An equipment operation and maintenance intelligent agent based on multi-source data fusion and a large model

Abstract

The application relates to the technical field of industrial internet and artificial intelligence, and discloses an equipment operation and maintenance intelligent agent based on multi-source data fusion and a large model, which adopts a layered architecture and comprises an equipment access layer, a data storage layer, an intelligent algorithm layer, a business application layer and a front-end display layer. Real-time operation state data of heterogeneous equipment is collected by the equipment access layer and is subjected to standardized processing; the data storage layer cooperatively stores time sequence data and a structured knowledge base; the intelligent algorithm layer utilizes a hybrid retrieval strategy, combines a large language model (LLM) and retrieval enhancement generation (RAG) technology, performs semantic understanding on fault description input by a user, simultaneously fuses trend analysis results of equipment historical operation data, and intelligently generates fault feature diagnosis and maintenance schemes. The application supports natural language interaction and dynamic updating of a knowledge base, can significantly improve the response speed and intelligent level of equipment operation and maintenance, and realizes effective transmission of operation and maintenance experience.

Description

Technical Field

[0001] This invention relates to the fields of industrial internet and artificial intelligence technology, and in particular to an intelligent device operation and maintenance agent based on multi-source data fusion and large model. Background Technology

[0002] Current equipment maintenance primarily relies on manual labor or traditional information systems to manage knowledge documents such as equipment maintenance manuals and historical failure cases, and uses independent data acquisition systems to store real-time and historical operational data. When equipment malfunctions or requires maintenance, maintenance personnel typically need to perform the following steps: First, they manually review uploaded documents in the equipment management system using combined conditions or keywords to obtain reference knowledge; second, they query the data acquisition system to obtain equipment operational data for the target time period; finally, they compare the data based on their personal experience and develop a maintenance plan.

[0003] However, the existing solutions mentioned above have significant drawbacks: First, knowledge retrieval efficiency is low. Due to the lack of an effective knowledge base search engine, searching through a large number of technical documents is slow and prone to omissions or errors, resulting in long response times. Second, data analysis has a high barrier to entry. Most equipment maintenance personnel lack professional data analysis capabilities, making it difficult to effectively correlate the trends of raw equipment operating data with specific faults, leading to a high rate of misjudgment. Third, there is a lack of closed-loop management and knowledge transfer. Because historical data is not effectively mined and relies on personal experience, expert experience is difficult to transfer. Technical documents are scattered and difficult to retrieve, often only allowing for emergency fault handling, and preventing preventative maintenance, which can easily lead to delivery delays or a chain reaction of failures. Summary of the Invention

[0004] This invention provides an intelligent device operation and maintenance agent based on multi-source data fusion and large model. By integrating RAG technology, large language model and IoT multi-source data, it can achieve accurate diagnosis of equipment faults, intelligent generation of maintenance plans and dynamic closed-loop update of knowledge base, significantly reducing the operation and maintenance threshold and improving fault response speed and decision accuracy.

[0005] This invention provides an intelligent device operation and maintenance agent based on multi-source data fusion and a large model, employing a layered architecture, specifically including: The device access layer is used to access heterogeneous devices and collect real-time operating status data. It performs protocol parsing and cleaning and transformation on the real-time operating status data to generate standardized operating status data and transmits it to the data storage layer. The data storage layer is used to receive the standardized operating status data and store it in time sequence as historical operating status data of the equipment. At the same time, it also pre-sets and stores equipment knowledge base data, which includes equipment technical documents, equipment maintenance knowledge documents and historical fault cases. The intelligent algorithm layer receives fault query commands, which include natural language descriptions of fault codes or fault characteristics. Based on the fault query commands, the intelligent algorithm layer extracts standardized operating status data for the current time period and compares it with historical operating status data of the equipment in the data storage layer to generate data anomaly fluctuation results. Simultaneously, it uses a large language model to transform the fault query commands into retrieval vectors, retrieves and matches relevant knowledge fragments in the equipment knowledge base data, and generates fault characteristic diagnosis and maintenance plans based on the data anomaly fluctuation results and the relevant knowledge fragments. The business application layer is used to receive the fault feature diagnosis and maintenance plan, perform microservice scheduling and interface encapsulation on the fault feature diagnosis and maintenance plan, generate encapsulated business display data, and push it to the front-end display layer. The front-end presentation layer is used to send the fault query command to the business application layer, so that the business application layer can schedule the intelligent algorithm layer to receive and visualize the encapsulated business display data, and provide a human-computer interaction interface for users to edit the fault feature diagnosis and maintenance plan, and feed the edited maintenance plan back to the data storage layer to dynamically update the equipment knowledge base data.

[0006] Furthermore, the device access layer includes a device access management module, a protocol parsing component, a data preprocessing component, a status monitoring and alarm module, and a data transmission channel module; The device access management module is used to be compatible with mainstream industrial communication protocols to access the heterogeneous devices, and to perform batch and single-point configuration of the accessed heterogeneous devices. The protocol parsing component is communicatively connected to the device access management module and is used to parse the real-time operating status data collected in different communication protocol formats in order to achieve standardized data collection; wherein, the real-time operating status data includes temperature, pressure and vibration frequency parameters of heterogeneous devices; The data preprocessing component is connected to the protocol parsing component and is used to clean and transform the parsed real-time running status data to ensure the accuracy and consistency of the data, thereby generating the standardized running status data and classifying and transmitting it to the data storage layer. The status monitoring and alarm module is used to establish a device status monitoring and abnormal alarm mechanism to monitor the operating status and access stability of the heterogeneous devices. The data transmission channel module is built on a high-performance network programming framework to ensure the stability of data transmission when a large number of heterogeneous devices access the network concurrently.

[0007] Furthermore, the data storage layer includes a time-series data storage module, a relational structured storage module, a caching acceleration and middleware module, a data governance module, and a data security and disaster recovery management module; The time-series data storage module is built on a time-series database and is used to receive standardized operating status data from the device access layer and persistently store the standardized operating status data as historical operating status data of the device to support the intelligent algorithm layer in reading, writing and trend querying of data within a specific time period. The relational structured storage module is built on a relational database and is used to centrally store and maintain the device knowledge base data. It uses a transaction mechanism to ensure the data consistency of the device knowledge base data during concurrent queries and dynamic updates. The cache acceleration and middleware module is used to cache frequently accessed maintenance knowledge fragments, dictionary data and real-time device status snapshots in memory, so as to reduce the query pressure on the underlying database and improve the response speed. The data governance module is used to standardize the management of master data, metadata and data quality of the data entering the database, and supports real-time and offline data processing. The data security and disaster recovery management module is used to achieve fine-grained control over data access permissions and to establish a scheduled backup and off-site disaster recovery mechanism to ensure the security of the device's historical operating status data and the device's knowledge base data.

[0008] Furthermore, the intelligent algorithm layer includes an intent recognition and instruction parsing module, a time-series data trend analysis module, a knowledge base semantic retrieval module, and an intelligent reasoning and solution generation module; The intent recognition and command parsing module is used to receive the fault query command. When the command is a natural language description, it uses a large language model to recognize the user's intent and convert it into a standardized retrieval vector or retrieval command. The time-series data trend analysis module is communicatively connected to the data storage layer. It is used to respond to the fault query command, extract the standardized operating status data, and call the time-series analysis algorithm to compare the trend with the historical operating status data of the equipment in order to identify abnormal fluctuations of key parameters and generate the abnormal fluctuation results of the data. The knowledge base semantic retrieval module is built based on the retrieval enhancement generation RAG technology. It is used to receive the retrieval vector or retrieval instruction, perform a hybrid retrieval in the device knowledge base data, and recall the associated knowledge fragments that match the fault characteristics and the abnormal fluctuation results of the data. The intelligent reasoning and solution generation module is used to fuse the fault query command, the abnormal data fluctuation results and the related knowledge fragments in context, input them into the large language model for logical reasoning, and generate the structured fault feature diagnosis and maintenance solution.

[0009] Furthermore, the intelligent algorithm layer also includes an algorithm scheduling engine, which is used to uniformly schedule the intent recognition and instruction parsing module, the time series data trend analysis module, and the intelligent reasoning and solution generation module. It automatically matches the appropriate algorithm model according to the type of the fault query instruction and monitors the usage of computing resources to ensure the real-time generation of the fault feature diagnosis and maintenance solution.

[0010] Furthermore, the business application layer includes a microservice scheduling and governance module, a business logic processing module, an interface security encapsulation module, and an asynchronous communication and task scheduling module; The microservice scheduling module is used to uniformly manage business services within the system, realize service registration and discovery, routing and forwarding, and circuit breaking and rate limiting, so as to dynamically schedule the generation tasks of fault feature diagnosis and maintenance plan; The business logic processing module is used to perform business formatting processing on the fault feature diagnosis results fed back by the intelligent algorithm layer, and to associate the generated maintenance plan with the equipment technical documents and equipment maintenance knowledge documents in the equipment knowledge base data. The interface security encapsulation module is used to perform user authentication and interface security encapsulation based on a security token authentication mechanism before pushing the fault feature diagnosis and maintenance plan to the front-end display layer, so as to generate the encapsulated business display data. The asynchronous communication and task scheduling module introduces message middleware to achieve asynchronous communication and service decoupling between various microservice modules, and integrates a task scheduling system to execute periodic business logic and automated tasks within the system.

[0011] Furthermore, the business logic processing module includes a fault diagnosis service unit, an intelligent maintenance solution generation unit, and a back-end management service unit; The fault diagnosis service unit is used to receive fault codes or descriptions from the front end, call the interface of the intelligent algorithm layer to perform calculations, and perform business formatting processing on the returned fault feature diagnosis results. The intelligent maintenance solution generation unit is used to process the maintenance solution generation logic, receive the solution content generated by the algorithm layer, and associate the generated solution with the equipment technical document management, equipment maintenance knowledge document management and equipment maintenance knowledge data management modules to form a complete business document. The back-end management service unit is used to support the basic operation and maintenance functions of the system, including user management, log management, permission management and data maintenance, to ensure the integrity of system data and the traceability of operations.

[0012] Furthermore, the front-end presentation layer includes an intelligent question-and-answer dialogue interaction module, a data visualization and display module, a user feedback and knowledge loop module, and a front-end status management and communication scheduling module; The intelligent question-and-answer dialogue interaction module is used to provide an intelligent question-and-answer dialogue interaction interface, receive the fault phenomenon described by the user in natural language or input the fault code, and convert the user's input information into the fault query instruction and send it to the intelligent algorithm layer. The data visualization and display module is used to receive the encapsulated business display data from the business application layer, and to visualize the fault feature diagnosis and maintenance plan in the form of data charts or three-dimensional scenes through the visualization component; The user feedback and knowledge loop module provides a human-computer interaction interface for users to like, dislike, or edit the fault feature diagnosis and maintenance plan. The edited maintenance plan is saved and sent back as feedback data to dynamically update the equipment knowledge base data in the data storage layer. The front-end state management and communication scheduling module is used to realize visualization and business operation of multiple terminals and multiple scenarios through front-end routing control and global state management mechanism, and to optimize the front-end and back-end data interaction with the business application layer by using network request components.

[0013] The beneficial effects of this invention are as follows: This invention constructs an intelligent device operation and maintenance agent based on multi-source data fusion and a large model. Utilizing RAG retrieval enhancement generation technology and an LLM large language model, it breaks through the limitations of traditional models that rely on pre-trained knowledge, achieving efficient and accurate retrieval of equipment technical documents and maintenance knowledge, significantly improving retrieval efficiency and accuracy. Simultaneously, by employing a hybrid retrieval strategy and deeply integrating real-time operating status data and historical data from IoT devices, it can automatically identify data anomaly patterns and uncover potential risks, solving the problems of high data analysis thresholds and difficulty in associating fault characteristics in traditional operation and maintenance. Furthermore, by supporting natural language interaction and user feedback on maintenance plans, the knowledge base can be dynamically updated, achieving intelligent closed-loop management of the entire process from fault diagnosis to plan generation, effectively promoting the accumulation and transmission of expert experience. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the structure of the equipment operation and maintenance intelligent agent based on multi-source data fusion and large model of the present invention.

[0015] Figure 2This is a flowchart of the intelligent interaction process of the present invention.

[0016] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0018] like Figure 1 , 2 As shown, this invention provides an intelligent device operation and maintenance agent based on multi-source data fusion and a large model, adopting a layered architecture, specifically including: The device access layer is used to access heterogeneous devices and collect real-time operating status data. It performs protocol parsing and cleaning and transformation on the real-time operating status data to generate standardized operating status data and transmits it to the data storage layer. The data storage layer is used to receive the standardized operating status data and store it in time sequence as historical operating status data of the equipment. At the same time, it also pre-sets and stores equipment knowledge base data, which includes equipment technical documents, equipment maintenance knowledge documents and historical fault cases. The intelligent algorithm layer receives fault query commands, which include natural language descriptions of fault codes or fault characteristics. Based on the fault query commands, the intelligent algorithm layer extracts standardized operating status data for the current time period and compares it with historical operating status data of the equipment in the data storage layer to generate data anomaly fluctuation results. Simultaneously, it uses a large language model to transform the fault query commands into retrieval vectors, retrieves and matches relevant knowledge fragments in the equipment knowledge base data, and generates fault characteristic diagnosis and maintenance plans based on the data anomaly fluctuation results and the relevant knowledge fragments. The business application layer is used to receive the fault feature diagnosis and maintenance plan, perform microservice scheduling and interface encapsulation on the fault feature diagnosis and maintenance plan, generate encapsulated business display data, and push it to the front-end display layer. The front-end presentation layer is used to send the fault query command to the business application layer, so that the business application layer can schedule the intelligent algorithm layer to receive and visualize the encapsulated business display data, and provide a human-computer interaction interface for users to edit the fault feature diagnosis and maintenance plan, and feed the edited maintenance plan back to the data storage layer to dynamically update the equipment knowledge base data.

[0019] (1) Device access layer In one embodiment, the device access layer serves as the underlying foundation of the entire device operation and maintenance intelligent agent, supporting multi-protocol compatibility and heterogeneous device access to build a stable and efficient data transmission and management channel, thereby opening up the data channel between the underlying industrial physical equipment and the upper-layer platform. Specifically, the device access layer includes a device access management module, a protocol parsing component, a data preprocessing component, a status monitoring and alarm module, and a data transmission channel module; The device access management module is designed to be compatible with mainstream industrial communication protocols, enabling unified access for various heterogeneous industrial devices. This module supports both batch configuration and flexible single-point configuration for device access, building a basic device network. After device access, the module continuously and in real-time collects operational status data from the heterogeneous devices, including but not limited to key parameters such as temperature, pressure, and vibration frequency.

[0020] The protocol parsing component is used to receive the operating status data collected by the device access management module. It uses a built-in parsing mechanism to uniformly parse data of different industrial communication protocol formats in order to achieve standardized data collection.

[0021] The data preprocessing component is connected to the protocol parsing component and is used to perform standardization processes such as cleaning and transformation on the parsed operating status data. This component ensures the accuracy and consistency of the data, ultimately processing the original equipment operating status data and generating standardized operating status data.

[0022] The status monitoring and alarm module is used to establish a device status monitoring and abnormal alarm mechanism, monitor the data communication status between the underlying device and the access layer in real time, and provide alarm feedback when the device is abnormal or communication is blocked.

[0023] The data transmission channel module is built on a high-performance network programming framework to ensure system stability and transmission efficiency when a large number of heterogeneous devices access the system concurrently. It stably transmits the standardized operating status data generated by the data preprocessing component to the upper data storage layer for classified storage.

[0024] (2) Data storage layer In one embodiment, the data storage layer serves as the data hub connecting the entire device operation and maintenance intelligent system. Its main function is to facilitate the collaborative operation of multiple storage engines, receiving collected data from the lower layers and providing dual support of time-series and structured data for the "hybrid retrieval strategy" of the upper-layer intelligent algorithm layer. Specifically, the data storage layer includes a time-series data storage module, a relational structured storage module, a caching acceleration and middleware module, a data governance module, and a data security and disaster recovery management module. The time-series data storage module uses a time-series database as its core underlying engine, specifically adapted to the characteristics of high-frequency reported data from industrial equipment. It receives standardized operational status data from the underlying device access layer and optimizes the read / write and range query performance of this massive amount of time-series data. This standardized operational status data is persistently stored as historical operational status data for the equipment, providing underlying data support for the intelligent algorithm layer to subsequently extract data for specific time periods, compare data trends, and monitor abnormal fluctuations.

[0025] The relational structured storage module uses a relational database as its core engine to centrally manage and maintain the system's core business data and knowledge assets. It pre-installs and stores equipment knowledge base data, which specifically includes structured equipment technical documents, equipment maintenance knowledge documents, and historical fault cases. This module leverages the transaction characteristics of relational databases to strictly ensure the consistency of the core data in the equipment knowledge base during concurrent queries and when receiving updates from the front-end presentation layer.

[0026] Caching acceleration and middleware module: In order to improve the system's response speed when dealing with emergency fault diagnosis, a caching middleware is also introduced in the data storage layer. This middleware is used to cache frequently accessed hot maintenance knowledge fragments, frequently called dictionary data, or real-time device status snapshots in memory, thereby significantly reducing the I / O query pressure of the underlying physical database and ensuring that the intelligent agent can achieve question-and-answer response in seconds.

[0027] The data governance module is responsible for the full lifecycle management of data, covering the standardized management of master data, metadata, data assets, and data quality. It employs a dual-track parallel mechanism of real-time and offline processing for incoming data to ensure the standardization of incoming data.

[0028] The data security and disaster recovery management module leverages the Java technology ecosystem to achieve fine-grained control over data access permissions. Simultaneously, this module establishes a scheduled backup and off-site disaster recovery mechanism to ensure the absolute security and high availability of the device's historical operating status data and the device's knowledge base data, preventing the loss of industrial data due to unforeseen circumstances.

[0029] (3) Intelligent Algorithm Layer In one embodiment, the intelligent algorithm layer, located above the data storage layer and below the business application layer, serves as the brain of the entire equipment operation and maintenance intelligence. Its core function lies in deeply integrating Large Language Modeling (LLM) and Retrieval Enhancement Generation (RAG) technologies to execute a hybrid retrieval strategy based on multi-source data, thereby achieving an intelligent closed loop from fault description to precise maintenance plan generation. Specifically, the intelligent algorithm layer includes an intent recognition and command parsing module, a time-series data trend analysis module, a knowledge base semantic retrieval module, an intelligent reasoning and plan generation module, and an algorithm scheduling engine module. The intent recognition and command parsing module serves as the entry point for the algorithm layer, receiving fault query commands sent from the front-end presentation layer. When the fault query command contains a fault code, the meaning of the code is directly parsed; when the fault query command is a natural language description (e.g., the device is noisy and has an abnormal temperature), the built-in large language model semantic understanding engine is used to convert the unstructured natural language into a computer-executable retrieval vector or standardized retrieval command.

[0030] The time-series data trend analysis module (data side of hybrid retrieval) is used for quantitative analysis of the objective operating status of the equipment. After receiving the fault query command, this module performs the following operations: Step a: Based on the timestamp of the instruction, extract the standardized operating status data (such as temperature, pressure, vibration frequency, etc.) of the device within the current time period from the data storage layer.

[0031] Step b: Call the time series analysis algorithm to compare the current data with the historical operating status data of the device stored in the data storage layer.

[0032] Step c: Monitor key parameters for unusual fluctuations (e.g., whether the temperature shows an abnormal upward trend before and after the fault occurs) and output the results of abnormal fluctuation data. This step ensures that the diagnosis relies not only on the user's subjective description but also on the objective physical performance of the equipment.

[0033] The knowledge base semantic retrieval module (the knowledge side of hybrid retrieval) utilizes RAG (Retrieval-Augmented Generation) technology to address the problems of large model illusion and knowledge lag. Specifically, based on the aforementioned retrieval instructions, high-dimensional vector matching is performed on the equipment knowledge base data in the data storage layer. It supports a hybrid mode of keyword retrieval and semantic retrieval, accurately recalling relevant knowledge fragments highly correlated with the current fault characteristics and the aforementioned abnormal data fluctuation results from massive amounts of equipment technical documents, maintenance manuals, and historical fault cases (for example, retrieving historical cases showing that when the temperature rises abnormally and is accompanied by vibration, it is mostly due to bearing wear).

[0034] The intelligent reasoning and solution generation module is the decision center that outputs the final result. This module performs the following operations: Step a: The fault description input by the user, the abnormal fluctuation results of the data output by the time series data trend analysis module, and the related knowledge fragments recalled by the knowledge base semantic retrieval module are used as the Prompt context input to the large language model.

[0035] Step b: Based on the multi-source information mentioned above, the large language model performs logical reasoning to generate a structured and executable fault characteristic diagnosis and maintenance plan. This plan not only includes the cause of the fault but also provides specific maintenance procedure suggestions, and is ultimately fed back to the business application layer.

[0036] Algorithm scheduling engine module: In order to ensure that the above modules work together, the intelligent algorithm layer also has an algorithm scheduling engine, which is responsible for automatically matching and adapting algorithm models, managing model versions, and monitoring the usage of computing resources to ensure that fault diagnosis tasks are completed within a few seconds (e.g., returning a solution with a matching degree of more than 90% within 10 seconds).

[0037] (4) Business Application Layer In one embodiment, the business application layer, built upon Java technology and a microservice architecture, serves as the core hub for business logic processing and service scheduling. Its primary responsibility is to encapsulate the technical results output by the intelligent algorithm layer into business logic and to achieve efficient reuse and secure scheduling of various functional modules through a microservice cluster. Specifically, the business application layer includes a microservice scheduling and governance module, a business logic processing module, an interface security encapsulation module, and an asynchronous communication and task scheduling module. The microservice scheduling and governance module is built on the Spring Cloud microservice architecture and is responsible for the unified management of various business services within the system. Specifically, it includes service registration and discovery, intelligent routing, configuration management, and circuit breaker / rate limiting mechanisms. Through this module, the business application layer can dynamically schedule fault diagnosis services and maintenance plan generation services based on the load of frontend requests, ensuring system stability and high availability under high concurrency.

[0038] The business logic processing module comprises several independent business microservice units (fault diagnosis service unit, intelligent maintenance solution generation unit, and backend management service unit) to execute specific business rules. Specifically, the fault diagnosis service unit receives fault codes or descriptions from the front end, calls the intelligent algorithm layer's interface for calculation, and performs business formatting processing on the returned fault feature diagnosis results (such as adding diagnostic timestamps and associating device IDs). The intelligent maintenance solution generation unit handles the generation logic of maintenance solutions, receiving not only the solution content generated by the algorithm layer but also associating the generated solutions with the equipment technical document management, equipment maintenance knowledge document management, and equipment maintenance knowledge data management modules to form complete business documents. The backend management service unit supports the system's basic operation and maintenance functions, including user management, log management, permission management, and data maintenance, ensuring the integrity of system data and the traceability of operations. Simultaneously, when executing the above business logic processing and interacting with underlying data, this module optimizes database operation efficiency and concurrent access performance through the MyBatis / MyBatis-Plus persistence layer framework and database connection pool technology.

[0039] Interface Security Encapsulation Module: To ensure the security of data transmission, this module is built based on Spring Security and the OAuth2 protocol. Before pushing fault diagnosis and maintenance solutions to the front-end presentation layer, this module performs interface security encapsulation on business data, generating encapsulated business display data. In specific implementation, a JWT (JSON Web Token) token mechanism is used for user authentication and authorization to prevent unauthorized interface calls and data leakage.

[0040] The asynchronous communication and task scheduling module introduces message middleware (such as RabbitMQ or Kafka) to enable asynchronous communication and decoupling between microservices, improving system response speed. Simultaneously, it integrates a task scheduling system to execute periodic business logic (such as periodically updating the knowledge base index and generating device health reports on a regular basis), supporting the automated operation of the system.

[0041] (5) Front-end presentation layer In one embodiment, the front-end presentation layer is built on Vue2 and its surrounding ecosystem, aiming to construct a lightweight, high-performance front-end application. This layer serves as the unified entry point for intelligent interaction between the entire device operation and maintenance intelligent agent and the user, supporting adaptation to multiple terminals such as PCs and mobile devices. Specifically, the front-end presentation layer includes an intelligent question-and-answer dialogue interaction module, a data visualization and display module, a user feedback and knowledge loop module, and a front-end status management and communication scheduling module. The intelligent question-and-answer dialogue interaction module provides end users with an intuitive dialogue interface that supports natural language dialogue. Users only need to describe the device's fault characteristics (e.g., fault symptoms) in natural language or directly input the fault code. After capturing the user's input information, the module packages it into the fault query command and sends it to the intelligent algorithm layer, thereby driving the underlying semantic intent recognition and subsequent business logic processing.

[0042] The data visualization and display module receives the encapsulated business display data from the business application layer and presents it to the user through multi-scenario visualization components. Specifically, this module integrates the Vue2 component library and encapsulates custom business components. It also employs visualization plugins to display complex fault diagnosis and maintenance solutions in intuitive data charts and even 3D scenes, greatly reducing the barrier to entry for ordinary operations and maintenance personnel.

[0043] The user feedback and knowledge loop module provides a specific human-computer interaction interface for users to interact with and provide feedback on the fault feature diagnosis and maintenance plan returned by the intelligent agent. Users can give feedback on the results by liking or disliking them. More importantly, it supports experts or senior maintenance personnel to directly edit and save the generated maintenance plan content on the interface. The user-edited maintenance plan is sent back as feedback data and finally fed back to the data storage layer to dynamically update the equipment knowledge base data, thereby realizing continuous optimization and intelligent closed-loop management of equipment management.

[0044] Front-end state management and communication scheduling module: To ensure smooth interaction across multiple scenarios, this layer implements page routing control through Vue Router and global state management through Vuex at the underlying engineering level. Simultaneously, it leverages the Axios network request library to optimize the front-end and back-end data interaction process with the business application layer, and uses build tools such as Webpack and Babel to ensure the compatibility and adaptation of the front-end code across different browsers and terminal devices.

[0045] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0046] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A device operation and maintenance intelligent agent based on multi-source data fusion and a large model, characterized in that, A layered architecture is adopted, specifically including: The device access layer is used to access heterogeneous devices and collect real-time operating status data. It performs protocol parsing and cleaning and transformation on the real-time operating status data to generate standardized operating status data and transmits it to the data storage layer. The data storage layer is used to receive the standardized operating status data and store it in time sequence as historical operating status data of the equipment. At the same time, it also pre-sets and stores equipment knowledge base data, which includes equipment technical documents, equipment maintenance knowledge documents and historical fault cases. The intelligent algorithm layer receives fault query commands, which include natural language descriptions of fault codes or fault characteristics. Based on the fault query commands, the intelligent algorithm layer extracts standardized operating status data for the current time period and compares it with historical operating status data of the equipment in the data storage layer to generate data anomaly fluctuation results. Simultaneously, it uses a large language model to transform the fault query commands into retrieval vectors, retrieves and matches relevant knowledge fragments in the equipment knowledge base data, and generates fault characteristic diagnosis and maintenance plans based on the data anomaly fluctuation results and the relevant knowledge fragments. The business application layer is used to receive the fault feature diagnosis and maintenance plan, perform microservice scheduling and interface encapsulation on the fault feature diagnosis and maintenance plan, generate encapsulated business display data, and push it to the front-end display layer. The front-end presentation layer is used to send the fault query command to the business application layer, so that the business application layer can schedule the intelligent algorithm layer to receive and visualize the encapsulated business display data, and provide a human-computer interaction interface for users to edit the fault feature diagnosis and maintenance plan, and feed the edited maintenance plan back to the data storage layer to dynamically update the equipment knowledge base data.

2. The equipment operation and maintenance intelligent agent based on multi-source data fusion and large model according to claim 1, characterized in that, The device access layer includes a device access management module, a protocol parsing component, a data preprocessing component, a status monitoring and alarm module, and a data transmission channel module; The device access management module is used to be compatible with mainstream industrial communication protocols to access the heterogeneous devices, and to perform batch and single-point configuration of the accessed heterogeneous devices. The protocol parsing component is communicatively connected to the device access management module and is used to parse the real-time operating status data collected in different communication protocol formats in order to achieve standardized data collection; wherein, the real-time operating status data includes temperature, pressure and vibration frequency parameters of heterogeneous devices; The data preprocessing component is connected to the protocol parsing component and is used to clean and transform the parsed real-time running status data to ensure the accuracy and consistency of the data, thereby generating the standardized running status data and classifying and transmitting it to the data storage layer. The status monitoring and alarm module is used to establish a device status monitoring and abnormal alarm mechanism to monitor the operating status and access stability of the heterogeneous devices. The data transmission channel module is built on a high-performance network programming framework to ensure the stability of data transmission when a large number of heterogeneous devices access the network concurrently.

3. The equipment operation and maintenance intelligent agent based on multi-source data fusion and large model according to claim 1, characterized in that, The data storage layer includes a time-series data storage module, a relational structured storage module, a caching acceleration and middleware module, a data governance module, and a data security and disaster recovery management module. The time-series data storage module is built on a time-series database and is used to receive standardized operating status data from the device access layer and persistently store the standardized operating status data as historical operating status data of the device to support the intelligent algorithm layer in reading, writing and trend querying of data within a specific time period. The relational structured storage module is built on a relational database and is used to centrally store and maintain the device knowledge base data. It uses a transaction mechanism to ensure the data consistency of the device knowledge base data during concurrent queries and dynamic updates. The caching acceleration and middleware module is used to cache frequently accessed maintenance knowledge fragments, dictionary data and real-time device status snapshots in memory, so as to reduce the query pressure on the underlying database and improve the response speed. The data governance module is used to standardize the management of master data, metadata and data quality of the data entering the database, and supports real-time and offline data processing. The data security and disaster recovery management module is used to achieve fine-grained control over data access permissions and to establish a scheduled backup and off-site disaster recovery mechanism to ensure the security of the device's historical operating status data and the device's knowledge base data.

4. The equipment operation and maintenance intelligent agent based on multi-source data fusion and large model according to claim 1, characterized in that, The intelligent algorithm layer includes an intent recognition and instruction parsing module, a time-series data trend analysis module, a knowledge base semantic retrieval module, and an intelligent reasoning and solution generation module; The intent recognition and command parsing module is used to receive the fault query command. When the command is a natural language description, it uses a large language model to recognize the user's intent and convert it into a standardized retrieval vector or retrieval command. The time-series data trend analysis module is communicatively connected to the data storage layer. It is used to respond to the fault query command, extract the standardized operating status data, and call the time-series analysis algorithm to compare the trend with the historical operating status data of the equipment in order to identify abnormal fluctuations of key parameters and generate the abnormal fluctuation results of the data. The knowledge base semantic retrieval module is built based on the retrieval enhancement generation RAG technology. It is used to receive the retrieval vector or retrieval instruction, perform a hybrid retrieval in the device knowledge base data, and recall the associated knowledge fragments that match the fault characteristics and the abnormal fluctuation results of the data. The intelligent reasoning and solution generation module is used to fuse the fault query command, the abnormal data fluctuation results and the related knowledge fragments in context, input them into the large language model for logical reasoning, and generate the structured fault feature diagnosis and maintenance solution.

5. The equipment operation and maintenance intelligent agent based on multi-source data fusion and large model according to claim 4, characterized in that, The intelligent algorithm layer also includes an algorithm scheduling engine, which is used to uniformly schedule the intent recognition and instruction parsing module, the time series data trend analysis module, and the intelligent reasoning and solution generation module. It automatically matches the appropriate algorithm model according to the type of the fault query instruction and monitors the usage of computing resources to ensure the real-time generation of the fault feature diagnosis and maintenance solution.

6. The equipment operation and maintenance intelligent agent based on multi-source data fusion and large model according to claim 1, characterized in that, The business application layer includes a microservice scheduling and governance module, a business logic processing module, an interface security encapsulation module, and an asynchronous communication and task scheduling module. The microservice scheduling module is used to uniformly manage business services within the system, realize service registration and discovery, routing and forwarding, and circuit breaking and rate limiting, so as to dynamically schedule the generation tasks of fault feature diagnosis and maintenance plan; The business logic processing module is used to perform business formatting processing on the fault feature diagnosis results fed back by the intelligent algorithm layer, and to associate the generated maintenance plan with the equipment technical documents and equipment maintenance knowledge documents in the equipment knowledge base data. The interface security encapsulation module is used to perform user authentication and interface security encapsulation based on a security token authentication mechanism before pushing the fault feature diagnosis and maintenance plan to the front-end display layer, so as to generate the encapsulated business display data. The asynchronous communication and task scheduling module introduces message middleware to achieve asynchronous communication and service decoupling between various microservice modules, and integrates a task scheduling system to execute periodic business logic and automated tasks within the system.

7. The equipment operation and maintenance intelligent agent based on multi-source data fusion and large model according to claim 6, characterized in that, The business logic processing module includes a fault diagnosis service unit, an intelligent maintenance solution generation unit, and a back-end management service unit. The fault diagnosis service unit is used to receive fault codes or descriptions from the front end, call the interface of the intelligent algorithm layer to perform calculations, and perform business formatting processing on the returned fault feature diagnosis results. The intelligent maintenance solution generation unit is used to process the maintenance solution generation logic, receive the solution content generated by the algorithm layer, and associate the generated solution with the equipment technical document management, equipment maintenance knowledge document management and equipment maintenance knowledge data management modules to form a complete business document. The back-end management service unit is used to support the basic operation and maintenance functions of the system, including user management, log management, permission management and data maintenance, to ensure the integrity of system data and the traceability of operations.

8. The equipment operation and maintenance intelligent agent based on multi-source data fusion and large model according to claim 1, characterized in that, The front-end presentation layer includes an intelligent question-and-answer dialogue interaction module, a data visualization and display module, a user feedback and knowledge loop module, and a front-end status management and communication scheduling module. The intelligent question-and-answer dialogue interaction module is used to provide an intelligent question-and-answer dialogue interaction interface, receive the fault phenomenon described by the user in natural language or input the fault code, and convert the user's input information into the fault query instruction and send it to the intelligent algorithm layer. The data visualization and display module is used to receive the encapsulated business display data from the business application layer, and to visualize the fault feature diagnosis and maintenance plan in the form of data charts or three-dimensional scenes through the visualization component; The user feedback and knowledge loop module provides a human-computer interaction interface for users to like, dislike, or edit the fault feature diagnosis and maintenance plan. The edited maintenance plan is saved and sent back as feedback data to dynamically update the equipment knowledge base data in the data storage layer. The front-end state management and communication scheduling module is used to realize visualization and business operation of multiple terminals and multiple scenarios through front-end routing control and global state management mechanism, and to optimize the front-end and back-end data interaction with the business application layer by using network request components.

Images