Knowledge graph-based vehicle-mounted train control data multi-dimensional association method and related device

Abstract

The application discloses a kind of based on knowledge graph's vehicle-mounted train control data multidimensional association method and related device, it is related to vehicle-mounted train control data analysis technical field, first obtain braking, speed core data and other associated auxiliary data, according to the correlation degree with core data is divided into core and auxiliary association layer;After defining entity, accurately extract each entity by the mixed mode of rule+machine learning, extract associated relationship in combination with attention mechanism model and field dictionary;Subsequently, build vehicle-mounted train control data knowledge graph after knowledge fusion and store in graph database;Realize braking and speed real-time association, abnormal traceability and other multidimensional analysis based on graph, output structured report and visual graph, adapt running monitoring, fault diagnosis and operation optimization scene, train stops after saving graph data.The method realizes the accurate association of multi-source heterogeneous data, greatly shortens fault processing time, reduces train energy consumption, effectively improves train operation safety, reliability and operation efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vehicle train control data analysis, in particular to a vehicle train control data multi-dimensional association method based on a knowledge graph and related devices. BACKGROUND

[0002] With the development of train operation control systems, a large amount of multi-source heterogeneous data is generated during the operation of vehicle train control equipment, covering brake system data, speed data, communication data, equipment state data, and fault log data. Current vehicle train control data processing mostly uses single dimension or simple association methods, which cannot fully mine the deep association of brake data and speed data, and the multi-source data integration capability is insufficient, resulting in low data value utilization rate. When there is abnormal speed fluctuation or emergency braking warning, it is difficult to quickly trace the associated data, and it is difficult to provide comprehensive support for fault positioning, affecting the safety and reliability of train operation. SUMMARY

[0003] The purpose of the present application is to provide a vehicle train control data multi-dimensional association method based on a knowledge graph and related devices, which can effectively improve the safety and reliability of train operation.

[0004] To achieve the above purpose, the present application provides the following solutions:

[0005] In a first aspect, the present application provides a vehicle train control data multi-dimensional association method based on a knowledge graph, comprising the following steps:

[0006] Cycle to determine the train operation state every preset time length, if the train is not stopped, obtain the multi-source vehicle train control data of the vehicle train control system; the multi-source vehicle train control data includes brake core data, speed core data and associated auxiliary data; the associated auxiliary data includes communication data, equipment state data, environment data and fault log data.

[0007] According to the data type of each data in the multi-source vehicle train control data and the association degree of the core data, each data is classified to obtain a core data layer and an auxiliary association layer; the core data is brake or speed related data.

[0008] Based on the core data layer and the auxiliary association layer, the core entity and the auxiliary entity are defined, and a hybrid method of mainly using rules and supplemented by machine learning is used to extract the core entity and the auxiliary entity from the multi-source vehicle train control data.

[0009] A relationship extraction model based on an attention mechanism is used to extract the association relationship between each entity in combination with a vehicle train control field dictionary, and the result is corrected by artificial rule verification to obtain the association relationship contained in the knowledge graph.

[0010] The extracted entities and the associated relationships are knowledge fused to construct a train control data knowledge graph and store the train control data knowledge graph in a graph database.

[0011] Based on the train control data knowledge graph, data multi-dimensional association analysis is performed, and the association analysis result is output and adapted to a specific application scenario.

[0012] If the train stops, the data collection is stopped and the latest train control data knowledge graph is saved.

[0013] Optionally, the core data layer includes brake core data and speed core data; the auxiliary association layer includes communication data, device state data, environment data, and fault log data; and a hybrid method of mainly using rules and supplemented by machine learning is used to extract core entities and auxiliary entities from multi-source train control data, including the following steps:

[0014] For structured data in multi-source train control data, core entities are directly extracted based on a preset extraction rule. The preset extraction rule is to extract entities of corresponding types according to field identifiers. The core entities include brake pressure entities, brake command level entities, brake actuator entities, real-time speed entities, target speed entities, and speed change rate entities.

[0015] For unstructured data in multi-source train control data, auxiliary entities are extracted using an entity extraction model. The entity extraction model is a model obtained by training an LSTM model using a plurality of pieces of multi-source train control data annotated with entity category labels as training corpus. The auxiliary entities include command entities, coordination information entities, on-board host entities, speed sensor entities, brake valve entities, track slope entities, external temperature entities, brake pressure abnormal entities, and speed sensor fault entities.

[0016] Optionally, the extracted entities and the associated relationships are knowledge fused, including the following steps:

[0017] Entity linking is used to match and merge repeated entities into the same entity based on attribute similarity or semantic similarity between the repeated entities.

[0018] Relationship reconciliation is used to correct the conflicting associated relationships based on a preset timestamp priority rule. The timestamp priority rule is that the timestamp priority of the core data layer is higher than that of the auxiliary association layer.

[0019] Optionally, a train control data knowledge graph is constructed and stored in a graph database, including the following steps:

[0020] Each entity is taken as a node of the train control data knowledge graph; the node attributes include collection timestamp, numerical value / state description, and data source.

[0021] Each association is taken as a directed edge between nodes of the train control data knowledge graph; the edge attributes include association confidence and historical matching times.

[0022] The graph database storing the train control data knowledge graph is deployed on the vehicle server to support real-time data multi-dimensional association analysis requests.

[0023] Optionally, the association relationships included in the knowledge graph include the association relationship between braking and speed, the association relationship between core entities and auxiliary entities, and the association relationship between auxiliary entities; the association relationship between braking and speed includes: braking pressure entity associated with real-time speed entity, braking instruction level entity associated with target speed entity, and braking actuator entity associated with speed change rate entity; the association relationship between core entities and auxiliary entities includes: instruction-related core entities associated with instruction entities and coordination information entities, and state-related core entities associated with each auxiliary entity except instruction entities and coordination information entities; instruction-related entities include braking instruction level entity and target speed entity; state-related core entities include braking pressure entity, braking actuator entity, real-time speed entity, and speed change rate entity.

[0024] Braking and speed real-time association, specifically including the following steps:

[0025] Real-time collection of braking pressure data, when the braking pressure data reaches the preset braking pressure threshold, the real-time speed node associated with the braking pressure node is traversed through the train control data knowledge graph.

[0026] Call the historical association rule, when the change rule of the real-time speed node is judged to be out of the rule range, mark it as "braking-speed association anomaly" and trigger the abnormal data association traceability process.

[0027] Speed deviation and braking instruction traceability, specifically including the following steps:

[0028] When the speed deviation value is detected to be greater than the preset speed deviation threshold, the braking actuator node and the braking instruction level node are traced back through the train control data knowledge graph.

[0029] Check whether the braking instruction is issued and the braking level is up to standard, and trace back the instruction node associated with the braking instruction level node to confirm whether the target speed is not updated due to instruction transmission delay.

[0030] Abnormal data correlation traceability, specifically: when the real-time speed fluctuates abnormally within 10s, the speed sensor state, brake actuator state and external temperature node are associated based on the vehicle train control data knowledge graph, and the root cause of the abnormality is determined through preset inference rules.

[0031] Historical data correlation mining, specifically: traversing the historical data of any slope line in the vehicle train control data knowledge graph within three months, mining the historical correlation rules of track slope, brake pressure and speed drop amplitude, as the data basis for brake strategy optimization of slope line.

[0032] Optionally, the data multi-dimensional correlation analysis includes brake and speed real-time correlation, speed deviation and brake instruction traceability, abnormal data correlation traceability and historical data correlation mining; the correlation analysis results include structured reports and visualized graphs; the specific application scenarios include train operation monitoring scenarios, fault diagnosis scenarios and operation optimization scenarios; outputting the correlation analysis results and adapting to specific application scenarios, specifically including the following steps:

[0033] Output the correlation analysis results in the form of structured reports, and output the visualized graphs; the structured reports include brake and speed correlation state reports and abnormal data traceability reports; in the visualized graphs, the nodes of core entities are red, the nodes of auxiliary entities are blue, the nodes of track slope entities, external temperature entities, brake pressure abnormal entities and speed sensor fault entities are green, and the directed edges are labeled with relationship types and confidence.

[0034] In the train operation monitoring scenario, when there is an abnormal mark, the vehicle terminal issues a sound and light alarm with beeping and flashing.

[0035] In the fault diagnosis scenario, output the abnormal data traceability report to clearly identify the fault-related components.

[0036] In the operation optimization scenario, output the historical correlation rules and support parameter adjustment of the vehicle train control system to reduce train energy consumption.

[0037] In a second aspect, the present application provides a vehicle train control data multi-dimensional correlation device based on a knowledge graph, comprising the following functional modules:

[0038] A multi-source vehicle train control data acquisition module is used to cyclically judge the train running state every preset time length, and if the train is not stopped, the multi-source vehicle train control data of the vehicle train control system is acquired; the multi-source vehicle train control data includes brake core data, speed core data and associated auxiliary data; the associated auxiliary data includes communication data, device state data, environment data and fault log data; if the train is stopped, data acquisition is stopped and the latest vehicle train control data knowledge graph is saved.

[0039] The multi-source data classification module is configured to classify each data in the multi-source vehicle-mounted train control data according to a degree of association between the data type of each data and the core data, to obtain a core data layer and an auxiliary association layer; the core data is brake or speed related data.

[0040] The entity definition and extraction module is configured to define core entities and auxiliary entities based on the core data layer and the auxiliary association layer respectively, and to extract the core entities and the auxiliary entities from the multi-source vehicle-mounted train control data in a hybrid manner mainly based on rules and supplemented by machine learning.

[0041] The association relationship extraction module is configured to extract association relationships between entities by using a relationship extraction model based on an attention mechanism in combination with a vehicle-mounted train control field dictionary, and to correct the results by manual rule checking, to obtain association relationships included in the knowledge graph.

[0042] The knowledge graph construction module is configured to perform knowledge fusion on the extracted entities and association relationships, to construct a vehicle-mounted train control data knowledge graph, and to store the vehicle-mounted train control data knowledge graph in a graph database; the knowledge fusion includes processing of entity duplication and association relationship conflict.

[0043] The data multi-dimensional association analysis module is configured to perform data multi-dimensional association analysis based on the vehicle-mounted train control data knowledge graph, to output association analysis results and to adapt to specific application scenarios.

[0044] In a third aspect, the present application provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the knowledge graph-based vehicle-mounted train control data multi-dimensional association method described above.

[0045] In a fourth aspect, the present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the steps of the knowledge graph-based vehicle-mounted train control data multi-dimensional association method described above.

[0046] In a fifth aspect, the present application provides a computer program product comprising a computer program, wherein the computer program is executed by a processor to implement the steps of the knowledge graph-based vehicle-mounted train control data multi-dimensional association method described above.

[0047] According to the embodiments provided in the present application, the following technical effects are disclosed:

[0048] The application provides a knowledge graph-based multi-dimensional association method for vehicle train control data and a related device. In the method, multi-source vehicle train control data of a vehicle train control system is first acquired, core data and auxiliary data are comprehensively covered, the problem of scattered data collection sources in the prior art is solved, and a complete data foundation is laid for subsequent deep association mining; then, each data is classified according to the correlation degree with the core data, a core data layer and an auxiliary association layer are divided, the data priority and the correlation weight are determined, the low association efficiency caused by disordered data is avoided, and the pertinence and accuracy of data processing are improved; then, core entities and auxiliary entities are extracted from the multi-source vehicle train control data by using a hybrid mode mainly based on rules and supplemented by machine learning, and then the core association between entities is focused by using an attention mechanism, the professional scene is adapted by using a domain dictionary, and errors are corrected by manual verification, so that the problems of insufficient deep association mining and low accuracy in the prior art are solved, and a precise entity association network is constructed; repeated entities are combined by entity linking, conflicts are corrected by relationship reconciliation, data redundancy and contradictions are eliminated, and the consistency and integrity of the knowledge graph are ensured; the graph database is deployed on the vehicle server, real-time association query is supported, the problem of slow response in traditional data storage and association query is solved, and an efficient carrier is provided for rapid data analysis; multi-dimensional association analysis is performed according to the knowledge graph, real-time monitoring of braking and speed, rapid tracing of abnormal sources and deep mining of historical rules are realized, structured reports and visual graphs make the results intuitive and easy to understand, and multi-scene adaptation meets the diversified needs of operation monitoring, fault diagnosis and operation optimization. The train operation state is cyclically judged every preset time length, if the train does not stop, the above steps are repeated to ensure real-time updating of data association and continuous dynamic support for train operation; if the train stops, data collection is stopped and the latest knowledge graph is saved, the knowledge graph is independently stored to support power recovery, complete data support is provided for subsequent historical data mining and system restart, and data continuity is ensured. BRIEF DESCRIPTION OF DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort.

[0050] Figure 1 A flowchart of a knowledge graph-based multi-dimensional association method for vehicle train control data provided by an embodiment of the present application.

[0051] Figure 2 A data structure diagram of multi-source vehicle train control data in a knowledge graph-based multi-dimensional association method for vehicle train control data provided by an embodiment of the present application.

[0052] Figure 3 A functional module schematic diagram of a knowledge graph-based vehicle-mounted train control data multi-dimensional association device is provided for an embodiment of the present application.

[0053] Figure 4 A structural schematic diagram of a computer device is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0054] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0055] Currently, the processing of vehicle-mounted train control data mostly adopts a single dimension or simple association manner, for example, only abnormal detection is performed on speed data, or only linear matching is performed between braking instructions and speed changes. Such processing manner has obvious defects: on the one hand, the deep association between braking data and speed data cannot be fully mined, for example, the nonlinear relationship between the speed change rate of braking pressure and the speed drop amplitude, the difference law of speed adjustment under different braking modes, etc., resulting in low data value utilization rate. On the other hand, the integration capability for multi-source data is insufficient, when there is abnormal fluctuation of train speed or emergency braking warning, it is difficult to quickly trace the associated communication instructions, device states and other data, and it is impossible to provide comprehensive data support for fault positioning, thereby affecting the safety and reliability of train operation.

[0056] And as a structured semantic network, the knowledge graph has strong entity association and relationship mining capability, and can convert multi-source heterogeneous data into knowledge nodes and relationship edges with semantic association. Applying the knowledge graph technology to the field of vehicle-mounted train control data association can effectively solve the problems of current data association fragmentation and insufficient deep relationship mining, especially constructing an association system with braking and speed data as the core, which can further improve the application value of vehicle-mounted train control data and provide intelligent support for train operation control. The present application aims to solve the problems of traditional data processing fragmentation and low value utilization rate by using the knowledge graph.

[0057] In order to make the above-mentioned purposes, features and advantages of the present application more apparent and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0058] A knowledge graph-based vehicle-mounted train control data multi-dimensional association method provided by an embodiment of the present application includes the following steps in one exemplary embodiment, as shown in Figure 1

[0059] ​A1, acquiring multi-source train control data of the train control system; the multi-source train control data includes braking core data, speed core data, and associated auxiliary data; the associated auxiliary data includes communication data, device state data, environmental data, and fault log data.

[0060] In an exemplary embodiment, the braking control unit, speed measurement unit, communication unit, and device state monitoring unit of the train control system collect data in real time and form multi-source train control data. All data are synchronously labeled based on a time axis, with a timestamp accuracy of within 100 ms. The specific data items include braking core data, speed core data, and associated auxiliary data.

[0061] The braking core data includes brake cylinder data, wheel cylinder data, brake command level data, emergency brake trigger signal data, brake valve switch state data, and brake response delay time data.

[0062] The speed core data includes train real-time speed data fused from multiple speed sensors, target speed data issued by the ground, train acceleration / deceleration data, and speed deviation value data (difference data between real-time speed and target speed).

[0063] The associated auxiliary data includes interactive instruction data with the ground, speed coordination information data communicated with the ground, vehicle host information data, speed sensor working data, track slope data, external temperature data, historical braking fault record data, and speed sensor abnormal log data.

[0064] A2, classifying each data according to the data type of each data in the multi-source train control data and the correlation degree of the core data, to obtain a core data layer and an auxiliary correlation layer; as shown in Figure 2 The core data is braking or speed-related data. The core data layer includes braking core data and speed core data. The auxiliary correlation layer includes communication data, device state data, environmental data, and fault log data.

[0065] Specifically, the core data layer is the core associated subject of the knowledge graph, and the correlation degree weight of the entire core data layer is set to 0.7. Specifically, in the core data layer, the weight of the braking core data is set to 0.6, and the weight of the speed core data is set to 0.4.

[0066] The auxiliary association layer includes communication data (including the interaction instruction data collected in step A1 and the speed coordination information data communicated), device state data (including the vehicle host information data collected in step A1 and the speed sensor working data), environmental data (including the track slope data collected in step A1 and the external temperature data), and fault log data (including the historical braking fault record data collected in step A1 and the speed sensor abnormal log data), as the associated supplement of the core data, and the association weight is set according to the association closeness with the core data.

[0067] In the auxiliary association layer, the communication data is closely associated with the target speed value, and the weight is set to 0.3. In the device state data, the sensor working voltage is closely associated with the collection accuracy of the real-time speed, and the weight is set to 0.2. In the environmental data, the track slope is closely associated with the braking pressure demand, and the weight is set to 0.2, while the external temperature in the environmental data has a lower association degree, and the weight is set to 0.1. The weight of the fault log data is set to 0.2.

[0068] A3, based on the core data layer and the auxiliary association layer, the core entity and the auxiliary entity are defined, and a mixed mode of rule-based and machine learning-based is adopted to extract the core entity and the auxiliary entity from the multi-source vehicle-mounted train control data; specifically, the extraction priority is that the core entity is prior to the auxiliary entity, and the structured data is prior to the unstructured data, the core entity includes the braking pressure entity (corresponding to the brake cylinder data and the wheel cylinder data collected in step A1), the braking instruction level entity (corresponding to the braking instruction level data and the braking response delay time data collected in step A1), the braking actuator entity (corresponding to the emergency braking trigger signal data and the brake valve switch state data collected in step A1), the real-time speed entity (corresponding to the train real-time running speed data collected in step A1), the target speed entity (corresponding to the target speed data collected in step A1), and the speed change rate entity (corresponding to the speed deviation value data and the train acceleration / deceleration data collected in step A1); the auxiliary entity includes the instruction entity, the coordination information entity, the vehicle host entity, the speed sensor entity, the brake valve entity, the track slope entity, the external temperature entity, the braking pressure abnormal entity, and the speed sensor fault entity.

[0069] Specifically in this embodiment, in step A3, the core entity and the auxiliary entity are extracted from the multi-source vehicle-mounted train control data by using a mixed mode of rule-based and machine learning-based, which specifically includes the following steps:

[0070] A31, for the structured data in the multi-source vehicle-mounted train control data, the core entity is directly extracted based on a preset extraction rule; the preset extraction rule is to extract the entity of a corresponding type according to a field identifier. The structured data is stored in a relational database, the field identifier is clear, and each core entity can be directly extracted based on the field identifier.

[0071] A32、For unstructured data in multi-source vehicle train control data, an entity extraction model is used to extract auxiliary entities; the entity extraction model is a model obtained by training an LSTM model with a training corpus of a plurality of pieces of multi-source vehicle train control data annotated with entity category labels.

[0072] In this embodiment, the training corpus is 10,000 pieces of multi-source vehicle train control data with completed annotations, the annotation format includes entity category labels, and after the LSTM model is trained, the entity extraction accuracy on the test set is ≥95%, and the recall rate is ≥90%.

[0073] In one exemplary embodiment, first, three brake entities are extracted according to the field identifier, including brake pressure entities (corresponding to brake cylinder data and wheel cylinder data), brake command level entities (corresponding to brake command level data and brake response delay time data), and brake actuator entities (corresponding to emergency brake trigger signal data and brake valve switch state data), and then three speed entities are extracted according to the remaining field identifier, including real-time speed entities (corresponding to train real-time running speed data), target speed entities (corresponding to target speed data issued by the ground instruction), and speed change rate entities (corresponding to speed deviation value data and train acceleration / deceleration data).

[0074] Second, unstructured data extraction is performed, including communication data (including the interaction instruction data with the ground collected in step A1 and the speed coordination information data communicated), device state data (including the vehicle host information data collected in step A1 and the speed sensor working data), environment data (including the track slope data collected in step A1 and the external temperature data), and fault log data (including the historical brake fault record data collected in step A1 and the speed sensor abnormal log data), as an associated supplement to the structured entities.

[0075] A4、A relationship extraction model based on an attention mechanism is used in combination with a vehicle train control field dictionary to extract the associated relationships between entities, and the results are corrected through artificial rule verification to obtain the associated relationships included in the knowledge graph; the associated relationships include the associated relationships between braking and speed, the associated relationships between core entities and auxiliary entities, and the associated relationships between auxiliary entities.

[0076] Specifically in this embodiment, the first type of associated relationship is the associated relationship between braking and speed, including: the brake pressure entity is associated with the real-time speed entity, the brake command level entity is associated with the target speed entity, and the brake actuator entity is associated with the speed change rate entity.

[0077] The second type of association relationship is the association relationship between the core entity and the auxiliary entity, including: the instruction-related core entity associated with the instruction entity and the coordination information entity, and the state-related core entity associated with each auxiliary entity (including the vehicle host entity, the speed sensor entity, the track slope entity, the external temperature entity, the brake pressure abnormal entity, and the speed sensor fault entity) except the instruction entity and the coordination information entity; the instruction-related entity includes the brake instruction level entity and the target speed entity; and the state-related core entity includes the brake pressure entity, the brake actuator entity, the real-time speed entity, and the speed change rate entity.

[0078] The third type of association relationship is the association relationship between the auxiliary entities, including the instruction entity associated with the coordination information entity, the vehicle host entity associated with the speed sensor entity, the track slope entity, the external temperature entity, the brake pressure abnormal entity, and the speed sensor fault entity.

[0079] Specifically, the input of the relationship extraction model is each entity extracted in step A3 and the context data. These entities with relationships are referred to as entity pairs. The local features of the entity pairs are extracted by the relationship extraction model, and the attention mechanism is used to focus on the various types of association relationships defined in step A4, such as the numerical matching degree of the brake change amplitude and the speed change rate, and other key association features.

[0080] After the extraction is completed, the results are corrected by manual rules. In an exemplary embodiment, the manual rule is set as: the brake instruction trigger time must be later than the speed deviation threshold time, and the time difference < 1s. For example, when the brake entity is triggered, the real-time speed entity in the speed entity appears a real-time speed increase within a certain small time difference (time difference < 1s), which is considered as incorrect data. By this manual rule, the incorrect relationships that do not meet the rule are removed, and the final association relationship extraction accuracy is ≥95%.

[0081] A5, knowledge fusion is performed on the extracted entities and association relationships to construct a train control data knowledge graph and store the train control data knowledge graph in a graph database; the knowledge fusion includes processing of entity duplication and association relationship conflict. For example, the brake instruction level entity and the coordination information entity issue similar brake instructions within a certain small time difference (time difference < 100ms), which is considered as the same repeated entity. In addition, when the brake pressure entity appears a brake pressure increase, the real-time speed entity in the speed entity appears a real-time speed increase within a certain small time difference (time difference < 1s), which is considered as a relationship conflict.

[0082] Specifically, in this embodiment, the “knowledge fusion of the extracted entities and association relationships” in step A5 specifically includes the following steps:

[0083] A51, adopt entity linking, match based on attribute similarity or semantic similarity between repeated entities, and merge repeated entities into the same entity. For example, based on entity attribute similarity matching, when the error of real-time speed values collected by two sensors is <1 km / h, it is determined that they are the same real-time speed entity. Monitoring feedback brake state normal and fault log record brake no abnormal, semantic similarity ≥ 90%, merged into the same brake state entity.

[0084] A52, adopt relationship reconciliation, modify the conflicting association relationship based on the preset timestamp priority rule; the timestamp priority rule is that the timestamp priority of the core data layer is higher than that of the auxiliary association layer. For example, the difference between the target speed displayed by the communication instruction and the record of the vehicle host is corrected based on the timestamp of the brake control unit (the timestamp priority of the core data layer is higher than that of the auxiliary association layer, specifically, the timestamp of the brake core data is better than that of the speed core data, and both are better than that of the auxiliary association data, and the timestamp of the high-precision acquisition equipment is used as the reference), and the relationship of the vehicle host data is corrected to the instruction associated target speed.

[0085] Further, the "constructing the vehicle train control data knowledge graph and storing the vehicle train control data knowledge graph in the graph database" in step A5 specifically includes the following steps:

[0086] A53, each entity is taken as a node of the vehicle train control data knowledge graph; the node attribute includes acquisition timestamp, numerical value / state description and data source.

[0087] A54, each association relationship is taken as a directed edge between nodes of the vehicle train control data knowledge graph; the edge attribute includes association confidence and historical matching times.

[0088] A55, the graph database storing the vehicle train control data knowledge graph is deployed on the vehicle server, supporting real-time data multi-dimensional association analysis request. The graph database is deployed on the vehicle server, supporting real-time association query request, and the response time is short.

[0089] A6, based on the vehicle train control data knowledge graph, data multi-dimensional association analysis is performed, and the association analysis result is output and adapted to specific application scenarios; the data multi-dimensional association analysis includes brake and speed real-time association, speed deviation and brake instruction traceability, abnormal data association traceability and historical data association mining; the association analysis result includes structured report and visual graph; the specific application scenarios include train operation monitoring scenario, fault diagnosis scenario and operation optimization scenario.

[0090] In an exemplary embodiment, brake and speed real-time association specifically includes the following steps:

[0091] B1. Real-time acquisition of braking pressure data. When the braking pressure data reaches the preset braking pressure threshold, the real-time speed nodes associated with the braking pressure node are traversed through the on-board train control data knowledge graph.

[0092] B2. Call historical correlation patterns. When the change pattern of the real-time speed node exceeds the pattern range, mark it as "braking-speed correlation anomaly" and trigger the abnormal data correlation tracing process.

[0093] Speed ​​deviation and braking command tracing specifically includes the following steps:

[0094] C1. When the detected speed deviation value is greater than the preset speed deviation threshold, the brake actuator node and brake command level node are traced back in reverse through the on-board train control data knowledge graph.

[0095] C2. Verify whether the braking command has been issued and whether the braking level meets the standard. At the same time, trace the command node associated with the braking command level node to confirm whether the target speed has not been updated due to command transmission delay.

[0096] The abnormal data correlation and source tracing process involves: when real-time speed fluctuates abnormally within 10 seconds, the system uses a knowledge graph of onboard train control data to correlate speed sensor status, brake actuator status, and external temperature nodes, and then uses preset inference rules to determine the root cause of the anomaly. For example, through rule-based inference, it is deduced that abnormal sensor voltage causes speed acquisition deviation, brake valve sticking causes unstable speed regulation, and low temperature exacerbates actuator malfunction, thus determining the root cause of the anomaly as sensor failure.

[0097] Historical data correlation mining specifically involves: traversing the historical data of any slope line within three months in the onboard train control data knowledge graph to mine historical correlation patterns of track gradient, braking pressure, and speed reduction, which serve as the data basis for optimizing braking strategies for slope lines.

[0098] In an exemplary embodiment, step A6, "outputting the correlation analysis results and adapting them to specific application scenarios," specifically includes the following steps:

[0099] The system outputs correlation analysis results in a structured report format, along with a visual graph. The structured report includes a braking and speed correlation status report (containing real-time correlation status, anomaly markers, and confidence levels) and an anomaly data tracing report (containing anomaly nodes, related entities, and fault cause inferences). In the visual graph, nodes representing core entities are red, nodes representing auxiliary entities are blue, and nodes representing track gradient, external temperature, abnormal braking pressure, and speed sensor faults are green. Directed edges are labeled with relationship types and confidence levels. The system supports real-time display on vehicle terminals and remote retrieval by ground control centers.

[0100] In the train operation monitoring scenario, when there is an abnormal mark, the on-board terminal issues a sound and light alarm of beeping and flashing to assist the driver in judging the running safety. In the fault diagnosis scenario, an abnormal data traceability report is output to clearly determine the associated components of the fault, guide the maintenance personnel to prioritize troubleshooting, and shorten the fault handling time by ≥60% compared with the traditional algorithm. In the operation optimization scenario, the historical association rules are output to support parameter adjustment of the on-board train control system and reduce train energy consumption.

[0101] A7, every preset time length (such as 100 ms) is circulated to determine whether the train running state is stopped, if the train is not stopped (real-time speed > 0 km / h), the above steps A1~A6 are repeated to ensure real-time update of data association, and dynamic support is continuously provided for train operation; if the train is stopped (real-time speed is 0 km / h, and lasts for 30s), step A8 is executed.

[0102] A8, stop data collection and save the latest on-board train control data knowledge graph, support power recovery.

[0103] In another exemplary embodiment, when the distribution information of the sleeper is used for speed measurement, the distribution information of the sleeper belongs to an absolute positioning method, which can directly update the position of the maglev train while accurately measuring the speed, and realize the positioning requirement of the maglev train.

[0104] The knowledge graph-based on-board train control data multi-dimensional association method provided by the above embodiments of the application constructs a knowledge graph based on braking and speed data, can accurately associate multi-source heterogeneous on-board train control data, has high data association accuracy, is more reliable in diagnosis results relying on entity association network and double mechanism, can mine historical association rules to support train control strategy optimization and reduce train energy consumption, can also be flexibly adapted to multiple scenes and compatible with existing hardware, has strong adaptability and low deployment cost.

[0105] Based on the same inventive concept, the embodiments of the application also provide a device for implementing the knowledge graph-based on-board train control data multi-dimensional association method described above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method. In one exemplary embodiment, as shown in Figure 3 , a knowledge graph-based on-board train control data multi-dimensional association device is provided, which includes the following functional modules:

[0106] The multi-source vehicle-mounted train control data acquisition module is configured to cyclically determine a train operation state every preset time length, and acquire multi-source vehicle-mounted train control data of a vehicle-mounted train control system if the train is not stopped; the multi-source vehicle-mounted train control data includes braking core data, speed core data, and associated auxiliary data; the associated auxiliary data includes communication data, device state data, environmental data, and fault log data; if the train is stopped, the data acquisition is stopped and the latest vehicle-mounted train control data knowledge graph is saved.

[0107] The multi-source data classification module is configured to classify each data in the multi-source vehicle-mounted train control data according to a data type of each data and an association degree of the data with core data, to obtain a core data layer and an auxiliary association layer; the core data is braking or speed related data; the core data layer includes braking core data and speed core data; and the auxiliary association layer includes communication data, device state data, environmental data, and fault log data.

[0108] The entity definition and extraction module is configured to define core entities and auxiliary entities based on the core data layer and the auxiliary association layer, respectively, and extract the core entities and the auxiliary entities from the multi-source vehicle-mounted train control data in a hybrid manner mainly based on rules and supplemented by machine learning; the core entities include braking pressure entities, braking instruction level entities, braking actuator entities, real-time speed entities, target speed entities, and speed change rate entities; and the auxiliary entities include instruction entities, coordination information entities, vehicle-mounted host entities, speed sensor entities, braking valve entities, track slope entities, external temperature entities, braking pressure abnormal entities, and speed sensor fault entities.

[0109] The association relationship extraction module is configured to extract association relationships between entities by using a relationship extraction model based on an attention mechanism in combination with a vehicle-mounted train control field dictionary, and correct the results by artificial rules, to obtain association relationships included in the knowledge graph; the association relationships include braking and speed association relationships, core entity and auxiliary entity association relationships, and auxiliary entity association relationships.

[0110] The knowledge graph construction module is configured to perform knowledge fusion on the extracted entities and association relationships, and construct a vehicle-mounted train control data knowledge graph, and store the vehicle-mounted train control data knowledge graph in a graph database; the knowledge fusion includes processing of entity duplication and association relationship conflicts.

[0111] The data multi-dimensional association analysis module is configured to perform data multi-dimensional association analysis based on the vehicle-mounted train control data knowledge graph, output association analysis results, and adapt to specific application scenarios; the data multi-dimensional association analysis includes braking and speed real-time association, speed deviation and braking instruction traceability, abnormal data association traceability, and historical data association mining; the association analysis results include structured reports and visualized graphs; and the specific application scenarios include train operation monitoring scenarios, fault diagnosis scenarios, and operation optimization scenarios.

[0112] Of course, Figure 3 The illustrated architecture is only exemplary, and in implementing different functions, according to actual needs, one or at least two components in the system can be omitted Figure 3 The illustrated architecture is only exemplary, and in implementing different functions, according to actual needs, one or at least two components in the system can be omitted

[0113] In an exemplary embodiment, a computer device is provided, which can be a server or a terminal, and its internal structure diagram can be as shown Figure 4 The computer device includes a processor, a memory, an input / output interface (I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through the system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capability. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The database of the computer device is used to store the train control data knowledge graph and other intermediate data. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with the terminal outside through network connection. The computer program is executed by the processor, which can realize the train control data multi-dimensional association method based on knowledge graph provided in the foregoing embodiment.

[0114] Those skilled in the art can understand, Figure 4 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0115] In an exemplary embodiment, a computer device is provided, which can be a server or a terminal, and its internal structure diagram can be as shown

[0116] In an exemplary embodiment, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by a processor to realize the steps in each of the method embodiments.

[0117] In an exemplary embodiment, a computer program product is provided, which includes a computer program, and the computer program is executed by a processor to realize the steps in each of the method embodiments.

[0118] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant regulations.

[0119] It can be understood by those skilled in the art that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing related hardware through a computer program, and the computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments of each method. Among them, any reference to memory, database or other medium used in each embodiment provided by the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (Read-Only Memory, ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric memory (Ferroelectric Random Access Memory, FRAM), phase change memory (Phase Change Memory, PCM), graphene memory, etc. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), etc.

[0120] The database involved in each embodiment provided by the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in each embodiment provided by the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0121] Each technical feature of the above embodiments can be combined arbitrarily. In order to make the description simple, not all possible combinations of each technical feature in the above embodiments are described, but as long as the combination of these technical features does not exist contradictory, it should be considered as the scope of the present application.

[0122] The principles and implementations of the present application are described in the specific examples herein, and the above examples are only used to help understand the method of the present application and its core idea; meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation and application range will be changed. In summary, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A multi-dimensional association method for onboard train control data based on knowledge graphs, characterized in that, include The train's operating status is checked cyclically at preset time intervals. If the train has not stopped, multi-source onboard train control data from the onboard train control system is acquired. The multi-source onboard train control data includes braking core data, speed core data, and associated auxiliary data. The associated auxiliary data includes communication data, equipment status data, environmental data, and fault log data. The data are classified according to the data type and correlation with the core data in the multi-source on-board train control data, resulting in a core data layer and an auxiliary correlation layer; the core data is braking or speed-related data. Core entities and auxiliary entities are defined based on the core data layer and the auxiliary association layer, respectively, and a hybrid approach of rule-based and machine learning-based methods is used to extract core entities and auxiliary entities from the multi-source vehicle-mounted train control data. An attention-based relation extraction model combined with a vehicle-mounted train control domain dictionary is used to extract the relationships between entities. The results are then manually verified and corrected to obtain the relationships contained in the knowledge graph. Knowledge fusion is performed on the extracted entities and their relationships to construct a vehicle-mounted train control data knowledge graph, which is then stored in a graph database. The knowledge fusion includes handling entity duplication and relationship conflicts. Based on the on-board train control data knowledge graph, perform multi-dimensional data correlation analysis, output correlation analysis results, and adapt them to specific application scenarios; If the train stops, data collection is stopped and the latest onboard train control data knowledge graph is saved; The core data layer includes braking core data and speed core data; the auxiliary association layer includes communication data, equipment status data, environmental data, and fault log data; a hybrid approach, primarily rule-based and supplemented by machine learning, is used to extract core and auxiliary entities from the multi-source on-board train control data, specifically including: For the structured data in the multi-source on-board train control data, core entities are directly extracted based on preset extraction rules; the preset extraction rules are to extract entities of the corresponding type according to field identifiers; the core entities include braking pressure entity, braking command level entity, braking actuator entity, real-time speed entity, target speed entity, and speed change rate entity. For the unstructured data in the multi-source onboard train control data, an entity extraction model is used to extract auxiliary entities. The entity extraction model is a model obtained by training an LSTM model with several multi-source onboard train control data labeled with entity category tags in advance. The auxiliary entities include command entities, coordination information entities, onboard host entities, speed sensor entities, brake valve entities, track gradient entities, external temperature entities, abnormal brake pressure entities, and speed sensor fault entities.

2. The knowledge graph-based multi-dimensional association method for onboard train control data according to claim 1, characterized in that, Knowledge fusion is performed on the extracted entities and their relationships, specifically including: Entity linking is used to match duplicate entities based on attribute similarity or semantic similarity, merging duplicate entities into a single entity; Relationship reconciliation is adopted, and conflicting association relationships are corrected based on a preset timestamp priority rule; the timestamp priority rule is that the timestamp of the core data layer has a higher priority than the timestamp of the auxiliary association layer.

3. The method for multi-dimensional association of onboard train control data based on knowledge graphs according to claim 1, characterized in that, Constructing a knowledge graph of onboard train control data and storing it in a graph database specifically includes: Each entity is used as a node in the vehicle-mounted train control data knowledge graph; node attributes include collection timestamp, numerical / status description, and data source; Each relationship is represented as a directed edge between nodes in the vehicle-mounted train control data knowledge graph; the edge attributes include the relationship confidence and the number of historical matches. The graph database storing the vehicle-mounted train control data knowledge graph is deployed on the vehicle-mounted server, supporting real-time multi-dimensional data correlation analysis requests.

4. The method for multi-dimensional association of onboard train control data based on knowledge graphs according to claim 1, characterized in that, The knowledge graph includes relationships between braking and speed, between core entities and auxiliary entities, and between auxiliary entities. The braking and speed relationship includes: the braking pressure entity is associated with the real-time speed entity, the braking command level entity is associated with the target speed entity, and the braking actuator entity is associated with the speed change rate entity. The core entity and auxiliary entity relationship includes: command-related core entities are associated with the command entity and the coordination information entity, and state-related core entities are associated with all auxiliary entities except the command entity and the coordination information entity. The command-related entities include the braking command level entity and the target speed entity; the state-related core entities include the braking pressure entity, the braking actuator entity, the real-time speed entity, and the speed change rate entity. The braking and speed are correlated in real time, specifically including: Real-time acquisition of braking pressure data; when the braking pressure data reaches a preset braking pressure threshold, the real-time speed nodes associated with the braking pressure nodes are traversed through the vehicle-mounted train control data knowledge graph. By calling historical correlation patterns, when the change pattern of the real-time speed node is determined to be outside the range of the pattern, it is marked as "braking-speed correlation anomaly" and the abnormal data correlation tracing process is triggered; The speed deviation and braking command tracking specifically includes: When the detected speed deviation value is greater than the preset speed deviation threshold, the braking actuator node and braking command level node are traced back in reverse through the on-board train control data knowledge graph; Verify whether the braking command has been issued and whether the braking level meets the standard. At the same time, trace the command nodes associated with the braking command level node to confirm whether the target speed has not been updated due to command transmission delay. The abnormal data association and source tracing specifically involves: when the real-time speed fluctuates abnormally within 10 seconds, the root cause of the abnormality is determined by inferring the speed sensor status, brake actuator status, and external temperature nodes based on the vehicle-mounted train control data knowledge graph and using preset inference rules. The historical data correlation mining specifically involves: traversing the historical data of any slope line within three months in the on-board train control data knowledge graph, and mining the historical correlation patterns of track gradient, braking pressure, and speed reduction, which serve as the data basis for optimizing the braking strategy of the slope line.

5. The method for multi-dimensional association of onboard train control data based on knowledge graphs according to claim 1, characterized in that, The multidimensional data correlation analysis includes real-time correlation between braking and speed, tracing speed deviation and braking command, tracing the source of abnormal data, and mining historical data correlations; the correlation analysis results include structured reports and visualization maps; the specific application scenarios include train operation monitoring, fault diagnosis, and operation optimization. Output the correlation analysis results and adapt them to specific application scenarios, including: The correlation analysis results are output in the form of a structured report, and a visualization map is also output. The structured report includes a braking and speed correlation status report and an anomaly data tracing report. In the visualization map, the nodes of the core entities are red, the nodes of the auxiliary entities are blue, and the nodes of the track slope entity, the external temperature entity, the braking pressure anomaly entity, and the speed sensor failure entity are green. The directed edges are labeled with the relationship type and confidence level. In train operation monitoring scenarios, when an abnormal marker is present, the on-board terminal will issue an audible and visual alarm with beeping and flashing. In fault diagnosis scenarios, an abnormal data tracing report is output to identify the components associated with the fault; In operational optimization scenarios, historical correlation patterns are output to support parameter adjustments in the onboard train control system, thereby reducing train energy consumption.

6. A knowledge graph-based multi-dimensional association device for onboard train control data, characterized in that, include: The multi-source onboard train control data acquisition module is used to cyclically determine the train's operating status at preset time intervals. If the train is not stopped, it acquires multi-source onboard train control data from the onboard train control system. The multi-source onboard train control data includes braking core data, speed core data, and associated auxiliary data. The associated auxiliary data includes communication data, equipment status data, environmental data, and fault log data. If the train stops, it stops data acquisition and saves the latest onboard train control data knowledge graph. The multi-source data classification module is used to classify each data according to the data type and correlation degree between each data in the multi-source on-board train control data and the core data, to obtain a core data layer and an auxiliary correlation layer; the core data is braking or speed-related data. The entity definition and extraction module is used to define core entities and auxiliary entities based on the core data layer and the auxiliary association layer respectively, and to extract core entities and auxiliary entities from the multi-source vehicle-mounted train control data using a hybrid approach that is primarily rule-based and secondarily machine learning-based. The relationship extraction module is used to extract the relationships between entities by using an attention-based relationship extraction model combined with a vehicle-mounted train control domain dictionary, and to obtain the relationships contained in the knowledge graph by manually verifying and correcting the results through rules. The knowledge graph construction module is used to perform knowledge fusion on the extracted entities and their relationships to construct a vehicle-mounted train control data knowledge graph and store the vehicle-mounted train control data knowledge graph in a graph database; the knowledge fusion includes handling entity duplication and relationship conflict. The data multidimensional correlation analysis module is used to perform multidimensional correlation analysis based on the on-board train control data knowledge graph, output the correlation analysis results and adapt them to specific application scenarios. The core data layer includes braking core data and speed core data; the auxiliary association layer includes communication data, equipment status data, environmental data, and fault log data; a hybrid approach, primarily rule-based and supplemented by machine learning, is used to extract core and auxiliary entities from the multi-source on-board train control data, specifically including: For the structured data in the multi-source on-board train control data, core entities are directly extracted based on preset extraction rules; the preset extraction rules are to extract entities of the corresponding type according to field identifiers; the core entities include braking pressure entity, braking command level entity, braking actuator entity, real-time speed entity, target speed entity, and speed change rate entity. For the unstructured data in the multi-source onboard train control data, an entity extraction model is used to extract auxiliary entities. The entity extraction model is a model obtained by training an LSTM model with several multi-source onboard train control data labeled with entity category tags in advance. The auxiliary entities include command entities, coordination information entities, onboard host entities, speed sensor entities, brake valve entities, track gradient entities, external temperature entities, abnormal brake pressure entities, and speed sensor fault entities.

7. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the knowledge graph-based multidimensional association method for vehicle-mounted train control data as described in any one of claims 1-5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the knowledge graph-based multidimensional association method for vehicle-mounted train control data as described in any one of claims 1-5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the knowledge graph-based multidimensional association method for vehicle-mounted train control data as described in any one of claims 1-5.

Images