Intelligent operation and maintenance work order processing method and device based on multi-modal data fusion, equipment and medium

Abstract

The invention provides an intelligent operation and maintenance work order processing method, device and equipment based on multi-modal data fusion and a medium, and the method comprises the steps: collecting alarm information of each business program or business subsystem in an IT operation and maintenance scene, the alarm information comprising a text log, a performance index, configuration information and network topology data; the collected alarm information is preprocessed; extracting key information according to the pre-processed alarm information; based on the extracted key information, generating a root cause hypothesis chain by analyzing an application relationship and a topological dependency relationship of each component in the plurality of service subsystems; according to the extracted key information and the root cause hypothesis chain, combining historical processing records in a knowledge base to generate a work order containing a suggested solution, and pushing the work order to a corresponding processor; and receiving a solution fed back by the processor, and completing work order processing.

Description

Technical Field

[0001] The present invention belongs to the cross - technical field of artificial intelligence and IT operation and maintenance, and specifically relates to an intelligent operation and maintenance work order processing method, device, equipment and medium based on multi - modal data fusion. Background Art

[0002] Traditional operation and maintenance systems rely on manual experience to process alarm information, suffering from problems such as poor timeliness and low knowledge reuse rate. Although existing technologies (such as CN118039100A) have achieved work order process management, they lack intelligent analysis capabilities based on large models; although an alarm handling prediction model has been proposed, the following core problems have not been solved: 1. Insufficient heterogeneous data processing ability: unable to efficiently fuse multi - modal information such as logs, performance metrics, and time - series data; 2. Lack of a dynamic knowledge base: historical work order experience has not formed a knowledge graph that can be iteratively optimized; 3. Limited model generalization: static models are difficult to adapt to the dynamic changes of complex business scenarios. Summary of the Invention

[0003] To solve the above problems, the present invention provides an intelligent operation and maintenance work order processing method, device, equipment and medium based on multi - modal data fusion.

[0004] The technical solution of the present invention is as follows: On the one hand, the present invention provides an intelligent operation and maintenance work order processing method based on multi - modal data fusion, including: Collect alarm information of each business program or business subsystem in the IT operation and maintenance scenario, where the alarm information includes text logs, performance metrics, configuration information, and network topology data; Pre - process the collected alarm information; Extract key information according to the pre - processed alarm information; Based on the extracted key information, generate a root cause hypothesis chain by analyzing the application relationships and topological dependencies of each component in multiple business subsystems; According to the extracted key information and root cause hypothesis chain, combine with historical processing records in the knowledge base to generate a work order containing a recommended solution and push it to the corresponding handler; Receive the solution feedback by the handler to complete the work order processing.

[0005] Preferably, the step of pre - processing the collected alarm information includes: Adopt the technology based on Few - shot Prompting. By designing a small number of examples, guide the pre - trained language model based on the Transformer architecture to extract business module names, common error fields, and / or key information of errors from text logs, and generate a standardized log template to achieve the standardization processing of unstructured logs.

[0006] Preferably, the steps of generating a work order including a proposed solution based on the extracted key information and root cause hypothesis chain, in combination with the historical processing records in the knowledge base, and pushing it to the corresponding handler include: Generating a work order including a proposed solution based on the extracted key information and root cause hypothesis chain, in combination with the historical processing records in the knowledge base; Dynamically selecting a handler according to the work order type, the handler's response speed, and the solution efficiency, and pushing the work order to the corresponding handler.

[0007] Preferably, the steps of receiving the solution feedback by the handler and completing the work order processing include: Inputting the solution feedback by the handler into the knowledge base for accuracy prediction; If the predicted accuracy is lower than the preset accuracy, triggering a preset review person to review the work order, and completing the work order processing according to the solution feedback by the review person; If the predicted accuracy is higher than the preset accuracy, completing the work order processing; The dynamic knowledge base stores the solutions of multiple historical work orders, and the accuracy of the solutions of each historical work order exceeds the preset accuracy.

[0008] Preferably, the steps of receiving the solution feedback by the handler and completing the work order processing further include: Updating the knowledge base according to the review result to optimize subsequent work order processing.

[0009] On the other hand, the present invention also provides an intelligent operation and maintenance work order processing device based on multi-modal data fusion, including: A collection module for collecting alarm information of each business program or business subsystem in the IT operation and maintenance scenario, where the alarm information includes text logs, performance metrics, configuration information, and network topology data; A preprocessing module for preprocessing the collected alarm information; A key information extraction module for extracting key information according to the preprocessed alarm information; A root cause hypothesis chain generation module for generating a root cause hypothesis chain based on the extracted key information by analyzing the application relationship and topology dependency relationship of each component in multiple business subsystems; A work order generation and push module for generating a work order including a proposed solution based on the extracted key information and root cause hypothesis chain, in combination with the historical processing records in the knowledge base, and pushing it to the corresponding handler; A work order processing module for receiving the solution feedback by the handler and completing the work order processing.

[0010] Preferably, the preprocessing module includes: A preprocessing unit, which is used to adopt the technology based on Few-shot Prompting. By designing a small number of examples, it guides a pre-trained language model based on the Transformer architecture to extract business module names, common error fields, and / or key information of errors from text logs, and generate a standardized log template to achieve the standardized processing of unstructured logs.

[0011] Preferably, the steps of generating a work order containing a suggested solution based on the extracted key information and root cause hypothesis chain, in combination with historical processing records in the knowledge base, and pushing it to the corresponding handler include: A work order generation unit, which is used to generate a work order containing a suggested solution based on the extracted key information and root cause hypothesis chain, in combination with historical processing records in the knowledge base; A work order pushing unit, which is used to dynamically select a handler according to the work order type and the response speed and solution efficiency of the handler, and push the work order to the corresponding handler.

[0012] On the other hand, the present invention also provides a control device, including a processor, a memory, and a program or instruction stored on the memory and executable on the processor. When the program or instruction is executed by the processor, the steps of the intelligent operation and maintenance work order processing method based on multi-modal data fusion as described above are implemented.

[0013] On the other hand, the present invention also provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by the processor, the steps of the intelligent operation and maintenance work order processing method based on multi-modal data fusion as described above are implemented.

[0014] The beneficial effects of the present invention are: By jointly processing multi-modal data such as text logs, performance metrics, and time series data, the data processing ability is significantly improved; by constructing a dynamic knowledge base, historical work order processing records are transformed into a structured knowledge graph, supporting real-time updates and semantic matching; a large number of reusable solutions can be accumulated, significantly improving the knowledge reuse rate; by dynamically adjusting model parameters according to the review results of work orders, it is ensured that the model can quickly adapt to new business scenarios. Description of the Drawings

[0015] Figure 1 It is a flowchart of the method in the embodiment of the present application; Figure 2 It is a structural block diagram of the device in the embodiment of the present application. Detailed Embodiment

[0016] For the convenience of those skilled in the art, the following further describes and explains the present invention patent through the accompanying drawings. The description is relatively detailed and complete, but it should not be construed as a limitation on the scope of the present invention patent. Obvious deformations and replacement forms of the following examples are all within the protection scope of this patent.

[0017] Referring to Figure 1 , an embodiment of the present application provides an intelligent operation and maintenance work order processing method based on multi-modal data fusion, which includes: Collect alarm information of each business program or business subsystem in the IT operation and maintenance scenario, and the alarm information includes text logs, performance indicators, configuration information, and network topology data; Preprocess the collected alarm information; Extract key information according to the preprocessed alarm information; Based on the extracted key information, generate a root cause hypothesis chain by analyzing the application relationship and topological dependency relationship of each component in multiple business subsystems; According to the extracted key information and the root cause hypothesis chain, combine the historical processing records in the knowledge base to generate a work order containing a recommended solution and push it to the corresponding handler; Receive the solution feedback by the handler to complete the work order processing.

[0018] In the IT operation and maintenance scenario, a complex IT operation and maintenance system usually consists of multiple business programs and business subsystems, and each business program or subsystem may generate alarm information. Among them, a business program refers to each independent application program or service in the IT operation and maintenance system, such as a payment system, a user authentication system, an order processing system, etc.; a business subsystem refers to a relatively independent module or component set in the IT operation and maintenance system, such as a database service subsystem, a network service subsystem, a middleware service subsystem, etc.

[0019] The alarm information includes multi-source heterogeneous data such as text logs, performance indicators, configuration information, and network topology data. Among them, the text log is a log file generated during the system operation, generated based on log information such as Syslog and Nginx, and the text log contains detailed error descriptions, timestamps, service names, etc.; the performance indicators are obtained by real-time monitoring using performance monitoring tools (such as Prometheus, Zabbix, etc.), such as CPU usage, memory usage, disk I / O, network bandwidth, etc.; the configuration information is generated by recording configuration changes and deploying the system through a configuration management system (such as Ansible, Puppet, etc.), and the configuration information can be in JSON or ML format, for example: service ports, database connection information, middleware configuration, etc.; the network topology data includes, for example, network device status, link status, traffic information, etc., and is obtained by monitoring with a network monitoring tool.

[0020] The steps for preprocessing the collected alarm information include: Adopt the technology based on Few-shot Prompting. By designing a small number of examples, guide the pre-trained language model based on the Transformer architecture to extract the business module name, common error fields, and / or key information of the error from the text log, and generate a standardized log template to achieve the standardized processing of unstructured logs.

[0021] Few-shot Prompting is a few-shot learning method that guides the pre-trained language model to complete specific tasks by designing a small number of examples (prompts). This method is particularly suitable for log parsing tasks because it can quickly adapt to log data in different formats without a large amount of labeled data.

[0022] In order to guide the model to identify and extract the key information in the log, it is necessary to design a small number of examples (prompts). These examples are natural language descriptions or structured templates used to guide the model to complete specific tasks. The specific steps are as follows: Analyze the text log: Analyze the text log to identify common error fields (such as "error", "timeout", "exception", etc.) and business module names (such as "payment system", "database service", etc.). By analyzing the structure and content of the text log, determine which fields and information are key and need to be extracted.

[0023] Define the task objective: Clearly define the objective of log parsing, such as extracting the business module name, error level, error information, etc.; according to the objective definition, design prompts that can guide the pre-trained language model based on the Transformer architecture to complete these tasks.

[0024] Design the prompt template: According to the analysis results, design a small number of prompt templates. These templates are natural language descriptions used to guide the model to identify and extract key information. The prompt templates are, for example: "Extract the business module name and error information from the text log.", "Identify the error level and business module name from the text log.", "Extract the key information from the text log, including the business module name, error level, and error description." After completing the example design, input the designed prompts into the pre-trained language model based on the Transformer architecture. The pre-trained language model uses the Few-shot Prompting technology to quickly learn and identify the business module name, common error fields, or error information in the text log according to the key information in the prompts.

[0025] The pre-trained language model converts the identified key information into a structured log template, facilitating subsequent processing and analysis. For example, the structured log template is: Timestamp: 2024-03-29 14:00:00, Business module name: payment system, Error level: ERROR, Error message: Database connection timeout.

[0026] By converting unstructured log data into a structured template form, it is convenient for subsequent processing and analysis. Through standardization processing, text logs can be more efficiently used for subsequent operation and maintenance analysis and work order generation.

[0027] With a small number of prompts and examples, the pre-trained language model can quickly adapt to the task of parsing text logs, reducing the dependence on a large amount of labeled data.

[0028] After completing log parsing, remove noise data and outliers to ensure the accuracy and reliability of the data.

[0029] Finally, convert the extracted key information into a unified format for subsequent processing.

[0030] The steps of generating a work order containing a recommended solution based on the extracted key information and the root cause hypothesis chain, combined with the historical processing records in the knowledge base, and pushing it to the corresponding handler include: Generate a work order containing a recommended solution based on the extracted key information and the root cause hypothesis chain, combined with the historical processing records in the knowledge base; Dynamically select the handler according to the work order type and the handler's response speed and resolution efficiency, and push the work order to the corresponding handler.

[0031] Extract key information such as the business module name (such as "payment system"), fault level (such as P0 (highest priority), P1 (high priority), etc.), impact scope (such as "affecting more than 1000 users", "affecting payment interfaces", etc.), and error message (such as "ERROR: Database connection timeout", "WARNING: High CPU usage", etc.) from the alarm information.

[0032] After extracting the key information, identify the dependencies between components in multiple subsystems. For example, the payment system depends on the database service, and the user authentication module depends on the payment system.

[0033] Based on the dependencies between components, analyze how the fault propagates in multiple subsystems. For example, if the database service fails, it may cause delays in the payment system, which in turn affects user transactions.

[0034] Generate a root cause hypothesis chain based on the analysis results. For example, "Database connection timeout → Payment service delay → User transaction failure".

[0035] After generating the root cause hypothesis chain, generate a work order based on the extracted key information and the root cause hypothesis chain. The content of the work order may include, for example: Business module name: The name of the affected business module; Fault level: The severity of the fault (e.g., P0, P1); Affected scope: The scope affected by the fault (e.g., the number of users, service interfaces); Root cause hypothesis chain: The propagation path and root cause of the problem; Initial handling suggestions: Initial handling suggestions proposed based on the dynamic knowledge base formed by historical work orders.

[0036] After generating the work order, count the average response time of each handler for processing the work order (for example, the average response time of handler A is 15 minutes, and the average response time of handler B is 30 minutes), count the success rate of each handler in resolving the work order (for example, the success rate of handler A in resolving the work order is 90%, and the success rate of handler B in resolving the work order is 80%), and count the historical work order type data processed by each handler (for example, handler A has been dealing with payment service work orders for a long time, and handler B has been dealing with database work orders for a long time). Dynamically select the most suitable handler according to the work order type, response speed, and resolution efficiency. For example, for a P0-level payment service work order, preferentially select handler A who has been dealing with payment service work orders for a long time, has a fast response speed, and high resolution efficiency, and then push the generated work order to the selected handler A.

[0037] The steps to complete the work order processing by receiving the solution feedback from the handler include: Input the solution feedback by the handler into the knowledge base for accuracy prediction; If the predicted accuracy is lower than the preset accuracy, trigger the preset review personnel to review the work order, and complete the work order processing according to the solution feedback by the review personnel; If the predicted accuracy is higher than the preset accuracy, complete the work order processing; The dynamic knowledge base stores the solutions of multiple historical work orders, and the accuracy of the solutions of each historical work order exceeds the preset accuracy.

[0038] After receiving the work order, the handler conducts fault handling according to the work order content and the proposed solution, and gives feedback on the solution. For example, the solution feedback by handler A may be as follows: Handling time: 2024-03-29 14:30:00, Solution: The database service has been restarted and the database connection has been restored.

[0039] The knowledge base classifies historical work order entities (fault types, solutions) and achieves semantic matching through keyword matching similarity (e.g., the similarity between "database connection failure" and "JDBC timeout" reaches 0.92). The knowledge graph update adopts an incremental learning strategy, only locally training the newly added work order data (such as about 500 per day) to avoid the resource consumption of full-scale reconstruction.

[0040] After the handler submits a solution, the following steps are taken: Use the knowledge graph (such as Neo4j graph database) to compare and verify the effectiveness of the solution feedback by the handler. The knowledge graph stores the processing records and solutions of historical work orders, as well as the dependency relationships between various components.

[0041] Through comparison, verify whether the solution submitted by the handler is effective. For example, check whether the database service has returned to normal, whether the payment system has returned to normal, and whether the user transaction is successful.

[0042] After verifying that the effectiveness of the solution submitted by the handler is insufficient, use the SHAP (SHapley Additive exPlanations) tool to calculate the contribution degree (SHAP value) of each feature to the model prediction, and display the calculated SHAP values to the operation and maintenance personnel through a visualization tool. The reviewer can quickly locate key problems based on the SHAP value graph and feature importance graph, improving the efficiency and accuracy of fault handling. Features include but are not limited to: Performance metrics: such as CPU usage rate, memory usage rate, disk I / O latency, network bandwidth usage rate, etc.

[0043] Log information: such as error logs, warning logs, etc.

[0044] Configuration information: such as service ports, database connection information, etc.

[0045] Topology information: such as the connection relationships of network devices, service dependency relationships, etc.; for example, the SHAP value of the feature "disk I / O latency" is 0.73, indicating that this feature has a relatively high contribution to fault prediction, and the SHAP value of the feature "CPU usage rate" is 0.25, indicating that this feature also makes a certain contribution to fault prediction.

[0046] The steps to complete the work order processing after receiving the solution feedback from the handler also include: Update the knowledge base according to the review results to optimize subsequent work order processing.

[0047] If the verification result shows that the solution is effective, add the solution to the dynamic knowledge base so that it can be quickly referred to when encountering similar problems in the future.

[0048] Referring to Figure 2 , an embodiment of the present invention further provides an intelligent operation and maintenance work order processing device based on multi-modal data fusion, including: A collection module 201 for collecting alarm information of each business program or business subsystem in the IT operation and maintenance scenario, where the alarm information includes text logs, performance metrics, configuration information, and network topology data; A preprocessing module 202 for preprocessing the collected alarm information; A key information extraction module 203 for extracting key information according to the preprocessed alarm information; A root cause hypothesis chain generation module 204 for generating a root cause hypothesis chain based on the extracted key information by analyzing the application relationship and topology dependency relationship of each component in multiple business subsystems; A work order generation and push module 205 for generating a work order containing a recommended solution according to the extracted key information and root cause hypothesis chain, in combination with the historical processing records in the knowledge base, and pushing it to the corresponding handler; A work order processing module 206 for receiving the solution feedback by the handler and completing the work order processing.

[0049] Preferably, the preprocessing module includes: A preprocessing unit for adopting the technology based on Few-shot Prompting, by designing a small number of examples, guiding the pre-trained language model based on the Transformer architecture to extract the business module name, common error fields, and / or key information of the error from the text log, and generating a standardized log template to realize the standardization processing of unstructured logs.

[0050] Preferably, the step of generating a work order containing a recommended solution according to the extracted key information and root cause hypothesis chain, in combination with the historical processing records in the knowledge base, and pushing it to the corresponding handler includes: A work order generation unit for generating a work order containing a recommended solution according to the extracted key information and root cause hypothesis chain, in combination with the historical processing records in the knowledge base; A work order push unit for dynamically selecting a handler according to the work order type and the response speed and solution efficiency of the handler, and pushing the work order to the corresponding handler.

[0051] On the other hand, the present invention also provides a control device, including a processor, a memory, and a program or instruction stored on the memory and executable on the processor. When the program or instruction is executed by the processor, the steps of the intelligent operation and maintenance work order processing method based on multi-modal data fusion as described above are implemented.

[0052] On the other hand, the present invention also provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by the processor, the steps of the intelligent operation and maintenance work order processing method based on multi-modal data fusion as described above are implemented.

[0053] It should be noted that the embodiments in this specification are all described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other.

[0054] Although the preferred embodiments of the embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications once they know the basic creative concept. Therefore, the appended claims are intended to be construed to include the preferred embodiments and all changes and modifications falling within the scope of the embodiments of the present invention.

[0055] It should also be noted that in this article, the orientation or positional relationship indicated by terms such as "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or component referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be construed as a limitation of the present invention. In addition, relational terms such as "first" and "second" are used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations, nor can they be construed as indicating or implying relative importance. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements does not include those elements, but also includes other elements not expressly listed, or elements inherent to such process, method, article or terminal device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article or terminal device including the element.

[0056] The above has introduced the technical solution provided by the present invention in detail. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the present invention, and the content of this specification should not be construed as a limitation to the present invention. At the same time, for those of ordinary skill in the art, according to the present invention, there will be various forms of changes in the specific implementation manner and application scope. It is not necessary and impossible to enumerate all the implementation manners here, and the obvious changes or variations derived therefrom still fall within the protection scope of the present invention.

Claims

1. An intelligent operation and maintenance work order processing method based on multi-modal data fusion, characterized in that, including: Collecting the alarm information of each business program or business subsystem in the IT operation and maintenance scenario, where the alarm information includes text logs, performance metrics, configuration information, and network topology data; Preprocessing the collected alarm information; Extracting key information based on the preprocessed alarm information; Based on the extracted key information, generating a root cause hypothesis chain by analyzing the application relationships and topological dependencies of each component in multiple business subsystems; Generating a work order containing a recommended solution based on the extracted key information and the root cause hypothesis chain, combined with the historical processing records in the knowledge base, and pushing it to the corresponding handler; Receiving the solution feedback by the handler and completing the work order processing.

2. The intelligent operation and maintenance work order processing method based on multi-modal data fusion according to claim 1, wherein, The steps of preprocessing the collected alarm information include: Adopting the technology based on Few-shot Prompting, by designing a small number of examples, guiding the pre-trained language model based on the Transformer architecture to extract the business module name, common error fields, and / or key information of the error from the text logs, and generating a standardized log template to achieve the standardized processing of unstructured logs.

3. The intelligent operation and maintenance work order processing method based on multi-modal data fusion according to claim 1, wherein The steps of generating a work order containing a recommended solution based on the extracted key information and the root cause hypothesis chain, combined with the historical processing records in the knowledge base, and pushing it to the corresponding handler include: Generating a work order containing a recommended solution based on the extracted key information and the root cause hypothesis chain, combined with the historical processing records in the knowledge base; Dynamically selecting a handler according to the work order type, the response speed, and the resolution efficiency of the handler, and pushing the work order to the corresponding handler.

4. The intelligent operation and maintenance work order processing method based on multi-modal data fusion according to claim 1, wherein The steps of receiving the solution feedback by the handler and completing the work order processing include: Inputting the solution feedback by the handler into the knowledge base for accuracy prediction; If the predicted accuracy is lower than the preset accuracy, triggering a preset reviewer to review the work order, and completing the work order processing according to the solution feedback by the reviewer; If the predicted accuracy is higher than the preset accuracy, completing the work order processing; The dynamic knowledge base stores the solutions of multiple historical work orders, and the accuracy of the solutions of each historical work order exceeds the preset accuracy.

5. The intelligent operation and maintenance work order processing method based on multi-modal data fusion according to claim 4, wherein, The steps of receiving the solution feedback by the handler and completing the work order processing further include: Updating the knowledge base according to the review result to optimize the subsequent work order processing.

6. An intelligent operation and maintenance work order processing device based on multi-modal data fusion, characterized in that, including: A collection module for collecting the alarm information of each business program or business subsystem in the IT operation and maintenance scenario, where the alarm information includes text logs, performance metrics, configuration information, and network topology data; A preprocessing module for preprocessing the collected alarm information; A key information extraction module for extracting key information based on the preprocessed alarm information; A root cause hypothesis chain generation module for generating a root cause hypothesis chain based on the extracted key information by analyzing the application relationships and topological dependencies of each component in multiple business subsystems; A work order generation and push module for generating a work order containing a recommended solution based on the extracted key information and the root cause hypothesis chain, combined with the historical processing records in the knowledge base, and pushing it to the corresponding handler; The work order processing module is used to receive the solution feedback by the processor and complete the work order processing.

7. The intelligent operation and maintenance work order processing device based on multi-modal data fusion according to claim 6, wherein The preprocessing module includes: The preprocessing unit is used to adopt the technology based on Few-shot Prompting. By designing a small number of examples, it guides the pre-trained language model based on the Transformer architecture to extract the business module name, common error fields, and / or key information of the error from the text log, and generate a standardized log template to achieve the standardized processing of unstructured logs.

8. The intelligent operation and maintenance work order processing device based on multi-modal data fusion according to claim 6, wherein, The steps of generating a work order containing a recommended solution based on the extracted key information and root cause hypothesis chain, combined with the historical processing records in the knowledge base, and pushing it to the corresponding processor include: The work order generation unit is used to generate a work order containing a recommended solution based on the extracted key information and root cause hypothesis chain, combined with the historical processing records in the knowledge base; The work order pushing unit is used to dynamically select a processor according to the work order type, the response speed, and the solution efficiency of the processor, and push the work order to the corresponding processor.

9. A control device, characterized in that, It includes a processor, a memory, and a program or instruction stored on the memory and executable on the processor. When the program or instruction is executed by the processor, it realizes the steps of the intelligent operation and maintenance work order processing method based on multi-modal data fusion according to any one of claims 1-5.

10. A readable storage medium, characterized in that, The program or instruction is stored on the readable storage medium. When the program or instruction is executed by the processor, it realizes the steps of the intelligent operation and maintenance work order processing method based on multi-modal data fusion according to any one of claims 1 to 5.

Images