A flight operation risk early warning method and system based on a knowledge graph

Abstract

The application relates to the technical field of knowledge graphs, and proposes a flight operation risk early warning method and system based on a knowledge graph, which comprises the following steps: fusing multiple sources of flight operation data to build a multilevel risk knowledge graph, generating structured instances from real-time flight data based on the graph ontology specification, and constraining the entity association direction; taking the instance as a search starting point to limit the link jump number to search a potential risk path, combining real-time data to check a transmission rule to screen effective risk links; combining the number of effective paths, the risk level, and the node semantic distance to quantitatively synthesize a risk index, matching a corresponding early warning level, and outputting early warning information containing a risk node and a disposal suggestion; and the application can improve the efficiency of flight operation risk early warning based on the knowledge graph.

Description

Technical Field

[0001] This invention relates to the field of knowledge graph technology, and in particular to a method and system for early warning of flight operation risks based on knowledge graphs. Background Technology

[0002] Current flight operation risk warnings largely rely on structured databases for threshold-based risk identification, which can only independently verify single-dimensional operational data and cannot integrate the inherent relationships between multi-source heterogeneous data such as aircraft technical status, airport operation procedures, meteorological environment, and historical risk events. Existing technologies lack unified domain entity association modeling methods, making it difficult to fully depict the cross-factor transmission logic of risks. In scenarios with massive real-time flight data input, it is easy to miss potential risk chains caused by the coupling of multiple factors, resulting in a significant shortcoming in the coverage of risk identification.

[0003] The existing early warning mechanism lacks a standardized system for verifying and quantifying risk transmission. When searching for potential risk paths, it fails to constrain the direction of entity relationship transmission and the number of search hops. It only performs simple numerical comparisons of risk triggering conditions, failing to distinguish the severity levels of different risk events or their proximity to target flights. The method of outputting warning levels with uniform thresholds ignores the attenuation effect of risk transmission paths, resulting in high distortion of comprehensive risk assessment results. Early warning information only outputs the risk level and cannot synchronously match corresponding historical response plans, leading to insufficient overall accuracy and practicality in decision support. Summary of the Invention

[0004] This invention provides a knowledge graph-based method and system for early warning of flight operation risks, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, this invention provides a knowledge graph-based method for flight operation risk early warning, comprising: Step 1: Based on historical flight operation data, aircraft technical documents, airport operation rules, and historical meteorological data, construct a multi-level risk knowledge graph for flight operations, including the data entities, entity attributes, and relationships between entities. Step 2: Convert real-time flight operation status data into structured state instances through the ontology layer definition in the multi-level risk knowledge graph, and define the relationship type and transmission direction between entities of the structured state instances in the multi-level risk knowledge graph; Step 3: Based on the relationship type and the transmission direction, take the structured state instance as the input query, take the corresponding graph node matched and located in the multi-level risk knowledge graph as the starting point, and search for the associated path within the preset number of hops using the risk propagation relationship edge in the multi-level risk knowledge graph. Step 4: Based on the attribute values ​​of the structured state instance, trigger verification of the risk transmission rules in the searched potential association paths, and mark the potential association paths that pass the verification as valid risk transmission paths; Step 5: Based on the number of effective risk transmission paths and the corresponding risk event nodes and risk level information, and combined with the semantic distance of the target flight in the multi-level risk knowledge graph, calculate the comprehensive risk index of the target flight, and output the risk warning information corresponding to the comprehensive risk index.

[0006] In a preferred embodiment, the step of constructing a multi-level risk knowledge graph for flight operations based on historical flight operation data, aircraft technical documents, airport operation rules, and historical meteorological data, including the data entities, entity attributes, and relationships between entities, includes: Historical flight operation data, aircraft technical documents, airport operation rules, and historical meteorological data, after data cleaning, are integrated into standardized text data for domain entity recognition to obtain a candidate entity set for flight operations; Multiple candidate entities that refer to the same objective object in the candidate entity set are merged into an entity node and assigned a globally unique identifier; Based on the event records in the historical flight operation data and the associated descriptions in the aircraft technical documents, the direct relationships between each entity node are extracted and assigned relationship type labels; Based on the program logic in the airport operation rules and the meteorological impact records in the historical meteorological data, the indirect relationships between the entity nodes are parsed out and assigned transmission attribute labels; Based on the entity nodes, the relationship type labels, and the transmission attribute labels, a multi-level risk knowledge graph is constructed according to a hierarchical structure from high to low risk levels.

[0007] In a preferred embodiment, the step of converting real-time flight operation status data into structured state instances through the ontology layer definition in the multi-level risk knowledge graph includes: The ontology layer definition of the multi-level risk knowledge graph includes the attribute field set corresponding to each entity type, the data format constraints of each attribute field, and the valid value range of each attribute field. Receive the real-time flight operation status data and parse out the data fields and corresponding field values ​​in the real-time flight operation status data; Based on the mapping relationship between the attribute field set and the data field in the ontology layer definition, the field values ​​are respectively filled into the attribute fields of the corresponding entity types in the multi-level risk knowledge graph; Based on the data format constraints, the format of the filled field value is converted, and the validity of the converted field value is verified according to the valid value range. The field values ​​that pass the validity check and the corresponding entity types are instantiated into structured state instances.

[0008] In a preferred embodiment, defining the relationship types and transmission directions between entities in the multi-level risk knowledge graph for the structured state instances includes: Based on the entity type of the structured state instance, the allowed relation type set is defined from the ontology layer; each relation type in the allowed relation type set contains a corresponding set of legal start entity types and a set of legal end entity types; Obtain the access location information of the structured state instance in the multi-level risk knowledge graph; the access location information includes the existing node identifier of the target flight corresponding to the structured state instance in the graph; Based on the set of allowed relation types, select the first type of available relation types that use the existing node identifier as the legal starting entity type and the structured state instance as the legal ending entity type. Based on the set of allowed relation types, a second type of available relation type is selected, which uses the entity type of the structured state instance as the legal starting entity type and the existing node identifier as the legal ending entity type. The first type of available relation type and the second type of available relation type are jointly defined as the set of relation types that the structured state instance is allowed to use when accessing the multi-level risk knowledge graph; The propagation direction from the existing node to the structured state instance is determined based on the first type of available relationship, and the propagation direction from the structured state instance to the existing node is determined based on the second type of available relationship.

[0009] In a preferred embodiment, the step of using the structured state instance as input query based on the relationship type and the transmission direction, and starting from the corresponding graph node matched and located in the multi-level risk knowledge graph, performing an association path search within a preset hop count range using the risk propagation relationship edges in the multi-level risk knowledge graph, includes: Using the attribute field values ​​of the structured state instance as matching conditions, the type and attributes of all entity nodes in the multi-level risk knowledge graph are matched and located to obtain the search starting point of the structured state instance; Based on the relationship type and the transmission direction, relevant edges that conform to the transmission direction and whose relationship type belongs to the range of the relationship type are selected from the multi-level risk knowledge graph and used as available risk propagation relationship edges; Taking the search starting point as the current node, all adjacent nodes directly connected to the current node through an available risk propagation relationship edge are recorded as one-hop reachable nodes, and the available risk propagation relationship edges traversed from the search starting point to each of the one-hop reachable nodes are recorded as a one-hop path set; Using each of the aforementioned one-hop reachable nodes as the new current node, repeatedly execute the operation of expanding to adjacent nodes through the available risk propagation relationship edge, expanding outward hop by hop until the current expansion hop count reaches the upper limit of the preset hop count range, and record the newly added reachable nodes and corresponding paths during each hop expansion process; Paths whose endpoints are risk event nodes are selected from all recorded paths and output as the potential associated paths.

[0010] In a preferred embodiment, the step of triggering verification of risk transmission rules in the searched potential association paths based on the attribute values ​​of the structured state instance includes: Obtain the edge identifier corresponding to each risk propagation relationship edge traversed by the potential association path, and the node pair identifier corresponding to the adjacent node pair in each potential association path; Using the edge identifier and the node pair identifier as indexes, extract the risk transmission rules corresponding to each potential association path from the relation attribute storage area of ​​the multi-level risk knowledge graph; The risk transmission rule includes a condition field and an action field; the condition field records the judgment criteria that must be met to trigger the current risk transmission rule and the corresponding range of benchmark values, and the action field records the transmission result type corresponding to the current risk transmission rule after it is triggered; The risk transmission rule is structured and parsed, and the condition field is parsed into at least one condition expression to be determined, and the action field is parsed into the corresponding transmission action to be executed; the condition expression to be determined contains the entity attribute name to be compared and the comparison logic corresponding to the benchmark value range.

[0011] In a preferred embodiment, the step of triggering verification of risk transmission rules in the searched potential association paths based on the attribute values ​​of the structured state instance, and marking the potential association paths that pass the verification as valid risk transmission paths, includes: Based on the entity attribute name in the condition expression to be determined, extract the corresponding real-time attribute value from the attribute field of the structured state instance; Substitute the real-time attribute value into the condition expression to be determined, and compare the real-time attribute value with the benchmark value range according to the comparison logic to determine whether the real-time attribute value falls within the benchmark value range; When all the comparison results of the condition expressions to be determined corresponding to a potential association path are found to be included, the potential association path is determined to pass the trigger verification. When any one of the condition expressions to be determined is not found, the potential association path is determined to fail the trigger verification. Each potential associated path that passes the trigger verification is marked as a valid risk transmission path, and the transmission action to be executed corresponding to each valid risk transmission path is recorded as the transmission result label of the valid risk transmission path.

[0012] In a preferred embodiment, the step of calculating the comprehensive risk index of the target flight based on the number of effective risk transmission paths and the corresponding risk event nodes and risk level information, combined with the semantic distance of the target flight in the multi-level risk knowledge graph, includes: The total number of effective risk transmission paths is counted, and the risk level information of the risk event nodes corresponding to the effective risk transmission paths is read. Obtain the node position of the target flight node in the multi-level risk knowledge graph corresponding to the target flight, and calculate the semantic distance between the target flight node and the risk event node corresponding to each effective risk transmission path; Based on the risk level information corresponding to each effective risk transmission path, a risk contribution weight coefficient is determined for each effective risk transmission path; the risk contribution weight coefficient is positively correlated with the risk level information. Based on the semantic distance corresponding to each effective risk transmission path, a distance attenuation coefficient is determined for each effective risk transmission path, and the distance attenuation coefficient is negatively correlated with the semantic distance; Based on the total number of effective risk transmission paths, the risk contribution weight coefficient, and the distance attenuation coefficient, the effective risk transmission paths are evaluated in multiple dimensions to obtain the comprehensive risk index of the target flight.

[0013] In a preferred embodiment, the step of outputting the risk warning information corresponding to the comprehensive risk index includes: The comprehensive risk index is compared with the preset risk warning level range to determine the target value range of the comprehensive risk index; Based on the current risk warning level determined by the risk warning level corresponding to the target value range, historical risk management measures records that match the current risk warning level are parsed from the risk event entities of the multi-level risk knowledge graph as suggested mitigation measures. The current risk warning level, the key risk event nodes that trigger the effective risk transmission path, and the suggested mitigation measures are combined into a risk warning information output.

[0014] To address the aforementioned problems, the present invention also provides a knowledge graph-based flight operation risk early warning system, the system comprising: The knowledge graph construction module is used to construct a multi-level risk knowledge graph for flight operations based on data entities, entity attributes, and relationships between entities in historical flight operation data, aircraft technical documents, airport operation rules, and historical meteorological data. The state instantiation module is used to convert real-time flight operation status data into structured state instances through the ontology layer definition in the multi-level risk knowledge graph, and to limit the relationship type and transmission direction between entities of the structured state instances in the multi-level risk knowledge graph. The risk path search module is used to query the structured state instance as input based on the relationship type and the transmission direction, and to search for associated paths within a preset number of hops, starting from the corresponding graph node matched and located in the multi-level risk knowledge graph and using the risk propagation relationship edges in the multi-level risk knowledge graph. The transmission path verification module is used to trigger verification of risk transmission rules in the searched potential association paths based on the attribute values ​​of the structured state instance, and mark the potential association paths that pass the verification as valid risk transmission paths. The comprehensive risk assessment module is used to calculate the comprehensive risk index of the target flight based on the number of effective risk transmission paths and the corresponding risk event nodes and risk level information, combined with the semantic distance of the target flight in the multi-level risk knowledge graph, so as to output the risk warning information corresponding to the comprehensive risk index.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention constructs a hierarchical knowledge graph of flight operation risks, unifies and integrates multi-source heterogeneous business data, and standardizes entities, relationships, and risk transmission attributes. Based on the ontology layer, it standardizes the conversion rules of real-time flight data instances, completes automated verification of data format and value legality, can fully sort out the bidirectional risk transmission logic between various operational elements, limit the type of compliance relationship and the direction of propagation, and realize the efficient integration and access of real-time flight status and the knowledge graph system, which greatly improves the ability of multi-source data fusion modeling and real-time data structuring processing.

[0016] 2. This invention mines potential risk transmission links through multi-path traversal retrieval with a limited number of hops, filters effective risk paths by verifying transmission rules one by one based on real-time entity attributes, and completes multi-dimensional comprehensive risk quantification calculation by combining risk level weights and semantic distance decay coefficients. It outputs graded early warning information that matches corresponding handling suggestions, fully covering the entire process of risk tracing, quantitative assessment, and auxiliary handling, effectively improving the completeness of flight operation risk identification, the accuracy of risk quantification, and the decision support value of early warning information. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a knowledge graph-based flight operation risk early warning method according to an embodiment of the present invention. Figure 2 This is a functional module diagram of a knowledge graph-based flight operation risk early warning system provided in an embodiment of the present invention.

[0018] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0020] This application provides a knowledge graph-based flight operation risk warning method. The executing entity of this knowledge graph-based flight operation risk warning method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application embodiment: a server, a terminal, etc. In other words, the knowledge graph-based flight operation risk warning method can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0021] Reference Figure 1 The diagram shown is a flowchart illustrating a knowledge graph-based flight operation risk warning method according to an embodiment of the present invention. In this embodiment, the knowledge graph-based flight operation risk warning method includes: Step 1: Based on historical flight operation data, aircraft technical documents, airport operation rules, and historical meteorological data, construct a multi-level risk knowledge graph for flight operations, including the data entities, entity attributes, and relationships between entities. In this embodiment of the invention, the step of constructing a multi-level risk knowledge graph for flight operations based on historical flight operation data, aircraft technical documents, airport operation rules, and historical meteorological data, including the data entities, entity attributes, and relationships between entities, includes: Historical flight operation data, aircraft technical documents, airport operation rules, and historical meteorological data, after data cleaning, are integrated into standardized text data for domain entity recognition to obtain a candidate entity set for flight operations; Multiple candidate entities that refer to the same objective object in the candidate entity set are merged into an entity node and assigned a globally unique identifier; Based on the event records in the historical flight operation data and the associated descriptions in the aircraft technical documents, the direct relationships between each entity node are extracted and assigned relationship type labels; Based on the program logic in the airport operation rules and the meteorological impact records in the historical meteorological data, the indirect relationships between the entity nodes are parsed out and assigned transmission attribute labels; Based on the entity nodes, the relationship type labels, and the transmission attribute labels, a multi-level risk knowledge graph is constructed according to a hierarchical structure from high to low risk levels.

[0022] The system reads four types of raw data sources: historical flight operation data, aircraft technical documents, airport operation rules, and historical meteorological data. It then removes data entries line by line that are missing fields, contain garbled characters, or have values ​​exceeding the permitted range. After cleaning, the four types of data are converted into standardized text data with unified character encoding, field separators, and paragraph segmentation identifiers. The standardized text data is then traversed character by character to perform domain entity recognition. This recognition operation uses a pre-defined vocabulary of flight operation domain entities as a matching benchmark, extracting text fragments that perfectly match the vocabulary. Once all extractions are complete, a candidate entity set for flight operations is formed.

[0023] Traverse all entity content within the candidate entity set for flight operations, using the objective physical objects, business processes, and risk events corresponding to flight operation business as a unified matching benchmark. Compare the business referents of every two entities in the candidate entity set in turn. If the two entities point to the same objective object, they are determined to be mergeable entities. Integrate all the multiple candidate entities determined to be mergeable into a single entity node. Assign a set of character sequences that do not repeat with other entity nodes as a globally unique identifier to each integrated entity node, thus completing the identification assignment work for all entity nodes.

[0024] Retrieve all entity nodes with completed identifier assignments, read the complete event records within the historical flight operation data and the complete association descriptions within the aircraft technical documents one by one, extract the related sentences in the text that simultaneously contain two entity nodes, define the content in the sentences that expresses the direct business interaction, status impact, and equipment linkage between entity nodes as the direct association between entity nodes, bind the corresponding business meaning text identifier to each extracted direct association as the relationship type label, and complete the label assignment operation for all direct associations.

[0025] Retrieve all entity nodes with configured globally unique identifiers, read the complete program logic text within the airport operation rules and the complete meteorological impact record text within the historical meteorological data, and parse the entity node interaction content within the text that contains intermediate business element connections and non-instantaneous action links. Define the entity node interaction content generated across intermediate elements as indirect associations between entity nodes, and bind a text identifier that can describe the risk transmission characteristics to each parsed indirect association as a transmission attribute label, and complete the label assignment operation for all indirect associations.

[0026] This process involves aggregating all entity nodes carrying globally unique identifiers, all direct relationships and corresponding relationship type tags bound to text identifiers, and all indirect relationships and corresponding transmission attribute tags bound to text identifiers. Using the risk level classification standards defined in the civil aviation operation specifications as the hierarchical sorting benchmark, entity nodes are arranged hierarchically in descending order of risk level values. Entity nodes with direct and indirect relationships are connected through corresponding relationship edges. Once the hierarchical arrangement and node connections are completed, a complete multi-level risk knowledge graph is generated.

[0027] The beneficial effects are as follows: relying on unified cleaning and standardized processing of multi-source business data, combined with a fixed domain entity vocabulary library to complete entity extraction, it can generate a complete candidate entity set covering all business dimensions of flights. Through unified business reference benchmarks, entity normalization is completed and exclusive identifiers are configured, achieving standardized modeling of all business entities without duplication. It distinguishes between two different types of interaction logic in business texts, extracts corresponding associations and configures exclusive identifiers and tags. Based on the established risk level standards of civil aviation, a hierarchical graph structure is built to completely construct a multi-level risk knowledge graph that can clearly distinguish between direct interaction and indirect risk transmission links, realizing the complete structured carrying of various business elements and risk transmission logic of flight operations.

[0028] Step 2: Convert real-time flight operation status data into structured state instances through the ontology layer definition in the multi-level risk knowledge graph, and define the relationship type and transmission direction between entities of the structured state instances in the multi-level risk knowledge graph; In this embodiment of the invention, the step of converting real-time flight operation status data into structured state instances through the ontology layer definition in the multi-level risk knowledge graph includes: The ontology layer definition of the multi-level risk knowledge graph includes the attribute field set corresponding to each entity type, the data format constraints of each attribute field, and the valid value range of each attribute field. Receive the real-time flight operation status data and parse out the data fields and corresponding field values ​​in the real-time flight operation status data; Based on the mapping relationship between the attribute field set and the data field in the ontology layer definition, the field values ​​are respectively filled into the attribute fields of the corresponding entity types in the multi-level risk knowledge graph; Based on the data format constraints, the format of the filled field value is converted, and the validity of the converted field value is verified according to the valid value range. The field values ​​that pass the validity check and the corresponding entity types are instantiated into structured state instances.

[0029] The definition of the structured state instance's relationship type and transmission direction between entities in the multi-level risk knowledge graph includes: Based on the entity type of the structured state instance, the allowed relation type set is defined from the ontology layer; each relation type in the allowed relation type set contains a corresponding set of legal start entity types and a set of legal end entity types; Obtain the access location information of the structured state instance in the multi-level risk knowledge graph; the access location information includes the existing node identifier of the target flight corresponding to the structured state instance in the graph; Based on the set of allowed relation types, select the first type of available relation types that use the existing node identifier as the legal starting entity type and the structured state instance as the legal ending entity type. Based on the set of allowed relation types, a second type of available relation type is selected, which uses the entity type of the structured state instance as the legal starting entity type and the existing node identifier as the legal ending entity type. The first type of available relation type and the second type of available relation type are jointly defined as the set of relation types that the structured state instance is allowed to use when accessing the multi-level risk knowledge graph; The propagation direction from the existing node to the structured state instance is determined based on the first type of available relationship, and the propagation direction from the structured state instance to the existing node is determined based on the second type of available relationship.

[0030] The multi-level risk knowledge graph ontology layer definition file pre-inputs a set of attribute fields specific to all entity types under the flight operation scenario. Simultaneously, it inputs the data format constraint standards for each attribute field and the valid value range of the attribute field defined by the civil aviation operation management specifications. The three types of content in the ontology layer definition file are permanently stored for subsequent real-time data standardization and conversion.

[0031] It continuously receives real-time flight operation status raw data pushed by external transmission channels, and extracts all data field contents carried by the raw data segment by segment from the internal storage unit of the raw data. Simultaneously, it extracts the original field value contents corresponding to each data field. After the splitting and extraction operation is completed, it forms the raw data pair content composed of data fields and field values.

[0032] The system retrieves the pre-defined matching mapping relationship between the attribute field set recorded in the ontology layer of the multi-level risk knowledge graph and the external real-time data fields. It uses the complete consistency of text content as the matching criterion to complete the one-to-one matching of the two types of fields. The field values ​​obtained from the original data splitting are then completely filled into the attribute field storage location corresponding to the matching entity type inside the multi-level risk knowledge graph.

[0033] The system retrieves the data format constraint standards for each attribute field recorded in the ontology layer definition of the multi-level risk knowledge graph. Following the character type, character arrangement, and data organization style specified in the constraint standards, it performs a unified format rewriting operation on the field values ​​that have already been filled. It then retrieves the valid value range of the corresponding attribute fields within the ontology layer definition as the verification benchmark. After format rewriting, it performs a complete comparison and verification against the value range boundaries. If all field values ​​fall within the value range boundaries, the verification is considered successful; if any field value exceeds the value range boundaries, the verification is considered unsuccessful.

[0034] Collect all field values ​​that have undergone format conversion and passed legality checks, bind the entity type identifier information that matches the field values, and combine the entity type identifier with the compliant field values ​​to generate a structured state instance with a complete standardized structure.

[0035] Read the entity type identifier bound to the structured state instance, retrieve the complete set of allowed relationship types stored within the multi-level risk knowledge graph ontology layer definition, and completely store the single type of relationship, the set of legal starting entity types associated with the relationship type, and the set of legal ending entity types associated with the relationship type for each independent record entry within the set of allowed relationship types. Completely retain these three types of information for subsequent relationship filtering work.

[0036] The system retrieves all globally unique identifiers corresponding to generated entity nodes within the multi-level risk knowledge graph, matches the target flight-specific identification information bound to the structured state instance, extracts the successfully matched node identifiers to form access location information, and uses the existing node identifiers of the target flight in the graph as the core identification content.

[0037] The system iterates through all records within the allowed relation type set, using the complete inclusion of the entity type corresponding to the existing node identifier in the legal starting entity type set within the entry and the complete inclusion of the entity type corresponding to the structured state instance in the legal ending entity type set within the entry as dual matching criteria. Relation types that simultaneously satisfy both matching criteria are uniformly grouped into the first type of usable relation type.

[0038] The system iterates through all records within the allowed relation type set, using the complete inclusion of the entity type corresponding to the structured state instance in the legal starting entity type set within the entry and the complete inclusion of the entity type corresponding to the existing node identifier in the legal ending entity type set within the entry as dual matching criteria. Relation types that simultaneously satisfy both matching criteria are uniformly grouped into a second type of usable relation type.

[0039] The first type of available relation type and the second type of available relation type are integrated and collected. All the integrated relation type entries are uniformly classified as the set of allowed relation types that can be called and used during the process of accessing the multi-level risk knowledge graph in a structured state instance. During the access operation, only the relation types recorded in this set can be called to complete the node association construction.

[0040] Read the entity node matching logic corresponding to the first type of available relationship, fix the entity node corresponding to the existing node identifier in the graph as the link start end and the entity node corresponding to the structured state instance as the link end end, forming a one-way transmission direction from the existing node to the structured state instance. Read the entity node matching logic corresponding to the second type of available relationship, fix the entity node corresponding to the structured state instance as the link start end and the entity node corresponding to the existing node identifier in the graph as the link end end, forming a one-way transmission direction from the structured state instance to the existing node.

[0041] Retrieve all entity node storage records within the multi-level risk knowledge graph, and compare the flight identification information bound to each node with the target flight identifier information corresponding to the structured state instance. If no node in the graph storage records is bound to the same flight identification information, it is determined that there is no existing node identifier. Retrieve the exclusive flight identifier information of the target flight, add a blank node storage record in the entity node storage area of ​​the multi-level risk knowledge graph, enter the flight identifier information into the blank record to complete the initial flight node creation operation, and directly use the globally unique identifier stored inside the newly created initial flight node as the existing node identifier required by the process for subsequent link screening.

[0042] The beneficial effects are as follows: relying on the unified specifications preset in the ontology layer, the real-time flight data is standardized and parsed for compliance verification, and structured state instances with unified structure are stably generated. Based on the built-in relation constraint set of the ontology layer, entity types are matched and compliant associations are screened in two directions. The association types and bidirectional transmission paths that can be used by instances to access the graph are accurately limited. For scenarios where there is no corresponding flight node in the graph, the node is created and access identifiers are supplemented automatically. The entire process relies on fixed business matching benchmarks to constrain entity association logic, ensuring that the node access and link building between structured state instances and multi-level risk knowledge graphs have a unified and reproducible execution standard, avoiding the problem of invalid link generation caused by unconstrained entity associations.

[0043] Step 3: Based on the relationship type and the transmission direction, take the structured state instance as the input query, take the corresponding graph node matched and located in the multi-level risk knowledge graph as the starting point, and search for the associated path within the preset number of hops using the risk propagation relationship edge in the multi-level risk knowledge graph. In this embodiment of the invention, the step of using the structured state instance as input query based on the relationship type and the transmission direction, and starting from the corresponding graph node matched and located in the multi-level risk knowledge graph, performing an association path search within a preset hop count range using the risk propagation relationship edges in the multi-level risk knowledge graph, includes: Using the attribute field values ​​of the structured state instance as matching conditions, the type and attributes of all entity nodes in the multi-level risk knowledge graph are matched and located to obtain the search starting point of the structured state instance; Based on the relationship type and the transmission direction, relevant edges that conform to the transmission direction and whose relationship type belongs to the range of the relationship type are selected from the multi-level risk knowledge graph and used as available risk propagation relationship edges; Taking the search starting point as the current node, all adjacent nodes directly connected to the current node through an available risk propagation relationship edge are recorded as one-hop reachable nodes, and the available risk propagation relationship edges traversed from the search starting point to each of the one-hop reachable nodes are recorded as a one-hop path set; Using each of the aforementioned one-hop reachable nodes as the new current node, repeatedly execute the operation of expanding to adjacent nodes through the available risk propagation relationship edge, expanding outward hop by hop until the current expansion hop count reaches the upper limit of the preset hop count range, and record the newly added reachable nodes and corresponding paths during each hop expansion process; Paths whose endpoints are risk event nodes are selected from all recorded paths and output as the potential associated paths.

[0044] Extract all stored attribute field values ​​within the structured state instance as the basis for complete matching. Traverse all completed entity nodes within the multi-level risk knowledge graph, and sequentially compare the entity type information bound to each entity node with the entity type information of the structured state instance, and the values ​​of all stored attribute fields of the entity node with the corresponding attribute field values ​​of the structured state instance. When a single entity node simultaneously meets both the criteria of completely identical entity type and completely matched values ​​of all corresponding attribute fields, the entity node is determined as the search starting point corresponding to the structured state instance.

[0045] The system retrieves the set of allowed relationship types and the two types of unidirectional transmission direction constraint rules determined in the pre-process stage. It then traverses all the risk propagation relationship edges that have been built within the multi-level risk knowledge graph. Each risk propagation relationship edge is checked to see if the relationship type recorded is completely included in the set of allowed relationship types and whether the link transmission direction recorded by each risk propagation relationship edge is consistent with any one of the two types of unidirectional transmission direction constraint rules. Risk propagation relationship edges that meet both of the verification criteria are uniformly collected and marked as usable risk propagation relationship edges.

[0046] The search starting point that has been located is set as the current node. All entity nodes in the multi-level risk knowledge graph that are directly connected to the current node through a single available risk propagation relationship edge are retrieved. All retrieved entity nodes are uniformly classified and recorded as one-hop reachable nodes. The corresponding available risk propagation relationship edges traversed by the search starting point to reach each one-hop reachable node are completely retained. The complete links formed by the combination of all starting points, relationship edges, and one-hop reachable nodes are uniformly summarized and stored as a one-hop path set.

[0047] Each reachable node recorded in the one-hop path set is individually set as a brand new current node. All entity nodes in the multi-level risk knowledge graph that are directly connected to the brand new current node through a single available risk propagation relationship edge are retrieved and the corresponding new links are recorded. The expansion operation of retrieving adjacent nodes from the current node is repeated continuously. Each round of adjacent node retrieval operation corresponds to a hop count update. The hop count upper limit defined in advance according to the flight risk transmission business scenario is used as the stopping criterion. The expansion operation is terminated when the hop count is updated to be completely consistent with the hop count upper limit. The newly identified reachable nodes and corresponding complete link content in each round of expansion operation are completely retained.

[0048] Retrieve all link storage contents of the complete record of the hop count expansion stage, and use the risk event entity type predefined in the multi-level risk knowledge graph as the endpoint node matching judgment benchmark. Check the entity type of the entity node bound to the end of each link one by one. Links whose endpoint node entity type and risk event entity type are completely matched are uniformly collected and organized. All links that have been collected are uniformly marked and output as potential related paths.

[0049] The beneficial effects are as follows: relying on the complete attribute content of structured state instances, the system accurately matches the graph entities to obtain a dedicated search starting point; based on the relationship type and transmission direction limited by the previous steps, the system filters compliance risk propagation relationships to narrow the search scope; it expands the node links layer by layer outward according to a fixed hop limit and retains the complete path content of each level; it only selects the links of the terminal related risk event entities as potential related paths as output; it relies on standardized matching benchmarks to constrain the path search scope throughout the process, filters irrelevant entities and invalid transmission links, reduces redundant path retention, ensures that the risk transmission link search process fits the flight operation risk transmission logic, and improves the targeting of risk-related path identification.

[0050] Step 4: Based on the attribute values ​​of the structured state instance, trigger verification of the risk transmission rules in the searched potential association paths, and mark the potential association paths that pass the verification as valid risk transmission paths; In this embodiment of the invention, the step of triggering verification of risk transmission rules in the searched potential association paths based on the attribute values ​​of the structured state instance includes: Obtain the edge identifier corresponding to each risk propagation relationship edge traversed by the potential association path, and the node pair identifier corresponding to the adjacent node pair in each potential association path; Using the edge identifier and the node pair identifier as indexes, extract the risk transmission rules corresponding to each potential association path from the relation attribute storage area of ​​the multi-level risk knowledge graph; The risk transmission rule includes a condition field and an action field; the condition field records the judgment criteria that must be met to trigger the current risk transmission rule and the corresponding range of benchmark values, and the action field records the transmission result type corresponding to the current risk transmission rule after it is triggered; The risk transmission rule is structured and parsed, and the condition field is parsed into at least one condition expression to be determined, and the action field is parsed into the corresponding transmission action to be executed; the condition expression to be determined contains the entity attribute name to be compared and the comparison logic corresponding to the benchmark value range.

[0051] The step of triggering verification of risk transmission rules in the searched potential association paths based on the attribute values ​​of the structured state instance, and marking the potential association paths that pass the verification as valid risk transmission paths, includes: Based on the entity attribute name in the condition expression to be determined, extract the corresponding real-time attribute value from the attribute field of the structured state instance; Substitute the real-time attribute value into the condition expression to be determined, and compare the real-time attribute value with the benchmark value range according to the comparison logic to determine whether the real-time attribute value falls within the benchmark value range; When all the comparison results of the condition expressions to be determined corresponding to a potential association path are found to be included, the potential association path is determined to pass the trigger verification. When any one of the condition expressions to be determined is not found, the potential association path is determined to fail the trigger verification. Each potential associated path that passes the trigger verification is marked as a valid risk transmission path, and the transmission action to be executed corresponding to each valid risk transmission path is recorded as the transmission result label of the valid risk transmission path.

[0052] Retrieve all potential association paths output from the previous step one by one, traverse all risk propagation relationship edges arranged sequentially within a single potential association path, extract the pre-bound and stored edge identifier text content of a single risk propagation relationship edge, split the entity nodes arranged in pairs within a single potential association path to form independent node pairs, extract the pre-bound and stored node pair identifier text content for each pair of node pairs, and collect and store all edge identifiers and all node pair identifiers corresponding to a single potential association path for index retrieval.

[0053] The collected edge identifiers and node pair identifiers are combined to form a unique retrieval index text. The independent relation attribute storage area within the multi-level risk knowledge graph is accessed. The matching criterion is that the retrieval index text is completely consistent with the index identifier of the record in the storage area. The complete rule text that successfully matches the index is extracted from the storage area one by one. Each extracted rule text is individually marked as a risk transmission rule specific to the corresponding potential associated path.

[0054] Read the complete text content of each extracted risk transmission rule. The rule text is divided into two independent storage areas as condition fields and action fields, respectively. The condition field storage area contains the text content of all judgment criteria that must be met to start this risk transmission rule, as well as the corresponding defined benchmark value range boundary content. The action field storage area contains the text content of the transmission result classification that will be generated after this risk transmission rule meets all judgment criteria. The content of both types of fields is uniformly and fixedly stored for subsequent analysis and judgment operations.

[0055] The complete text content stored in the condition field of a single risk transmission rule is split and decomposed segment by segment. According to the fixed field separators within the text, at least one independent condition expression to be determined is generated. The fixed record within a single condition expression is the entity attribute name text that needs to be extracted for comparison, and the fixed comparison logic text that performs the comparison operation based on the matching benchmark value range. Simultaneously, the complete text content stored in the action field of the risk transmission rule is split and decomposed segment by segment. After splitting, the text of the transmission action to be executed, which can represent the output content of risk transmission, is obtained. The split condition expression to be determined and the transmission action to be executed are uniformly bound to the corresponding risk transmission rule.

[0056] Read the entity attribute name text of the complete record inside a single condition expression to be judged, traverse all the attribute field entries stored inside the structured state instance, and use the complete consistency between the field name text of the attribute field and the entity attribute name text as the matching judgment benchmark to locate the storage location of the successfully matched attribute field, extract the complete real-time attribute value content stored in the storage location and store it separately for subsequent comparison operations.

[0057] The system retrieves the fixed comparison logic text recorded within the condition expression to be judged and the baseline value range boundary content stored in conjunction with the risk transmission rule condition field. The extracted real-time attribute values ​​are then substituted into the condition expression to be judged to complete the content filling. The content comparison operation is carried out according to the fixed comparison rules recorded in the comparison logic text. If the real-time attribute value is completely within the upper and lower boundaries of the baseline value range, a comparison result marked as falling within the range is generated. If the real-time attribute value is outside the upper and lower boundaries of the baseline value range, a comparison result marked as not falling within the range is generated.

[0058] The comparison results corresponding to all the conditional expressions bound to a single potential related path are collected. The marked content of all the collected comparison results is checked one by one. If all comparison results are marked as falling into the category, the potential related path is judged to have passed the verification. If any comparison result in the collected content is marked as not falling into the category, the potential related path is judged to have failed the verification. A single potential related path can only produce two fixed verification judgment conclusions: pass or fail.

[0059] Retrieve all potential associated paths marked as having passed the verification and judgment conclusions, add a unique identifier to each such path to complete the marking process, and name the marked paths as valid risk transmission paths. Simultaneously retrieve the text of the transmission action to be executed generated by parsing the risk transmission rules corresponding to each valid risk transmission path, and bind this text as a unique identifier to the storage entry of the corresponding valid risk transmission path, and permanently retain it as the transmission result label associated with this path.

[0060] The beneficial effects include: accurately indexing and retrieving corresponding risk transmission rules by relying on the built-in edge identifiers and node pair identifiers of the path; separating independent judgment conditions and transmission actions by splitting the rule text; completing condition comparison and verification by relying on the built-in real-time attributes of the structured state instance; uniformly judging the validity of the path based on the matching results of all conditions; binding exclusive transmission result tags to compliant paths; completing risk transmission logic verification by relying on the built-in standardized rules of the graph throughout the process; strictly constraining the risk path screening criteria; eliminating invalid transmission links that do not meet the operating conditions; ensuring that the selected effective risk transmission paths are all in line with the real-time flight operation status; and improving the standardization and reliability of the risk transmission link judgment work.

[0061] Step 5: Based on the number of effective risk transmission paths and the corresponding risk event nodes and risk level information, and combined with the semantic distance of the target flight in the multi-level risk knowledge graph, calculate the comprehensive risk index of the target flight, and output the risk warning information corresponding to the comprehensive risk index.

[0062] In this embodiment of the invention, the step of calculating the comprehensive risk index of the target flight based on the number of effective risk transmission paths and the corresponding risk event nodes and risk level information, combined with the semantic distance of the target flight in the multi-level risk knowledge graph, includes: The total number of effective risk transmission paths is counted, and the risk level information of the risk event nodes corresponding to the effective risk transmission paths is read. Obtain the node position of the target flight node in the multi-level risk knowledge graph corresponding to the target flight, and calculate the semantic distance between the target flight node and the risk event node corresponding to each effective risk transmission path; Based on the risk level information corresponding to each effective risk transmission path, a risk contribution weight coefficient corresponding to each effective risk transmission path is determined, and the risk contribution weight coefficient is positively correlated with the risk level information. Based on the semantic distance corresponding to each effective risk transmission path, a distance attenuation coefficient is determined for each effective risk transmission path, and the distance attenuation coefficient is negatively correlated with the semantic distance; Based on the total number of effective risk transmission paths, the risk contribution weight coefficient, and the distance attenuation coefficient, the effective risk transmission paths are evaluated in multiple dimensions to obtain the comprehensive risk index of the target flight.

[0063] The formula for calculating the comprehensive risk index is as follows: ; In the formula, As a comprehensive risk index, The total number of effective risk transmission paths. For the first Risk level information of the risk event nodes corresponding to each effective risk transmission path. For the first The semantic distance corresponding to each effective risk transmission path For the first Risk contribution weight coefficients corresponding to each effective risk transmission path For the first The distance attenuation coefficient corresponding to an effective risk transmission path.

[0064] The output of the risk warning information corresponding to the comprehensive risk index includes: The comprehensive risk index is compared with the preset risk warning level range to determine the target value range of the comprehensive risk index; Based on the current risk warning level determined by the risk warning level corresponding to the target value range, historical risk management measures records that match the current risk warning level are parsed from the risk event entities of the multi-level risk knowledge graph as suggested mitigation measures. The current risk warning level, the key risk event nodes that trigger the effective risk transmission path, and the suggested mitigation measures are combined into a risk warning information output.

[0065] Iterate through all the valid risk transmission path entries marked in the previous step, count each one, and summarize all entries to obtain the total number of valid risk transmission paths. Retrieve the stored content of the risk event node bound to the end of a single valid risk transmission path, read the risk level text information pre-entered and stored inside the risk event node, and bind and store the risk level information matched by a single path with each path entry.

[0066] The complete storage location record of the target flight node is located within the multi-level risk knowledge graph. The complete storage content of the connection links between the recorded nodes in the graph is retrieved. The complete number of risk propagation relationship edges connected between two nodes is used as the semantic distance judgment benchmark. The total number of risk propagation relationship edges connected between the target flight node and the risk event node at the end of a single effective risk transmission path is calculated separately. The calculated total number of connected edges is used as the corresponding semantic distance content between the two types of nodes and is bound to the corresponding path for storage.

[0067] The risk level ranking benchmark predefined in the civil aviation operation specifications is retrieved. Based on a fixed ranking of risk levels from low to high corresponding hazard levels, the risk contribution weights of each level are matched. A matching relationship is established using the positive mapping standard between the five risk levels and their weights pre-stored in the map ontology layer. The higher the hazard level corresponding to a risk level, the higher the risk contribution weight obtained from the table. The larger the specific value, the more the risk level information bound to a single effective risk transmission path will match the corresponding level content, generate a risk contribution weight coefficient specific to that path, and complete the binding and storage.

[0068] A matching relationship is established based on the inverse mapping standard of semantic distance and attenuation coefficient values ​​pre-stored in the graph ontology layer. The more risk propagation relationship edges connected between the target flight node and the risk event node, the greater the semantic distance. The larger the value, the higher the distance attenuation coefficient obtained from the table. The smaller the specific value, the more likely it is to match the corresponding level of content based on the semantic distance bound to a single effective risk transmission path, generate a distance attenuation coefficient specific to this path, and complete the binding and storage.

[0069] The system retrieves three complete categories of stored information: the total number of effective risk transmission paths, the risk contribution weight coefficient for each path, and the distance attenuation coefficient for each path. The two types of coefficients corresponding to a single path are used to characterize the impact of a single path on the overall risk of the target flight, and the total number of paths is used to characterize the richness of risk sources. The three categories of information are then layered and integrated according to a fixed multi-dimensional evaluation logic. After integration, a comprehensive risk index for the target flight is generated and stored separately for subsequent early warning classification.

[0070] Standardized parameter mapping rules are pre-built into the ontology layer of the multi-level risk knowledge graph to determine... and The only numerical value. Weighting based on risk contribution. This invention adopts the five-level risk classification standard commonly used in the civil aviation field, dividing all risk event nodes into five levels: Level 1 Minor, Level 2 General, Level 3 Significant, Level 4 Major, and Level 5 Extremely Major. A fixed positive mapping relationship is preset: Level 1 Minor risk has a weight of 0.2, Level 2 General risk has a weight of 0.4, Level 3 Significant risk has a weight of 0.6, Level 4 Major risk has a weight of 0.8, and Level 5 Extremely Major risk has a weight of 1.0. After obtaining the risk level corresponding to a single path, the determined weight value can be directly read by referring to this preset standard. (Regarding distance attenuation coefficients...) This invention will use semantic distance Defined as the total number of risk propagation relationship edges connected between the target flight node and the risk event node. Combined with the preset maximum search hop count of 5, the semantic distance is limited to positive integers from 1 to 5. At the same time, the preset reverse mapping rules are: the attenuation coefficient is 1.0 when the semantic distance is 1, 0.8 when the semantic distance is 2, 0.6 when the semantic distance is 3, 0.4 when the semantic distance is 4, and 0.2 when the semantic distance is 5. After counting the total number of hops of the path links, a unique attenuation coefficient value can be matched. The following is a complete demonstration of the comprehensive risk index calculation process using a specific scenario. Assume that three effective risk transmission paths are identified. The first path corresponds to a level 5 extremely high risk with a semantic distance of 1, a weight of 1.0, and an attenuation coefficient of 1.0, resulting in a single path contribution of 1.0 multiplied by 1.0 equaling 1.0. The second path corresponds to a level 3 relatively high risk with a semantic distance of 2, a weight of 0.6, and an attenuation coefficient of 0.8, resulting in a single path contribution of 0.6 multiplied by 0.8 equaling 0.48. The third path corresponds to a level 2 general risk with a semantic distance of 3, a weight of 0.4, and an attenuation coefficient of 0.6, resulting in a single path contribution of 0.4 multiplied by 0.6 equaling 0.24. The sum of the three path contributions is 1.72, which is then divided by the total number of effective paths (3), resulting in a final comprehensive risk index of approximately 0.573 for the target flight.

[0071] The parameters corresponding to the comprehensive risk index are derived from the supporting business data of the multi-level risk knowledge graph and the standardized processing results of the entire process. All parameters are defined and extracted based on the civil aviation operation management specifications and the statistical records of historical flight risk events, and there is no situation of random human assignment.

[0072] The comprehensive risk index is generated by averaging the weighted summaries of all effective risk transmission paths. This value is used to uniformly quantify the overall risk level of the target flight and fully incorporates the three core assessment elements: effective risk transmission path, risk level, and node semantic distance.

[0073] The total number of effective risk transmission paths is obtained by a complete statistical analysis of the pre-verification process. The statistical process traverses all stored entries marked as effective risk transmission paths, counts them one by one, and summarizes them to obtain a fixed value. The statistical basis is the marking results after the potential related paths complete the risk transmission rule trigger verification.

[0074] The risk level information corresponding to a single effective risk transmission path is stored in a fixed storage area of ​​the risk event node within the multi-level risk knowledge graph. This content is defined based on the flight operation risk classification standards issued by the Civil Aviation Administration of China, combined with decades of historical flight accident and incident records, and is simultaneously bound to the corresponding path entry during the retrieval of potential related paths.

[0075] The semantic distance corresponding to a single effective risk transmission path is determined by the number of risk propagation relationship edges connected between the target flight node and the risk event node. The calculation process fully reads all the associated link records built between nodes in the multi-level risk knowledge graph, and uses the actual total number of linked edges as the sole basis for determining the semantic distance.

[0076] The risk contribution weight coefficient matching rule for a single effective risk transmission path is formed based on historical flight risk handling data. The same risk level corresponds to a fixed weight level. The higher the risk level, the higher the matching weight level. During matching, the preset weight content corresponding to the risk level information is directly read to complete the assignment.

[0077] The distance attenuation coefficient matching rule for a single effective risk transmission path is set based on the flight risk transmission and propagation law. The larger the semantic distance value, the longer the link between the risk event and the target flight, and the lower the matching attenuation level. During matching, the preset attenuation content corresponding to the semantic distance is directly read to complete the assignment.

[0078] Retrieve the risk warning level interval division records that have been pre-divided and fixedly stored in accordance with the Civil Aviation Flight Operation Risk Management and Control Standards. Each interval record has a fixed lower limit boundary and a fixed upper limit boundary. Compare the comprehensive risk index generated by the previous process with all interval boundary contents item by item. If the comprehensive risk index value is simultaneously greater than or equal to the lower limit boundary of a single interval and less than or equal to the upper limit boundary of a single interval, then the interval is determined to be the target value interval.

[0079] Retrieve the exclusive risk warning level identifier text bound to the target value range and define this text as the current risk warning level. Traverse all risk event entity storage entries within the multi-level risk knowledge graph, and read the historical risk handling measures records bound to each risk event entity. Use the complete consistency between the warning level identifier associated with the risk event entity and the current risk warning level text as the matching criterion, collect all successfully matched historical risk handling measures records, and uniformly mark them as suggested mitigation measures.

[0080] Extract the text of the current risk warning level confirmed within the process, the key risk event node identification content corresponding to all generated effective risk transmission paths, and the complete text of the collected suggested mitigation measures. Integrate these three types of content into the same information storage unit according to a fixed information layout structure. The integrated standardized information storage unit is then used as the complete risk warning information to be output externally.

[0081] The beneficial effects include: comprehensively statistically analyzing all compliance risk transmission links and extracting corresponding risk levels; determining the semantic correlation of nodes based on the total number of links between entities in the graph; establishing positive weight matching rules and reverse attenuation matching rules based on civil aviation regulations and historical risk data; and achieving a standardized comprehensive risk index by combining the total number of links, weights, and attenuation content across multiple dimensions. All assessment elements are taken from standardized business records in the graph, avoiding assessment bias caused by subjective manual assignment. The system matches corresponding warning levels based on preset standard intervals, automatically retrieves historical handling plans of the same level, and integrates risk levels, risk source nodes, and handling suggestions to form complete warning information. This enables integrated processing of flight risk quantitative assessment, classification judgment, and handling plan output. The entire process relies on a unified business benchmark to complete the matching and judgment of each link, ensuring that the risk assessment and warning output content has a stable and unified execution standard.

[0082] like Figure 2 The diagram shown is a functional block diagram of a knowledge graph-based flight operation risk early warning system provided in an embodiment of the present invention.

[0083] The knowledge graph-based flight operation risk early warning system M00 described in this invention can be installed in an electronic device. Depending on the functions implemented, the knowledge graph-based flight operation risk early warning system M00 may include a knowledge graph construction module M01, a state instantiation module M02, a risk path search module M03, a transmission path verification module M04, and a comprehensive risk assessment module M05. The modules described in this invention can also be referred to as units, which are a series of computer program segments that can be executed by the processor of an electronic device and perform a fixed function, stored in the memory of the electronic device.

[0084] In this embodiment, the functions of each module / unit are as follows: The knowledge graph construction module M01 is used to construct a multi-level risk knowledge graph of flight operations based on the data entities, entity attributes and relationships between entities in historical flight operation data, aircraft technical documents, airport operation rules and historical meteorological data. The state instantiation module M02 is used to convert real-time flight operation status data into structured state instances through the ontology layer definition in the multi-level risk knowledge graph, and to limit the relationship type and transmission direction between entities in the structured state instances in the multi-level risk knowledge graph. The risk path search module M03 is used to perform a query based on the relationship type and the transmission direction, taking the structured state instance as input, starting from the corresponding graph node matched and located in the multi-level risk knowledge graph, and searching for associated paths within a preset number of hops using the risk propagation relationship edges in the multi-level risk knowledge graph. The transmission path verification module M04 is used to trigger verification of risk transmission rules in the searched potential association paths based on the attribute values ​​of the structured state instance, and mark the potential association paths that pass the verification as valid risk transmission paths. The comprehensive risk assessment module M05 is used to calculate the comprehensive risk index of the target flight based on the number of effective risk transmission paths and the corresponding risk event nodes and risk level information, combined with the semantic distance of the target flight in the multi-level risk knowledge graph, so as to output the risk warning information corresponding to the comprehensive risk index.

[0085] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0086] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0087] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0088] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0089] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A knowledge graph-based method for early warning of flight operation risks, characterized in that, The method includes: Step 1: Based on historical flight operation data, aircraft technical documents, airport operation rules, and historical meteorological data, construct a multi-level risk knowledge graph for flight operations, including the data entities, entity attributes, and relationships between entities. Step 2: Convert real-time flight operation status data into structured state instances through the ontology layer definition in the multi-level risk knowledge graph, and define the relationship type and transmission direction between entities of the structured state instances in the multi-level risk knowledge graph; Step 3: Based on the relationship type and the transmission direction, take the structured state instance as the input query, take the corresponding graph node matched and located in the multi-level risk knowledge graph as the starting point, and search for the associated path within the preset number of hops using the risk propagation relationship edge in the multi-level risk knowledge graph. Step 4: Based on the attribute values ​​of the structured state instance, trigger verification of the risk transmission rules in the searched potential association paths, and mark the potential association paths that pass the verification as valid risk transmission paths; Step 5: Based on the number of effective risk transmission paths and the corresponding risk event nodes and risk level information, and combined with the semantic distance of the target flight in the multi-level risk knowledge graph, calculate the comprehensive risk index of the target flight, and output the risk warning information corresponding to the comprehensive risk index.

2. The knowledge graph-based flight operation risk early warning method as described in claim 1, characterized in that, The process involves constructing a multi-level risk knowledge graph for flight operations based on historical flight operation data, aircraft technical documents, airport operation rules, and historical meteorological data, including the relationships between data entities, entity attributes, and entities. Historical flight operation data, aircraft technical documents, airport operation rules, and historical meteorological data, after data cleaning, are integrated into standardized text data for domain entity recognition to obtain a candidate entity set for flight operations; Multiple candidate entities that refer to the same objective object in the candidate entity set are merged into an entity node and assigned a globally unique identifier; Based on the event records in the historical flight operation data and the associated descriptions in the aircraft technical documents, the direct relationships between each entity node are extracted and assigned relationship type labels; Based on the program logic in the airport operation rules and the meteorological impact records in the historical meteorological data, the indirect relationships between the entity nodes are parsed out and assigned transmission attribute labels; Based on the entity nodes, the relationship type labels, and the transmission attribute labels, a multi-level risk knowledge graph is constructed according to a hierarchical structure from high to low risk levels.

3. The knowledge graph-based flight operation risk early warning method as described in claim 1, characterized in that, The process of converting real-time flight operation status data into structured state instances through the ontology layer definition in the multi-level risk knowledge graph includes: The ontology layer definition of the multi-level risk knowledge graph includes the attribute field set corresponding to each entity type, the data format constraints of each attribute field, and the valid value range of each attribute field. Receive the real-time flight operation status data and parse out the data fields and corresponding field values ​​in the real-time flight operation status data; Based on the mapping relationship between the attribute field set and the data field in the ontology layer definition, the field values ​​are respectively filled into the attribute fields of the corresponding entity types in the multi-level risk knowledge graph; Based on the data format constraints, the format of the filled field value is converted, and the validity of the converted field value is verified according to the valid value range. The field values ​​that pass the validity check and the corresponding entity types are instantiated into structured state instances.

4. The knowledge graph-based flight operation risk early warning method as described in claim 3, characterized in that, The definition of the structured state instance's relationship type and transmission direction between entities in the multi-level risk knowledge graph includes: Based on the entity type of the structured state instance, the allowed relation type set is defined from the ontology layer; each relation type in the allowed relation type set contains a corresponding set of legal start entity types and a set of legal end entity types; Obtain the access location information of the structured state instance in the multi-level risk knowledge graph; the access location information includes the existing node identifier of the target flight corresponding to the structured state instance in the graph; Based on the set of allowed relation types, select the first type of available relation types that use the existing node identifier as the legal starting entity type and the structured state instance as the legal ending entity type. Based on the set of allowed relation types, a second type of available relation type is selected, which uses the entity type of the structured state instance as the legal starting entity type and the existing node identifier as the legal ending entity type. The first type of available relation type and the second type of available relation type are jointly defined as the set of relation types that the structured state instance is allowed to use when accessing the multi-level risk knowledge graph; The propagation direction from the existing node to the structured state instance is determined based on the first type of available relationship, and the propagation direction from the structured state instance to the existing node is determined based on the second type of available relationship.

5. The knowledge graph-based flight operation risk early warning method as described in claim 1, characterized in that, The process, based on the relationship type and the transmission direction, uses the structured state instance as input for querying, and takes the corresponding graph node matched and located in the multi-level risk knowledge graph as the starting point, and searches for associated paths within a preset number of hops along the risk propagation relationship edges in the multi-level risk knowledge graph, including: Using the attribute field values ​​of the structured state instance as matching conditions, the type and attributes of all entity nodes in the multi-level risk knowledge graph are matched and located to obtain the search starting point of the structured state instance; Based on the relationship type and the transmission direction, relevant edges that conform to the transmission direction and whose relationship type belongs to the range of the relationship type are selected from the multi-level risk knowledge graph and used as available risk propagation relationship edges; Taking the search starting point as the current node, all adjacent nodes directly connected to the current node through an available risk propagation relationship edge are recorded as one-hop reachable nodes, and the available risk propagation relationship edges traversed from the search starting point to each of the one-hop reachable nodes are recorded as a one-hop path set; Using each of the aforementioned one-hop reachable nodes as the new current node, repeatedly execute the operation of expanding to adjacent nodes through the available risk propagation relationship edge, expanding outward hop by hop until the current expansion hop count reaches the upper limit of the preset hop count range, and record the newly added reachable nodes and corresponding paths during each hop expansion process; Paths whose endpoints are risk event nodes are selected from all recorded paths and output as the potential associated paths.

6. The knowledge graph-based flight operation risk early warning method as described in claim 5, characterized in that, The step of triggering verification of risk transmission rules in the searched potential association paths based on the attribute values ​​of the structured state instance includes: Obtain the edge identifier corresponding to each risk propagation relationship edge traversed by the potential association path, and the node pair identifier corresponding to the adjacent node pair in each potential association path; Using the edge identifier and the node pair identifier as indexes, extract the risk transmission rules corresponding to each potential association path from the relation attribute storage area of ​​the multi-level risk knowledge graph; The risk transmission rule includes a condition field and an action field; the condition field records the judgment criteria that must be met to trigger the current risk transmission rule and the corresponding range of benchmark values, and the action field records the transmission result type corresponding to the current risk transmission rule after it is triggered; The risk transmission rule is structured and parsed, and the condition field is parsed into at least one condition expression to be determined, and the action field is parsed into the corresponding transmission action to be executed; the condition expression to be determined contains the entity attribute name to be compared and the comparison logic corresponding to the benchmark value range.

7. The knowledge graph-based flight operation risk early warning method as described in claim 6, characterized in that, The step of triggering verification of risk transmission rules in the searched potential association paths based on the attribute values ​​of the structured state instance, and marking the potential association paths that pass the verification as valid risk transmission paths, includes: Based on the entity attribute name in the condition expression to be determined, extract the corresponding real-time attribute value from the attribute field of the structured state instance; Substitute the real-time attribute value into the condition expression to be determined, and compare the real-time attribute value with the benchmark value range according to the comparison logic to determine whether the real-time attribute value falls within the benchmark value range; When all the comparison results of the condition expressions to be determined corresponding to a potential association path are found to be included, the potential association path is determined to pass the trigger verification. When any one of the condition expressions to be determined is not found, the potential association path is determined to fail the trigger verification. Each potential associated path that passes the trigger verification is marked as a valid risk transmission path, and the transmission action to be executed corresponding to each valid risk transmission path is recorded as the transmission result label of the valid risk transmission path.

8. The knowledge graph-based flight operation risk early warning method as described in claim 1, characterized in that, The step of calculating the comprehensive risk index of the target flight based on the number of effective risk transmission paths and the corresponding risk event nodes and risk level information, combined with the semantic distance of the target flight in the multi-level risk knowledge graph, includes: The total number of effective risk transmission paths is counted, and the risk level information of the risk event nodes corresponding to the effective risk transmission paths is read. Obtain the node position of the target flight node in the multi-level risk knowledge graph corresponding to the target flight, and calculate the semantic distance between the target flight node and the risk event node corresponding to each effective risk transmission path; Based on the risk level information corresponding to each effective risk transmission path, a risk contribution weight coefficient is determined for each effective risk transmission path; the risk contribution weight coefficient is positively correlated with the risk level information. Based on the semantic distance corresponding to each effective risk transmission path, a distance attenuation coefficient is determined for each effective risk transmission path, and the distance attenuation coefficient is negatively correlated with the semantic distance; Based on the total number of effective risk transmission paths, the risk contribution weight coefficient, and the distance attenuation coefficient, the effective risk transmission paths are evaluated in multiple dimensions to obtain the comprehensive risk index of the target flight.

9. The knowledge graph-based flight operation risk early warning method as described in claim 8, characterized in that, The output of the risk warning information corresponding to the comprehensive risk index includes: The comprehensive risk index is compared with the preset risk warning level range to determine the target value range of the comprehensive risk index; Based on the current risk warning level determined by the risk warning level corresponding to the target value range, historical risk management measures records that match the current risk warning level are parsed from the risk event entities of the multi-level risk knowledge graph as suggested mitigation measures. The current risk warning level, the key risk event nodes that trigger the effective risk transmission path, and the suggested mitigation measures are combined into a risk warning information output.

10. A knowledge graph-based flight operation risk early warning system, characterized in that, The system for implementing the knowledge graph-based flight operation risk early warning method as described in claim 1 includes: The knowledge graph construction module is used to construct a multi-level risk knowledge graph for flight operations based on data entities, entity attributes, and relationships between entities in historical flight operation data, aircraft technical documents, airport operation rules, and historical meteorological data. The state instantiation module is used to convert real-time flight operation status data into structured state instances through the ontology layer definition in the multi-level risk knowledge graph, and to limit the relationship type and transmission direction between entities of the structured state instances in the multi-level risk knowledge graph. The risk path search module is used to query the structured state instance as input based on the relationship type and the transmission direction, and to search for associated paths within a preset number of hops, starting from the corresponding graph node matched and located in the multi-level risk knowledge graph and using the risk propagation relationship edges in the multi-level risk knowledge graph. The transmission path verification module is used to trigger verification of risk transmission rules in the searched potential association paths based on the attribute values ​​of the structured state instance, and mark the potential association paths that pass the verification as valid risk transmission paths. The comprehensive risk assessment module is used to calculate the comprehensive risk index of the target flight based on the number of effective risk transmission paths and the corresponding risk event nodes and risk level information, combined with the semantic distance of the target flight in the multi-level risk knowledge graph, so as to output the risk warning information corresponding to the comprehensive risk index.

Images