Intelligent operation and maintenance method and device applied to subway faults and electronic equipment

Abstract

The invention provides an intelligent operation and maintenance method and device applied to subway faults and electronic equipment, and relates to the field of data processing. In the method, multi-source data for a target subway are obtained, and the multi-source data comprise sensor data, maintenance records and equipment logs; constructing a fault knowledge graph according to the multi-source data; fusing the multi-source data and the fault knowledge graph by adopting a deep learning model to obtain feature data and a fusion model; generating a fault tree based on the feature data and the fusion model; acquiring real-time monitoring data for the target subway; and performing fault prediction on the real-time monitoring data through the fault tree to generate a multi-dimensional fault report. By implementing the technical scheme provided by the invention, intelligent operation and maintenance of subway faults are facilitated.

Description

Technical Field

[0001] The present application relates to the technical field of data processing, and in particular to an intelligent operation and maintenance method, device and electronic equipment applied to subway failures. Background Art

[0002] In modern urban transportation, subways, as an important means of public transportation, carry the growing passenger demand. However, with the increase in the number of operating lines and trains, the risk of failure in the subway system has also increased significantly. Train delays caused by failures not only affect passengers' travel experience, but also have a negative impact on the overall efficiency of urban transportation. Therefore, the failure operation and maintenance of subway equipment, including detection and prediction, has become a key issue that needs to be solved urgently.

[0003] At present, the traditional subway fault operation and maintenance methods mainly rely on regular manual inspections and empirical judgments. First, the frequency and quality of manual inspections vary, making it difficult to ensure the timely discovery of potential faults. Second, many faults are hidden and often cannot be discovered through simple observation or regular inspections in the early stages, resulting in the continued operation of equipment without sufficient evaluation, which poses a huge safety hazard. Therefore, the above-mentioned regular manual inspections and empirical judgment methods are not conducive to intelligent operation and maintenance of subway faults.

[0004] Therefore, there is an urgent need for an intelligent operation and maintenance method, device and electronic equipment applied to subway failures. Summary of the invention

[0005] The present application provides an intelligent operation and maintenance method, device and electronic equipment applied to subway failures, which facilitates intelligent operation and maintenance of subway failures.

[0006] In a first aspect of the present application, an intelligent operation and maintenance method for subway faults is provided, the method comprising: obtaining multi-source data for a target subway, the multi-source data comprising sensor data, maintenance records and equipment logs; constructing a fault knowledge graph based on the multi-source data; using a deep learning model to fuse the multi-source data and the fault knowledge graph to obtain feature data and a fusion model; generating a fault tree based on the feature data and the fusion model; obtaining real-time monitoring data for the target subway; performing fault prediction on the real-time monitoring data through the fault tree to generate a multi-dimensional fault report.

[0007] By adopting the above technical solutions, by collecting multiple data sources such as sensor data, maintenance records and equipment logs, the operating status of subway equipment can be fully understood. The fusion of multi-source data helps to improve the accuracy of fault analysis and ensure all-round monitoring of equipment status. The knowledge graph can systematically organize and express the relevant knowledge of equipment failure. This structured information can help engineers find and understand the fault background more quickly and improve the efficiency of troubleshooting. Using deep learning models for feature extraction and data fusion can automatically identify complex nonlinear relationships, extract more representative features, and improve the accuracy of prediction models. This intelligent processing can reduce manual intervention and reduce labor costs. Generating a fault tree based on feature data and fusion models can systematically display the hierarchical relationship and causes of potential faults. This visualized information makes fault analysis more intuitive and facilitates operation and maintenance personnel to identify key fault nodes, thereby making more effective maintenance decisions. By obtaining real-time monitoring data, the system can timely reflect the operating status of the equipment and make predictions and warnings as soon as the problem occurs. This real-time performance greatly enhances the initiative and effectiveness of maintenance work. Generating multi-dimensional fault reports can help decision makers analyze fault conditions from different angles. The visualization of this information can provide a scientific basis for operation and maintenance management. Therefore, it is convenient to carry out intelligent operation and maintenance of subway failures.

[0008] Optionally, the acquiring of multi-source data for the target subway specifically includes: receiving original operation data sent by a sensor network for multiple operating devices of the target subway, the original operation data including temperature data, vibration data, voltage data and current data; acquiring original maintenance records from a maintenance management system corresponding to the target subway through an ETL tool, the original maintenance records including a description of each equipment failure, cause, maintenance time and treatment measures; collecting original log information of the target subway from a subway system, the original log information including operation logs and alarm information; preprocessing the original operation data, the original maintenance records and the original log information to obtain the multi-source data, the preprocessing including denoising, filtering and normalization.

[0009] By adopting the above technical solution, by receiving the original operation data from the sensor network, obtaining the maintenance records in the maintenance management system, and collecting the operation logs and alarm information, it is possible to ensure a comprehensive and systematic understanding of the operation status of the target subway equipment. This multi-dimensional data collection provides a solid foundation for subsequent analysis. Preprocessing the original operation data, maintenance records and log information can significantly improve the data quality. Steps such as denoising, filtering and normalization can eliminate interference and outliers in the data, ensuring that the data used for analysis is more accurate and reliable. Preprocessing steps such as normalization help eliminate the dimension and range differences between different data sources, so that various types of data are consistent when analyzed. This consistency is crucial for subsequent data fusion and model training. By denoising and filtering the data, the complexity of subsequent analysis can be simplified and the errors that may occur during data processing can be reduced. This makes the model and algorithm more efficient in processing data, and improves the feasibility and accuracy of the overall analysis. The collected real-time sensor data and detailed maintenance records provide the necessary information basis for real-time monitoring and fault prediction. Through the analysis of real-time data, potential fault risks can be identified in a timely manner, thereby improving the initiative of operation and maintenance. After preprocessing, multi-source data can be directly applied to data analysis and model building, which helps improve the training effect and accuracy of deep learning and other analytical models. The data integration and processing process provides good input data for subsequent fault detection and prediction models.

[0010] Optionally, constructing a fault knowledge graph based on the multi-source data specifically includes: performing named entity recognition and relationship extraction on the maintenance record and the equipment log to obtain a first entity, a second entity and an association relationship between the first entity and the second entity, the first entity and the second entity being any two of a plurality of devices included in the target subway; constructing a triple with the first entity, the second entity and the association relationship, and gradually organizing the information to obtain an initial knowledge graph; labeling the sensor data to obtain a fault event, the fault event being used to represent the association between key data and the cause of the fault; and adding the fault event to the initial knowledge graph to obtain the fault knowledge graph.

[0011] By adopting the above technical solution, through the analysis of maintenance records and equipment logs, the information of equipment failure can be systematically identified and recorded. This systematic approach makes the fault information more structured and convenient for subsequent query and analysis. Through named entity recognition and relationship extraction, key equipment and the relationship between them can be identified, which enhances the availability of fault knowledge. This enables operation and maintenance personnel to quickly understand the mutual influence between equipment when analyzing faults, and improves the efficiency of decision-making. The association relationship between the first entity and the second entity is converted into a triple to form a causal relationship network, which is convenient for understanding and analyzing the cause and impact of the fault. This relationship network helps to identify potential failure modes, so as to better conduct risk assessment. By integrating sensor data, maintenance records and fault events into the same knowledge graph, a comprehensive data view can be formed. This integration makes fault analysis not only rely on historical maintenance records, but also combines real-time sensor data to improve the accuracy of fault detection. The fault knowledge graph can provide a knowledge basis for intelligent fault diagnosis. By annotating fault events, the system can identify the association between key data and fault causes, promote automated reasoning and analysis based on knowledge graphs, and improve the efficiency and accuracy of fault diagnosis.

[0012] Optionally, the deep learning model is used to fuse the multi-source data and the fault knowledge graph to obtain feature data and a fusion model, specifically including: using an LSTM model to extract features from the sensor data to obtain a feature matrix; using a BERT model to generate feature vectors for the maintenance records and the equipment logs to obtain text vectors; using a GNN graph neural network to perform feature learning on the fault knowledge graph to obtain graph structure features; and comprehensively encoding the feature matrix, the text vector and the graph structure features through a combined model of a graph convolutional network and a Transformer to obtain the feature data and the fusion model.

[0013] By adopting the above technical solutions, the time series features in time series data can be effectively captured by using the LSTM model to extract features from sensor data. This capability helps to identify patterns and trends in data and improve the accuracy of fault detection and prediction. The BERT model is used to generate feature vectors for maintenance records and equipment logs, which can fully mine the semantic information in text data. This two-way context understanding capability enables the model to better capture the relationship between fault descriptions and maintenance measures, thereby improving the depth of fault analysis. The GNN graph neural network is used to learn features of the fault knowledge graph, which can effectively mine the complex relationships and structural information between devices. The introduction of graph structure features can enhance the model's understanding of the mutual influence between devices, thereby providing a more comprehensive perspective on fault analysis. By combining time series features, text vectors, and graph structure features, the fusion model can make comprehensive use of multiple types of data. This fusion of multimodal data allows the model to analyze faults from different angles and provide more comprehensive and accurate fault prediction and diagnosis. Combining multiple deep learning architectures for feature encoding can enhance the adaptability and robustness of the model to different data types. The combination of different models can complement their advantages and disadvantages and improve the performance of the overall model. By integrating multiple feature information, the fusion model can more accurately capture potential fault signals and improve the accuracy of fault prediction. This accurate prediction helps to identify potential problems in advance and reduce equipment downtime and maintenance costs.

[0014] Optionally, generating a fault tree based on the feature data and the fusion model specifically includes: obtaining equipment failure, failure cause and specific failure type according to the feature data; constructing root nodes, child nodes and leaf nodes corresponding to the equipment failure, failure cause and specific failure type to obtain a basic framework; processing the fusion model with a graph algorithm to obtain a weight relationship; generating the fault tree based on the basic framework and the weight relationship.

[0015] By adopting the above technical solution, by clearly distinguishing equipment failures, failure causes and specific failure types, fault information can be systematically analyzed and recorded. This structured approach makes the understanding and handling of faults clearer and helps operation and maintenance personnel quickly identify the root cause of the problem. By constructing the root node, child nodes and leaf nodes, the basic framework of the fault tree is generated, which can intuitively display the hierarchical relationship of the fault. This hierarchical structure is not only convenient for visualizing fault information, but also helps users quickly locate the specific fault type and cause. By processing the weight relationship obtained by the fusion model through the graph algorithm, we can deeply understand the mutual influence and relationship between different faults. This understanding helps to identify the fault propagation path and reduce potential risks. The generated fault tree provides a basis for decision-making. The operation and maintenance team can formulate corresponding maintenance and response strategies based on the information in the fault tree to effectively prevent the occurrence and spread of faults. The fault tree can systematically organize potential faults and their causes, help predict possible future faults, and take corresponding preventive measures. This proactive maintenance can reduce the incidence of equipment failures. Through the structured display of the fault tree, operation and maintenance personnel can quickly identify and handle faults. The clear fault hierarchy enables technicians to locate and repair more efficiently when troubleshooting.

[0016] Optionally, the fault prediction of the real-time monitoring data is performed through the fault tree to generate a multi-dimensional fault report, specifically including: inputting the real-time monitoring data into the fault tree, and using the Bayesian network to perform probability inference to obtain a potential fault probability; if it is determined that the potential fault probability is greater than or equal to a preset threshold, generating the multi-dimensional fault report, the multi-dimensional fault report including a fault trend graph and a cause distribution graph.

[0017] By adopting the above technical solution, the real-time monitoring of equipment status can be achieved by inputting real-time monitoring data into the fault tree. This real-time performance ensures that the operation and maintenance team can quickly identify potential faults and reduce the impact of equipment failures. Using Bayesian networks for probabilistic inference, the probability of potential faults can be quantitatively evaluated. This probability-based approach helps to assess risks more accurately, so that fault prediction does not only rely on historical data, but is dynamically adjusted in combination with the current status. Setting preset thresholds can be flexibly adjusted according to actual needs and operating conditions, making fault warnings more personalized and accurate. When the probability of potential faults reaches or exceeds the threshold, timely generating reports can effectively reduce the risk of missing key fault signals. The generated multi-dimensional fault report includes fault trend graphs and cause distribution graphs, which can provide operation and maintenance personnel with a more comprehensive perspective on faults. This diverse data display helps to quickly identify fault modes, understand fault development trends, and analyze fault causes.

[0018] Optionally, the method further includes: determining a fault level according to the multi-dimensional fault report; and sending the multi-dimensional fault report to operation and maintenance personnel according to the fault level.

[0019] By adopting the above technical solutions, by determining the fault level, the operation and maintenance personnel can clearly understand the severity and priority of the fault. This classification helps to prioritize the most urgent and serious faults within limited resources and time, ensuring the stable operation of key equipment. Reasonable allocation of resources according to the fault level enables the operation and maintenance team to invest manpower and material resources in fault handling in a targeted manner. Effective resource allocation can reduce ineffective resource waste and improve maintenance efficiency. Send multi-dimensional fault reports to operation and maintenance personnel in a timely manner, so that they can quickly obtain information about the fault. This rapid information transmission can shorten the fault response time and ensure that the problem is resolved as soon as possible. According to the classification of fault levels, the operation and maintenance team can formulate corresponding processing processes and standard operating procedures. Clear processes can improve team collaboration efficiency and reduce confusion and uncertainty in fault handling.

[0020] In a second aspect of the present application, an intelligent operation and maintenance device applied to subway faults is provided, and the intelligent operation and maintenance device includes an acquisition module and a processing module, wherein the acquisition module is used to acquire multi-source data for a target subway, and the multi-source data includes sensor data, maintenance records, and equipment logs; the processing module is used to construct a fault knowledge graph based on the multi-source data; the processing module is also used to use a deep learning model to fuse the multi-source data and the fault knowledge graph to obtain feature data and a fusion model; the processing module is also used to generate a fault tree based on the feature data and the fusion model; the acquisition module is also used to acquire real-time monitoring data for the target subway; the processing module is also used to perform fault prediction on the real-time monitoring data through the fault tree to generate a multi-dimensional fault report.

[0021] In the third aspect of the present application, an electronic device is provided, which includes a processor, a memory, a user interface and a network interface, the memory is used to store instructions, the user interface and the network interface are both used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device performs the method described above.

[0022] In a fourth aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores instructions, and when the instructions are executed, the method described above is executed.

[0023] In summary, one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: By collecting multiple data sources such as sensor data, maintenance records, and equipment logs, the operating status of subway equipment can be fully understood. The fusion of multi-source data helps to improve the accuracy of fault analysis and ensure all-round monitoring of equipment status. Knowledge graphs can systematically organize and express relevant knowledge about equipment failures. This structured information can help engineers find and understand the fault background more quickly and improve the efficiency of troubleshooting. Using deep learning models for feature extraction and data fusion can automatically identify complex nonlinear relationships, extract more representative features, and improve the accuracy of prediction models. This intelligent processing can reduce manual intervention and reduce labor costs. Generating a fault tree based on feature data and fusion models can systematically display the hierarchical relationship and causes of potential faults. This visualized information makes fault analysis more intuitive and facilitates operation and maintenance personnel to identify key fault nodes, thereby making more effective maintenance decisions. By obtaining real-time monitoring data, the system can promptly reflect the operating status of the equipment and make predictions and warnings as soon as the problem occurs. This real-time performance greatly enhances the initiative and effectiveness of maintenance work. Generating multi-dimensional fault reports can help decision makers analyze fault conditions from different angles, and the visualization of this information can provide a scientific basis for operation and maintenance management. Therefore, it is convenient to perform intelligent operation and maintenance of subway faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A flow chart of an intelligent operation and maintenance method for subway failures provided in an embodiment of the present application.

[0025] Figure 2 Another flow chart of an intelligent operation and maintenance method applied to subway failures provided in an embodiment of the present application.

[0026] Figure 3 A module diagram of an intelligent operation and maintenance device for subway failures provided in an embodiment of the present application.

[0027] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0028] Explanation of the reference numerals: 31, acquisition module; 32, processing module; 41, processor; 42, communication bus; 43, user interface; 44, network interface; 45, memory. DETAILED DESCRIPTION

[0029] In order to enable technicians in this field to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments.

[0030] In the description of the embodiments of the present application, words such as "for example" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "for example" or "for example" is intended to present related concepts in a specific way.

[0031] In the description of the embodiments of the present application, the meaning of the term "multiple" refers to two or more. For example, multiple systems refer to two or more systems, and multiple screen terminals refer to two or more screen terminals. In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. The terms "include", "comprise", "have" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0032] In the modern urban transportation system, the subway, as a vital public transportation tool, carries more and more passenger demand and has become the first choice for citizens' daily travel. However, with the continuous expansion of subway operating lines and the increase in the number of trains, the risk of failure faced by the subway system is also increasing significantly. Once these failures occur, they often lead to train delays, directly affecting the travel experience of passengers, and also have a negative impact on the efficiency and order of the entire city's transportation. Therefore, how to effectively carry out fault operation and maintenance of subway equipment, including fault detection and prediction, has become a key issue that needs to be solved urgently.

[0033] At present, traditional subway fault operation and maintenance methods mainly rely on regular manual inspections and empirical judgments. Although these methods can work to some extent, their limitations are also very obvious. First, the frequency and quality of manual inspections are often uneven, and there is a certain degree of randomness and uncertainty. This makes it more difficult to detect and deal with potential faults in a timely manner. Secondly, many equipment failures are hidden and difficult to detect through simple visual observation or regular inspections in the early stages. This leads to some equipment continuing to be put into operation without sufficient evaluation and testing, thereby increasing safety hazards and failure risks.

[0034] In addition, the method of fault identification based on empirical judgment often lacks scientificity and systematicity. Although the personal experience of operation and maintenance personnel can provide help in some cases, it is also easily affected by subjective factors and leads to misjudgment. In summary, this traditional method makes intelligent operation and maintenance of subway faults more difficult.

[0035] In order to solve the above technical problems, this application provides an intelligent operation and maintenance method for subway failures. Figure 1 , Figure 1 A flowchart of an intelligent operation and maintenance method for subway failures provided in an embodiment of the present application. The intelligent operation and maintenance method is applied to a server and includes steps S110 to S160, which are as follows: S110 . Acquire multi-source data for a target subway, where the multi-source data includes sensor data, maintenance records, and equipment logs.

[0036] Specifically, a server refers to a centralized data management system that is responsible for collecting and storing data from different sources. In the process of subway operation and maintenance, a server can be a dedicated computer system or cloud platform that can process large amounts of data and conduct subsequent analysis. The target subway refers to a specific subway line or train system, which can be a specific subway network in a certain city. This specificity helps to focus the collected data on a specific operation or maintenance need. Multi-source data means that the data comes from different channels and types. Such a data set can provide a more comprehensive perspective and help with accurate analysis and judgment. For the subway system, these data sources include: sensor data, that is, real-time collection of various physical status information from subway operating equipment, such as temperature, pressure, vibration, etc., for monitoring the health status of the equipment; maintenance records, that is, the maintenance and repair history of the equipment, including fault description, maintenance time, maintenance measures and other information, which helps to understand the maintenance background and failure mode of the equipment; equipment logs, which record various operation information during the operation of the equipment, such as operator operations, alarm information, etc., for analyzing the working conditions and abnormal conditions of the equipment.

[0037] In a possible implementation, multi-source data for a target subway is obtained, specifically including: receiving original operation data sent by a sensor network for multiple operating devices of the target subway, the original operation data including temperature data, vibration data, voltage data, and current data; obtaining original maintenance records from a maintenance management system corresponding to the target subway through an ETL tool, the original maintenance records including a description of each equipment failure, cause, maintenance time, and treatment measures; collecting original log information of the target subway from the subway system, the original log information including operation logs and alarm information; preprocessing the original operation data, original maintenance records, and original log information to obtain multi-source data, the preprocessing including denoising, filtering, and normalization.

[0038] Specifically, the received sensor data refers to the raw operation data collected through the sensor network. These sensors are installed on various equipment in the subway, such as motors, carriages, tracks, etc., and can monitor the status of the equipment in real time. The above raw operation data includes: temperature data, that is, monitoring the temperature of equipment such as motors and transformers during operation to avoid overheating; vibration data, that is, detecting the vibration of the equipment to determine whether there is an abnormality, such as looseness or wear; voltage data, that is, recording the power supply and equipment working voltage to ensure that the equipment operates within the normal voltage range; current data, that is, monitoring the current of the equipment to analyze its energy efficiency and load conditions. ETL tools are used to extract data from the maintenance management system, convert the data format and load it into the database. The original maintenance records include: description of each equipment failure, such as "motor overheating" or "brake failure"; cause of failure, which may be "parts aging" or "improper operation"; maintenance time, that is, recording the specific maintenance date and time; treatment measures, specific measures taken during the maintenance process, such as replacing parts or adjusting settings.

[0039] The above-mentioned raw log information includes: operation log, which records all the operations of the operator during the operation process, such as "train start", "train stop", etc. Alarm information, that is, when the equipment is abnormal, the system will generate an alarm, such as "motor overload alarm", and record it in the log. Before the collected data is used for analysis, it needs to be preprocessed to ensure the quality and validity of the data. The preprocessing steps include: denoising, cleaning the noise in the data, that is, unnecessary interference information, to ensure the accuracy of the data. Filtering, by using algorithms to filter out some useless data points and retain only important information. Normalization processing, by converting data from different data sources to the same standard range for easy comparison and analysis.

[0040] For example, the subway motor is equipped with temperature and vibration sensors. The temperature data recorded by the temperature sensor during operation is 70°C, while the vibration data recorded by the vibration sensor is within the normal range. At this time, the server will receive this data in real time to monitor the operating status of the motor. After a certain equipment failure, the operation and maintenance team used the ETL tool to extract relevant records from the maintenance management system and found that "On January 29, 2025, the motor overheated, the reason: cooling system failure, treatment measures: replace coolant." These data are converted into a structured format for subsequent analysis. Suppose that in a shift, the operator recorded the "train start" and "train stop" events, and triggered the "brake failure" alarm after stopping. All operation and alarm information will be recorded in the log for subsequent analysis and troubleshooting. In data preprocessing, if there are sudden abnormal values ​​in the temperature data, such as 150°C recorded instantaneously, this will be regarded as noise and removed. At the same time, the temperature data is filtered to eliminate rapid fluctuations in a short period of time, and finally all temperature data are converted to the range of 0 to 1 so that the subsequent machine learning model can effectively use this data.

[0041] S120. Construct a fault knowledge graph based on multi-source data.

[0042] Specifically, a knowledge graph is a graphical structure that uses nodes and edges to represent entities and their relationships. In the fault knowledge graph, nodes represent different devices, fault types, fault causes, etc., while edges represent the associations between them, such as "causal relationship" and "similar relationship". By building a knowledge graph, the server can better understand and analyze the failure mode, cause of failure and their mutual correlation of the equipment. This structured information storage method helps with subsequent fault prediction and decision support.

[0043] In a possible implementation, a fault knowledge graph is constructed based on multi-source data, specifically including: performing named entity recognition and relationship extraction on maintenance records and equipment logs to obtain a first entity, a second entity, and an association relationship between the first entity and the second entity, where the first entity and the second entity are any two devices among multiple devices included in the target subway; constructing triples with the first entity, the second entity, and the association relationship, and gradually organizing the information to obtain an initial knowledge graph; labeling sensor data to obtain fault events, which are used to represent the association between key data and fault causes; and adding the fault events to the initial knowledge graph to obtain a fault knowledge graph.

[0044] Specifically, named entity recognition is a natural language processing technology that aims to identify meaningful entities from text, such as equipment names, fault types, etc. For example, equipment entities such as "motor" and "brake system" are identified from maintenance records. After identifying the entities, it is necessary to extract the relationships between them, such as the relationship between "motor" and "motor overheating". Through this process, the association information between multiple devices can be obtained. Through named entity recognition and relationship extraction, two specific devices, such as the first entity and the second entity, and the relationship between them are identified. For example, the first entity is a motor, the second entity is a cooling system, and the association relationship is that the motor overheating is caused by a cooling system failure.

[0045] Among them, a triple is a form of expressing a relationship, which consists of three parts: subject, predicate and object. In an embodiment of the present application, the constructed triples are as follows: (motor, causes, motor overheating), (cooling system, affects, motor). According to the constructed triples, information is gradually integrated and organized to form a preliminary knowledge graph. At this stage, the server will integrate all identified entities and relationships into a structured graphical representation. The server annotates the data collected from the sensor, identifies key data related to the fault, and forms a fault event. For example, an event in which the temperature recorded by a sensor exceeds a set threshold can be labeled as a "motor overheating event." The server integrates the annotated fault event information into the initial knowledge graph to further enrich and improve the content of the knowledge graph. This can reflect the association between the fault event and the equipment and the cause of the fault, forming a final fault knowledge graph.

[0046] S130. Use a deep learning model to fuse multi-source data and a fault knowledge graph to obtain feature data and a fusion model.

[0047] Specifically, deep learning is a machine learning method that uses a neural network model to learn features and patterns from data. In an embodiment of the present application, the server will apply one or more deep learning models to analyze multi-source data and fault knowledge graphs. For example, the model may learn the possibility of a fault occurring under a specific sensor reading, as well as the potential causes of the fault. Feature data is useful information extracted from multi-source data and fault knowledge graphs in numerical form for subsequent analysis. The fusion model is a model trained through deep learning that can effectively process and predict faults. This model not only takes into account the characteristics of the original data, but also the knowledge and relationships implicit in the knowledge graph.

[0048] In one possible implementation, a deep learning model is used to fuse multi-source data and a fault knowledge graph to obtain feature data and a fusion model, specifically including: using an LSTM model to extract features from sensor data to obtain a feature matrix; using a BERT model to generate feature vectors for maintenance records and equipment logs to obtain text vectors; using a GNN graph neural network to learn features of the fault knowledge graph to obtain graph structure features; and using a combined model of a graph convolutional network and a Transformer to comprehensively encode the feature matrix, text vectors, and graph structure features to obtain feature data and a fusion model.

[0049] Specifically, the server combines multi-source data, such as sensor data, maintenance records, and equipment logs, with the fault knowledge graph to extract useful features and information. Deep learning models are able to process these complex data structures and learn from them. LSTM is a recurrent neural network for time series data, which is particularly good at processing sequence data with time dependencies. In the embodiment of the present application, LSTM will be used to extract features of sensor data, such as temperature, vibration, and other data that change over time. These features are integrated into a feature matrix, and each row in the matrix may represent a data feature of a time step. BERT is a powerful pre-trained language model suitable for processing natural language text. In the embodiment of the present application, BERT is used to analyze maintenance records and equipment logs, extract semantic information therein, and generate text vectors. Text vectors can capture context-related semantic information, such as keywords and sentence structures in equipment fault descriptions. GNN is a deep learning model for graph-structured data that can handle complex relationships between nodes and edges. In this step, GNN will perform feature learning on the fault knowledge graph and extract structural features from it. These features may include connectivity between devices and fault causality, which can better reflect the mutual influence of devices. Graph convolutional networks and Transformers are both deep learning models used to process complex data structures. In this part, we first use graph convolutional networks to process the graph structure features obtained from GNN, and combine the feature matrix extracted by LSTM and the text vector generated by BERT. This can integrate information from different sources. The feature data finally generated will be a high-dimensional vector that can fully express sensor data, maintenance records and their corresponding knowledge graph information. The fusion model is a trained deep learning model that can be used for subsequent fault prediction and analysis.

[0050] For example, suppose there is sensor data of a motor, which records its temperature over a period of time, such as 85°C, 90°C, 92°C, and vibration, such as 0.5g, 0.7g, 0.9g, etc. The LSTM model inputs this data and extracts time-related features, such as the rising trend of temperature and abnormal changes in vibration, by analyzing the time series, thereby generating a feature matrix. In addition, a maintenance record is "On January 15, 2024, the motor was repaired due to overheating failure." The server uses the BERT model to analyze the text, which can generate a text vector that captures the semantic information of "motor", "overheating" and "repair" and converts it into digital form. Assume that the fault knowledge graph contains nodes "motor", "cooling system" and edges "cooling system failure causes motor overheating". GNN will extract graph structure features, such as the fault causal relationship between the motor and the cooling system, by analyzing these nodes and their relationships. Finally, the feature matrix, text vector and graph structure features obtained from LSTM, BERT and GNN are combined and processed by graph convolutional network and Transformer to generate the final feature data and fusion model. This model can effectively capture the temporal features of sensor data, the semantic features of text information and the structural features in the knowledge graph. Therefore, by using different types of deep learning models to fuse multi-source data and fault knowledge graphs, rich feature data and fusion models are finally obtained. This model has the ability to process complex data structures, providing a solid foundation for subsequent fault prediction and intelligent operation and maintenance, thereby improving the reliability and responsiveness of the system.

[0051] S140: Generate a fault tree based on the characteristic data and the fusion model.

[0052] Specifically, feature data refers to high-dimensional data extracted and fused from multi-source data through deep learning models, including LSTM, BERT and GNN. Feature data contains information related to equipment performance, fault history and fault relationships. A fusion model refers to a trained deep learning model that combines the features of multiple data types to more accurately predict and analyze faults. This model can understand the relationship between different data and provide a more comprehensive basis for fault analysis. A fault tree is a fault analysis tool presented in a graphical manner, including a root node, representing a top-level fault, and multiple child nodes, representing various causes of root node failures. In the process of generating a fault tree, the server maps the information in the feature data to the structure of the fault tree, clarifying the potential faults, fault causes and specific fault types of each device. In an embodiment of the present application, the root node represents a major fault of the system, and the child nodes represent the various factors that cause this fault.

[0053] In a possible implementation, a fault tree is generated based on feature data and a fusion model, specifically including: obtaining equipment failure, failure cause and specific failure type according to the feature data; constructing root nodes, child nodes and leaf nodes according to the equipment failure, failure cause and specific failure type to obtain a basic framework; processing the fusion model with a graph algorithm to obtain a weight relationship; and generating a fault tree based on the basic framework and the weight relationship.

[0054] Specifically, at this stage, by analyzing feature data, the server can identify equipment failures, such as "motor overheating", as well as the cause of the failure, such as "insufficient coolant", and the specific type of failure, such as "fan failure". The server uses graph algorithms, such as graph convolutional networks, to analyze the features in the fusion model, identify the degree of correlation between different faults, and obtain weight relationships. These weights can help determine which factors have a greater impact on the failure. Combined with the basic framework, that is, the node structure, and the weight relationship, a fault tree is finally generated. The fault tree graphically displays the hierarchical structure of equipment failures, which is convenient for operation and maintenance personnel to analyze.

[0055] For example, suppose there is a motor in a subway system, and its sensor data and historical maintenance records are as follows: Equipment status: normal, overheating, fault; Fault cause: insufficient coolant, fan failure, power failure; Fault type: motor overheating, short circuit failure. The basic framework is constructed to obtain: Root node: motor failure, child nodes: overheating, power failure, leaf nodes: for "overheating": insufficient coolant, fan failure; for "power failure": short circuit, unstable voltage. Suppose that through graph algorithm analysis, it is concluded that the weights of insufficient coolant and fan failure on overheating are 0.6 and 0.4 respectively, while the weight of power failure on the overall system failure is 0.5.

[0056] S150: Acquire real-time monitoring data for the target subway.

[0057] Specifically, real-time monitoring data refers to information about the operation status of the subway collected in real time through sensors, monitoring systems, etc. This data includes the status of the equipment, the operating environment, and other relevant operation information.

[0058] For example, suppose a subway line is the target subway, such as "Urban Subway Line 1". The following are some possible real-time monitoring data types and their sources: Temperature sensor: monitors the temperature of motors, transformers or inside the carriage to ensure that the equipment does not overheat; Vibration sensor: detects vibration during train operation and promptly detects abnormal wear or failure of mechanical parts; Current / voltage sensor: monitors the normal operation of the power system and promptly detects power anomalies. In addition, there is surveillance camera data: the video surveillance system transmits images of train platforms and carriages in real time to monitor passenger flow and safety; operation log: records the train's operating speed, stop time and fault alarm information. These log data can help operation and maintenance personnel analyze problems in operation; Communication system data: collects communication information between the train and the dispatching center through the subway's wireless communication system, including the train's positioning and operation status information.

[0059] S160. Perform fault prediction on the real-time monitoring data through a fault tree and generate a multi-dimensional fault report.

[0060] Specifically, by analyzing the real-time monitoring data and combining it with the structure of the fault tree, the server can perform fault prediction and identify potential fault risks. Fault prediction is based on logical reasoning based on historical data and existing real-time data to determine whether changes in certain parameters will cause equipment failure. For example, if the temperature is too high and the current is abnormal, it may cause a motor failure. Multi-dimensional fault reports refer to the presentation of fault prediction results in the form of graphics or tables. This report contains information from multiple aspects, such as the type of fault, severity, time trend of fault occurrence, possible cause distribution, etc. With this information, operation and maintenance personnel can better understand the risk of failure and take corresponding preventive measures.

[0061] In one possible implementation, fault prediction is performed on real-time monitoring data through a fault tree to generate a multi-dimensional fault report, specifically including: inputting the real-time monitoring data into the fault tree, and using the Bayesian network to perform probability inference to obtain the potential fault probability; if it is determined that the potential fault probability is greater than or equal to a preset threshold, a multi-dimensional fault report is generated, and the multi-dimensional fault report includes a fault trend graph and a cause distribution graph.

[0062] Specifically, a Bayesian network is a probabilistic graphical model that expresses conditional dependencies between variables. In fault prediction, a Bayesian network can infer the probability of a potential fault based on the input data. The server uses the Bayesian network to calculate the probability of each fault based on the relationships defined in the fault tree and the status of the real-time monitoring data. For example, if the values ​​of temperature and current exceed the normal range, the Bayesian network will update the probability of the fault to reflect the impact of these new data. Through the above steps, the server will derive the potential probability of a certain fault. If this probability is higher than a preset threshold, it means that the risk of the fault is high and action is required. Conversely, when the probability is less than or equal to the threshold, it means that the risk of the fault is low. Once it is determined that the probability of a potential fault is higher than the preset threshold, the server will automatically generate a multi-dimensional fault report. The report will provide key information to help operation and maintenance personnel understand the risk of the fault and its possible causes. The report content can include a fault trend chart, which can show the change in the probability of faults over a period of time, and a cause distribution chart, which can show the proportion and distribution of possible causes of faults.

[0063] In one possible implementation, refer to Figure 2 , Figure 2 Another flow chart of an intelligent operation and maintenance method applied to subway faults provided in an embodiment of the present application includes steps S210 to S220, and the above steps are as follows: S210, determining the fault level according to the multi-dimensional fault report; S220, sending the multi-dimensional fault report to the operation and maintenance personnel according to the fault level.

[0064] Specifically, the fault level is a classification of the severity and urgency of the fault, which can be divided into multiple levels, for example: Level 1 (low): The fault risk is low and will not affect normal operation. It can be handled within the planned maintenance time. Level 2 (medium): The fault risk is medium and needs to be handled as soon as possible, but it will not immediately affect safety or operation. Level 3 (high): The fault risk is high and must be handled immediately. If it is not resolved, it may cause an accident or serious operational interruption. The server will evaluate and determine the level of the fault based on the data in the multi-dimensional fault report, such as the probability of potential failure, the severity of the cause of the failure, etc. Once the fault level is determined, the server will send the multi-dimensional fault report to the corresponding operation and maintenance personnel. This process can be carried out through email, SMS, mobile application push notifications, etc. The sending of the report ensures that the operation and maintenance personnel are aware of the current fault status in a timely manner and take appropriate measures to deal with it.

[0065] For example, suppose a multi-dimensional fault report generated by the server shows: Fault type: Car air conditioning fault, Potential fault probability: 90%, Fault trend chart: The frequency of air conditioning faults has gradually increased in the past few days, Cause distribution chart: The main causes include "power supply problems" accounting for 70% and "component aging" accounting for 30%. The server compares the potential fault probability of 90% with the preset fault level standard. Since 90% is higher than the preset high-risk threshold, such as 80%, the server determines this fault level as level 3 (high). The server sends a detailed multi-dimensional fault report to the operation and maintenance personnel based on the determined fault level. The report contains an indication of the urgency of the air conditioning fault and recommends repairs as soon as possible, and even arranges for maintenance personnel to go to the site immediately for inspection and processing.

[0066] The present application also provides an intelligent operation and maintenance device for subway failures, referring to Figure 3 , Figure 3 A module diagram of an intelligent operation and maintenance device for subway faults provided in an embodiment of the present application. The intelligent operation and maintenance device includes an acquisition module 31 and a processing module 32, wherein the acquisition module 31 acquires multi-source data for a target subway, the multi-source data including sensor data, maintenance records, and equipment logs; the processing module 32 constructs a fault knowledge graph based on the multi-source data; the processing module 32 uses a deep learning model to fuse the multi-source data and the fault knowledge graph to obtain feature data and a fusion model; the processing module 32 generates a fault tree based on the feature data and the fusion model; the acquisition module 31 acquires real-time monitoring data for a target subway; the processing module 32 predicts faults on the real-time monitoring data through a fault tree and generates a multi-dimensional fault report.

[0067] In a possible implementation, the acquisition module 31 acquires multi-source data for the target subway, specifically including: the acquisition module 31 receives the original operation data sent by the sensor network to multiple operating devices of the target subway, the original operation data including temperature data, vibration data, voltage data and current data; the processing module 32 obtains the original maintenance record from the maintenance management system corresponding to the target subway through the ETL tool, the original maintenance record including the description, cause, maintenance time and treatment measures of each equipment failure; the processing module 32 collects the original log information of the target subway from the subway system, the original log information including operation logs and alarm information; the processing module 32 pre-processes the original operation data, original maintenance records and original log information to obtain multi-source data, and the pre-processing includes denoising, filtering and normalization.

[0068] In a possible implementation, the processing module 32 constructs a fault knowledge graph based on multi-source data, specifically including: the processing module 32 performs named entity recognition and relationship extraction on maintenance records and equipment logs to obtain a first entity, a second entity, and an association relationship between the first entity and the second entity, where the first entity and the second entity are any two devices among multiple devices included in the target subway; the processing module 32 constructs a triple with the first entity, the second entity, and the association relationship, and gradually organizes the information to obtain an initial knowledge graph; the processing module 32 labels the sensor data to obtain a fault event, which is used to represent the association between key data and the cause of the fault; the processing module 32 adds the fault event to the initial knowledge graph to obtain a fault knowledge graph.

[0069] In one possible implementation, the processing module 32 uses a deep learning model to fuse multi-source data and the fault knowledge graph to obtain feature data and a fusion model, specifically including: the processing module 32 uses an LSTM model to extract features from sensor data to obtain a feature matrix; the processing module 32 uses a BERT model to generate feature vectors for maintenance records and equipment logs to obtain text vectors; the processing module 32 uses a GNN graph neural network to perform feature learning on the fault knowledge graph to obtain graph structure features; the processing module 32 uses a combined model of a graph convolutional network and a Transformer to comprehensively encode the feature matrix, text vectors, and graph structure features to obtain feature data and a fusion model.

[0070] In a possible implementation, the processing module 32 generates a fault tree based on the feature data and the fusion model, specifically including: the processing module 32 obtains the equipment failure, the cause of the failure and the specific failure type according to the feature data; the processing module 32 constructs the root node, child nodes and leaf nodes according to the equipment failure, the cause of the failure and the specific failure type to obtain the basic framework; the processing module 32 uses a graph algorithm to process the fusion model to obtain a weight relationship; the processing module 32 generates a fault tree based on the basic framework and the weight relationship.

[0071] In one possible implementation, the processing module 32 performs fault prediction on the real-time monitoring data through a fault tree and generates a multi-dimensional fault report, which specifically includes: the processing module 32 inputs the real-time monitoring data into the fault tree, and uses the Bayesian network to perform probability inference to obtain the potential fault probability; if the processing module 32 determines that the potential fault probability is greater than or equal to a preset threshold, it generates a multi-dimensional fault report, and the multi-dimensional fault report includes a fault trend graph and a cause distribution graph.

[0072] In a possible implementation, the processing module 32 determines the fault level according to the multi-dimensional fault report; the processing module 32 sends the multi-dimensional fault report to the operation and maintenance personnel according to the fault level.

[0073] It should be noted that: when the device provided in the above embodiment realizes its function, only the division of the above functional modules is used as an example. In actual application, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0074] The present application also provides an electronic device, referring to Figure 4 , Figure 4 The electronic device may include: at least one processor 41 , at least one network interface 44 , a user interface 43 , a memory 45 , and at least one communication bus 42 .

[0075] The communication bus 42 is used to realize the connection and communication between these components.

[0076] The user interface 43 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 43 may also include a standard wired interface and a wireless interface.

[0077] The network interface 44 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0078] Among them, the processor 41 may include one or more processing cores. The processor 41 uses various interfaces and lines to connect various parts in the entire server, and executes various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 45, and calling data stored in the memory 45. Optionally, the processor 41 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 41 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 41, and it can be implemented separately through a chip.

[0079] Among them, the memory 45 may include a random access memory (RAM) or a read-only memory (Read-Only Memory). Optionally, the memory 45 includes a non-transitory computer-readable storage medium. The memory 45 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 45 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned method embodiments, etc.; the data storage area may store data involved in the above-mentioned method embodiments, etc. The memory 45 may also be optionally at least one storage device located away from the aforementioned processor 41. As Figure 4 As shown, the memory 45 as a computer storage medium may include an operating system, a network communication module, a user interface module, and an application program for an intelligent operation and maintenance method for subway failures.

[0080] exist Figure 4In the electronic device shown, the user interface 43 is mainly used to provide an input interface for the user and obtain data input by the user; and the processor 41 can be used to call an application program stored in the memory 45 for an intelligent operation and maintenance method applied to subway failures. When executed by one or more processors, the electronic device executes one or more methods in the above-mentioned embodiments.

[0081] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the order of the actions described, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required for the present application.

[0082] The present application also provides a computer-readable storage medium, which stores instructions. When executed by one or more processors, the electronic device executes one or more of the methods described in the above embodiments.

[0083] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0084] In the several embodiments provided in this application, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some service interfaces, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0085] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0086] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0087] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a memory and includes several instructions for a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned memory includes: various media that can store program codes, such as USB flash drives, mobile hard drives, magnetic disks or optical disks.

[0088] The above is only an exemplary embodiment of the present disclosure, and the scope of the present disclosure cannot be limited thereto. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the disclosure of the specification and practice, those skilled in the art will easily think of other embodiments of the present disclosure. This application is intended to cover any variation, use or adaptive change of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the technical field not recorded in the present disclosure. The description and examples are regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. An intelligent operation and maintenance method applied to subway failures, characterized in that: The method comprises: Acquire multi-source data for a target subway, wherein the multi-source data includes sensor data, maintenance records, and equipment logs; Constructing a fault knowledge graph based on the multi-source data; Using a deep learning model to fuse the multi-source data and the fault knowledge graph to obtain feature data and a fusion model; Generate a fault tree based on the feature data and the fusion model; Acquire real-time monitoring data for the target subway; Fault prediction is performed on the real-time monitoring data through the fault tree to generate a multi-dimensional fault report.

2. The intelligent operation and maintenance method for subway failures according to claim 1 is characterized in that: The obtaining of multi-source data for the target subway specifically includes: Receiving original operation data sent by a sensor network to a plurality of operation devices of the target subway, wherein the original operation data includes temperature data, vibration data, voltage data, and current data; Obtaining original maintenance records from the maintenance management system corresponding to the target subway through an ETL tool, wherein the original maintenance records include descriptions, causes, maintenance time, and treatment measures for each equipment failure; Collecting original log information of the target subway from the subway system, wherein the original log information includes operation logs and alarm information; The original operation data, the original maintenance record and the original log information are preprocessed to obtain the multi-source data, and the preprocessing includes denoising, filtering and normalization processing.

3. The intelligent operation and maintenance method for subway failure according to claim 1 is characterized in that: The constructing of a fault knowledge graph based on the multi-source data specifically includes: Performing named entity recognition and relationship extraction on the maintenance record and the equipment log to obtain a first entity, a second entity, and an association relationship between the first entity and the second entity, wherein the first entity and the second entity are any two devices among a plurality of devices included in the target subway; Constructing a triple with the first entity, the second entity and the association relationship, and gradually organizing the information to obtain an initial knowledge graph; Annotating the sensor data to obtain a fault event, wherein the fault event is used to indicate the association between the key data and the cause of the fault; The fault event is added to the initial knowledge graph to obtain the fault knowledge graph.

4. The intelligent operation and maintenance method for subway failure according to claim 1 is characterized in that: The deep learning model is used to fuse the multi-source data and the fault knowledge graph to obtain feature data and a fusion model, specifically including: Using the LSTM model to extract features from the sensor data to obtain a feature matrix; Using the BERT model to generate feature vectors for the maintenance record and the equipment log to obtain text vectors; Using a GNN graph neural network to perform feature learning on the fault knowledge graph to obtain graph structure features; The feature matrix, the text vector and the graph structure features are comprehensively encoded through a combined model of a graph convolutional network and a Transformer to obtain the feature data and the fusion model.

5. The intelligent operation and maintenance method for subway failure according to claim 1 is characterized in that: The generating a fault tree based on the feature data and the fusion model specifically includes: According to the characteristic data, the equipment failure, the cause of the failure and the specific failure type are obtained; According to the device failure, the failure cause and the specific failure type, a root node, a child node and a leaf node are constructed to obtain a basic framework; Processing the fusion model using a graph algorithm to obtain a weight relationship; The fault tree is generated based on the basic framework and the weight relationship.

6. The intelligent operation and maintenance method for subway failure according to claim 1 is characterized in that: The method of performing fault prediction on the real-time monitoring data by using the fault tree to generate a multi-dimensional fault report specifically includes: Inputting the real-time monitoring data into the fault tree, and performing probability inference using a Bayesian network to obtain a potential fault probability; If it is determined that the potential failure probability is greater than or equal to a preset threshold, the multi-dimensional failure report is generated, and the multi-dimensional failure report includes a failure trend graph and a cause distribution graph.

7. The intelligent operation and maintenance method for subway failure according to claim 1 is characterized in that: The method further comprises: Determining a fault level according to the multi-dimensional fault report; According to the fault level, the multi-dimensional fault report is sent to the operation and maintenance personnel.

8. An intelligent operation and maintenance device applied to subway failures, characterized in that: The intelligent operation and maintenance device comprises an acquisition module (31) and a processing module (32), wherein: The acquisition module (31) is used to acquire multi-source data for a target subway, wherein the multi-source data includes sensor data, maintenance records, and equipment logs; The processing module (32) is used to construct a fault knowledge graph based on the multi-source data; The processing module (32) is further used to fuse the multi-source data and the fault knowledge graph using a deep learning model to obtain feature data and a fusion model; The processing module (32) is further used to generate a fault tree based on the characteristic data and the fusion model; The acquisition module (31) is also used to acquire real-time monitoring data for the target subway; The processing module (32) is also used to perform fault prediction on the real-time monitoring data through the fault tree and generate a multi-dimensional fault report.

9. An electronic device, characterized in that: The electronic device comprises a processor (41), a memory (45), a user interface (43) and a network interface (44), wherein the memory (45) is used to store instructions, the user interface (43) and the network interface (44) are both used to communicate with other devices, and the processor (41) is used to execute the instructions stored in the memory (45) so that the electronic device executes the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed, the method according to any one of claims 1 to 7 is performed.

Images