Signal equipment fault diagnosis method, device, equipment and medium

Abstract

The invention relates to a signal equipment fault diagnosis method and device, equipment and a medium, and the method comprises the following steps: carrying out the systematic processing of project data, constructing a project-level knowledge graph, and carrying out the fine adjustment of a first large model; performing systematization processing on product requirements and design files, constructing a product-level knowledge graph, and performing fine adjustment on the second large model; when fault information of the signal equipment is obtained, sending the fault information to the first large model; the first large model carries out fault diagnosis, and if the first large model can output an accurate fault reason, the fault reason is fed back to a user; otherwise, downloading the log in the specified time period, and sending the fault information and the equipment log to the second large model; and the second large model generates a retrieval code through an intelligent agent according to the fault information and the equipment log in combination with product-level knowledge, analyzes related functional variables in the log, realizes fault diagnosis, outputs fault causes and generates a fault report. Compared with the prior art, the method has the advantages of being high in reasoning accuracy, improving the fault diagnosis speed and the like.

Description

Technical Field

[0001] The present invention relates to automation control technology in the field of rail transit, and in particular to a signal equipment fault diagnosis method, device, equipment and medium. Background Art

[0002] With the rapid development of rail transit systems, higher requirements are placed on the safety, reliability and maintenance efficiency of signal equipment. At present, the fault diagnosis of rail transit signal equipment mainly adopts the following methods: rule-based expert system, statistical data analysis method, and intelligent diagnosis method based on machine learning. However, these traditional methods have obvious limitations: rule-based expert system requires a lot of manual writing and maintenance of rule base, which is difficult to cover all fault scenarios, and the cost of updating and maintaining rules is high; statistical methods have high requirements on data quality and are difficult to handle complex multivariate correlation analysis; traditional machine learning-based methods require a lot of labeled data for training, and have limited generalization capabilities. These methods usually require feedback from the project site to the supplier's product development department step by step to clarify the cause of the fault. This process is not only time-consuming and labor-intensive, but also easily affected by subjective factors, which is not conducive to the safe and efficient operation of the rail transit system.

[0003] In recent years, large language models have made breakthrough progress in the field of natural language processing, demonstrating powerful knowledge understanding and reasoning capabilities. However, there are still many challenges in directly applying large language models to fault diagnosis in the field of rail transit: first, general large language models lack an in-depth understanding of rail transit expertise, and it is difficult to accurately understand and analyze professional terms and technical details; second, the reasoning process of the model lacks interpretability, and it is difficult to meet the reliability requirements of safety-critical systems for fault diagnosis; third, the ability of large models to process structured data is limited, and it is difficult to effectively analyze time series data and multidimensional parameters in equipment logs. In addition, the rail transit industry has strict requirements for data security and privacy protection. How to achieve efficient fault diagnosis while protecting core technical data is also an urgent problem to be solved. For example, CN118133095A discloses a rail transit equipment fault prediction method based on a large language model, which converts massive structured operation and maintenance data and maintenance records of urban rail transit into natural language data sets, and uses data sets in the direction of rail transit operation and maintenance to fine-tune the BERT model. However, this method does not take data security into consideration, and the effectiveness of the reasoning results needs to be improved.

[0004] At the same time, the rail transit industry is developing towards intelligence and automation, which puts forward new requirements for fault diagnosis systems: on the one hand, it is necessary to improve the accuracy and efficiency of diagnosis and shorten the fault response time; on the other hand, it is necessary to enhance the system's adaptive ability to handle new equipment and new types of faults. In addition, with the increase in the types and number of equipment, the failure mode has become more complex, and traditional manual diagnosis methods can no longer meet the needs. Therefore, it is particularly important to develop a technology that can automatically, accurately and efficiently perform fault diagnosis. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and solve the problem that traditional fault diagnosis methods require a lot of manpower to analyze fault logs or write complex fault diagnosis scripts, and to provide a signal equipment fault diagnosis method, device, equipment and medium, which improves the efficiency and accuracy of fault analysis and reduces the labor cost in operation and implementation through natural language processing technology and retrieval enhancement generation mechanism of large models; provides a more systematic and standardized fault diagnosis process, helps to accumulate historical case experience, and continuously optimizes the performance of diagnostic algorithms, thereby further improving the overall reliability and safety of the rail transit system.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] According to a first aspect of the present invention, a signal device fault diagnosis method is provided, the method comprising the following steps:

[0008] S1, build the first large model, systematically process the project data, build a project-level knowledge graph, define functional variables and equipment components as entity nodes, formally express the dependencies and temporal associations between entities, and fine-tune the pre-trained first large model based on the project-level knowledge graph using retrieval-enhanced generation to enable it to have project-level knowledge;

[0009] S2, builds the second largest model, systematically processes product requirements and design documents, builds a product-level knowledge graph, establishes multi-dimensional associations and state transition rules between variables, realizes the dynamic feature expression of the equipment operation process, and fine-tunes the pre-trained second largest model based on the product-level knowledge graph using retrieval enhancement generation to enable it to have product-level knowledge;

[0010] S3, when the maintenance monitoring system obtains the fault information of the signal equipment, it sends it to the first large model;

[0011] S4, the first large model uses project-level knowledge to perform fault diagnosis based on the acquired fault information. If the first large model outputs an accurate fault cause, the fault cause is fed back to the user, and the process ends based on the user's feedback;

[0012] S5, if the first model cannot determine the exact cause of the fault, the maintenance monitoring system is fed back to download the log of the corresponding faulty board in the specified period, and the fault information and device log are sent to the second model through the function interface;

[0013] S6, the second largest model, based on the fault information and equipment logs, combined with product-level knowledge, generates retrieval code through the intelligent agent, analyzes the relevant functional variables in the logs, realizes fault diagnosis, and outputs the fault cause;

[0014] S7, the second largest model feeds back the cause of the fault to the engineering design personnel. After the fault diagnosis result is manually confirmed, the second largest model generates a fault report.

[0015] As a preferred technical solution, the step S1 is specifically as follows:

[0016] Data preprocessing: Perform text preprocessing operations on project data, including word segmentation, stop word removal, and standardization. At the same time, extract device types, functional descriptions, interface definitions, performance indicators, and build a project knowledge graph to store the association relationship between devices;

[0017] Training data construction: Collect historical failure cases including failure phenomena, diagnostic processes and solutions, pair the historical failure cases according to "problem-solution" pairs, and use prompt word engineering technology to construct training samples;

[0018] Vectorization processing: Fine-tune the pre-trained first model based on the training samples, and use the pre-trained model to convert the text into a vector representation, establish a vector index, and store the processed vector into the project knowledge base.

[0019] As a preferred technical solution, the step S2 is specifically as follows:

[0020] Data preprocessing: Systematically process product requirements and design documents, extract function point definitions, variable names, and state definitions, and build a product knowledge graph to establish a dependency graph between function variables;

[0021] Training data construction: Classify product files by functional modules, construct training samples of variable reference relationships and execution order, and design specific prompt word templates for fault diagnosis scenarios;

[0022] Vectorization processing: Fine-tune the pre-trained second largest model based on the training samples, and use the pre-trained model for vector conversion, build a multi-dimensional index that supports multi-angle retrieval by function, module, and variable, and store the processed vectors in the product knowledge base.

[0023] As a preferred technical solution, in the product requirements and design documents, the function variable name defined at each function point is the same as the variable name in the log, and the reference relationship and execution order relationship between the function variables are defined.

[0024] As a preferred technical solution, the construction of the product knowledge graph is specifically as follows:

[0025] Entity construction: define the functional variables as entity nodes in the graph, set the corresponding attribute information according to the different types of variables; for enumeration type variables representing states, define their possible values ​​as independent state entities; for numerical type variables representing physical quantities, define their attribute information, including numerical range and precision; define related equipment components as equipment entities;

[0026] Relationship construction: construct computational dependencies between variables, state transition relationships between variables, subordination between variables and equipment, and causal relationships related to fault diagnosis;

[0027] Rule expression: Use formal language to describe various rules, including the judgment logic of state transition, the judgment conditions of variable validity, the conditions for fault triggering, and the calculation formula of variables;

[0028] Temporal relationships: Extract the associations between variables in different operating cycles and determine the temporal relationships, including the recording and use of historical states, constraints on state duration, analysis rules for variable change trends, and the temporal logic of event sequences.

[0029] As a preferred technical solution, the fault information is obtained through the fault meaning and troubleshooting method defined in the equipment maintenance manual.

[0030] As a preferred technical solution, the first large model interacts with the user by obtaining prompt words input by the user, and the interaction includes asking questions, following up questions and providing more fault phenomena.

[0031] As a preferred technical solution, the first large model comprehensively analyzes input information of multiple dimensions to determine the cause of the fault, wherein the input information of multiple dimensions includes: a fault description input by the user, including the time, location, and fault phenomenon of the fault, the fault phenomenon including a natural language description and a screenshot of the fault phenomenon; equipment log data, including equipment operating status, control instruction execution status, key performance indicators, and current operating mode; fault information from a maintenance monitoring system, including fault type, occurrence time, and level; and historical maintenance records.

[0032] As a preferred technical solution, the first large model generates fault diagnosis results based on input information in multiple dimensions, including a detailed analysis of possible fault causes and an assessment of the scope of impact, while ranking the probability of various possible fault causes and providing recommended treatment plans including emergency measures and long-term solutions.

[0033] As a preferred technical solution, in step S4, the first large model determines whether it can output the accurate cause of the fault by outputting a diagnostic confidence index. When the diagnostic confidence is lower than a preset threshold, the first large model cannot give the accurate cause of the fault; when the diagnostic confidence is higher than the preset threshold, the first large model outputs the accurate cause of the fault; wherein the threshold is dynamically adjusted according to the operating environment.

[0034] As a preferred technical solution, the second largest model performs fault diagnosis based on fault information and equipment logs combined with product knowledge. The diagnosis process includes the following steps:

[0035] The agent generates code by analyzing fault information and equipment logs to extract keywords and time ranges;

[0036] Match relevant functional modules and variables according to the product knowledge base, and use predefined code templates to generate search statements, while dynamically adjusting search parameters and scope as needed.

[0037] As a preferred technical solution, the second largest model analyzes the following functional variables in the log: device status flags, including operating mode, control status, and fault status; communication interface status, including communication quality, data integrity, and latency; control instruction sequence, including instruction type, execution result, and response time; performance indicator data, including speed, position, and acceleration; and error code information, including error type, occurrence time, and duration.

[0038] As a preferred technical solution, in step S6, multiple methods are used to analyze the functional variables in the log: tracking the changing trend of variables over time and identifying abnormal points through time series analysis, studying the mutual influence and causal relationship between multiple variables through correlation analysis, identifying data points and abnormal patterns that deviate from the normal range through anomaly detection, and comparing with known fault patterns through pattern matching to find similar cases.

[0039] As a preferred technical solution, in step S7, the process of generating a fault report is as follows:

[0040] Describe the fault phenomenon, extract key log records when the fault occurs, analyze the equipment status change sequence, and summarize the fault manifestation characteristics;

[0041] Determine the exact time and location of the failure, including the precise timestamp of when the failure occurred, the physical location of the failed device, as well as environmental conditions and operating scenarios;

[0042] Conduct relevant variable analysis, extract variable state changes before and after the failure, analyze variable change trends and abnormal points, and identify key influencing factors;

[0043] Conduct root cause analysis to infer the cause of the failure based on product knowledge, verify the cause-effect chain, and assess the scope of the failure impact;

[0044] Proposes recommendations for improvement, including providing short-term emergency solutions, suggesting long-term improvements, and developing preventive strategies and monitoring recommendations.

[0045] According to a second aspect of the present invention, a signal equipment fault diagnosis device is provided for implementing the method described, the device comprising a maintenance monitoring system, a first large model server, a second large model server, an access terminal and a network communication device, wherein:

[0046] The first large model server includes a first operation processing unit and a first graphics processing unit, the first operation processing unit is used to obtain fault information and equipment logs from the maintenance monitoring system and perform human-computer interaction functions, and the first graphics processing unit is used to perform fault diagnosis reasoning operations of the first large model;

[0047] The second large model server includes a second operation processing unit and a second graphics processing unit. The second operation processing unit is used to obtain fault descriptions and equipment logs from the first large model and perform human-computer interaction functions. The second graphics processing unit is used to perform fault diagnosis reasoning operations on the second large model.

[0048] According to a third aspect of the present invention, there is provided an electronic device, comprising a memory and a processor, wherein a computer program is stored in the memory, and the method described above is implemented when the processor executes the program.

[0049] According to a fourth aspect of the present invention, there is provided a computer-readable storage medium having a computer program stored thereon, wherein the program implements the method described when executed by a processor.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] 1. Innovative dual-model hierarchical architecture design: This invention proposes a dual-layer large model architecture at the project level and product level. This design has multiple innovations: First, the hierarchical design realizes the reasonable distribution of knowledge. The project-level model masters general knowledge and common fault diagnosis capabilities, while the product-level model focuses on deep technical details; second, this architecture naturally forms a data security barrier, effectively protecting the core technical information of the product; third, the collaborative working mechanism of the dual models provides dual protection for fault diagnosis, significantly improving the accuracy of diagnosis. The beneficial effects brought about by this innovative architecture include: improving the accuracy of fault diagnosis, strengthening data security, and optimizing the efficiency of computing resources.

[0052] 2. Improved search enhancement generation method: This invention makes specific optimizations based on the traditional search enhancement generation method: First, a special text preprocessing process is designed for the rail transit field to improve the recognition and understanding of professional terms; second, a vectorization solution that adapts to multi-source heterogeneous data is developed, which can simultaneously process text documents, structured logs and time series data; third, a hierarchical knowledge retrieval strategy is designed, which can intelligently adjust the search scope and depth according to the fault type. The beneficial effects brought about by these improvements include: improving search accuracy, shortening response time, and enhancing the adaptability of the system.

[0053] 3. Innovative combination of knowledge graph and big model: This invention creatively combines knowledge graph technology with big model: First, a special knowledge graph in the field of rail transit is constructed to accurately express the relationship between equipment and the fault propagation path; second, the knowledge graph is used as the reasoning support of the big model to provide a structured knowledge background. The beneficial effects brought by this combination include: enhancing the explainability of fault diagnosis, improving the accuracy of reasoning, and realizing the continuous accumulation of knowledge.

[0054] 4. Innovative variable analysis method: This invention proposes a set of innovative variable analysis methods: first, a multi-dimensional variable feature extraction algorithm is designed to fully capture the dynamic changes of equipment status; second, a variable association analysis method based on causal reasoning is developed to accurately identify the root cause of the fault; third, an adaptive anomaly detection mechanism is implemented, which can dynamically adjust the judgment threshold according to the operating environment. The beneficial effects brought by these innovations include: improving the accuracy of fault location, reducing false positives and false negatives, and accelerating the speed of fault diagnosis.

[0055] 5. Security and privacy protection innovation: This invention has significant innovations in data security and privacy protection: First, hierarchical control of data access is achieved through a dual-model architecture; second, a data desensitization and secure transmission mechanism is developed to protect sensitive information. The beneficial effects brought about by these innovations include: effectively protecting core technical data, meeting industry security requirements, and improving the credibility of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 is a flow chart of a signal device fault diagnosis method according to an embodiment of the present invention;

[0057] Figure 2 An application method of signal equipment fault diagnosis in an embodiment of the present invention;

[0058] Figure 3 Schematic diagram of a knowledge graph for fault diagnosis in an embodiment of the present invention;

[0059] Figure 4 Schematic diagram of the structure of the signal equipment fault diagnosis device in an embodiment of the present invention. DETAILED DESCRIPTION

[0060] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0061] Unless otherwise defined, the technical terms or scientific terms involved in this application should be understood by people with ordinary skills in the technical field to which this application belongs. The words "one", "a", "a", "the" and the like involved in this application do not indicate a quantitative limitation, and may represent the singular or plural. The terms "include", "comprise", "have" and any of their variations involved in this application are intended to cover non-exclusive inclusions; for example, a process, method, system, product or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units that are not listed, or may also include other steps or units inherent to these processes, methods, products or devices. The words "connect", "connected", "coupled" and the like involved in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The "multiple" involved in this application refers to two or more. "And / or" describes the association relationship of associated objects, indicating that there may be three relationships, for example, "A and / or B" can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects before and after are in an "or" relationship. The terms "first", "second", "third", etc. involved in this application are only used to distinguish similar objects and do not represent a specific ordering of the objects.

[0062] This embodiment provides a signal device fault diagnosis method, such as Figure 1As shown, the method comprises the following steps:

[0063] S1, build the first large model, systematically process the project data, build a project-level knowledge graph, define functional variables and equipment components as entity nodes, formally express the dependencies and temporal associations between entities, and based on the project-level knowledge graph, use retrieval-enhanced generation to fine-tune the pre-trained first large model to enable it to have project-level knowledge.

[0064] Step S1 is specifically as follows:

[0065] Data preprocessing: Comprehensive data preprocessing of project materials, including text cleaning of project functional requirements, equipment maintenance manuals, project operation procedures, historical fault reports and other documents, and improving text quality through word segmentation, stop word removal, standardization and other text preprocessing operations. On this basis, extract key information such as equipment type, function description, interface definition, performance indicators, and build a project knowledge graph to store the relationship between devices;

[0066] Training data construction: Collect historical failure cases including failure phenomena, diagnostic processes and solutions, pair the historical failure cases according to "problem-solution" pairs, and use prompt word engineering technology to construct training samples;

[0067] Vectorization processing: Based on the training samples, the pre-trained first large model is fine-tuned using a small sample learning method, and pre-trained models such as BERT are used to convert text into vector representations, establish efficient vector indexes, and store the processed vectors in the project knowledge base to provide a basis for subsequent retrieval and diagnosis.

[0068] S2, builds the second largest model, systematically processes product requirements and design documents, builds a product-level knowledge graph, establishes multi-dimensional associations and state transition rules between variables, realizes the dynamic feature expression of the equipment operation process, and uses retrieval-enhanced generation based on the product-level knowledge graph to fine-tune the pre-trained second largest model so that it has product-level knowledge.

[0069] Step S2 is specifically as follows:

[0070] Data preprocessing: Systematically process product requirements and design documents, extract key information such as the definition of each function point, variable name, state definition, etc., build a product knowledge graph to establish a dependency graph between functional variables, and accurately express the relationship between various functional modules;

[0071] Training data construction: Classify product files by functional modules, construct training samples of variable reference relationships and execution order, and design specific prompt word templates for fault diagnosis scenarios;

[0072] Vectorization processing: Fine-tune the pre-trained second largest model based on the training samples, and use the pre-trained model for vector conversion to build a multi-dimensional index that supports multi-angle retrieval by function, module, and variable, so that subsequent accurate queries can be performed from multiple angles such as function, module, and variable, and the processed vectors are stored in the product knowledge base.

[0073] Among them, the function variable name defined in each function point in the product requirements and design files is the same as the variable name in the log, and the reference relationship and execution order relationship between the function variables are defined.

[0074] The specific steps to build a product knowledge graph are:

[0075] Entity construction: Define functional variables as entity nodes in the graph, and set corresponding attribute information according to different types of variables; for enumeration type variables representing states, define their possible values ​​as independent state entities; for numerical type variables representing physical quantities, define their attribute information, which includes numerical range and precision; define related equipment components as equipment entities to form a complete entity hierarchy.

[0076] Relationship construction: construct computational dependencies between variables, state transition relationships between variables, subordination between variables and devices, and causal relationships related to fault diagnosis. Among them, computational dependencies between variables are used to express how certain variables are calculated from other variables; state transition relationships between variables are used to describe how a certain state variable changes according to changes in other variables; subordination between variables and devices is used to indicate which device's output a certain variable belongs to; and causal relationships related to fault diagnosis are used to describe how certain states lead to specific faults.

[0077] Rule expression: Formal language is used to describe various rules, including the judgment logic of state transition, the judgment conditions of variable validity, the conditions for fault triggering, and the calculation formulas of variables; these rules are systematically organized and stored to facilitate reasoning and analysis of large models.

[0078] Time series relationship: Extract the association between variables in different operation cycles and determine the time series relationship, including the recording and use of historical states, constraints on state duration, analysis rules for variable change trends, and time series logic of event sequences. The establishment of these time series relationships enables the system to understand and analyze the dynamic change characteristics of the equipment during operation.

[0079] S3, when the maintenance monitoring system obtains the fault information of the signal equipment, it sends it to the first large model.

[0080] Fault information is obtained through the fault meaning and troubleshooting methods defined in the equipment maintenance manual.

[0081] S4, the first large model uses project-level knowledge to perform fault diagnosis based on the acquired fault information. If the first large model outputs an accurate fault cause, the fault cause is fed back to the user, and the process ends based on the user's feedback.

[0082] The first model interacts with the user by obtaining prompt words input by the user, including but not limited to asking questions, following up questions, and providing more fault phenomena.

[0083] The first large model comprehensively analyzes input information from multiple dimensions to determine the cause of the fault, where the input information from multiple dimensions includes: a fault description input by the user, including the time, location, and fault phenomenon of the fault, the fault phenomenon including a natural language description and a screenshot of the fault phenomenon; real-time log data from the equipment, including equipment operating status, control instruction execution status, key performance indicators, and current operating mode; fault information from the maintenance monitoring system, including fault type, occurrence time, and level; and historical maintenance records, from which to find experience in handling similar faults.

[0084] The first model generates fault diagnosis results based on input information from multiple dimensions, including a detailed analysis of possible fault causes and an assessment of the scope of impact. It also ranks the probability of various possible fault causes and provides recommended treatment plans including emergency measures and long-term solutions.

[0085] To ensure the reliability of the diagnosis, the system will give a diagnostic confidence index. In some special cases, such as encountering a new fault mode that is beyond the coverage of historical cases, or multiple faults occurring simultaneously resulting in feature confusion, or lack of key status information due to incomplete log data, and when the diagnostic confidence is lower than a preset threshold (for example, lower than 0.8), the first large model may not be able to give an accurate diagnostic result. If the cause of the fault can be determined, the user will be informed of the diagnostic result, and whether to end the current diagnostic process will be determined based on the user's feedback. If the cause of the fault cannot be determined, the maintenance monitoring system will be instructed to download the log data of the faulty board within a specific time period. In one embodiment, the threshold can also be dynamically adjusted according to the operating environment instead of using a fixed value.

[0086] S5, if the first model cannot determine the exact cause of the fault, the maintenance monitoring system is fed back to download the log of the corresponding faulty board in the specified period, and the fault information and device log are sent to the second model through the function interface.

[0087] S6, the second largest model, generates retrieval code through the intelligent agent based on the fault information and equipment logs, combined with product-level knowledge, analyzes the relevant functional variables in the logs, realizes fault diagnosis, and outputs the cause of the fault.

[0088] The second model performs fault diagnosis based on fault information and device logs combined with product knowledge. The diagnosis process includes the following steps:

[0089] The agent generates code by analyzing fault information and equipment logs to extract keywords and time ranges;

[0090] Match relevant functional modules and variables according to the product knowledge base, and use predefined code templates to generate search statements, while dynamically adjusting search parameters and scope as needed.

[0091] The second largest model analyzes the following functional variables in the log: device status flags, including operating mode, control status, and fault status; communication interface status, including communication quality, data integrity, and latency; control instruction sequence, including instruction type, execution result, and response time; performance indicator data, including speed, position, and acceleration; and error code information, including error type, occurrence time, and duration.

[0092] A variety of methods are used to analyze the functional variables in the logs: time series analysis is used to track the changing trends of variables over time and identify anomalies, correlation analysis is used to study the mutual influence and causal relationship between multiple variables, anomaly detection is used to identify data points and abnormal patterns that deviate from the normal range, and pattern matching is used to compare with known failure modes to find similar cases.

[0093] S7, the second largest model feeds back the cause of the fault to the engineering design personnel. After the fault diagnosis result is manually confirmed, the second largest model generates a fault report.

[0094] The process of generating a fault report is as follows:

[0095] Describe the fault phenomenon, extract key log records when the fault occurs, analyze the equipment status change sequence, and summarize the fault manifestation characteristics;

[0096] Determine the exact time and location of the failure, including the precise timestamp of when the failure occurred, the physical location of the failed device, as well as environmental conditions and operating scenarios;

[0097] Conduct relevant variable analysis, extract variable state changes before and after the failure, analyze variable change trends and abnormal points, and identify key influencing factors;

[0098] Conduct root cause analysis to infer the cause of the failure based on product knowledge, verify the cause-effect chain, and assess the scope of the failure impact;

[0099] Proposes recommendations for improvement, including providing short-term emergency solutions, suggesting long-term improvements, and developing preventive strategies and monitoring recommendations.

[0100] Through the above process, the present invention realizes automatic diagnosis of signal equipment faults, improves the efficiency and accuracy of fault location, reduces the need for manual intervention, and reduces operation and implementation costs.

[0101] In summary, the present invention has significantly improved the intelligent level of fault diagnosis of rail transit signal equipment through the organic combination of the above innovations, and has higher diagnostic accuracy, faster response speed, better interpretability and stronger safety, providing a strong guarantee for the safe operation of the rail transit system. At the same time, these innovative technical solutions of the present invention also provide a reference example for intelligent fault diagnosis in other industrial fields.

[0102] Figure 2 The figure is a schematic diagram of a possible application mode of the present invention. The first large model is deployed at the project level and connected to the maintenance monitoring system. When the maintenance monitoring system receives an alarm from the on-site signal equipment, it notifies the first large model and makes a preliminary diagnosis. The user can interact with the first large model to determine whether its diagnosis result is correct. If the first large model cannot make a diagnosis, or the user thinks it is not accurate enough, the fault description and log can be sent to the second large model deployed on the product side. The second large model has product-level knowledge and can generate retrieval code through the intelligent agent to conduct in-depth analysis of key functional variables in the log to find the root cause of the fault. The second large model submits the diagnosis results to the engineering design personnel for review and confirmation. In this process, the engineering design personnel can correct or supplement the diagnosis results according to their professional knowledge and experience, and guide the second large model to perform correct fault diagnosis through interaction. The confirmed fault cause is summarized into a detailed fault report by the second large model, and the report content includes a description of the fault phenomenon, the time and location of the fault, the state and change trend of the fault-related variables, and possible solution suggestions. The report is submitted to the user by the engineering design personnel.

[0103] Figure 3 It is a schematic diagram of the knowledge graph in the present invention. This figure takes the fault diagnosis of emergency braking caused by wheel slip as an example. When the system needs to analyze the cause of emergency braking, the state variables related to emergency braking, such as positioning state and wheel movement state, can be quickly located through the knowledge graph. By analyzing the state transition rules of these variables and the causal relationship between them, the system can accurately infer the specific reason for triggering the emergency braking. For example, the system can find that the train lost its positioning due to continuous wheel slippage, which in turn triggered the emergency brake. This reasoning method based on the knowledge graph can not only quickly locate the cause of the fault, but also provide a clear explanation link to help engineers understand the mechanism of fault occurrence.

[0104] The above is an introduction to a method embodiment. The following is a further explanation of the solution of the present invention through an apparatus embodiment.

[0105] like Figure 4 As shown, a signal equipment fault diagnosis device includes a maintenance monitoring system, a first large model server, a second large model server, an access terminal and a network communication device, wherein:

[0106] The first large model server includes a first operation processing unit and a first graphics processing unit, the first operation processing unit is used to obtain fault information and equipment logs from the maintenance monitoring system and perform human-computer interaction functions, and the first graphics processing unit is used to perform fault diagnosis reasoning operations of the first large model;

[0107] The second large model server includes a second operation processing unit and a second graphics processing unit. The second operation processing unit is used to obtain fault descriptions and equipment logs from the first large model and perform human-computer interaction functions. The second graphics processing unit is used to perform fault diagnosis reasoning operations on the second large model.

[0108] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the described module can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0109] The electronic device of the present invention includes a central processing unit (CPU), which can perform various appropriate actions and processes according to computer program instructions stored in a read-only memory (ROM) or loaded from a storage unit into a random access memory (RAM). In the RAM, various programs and data required for device operation can also be stored. The CPU, ROM and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0110] Multiple components in the device are connected to the I / O interface, including: input units, such as keyboards, mice, etc.; output units, such as various types of displays, speakers, etc.; storage units, such as disks, optical disks, etc.; and communication units, such as network cards, modems, wireless communication transceivers, etc. The communication unit allows the device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunication networks.

[0111] The processing unit performs the various methods and processes described above, such as methods S1 to S7. For example, in some embodiments, methods S1 to S7 may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed on the device via a ROM and / or a communication unit. When the computer program is loaded into the RAM and executed by the CPU, one or more steps of methods S1 to S7 described above may be performed. Alternatively, in other embodiments, the CPU may be configured to execute methods S1 to S7 in any other appropriate manner (e.g., by means of firmware).

[0112] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0113] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0114] In the context of the present invention, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0115] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A signal equipment fault diagnosis method, characterized in that: The method comprises the following steps: S1, build the first large model, systematically process the project data, build a project-level knowledge graph, define functional variables and equipment components as entity nodes, formally express the dependencies and temporal associations between entities, and fine-tune the pre-trained first large model based on the project-level knowledge graph using retrieval-enhanced generation to enable it to have project-level knowledge; S2, builds the second largest model, systematically processes product requirements and design documents, builds a product-level knowledge graph, establishes multi-dimensional associations and state transition rules between variables, realizes the dynamic feature expression of the equipment operation process, and fine-tunes the pre-trained second largest model based on the product-level knowledge graph using retrieval enhancement generation to enable it to have product-level knowledge; S3, when the maintenance monitoring system obtains the fault information of the signal equipment, it sends it to the first large model; S4, the first large model uses project-level knowledge to perform fault diagnosis based on the acquired fault information. If the first large model outputs an accurate fault cause, the fault cause is fed back to the user, and the process ends based on the user's feedback; S5, if the first model cannot determine the exact cause of the fault, the maintenance monitoring system is fed back to download the log of the corresponding faulty board in the specified period, and the fault information and device log are sent to the second model through the function interface; S6, the second largest model, based on the fault information and equipment logs, combined with product-level knowledge, generates retrieval code through the intelligent agent, analyzes the relevant functional variables in the logs, realizes fault diagnosis, and outputs the fault cause; S7, the second largest model feeds back the cause of the fault to the engineering design personnel. After the fault diagnosis result is manually confirmed, the second largest model generates a fault report.

2. A signal equipment fault diagnosis method according to claim 1, characterized in that: The step S1 is specifically as follows: Data preprocessing: Perform text preprocessing operations on project data, including word segmentation, stop word removal, and standardization. At the same time, extract device types, functional descriptions, interface definitions, performance indicators, and build a project knowledge graph to store the association relationship between devices; Training data construction: Collect historical failure cases including failure phenomena, diagnostic processes and solutions, pair the historical failure cases according to "problem-solution" pairs, and use prompt word engineering technology to construct training samples; Vectorization processing: Fine-tune the pre-trained first model based on the training samples, and use the pre-trained model to convert the text into a vector representation, establish a vector index, and store the processed vector into the project knowledge base.

3. A signal equipment fault diagnosis method according to claim 1, characterized in that: The step S2 is specifically as follows: Data preprocessing: Systematically process product requirements and design documents, extract function point definitions, variable names, and state definitions, and build a product knowledge graph to establish a dependency graph between function variables; Training data construction: Classify product files by functional modules, construct training samples of variable reference relationships and execution order, and design specific prompt word templates for fault diagnosis scenarios; Vectorization processing: Fine-tune the pre-trained second largest model based on the training samples, and use the pre-trained model for vector conversion, build a multi-dimensional index that supports multi-angle retrieval by function, module, and variable, and store the processed vectors in the product knowledge base.

4. A signal equipment fault diagnosis method according to claim 1, characterized in that: In the product requirements and design documents, the function variable name defined in each function point is the same as the variable name in the log, and the reference relationship and execution order relationship between the function variables are defined.

5. A signal equipment fault diagnosis method according to claim 3, characterized in that: The construction of the product knowledge graph is specifically as follows: Entity construction: define the functional variables as entity nodes in the graph, set the corresponding attribute information according to the different types of variables; for enumeration type variables representing states, define their possible values ​​as independent state entities; for numerical type variables representing physical quantities, define their attribute information, including numerical range and precision; define related equipment components as equipment entities; Relationship construction: construct computational dependencies between variables, state transition relationships between variables, subordination between variables and equipment, and causal relationships related to fault diagnosis; Rule expression: Use formal language to describe various rules, including the judgment logic of state transition, the judgment conditions of variable validity, the conditions for fault triggering, and the calculation formula of variables; Temporal relationships: Extract the associations between variables in different operating cycles and determine the temporal relationships, including the recording and use of historical states, constraints on state duration, analysis rules for variable change trends, and the temporal logic of event sequences.

6. A signal equipment fault diagnosis method according to claim 1, characterized in that: The fault information is obtained through the fault meaning and troubleshooting method defined in the equipment maintenance manual.

7. A signal equipment fault diagnosis method according to claim 1, characterized in that: The first large model interacts with the user by obtaining prompt words input by the user, and the interaction includes asking questions, further questions and providing more fault phenomena.

8. A signal equipment fault diagnosis method according to claim 1, characterized in that: The first large model comprehensively analyzes input information of multiple dimensions to determine the cause of the fault, wherein the input information of multiple dimensions includes: a fault description input by a user, including the time, location, and fault phenomenon of the fault, wherein the fault phenomenon includes a natural language description and a screenshot of the fault phenomenon; log data of the equipment, including equipment operating status, control instruction execution status, key performance indicators, and current operating mode; fault information from a maintenance monitoring system, including fault type, occurrence time, and level; and historical maintenance records.

9. A signal equipment fault diagnosis method according to claim 8, characterized in that: The first model generates fault diagnosis results based on input information of multiple dimensions, including a detailed analysis of possible fault causes and an assessment of the scope of impact. It also ranks the probability of various possible fault causes and provides recommended treatment plans including emergency measures and long-term solutions.

10. A signal equipment fault diagnosis method according to claim 1, characterized in that: In step S4, the first large model determines whether it can output the accurate cause of the fault by outputting a diagnostic confidence index. When the diagnostic confidence is lower than a preset threshold, the first large model cannot give the accurate cause of the fault; when the diagnostic confidence is higher than the preset threshold, the first large model outputs the accurate cause of the fault; wherein the threshold is dynamically adjusted according to the operating environment.

11. A signal equipment fault diagnosis method according to claim 1, characterized in that: The second model performs fault diagnosis based on fault information and device logs combined with product knowledge. The diagnosis process includes the following steps: The agent generates code by analyzing fault information and equipment logs to extract keywords and time ranges; Match relevant functional modules and variables according to the product knowledge base, and use predefined code templates to generate search statements, while dynamically adjusting search parameters and scope as needed.

12. A signal equipment fault diagnosis method according to claim 1, characterized in that: The second largest model analyzes the following functional variables in the log: device status flags, including operating mode, control status, and fault status; communication interface status, including communication quality, data integrity, and latency; control instruction sequence, including instruction type, execution result, and response time; performance indicator data, including speed, position, and acceleration; and error code information, including error type, occurrence time, and duration.

13. A signal equipment fault diagnosis method according to claim 1, characterized in that: In step S6, a variety of methods are used to analyze the functional variables in the log: tracking the changing trend of variables over time and identifying abnormal points through time series analysis, studying the mutual influence and causal relationship between multiple variables through correlation analysis, identifying data points and abnormal patterns that deviate from the normal range through anomaly detection, and comparing with known fault patterns through pattern matching to find similar cases.

14. A signal equipment fault diagnosis method according to claim 1, characterized in that: In step S7, the process of generating a fault report is as follows: Describe the fault phenomenon, extract key log records when the fault occurs, analyze the equipment status change sequence, and summarize the fault manifestation characteristics; Determine the exact time and location of the failure, including the precise timestamp of when the failure occurred, the physical location of the failed device, as well as environmental conditions and operating scenarios; Conduct relevant variable analysis, extract variable state changes before and after the failure, analyze variable change trends and abnormal points, and identify key influencing factors; Conduct root cause analysis to infer the cause of the failure based on product knowledge, verify the cause-effect chain, and assess the scope of the failure impact; Proposes recommendations for improvement, including providing short-term emergency solutions, suggesting long-term improvements, and developing preventive strategies and monitoring recommendations.

15. A signal equipment fault diagnosis device, characterized in that: For implementing the method according to any one of claims 1 to 14, the device comprises a maintenance monitoring system, a first large model server, a second large model server, an access terminal and a network communication device, wherein: The first large model server includes a first operation processing unit and a first graphics processing unit, the first operation processing unit is used to obtain fault information and equipment logs from the maintenance monitoring system and perform human-computer interaction functions, and the first graphics processing unit is used to perform fault diagnosis reasoning operations of the first large model; The second large model server includes a second operation processing unit and a second graphics processing unit. The second operation processing unit is used to obtain fault descriptions and equipment logs from the first large model and perform human-computer interaction functions. The second graphics processing unit is used to perform fault diagnosis reasoning operations on the second large model.

16. An electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the program, the method according to any one of claims 1 to 14 is implemented.

17. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 14 is implemented.

Images