Expert knowledge graph-based time sequence large model and small model multi-agent collaborative diagnosis method

Abstract

The application belongs to the technical field of industrial equipment diagnosis, and relates to a time sequence large model and small model multi-agent collaborative diagnosis method based on an expert knowledge graph. The preliminary diagnosis result of the small model is sent to the knowledge graph for first round logic and mechanism verification. The physical rules of the graph are used to pre-filter the fast response result, so as to avoid that the preliminary conclusion with low reliability or contrary to the mechanism is directly adopted. According to the first round verification result and the confidence of the preliminary diagnosis result, the large model is selectively triggered for deep reasoning, so that the calling of the large model is focused on the complex situation that cannot be determined by the small model or that does not pass the graph verification, the fast response advantage of the small model is exerted, and the reasoning capability of the large model is utilized as needed. The reasoning diagnosis result of the large model is sent to the knowledge graph again for second round logic and mechanism verification, and the final diagnosis result is confirmed according to the verification result or a correction and marking operation is performed according to the graph rules.

Description

Technical Field

[0001] This invention belongs to the field of industrial equipment diagnostic technology and relates to a multi-agent collaborative diagnostic method based on a time-series large model and a small model using expert knowledge graphs. Background Technology

[0002] Thermal power equipment operates in a complex environment with tightly coupled mechanisms, making fault diagnosis crucial for ensuring power system safety. Existing diagnostic systems are mainly divided into two categories: The first category uses small, purely data-driven models (such as traditional neural networks and support vector machines). These models are typically deployed at the edge and have the advantages of fast response speed and low computational resource requirements. However, they heavily rely on the distribution of training data, have poor generalization ability, are difficult to handle unknown operating conditions or complex coupled faults, and are prone to misjudgment.

[0003] The second approach utilizes large language models for diagnosis. Large models possess powerful semantic understanding and knowledge reasoning capabilities, but purely data-driven large models suffer from the "illusion" problem, meaning they may generate conclusions that do not conform to physical laws or device mechanisms, leading to unreliable diagnostic results. Furthermore, in existing technologies, the collaboration between large and small models often involves simple "voting" or "chaining," lacking an effective constraint mechanism to ensure that the decisions of both always adhere to the physical laws of the domain. Therefore, how to complement the advantages of large and small models and ensure that collaborative decisions conform to physical laws is a key challenge that urgently needs to be addressed. Summary of the Invention

[0004] The purpose of this invention is to address the problems in the prior art by providing a multi-agent collaborative diagnosis method and system based on expert knowledge graphs for time-series large and small models. This aims to solve the problems of poor generalization ability of purely data-driven small models, the susceptibility of large models to illusions, and the potential for existing large-small model collaborative decision-making to violate physical laws, thereby improving the reliability, accuracy, and interpretability of fault diagnosis for thermal power equipment.

[0005] To achieve the above objectives, the present invention employs the following technical solution: A multi-agent collaborative diagnostic method for time-series large and small models based on expert knowledge graphs includes the following steps: Acquire structural knowledge, fault knowledge, and diagnostic rule knowledge of the target equipment, and construct an expert knowledge graph based on the structural knowledge, fault knowledge, and diagnostic rule knowledge of the target equipment; Deploy a small temporal model and a large language model respectively. Define the small temporal model as the first agent and the large language model as the second agent. The first agent is used for fast response and feature extraction, and the second agent is used for deep knowledge reasoning. A collaborative middleware is constructed based on a temporal small model and a large language model, which is used to connect the temporal small model and the large language model. The following collaborative diagnostics are performed based on the collaborative middleware: Preliminary diagnostic results are obtained based on a time-series small model, and then sent to an expert knowledge graph for the first round of logic and mechanism verification. Based on the confidence levels of the first round of verification results and preliminary diagnostic results, the large language model is selectively triggered to perform deep reasoning to obtain reasoning diagnostic results; The reasoning and diagnosis results are sent to the expert knowledge graph for a second round of logical and mechanistic verification. The final diagnosis result is confirmed based on the results of the second round of verification, or correction or labeling operations are performed according to the rules in the expert knowledge graph.

[0006] A further improvement of the present invention is that: The structural knowledge includes the physical connections and hierarchical relationships between equipment components; The fault knowledge includes fault modes, fault phenomena, and fault causes. The diagnostic rule knowledge includes physical laws, mechanistic constraints, and expert experience.

[0007] The time-series small model is deployed at the edge computing end, and the large language model is deployed at the cloud computing center.

[0008] The collaborative middleware is also used to execute a task layering mechanism, including: Based on the complexity of the input data or the confidence level of the preliminary diagnostic results, tasks are categorized into routine tasks, complex tasks, or unknown tasks. Regular tasks are assigned to the small temporal model for processing, complex tasks are assigned to the large language model for processing, and unknown tasks are assigned to both the small temporal model and the large language model for parallel processing.

[0009] The step of performing correction or labeling operations based on rules in the expert knowledge graph includes: When the reasoning diagnosis result fails the second round of logic and mechanism verification, the expert knowledge graph generates correction suggestions based on preset rules. The large language model performs deep reasoning again based on the correction suggestions to obtain the corrected reasoning diagnosis results; If the revised reasoning diagnosis result still fails to pass verification, then the result will be marked as pending manual confirmation.

[0010] The collaborative middleware maintains a global session context for recording the preliminary diagnostic results, the first round of verification results, the reason for triggering deep reasoning in the large language model, the reasoning diagnostic results, the second round of verification results, and information on correction or marking operations. A time-series large-scale and small-scale multi-agent collaborative diagnostic system based on expert knowledge graphs includes: The expert knowledge graph construction module is used to acquire structural knowledge, fault knowledge, and diagnostic rule knowledge of the target equipment, and to construct an expert knowledge graph based on the structural knowledge, fault knowledge, and diagnostic rule knowledge of the target equipment. The multi-agent deployment module is used to deploy a small temporal model and a large language model respectively. The small temporal model is defined as the first agent and the large language model is defined as the second agent. The first agent is used for fast response and feature extraction, and the second agent is used for deep knowledge reasoning. A collaborative middleware construction module is used to build collaborative middleware based on a temporal small model and a large language model. The collaborative middleware is used to connect the temporal small model and the large language model. The collaborative diagnostics module is used to perform the following collaborative diagnostics based on the collaborative middleware: Preliminary diagnostic results are obtained based on a time-series small model, and then sent to an expert knowledge graph for the first round of logic and mechanism verification. Based on the confidence levels of the first round of verification results and preliminary diagnostic results, the large language model is selectively triggered to perform deep reasoning to obtain reasoning diagnostic results; The reasoning and diagnosis results are sent to the expert knowledge graph for a second round of logical and mechanistic verification. The final diagnosis result is confirmed based on the results of the second round of verification, or correction or labeling operations are performed according to the rules in the expert knowledge graph.

[0011] A computer program product includes a computer program that, when executed by a processor, implements any one of the methods described.

[0012] A terminal device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of any of the methods described above.

[0013] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the methods described herein.

[0014] Compared with the prior art, the present invention has the following beneficial effects: This invention discloses a multi-agent collaborative diagnosis method based on an expert knowledge graph, comprising a temporal large model and a small model. It constructs an expert knowledge graph based on equipment structure knowledge, fault knowledge, and diagnostic rule knowledge, and deploys a temporal small model for rapid response and feature extraction, and a large language model for deep knowledge reasoning. Further, a collaborative middleware executes the following constrained collaborative diagnosis: the preliminary diagnostic results of the small model are sent to the knowledge graph for the first round of logical and mechanistic verification. This utilizes the physical rules of the knowledge graph to pre-screen the rapid response results, effectively preventing the direct adoption of unreliable or mechanistically inconsistent preliminary conclusions. Then, based on the confidence levels of the first round of verification results and the preliminary diagnostic results... This method selectively triggers deep reasoning in the large model, focusing its calls on complex situations where the small model cannot determine the outcome or the graph verification fails. This leverages the rapid response advantage of the small model while utilizing the reasoning capabilities of the large model as needed. Finally, the reasoning and diagnostic results of the large model are sent back to the knowledge graph for a second round of logic and mechanism verification. Based on the verification results, the final diagnostic result is confirmed, or correction and labeling operations are performed according to the graph rules. This method imposes mandatory physical constraints on the output of the large model, which is most prone to illusion. This approach addresses the shortcomings of purely data-driven collaboration that may violate the physical laws of the device, while also suppressing the illusion problem of the large model and improving the generalization ability of the small model. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a system diagram disclosed in an embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0018] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0019] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0020] In the description of the embodiments of the present invention, it should be noted that if terms such as "upper," "lower," "horizontal," or "inner" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of the invention is in use, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. Furthermore, terms such as "first" and "second" are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0021] Furthermore, the use of the term "horizontal" does not imply that the component must be absolutely horizontal, but rather that it can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal than "vertical," and does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0022] In the description of the embodiments of the present invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in the present invention according to the specific circumstances.

[0023] The present invention will now be described in further detail with reference to the accompanying drawings: Example 1 See Figure 1 This embodiment discloses a multi-agent collaborative diagnosis method for time-series large and small models based on expert knowledge graphs, including the following steps: Step 1: Construct a knowledge graph of experts in the thermal power field Acquire structural knowledge, fault knowledge, and diagnostic rule knowledge of thermal power equipment, and construct a knowledge graph containing entities, relationships, and attributes as the logical constraint hub. Among them, structural knowledge describes the physical connections and hierarchical relationships between equipment components; fault knowledge describes fault modes, phenomena, and causes; and diagnostic rule knowledge includes the physical laws, mechanistic constraints, and expert experience rules followed by equipment operation.

[0024] Step 2: Deploy the multi-agent diagnostic model Multiple lightweight time-series mini-models are deployed as the first intelligent agent at the edge computing end to receive real-time sensing data, perform rapid response and preliminary feature extraction, and output preliminary diagnostic conclusions and their confidence levels.

[0025] A large language model is deployed as a second intelligent agent in a cloud computing center to handle complex and nondeterministic faults that cannot be determined by the small model, perform deep knowledge reasoning, and output reasoning and diagnostic conclusions.

[0026] Step 3: Build the collaboration middleware Establish a collaborative middleware to connect the edge-end small model agents, large model agents, and the knowledge graph from step 1.

[0027] The collaborative middleware is responsible for task distribution, result collection, confidence assessment, and constraint verification by calling the knowledge graph.

[0028] Furthermore, in this embodiment, the collaboration middleware maintains a global session context to record the decision-making links and intermediate inference states between different agents.

[0029] Step 4: Collaborative Diagnosis of Large and Small Models Based on Task Hierarchy and Graph Constraints The collaborative middleware performs the collaboration of large and small models and the mandatory verification of the knowledge graph according to the following process: Step 4.1: Rapid diagnosis and confidence assessment of small models The collaborative middleware sends real-time sensor data into the edge time series mini-model, and the mini-model outputs preliminary diagnostic conclusions and their confidence levels. Step 4.2: The knowledge graph is used to perform preliminary verification of the conclusions of the small model. The collaborative middleware sends the preliminary diagnostic conclusions of the small model to the expert knowledge graph for logical and mechanistic verification: If the preliminary diagnosis passes the atlas verification and the confidence level is higher than the preset high confidence threshold, the preliminary diagnosis will be directly output as the final diagnosis result, and the process will end. In this embodiment, the high confidence threshold is set to 0.85.

[0030] If the preliminary diagnostic conclusion fails the graph verification or the confidence level is lower than the low confidence threshold, deep inference is triggered by the large cloud model. In this embodiment, the low confidence threshold is set to 0.6.

[0031] Step 4.3: Large Model Deep Inference and Graph Revalidation The collaborative middleware sends the raw data, feature vectors extracted by the small model, and the preliminary conclusions of the small model to the large model in the cloud. The large model then combines its internal knowledge to perform cross-condition, complex coupled fault reasoning and outputs reasoning conclusions. The collaborative middleware then sends the reasoning conclusion to the expert knowledge graph for final mandatory verification. If the reasoning conclusion passes the graph verification, it is confirmed as the final diagnosis result.

[0032] If the reasoning conclusion fails the verification, i.e. violates the physical laws or mechanism rules in the knowledge graph, a correction operation is performed: the knowledge graph provides correction suggestions based on the rule base, and the middleware guides the large model to re-reason based on the correction suggestions; if the repeated reasoning still fails, it is marked as pending manual confirmation and terminated.

[0033] Step 4.4: Parallel processing of special tasks For unknown tasks, the collaborative middleware can distribute the task to both the small model and the large model at the same time. After obtaining the conclusions, the graph verification steps 4.2 and 4.3 are executed in sequence, and the conclusion that passes the verification and has higher confidence is taken as the standard.

[0034] Through the above steps, the expert knowledge graph acts as a global constraint judge, performing mandatory logical and mechanistic verification on every decision output of both the small and large models, ensuring that all collaborative decisions strictly conform to physical laws. Example 2 This embodiment focuses on fault diagnosis of bearings in thermal power turbines, and specifically includes the following steps: Step 1: Construct a knowledge graph of experts in the thermal power field Acquire structural knowledge, fault knowledge, and diagnostic rule knowledge of thermal power equipment, and construct a knowledge graph containing entities, relationships, and attributes as the logical constraint hub. Among them, structural knowledge describes the physical connections and hierarchical relationships between equipment components; fault knowledge describes fault modes, phenomena, and causes; and diagnostic rule knowledge includes the physical laws, mechanistic constraints, and expert experience rules followed by equipment operation.

[0035] Step 2: Deploy the model Deploy a small LSTM-based time-series model at the edge, taking the vibration, temperature, and rotational speed time-series data of the most recent 30 seconds as input, and outputting the fault category and confidence level. Deploy a GPT-like large language model in the cloud.

[0036] Step 3: Collaborative Middleware Execution. This step specifically discloses three scenarios, including: Scenario 1: Small model with high confidence and passed graph validation Small model output: Bearing condition is normal, confidence level 0.92; The collaborative middleware sends the conclusion to the knowledge graph for verification. Spectrum rule query: Vibration amplitude <50μm, temperature rise <5℃, spectrum is mainly power frequency, which is determined to be in line with normal mechanism, and the verification is passed; Output: The bearing is in normal condition; no need to call the large model.

[0037] Scenario 2: The output of the small model is rejected by the graph, triggering the output of the large model. Small model output: Suspected rubbing fault, confidence level 0.65; The collaborative middleware sends the collision fault data to the graph for verification. The spectrum analysis revealed that at the current rotational speed of 3000 rpm, the power frequency component in the vibration spectrum did not decrease, and there were no high-order harmonics with rubbing characteristics. Therefore, the rubbing fault was determined to be contrary to the physical mechanism, and the verification failed. The collaborative middleware triggers a large cloud model, sending the original data, features extracted from the small model, and rejection conclusions along with the fault detection. The output of the large model after comprehensive reasoning is: based on the vibration spectrum characteristics and load change trend, the diagnosis is mass imbalance.

[0038] The collaborative middleware will send the "quality imbalance" back to the graph for verification. Spectrum rules: Power frequency vibration is dominant, and the amplitude is proportional to the square of the rotational speed. It is determined to conform to the unbalance mechanism and the verification is passed. Output: Quality imbalance, and also record: The small model was rejected by the graph because it lacked rubbing features.

[0039] Scenario 3: The conclusions of the large model are also rejected by the graph, and a correction is performed. Large model error output: thermal bending failure; The graph verification revealed that under the current operating conditions, the vibration phase did not drift slowly and there were no abnormal temperature gradient records, indicating that the thermal bending mechanism was violated. The graph provides correction suggestions: based on the existing characteristics, it is recommended to change it to mass imbalance or loose foundation; The collaborative middleware guides the large model to re-infer based on the correction suggestions, and the large model re-outputs: quality imbalance, verified by the graph, and outputs the conclusion; If the second inference still fails, mark it as awaiting manual confirmation.

[0040] Example 3 This embodiment also discloses a multi-agent collaborative diagnostic system for time-series large and small models based on expert knowledge graphs, including: The expert knowledge graph construction module is used to acquire structural knowledge, fault knowledge, and diagnostic rule knowledge of the target equipment, and to construct an expert knowledge graph based on the structural knowledge, fault knowledge, and diagnostic rule knowledge of the target equipment. The multi-agent deployment module is used to deploy a small temporal model and a large language model respectively. The small temporal model is defined as the first agent and the large language model is defined as the second agent. The first agent is used for fast response and feature extraction, and the second agent is used for deep knowledge reasoning. A collaborative middleware construction module is used to build collaborative middleware based on a temporal small model and a large language model. The collaborative middleware is used to connect the temporal small model and the large language model. The collaborative diagnostics module is used to perform the following collaborative diagnostics based on the collaborative middleware: Preliminary diagnostic results are obtained based on a time-series small model, and then sent to an expert knowledge graph for the first round of logic and mechanism verification. Based on the confidence levels of the first round of verification results and preliminary diagnostic results, the large language model is selectively triggered to perform deep reasoning to obtain reasoning diagnostic results; The reasoning and diagnosis results are sent to the expert knowledge graph for a second round of logical and mechanistic verification. The final diagnosis result is confirmed based on the results of the second round of verification, or correction or labeling operations are performed according to the rules in the expert knowledge graph.

[0041] A schematic diagram of a terminal device according to an embodiment of the present invention. The terminal device of this embodiment includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the various method embodiments described above. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the various device embodiments described above.

[0042] The computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention.

[0043] The terminal device can be a desktop computer, laptop computer, cloud server, or other device with strong computing power. The terminal device may include, but is not limited to, a processor and memory.

[0044] The optimal choice for the processor is a multi-core high-speed central processing unit (CPU).

[0045] The memory can be used to store the computer program and / or module. The processor implements various functions of the terminal device by running or executing the computer program and / or module stored in the memory and calling the data stored in the memory.

[0046] If the modules / units integrated into the terminal device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0047] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A multi-agent collaborative diagnostic method for time-series large and small models based on expert knowledge graphs, characterized in that, Includes the following steps: Acquire structural knowledge, fault knowledge, and diagnostic rule knowledge of the target equipment, and construct an expert knowledge graph based on the structural knowledge, fault knowledge, and diagnostic rule knowledge of the target equipment; Deploy a small temporal model and a large language model respectively. Define the small temporal model as the first agent and the large language model as the second agent. The first agent is used for fast response and feature extraction, and the second agent is used for deep knowledge reasoning. A collaborative middleware is constructed based on a temporal small model and a large language model, which is used to connect the temporal small model and the large language model. The following collaborative diagnostics are performed based on the collaborative middleware: Preliminary diagnostic results are obtained based on a time-series small model, and then sent to an expert knowledge graph for the first round of logic and mechanism verification. Based on the confidence levels of the first round of verification results and preliminary diagnostic results, the large language model is selectively triggered to perform deep reasoning to obtain reasoning diagnostic results; The reasoning and diagnosis results are sent to the expert knowledge graph for a second round of logical and mechanistic verification. The final diagnosis result is confirmed based on the results of the second round of verification, or correction or labeling operations are performed according to the rules in the expert knowledge graph.

2. The method for multi-agent collaborative diagnosis of time-series large and small models based on expert knowledge graphs according to claim 1, characterized in that, The structural knowledge includes the physical connections and hierarchical relationships between equipment components; The fault knowledge includes fault modes, fault phenomena, and fault causes. The diagnostic rule knowledge includes physical laws, mechanistic constraints, and expert experience.

3. The method for collaborative diagnosis of multiple agents in a time-series large and small model based on expert knowledge graphs according to claim 1, characterized in that, The time-series small model is deployed at the edge computing end, and the large language model is deployed at the cloud computing center.

4. The method for multi-agent collaborative diagnosis of time-series large and small models based on expert knowledge graphs according to claim 1, characterized in that, The collaborative middleware is also used to execute a task layering mechanism, including: Based on the complexity of the input data or the confidence level of the preliminary diagnostic results, tasks are categorized into routine tasks, complex tasks, or unknown tasks. Regular tasks are assigned to the small temporal model for processing, complex tasks are assigned to the large language model for processing, and unknown tasks are assigned to both the small temporal model and the large language model for parallel processing.

5. The method for collaborative diagnosis of multiple agents in a time-series large and small model based on expert knowledge graphs according to claim 1, characterized in that, The step of performing correction or labeling operations based on rules in the expert knowledge graph includes: When the reasoning diagnosis result fails the second round of logic and mechanism verification, the expert knowledge graph generates correction suggestions based on preset rules. The large language model performs deep reasoning again based on the correction suggestions to obtain the corrected reasoning diagnosis results; If the revised reasoning diagnosis result still fails to pass verification, then the result will be marked as pending manual confirmation.

6. The method for multi-agent collaborative diagnosis of time-series large and small models based on expert knowledge graphs according to claim 1, characterized in that, The collaborative middleware maintains a global session context for recording the preliminary diagnostic results, the first round of verification results, the reason for triggering deep reasoning in the large language model, the reasoning diagnostic results, the second round of verification results, and information on correction or marking operations.

7. A multi-agent collaborative diagnostic system for time-series large and small models based on expert knowledge graphs, characterized in that, include: The expert knowledge graph construction module is used to acquire structural knowledge, fault knowledge, and diagnostic rule knowledge of the target equipment, and to construct an expert knowledge graph based on the structural knowledge, fault knowledge, and diagnostic rule knowledge of the target equipment. The multi-agent deployment module is used to deploy a small temporal model and a large language model respectively. The small temporal model is defined as the first agent and the large language model is defined as the second agent. The first agent is used for fast response and feature extraction, and the second agent is used for deep knowledge reasoning. A collaborative middleware construction module is used to build collaborative middleware based on a temporal small model and a large language model. The collaborative middleware is used to connect the temporal small model and the large language model. The collaborative diagnostics module is used to perform the following collaborative diagnostics based on the collaborative middleware: Preliminary diagnostic results are obtained based on a time-series small model, and then sent to an expert knowledge graph for the first round of logic and mechanism verification. Based on the confidence levels of the first round of verification results and preliminary diagnostic results, the large language model is selectively triggered to perform deep reasoning to obtain reasoning diagnostic results; The reasoning and diagnosis results are sent to the expert knowledge graph for a second round of logical and mechanistic verification. The final diagnosis result is confirmed based on the results of the second round of verification, or correction or labeling operations are performed according to the rules in the expert knowledge graph.

8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-6.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1-6.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1-6.

Images