Fault diagnosis method and system based on multi-agent architecture and dynamic prompt pool

Abstract

The invention discloses a fault diagnosis method and system based on a multi-agent architecture and a dynamic prompt pool, and relates to the technical field of artificial intelligence, and the method comprises the steps: receiving sensor data of equipment through a main agent, and carrying out the preprocessing of the sensor data; distributing different types of fault diagnosis requests to corresponding sub-agents by using the main agent; respectively retrieving related information in a dynamic prompt pool of the shared knowledge base by utilizing each sub-agent; diagnosing the preprocessed sensor data by using each sub-agent according to the corresponding related information to obtain a corresponding diagnosis result; and integrating the diagnosis results by using the main agent and then outputting a diagnosis report. According to the method and the device, the skilled tasks are processed in a targeted manner through the sub-agents, so that the overall performance and accuracy of diagnosis are improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and particularly to a fault diagnosis method and system based on a multi-agent architecture and a dynamic prompt pool. Background Art

[0002] With the rapid development of industrial automation and Internet of Things (IoT) technology, modern device systems have become increasingly complex. To ensure the efficient operation of these systems, it is crucial to perform real-time monitoring and fault diagnosis on each component in the device. Traditional fault diagnosis methods rely on a single model or algorithm, which may be effective in dealing with simple and linear problems, but are inadequate when faced with complex systems with multi-variable, non-linear dynamic changes. In addition, modern devices often consist of multiple interconnected but functionally diverse subsystems, and each subsystem may contain different types of sensors and data sources. Therefore, a method that can integrate multiple diagnostic tools and technologies is needed to achieve comprehensive and accurate monitoring and maintenance of the entire device system. In recent years, artificial intelligence (AI) technology, especially agent-based methods, has gradually attracted attention. These methods use agents to simulate and optimize complex system behaviors, thus providing a new solution for fault diagnosis.

[0003] Although existing fault diagnosis technologies have made certain progress, they still face many challenges in practical applications. First, traditional methods are usually designed for specific types of data or fault patterns, lacking flexibility and generality, and are difficult to adapt to different types of devices and their diverse fault patterns. Second, due to the large amount and wide source of data in modern device systems, how to effectively extract useful information and conduct comprehensive analysis has become a major problem. Existing single architectures cannot fully exploit the potential value in multi-source heterogeneous data, resulting in inaccurate and unreliable diagnostic results. In addition, many fault diagnosis processes rely on manual intervention, which not only increases labor costs but also may lead to low diagnostic efficiency. Summary of the Invention

[0004] The main purpose of the embodiments of this application is to propose a fault diagnosis method and system based on a multi-agent architecture and a dynamic prompt pool to improve the accuracy of fault diagnosis.

[0005] To achieve the above object, on the one hand, an embodiment of this application proposes a fault diagnosis method based on a multi-agent architecture and a dynamic prompt pool, and the method includes the following steps:

[0006] Use a main agent to receive sensor data of a device and preprocess the sensor data;

[0007] Use the main agent to allocate different types of fault diagnosis requests to corresponding sub-agents;

[0008] Each of the sub - agents retrieves relevant information from the dynamic hint pool of the shared knowledge base respectively;

[0009] Each of the sub - agents diagnoses the pre - processed sensor data according to the corresponding relevant information to obtain corresponding diagnosis results;

[0010] The main agent integrates each of the diagnosis results and then outputs a diagnosis report.

[0011] In some embodiments, the pre - processing of the sensor data includes the following steps:

[0012] Determine a binary mask matrix \(M\in\{0,1\}\), with size \(N\times T\); where \(M_{i,t} = 0\) represents a missing value \(X_{i,t}\); \(M_{i,t}=1\) represents an observed value;

[0013] Use the binary mask matrix to distinguish the observed values and missing values in the sensor data;

[0014] The expression for the observed values is: \(X_{obs}=X\odot M\);

[0015] The expression for the missing values is: \(X_{miss}=X\odot(1 - M)\);

[0016] where \(\odot\) represents element - wise multiplication;

[0017] Use a sliding window to construct multiple subsequences according to the observed values in the sensor data;

[0018] The expression for the subsequence is:

[0019] where \(\tau\) is the length of the sliding window.

[0020] In some embodiments, the step of each of the sub - agents retrieving relevant information from the dynamic hint pool of the shared knowledge base respectively includes the following steps:

[0021] Adopt an unsupervised K - means clustering method to identify the optimal cluster or state \(K\), and then assign the first cluster label to each time point in the pre - processed sensor data;

[0022] Use the sliding window to predict the second cluster label for the next \(v\) steps based on the samples observed in the previous \(\tau\) time steps;

[0023] The expression for the samples observed in the previous \(\tau\) time steps is: \(S\) t =X t-τ+1:t \(\in R\) N×τ ;

[0024] The expression of the second cluster of tags in the next v steps is: S t+1 = X t+1:t+v ∈ R N×v ;

[0025] Use each of the sub - agents to respectively retrieve the relevant information corresponding to the second cluster of tags in the dynamic hint pool of the shared knowledge base.

[0026] In some embodiments, the use of the main agent to allocate different types of fault diagnosis requests to corresponding sub - agents includes the following steps:

[0027] Use the main agent to allocate to the anomaly detection sub - agent, the fault prediction sub - agent, and the component status evaluation sub - agent according to the type of the fault diagnosis request.

[0028] In some embodiments, the use of each of the sub - agents to diagnose the pre - processed sensor data according to the corresponding relevant information to obtain corresponding diagnosis results includes the following steps:

[0029] Use the training set to calculate the robust normalized anomaly score of each variable at all time steps;

[0030] The expression of the robust normalized anomaly score is:

[0031]

[0032] Where, represents the robust normalized anomaly score;

[0033] Calculate the moving average of the maximum robust normalized anomaly score in the validation set;

[0034] The expression of the moving average is:

[0035]

[0036] Where, Th represents the moving average, wa is the number of time points in the moving average calculation, and Tval is the time point in the validation set;

[0037] Use the anomaly detection sub - agent to diagnose the pre - processed sensor data according to the moving average to obtain an anomaly diagnosis result.

[0038] In some embodiments, the method further includes the following steps:

[0039] Use each of the sub - agents to interact with the external environment and then continuously learn the optimal strategy;

[0040] Adjust the behaviors of the sub - agents according to the feedback information of each sub - agent by using the main agent.

[0041] In some embodiments, the method further includes the following steps:

[0042] Dynamically monitor the state parameters of the device and the environmental change parameters;

[0043] Adjust the structures and parameters of the main agent and each sub - agent according to the state parameters of the device and the environmental change parameters.

[0044] To achieve the above object, on the other hand, an embodiment of the present application proposes a fault diagnosis system based on a multi - agent architecture and a dynamic hint pool, and the system includes:

[0045] A data processing module, configured to receive sensor data of a device by using a main agent and pre - process the sensor data;

[0046] A request allocation module, configured to allocate different types of fault diagnosis requests to corresponding sub - agents by using the main agent;

[0047] An information retrieval module, configured to retrieve relevant information respectively in the dynamic hint pool of a shared knowledge base by using each sub - agent;

[0048] A data diagnosis module, configured to diagnose the pre - processed sensor data according to the corresponding relevant information by using each sub - agent to obtain corresponding diagnosis results;

[0049] A diagnosis integration module, configured to integrate each diagnosis result by using the main agent and then output a diagnosis report.

[0050] To achieve the above object, on the other hand, an embodiment of the present application proposes an electronic device, which includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the above - mentioned method is implemented.

[0051] To achieve the above object, on the other hand, an embodiment of the present application proposes a computer - readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the above - mentioned method is implemented.

[0052] The embodiments of the present application at least include the following beneficial effects:

[0053] The solution of this application includes: using the main agent to receive the sensor data of the device and preprocess the sensor data; using the main agent to allocate different types of fault diagnosis requests to the corresponding sub-agents; using each sub-agent to respectively retrieve relevant information in the dynamic hint pool of the shared knowledge base; using each sub-agent to diagnose the preprocessed sensor data according to the corresponding relevant information to obtain the corresponding diagnosis results; using the main agent to integrate each diagnosis result and then output a diagnosis report. This application improves the overall performance and accuracy of diagnosis by having each sub-agent specifically handle the tasks it is good at. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions in the embodiments of this application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of this application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0055] Figure 1 It is a schematic flowchart of the fault diagnosis method based on a multi-agent architecture and a dynamic hint pool provided by an embodiment of this application;

[0056] Figure 2 It is an example architecture diagram of the fault diagnosis method based on a multi-agent architecture and a dynamic hint pool provided by an embodiment of this application;

[0057] Figure 3 It is an example flowchart of the fault diagnosis method based on a multi-agent architecture and a dynamic hint pool provided by an embodiment of this application;

[0058] Figure 4 It is a schematic structural diagram of the fault diagnosis system based on a multi-agent architecture and a dynamic hint pool provided by an embodiment of this application;

[0059] Figure 5 It is a schematic hardware structure diagram of an electronic device provided by an embodiment of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0060] In order to make the purpose, technical solutions and advantages of this application more clear, the following will further describe this application in detail in combination with the drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain this application and are not used to limit this application. When the following description involves the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with the embodiments of this application. They are only examples of devices and methods consistent with some aspects of the embodiments of this application detailed in the appended claims.

[0061] It can be understood that the terms "first", "second", etc. used in this application may be used herein to describe various concepts, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of this application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, words such as "if" and "when" used herein may be interpreted as "when...", "while...", or "in response to determining".

[0062] The terms "at least one", "a plurality of", "each", "any one", etc. used in this application, at least one includes one, two, or more than two, a plurality of includes two or more than two, each refers to each one in the corresponding plurality, and any one refers to any one in the plurality.

[0063] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0064] Before elaborating on the embodiments of this application in detail, some related technologies involved in the embodiments of this application are described first as follows:

[0065] Related technologies: Existing agent-based fault diagnosis frameworks have, to a certain extent, improved the above problems. This framework realizes the efficient management and optimization of complex systems through a hierarchical multi-agent architecture. Specifically, each agent is responsible for a specific task, such as data collection, feature extraction, fault prediction, etc., and can call different tools and algorithm models to complete their respective tasks. The figure shows the workflow of this hierarchical multi-agent architecture: First, the master agent receives data from the device and selects the most suitable sub-agent according to the task requirements. Each sub-agent can retrieve relevant information from the shared prompt pool and generate corresponding outputs. These knowledge bases, as dynamic knowledge bases, provide the necessary historical background and understanding to effectively analyze newly input data.

[0066] Disadvantages of related technologies: Relying on a single model or algorithm during the fault diagnosis process, which results in poor performance when dealing with complex multivariable and non-linear dynamic changes. Especially when facing diverse fault modes in modern devices, the diagnostic accuracy is relatively low. Since these methods usually need to be designed for specific types of data or fault modes, they lack flexibility and generality and are difficult to adapt to different types of devices and their diverse fault modes. In addition, the collaboration mechanism between sub-agents in existing fault diagnosis systems is not perfect, resulting in poor information transmission between sub-agents and the overall performance not reaching the optimal level. This inefficient collaboration mechanism not only affects the accuracy of the diagnostic results but also may lead to a high error rate, thus reducing the reliability of the system.

[0067] To solve these problems, some embodiments of the present application propose a fault diagnosis method and system based on a multi-agent architecture and a dynamic hint pool. By introducing a hierarchical multi-agent architecture, each agent can call different tools and algorithm models and is specifically responsible for the diagnostic tasks of specific components. The modular design of the present application not only improves the flexibility and scalability of the system but also can more accurately identify and locate faults, greatly enhancing the overall efficiency of fault diagnosis.

[0068] Specifically, the present application aims to solve multiple key technical and performance problems existing in the prior art to improve the overall efficiency and reliability of the fault diagnosis system. First, aiming at the problem that a single model or algorithm in the prior art performs poorly when dealing with complex multivariable and non-linear dynamic changes, the embodiments of the present application propose a multi-agent architecture integrating multiple tools and algorithms. By flexibly calling different sub-agents, it can adapt to different types of devices and their diverse fault modes, thus significantly improving the diagnostic accuracy and flexibility. Second, the collaboration mechanism between sub-agents in the prior art is not perfect, resulting in poor information transmission and the overall performance not reaching the optimal level. The embodiments of the present application ensure seamless cooperation between sub-agents by introducing an optimized collaboration mechanism and an efficient information transmission protocol, improving the overall performance of the system.

[0069] In addition, the existing technology is less efficient in data processing and analysis, and requires a large amount of manual intervention for data preprocessing and result interpretation. This not only increases the labor cost but also significantly reduces the response speed of the system. To solve this problem, this application designs a more automated fault diagnosis solution, reduces the need for manual intervention, and significantly improves the response speed and processing efficiency of the system through efficient algorithms and optimized data processing flows. Specifically, this application adopts a Dynamic Prompting Mechanism (DPM) and Direct Preference Optimization (DPO), enabling the system to quickly and accurately perform fault prediction and diagnosis, and further enhancing the real-time monitoring ability of the system.

[0070] In summary, the embodiments of this application provide a fault diagnosis method and system based on a multi-agent architecture and a dynamic prompt pool. The technical solution of this application includes: using a main agent to receive sensor data of a device and preprocess the sensor data; using the main agent to allocate different types of fault diagnosis requests to corresponding sub-agents; using each sub-agent to separately retrieve relevant information from the dynamic prompt pool of the shared knowledge base; using each sub-agent to diagnose the preprocessed sensor data according to the corresponding relevant information to obtain corresponding diagnosis results; using the main agent to integrate each diagnosis result and then output a diagnosis report. This application improves the overall performance and accuracy of diagnosis by enabling each sub-agent to specifically handle the tasks it is good at.

[0071] The embodiments of this application provide a fault diagnosis method and system based on a multi-agent architecture and a dynamic prompt pool, which relates to the field of artificial intelligence technology. The fault diagnosis method and system based on a multi-agent architecture and a dynamic prompt pool provided by the embodiments of this application can be applied to terminals, servers, or software running on terminals or servers. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, a vehicle-mounted terminal, etc., but is not limited thereto; the server side can be configured as an independent physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network; the software can be an application implementing the fault diagnosis method based on a multi-agent architecture and a dynamic prompt pool, etc., but is not limited to the above forms.

[0072] This application can be used in numerous general-purpose or special-purpose computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and so on. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. This application can also be practiced in a distributed computing environment where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0073] Referring to Figure 1 , an embodiment of this application provides a fault diagnosis method based on a multi-agent architecture and a dynamic hint pool. This method can include but is not limited to S100 to S140, specifically as follows:

[0074] S100: Use the main agent to receive sensor data of the device and preprocess the sensor data.

[0075] Further, the preprocessing of the sensor data in S100 can include the following steps S101 to S103:

[0076] S101: Determine a binary mask matrix M ∈ {0, 1}, with a size of N × T; where Mi,t = 0 represents a missing value Xi,t; Mi,t = 1 represents an observed value;

[0077] S102: Use the binary mask matrix to distinguish the observed values and missing values in the sensor data;

[0078] The expression for the observed value is: Xobs = X ⊙ M;

[0079] The expression for the missing value is: Xmiss = X ⊙ (1 - M);

[0080] where ⊙ represents element-wise multiplication;

[0081] S103: Use a sliding window to construct multiple subsequences based on the observed values in the sensor data;

[0082] The expression for the subsequence is:

[0083] where τ is the length of the sliding window.

[0084] S110: Use the main agent to allocate different types of fault diagnosis requests to corresponding sub-agents.

[0085] Further, S110 may include the following steps S111:

[0086] S111: Use the main agent to allocate according to the type of the fault diagnosis request to an anomaly detection sub-agent, a fault prediction sub-agent, and a component status evaluation sub-agent.

[0087] S120: Use each of the sub-agents to separately retrieve relevant information from the dynamic hint pool of the shared knowledge base.

[0088] Further, S120 may include the following steps S121 to S123:

[0089] S121: Use the unsupervised K-means clustering method to identify the optimal cluster or state K, and then assign a first cluster label to each time point in the preprocessed sensor data;

[0090] S122: Use the sliding window to predict the second cluster label for the next v steps based on the samples observed in the previous τ time steps;

[0091] The expression for the samples observed in the previous τ time steps is: S t = X t-τ+1：t ∈ R N×τ ;

[0092] The expression for the second cluster label for the next v steps is: S t+1 = X t+1：t+v ∈ R N×v ;

[0093] S123: Use each of the sub-agents to separately retrieve the relevant information corresponding to the second cluster label from the dynamic hint pool of the shared knowledge base.

[0094] S130: Use each of the sub-agents to diagnose the preprocessed sensor data according to the corresponding relevant information to obtain corresponding diagnosis results.

[0095] Further, S130 may include the following steps S131 to S133:

[0096] S131: Use the training set to calculate the robust normalized anomaly score of each variable at all time steps;

[0097] The expression for the robust normalized anomaly score is:

[0098]

[0099] Where, representing the robust normalized anomaly score;

[0100] S132: Calculate the moving average of the maximum robust normalized anomaly score in the validation set;

[0101] The expression of the moving average is:

[0102]

[0103] where Th represents the moving average, wa is the number of time points in the moving average calculation, and Tval is the time point in the validation set;

[0104] S133: Use the anomaly detection sub-agent to diagnose the preprocessed sensor data according to the moving average to obtain an anomaly diagnosis result.

[0105] S140: Use the main agent to integrate each diagnosis result and then output a diagnosis report.

[0106] To further improve the accuracy of diagnosis, the embodiments of the present application may further include the following steps S151 - S152:

[0107] S151: Dynamically monitor the state parameters and environmental change parameters of the device;

[0108] S152: Adjust the structures and parameters of the main agent and each sub-agent according to the state parameters of the device and the environmental change parameters.

[0109] As another alternative embodiment, the embodiments of the present application may further include the following steps S161 - S162:

[0110] S161: Use the interaction between each sub-agent and the external environment to continuously learn the optimal strategy;

[0111] S162: Use the main agent to adjust the behaviors of each sub-agent according to the feedback information of each sub-agent.

[0112] Next, the solution of the embodiments of the present application will be introduced and described in detail in combination with specific application examples.

[0113] The overall architecture of this embodiment is as Figure 2As shown, it includes a Master Agent, multiple Sub-Agents, and a Shared Knowledge Base. The Master Agent is responsible for receiving various sensor data from devices and selecting the most suitable Sub-Agent according to task requirements. Each Sub-Agent can retrieve relevant information from the shared knowledge base and generate corresponding outputs. The operation of the entire system depends on an efficient collaboration mechanism and an optimized information transfer protocol to ensure seamless cooperation among Sub-Agents.

[0114] 1. Data Processing:

[0115] Define a binary mask matrix M ∈ {0, 1}^(N×T), where M_{i,t} = 0 represents the missing value X_{i,t}, and M_{i,t} = 1 represents the observed value in the data matrix X ∈ R^(N×T). Use this matrix to distinguish known and unknown data points:

[0116] X_obs = X ⊙ M;

[0117] X_miss = X ⊙ (1 - M);

[0118] where X_obs represents the observed values, X_miss represents the missing values, and ⊙ represents element-wise multiplication.

[0119] Construct subsequences using a sliding window of size τ These subsequences are constructed based on the samples observed in the previous τ steps and are used to predict the missing values in the next v steps

[0120] 2. Anomaly Detection:

[0121] Assume that the time series dataset exhibits normal behavior within the initial T_train time stamps. Then, any pattern that deviates from this normal behavior at subsequent time stamps t > T_train is considered an anomaly. Calculate the robust normalized anomaly score for each variable i at all time steps using the training set T_train:

[0122]

[0123] Calculate the simple moving average Th of the maximum anomaly score A_{t+1}^i in the validation set:

[0124]

[0125] where w_a is the number of time points in the moving average calculation, and T_val is the number of time points in the validation set.

[0126] 3. Classification:

[0127] Using the unsupervised K - means clustering method, identify the optimal cluster or state K, and assign cluster labels C ∈ R N×T to each time point in the data matrix X ∈ R T . Then, use a sliding window of size τ to predict the cluster labels S for the next v steps t+1 = X t+1：t+v ∈ R N×v , based on the samples S observed in the previous τ time steps t = X t-τ+1：t ∈ R N×τ .

[0128] Referring to Figure 3 , the specific description of this embodiment may include:

[0129] 1. The main agent is the core and is responsible for coordinating the work of each sub - agent:

[0130] Data reception and pre - processing: The main agent collects real - time data from device sensors and performs preliminary data cleaning and formatting. Task assignment: According to the received data type and task requirements, the main agent selects the most suitable sub - agent for processing. Result aggregation and feedback: The main agent aggregates the outputs of each sub - agent and generates a final diagnostic report, which is fed back to the user or the control system.

[0131] 2. Sub - agents:

[0132] Each sub - agent is responsible for a specific type of fault diagnosis task, such as time - series prediction, anomaly detection, etc. Data acquisition: Receive pre - processed data from the main agent. Feature extraction: Use feature extraction algorithms to extract useful feature information from the original data. Model invocation and analysis: According to the task requirements, call different tools and algorithm models for data analysis and fault prediction. Result generation and feedback: Feed back the analysis results to the main agent for further processing.

[0133] 3. Shared knowledge base:

[0134] Collaboration mechanism and information transfer protocol. To ensure the efficient collaboration among sub - agents, the embodiment of this application designs an optimized collaboration mechanism and information transfer protocol. Task scheduling: The main agent dynamically schedules the tasks of each sub - agent according to task priorities and resource availability. Information synchronization: Ensure information consistency among sub - agents through a real - time synchronization mechanism. Result feedback: Each sub - agent timely feeds back the processing results to the main agent for subsequent processing.

[0135] 4. Collaboration mechanism and information transfer protocol:

[0136] Adaptive Model Adjustment Mechanism. To address the problems brought about by the fixed model structure and parameter settings in the prior art, this embodiment introduces an adaptive model adjustment mechanism. Model Monitoring: Continuously monitor the device status and environmental changes in real time, and evaluate the effectiveness of the current model. Model Update: Automatically adjust the model structure and parameters according to the monitoring results to ensure that the model is always in an optimal state. Performance Evaluation: Regularly evaluate the model performance to ensure the stability and reliability of the system.

[0137] In summary, this embodiment includes the following technical solutions:

[0138] Hierarchical Multi-Agent Architecture: This embodiment adopts a hierarchical multi-agent architecture. The main agent is responsible for coordinating the work of each sub-agent and selecting the most suitable sub-agent for processing according to the task requirements. This way of clear division of labor and each performing its own duties enables each sub-agent to focus on the tasks it is good at, such as time series prediction, anomaly detection, or feature extraction, etc., thus improving the overall performance of the system.

[0139] Shared Knowledge Base and Dynamic Hint Pool: The shared knowledge base in the system exists in the form of a dynamic hint pool, providing the necessary historical background and understanding to analyze new input data more effectively. In this way, the sub-agents can quickly obtain relevant information and generate accurate outputs, greatly improving the diagnostic accuracy and flexibility.

[0140] Adaptive Model Adjustment Mechanism: To cope with the challenges brought about by device or environmental changes, this embodiment introduces an adaptive model adjustment mechanism. This mechanism can monitor the device status and environmental changes in real time and automatically adjust the model structure and parameters to ensure that the model is always in an optimal state. This not only improves the flexibility and versatility of the system but also reduces the maintenance cost and operation complexity.

[0141] 1. Design and Implementation of Hierarchical Multi-Agent Architecture: The uniqueness of this embodiment lies in the design and implementation of its hierarchical multi-agent architecture. Through the collaborative work of the main agent and multiple sub-agents, this architecture realizes the efficient monitoring and fault diagnosis of complex multi-variable and non-linear dynamic change systems.

[0142] 2. Management of Dynamic Hint Pool and Shared Knowledge Base: The management method of the dynamic hint pool and shared knowledge base in this embodiment is another important protection point. Through the dynamic update and efficient retrieval mechanism, the system can make full use of historical data and background information to provide more accurate and reliable fault diagnosis results.

[0143] 3. Adaptive Model Adjustment Mechanism: The adaptive model adjustment mechanism proposed in this embodiment can automatically update and optimize the model when the device or environment changes, ensuring the stability and reliability of the system. This mechanism greatly reduces the maintenance cost and operation complexity and has significant practical value.

[0144] Beneficial effects brought by the technical solution of this embodiment:

[0145] By introducing a hierarchical multi-agent architecture, the performance and efficiency of time series analysis tasks are significantly improved. Specifically, the main agent is responsible for coordinating each sub-agent, selecting the most suitable sub-agent to execute a specific task, and assigning the task to it. This way of clear division of labor and each performing its own duties enables each sub-agent to focus on the tasks it is good at, such as time series prediction, anomaly detection, or feature extraction, etc. In addition, the sub-agents can obtain relevant information from the shared knowledge base, which exists in the form of a dynamic prompt pool, providing the sub-agents with the necessary historical background and understanding to more effectively analyze the newly input data.

[0146] For the specific application of time series prediction, a sliding window method is adopted to construct time series subsequences. This method can use the data of the past τ time steps to predict the data of the future ν time steps, thereby improving the accuracy and reliability of the prediction. Experimental results show that in the tests on multiple benchmark datasets, the performance of the SelfExtend-Agentic-RAG framework is better than or at least reaches the current state-of-the-art methods. For example, on the METR-LA and PEMS-BAY datasets, using different evaluation metrics (such as RMSE, MAE, MAPE), the Agentic-RAG model performs better than other comparison models in different prediction ranges. Especially in the prediction range of Horizon@12, the error metrics of this model are significantly lower than those of other models, which proves its superiority in long-term prediction.

[0147] In addition, this application also provides alternative solutions to improve the accuracy of diagnosis:

[0148] Multi-agent architecture based on reinforcement learning: In addition to using the method of combining a hierarchical multi-agent architecture with a dynamic prompt pool, a multi-agent architecture based on reinforcement learning (RL) can also be adopted to implement the fault diagnosis system. In this architecture, each sub-agent continuously learns the optimal policy through interaction with the environment to improve the accuracy and efficiency of fault diagnosis. The main agent is responsible for coordinating each sub-agent and adjusting their behaviors according to the feedback. The advantage of this method is that it can adaptively optimize the behavior strategies of the sub-agents, especially performing better in complex and changeable environments.

[0149] The specific implementation steps of the alternative solution are as follows:

[0150] Initialization: Define the state space, action space, and reward function.

[0151] Training phase: Each sub-agent interacts with the environment and continuously updates its policy using algorithms such as Q-learning or deep reinforcement learning (DQN).

[0152] Collaboration mechanism: Through a shared reward mechanism, ensure collaboration among sub-agents to jointly optimize the overall performance.

[0153] Application phase: Deploy the trained sub-agents into the actual system for real-time monitoring and fault diagnosis.

[0154] Refer to Figure 4 , the embodiment of the present application also provides a fault diagnosis system based on a multi-agent architecture and a dynamic hint pool, which can implement the above-mentioned fault diagnosis method based on a multi-agent architecture and a dynamic hint pool. The system includes:

[0155] A data processing module, configured to use the main agent to receive sensor data of the device and preprocess the sensor data;

[0156] A request allocation module, configured to use the main agent to allocate different types of fault diagnosis requests to corresponding sub-agents;

[0157] An information retrieval module, configured to use each of the sub-agents to respectively retrieve relevant information in the dynamic hint pool of the shared knowledge base;

[0158] A data diagnosis module, configured to use each of the sub-agents to diagnose the preprocessed sensor data according to the corresponding relevant information to obtain corresponding diagnosis results;

[0159] A diagnosis integration module, configured to use the main agent to integrate each of the diagnosis results and then output a diagnosis report.

[0160] It can be understood that the content in the above method embodiments is applicable to the system embodiments of the present application. The functions specifically implemented by the system embodiments of the present application are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0161] The embodiment of the present application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the method of the embodiment of the present application. The electronic device can be any intelligent terminal including a tablet computer, an in-vehicle computer, etc.

[0162] It can be understood that the content in the above method embodiments is applicable to the device embodiments of the present application. The functions specifically implemented by the device embodiments of the present application are the same as those of the method of the present application, and the beneficial effects achieved are also the same as those of the method of the present application.

[0163] Please refer to Figure 5 ,Figure 5 Schematically shows the hardware structure of an electronic device according to another embodiment. The electronic device includes:

[0164] A processor 501, which can be implemented in ways such as a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0165] A memory 502, which can be implemented in forms such as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 502 can store an operating system and other application programs. When implementing the technical solutions provided in the embodiments of this specification through software or firmware, the relevant program codes are stored in the memory 502 and are called by the processor 501 to execute the methods of the embodiments of the present application;

[0166] An input / output interface 503, which is used to implement information input and output;

[0167] A communication interface 504, which is used to implement communication interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.);

[0168] A bus 505, which transmits information between various components of the device (such as the processor 501, the memory 502, the input / output interface 503, and the communication interface 504);

[0169] Among them, the processor 501, the memory 502, the input / output interface 503, and the communication interface 504 achieve communication connections with each other inside the device through the bus 505.

[0170] The embodiments of the present application also provide a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the method of the present application.

[0171] It can be understood that the content in the above method embodiments is applicable to the embodiments of this storage medium. The functions specifically implemented by the embodiments of this storage medium are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0172] As a non-transitory computer-readable storage medium, the memory can be used to store non-transitory software programs and non-transitory computer-executable programs. In addition, the memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories can be connected to the processor through a network. Examples of the above networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0173] The embodiments described in the embodiments of the present application are for more clearly illustrating the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those skilled in the art will know that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of the present application are equally applicable to similar technical problems.

[0174] Those skilled in the art can understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or combine certain steps, or different steps.

[0175] The system embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separated, that is, they may be located in one place, or may be distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0176] Those of ordinary skill in the art can understand that all or some of the steps in the methods disclosed above, and the functional units / modules in the systems and devices can be implemented as software, firmware, hardware, and appropriate combinations thereof.

[0177] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or modules does not have to be limited to those steps or modules clearly listed, but may include other steps or modules not clearly listed or inherent to these processes, methods, products, or devices.

[0178] It should be understood that in this application, "at least one (item)" means one or more, and "a plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that there can be three relationships. For example, "A and / or B" can mean: only A exists, only B exists, and both A and B exist at the same time. Among them, A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects before and after. "At least one (item) of the following" or its similar expression refers to any combination of these items, including any combination of single item (item) or plural items (items). For example, at least one (item) of a, b, or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0179] In several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of the above-mentioned modules is only a logical function division. In actual implementation, there can be other division methods. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the coupling or direct coupling or communication connection shown or discussed with each other can be through some interfaces. The indirect coupling or communication connection of systems or modules can be in electrical, mechanical or other forms.

[0180] The modules described above as separate components may or may not be physically separated. The components shown as modules may or may not be physical modules, that is, they can be located in one place, or they can be distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0181] In addition, each functional module in various embodiments of this application can be integrated in a processing module, or each module can exist physically alone, or two or more modules can be integrated in one module. The above-mentioned integrated module can be implemented in the form of hardware or in the form of a software functional module.

[0182] If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes multiple instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The foregoing storage medium includes: various media that can store programs, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs.

[0183] The preferred embodiments of the embodiments of the present application have been described above with reference to the accompanying drawings. However, this does not limit the scope of the rights of the embodiments of the present application. Any modifications, equivalent replacements, and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of the present application shall fall within the scope of the rights of the embodiments of the present application.

Claims

1. A fault diagnosis method based on a multi-agent architecture and a dynamic hint pool, characterized in that The method includes the following steps: Using the main agent to receive the sensor data of the device and preprocess the sensor data; Using the main agent to allocate different types of fault diagnosis requests to corresponding sub-agents; Using each of the sub-agents to respectively retrieve relevant information in the dynamic hint pool of the shared knowledge base; Using each of the sub-agents to diagnose the preprocessed sensor data according to the corresponding relevant information to obtain corresponding diagnosis results; Using the main agent to integrate each of the diagnosis results and then output a diagnosis report.

2. The fault diagnosis method based on a multi-agent architecture and a dynamic hint pool according to claim 1, wherein The preprocessing of the sensor data includes the following steps: Determine a binary mask matrix M ∈ {0, 1}, with a size of N×T; where Mi,t = 0 represents a missing value Xi,t; Mi,t = 1 represents an observed value; Using the binary mask matrix to distinguish the observed values and missing values in the sensor data; The expression for the observed value is: Xobs = X ⊙ M; The expression for the missing value is: Xmiss = X ⊙ (1 - M); where ⊙ represents element-wise multiplication; Using a sliding window to construct multiple subsequences according to the observed values in the sensor data; The expression of the subsequence is as follows: where τ is the length of the sliding window.

3. The fault diagnosis method based on a multi-agent architecture and a dynamic hint pool according to claim 2, characterized in that, The step of using each of the sub-agents to respectively retrieve relevant information in the dynamic hint pool of the shared knowledge base includes the following steps: Adopting an unsupervised K-means clustering method to identify the optimal cluster or state K, and then assigning a first cluster label to each time point in the preprocessed sensor data; Predicting the second cluster label for the next v steps using the sliding window based on the samples observed in the previous τ time steps; The expression for the samples observed in the previous τ time steps is: S t = X t-τ+1 : t ∈ R N×τ ; The expression of the second cluster of tags in the next ν steps is: S t+1 = X t+1：t+v ∈ R N×v ; Using each of the sub-agents to respectively retrieve the corresponding relevant information corresponding to the second cluster label in the dynamic hint pool of the shared knowledge base.

4. The fault diagnosis method based on a multi-agent architecture and a dynamic hint pool according to claim 1, characterized in that The step of using the main agent to allocate different types of fault diagnosis requests to corresponding sub-agents includes the following steps: Using the main agent to allocate according to the type of the fault diagnosis request to an anomaly detection sub-agent, a fault prediction sub-agent, and a component status evaluation sub-agent.

5. The fault diagnosis method based on a multi-agent architecture and a dynamic hint pool according to claim 4, wherein The step of using each of the sub-agents to diagnose the preprocessed sensor data according to the corresponding relevant information to obtain corresponding diagnosis results includes the following steps: Calculating the robust normalized anomaly score of each variable at all time steps using a training set; The expression for the robust normalized anomaly score is: Among them, represents the robust normalized anomaly score; Calculating the moving average of the maximum robust normalized anomaly score in the validation set; The expression for the moving average is: where Th represents the moving average, wa is the number of time points in the moving average calculation, and Tval is the time point in the validation set; Using the anomaly detection sub-agent to diagnose the preprocessed sensor data according to the moving average to obtain an anomaly diagnosis result.

6. The fault diagnosis method based on a multi-agent architecture and a dynamic hint pool according to claim 1, wherein The method further includes the following steps: Using each of the sub-agents to interact with the external environment and then continuously learn the optimal strategy; Using the main agent to adjust the behavior of each of the sub-agents according to the feedback information of each of the sub-agents.

7. The fault diagnosis method based on a multi-agent architecture and a dynamic hint pool according to any one of claims 1 to 6, characterized in that, The method further includes the following steps: Dynamically monitoring the state parameters and environmental change parameters of the device. Adjust the structures and parameters of the main agent and each of the sub-agents according to the status parameters of the device and the environmental change parameters.

8. A fault diagnosis system based on a multi-agent architecture and a dynamic hint pool, characterized in that The system includes: A data processing module, configured to use the main agent to receive sensor data of the device and preprocess the sensor data; A request allocation module, configured to use the main agent to allocate different types of fault diagnosis requests to corresponding sub-agents; An information retrieval module, configured to use each of the sub-agents to respectively retrieve relevant information in the dynamic hint pool of the shared knowledge base; A data diagnosis module, configured to use each of the sub-agents to diagnose the preprocessed sensor data according to the corresponding relevant information to obtain corresponding diagnosis results; A diagnosis integration module, configured to use the main agent to integrate each of the diagnosis results and then output a diagnosis report.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the method described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, the method described in any one of claims 1 to 7 is implemented.

Images