Urban rail emergency command and driving maintenance method and equipment based on intelligent linkage

Abstract

The invention relates to an urban rail emergency command and driving maintenance method based on intelligent linkage, and the method comprises the steps: obtaining the equipment state data of each subsystem in real time, carrying out the preprocessing, and obtaining a real-time feature vector; performing real-time analysis on the real-time feature vector by using a preset fault rule base, a rail transit knowledge graph and an anomaly detection model based on machine learning to realize automatic identification, classification and positioning of faults; fusing the fault information, the real-time train position, the station passenger flow data and the line topological structure, automatically evaluating the potential influence range and degree of the fault on the driving order, the passenger service and the subsequent working diagram based on a preset simulation deduction model, and generating a visual report; and generating a co-processing scheme based on the visual report, issuing the report and the scheme together, obtaining an actual situation fed back on site in real time, and dynamically adjusting the co-processing scheme and the visual report. Compared with the prior art, intelligent linkage of driving and maintenance is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of information operation and maintenance monitoring, in particular to a city rail emergency command and driving maintenance method and device based on intelligent linkage. BACKGROUND

[0002] Currently, the operation command and equipment maintenance of urban rail transit are usually in different systems, and the business processes are independent of each other. When a device fails, the following disadvantages exist: 1. Information island: the information transmission between driving command personnel and maintenance personnel relies on manual communication, which is inefficient and prone to misjudgment.

[0003] 2. Disposal lag: driving command personnel cannot intuitively assess the overall impact of the failure on operation; maintenance personnel cannot obtain driving scheduling intentions in time and respond passively.

[0004] 3. Decision isolation: driving adjustment scheme and maintenance disposal scheme are formulated separately, lack of coordination, which may lead to prolonged operation interruption time or waste of maintenance resources.

[0005] The traditional technology lacks an integrated platform that can realize real-time data sharing and intelligent linkage decision-making, and cannot realize efficient and closed-loop management of the whole process from fault occurrence to operation recovery.

[0006] Chinese patent CN116443080B discloses a rail transit driving scheduling command method, system, device and medium to solve the problems of information island, disposal lag and decision isolation. By obtaining comprehensive operation state information of the rail transit system, the rail transit system information is comprehensively and accurately perceived. By identifying abnormal events and determining the influence of abnormal events on driving and train operation situation, support is provided for optimized decision-making. By formulating a dynamic train operation diagram according to the basic line parameters, planned operation data, the influence of abnormal events on driving and train operation situation, a scheduling strategy can be automatically generated, replacing the manual train operation adjustment mode of driving dispatchers, reducing the work intensity of dispatchers. By generating and issuing scheduling instructions according to the dynamic train operation diagram, the risk of further expansion caused by the delay of manual scheduling instructions through telephone is avoided. However, the technical architecture of this invention still takes driving scheduling as a single core, only integrates basic data such as signals, power supply and passenger flow, does not form a multi-system deep linkage mechanism, and focuses on driving efficiency as a single target, without considering global costs such as resource consumption and operation loss.

[0007] In summary, there is still a need for an emergency command method that can solve the problems of information island, disposal lag and decision isolation, and realize intelligent linkage and global multi-target coordination of the whole process. SUMMARY

[0008] The purpose of the present application is to overcome the defects of the prior art and provide a smart linkage-based urban rail emergency command and driving maintenance method and device, which realizes the smart linkage of driving and maintenance.

[0009] The purpose of the present application can be achieved by the following technical solutions: A smart linkage-based urban rail emergency command and driving maintenance method, the method comprising: Real-time acquisition of device state data of each subsystem through a standard protocol interface and preprocessing to obtain real-time feature vectors; Real-time analysis of the real-time feature vectors using a pre-set fault rule library, a rail transit knowledge graph, and an abnormality detection model based on machine learning to realize automatic identification, classification, and positioning of faults and obtain fault information; Fusion of the fault information, real-time train position, station passenger flow data, and line topology structure, automatic evaluation of the potential impact range and degree of the fault on driving order, passenger service, and subsequent running graph based on a pre-set simulation deduction model, and generation of a visual report; Generation of a collaborative disposal scheme based on the visual report, issuance of the report and the scheme together, real-time acquisition of actual conditions fed back on site, and dynamic adjustment of the collaborative disposal scheme and the visual report.

[0010] Further, when the rail transit system is started, basic data is loaded, a rail transit knowledge graph is constructed based on the basic data, and the correlation between devices, faults, impacts, and disposal measures is established.

[0011] Further, the basic data includes line device account, line operation topology graph, real-time running graph, historical fault library, maintenance procedure, and emergency disposal preplan; The line device account includes device ID, physical location, model, and subsystem to which the device belongs; The line operation topology graph includes interval distribution, station location, and standby line configuration; The real-time running graph includes train frequency, stop time, and route planning; The historical fault library includes fault type in recent years, occurrence time, and disposal result; The maintenance procedure includes device maintenance steps and safety requirements; The emergency disposal preplan includes standard response processes for different fault scenarios.

[0012] Further, the process of constructing a rail transit knowledge graph based on the basic data includes: Based on the business needs of urban rail emergency command and driving maintenance, obtain a knowledge classification ontology, form a knowledge graph framework according to the core correlation logic between each ontology, and the knowledge classification ontology includes a device ontology, a fault ontology, an influence ontology, a disposal ontology and a case ontology; Based on the knowledge graph framework, extract knowledge graph data from the basic data and perform preprocessing; Based on the knowledge graph data and the knowledge graph framework after preprocessing, obtain the knowledge nodes and their correlation of the knowledge graph; Import the knowledge nodes, knowledge graph data and their correlation into a database, generate visual nodes and edges of the knowledge graph, and establish an index to complete the construction of the rail transit knowledge graph.

[0013] Further, the process of mapping the preprocessed knowledge graph data into knowledge nodes and their correlation of the knowledge graph according to the knowledge graph framework one by one includes: For structured data in the knowledge graph data, map one by one according to the knowledge graph framework to obtain the entities and their correlation of the structured data knowledge graph; For unstructured data in the knowledge graph data, combine NLP tools and domain rules to obtain the entities and their correlation of the unstructured data knowledge graph; Eliminate duplicate entities through normalization processing, and eliminate relationship conflicts or repetitions between different entities through priority rules or voting mechanisms; Combine the entities and their correlation of the structured data knowledge graph after processing and the entities and their correlation of the unstructured data knowledge graph obtained, and output them as knowledge nodes and their correlation of the knowledge graph.

[0014] Further, the fault rule base is constructed based on the experience of domain experts and historical fault cases, and contains IF-THEN logic.

[0015] Further, the process of automatic identification, classification and positioning of the fault includes: Automatic identification: match the real-time feature vector with the preset fault rule base to quickly identify known and pattern-specific faults; for potential faults that cannot be covered by the rule base or have unclear patterns, use an anomaly detection model based on machine learning to analyze, the anomaly detection model identifies data points that deviate significantly from the normal operation mode through unsupervised learning, and marks them as abnormal states, achieving the identification of unknown faults or early hidden faults; Classification: based on the automatic identification result, accurately classify the fault events through support vector machines or decision trees to determine their belonging to a large class and a subclass; Positioning: based on the line topology and the constructed rail transit knowledge graph, the positioning of the fault equipment in the network and its association with other equipment are analyzed; the alarm and status information from different subsystems are fused to cross-verify the fault positioning, and the positioning result is output; if the fault cannot be directly positioned, a Bayesian network-based probability model is used to comprehensively consider the historical failure rate of each device, the current state, environmental factors, and the conditional probability between the alarm information, calculate the probability of each device being the fault source, and list one or several devices with the highest probability as the positioning result.

[0016] Further, the simulation deduction model is a hybrid simulation model constructed based on digital twinning technology, integrating line topology, train operation characteristics, signal logic and passenger flow rules, and calibrated using historical operation data.

[0017] Further, the process of automatically calculating the potential impact range and degree of the fault on train operation order, passenger service and subsequent running graph based on the simulation deduction model includes: Based on the fault information, real-time train position, station passenger flow data and line topology structure, the simulation parameters of the simulation deduction model are set, and the simulation parameters are calibrated synchronously based on historical operation data; Based on the simulation deduction model, the potential impact range and degree of the fault on train operation order, passenger service and subsequent running graph are evaluated; wherein, The potential impact on the train operation order includes single train direct delay calculation and subsequent train chain delay calculation; the evaluation indexes of the potential impact range and degree on the train operation order include the total number of affected trains, the maximum delay of a single train, the total delay time and the running graph deviation rate; The potential impact on the passenger service includes station passenger backlog and total number of affected passengers, and the evaluation indexes of the potential impact range and degree on the passenger service include station passenger backlog rate and proportion of affected passengers; The potential impact on the subsequent running graph includes running graph recovery time, and the evaluation index of the potential impact range and degree on the subsequent running graph includes subsequent running graph deviation accumulation.

[0018] Further, the collaborative disposal scheme includes a train operation scheme and a maintenance scheme, and the process of generating a collaborative disposal scheme based on the visualization report includes: Train operation scheme generation: based on the visualization report, taking safety and maximum maintenance of operation as hard constraints, combining a preset train operation adjustment strategy library, generating a plurality of preliminary train operation plans through a constraint planning algorithm; simulating and calculating the key KPIs of each preliminary train operation plan through the simulation deduction model, and retaining the preliminary train operation plans with key KPIs higher than a preset threshold as the train operation scheme; Maintenance scheme generation: based on fault type and location, matching corresponding maintenance procedures and historical similar cases from the knowledge graph; combined with real-time maintenance resource data, determine resource schedulability, and generate a standardized maintenance scheme including fault location, maintenance steps, resource list and estimated repair time; Scheme linkage optimization: taking the shortest operation interruption time and the lowest overall resource consumption as the core target, the adjustment time / amplitude of the train operation scheme and the operation window / resource scheduling of the maintenance scheme are included in the same model; genetic algorithm or particle swarm algorithm is adopted to iteratively calculate the global benefit of different combinations of train operation scheme and maintenance scheme, and to evaluate the conflict points; the globally optimal combination of train operation scheme and maintenance scheme is output as the collaborative disposal scheme.

[0019] An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the intelligent linkage-based urban rail emergency command and train maintenance method as described above when executing the computer program.

[0020] 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 intelligent linkage-based urban rail emergency command and train maintenance method as described above.

[0021] Compared with the prior art, the beneficial effects of the present application include: 1. The present application provides data support for train adjustment and maintenance disposal through accurate fault identification and dynamic impact deduction, reduces misjudgment and resource waste caused by reliance on manual experience, improves decision-making scientificity and reduces disposal blindness; the present application acquires device state data of each subsystem in real time, and identifies faults through a dual fault identification mechanism of rule matching and machine learning; rule matching quickly identifies known faults, an anomaly detection model discovers unknown faults and hidden faults, avoids missed reports, and a composite fault identification method is suitable for complex scenarios and can handle multiple types of faults such as signals, vehicles and power supply, and adapts to the characteristics of devices in different subsystems; the present application fuses fault information with various information, integrates line topology, train characteristics and passenger flow rules, and calibrates historical data to accurately reproduce real operation states and deduce multiple scenarios in parallel, quantify train delay and passenger backlog, and provide clear basis for decision-making.

[0022] 2. The fault rule base of the present application contains IF-THEN logic, which can quickly identify known faults with clear patterns, and through lightweight classifier and multi-source information fusion positioning, accurate classification and accurate positioning of faults can be realized, and the misjudgment rate is reduced.

[0023] 3. The rail transit professional knowledge graph of the application clearly associates the correlation of equipment, faults, influences, treatments and resources, realizes the rapid retrieval of maintenance procedures and case experience, provides standardized operation steps and resource lists for the maintenance scheme, and reduces the cost of manual retrieval; the knowledge graph also supports continuous learning, stores historical treatment data, provides structured knowledge deposition for model optimization, and promotes system capability iteration.

[0024] 4. The application integrates train operation adjustment and equipment maintenance into the same optimization model, and collaboratively plans for global benefits to achieve efficient and beneficial treatment effect, which is a fundamental upgrade to the traditional scattered management mode.

[0025] 5. The visual report generated by the application directly presents the fault influence and treatment scheme, which is convenient for command personnel to quickly understand; the collaborative treatment scheme clearly presents the operation adjustment details, maintenance steps, resource list and other operable contents to ensure the accuracy of on-site execution and improve the technical landing effect.

[0026] 6. The application generates a train operation plan that meets the safety priority and maintains the operation hard constraint through a constraint planning algorithm, and generates a standardized maintenance scheme by matching the maintenance procedures and historical cases with the knowledge graph, avoiding the problems of scheme lag and incomplete consideration caused by traditional reliance on manual experience; in addition, the application takes the shortest operation interruption time and the lowest resource consumption as the target, and optimizes the combined scheme of train operation adjustment and maintenance work through genetic algorithm / particle swarm optimization algorithm, solves the operation conflict and resource waste caused by the traditional emergency and maintenance, and maximizes the reduction of operation loss caused by faults BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 is a flowchart of the method of the application; Figure 2 is a system structure diagram of the application. DETAILED DESCRIPTION

[0028] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor should belong to the protection scope of the application.

[0029] Embodiment 1 The embodiment discloses a city rail emergency command and train maintenance method based on intelligent linkage, as shown in Figure 1 The steps are specifically described as follows: Step S1, real-time acquisition of equipment state data of each subsystem through a standard protocol interface and preprocessing to obtain real-time feature vectors.

[0030] After obtaining the device state data of each subsystem in real time through the standard protocol interface, the preprocessing process includes: cleaning, aligning and standardizing the device state data of the subsystem accessed in real time; for different subsystem data characteristics, extracting key features such as the state switching sequence of the signal device, the current / voltage waveform characteristics of the power supply system, and the vibration spectrum of the vehicle bearing, forming a unified format time sequence feature vector, and outputting the real-time feature vector.

[0031] Step S2, using a preset fault rule library, an urban rail transit knowledge graph, and an abnormality detection model based on machine learning, the real-time feature vector is analyzed in real time to realize automatic identification, classification and positioning of faults.

[0032] The fault rule library is constructed based on the experience of domain experts and historical fault cases, and contains IF-THEN logic.

[0033] For example, "IF track circuit red light band AND associated signal machine open signal THEN determine as track section occupation fault". This step is used to quickly identify known and pattern-specific faults.

[0034] When the urban rail transit system is started, the basic data is loaded, and the urban rail transit knowledge graph is constructed based on the basic data, and the correlation between the equipment, faults, influences and disposal measures is established.

[0035] The basic data includes line equipment account, line operation topology graph, real-time operation graph, historical fault library, maintenance procedure and emergency disposal plan; The line equipment account includes equipment ID, physical location, model and subsystem to which it belongs; The line operation topology graph includes interval distribution, station location and spare line configuration; The real-time operation graph includes train schedule, stop time and route planning; The historical fault library includes fault types in recent years, occurrence time and disposal results; The maintenance procedure includes equipment maintenance steps and safety requirements; The emergency disposal plan includes standard response processes for different fault scenarios.

[0036] The process of constructing the urban rail transit knowledge graph based on the basic data includes: Based on the business needs of urban rail emergency command and train maintenance, obtain the knowledge classification ontology, form the knowledge graph framework according to the core correlation logic between each ontology, and the knowledge classification ontology includes equipment ontology, fault ontology, influence ontology, disposal ontology and case ontology; Extract knowledge graph data from the basic data based on the knowledge graph framework, and pre-process it; Based on the pre-processed knowledge graph data and knowledge graph framework, obtain the knowledge nodes of the knowledge graph and their associated relationships; Import the knowledge nodes, knowledge graph data, and their associated relationships into a database, generate visual nodes and edges of the knowledge graph, and establish an index to complete the construction of the rail transit knowledge graph.

[0037] In this embodiment, the device ontology: key device information of each subsystem of urban rail transit, including device ID, physical location (line / section / site), device type (signal / switch / train traction system / supply switch cabinet, etc.), device model, subsystem (ATS / TCMS / SCADA, etc.), device account (date of manufacture / maintenance period); Fault ontology: fault-related attributes, including fault ID, fault type (major categories: signal / vehicle / power supply / communication fault; subcategories: signal light off / switch stuck / traction system failure, etc.), fault characteristics (such as the combination of signal communication interruption and train speed limitation), fault causes (such as device aging / environmental interference / human error), historical failure rate; Impact ontology: impact dimensions of faults on operation, including impact objects (traffic order / passenger service / device chain), impact indicators (train delay time / affected passenger number / congested station number), and impact degree classification (minor / average / severe / extreme); Disposal ontology: fault response measures related knowledge, including disposal type (train adjustment / maintenance operation / passenger organization), specific measures (train decoupling / jump stop / spare parts replacement / passenger flow restriction), operation procedures (maintenance steps / safety protection requirements), resource requirements (personnel skill requirements / spare parts model / tool list), and estimated disposal time (MTTR); Case ontology: historical fault disposal cases, including case ID, fault scenario (time / location / initial state), disposal plan, execution process, actual effect (delay time / resource consumption / passenger feedback), and experience summary.

[0038] Knowledge graph data has structured, semi-structured, and unstructured data. After obtaining the knowledge graph data, pre-process it according to the data type, including: Structured data cleaning: through SQL statements or ETL tools, remove duplicate data, correct incorrect data, and complete missing data; Semi-structured data analysis: using Python's BeautifulSoup, XPath, and other tools, extract fault type-disposal steps and device-fault key information from Word / PDF manuals, and convert them into key-value pairs or table form; Unstructured data structuring: for text data, natural language processing (NLP) techniques are used to extract ontology entities and associated relationships; for log data, key word matching is used to extract key time nodes and execution status, and the data is converted into structured disposal process records.

[0039] The process of mapping the preprocessed knowledge graph data according to the knowledge graph framework to the knowledge nodes and their associated relationships of the knowledge graph includes: For structured data in the knowledge graph data, map one by one according to the knowledge graph framework to obtain the entities and their associated relationships of the structured data knowledge graph; For unstructured data in the knowledge graph data, use NLP tools combined with domain rules to obtain the entities and their associated relationships of the unstructured data knowledge graph; Eliminate duplicate entities through normalization processing, and eliminate relationship conflicts or repetitions between different entities through priority rules or voting mechanisms; Combine the processed entities and their associated relationships of the structured data knowledge graph and the obtained entities and their associated relationships of the unstructured data knowledge graph, and output them as knowledge nodes and their associated relationships of the knowledge graph.

[0040] The process of obtaining the entities and their associated relationships of the unstructured data knowledge graph through NLP tools combined with domain rules includes: Entity extraction: use a BERT-based named entity recognition model to extract device names, fault types, and disposal measures from text; Relationship extraction: use a combination of rule matching and machine learning to extract relationships between entities: Rule matching: based on keyword templates developed by domain experts, match relationships from text; Machine learning: for complex text, use a relationship extraction model such as RE-BERT to automatically identify implicit relationships such as fault-impact and disposal-effect; Attribute extraction: extract supplementary attributes of entities from text, such as "2023 XX turnout fault maintenance time 45 minutes" to extract "turnout fault disposal time: 45 minutes" and supplement the expected disposal time attribute of the disposal ontology instance.

[0041] Due to different data sources, such as inconsistent device names in device account and maintenance work order, normalization processing of extracted entities is required to avoid duplicate nodes in the knowledge graph.

[0042] The process of automatic identification, classification, and positioning of faults includes: Automatic identification: match the real-time feature vector with the preset fault rule library to quickly identify known and pattern-specific faults; for potential faults that cannot be covered by the rule library or have unclear patterns, use a machine learning-based anomaly detection model for analysis, which identifies data points that deviate significantly from the normal operation mode through unsupervised learning and marks them as abnormal states, enabling the identification of unknown faults or early implicit faults; Classification: based on the automatic identification results, accurately classify the fault events through support vector machines or decision trees to determine their belonging to a large class and a subclass; Positioning: based on the line topology structure and the constructed rail transit knowledge graph, analyze the positioning of the fault equipment in the network and its association with other equipment; integrate alarm and state information from different subsystems to cross-verify the fault positioning and output the positioning result; if the fault cannot be directly positioned, use a Bayesian network-based probability model to comprehensively consider the conditional probability between the historical fault rate, current state, environmental factors, and alarm information of each device, calculate the probability of each device being the fault source, and list the device or devices with the highest probability as the positioning result.

[0043] Step S3: Integrate the fault information, real-time train position, station passenger flow data, and line topology structure, and based on a preset simulation deduction model, automatically evaluate the potential impact range and degree of the fault on train operation order, passenger service, and subsequent running diagram, and generate a visual report.

[0044] The simulation deduction model is a hybrid simulation model based on digital twinning technology, integrating line topology, train operation characteristics, signal logic, and passenger flow rules, and calibrated using historical operation data.

[0045] Based on the simulation deduction model, the process of automatically calculating the potential impact range and degree of the fault on train operation order, passenger service, and subsequent running diagram includes: Based on the fault information, real-time train position, station passenger flow data, and line topology structure, set the simulation parameters of the simulation deduction model, and based on historical operation data, synchronize the calibration of the simulation parameters; Based on the simulation deduction model, evaluate the potential impact range and degree of the fault on train operation order, passenger service, and subsequent running diagram; wherein, The potential impact on train operation order includes single-train direct delay calculation and subsequent train chain delay calculation; the evaluation indicators of the potential impact range and degree on train operation order include the total number of affected trains, the maximum delay of a single train, the total delay time, and the running diagram deviation rate; The potential impact on passenger service includes station passenger backlog and the total number of affected passengers; the evaluation indicators of the potential impact range and degree on passenger service include the station passenger backlog rate and the proportion of affected passengers; The potential impact on the subsequent train diagram includes a train diagram recovery time, and the evaluation index of the range and degree of the potential impact on the subsequent train diagram includes a subsequent train diagram deviation accumulation.

[0046] In step S4, a collaborative treatment scheme is generated based on the visualization report, and the report is issued together with the scheme. Real-time feedback on the actual situation is obtained, and the collaborative treatment scheme and the visualization report are dynamically adjusted.

[0047] The collaborative treatment scheme includes a train operation scheme and a maintenance scheme. The process of generating the collaborative treatment scheme based on the visualization report includes: Train operation scheme generation: Based on the visualization report, taking safety and maximum maintenance of operation as hard constraints, combining a preset train operation adjustment strategy library, and generating a plurality of preliminary train operation plans through a constraint planning algorithm; key KPIs of each preliminary train operation plan are quickly simulated and calculated through a simulation deduction model, and the preliminary train operation plan with a key KPI higher than a preset threshold is reserved as a train operation scheme; Maintenance scheme generation: Based on the fault type and positioning, the corresponding repair procedures and historical similar cases are matched from the knowledge graph; combined with real-time repair resource data, resource schedulability is determined, and a standardized repair scheme including fault positioning, repair steps, resource list and estimated repair time is generated; Scheme linkage optimization: Taking the shortest operation interruption time and the lowest overall resource consumption as the core target, the adjustment time / amplitude of the train operation scheme and the operation window / resource scheduling of the maintenance scheme are included in the same model; genetic algorithm or particle swarm algorithm is adopted to iteratively calculate the global benefits of different combinations of train operation schemes and maintenance schemes, and to evaluate conflict points; the globally optimal combination of train operation scheme and maintenance scheme is output as the collaborative treatment scheme.

[0048] The key KPI is a weighted comprehensive number of the potential impact range and degree of each preliminary train operation plan on train order, passenger service and subsequent train diagram. The specific calculation process is the same as the potential impact range and degree of the fault on the train order, passenger service and subsequent train diagram. After obtaining each potential impact range and degree, the key KPI is recalculated based on a preset weight.

[0049] In this embodiment, the scheme is generated by a pre-trained scheme generation model. The process of generating the collaborative treatment scheme based on the visualization report by the scheme generation model is as shown above.

[0050] In this embodiment, step S4 is specifically: Based on the visualization report, a scheme generation model is used to generate a collaborative treatment scheme, and the optimized collaborative treatment scheme is pushed to the work terminal of the train dispatcher and the maintenance dispatcher.

[0051] The dispatcher can confirm, fine-tune or one-key issue the scheme. The system automatically issues train control instructions to the automatic train supervision system (ATS), passenger transport organization instructions to the passenger information system (PIS) and broadcast system (PA), and maintenance work orders, operation guidelines and electronic safety permissions to the mobile intelligent terminal of the on-site maintenance personnel. The system tracks the execution status of the train control instructions and the execution progress of the maintenance work orders in real time, and realizes visual monitoring of the whole process.

[0052] According to the actual situation fed back on site, such as the complexity of the fault exceeding the expectation or the disposal being blocked, the system can dynamically adjust the disposal scheme and re-evaluate and optimize it. After the maintenance is completed, the on-site personnel feed back the completion status through the terminal, and the system generates an operation recovery suggestion scheme after verification, guiding the train dispatcher to orderly restore the normal diagram.

[0053] In addition, the method also constructs a continuous learning engine, uses the whole process data (from fault occurrence to operation recovery) of each emergency disposal, and optimizes the key models in a closed loop. The specific implementation is as follows: Model construction and training basis: (1) Collect the complete data chain of each emergency event, including: system initial evaluation and prediction, finally executed disposal scheme, real-time feedback (such as instruction execution status, on-site maintenance progress) in the scheme execution, and actual achieved operation recovery effect indicators.

[0054] (2) Differentiated machine learning methods are adopted for different optimization targets: Abnormality detection model: the core is a deep neural network classifier.

[0055] Simulation and deduction model: the core is a parameter-adjustable simulation and deduction model.

[0056] Scheme generation model: the core is a reinforcement learning strategy model.

[0057] Specific optimization process: (1) Fault identification model optimization: the newly generated and on-site verified fault cases are used as new labeled samples to incrementally train the original deep neural network classifier. This process enables the model to identify new fault patterns, reduces false positives and false negatives for known faults based on more abundant samples, and improves recognition accuracy and coverage.

[0058] (2) Impact assessment model optimization: Compare the simulation prediction results (e.g., predicted delay time) of historical events with the actual values. Based on the prediction bias, use the Bayesian optimization algorithm to automatically adjust the key behavior parameters (e.g., passenger evacuation speed, average driver reaction time after failure) in the simulation model. This process continuously improves the realism and prediction reliability of the simulation model, making the impact assessment more realistic.

[0059] (3) Scheme generation model optimization: Treat each complete emergency disposal as a decision sequence, and use the actual comprehensive benefits (e.g., operation interruption time, number of affected passengers) achieved as the reward signal to train the reinforcement learning strategy model. Through extensive case learning, the model gradually masters the best strategy for balancing train adjustment and maintenance work under complex constraints, and pushes the generated scheme from safe and feasible to optimal comprehensive benefits.

[0060] Embodiment 2 This embodiment is based on the above-mentioned embodiment 1, and discloses a smart linkage-based urban rail emergency command and train maintenance system. As shown in Figure 2 , the hardware foundation includes servers, network devices, dispatcher workstations, field mobile terminals, etc., and the software system includes the following core modules: Data access and fusion module: responsible for connecting with each external subsystem (ATS, TCMS, SCADA, etc.) through standard protocol interfaces, real-time acquisition of device state data of each subsystem and preprocessing, real-time feature vector acquisition, and realization of real-time collection, analysis and standardization of multi-source heterogeneous data.

[0061] Fault intelligent identification module: built-in fault rule library, rail transit knowledge graph and machine learning-based anomaly detection model, analyzes real-time feature vectors, realizes automatic identification, classification and positioning of faults.

[0062] Impact assessment and simulation deduction module: fusion of fault information, real-time train position, station passenger flow data and line topology structure, use of simulation deduction model, automatic assessment of potential impact range and degree of fault on train order, passenger service and subsequent train diagram, and generation of visual report.

[0063] Collaborative decision-making and scheme generation module: generates collaborative disposal scheme based on visual report.

[0064] Instruction issuance and execution control module: responsible for decomposing the scheme into specific instructions and reliably issuing them to the corresponding execution system (ATS, PIS, mobile terminal, etc.).

[0065] Process tracking and visualization module: real-time display of fault location, impact range, disposal progress, etc. through graphical interface (GIS, train diagram).

[0066] Knowledge base and learning optimization module: store historical data, cases and models, support self-learning and continuous improvement of the system.

[0067] Driving terminal and maintenance terminal: provide a human-computer interaction interface for dispatchers, for information presentation, scheme confirmation and manual intervention.

[0068] The modules are connected and exchange data through a high-speed internal communication bus and a standardized data interface. The collaborative decision-making module serves as the core and interacts with the influence evaluation module, the instruction issuing module and the knowledge base module bidirectionally, forming a closed-loop intelligent decision-making and control system.

[0069] The specific details of the above modules can be understood by referring to the related descriptions and effects in Embodiment 1.

[0070] Embodiment 3 On the basis of Embodiment 1, the electronic device provided in this embodiment includes one or more processors and a memory, the memory stores one or more programs, and the one or more programs include instructions for executing the intelligent linkage-based urban rail emergency command and driving maintenance method as described above.

[0071] At the hardware level, the electronic device includes a processor, an internal bus, a network interface, a memory and a non-volatile memory, and of course, it can also include other hardware required by the business. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs to implement the intelligent linkage-based urban rail emergency command and driving maintenance method described above. Of course, in addition to the software implementation, the present application does not exclude other implementation manners, such as logic devices or a combination of software and hardware, etc. That is, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or a logic device.

[0072] The memory can include a non-permanent memory in a computer readable medium, a random access memory (RAM) and / or a non-volatile memory such as a read-only memory (ROM) or a flash memory (flash RAM). The memory is an example of a computer readable medium.

[0073] Computer-readable media includes permanent and non-permanent, movable and non-movable media that can implement information storage by any method or technology. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible by a computing device. According to the definition herein, computer-readable media does not include transitory media such as modulated data signals and carriers.

[0074] The above description is merely a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present application, and these modifications or replacements should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A smart linkage-based urban rail emergency command and driving maintenance method, characterized in that, The method comprises: Real-time acquisition of device state data of each subsystem through a standard protocol interface and preprocessing to obtain real-time feature vectors; Real-time analysis of the real-time feature vectors using a preset fault rule library, a rail transit knowledge graph, and an abnormality detection model based on machine learning to realize automatic identification, classification, and positioning of faults and obtain fault information; Fusion of the fault information, real-time train position, station passenger flow data, and line topology structure, automatic evaluation of the potential impact range and degree of the fault on train operation order, passenger service, and subsequent train operation diagram based on a preset simulation deduction model, and generation of a visual report; Based on the visual report, a collaborative disposal scheme is generated, and the report and the scheme are issued together, real-time acquisition of actual situations fed back on site, and dynamic adjustment of the collaborative disposal scheme and the visual report.

2. The urban rail emergency command and driving maintenance method based on intelligent linkage according to claim 1, characterized in that, When the rail transit system is started, basic data is loaded, a rail transit knowledge graph is constructed based on the basic data, and the correlation between devices, faults, impacts, and disposal measures is established.

3. The smart linkage-based urban rail emergency command and driving maintenance method according to claim 2, characterized in that, The basic data includes line device account, line operation topology graph, real-time operation diagram, historical fault library, maintenance procedure, and emergency disposal plan; The line device account includes device ID, physical location, model, and subsystem to which the device belongs; The line operation topology graph includes interval distribution, station location, and spare line configuration; The real-time operation diagram includes train schedule, stop time, and route planning; The historical fault library includes fault type in recent years, occurrence time, and disposal result; The maintenance procedure includes maintenance steps and safety requirements for each device; The emergency disposal plan includes standard response processes for different fault scenarios.

4. The smart linkage-based urban rail emergency command and driving maintenance method according to claim 2, characterized in that, The process of constructing the rail transit knowledge graph based on the basic data comprises: Based on the business needs of urban rail emergency command and train maintenance, knowledge classification ontology is obtained, a knowledge graph framework is formed according to the core correlation logic between each ontology, and the knowledge classification ontology includes device ontology, fault ontology, impact ontology, disposal ontology, and case ontology; Knowledge graph data is extracted from the basic data based on the knowledge graph framework and is preprocessed; Based on the preprocessed knowledge graph data and the knowledge graph framework, knowledge nodes of the knowledge graph and their correlation are obtained; The knowledge nodes, knowledge graph data, and their correlation are imported into a database to generate visual nodes and edges of the knowledge graph and establish an index, and the construction of the rail transit knowledge graph is completed.

5. The smart linkage-based urban rail emergency command and driving maintenance method according to claim 4, characterized in that, The process of mapping the preprocessed knowledge graph data to knowledge nodes of the knowledge graph and their correlation according to the knowledge graph framework comprises: For structured data in the knowledge graph data, structured data knowledge graph entities and their correlation are obtained by mapping according to the knowledge graph framework; For unstructured data in the knowledge graph data, unstructured data knowledge graph entities and their correlation are obtained by combining NLP tools and domain rules; Duplicate entities are eliminated through normalization processing, and relationship conflicts or repetitions between different entities are eliminated through priority rules or voting mechanisms. The entity and its associated relationship of the processed structured data knowledge graph and the entity and its associated relationship of the acquired unstructured data knowledge graph are merged as knowledge nodes and their associated relationship of the knowledge graph and output.

6. The smart linkage-based urban rail emergency command and driving maintenance method according to claim 1, characterized in that, The fault rule base is constructed based on domain expert experience and historical fault cases, and contains IF-THEN logic.

7. The smart linkage-based urban rail emergency command and driving maintenance method according to claim 1, characterized in that, The process of automatic identification, classification and positioning of the fault includes: Automatic identification: matching the real-time feature vector with the preset fault rule base to quickly identify known and pattern-specific faults; for potential faults that cannot be covered by the rule base or whose patterns are not specific, an abnormality detection model based on machine learning is used for analysis, the abnormality detection model identifies data points that deviate significantly from the normal operation mode through unsupervised learning and marks them as abnormal states, realizing the identification of unknown faults or early implicit faults; Classification: based on the automatic identification result, the fault event is accurately classified by support vector machine or decision tree to determine its major category and subcategory; Positioning: based on the line topology structure and the constructed rail transit knowledge graph, the positioning of the fault equipment in the network and its associated relationship with other equipment are analyzed; the alarm and state information from different subsystems are fused to cross-verify the fault positioning and output the positioning result; if the fault cannot be directly positioned, a probability model based on Bayesian network is used to comprehensively consider the conditional probability between the historical fault rate, current state, environmental factors and alarm information of each device, calculate the probability of each device being the fault source, and list one or several devices with the highest probability as the positioning result.

8. The smart linkage-based urban rail emergency command and driving maintenance method according to claim 1, characterized in that, The simulation deduction model is a hybrid simulation model constructed based on digital twinning technology, integrating line topology, train operation characteristics, signal logic and passenger flow rules, and calibrated using historical operation data.

9. The smart linkage-based urban rail emergency command and driving maintenance method according to claim 1, characterized in that, Based on the simulation deduction model, the process of automatically calculating the potential impact range and degree of the fault on train operation order, passenger service and subsequent running graph includes: Based on the fault information, real-time train position, station passenger flow data and line topology structure, the simulation parameters of the simulation deduction model are set, and the simulation parameters are calibrated based on historical operation data; Based on the simulation deduction model, the potential impact range and degree of the fault on train operation order, passenger service and subsequent running graph are evaluated; wherein, The potential impact on the train operation order includes single train direct delay calculation and subsequent train chain delay calculation; the evaluation indexes of the potential impact range and degree on the train operation order include the total number of affected trains, the maximum delay of a single train, the total delay time and the running graph deviation rate; The potential impact on the passenger service includes station passenger backlog and total number of affected passengers; the evaluation indexes of the potential impact range and degree on the passenger service include station passenger backlog rate and proportion of affected passengers; The potential impact on the subsequent running graph includes running graph recovery time; the evaluation indexes of the potential impact range and degree on the subsequent running graph include subsequent running graph deviation accumulation.

10. The smart linkage-based urban rail emergency command and driving maintenance method according to claim 1, characterized in that, The collaborative disposal scheme includes train operation scheme and maintenance scheme, and the process of generating the collaborative disposal scheme based on the visual report includes: The driving scheme generation is generated based on the visual report, with safety and maximum operation maintenance as hard constraints, combined with a preset driving adjustment strategy library, through a constraint planning algorithm to generate several preliminary driving plans; the simulation deduction model is used for simulation calculation of key KPIs of each preliminary driving plan, and the preliminary driving plan with a key KPI higher than a preset threshold is reserved as a driving scheme; The maintenance scheme generation is based on fault types and positioning, and matches corresponding maintenance procedures and historical similar cases from the knowledge graph; combined with real-time maintenance resource data, resource schedulability is determined, and a standardized maintenance scheme including fault positioning, maintenance steps, resource list and estimated repair time is generated; The scheme linkage optimization takes the shortest operation interruption time and the lowest overall resource consumption as the core target, and integrates the adjustment time / amplitude of the driving scheme and the operation window / resource scheduling of the maintenance scheme into the same model; a genetic algorithm or a particle swarm algorithm is used for iterative calculation of the global benefits of different driving schemes and maintenance scheme combinations, and conflict points are evaluated; the globally optimal driving scheme and maintenance scheme combination is output as a collaborative disposal scheme. 11.An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein, The processor implements the steps of the intelligent linkage-based urban rail emergency command and driving maintenance method according to any one of claims 1-10 when executing the computer program. 12.A computer readable storage medium having a computer program stored thereon, wherein, The computer program implements the steps of the intelligent linkage-based urban rail emergency command and driving maintenance method according to any one of claims 1-10 when executed by a processor.

Images