processed data is segmented through the BiLSTM neural network model to generate a fault entity corpus in the railway safety field; Annotating the railway safety field fault entity corpus according to predefined entity types, obtaining annotation information, and converting it into BIO format to generate railway safety fault training data; The BERT model was used to process the training data of railway safety faults in Siping City and generate word vectors for railway safety faults. Process the fault word vectors in the railway safety field through a bidirectional neural network BiLSTM model to extract deep semantic information; Processing the deep semantic information through a CRF model to obtain a maximum probability annotation sequence, and outputting a railway safety field fault entity according to the maximum probability annotation sequence; Extracting features of the railway safety field fault entity based on the railway safety fault word vector to obtain a feature vector; Performing a convolution operation on the feature vector through a convolutional neural network to obtain a convolution result; The maximum eigenvalue in the convolution result is extracted through a maximum pooling operation, and the fault entity relationship in the railway safety field is output by a convolutional neural network.

9. A railway fault tracing system based on a weighted path sorting algorithm, characterized in that: include: The acquisition module is used to obtain the historical fault data, equipment operation data and maintenance records of the railway system for triple construction; The knowledge graph module is used to store triple information in the graph database and build a traceability knowledge graph; A Bayesian network module, used to construct a subgraph of the traceability knowledge graph as a fault causal chain, and extract a Bayesian network model corresponding to the fault causal chain; A weighting module, used for weighting the fault causal chain according to the Bayesian network model, the random walk algorithm and the attention mechanism model; The sorting module is used to input the traceability knowledge graph into the weighted path sorting algorithm, execute the weighted path sorting algorithm according to the weighted fault causal chain, output the score of each path, and sort the paths according to the scores of the paths; The tracing module is used to use the highest ranked path or paths as the tracing result of the faulty node.

Images