A Method and System for Constructing an Ontology-Based Knowledge Graph for Active Safety in Urban Rail Transit Operation

Abstract

This invention provides a method and system for constructing an ontology-based knowledge graph for proactive safety in urban rail transit operations, belonging to the field of urban planning, transportation operation, and maintenance technology. The method involves identifying entities and defining the types and categories of relationships between them; constructing an ontology model for the proactive safety knowledge graph of urban rail transit operations based on the identified entities and the types and categories of relationships between them; preprocessing text data from typical accident scenarios based on the ontology model, and performing topic mining to obtain key risk information; extracting knowledge triples from typical accident scenarios using dependency parsing; and importing the knowledge graph into a graph database using <entity, relation, entity> as its standard tuple structure to complete the construction of the proactive safety knowledge graph for urban rail transit operations oriented towards typical accident scenarios. The method for constructing a proactive safety knowledge graph for urban rail transit operations provided by this invention can support better organization and implementation of urban rail transit operation safety management and proactive safety decision-making.

Description

Technical Field

[0001] This invention relates to the field of rail transit operation management technology, specifically to an ontology-based method and system for constructing an active safety knowledge graph for urban rail transit operations. Background Technology

[0002] Urban rail transit systems are typically large-scale, with complex structures, functions, and operating conditions. They involve strong coupling between different parts and a high density of personnel and equipment. Their operational safety encompasses human factors, as well as equipment, facilities, mobile machinery, the environment, and management. Furthermore, the large-scale, networked operation of rail transit systems in many large and medium-sized cities today places even higher demands on safe operation.

[0003] Currently, urban rail transit operators largely adopt a "passive" safety assurance model, primarily based on "experience-driven and emergency response," which makes it difficult to achieve an efficient closed-loop and information-based proactive safety management system. To realize a proactive safety control model of "risk management and advance prevention," higher demands are placed on urban rail transit operators to make proactive and effective safety decisions.

[0004] Knowledge graphs and their construction are emerging technologies in recent years. With the development of the times, knowledge graphs have been gradually developed by various industries and are widely used in education, power grids, finance, intelligent manufacturing, healthcare, law, humanities, and scientific literature. They not only provide rich knowledge sources but also offer diverse functions, playing a significant role in promoting the transformation and upgrading of various industries and related enterprises and providing technological support. Essentially, a knowledge graph is a structured semantic knowledge base. It uses symbols to express external concepts and the various relationships between them. Its basic building blocks are the "entity-relationship-entity" triple, along with entities and their related attributes. Entities are linked together through relationships, forming a network-like knowledge structure. An entity is a semantic ontology in a knowledge graph, sometimes also called an object or instance; attributes describe a class of entities or a certain characteristic of an entity; and relationships are the semantic ontology's connections, used to link entities. Building a knowledge graph is a systematic project, closely related to intelligent technologies such as databases, information retrieval, natural language processing, visualization analysis, and machine learning.

[0005] Domain-specific knowledge graphs refer to knowledge graphs focused on specific sub-fields, such as those in medicine, education, and finance. They involve more domain-specific knowledge, exhibiting greater depth and sophistication in knowledge representation. In terms of knowledge application, the density of specialized knowledge results in longer and more complex reasoning chains. Constructing a domain-specific knowledge graph requires not only drawing upon general knowledge graph construction methods but also designing targeted patterns, utilizing natural language processing techniques to extract reliable industry information, and integrating this information to ultimately form a usable knowledge graph.

[0006] In the field of urban rail transit system operation management and safety assurance, a complete knowledge graph has not yet been formed to support the realization of proactive safety control models. Summary of the Invention

[0007] The purpose of this invention is to provide a method and system for constructing an ontology-based knowledge graph for proactive safety in urban rail transit operations, in order to solve at least one of the technical problems existing in the background art.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] On the one hand, this invention provides a method for constructing an ontology-based knowledge graph for proactive safety in urban rail transit operations, including:

[0010] Identify entities and define the types and categories of relationships between them; entities include risk points, risk events, feedforward signals, incidents, and measures.

[0011] Based on the identified entities, the types and categories of relationships between entities, a knowledge graph ontology model for proactive safety in urban rail transit operations is constructed.

[0012] Based on the knowledge graph ontology model of proactive safety in urban rail transit operations, text data preprocessing of typical accident scenarios is performed, and key risk information is obtained by topic mining of the text.

[0013] Knowledge triples for typical accident scenarios are extracted based on dependency parsing.

[0014] Using <entity, relation, entity> as its standard tuple structure, a graph database was imported to complete the construction of an active safety knowledge graph for urban rail transit operations oriented towards typical accident scenarios.

[0015] Preferably, the constructed urban rail transit operation proactive safety knowledge graph network is divided into a four-layer chain network structure: risk point layer, risk event layer, feedforward signal layer, and measure layer.

[0016] Based on the external relationships and network structure of risk points, and taking into account the degree of participation, role, and importance of the points, three weight indicators are established to represent the relevant weights of risk points to other levels.

[0017] Based on the theory of set pair analysis, we establish the correlation degree between risk points and indicators of the entire knowledge graph, quantify the correlation degree of a certain risk point to different layers of the knowledge graph, and then determine the key connections.

[0018] Preferably, the relationships between entities include: risk point-risk event, risk event-risk point, risk point-feedforward signal, and feedforward signal-accident and action-risk point.

[0019] Preferably, a knowledge graph ontology model for proactive safety in urban rail transit operations is constructed, including:

[0020] The scope of the ontology for urban rail transit operation safety is defined as all relevant safety risk factors that lead to urban rail operation accidents; and important terms in the ontology are determined based on the reusability of existing ontologs.

[0021] Risk points are categorized into personnel, equipment and facilities, management, and environment. The attributes of each category refer to the level and type of risk points in the operation of the urban rail system, the information attributes of accidents, and the information attributes of measures.

[0022] Preferably, based on the knowledge graph ontology model of urban rail transit operation proactive safety, text data preprocessing of typical accident scenarios is performed, and topic mining is conducted on the text to obtain key risk information, including:

[0023] Accident text data preprocessing: For a typical accident scenario, a text set of accident categories is selected as the object of analysis and processing, and data processing such as word segmentation and sentence breaking is performed; words with the same meaning in the text are grouped under the same professional concept to complete the concept disambiguation;

[0024] Risk information theme mining for typical accident scenarios: For urban rail operation accident texts, a word and probability extraction method based on the BTM topic model is used; by applying the topic model, key knowledge in the accident text is obtained, providing some basis for the extraction of risk points and triples. The potential risk patterns contained in all information of a certain type of typical accident are abstracted into different text themes. The text themes with higher probabilities in the information reflect the main risk patterns in this type of unsafe event, and the types of feedforward signals are summarized.

[0025] Preferably, the knowledge triples of typical accident scenarios are extracted semi-automatically based on the dependency parsing of accident texts, and knowledge triples centered on predicates are extracted based on the results of dependency parsing.

[0026] Preferably, the standard tuple structure <entity, relation, entity> is used to import the knowledge into a graph database to complete the construction of an urban rail transit operation proactive safety knowledge graph for typical accident scenarios. This includes: using Neo4j to store the graph knowledge, using <entity, relation, entity> as its standard tuple structure, and using Python to import the tuple structure to complete the construction of an urban rail transit operation proactive safety knowledge graph for typical accident scenarios.

[0027] Preferably, the constructed knowledge graph network displays the content structure and interrelationships between entities, consisting of a four-layer chain network structure: risk point layer P, risk event layer E, feedforward signal layer F, and measure layer M. The risk points in the risk point layer are the core of the entire knowledge graph network. Based on the external relationships and network structure of the risk points, and comprehensively considering the degree of participation, role, and importance of the points, three weight indicators are established to represent the correlation between risk points and other layers: A1(x) represents the correlation weight between risk point x and the risk event layer, A2(x) represents the correlation weight between risk point x and the feedforward signal layer, and A3(x) represents the correlation weight between risk point x and the measure layer. The correlation degree between risk points and the entire knowledge graph is established to quantify the correlation degree of a certain risk point x with different layers of the knowledge graph, thereby identifying key connections.

[0028] Secondly, this invention provides an ontology-based knowledge graph construction system for proactive safety in urban rail transit operations, comprising:

[0029] The identification module is used to identify entities and define the types and relationships between entities; entities include risk points, risk events, feedforward signals, incidents, and measures.

[0030] The construction module is used to build an ontology model of the active safety knowledge graph of urban rail transit operation based on the determined entities, the types of relationships between entities, and the relationship types.

[0031] The preprocessing module is used to preprocess text data of typical accident scenarios based on the knowledge graph ontology model of urban rail operation active safety, and to perform topic mining on the text to obtain key risk information.

[0032] The extraction module is used to extract knowledge triples from typical accident scenarios based on dependency parsing.

[0033] The import module is used to import graph databases with <entity, relation, entity> as its standard tuple structure, and complete the construction of an active safety knowledge graph for urban rail transit operation oriented towards typical accident scenarios.

[0034] Thirdly, the present invention provides a non-transitory computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the simulated human driving evidence dynamic accumulation braking method for human-machine hybrid driving safety interaction as described above.

[0035] Fourthly, the present invention provides a computer program product, including a computer program that, when run on one or more processors, is used to implement the simulated human driving evidence dynamic cumulative braking method for human-machine hybrid driving safety interaction as described above.

[0036] Fifthly, the present invention provides an electronic device, comprising: a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing the simulated human-machine hybrid driving safety interaction dynamic cumulative braking method as described above.

[0037] The beneficial effects of this invention are as follows: The entire knowledge graph network is divided into a four-layer chain structure: a risk point layer, a risk event layer, a feedforward signal layer (connected to accident entities), and a measure layer. Based on the external relationships and network structure of risk points, and comprehensively considering the participation, role, and importance of points, three weighted indicators are established to represent the relevance of risk points to other layers. Using set pair analysis theory, the correlation degree between risk points and the entire knowledge graph is established, quantifying the correlation degree (trend degree) of a risk point to different layers of the knowledge graph. This allows for the determination of which connections are more important and worthy of attention, providing support for better organization and implementation of urban rail transit operation safety management and proactive safety decision-making.

[0038] The advantages of additional aspects of the invention will be set forth more clearly in the following description or will be learned by practice of the invention. Attached Figure Description

[0039] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart illustrating the knowledge graph construction process described in an embodiment of the present invention.

[0041] Figure 2 This is a schematic diagram of the urban rail transit operation proactive safety knowledge graph ontology model according to an embodiment of the present invention.

[0042] Figure 3 This is a schematic diagram of the graphical model representation of BTM according to an embodiment of the present invention.

[0043] Figure 4 This is a schematic diagram of the perplexity curve according to an embodiment of the present invention.

[0044] Figure 5 This is a fire scenario knowledge graph for proactive safety in urban rail transit operations, as described in an embodiment of the present invention.

[0045] Figure 6 This is a schematic diagram illustrating the hierarchical relationship of the knowledge graph as described in an embodiment of the present invention.

[0046] Figure 7 The query results are for the risk point "cable" described in the embodiments of the present invention.

[0047] Figure 8 This is a link query diagram for the risk point "cable" described in an embodiment of the present invention. Detailed Implementation

[0048] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0049] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0050] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, and / or groups thereof.

[0051] To facilitate understanding of the present invention, the present invention will be further explained and described below with reference to the accompanying drawings and specific embodiments. However, the specific embodiments do not constitute a limitation on the embodiments of the present invention.

[0052] Those skilled in the art should understand that the accompanying drawings are merely schematic diagrams of embodiments, and the components in the drawings are not necessarily essential for implementing the present invention.

[0053] Example 1

[0054] In this embodiment 1, an ontology-based knowledge graph construction system for proactive safety in urban rail transit operation is first provided. This system includes: a determination module for determining entities and defining the types and categories of relationships between entities; entities include risk points, risk events, feedforward signals, accidents, and measures; a construction module for constructing an ontology model of the proactive safety knowledge graph for urban rail transit operation based on the determined entities and the types and categories of relationships between entities; a preprocessing module for performing text data preprocessing of typical accident scenarios based on the ontology model of the proactive safety knowledge graph for urban rail transit operation, and performing topic mining to obtain key risk information from the text; an extraction module for extracting knowledge triples from typical accident scenarios based on dependency parsing; and an import module for importing data into a graph database with <entity, relation, entity> as its standard tuple structure, thus completing the construction of the proactive safety knowledge graph for urban rail transit operation oriented towards typical accident scenarios.

[0055] In this embodiment, the above-described system is used to implement an ontology-based method for constructing an active safety knowledge graph for urban rail transit operations. The method includes: using a determination module to determine entities and define the types and categories of relationships between entities; entities include risk points, risk events, feedforward signals, accidents, and measures; using a construction module to construct an ontology model of the active safety knowledge graph for urban rail transit operations based on the determined entities and the types and categories of relationships between entities; using a preprocessing module to preprocess text data of typical accident scenarios based on the ontology model of the active safety knowledge graph for urban rail transit operations, and performing topic mining to obtain key risk information from the text; using an extraction module to extract knowledge triples of typical accident scenarios based on dependency parsing; and using an import module to import the graph database with <entity, relation, entity> as its standard tuple structure, thus completing the construction of the active safety knowledge graph for urban rail transit operations oriented towards typical accident scenarios.

[0056] It also includes dividing the constructed urban rail operation proactive safety knowledge graph network into a four-layer chain network structure: risk point layer, risk event layer, feedforward signal layer, and measure layer;

[0057] Based on the external relationships and network structure of risk points, and taking into account the degree of participation, role, and importance of the points, three weight indicators are established to represent the relevant weights of risk points to other levels.

[0058] Based on the theory of set pair analysis, we establish the correlation degree between risk points and indicators of the entire knowledge graph, quantify the correlation degree of a certain risk point to different layers of the knowledge graph, and then determine the key connections.

[0059] The relationships between entities include: risk point-risk event, risk event-risk point, risk point-feedforward signal, and feedforward signal-accident and action-risk point.

[0060] Constructing a knowledge graph ontology model for proactive safety in urban rail transit operations, including:

[0061] The scope of the ontology for urban rail transit operation safety is defined as all relevant safety risk factors that lead to urban rail operation accidents; and important terms in the ontology are determined based on the reusability of existing ontologs.

[0062] Risk points are categorized into personnel, equipment and facilities, management, and environment. The attributes of each category refer to the level and type of risk points in the operation of the urban rail system, the information attributes of accidents, and the information attributes of measures.

[0063] Based on the knowledge graph ontology model of proactive safety in urban rail transit operations, text data preprocessing of typical accident scenarios is performed, and topic mining is conducted on the text to obtain key risk information, including:

[0064] Accident text data preprocessing: For a typical accident scenario, a text set of accident categories is selected as the object of analysis and processing, and data processing such as word segmentation and sentence breaking is performed; words with the same meaning in the text are grouped under the same professional concept to complete the concept disambiguation;

[0065] Risk information theme mining for typical accident scenarios: For urban rail operation accident texts, a word and probability extraction method based on the BTM topic model is used; by applying the topic model, key knowledge in the accident text is obtained, providing some basis for the extraction of risk points and triples. The potential risk patterns contained in all information of a certain type of typical accident are abstracted into different text themes. The text themes with higher probabilities in the information reflect the main risk patterns in this type of unsafe event, and the types of feedforward signals are summarized.

[0066] Dependency parsing based on accident texts is used to semi-automatically extract knowledge triples for typical accident scenarios. Based on the results of dependency parsing, knowledge triples centered on predicates are extracted.

[0067] Using <entity, relation, entity> as its standard tuple structure, the graph database is imported to complete the construction of an active safety knowledge graph for urban rail transit operations oriented towards typical accident scenarios. This includes: using Neo4j to store the graph knowledge, using <entity, relation, entity> as its standard tuple structure, and using Python to import the tuple structure to complete the construction of an active safety knowledge graph for urban rail transit operations oriented towards typical accident scenarios.

[0068] The constructed knowledge graph network displays the content structure and interrelationships between entities, consisting of a four-layer chain structure: risk point layer P, risk event layer E, feedforward signal layer F, and measure layer M. The risk points in the risk point layer are the core of the entire knowledge graph network. Based on the external relationships and network structure of the risk points, and considering the degree of participation, role, and importance of the points, three weight indicators are established to represent the correlation between risk points and other layers: A1(x) represents the correlation weight between risk point x and the risk event layer, A2(x) represents the correlation weight between risk point x and the feedforward signal layer, and A3(x) represents the correlation weight between risk point x and the measure layer. The evaluation objectives and evaluation criteria are formulated as a set pair. Based on the extensibility theory of set pair analysis, the correlation degree between risk points and the indicators of the entire knowledge graph is established, quantifying the correlation degree of a certain risk point x with different layers of the knowledge graph, thereby identifying key connections.

[0069] Example 2

[0070] In this embodiment 2, in order to promote the informatization and proactive safety decision-making of the current urban rail transit system, an ontology-based method for constructing a proactive safety knowledge graph for urban rail transit operation is proposed, which includes three steps: domain-oriented pattern and ontology construction, domain knowledge extraction, and domain graph construction. It involves several key contents such as knowledge expression and model building, knowledge acquisition, knowledge storage, retrieval, and reasoning.

[0071] The ontology-based method for constructing a knowledge graph for proactive safety in urban rail transit operations includes the following steps:

[0072] Step 1: Determine the research field—urban rail transit operation safety; identify the entities within the field—five categories of entities: risk points, risk events, feedforward signals, accidents, and measures; and define the types and relationships between entities.

[0073] Step 2: Construct an ontology model of proactive safety knowledge graph for urban rail transit operations using a seven-step method;

[0074] Step 3: Perform text data preprocessing for typical accident scenarios, and conduct topic mining to obtain key risk information from the text;

[0075] Step 4: Extract knowledge triples from typical accident scenarios based on dependency parsing, and consult industry experts after extraction is completed;

[0076] Step 5: Import the graph database using <entity, relation, entity> as its standard tuple structure to complete the construction of the urban rail transit operation proactive safety knowledge graph for typical accident scenarios. The entire knowledge graph network is divided into a four-layer chain network structure: risk point layer, risk event layer, feedforward signal layer (this layer is connected to the accident entity), and measure layer.

[0077] Step 6: Based on the external relationships and network structure of risk points, and comprehensively considering the degree of participation, role, and importance of each point, establish three weighted indicators to represent the relevance of risk points to other levels. Using set pair analysis, establish the correlation between risk points and the entire knowledge graph, quantify the correlation (trend) of a risk point x to different layers of the knowledge graph, and then determine which connections are more important and worthy of attention.

[0078] Step 1 specifically involves:

[0079] Five types of entities are introduced: risk points, risk events, feedforward signals, accidents, and measures. Based on the introduction of entity types, and with the core concept of risk points generating events, the types of relationships between entities are defined as follows: risk point-risk event, risk event-risk point, risk point-feedforward signal, feedforward signal-accident, and measures-risk point.

[0080] Step 2 specifically involves:

[0081] We will use the relatively mature seven-step method to construct an ontology model of proactive safety knowledge graph for urban rail transit operations.

[0082] ① Determine the subject's professional field and scope.

[0083] This paper studies the field of urban rail transit operation safety; the purpose of the study is to clarify and systematize the knowledge of safety risks in urban rail operation; the scope of the subject is all relevant safety risk factors that may lead to urban rail operation accidents.

[0084] ② Examine the feasibility of reusing existing ontologies

[0085] There is no reusable ontology. However, ontologies from other industries can be used as a reference.

[0086] ③ List the important terms within the domain (ontology).

[0087] ④ Define classes and class hierarchy

[0088] Risk points are categorized into four main types: personnel, equipment and facilities, management, and environment, involving four levels. Accidents are classified into 10 categories, with feedforward signals further refined based on accident category. Measures are categorized into three types: prevention measures, governance measures, and rescue measures. Except for risk points, the categories of other entities do not involve levels.

[0089] ⑤ Define class attributes

[0090] Risk points and accidents are the core components of the ontology concept layer in this field. The attributes of the class mainly refer to the level and category attributes of risk points, the information attributes of accidents, and the information attributes of measures in the operation of urban rail transit systems.

[0091] ⑥ Define the facets of the attribute

[0092] In the field of urban rail transit operation safety, the main focus is on character-based data.

[0093] ⑦ Create an instance

[0094] Construct an ontology model of proactive safety knowledge graph for urban rail transit operations, with the core concept of "what to prevent and how to prevent it".

[0095] Step 3 specifically involves:

[0096] ① Accident text data preprocessing

[0097] For a specific typical accident scenario, a text set of accident categories is selected as the object of analysis and processing. Data processing, including word segmentation and sentence division, is performed. Words in the text expressing the same meaning are grouped under consistent professional concepts to achieve concept disambiguation.

[0098] ② Risk information theme mining of typical accident scenarios

[0099] For urban rail transit operation accident texts, a word and probability extraction method based on the BTM topic model is used. By applying the topic model, key knowledge (keywords) in the accident text can be obtained, providing some basis for the extraction of risk points and triples. Furthermore, the potential risk patterns contained in all information of a typical accident can be abstracted into different text topics. The text topics with higher probabilities reflect several major risk patterns in this type of unsafe event. Through the analysis of these major risk patterns, this paper can summarize the main types of feedforward signals obtained.

[0100] Step 4 specifically involves:

[0101] Dependency parsing based on accident text is used for semi-automatic extraction of knowledge triples from typical accident scenarios. This primarily relies on the results of dependency parsing to extract knowledge triples centered around predicates (which trigger the entire triple, typically verbs). After extraction, industry experts are consulted, focusing on entity concepts such as risk points, risk events, and measures, and the relationships between them. All triple content is then manually revised.

[0102] Step 5 specifically involves:

[0103] Neo4j is a popular native graph database system that stores structured data in graphs and visualizes entities and relationships using nodes and edges, possessing all the features of a mature database. This study uses Neo4j to store graph knowledge, employing <entity, relation, entity> as its standard tuple structure. Python is used to import this tuple structure to construct a knowledge graph for proactive safety in urban rail transit operations, oriented towards typical accident scenarios. Based on the entities in the constructed knowledge graph, the entire knowledge graph network can be divided into a four-layer chain structure: a risk point layer, a risk event layer, a feedforward signal layer (connected to accident entities), and a measure layer.

[0104] Step 6 specifically involves:

[0105] The constructed knowledge graph network can display the content structure and interrelationships between entities. It consists of a four-layer chain structure: risk point layer P, risk event layer E, feedforward signal layer F, and measure layer M, exhibiting strong complex network characteristics. The risk points in the risk point layer are the core of the entire knowledge graph network. Based on the external relationships and network structure of the risk points, and considering their participation, role, and importance, three weight indicators representing the correlation between risk points and other layers are established: A1(x) – the correlation weight between risk point x and the risk event layer; A2(x) – the correlation weight between risk point x and the feedforward signal layer; and A3(x) – the correlation weight between risk point x and the measure layer. Formulating the review objectives and review criteria into a set pair to objectively reflect their similarities, differences, and oppositions is the basic idea of ​​set pair analysis. To solve the practical problems in this paper, based on the extensibility theory of set pair analysis, the correlation degree between risk points and the entire knowledge graph is established. The correlation degree (trendiness) of a certain risk point x with different layers of the knowledge graph is quantified, thereby determining which connections are more important and worthy of attention.

[0106] Example 3

[0107] like Figure 1 As shown in Embodiment 3, an ontology-based method for constructing an active safety knowledge graph for urban rail transit operations is provided. This method includes:

[0108] Step 1: Introduce five main types of entities: risk points, risk events, feedforward signals, accidents, and measures. Based on the entity types, and with the core concept of risk points generating events, define the types of relationships between entities: risk point-risk event, risk event-risk point, risk point-feedforward signal, feedforward signal-accident, and measures-risk point, as shown in Table 1.

[0109] The International Organization for Standardization (ISO) defines risk in its ISO Guidelines on Risk Management (ISO 31000:2018) as: the impact of uncertainty on objectives. This impact refers to deviations from expectations, which can be positive, negative, or both, and can identify, create, or lead to opportunities and threats. Objectives can be of different aspects and categories and can be applied at different levels. However, this definition has certain limitations. Based on this, a more reasonable definition of risk for urban rail transit systems is provided.

[0110] Definition 1: Risk: Risk is an inherent attribute of things themselves, usually used to describe the combination of the probability of a harmful event occurring and the severity of its potential consequences. People's perception of risk is subject to uncertainty and subjectivity. Everything carries a certain degree of risk; risk is objective and cannot be eliminated. The uncertainty and subjectivity mean that different people have different levels of understanding of risk, and based on certain perceptions, the negative effects of some risks can be reduced or eliminated by human means.

[0111] Definition 2: Risk Point: A system component node that may generate or propagate risks during the operation and production of an urban rail transit system is called a risk point. A risk point may have several risk attributes.

[0112] Risks within a system have the potential to propagate along risk propagation paths, adversely affecting people or property and potentially causing accidents. Based on this fundamental phenomenon, and to facilitate the construction of an active safety knowledge graph ontology model for urban rail transit operations, the concepts of risk events and feedforward signals are proposed, and accidents are defined more accurately.

[0113] In the field of information extraction, an event is an event or change of state that occurs within a specific time period and geographical area, resulting from one or more actions by one or more actors. Therefore, the action or phenomenon of a risk point becoming in a failed (damaged) state and propagating risk to other risk points can be described as an event occurring at that risk point. Based on the above analysis, a definition of a risk event is proposed.

[0114] Definition 3: Risk event: A series of events, either individually or in combination, caused by a change in the state of a risk point, that transform a potential hazard into an actual loss. A risk event is a prerequisite for causing or exacerbating an accident.

[0115] As risk propagates within the system, when a risk event (a state change and risk propagation) occurs at a risk point at the end of the propagation path, the entire system reaches a dangerous state—a critical state before an accident. The risk event generated at this risk point at the end of the propagation path is extremely critical; preventing this event means cutting off the possibility of an accident at the last moment. Therefore, this risk event is considered a feedforward signal for an accident, and the following definition is proposed:

[0116] Definition 4: Feedforward Signals: Feedforward signals are events that exceed a specific safety threshold, indicating an impending accident. In the process of developing into an accident, the appearance of a feedforward signal indicates that the hazard has progressed to a critical state approaching an accident. Precursor information generally signifies the complete failure of one or more safety systems, or the partial failure of multiple safety systems.

[0117] Furthermore, the concepts of accidents and measures in the operation of urban rail transit systems are proposed.

[0118] Definition 5 Accident: A sequence of events that occurs during production activities and is contrary to human will, due to limitations in scientific knowledge and technology, or due to cognitive limitations, and which cannot currently be prevented or effectively controlled. Accidents may force a system to temporarily or for a longer period of time to cease operation, and may also cause casualties, property damage, or environmental harm.

[0119] Definition 6: Measures: These refer to actions taken to address risk points within a system, including prevention measures, governance measures, and rescue measures. Measures are adopted and implemented on risk points; the object of the relationship is the risk point, therefore, the two have an implementation relationship.

[0120] Therefore, five main entity categories are introduced: risk points, risk events, feedforward signals, accidents, and measures. Based on the introduction of entity types, and with the core concept of risk points generating events, the types of relationships between entities are defined, as shown in Table 1.

[0121] Table 1 Relationship Type Information Table

[0122]

[0123]

[0124] Step 2: Construct an ontology model of proactive safety knowledge graph for urban rail transit operations using a seven-step method.

[0125] like Figure 2 As shown, step 2 specifically involves:

[0126] The ontology model uses risk points, risk events, feedforward signals, accidents, and measures as its main entities, connecting these entities through key relationships (including action relationships, causal relationships, etc.). The relatively mature seven-step method is chosen to construct the urban rail transit operation proactive safety knowledge graph ontology model. The steps are as follows:

[0127] (1) Determine the professional field and scope of the subject.

[0128] The research field is urban rail transit operation safety; the research aims to clarify and systematize the knowledge of safety risks in urban rail operation, and provide support for building an active safety knowledge graph for urban rail operation; the scope of the ontology is all relevant safety risk factors that may lead to urban rail operation accidents.

[0129] (2) Examine the possibility of reusing existing ontologies

[0130] The main functions of constructing a knowledge graph are knowledge reasoning and decision support. Based on a literature review, existing ontology structures in the field of urban rail transit operation safety do not adequately meet the requirements of this paper, therefore no reusable ontology exists. However, ontology structures in other industries, such as medicine and chemistry, offer valuable references, which this study can draw upon.

[0131] (3) List the important terms in the domain (ontology).

[0132] See step 1 for important terms in the ontology.

[0133] (4) Define classes and class hierarchy

[0134] Risk points are categorized into four main types: personnel, equipment and facilities, management, and environment, involving four levels: Level 1, Level 2, Level 3, and Level 4. Accidents are classified into 10 categories, with feedforward signals further refined based on accident category. Measures are categorized into three types: prevention measures, governance measures, and rescue measures. Except for risk points, the categories of other entities do not involve levels.

[0135] (5) Define class attributes

[0136] Risk points and accidents are the core components of the ontology concept layer in this field. The attributes of the class mainly refer to the level and category attributes of risk points, the information attributes of accidents, and the information attributes of measures in the operation of urban rail transit systems.

[0137] (6) Define the facets of the attribute

[0138] An attribute facet refers to the attributes of an attribute, such as the attribute's type, allowed values, and number of possible values. In the computer field, this is used to describe attribute types such as character, numeric, date, object instance, object class, and file path. In the field of urban rail transit operation safety, it mainly involves character data.

[0139] (7) Create an instance

[0140] Using the first six steps of the "seven-step method," the main concepts, terms, relationships, and attributes in the field of urban rail transit operation safety and accidents were defined and established, completing the top-level design of the knowledge graph ontology concept layer, enabling the creation of instances and the filling of relationships between instances. For example... Figure 4 As shown, a knowledge graph ontology model for proactive safety in urban rail transit operations was constructed with the core concept of "what to prevent and how to prevent it".

[0141] Step 3: Perform text data preprocessing for typical accident scenarios, and conduct topic mining to obtain key risk information from the text.

[0142] Step 3 specifically involves:

[0143] Taking urban rail transit fires (including explosions) as the research object, this study performs word segmentation and sentence breaking on the text set, and groups words with consistent meanings under a consistent professional concept to achieve concept disambiguation. The risk knowledge and information contained in the fire (including explosion) accident texts can serve as an important basis for constructing a knowledge graph knowledge base. Topic modeling is a probability-based statistical algorithm model that can analyze the implicit topic structure in text. Generally, in topic modeling, documents are modeled as a mixture of topics, and topics are the probability distributions of words. Topic modeling can well express the inherent relationships between lexical features in text. Knowledge triples in the graph are also regarded as multiple word feature items. Using topic modeling, knowledge triples are organized, their relationships are depicted, and potential patterns are discovered. Through the application of topic modeling, key knowledge (keywords) in accident texts can be obtained, providing a basis for the extraction of risk points and triples. It can also abstract the potential risk patterns contained in all the information of a certain type of typical accident into different text topics. The text topics with higher probabilities in the information reflect several major risk patterns in this type of unsafe event. Through the analysis of major risk patterns, this paper can summarize the main types of feedforward signals.

[0144] The Biterm Topic Model (BTM) views the entire corpus as a mixture of various topics and performs topic learning on short texts from the perspective of the entire corpus to address the sparsity problem. A biterm refers to an unordered pair of words that appear simultaneously in the context, each biterm originating from a specific topic. The author extracts any two different words from a short text to form a biterm. The BTM model models word co-occurrence patterns rather than document-level patterns, utilizing the aggregation of the entire corpus for topic learning. This allows it to obtain the distribution of topic components and global topics within the corpus, solving the document-level pattern sparsity problem and enhancing the effectiveness of topic learning. The graphical representation of BTM is as follows: Figure 3 As shown, α and β are hyperparameters of the Dirichle prior distribution; θ is the topic probability distribution in the BTM text corpus; Ф k This is the topic-word pair probability distribution; z is the topic number of the word pair bitterm, and T is the number of topics; w i w j B represents the two words in the biterm; B is the number of word pairs in the BTM corpus.

[0145] This paper uses perplexity to measure the effectiveness of topic models in extracting text topics. A lower perplexity indicates a stronger topic model, better generalization ability, and higher overall performance. The formula is defined as follows:

[0146]

[0147] In the formula D test —The text set of the test corpus;

[0148] —In the document The word;

[0149] For document The number of words in the text;

[0150] M—The size of the test corpus text set (the number of documents);

[0151]

[0152] For models trained with different numbers of topics, the optimal number of topics (and other parameters) corresponds to the lowest perplexity value. However, more topics are not always better; a large number of topics often leads to overfitting. A more reliable method is to observe the curve of perplexity as a function of the number of topics, find the minimum value within a reasonable range of curve fitting, and then select an appropriate number of topics. Taking fire (including explosion) accidents as the research object, after data preprocessing, BTM topic analysis was performed. The lowest perplexity point was obtained when the number of topics equaled 8. Figure 8 As shown. Based on the theme classification results, all themes and their corresponding meanings are summarized. The main content of theme 6 is emergency rescue. Among the remaining results, it is concluded that there are five main types of feedforward signals for urban rail operation fire scenarios. The sum of the theme intensities corresponding to the feedforward signals is expressed as a probability, and the ratio of each probability to the sum of all probabilities is taken as the feedforward signal f. k The influence coefficients are shown in Table 2.

[0153] Table 2 Feedforward Signals

[0154]

[0155] Step 4: Extract knowledge triples from typical accident scenarios based on dependency parsing, and consult industry experts after extraction is completed.

[0156] Step 4 is as follows:

[0157] Using typical accident scenario texts as the research object cannot meet the needs of deep learning methods that require a large amount of labeled data. Dependency parsing-based extraction methods can lay the foundation for triple extraction and achieve good results. To obtain entities and the relationships between them, knowledge triples from typical accident scenarios are extracted using dependency parsing. After extraction, industry experts are consulted, focusing on entity concepts such as risk points, risk events, and measures, and the relationships between them. Generally, the types of relationships labeled by dependency parsing are shown in Table 3.

[0158] Table 3. Dependency Parsing Tagging Relations

[0159]

[0160] This study uses dependency parsing based on accident text to semi-automatically extract knowledge triples from typical accident scenarios. It primarily relies on the results of dependency parsing to extract knowledge triples centered around predicates (which trigger the entire triple, typically verbs). Referring to previous research, the basic premises and extraction rules are as follows:

[0161] 1) Dependency parsing will yield a core predicate (verb) of a sentence, but there may be multiple predicates in the sentence that have other relationships with the core predicate. Each predicate (verb) in the sentence corresponds to a risk event or measure in a triple.

[0162] 2) If there are predicates in a sentence that are in a parallel relationship, the predicates (verbs) that are in a parallel relationship can form a predicate (verb) chain and share the same subject.

[0163] 3) Taking the predicate in the sentence as the core, if it is determined that there is a subject-predicate relationship and the nouns in the relationship are parallel, then these nouns share the same predicate.

[0164] 4) Taking the predicate in the sentence as the core, if there is a verb-object relationship in the sentence and the nouns in the relationship are parallel, then these nouns share the same predicate.

[0165] 5) If there are subject-verb and verb-object relationships in the sentence, directly extract the corresponding triplet <subject, verb, object>.

[0166] 6) If the sentence contains only a subject-predicate relationship, the core content of the verb-complement structure or prepositional phrase relationship is identified as supplementary knowledge, and the knowledge triple <subject, predicate, supplementary knowledge> is extracted.

[0167] 7) If the sentence contains only a verb-object relationship, then the core content of the attributive-head relationship is extracted as supplementary knowledge to form a triplet of <supplementary knowledge, predicate, object>.

[0168] Due to potential descriptive issues in the accident text corpus and errors in natural language extraction methods, after obtaining all knowledge triples, consultations were conducted with multiple experts and operational personnel from the urban rail transit industry and operating companies. Based on expert opinions and the urban rail transit operation safety professional thesaurus constructed in this paper, all triple content was manually corrected, with a focus on obtaining triple results for <risk point, risk event, risk point> and <implementation position, measure, risk point>. After preprocessing all fire accident text data, semantic role labeling was performed on the preprocessed fire category text set using Harbin Institute of Technology's PyLTP. Knowledge triples were obtained based on the dependency parsing results of all accident texts. Based on the schema and concept relationship settings, and after expert consultation and manual correction, all triples (entities and relations) were obtained.

[0169] Step 5: Import the graph database using <entity, relation, entity> as its standard tuple structure to complete the construction of the urban rail transit operation proactive safety knowledge graph for typical accident scenarios. The entire knowledge graph network is divided into a four-layer chain structure: risk point layer, risk event layer, feedforward signal layer (this layer is connected to the accident entity), and measure layer.

[0170] Step 5 specifically involves:

[0171] The constructed knowledge graph network can display the content structure and relationships between entities. It consists of a four-layer chain network structure: risk point layer (P), risk event layer (E), feedforward signal layer (F), and measure layer (M), exhibiting strong complex network characteristics. Knowledge graph triples (entities and relationships) are imported into Neo4j to construct a fire scenario knowledge graph for proactive safety in urban rail transit operations, such as... Figure 5 As shown, the nodes within the three boxes from left to right represent the measure layer (M), the risk event layer (E), and the feedforward signal layer (F), respectively. The rightmost node represents a fire accident, and the remaining nodes represent the risk point layer (P). This knowledge graph contains 207 nodes and 284 relationships, including 87 risk points, 89 risk events, 22 measures, 5 feedforward signals, and 1 accident, as well as several hierarchical, causal, and implementation relationships. Each risk point and measure carries attribute information.

[0172] Step 6: Based on the external relationships and network structure of risk points, and comprehensively considering the degree of participation, role, and importance of each point, establish three weighted indicators to represent the relevance of risk points to other levels. Using set pair analysis, establish the correlation between risk points and the entire knowledge graph, quantify the correlation (trend) of a risk point x to different layers of the knowledge graph, and then determine which connections are more important and worthy of attention.

[0173] Step 6 specifically involves:

[0174] The risk points in the risk point layer are the core of the entire knowledge graph network. Based on the external relationships and network structure of the risk points, and considering the degree of participation, role, and importance of the points, three weight indicators representing the relevant weights of the risk points to other levels are established, as shown in equations (1) to (3). Combining the relevant weight indicators, with the risk points as the core, the relationships between the various layers of the knowledge graph are represented as follows: Figure 6 As shown.

[0175] A1(x)=α1*l PE (x)+α2*l EP (x) (1)

[0176] In the formula, l PE (x) represents the frequency of other risk events connected to risk point x; EP (x) represents the frequency of risk point x being connected to other risk events; α1 and α2 are adjustment coefficients, α1+α2=1; A1(x) represents the correlation weight between risk point x and the risk event layer.

[0177]

[0178] In the formula, β k Indicates the feedforward signal f k The influence coefficients of all βk The sum is 1; This indicates that the risk point x is connected to the feedforward signal f. k The frequency of is 1 or 0; n represents the total number of feedforward signals; A2(x) represents the correlation weight between risk point x and the feedforward signal layer.

[0179] A3(x)=ρ1l PM1 (x)+ρ2l PM2 (x)+ρ3l PM3 (x) (3)

[0180] In the formula, A3(x) represents the correlation weight between risk point x and the measure layer; ρ1 represents the influence coefficient of prevention and control measures; l PM1 (x) represents the frequency of prevention and control measures connected to risk point x; ρ2 represents the impact coefficient of governance measures, l PM2 (x) represents the frequency of governance measures connected to risk point x; ρ3 represents the impact coefficient of rescue measures, l PM3 (x) represents the frequency of rescue measures connected to risk point x; the sum of ρ1, ρ2, and ρ3 is 1.

[0181] The basic idea of ​​set pair analysis is to formulate the evaluation objectives and evaluation criteria into a set pair, objectively reflecting the similarities, differences, and oppositions between the two. Based on the extensibility theory of set pair analysis, the method is improved to establish the correlation between risk points and the indicators of the entire knowledge graph, quantify the correlation (degree of tendency) of a certain risk point x to different layers of the knowledge graph, and then determine which connections are more important and worthy of attention.

[0182]

[0183] N = A1(x) + A2(x) + A3(x) (5)

[0184] In the formula, μ(x) represents the set-pair connectivity, and the coefficients of i1, i2, and i3 represent the degree of tendency of risk point x towards the risk event layer, feedforward signal layer, and measure layer, respectively. Based on this, it can be determined which links of the risk point require focused attention, and the content of the entities (and entity attributes) and entity relationships at both ends of the link can be used to assist in security decision-making. Alternatively, expert opinions can be used to assign values ​​to i1, i2, and i3, thereby evaluating and comparing the overall importance of risk points in the knowledge graph and determining the risk points to be prioritized for control.

[0185] To reflect the importance of the "proactive" nature of proactive safety decision support, this paper, based on expert opinions, argues that the influence coefficients of prevention, governance, and rescue measures in the knowledge graph network decrease sequentially, with ρ1, ρ2, and ρ3 taking values ​​of 0.5, 0.3, and 0.2, respectively.

[0186] The Cypher statement queries and exports the relationships and entities associated with a specific risk point, and statistically summarizes the frequency of links connecting the risk point to other layers. Taking the risk point "cable" as an example, the statement "match(p:risk point{fxmc:"passengers"})-[r]-(n)returnr,n" retrieves the relevant links for the risk point. Figure 7 As shown.

[0187] Based on equations (1) to (5), the final calculation result is shown in equation (6).

[0188] μ=0.65i1+0.25i2+0.1i3 (6)

[0189] Further inputting the query statement "MATCH p=(a:`risk point`{fxmc:'cable'})-[r]-(c:`risk event`)-[m]-(b:`risk point`)RETURN p" will yield the following results. Figure 8 The results are shown.

[0190] Based on the results, the risk point "cable" shows a high degree of tendency towards the risk event layer. Therefore, the most priority considerations for cables are: cables located in tunnels are a key focus for operators; the insulation and sheathing of cables throughout the urban rail system should be repaired and replaced promptly during routine safety inspections; and inspection personnel should take measures to prevent cables from igniting and forming a flammable solution that could affect the train's braking system.

[0191] Example 4

[0192] This embodiment 4 provides a non-transitory computer-readable storage medium for storing computer instructions. When these computer instructions are executed by a processor, they implement the ontology-based urban rail transit operation proactive safety knowledge graph construction method described above. The method includes:

[0193] Identify entities and define the types and categories of relationships between them; entities include risk points, risk events, feedforward signals, incidents, and measures.

[0194] Based on the identified entities, the types and categories of relationships between entities, a knowledge graph ontology model for proactive safety in urban rail transit operations is constructed.

[0195] Based on the knowledge graph ontology model of proactive safety in urban rail transit operations, text data preprocessing of typical accident scenarios is performed, and key risk information is obtained by topic mining of the text.

[0196] Knowledge triples for typical accident scenarios are extracted based on dependency parsing.

[0197] Using <entity, relation, entity> as its standard tuple structure, a graph database was imported to complete the construction of an active safety knowledge graph for urban rail transit operations oriented towards typical accident scenarios.

[0198] Example 5

[0199] This embodiment 5 provides a computer program product, including a computer program that, when run on one or more processors, is used to implement the ontology-based urban rail transit operation proactive safety knowledge graph construction method described above. The method includes:

[0200] Identify entities and define the types and categories of relationships between them; entities include risk points, risk events, feedforward signals, incidents, and measures.

[0201] Based on the identified entities, the types and categories of relationships between entities, a knowledge graph ontology model for proactive safety in urban rail transit operations is constructed.

[0202] Based on the knowledge graph ontology model of proactive safety in urban rail transit operations, text data preprocessing of typical accident scenarios is performed, and key risk information is obtained by topic mining of the text.

[0203] Knowledge triples for typical accident scenarios are extracted based on dependency parsing.

[0204] Using <entity, relation, entity> as its standard tuple structure, a graph database was imported to complete the construction of an active safety knowledge graph for urban rail transit operations oriented towards typical accident scenarios.

[0205] Example 6

[0206] This embodiment 6 provides an electronic device, including: a processor, a memory, and a computer program; wherein, the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device executes instructions to implement the ontology-based urban rail transit operation proactive safety knowledge graph construction method described above, the method including:

[0207] Identify entities and define the types and categories of relationships between them; entities include risk points, risk events, feedforward signals, incidents, and measures.

[0208] Based on the identified entities, the types and categories of relationships between entities, a knowledge graph ontology model for proactive safety in urban rail transit operations is constructed.

[0209] Based on the knowledge graph ontology model of proactive safety in urban rail transit operations, text data preprocessing of typical accident scenarios is performed, and key risk information is obtained by topic mining of the text.

[0210] Knowledge triples for typical accident scenarios are extracted based on dependency parsing.

[0211] Using <entity, relation, entity> as its standard tuple structure, a graph database was imported to complete the construction of an active safety knowledge graph for urban rail transit operations oriented towards typical accident scenarios.

[0212] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0213] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0214] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0215] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment, whereby a series of operational steps are performed to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0216] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that, based on the technical solutions disclosed in the present invention, various modifications or variations that can be made by those skilled in the art without creative effort should be included within the scope of protection of the present invention.

Claims

1. A method for constructing an ontology-based knowledge graph for proactive safety in urban rail transit operations, characterized in that, include: Identify entities and define the types and categories of relationships between them; Entities include risk points, risk events, feedforward signals, incidents, and measures; Based on the identified entities, the types and categories of relationships between entities, a knowledge graph ontology model for proactive safety in urban rail transit operations is constructed. This model includes: defining the ontology scope of urban rail transit operation safety as all relevant safety risk factors leading to urban rail transit operation accidents; determining important terms within the ontology based on the reusability of existing ontologs; and categorizing risk points into personnel, equipment and facilities, management, and environment categories, with the category attributes referring to the level and category attributes of risk points in urban rail transit system operation, the information attributes of accidents, and the information attributes of measures. Based on the knowledge graph ontology model of proactive safety in urban rail transit operations, text data preprocessing of typical accident scenarios is performed, and key risk information is obtained by topic mining of the text. Knowledge triples for typical accident scenarios are extracted based on dependency parsing. Using <entity, relation, entity> as its standard tuple structure, a graph database was imported to construct a knowledge graph for proactive safety in urban rail transit operations oriented towards typical accident scenarios. The constructed knowledge graph network was divided into a four-layer chain structure: risk point layer, risk event layer, feedforward signal layer, and measure layer. Based on the external relationships and network structure of risk points, and comprehensively considering the participation, role, and importance of points, three weight indicators representing the relevance of risk points to other layers were established. Based on the theory of set pair analysis, the correlation degree between risk points and the entire knowledge graph was established, quantifying the correlation degree of a certain risk point to different layers of the knowledge graph, and thus identifying key connections.

2. The method for constructing an ontology-based proactive safety knowledge graph for urban rail transit operation according to claim 1, characterized in that, The relationships between entities include: risk point-risk event, risk event-risk point, risk point-feedforward signal, feedforward signal-accident and action-risk point.

3. The method for constructing an ontology-based proactive safety knowledge graph for urban rail transit operation according to claim 1, characterized in that, Based on the knowledge graph ontology model of proactive safety in urban rail transit operations, text data preprocessing of typical accident scenarios is performed, and topic mining is conducted on the text to obtain key risk information, including: Accident text data preprocessing: For a typical accident scenario, a text set of accident categories is selected as the object of analysis and processing, and data processing such as word segmentation and sentence breaking is performed; words with the same meaning in the text are grouped under the same professional concept to complete the concept disambiguation; Risk information theme mining for typical accident scenarios: For urban rail operation accident texts, a word and probability extraction method based on the BTM topic model is used; by applying the topic model, key knowledge in the accident text is obtained, providing some basis for the extraction of risk points and triples. The potential risk patterns contained in all information of a certain type of typical accident are abstracted into different text themes. The text themes with higher probabilities in the information reflect the main risk patterns in this type of unsafe event, and the types of feedforward signals are summarized.

4. The method for constructing an ontology-based proactive safety knowledge graph for urban rail transit operation according to claim 1, characterized in that, Dependency parsing based on accident texts is used to semi-automatically extract knowledge triples for typical accident scenarios. Based on the results of dependency parsing, knowledge triples centered on predicates are extracted.

5. The method for constructing an ontology-based proactive safety knowledge graph for urban rail transit operation according to claim 1, characterized in that, Using <entity, relation, entity> as its standard tuple structure, the graph database is imported to complete the construction of an active safety knowledge graph for urban rail transit operations oriented towards typical accident scenarios. This includes: using Neo4j to store the graph knowledge, using <entity, relation, entity> as its standard tuple structure, and using Python to import the tuple structure to complete the construction of an active safety knowledge graph for urban rail transit operations oriented towards typical accident scenarios.

6. The method for constructing an ontology-based proactive safety knowledge graph for urban rail transit operation according to claim 1, characterized in that, The constructed knowledge graph network displays the content structure and interrelationships between entities, consisting of a four-layer chain network structure: risk point layer P, risk event layer E, feedforward signal layer F, and measure layer M. Risk points at the risk point layer are the core of the entire knowledge graph network. Based on the external relationships and network structure of risk points, and comprehensively considering the degree of participation, role, and importance of points, three weight indicators are established to represent the relevance of risk points to other levels: This indicates the correlation weight between risk point x and the risk event layer. This indicates the correlation weight between the risk point x and the feedforward signal layer. This represents the weight of risk point x in relation to the measures layer; the evaluation objectives and evaluation criteria are formulated as a set pair, and based on the extensibility theory of set pair analysis, the correlation between risk point and the indicators of the entire knowledge graph is established, the correlation between a certain risk point x and different layers of the knowledge graph is quantified, and then the key connections are determined.

7. A knowledge graph construction system for proactive safety in urban rail transit operations based on ontology, characterized in that, include: The determination module is used to determine entities and define the types and categories of relationships between entities; Entities include risk points, risk events, feedforward signals, incidents, and measures; The construction module is used to build an ontology model of proactive safety knowledge graph for urban rail transit operation based on the identified entities, the types of relationships between entities, and the relationship types. The construction of the ontology model of proactive safety knowledge graph for urban rail transit operation includes: determining the ontology scope of urban rail transit operation safety as all relevant safety risk factors that lead to urban rail transit operation accidents; determining important terms in the ontology based on the reusability of existing ontology; risk points are divided into personnel, equipment and facilities, management, and environment categories, and the attributes of the categories refer to the level and category attributes of risk points in urban rail transit system operation, the information attributes of accidents, and the information attributes of measures. The preprocessing module is used to preprocess text data of typical accident scenarios based on the knowledge graph ontology model of urban rail operation active safety, and to perform topic mining on the text to obtain key risk information. The extraction module is used to extract knowledge triples from typical accident scenarios based on dependency parsing. The import module is used to import graph databases with <entity, relation, entity> as its standard tuple structure to complete the construction of an active safety knowledge graph for urban rail transit operations oriented towards typical accident scenarios. The constructed active safety knowledge graph network for urban rail transit operations is divided into a four-layer chain network structure: risk point layer, risk event layer, feedforward signal layer, and measure layer. Based on the external relationships and network structure of risk points, and comprehensively considering the participation degree, role relationship, and importance of points, three weight indicators representing the relevance of risk points to other layers are established. Based on the theory of set pair analysis, the correlation degree between risk points and the entire knowledge graph is established, quantifying the correlation degree of a certain risk point to different layers of the knowledge graph, and thus identifying key connections.

8. An electronic device, characterized in that, include: The device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions that implement the ontology-based urban rail transit operation active safety knowledge graph construction method as described in any one of claims 1-6.

Images