Equipment operation fault diagnosis method and device, medium, equipment and program product
By combining the fault diagnosis model of Monte Carlo tree search and large language model in industrial process control, the problems of long analysis time and poor interpretation of traditional fault diagnosis models are solved, and fast and accurate fault pattern recognition is achieved, suitable for complex industries. environment.
Patent Information
- Application Number
- CN202411073467.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-06
- Publication Date
- 2024-11-15
AI Technical Summary
Traditional fault diagnosis models are long-term in industrial process control, poorly interpretable, and are susceptible to data loss, resulting in inaccurate fault detection and affecting production quality and efficiency.
The fault diagnosis model is constructed using Monte Carlo tree search and large language model. By obtaining device status information, using Monte Carlo tree search to determine the decision path, and generating diagnostic results through large language models, improving the accuracy and interpretability of the diagnosis.
It realizes the rapid and accurate identification of potential equipment failure modes in real-time production control, improves the accuracy and interpretability of fault diagnosis, and adapts to complex and changeable industrial environments.
Smart Images

Figure CN118963324A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the field of industrial process control, and in particular, to a method, apparatus, medium, equipment and program product for diagnosing equipment operation faults. Background Art
[0002] Fault diagnosis in the field of industrial process control is a critical task, which aims to detect and repair faults in a timely manner by monitoring, analyzing and identifying abnormal conditions in the system to ensure the stability and reliability of the production process. The first step in fault diagnosis is to monitor the status of the industrial process control system. Fault diagnosis systems usually need to be able to respond in real time, provide timely alarms, and automatically take measures to repair the fault or reduce the impact of the fault on the production process when possible.
[0003] Traditional fault diagnosis models usually involve complex data processing methods and complex pattern definitions. These methods have long analysis times, poor interpretability, and are not suitable for real-time production control. Fault detection is often affected by missing data, which may lead to inaccurate results and thus affect production quality and efficiency. Summary of the invention
[0004] The purpose of the present disclosure is to provide a method, apparatus, medium, equipment and program product for diagnosing equipment operation faults, which can accurately diagnose faults of equipment in actual operation.
[0005] In order to achieve the above objectives, in a first aspect, the present disclosure provides a method for diagnosing a device operation fault, comprising: Get device status information; The state information is input into a fault diagnosis model to obtain a diagnosis result output by the fault diagnosis model, wherein the diagnosis result includes the probability of the device corresponding to different fault types, and the fault diagnosis model is constructed based on Monte Carlo tree search and a large language model.
[0006] In a second aspect, the present disclosure provides a device for diagnosing equipment operation faults, comprising: An acquisition module, configured to acquire status information of a device; The execution module is configured to input the state information into a fault diagnosis model to obtain a diagnosis result output by the fault diagnosis model, wherein the diagnosis result includes the probability of the device corresponding to different fault types, and the fault diagnosis model is constructed based on Monte Carlo tree search and a large language model.
[0007] In a third aspect, the present disclosure provides a non-temporary computer-readable storage medium having a computer program stored thereon, which implements the method described in the first aspect when executed by a processor.
[0008] In a fourth aspect, the present disclosure provides an electronic device, including: a memory having a computer program stored thereon; A processor is used to execute the computer program in the memory to implement the method described in the first aspect.
[0009] In a fifth aspect, the present disclosure provides an extreme machine program product, including a computer program, which implements the method described in the first aspect when executed by a processor.
[0010] Through the above technical scheme, a fault diagnosis model is constructed based on the Monte Carlo tree search and the large language model. The fault diagnosis model has the advantages of both the Monte Carlo tree search and the large language model. It can more accurately identify the potential failure modes of the equipment while having good interpretability, thereby accurately diagnosing the faults of the equipment in actual operation.
[0011] Other features and advantages of the present disclosure will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS
[0012] The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present disclosure but do not constitute a limitation of the present disclosure. In the accompanying drawings: Figure 1 The figure is a flow chart of a method for diagnosing a device operation fault according to an exemplary embodiment of the present disclosure.
[0013] Figure 2 It is a block diagram of a fault diagnosis model according to an exemplary embodiment of the present disclosure.
[0014] Figure 3 It is a schematic diagram of test results obtained according to the device operation fault diagnosis method shown in an exemplary embodiment of the present disclosure.
[0015] Figure 4 The present invention is a block diagram of a device for diagnosing a fault in equipment operation according to an exemplary embodiment of the present disclosure.
[0016] Figure 5 is a block diagram of an electronic device according to an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION
[0017] The specific implementation of the present disclosure is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described herein is only used to illustrate and explain the present disclosure, and is not used to limit the present disclosure.
[0018] It should be noted that all actions of acquiring signals, information or data in the present disclosure are carried out in compliance with the relevant data protection laws and policies of the country where the device is located and with the authorization given by the owner of the corresponding device.
[0019] As mentioned in the background, the first step in fault diagnosis in the field of industrial process control is to monitor the state of the industrial process control system. This involves the use of sensors, such as obtaining real-time data by measuring parameters such as temperature, pressure, and flow. These real-time data are compared with the expected normal operating range to detect any abnormalities that may occur in the equipment. For example, during the normal operation of the equipment, a large amount of data is collected and a baseline model is established. The baseline model is used to represent the normal operating state, so that the system can detect any abnormalities in the equipment that deviate from the normal operating state while the equipment is running.
[0020] By monitoring data in real time and comparing it to a baseline model, the system can identify any anomalies or deviations from normal operating conditions. Anomaly detection is a key step in early fault diagnosis. Once an anomaly is detected, the next step is to identify the specific fault that caused the anomaly. This involves a deeper level of analysis and may require a combination of expertise, model interpretation techniques, and the experience of domain experts. Once a fault is identified, further diagnosis and root cause analysis are required. This requires a deep understanding of the physical principles of industrial processes, chemical processes, and other aspects of knowledge. Methods such as expert systems and expertise bases can also be used to assist this process. Traditional fault diagnosis systems typically need to be able to respond in real time, provide timely alerts, and automatically take measures to repair the fault or reduce the impact of the fault on the production process when possible.
[0021] However, the inventors found that traditional fault diagnosis models usually involve complex data processing methods and complex pattern definitions. These methods have long analysis times, poor interpretability, and are not suitable for real-time production control. Fault detection is often affected by missing data, which may lead to inaccurate results and thus affect production quality and efficiency.
[0022] In view of this, the present disclosure provides a method, apparatus, medium, device and program product for diagnosing equipment operation faults, which can accurately diagnose faults of equipment in actual operation.
[0023] Figure 1 FIG. 1 is a flow chart of a method for diagnosing a device operation fault according to an exemplary embodiment of the present disclosure. Figure 1 As shown, the device operation fault diagnosis method may include the following steps: In step S11, the status information of the device is obtained.
[0024] It is worth noting that the status information may include information corresponding to the normal status and / or fault status of the device in different operating modes.
[0025] In step S12, the state information is input into a fault diagnosis model to obtain a diagnosis result output by the fault diagnosis model, wherein the diagnosis result includes the probability of the device corresponding to different fault types, and the fault diagnosis model is constructed based on Monte Carlo tree search and a large language model.
[0026] It is worth noting that the equipment disclosed in the present invention includes but is not limited to large-scale coal mining equipment. Accordingly, text data can be run through large-scale coal mining equipment as training and testing experimental data. The fault diagnosis model in the present invention supports the analysis and processing capabilities of no less than tens of millions of original text information.
[0027] It is worth mentioning that traditional fault diagnosis models can be divided into two categories: rule-based models and deep learning-based models.
[0028] Rule-based models use pre-defined rules and logic to perform fault diagnosis. Rule-based models often use expert systems, combined with the knowledge and reasoning ability of domain experts, to perform fault diagnosis. This type of model usually includes a rule engine and a knowledge base, which can judge the system status and possible faults based on the experience of experts. It has good interpretability, but is easily affected by various factors such as the quality of expert knowledge and environmental factors. Among them, the rules can be pre-defined based on domain expertise, such as the experience of process engineers, and are usually expressed in the form of "if-then", which is used to judge whether there is a fault based on the observed status and events.
[0029] Fault diagnosis models based on deep learning are generally trained through large-scale data, which gives such models strong generalization capabilities. Such models can automatically learn complex features and relationships from large amounts of data. They are suitable for processing nonlinear and highly complex systems and can adapt to changing industrial environments, including system evolution and changes. They may face constantly changing working conditions and operating conditions, which is particularly critical for long-term industrial processes and equipment.
[0030] The fault diagnosis model in the disclosed embodiment is constructed based on Monte Carlo tree search and large language model. The fault diagnosis model can update the diagnosis model and the search tree of the diagnosis model according to the actual operating status of the equipment and the diagnosis results, and guide the large language model to perform self-learning, so that the fault diagnosis model can gradually improve its performance and adapt to changing environments and problems. This combination can make up for the shortcomings of traditional deep learning models in terms of interpretability, making the model output easier to understand and verify.
[0031] Among them, the integration of prior knowledge of large language models brings a new level to the fault diagnosis model. By integrating domain expertise into the model, we can better understand the working principle of the system and thus identify potential failure modes more accurately.
[0032] In the disclosed embodiment, the state information of the device is input into the fault diagnosis model. The fault diagnosis model uses Monte Carlo tree search to determine the operation logic of the device in the corresponding state according to the state information. The Monte Carlo tree search can quickly determine the decision path of the diagnosis process, thereby shortening the time; and the operation logic is executed through a large language model and a corresponding diagnosis result is generated. Then, the fault type of the device is determined according to the diagnosis result, and the corresponding state change is generated through the large language model and outputs natural language that is easy for users to understand, thereby improving the interpretability of the diagnosis result. At the same time, by combining the advantages of Monte Carlo tree search and large prediction model, the potential failure mode of the device can be quickly and accurately identified while having good interpretability, thereby accurately diagnosing the fault of the device in actual operation.
[0033] In order to facilitate those skilled in the art to better understand the equipment operation fault diagnosis method provided by the present disclosure, the method is described in detail below.
[0034] In a feasible implementation manner, the diagnosis process of the fault diagnosis model includes: Determining a target decision path through the Monte Carlo tree search according to the state information; The target decision path is executed through the large language model, and a diagnosis result is generated.
[0035] It is worth noting that the target decision path includes the optimal decision path that may lead to equipment failure, and the diagnosis result includes the fault type and fault cause corresponding to the target decision path.
[0036] For example, according to the status information, a target decision path corresponding to the device status information is called through Monte Carlo tree search, and the target decision path is simulated through a large language model, and the fault type and fault cause corresponding to the target decision path are generated.
[0037] The state interpretation mechanism is introduced in the embodiment of the present disclosure, and an explanatory description of the corresponding evaluation results and device state changes is generated through a large language model, which helps users understand the system's diagnostic basis for device failures.
[0038] In a feasible implementation manner, the process of constructing the fault diagnosis model includes: Constructing a search space corresponding to the device operation failure problem, wherein each node in the search space represents a state of the device, and each edge in the search space represents a transition action from one state to another state; According to the search space and the historical state information of the device, simulating an operation logic tree of the device in a preset state by Monte Carlo tree search, the operation logic tree including at least one of the operation state, fault type and operation process of the device; The large language model is used to simulate each operation logic in the operation logic tree and generate corresponding simulation diagnosis results to obtain a fault diagnosis model.
[0039] It is worth noting that the search space may be a problem space including all possible fault problems of the device. The operation logic tree may also include possible fault modes and causes of the device. The transfer action may be a different setting or detection method of the device operation.
[0040] It should be understood that the fault diagnosis model can understand the context of equipment operation, diagnose equipment operation, and provide a more accurate explanation of abnormal situations that occur in the simulation. It can better adapt and respond to complex and changing fault scenarios, provide real-time decision support for equipment operation and maintenance personnel in industrial production processes, and provide a new way to help them quickly and accurately diagnose and solve equipment failures.
[0041] It is worth noting that the data set of the text description of the equipment status is determined based on the historical status information, and the data set is preprocessed to improve the quality of the text description of the data set. The possible decision paths of the equipment in the preset state are searched iteratively through the Monte Carlo tree search to simulate the operation logic tree of the equipment in the preset state, including but not limited to simulating the operation of the equipment, detecting faults, and executing operations, etc. The knowledge graph can be used to cover the elements of each stage of equipment production and management. In each Monte Carlo tree search iteration process, the large language model is used to simulate the operation of the equipment in the current state, and the large language model is used to generate the results of the simulation operation, including possible faults and other key information. Finally, the simulation results are evaluated, and an appropriate score is assigned to each state to indicate its value in solving the problem, thereby determining the optimal fault decision path as the target decision path and obtaining the fault diagnosis model.
[0042] For example, the possible operating states of the device and the different possible causes of failure are defined as states, which may include the device's sensor data, operation logs, and possible failure modes and causes. Possible actions for transitioning between states are defined as transition actions, which may be different settings or detection methods for device operations. Based on the defined states and transition actions, the device operation failure problem is modeled as a search space, where each node in the search space represents a possible state or decision of the device, and each edge in the search space represents a transition action from one state to another.
[0043] Determine the data set T=[t1, t2, …, t n ], T∈R N·M , the data set includes the number of samples N of the clustered text descriptions of the device status, and the number of text description data M, and the data set is cleaned and denoised. The possible decision paths of the device in the preset state are iteratively searched in the search space through the Monte Carlo tree search, including but not limited to simulating the operation of the device, detecting faults, and executing operations. In each Monte Carlo tree search iteration, the large language model is used to simulate the operation of the device in the current state, and the large language model is used to generate the results of the simulation operation, including possible faults and other key information. Finally, the simulation results are evaluated and an appropriate score is assigned to each state to indicate its value in solving the problem. When a certain number of iterations or other termination conditions are reached, the optimal decision path with the largest score is selected as the final fault diagnosis decision to obtain the fault diagnosis model.
[0044] In the disclosed embodiment, an efficient search is performed in a large-scale decision space through Monte Carlo tree search to find multiple decision paths that are most likely to cause equipment failure, and each decision path is simulated through a large language model, and the fault type and probability corresponding to each decision path are generated. By utilizing the prior knowledge in the large language model, the Monte Carlo tree search can be made more targeted, thereby guiding the expansion and selection of the next search of the Monte Carlo tree. The obtained fault diagnosis model can better understand the working principle of the equipment, thereby more accurately identifying the potential failure mode of the equipment.
[0045] In a feasible implementation manner, simulating the operation logic tree of the device in a preset state through Monte Carlo tree search according to the search space and the historical state information of the device includes: Constructing an initial operation logic tree according to the historical status information of the device; According to the search space, the decision path of the Monte Carlo tree search is iteratively expanded to update the initial running logic tree until the number of iterations reaches a threshold or a preset iteration stop condition is met, thereby obtaining the running logic tree.
[0046] It is worth noting that the number threshold or preset iteration stop condition can be preset according to the accuracy requirement of the diagnosis result, or can be preset by the user according to actual diagnosis needs, and the present disclosure does not limit this.
[0047] In the disclosed embodiment, the search tree is expanded by using the Monte Carlo tree search algorithm, possible decision paths are explored in depth, and the results of each decision path are scored by using a large language model, thereby eliminating decision paths with lower probability, thereby compressing the state space and significantly reducing the complexity of the search.
[0048] In a feasible implementation manner, the iteratively expanding the decision path of the Monte Carlo tree search according to the search space includes: For each iteration process, determining a target node from the initial running logic tree by selecting an algorithm; According to the target node, randomly select a new node from the search space, and add the new node to the initial running logic tree; According to the new node, random simulation is performed through a large language model until a preset depth is reached to obtain a simulated diagnosis result; According to the simulation diagnosis result, the node information on the path of the initial operation logic tree during the current expansion process is reversely updated, and the node information includes the reward value and the number of visits of the node.
[0049] In the disclosed embodiment, a node is selected from the initial running logic tree through a selection algorithm, and a new node is randomly selected from the search space according to the target node to guide the Monte Carlo tree search to actually perform the next step, and the new node is added to the initial running logic tree to embed the state, thereby reducing the number of states that need to be considered during the Monte Carlo tree search, thereby improving the diagnostic efficiency.
[0050] In a feasible implementation manner, determining the target node from the initial running logic tree by selecting an algorithm includes: Starting from the root node in the initial operation logic tree, for each node, determine the average reward, number of selections, and total number of visits corresponding to the node; Substituting the preset exploration parameter, the average reward, the number of selections and the total number of visits into the selection calculation formula to obtain the probability of the corresponding node; Recursively select child nodes until a leaf node in the initial running logic tree is reached, and the node with the highest probability is used as the target node; Wherein, the selection calculation formula includes: , UCT(s, a) represents the probability of the transition action a corresponding to the node in state s, Q(s, a) represents the average reward of the transition action a in state s, N(s, a) represents the number of times the transition action a is selected in state s, N(s) represents the number of times state s is visited, and C represents the exploration parameter.
[0051] It is worth noting that the Monte Carlo tree search process mainly includes four stages: selection, expansion, simulation, and back propagation. In the selection stage, the target node is selected from the initial running logic tree by selecting the calculation formula; in the expansion stage, when the selected node (state-action pair) has unexplored child nodes, an unexplored action is selected for expansion; in the simulation stage, the operation results and the decision-making process are simulated through the large language model until the preset iteration stop condition is reached; in the back propagation stage, the node information on the selected path is updated, including the reward value and the number of visits.
[0052] In the disclosed embodiments, by simulating the operation results and simulating the decision by using a large language model, the experimental cost can be reduced and the search efficiency can be improved.
[0053] For example, Figure 2 As shown, the construction of the fault diagnosis model and the diagnosis process in the embodiment of the present disclosure may include the following steps: A. The user uses natural language to raise relevant diagnostic requirements. The model combines the prompt examples provided in the text knowledge base and uses the large language model to make the first step of information update. The model also uses the results of the large language model simulation in the Monte Carlo tree search algorithm to evaluate the simulation results and assign an appropriate score to each state to indicate its value in solving the problem.
[0054] B. Based on the defined actions and states, the search tree is continuously expanded to deeply explore possible decision paths.
[0055] Use the selection algorithm to select an action from the tree, verify the execution in the actual environment, and synchronize the current observations to update and return information.
[0056] C. The large language model makes the next simulation run and evaluation based on the prompt examples, natural language, current observations and historical information.
[0057] D. Repeat steps B and C above, gradually optimize the decision strategy through simulation and tree search, and select the optimal decision path as the final fault diagnosis decision when a certain number of iterations or the preset iteration stop condition is reached.
[0058] E. Finally, combined with the production capacity of the large language model, the diagnostic explanations and suggestions are output in natural language, making it easier for users to understand.
[0059] It is worth noting that if Figure 2As shown, the accuracy Acc and task success rate Tsr of the Llama2 method, the FLlama2 (Finetuned Llama2) method, the traditional Monte Carlo method (Monte Carlo Tree Search, MCTS) and the fault diagnosis model Smart-MCTS provided in the present invention can be seen in the following Table 1.
[0060] Table 1
[0061] According to Table 1, the fault diagnosis accuracy of Smart-MCTS is 94.5%, and the task success rate is 92.6%, which is 5.9% and 9.3% higher than the traditional Monte Carlo method MCTS, 12.2% and 21% higher than the fine-tuned Llama2 method, and 23.3% and 28% higher than the Llama2 method.
[0062] In summary, the fault diagnosis model provided by the present disclosure can be trained on large-scale data, has strong generalization capabilities, and can automatically learn complex features and relationships from large amounts of data. It is suitable for processing nonlinear and highly complex systems, so that the fault diagnosis model can adapt to changing industrial environments, including the evolution and changes of the system, which may face changing working conditions and operating conditions, which is particularly critical for long-term industrial processes and equipment. And the integration of prior knowledge of large language models brings a new level to the fault diagnosis model. By integrating domain expertise into the fault diagnosis model, the fault diagnosis model can better understand the working principle of the system, thereby more accurately identifying potential fault modes. It can make up for the shortcomings of traditional deep learning models in interpretability and make model output easier.
[0063] Based on the same inventive concept, Figure 4 As shown, the present disclosure also provides a device operation fault diagnosis apparatus, including an acquisition module 401 and a control module 402.
[0064] Wherein, the acquisition module 401 is configured to acquire the status information of the device; The execution module 402 is configured to input the state information into a fault diagnosis model to obtain a diagnosis result output by the fault diagnosis model, wherein the diagnosis result includes the probability of the device corresponding to different fault types, and the fault diagnosis model is constructed based on Monte Carlo tree search and a large language model.
[0065] In the disclosed embodiment, the state information of the device is input into the fault diagnosis model. The fault diagnosis model uses Monte Carlo tree search to determine the operation logic of the device in the corresponding state according to the state information. The Monte Carlo tree search can quickly determine the decision path of the diagnosis process, thereby shortening the time; and the operation logic is executed through a large language model and a corresponding diagnosis result is generated. Then, the fault type of the device is determined according to the diagnosis result, and the corresponding state change is generated through the large language model and outputs natural language that is easy for users to understand, thereby improving the interpretability of the diagnosis result. At the same time, by combining the advantages of Monte Carlo tree search and large prediction model, the potential failure mode of the device can be quickly and accurately identified while having good interpretability, thereby accurately diagnosing the fault of the device in actual operation.
[0066] In a feasible implementation manner, it further includes a construction module, wherein the construction module is configured to construct a search space corresponding to the device operation failure problem, wherein each node in the search space represents a state of the device, and each edge in the search space represents a transition action from one state to another state; According to the search space and the historical state information of the device, simulating an operation logic tree of the device in a preset state by Monte Carlo tree search, the operation logic tree including at least one of the operation state, fault type and operation process of the device; The large language model is used to simulate each operation logic in the operation logic tree and generate corresponding simulation diagnosis results to obtain a fault diagnosis model.
[0067] In a feasible implementation manner, the construction module is configured to construct an initial operation logic tree according to the historical status information of the device; According to the search space, the decision path of the Monte Carlo tree search is iteratively expanded to update the initial running logic tree until the number of iterations reaches a threshold or a preset iteration stop condition is met, thereby obtaining the running logic tree.
[0068] In a feasible implementation, the construction module is configured to determine the target node from the initial running logic tree by selecting an algorithm for each iteration process; According to the target node, randomly select a new node from the search space, and add the new node to the initial running logic tree; According to the new node, random simulation is performed through a large language model until a preset depth is reached to obtain a simulated diagnosis result; According to the simulation diagnosis result, the node information on the path of the initial operation logic tree during the current expansion process is reversely updated, and the node information includes the reward value and the number of visits of the node.
[0069] In a feasible implementation manner, the construction module is configured to start from the root node in the initial operation logic tree, and for each node, determine the average reward, the number of selections, and the total number of visits corresponding to the node; Substituting the preset exploration parameter, the average reward, the number of selections and the total number of visits into the selection calculation formula to obtain the probability of the corresponding node; Recursively select child nodes until a leaf node in the initial running logic tree is reached, and the node with the highest probability is used as the target node; Wherein, the selection calculation formula includes: , UCT(s, a) represents the probability of the transition action a corresponding to the node in state s, Q(s, a) represents the average reward of the transition action a in state s, N(s, a) represents the number of times the transition action a is selected in state s, N(s) represents the number of times state s is visited, and C represents the exploration parameter.
[0070] In a feasible implementation manner, the fault diagnosis model is configured to determine a target decision path through the Monte Carlo tree search according to the state information; The target decision path is executed through the large language model, and a diagnosis result is generated.
[0071] Regarding the device operation fault diagnosis apparatus in the above-mentioned embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.
[0072] Based on the same inventive concept, the present disclosure also provides an electronic device, including: a memory having a computer program stored thereon; A processor is used to execute the computer program in the memory to implement the above-mentioned device operation fault diagnosis method.
[0073] In the disclosed embodiment, the state information of the device is input into the fault diagnosis model. The fault diagnosis model uses Monte Carlo tree search to determine the operation logic of the device in the corresponding state according to the state information. The Monte Carlo tree search can quickly determine the decision path of the diagnosis process, thereby shortening the time; and the operation logic is executed through a large language model and a corresponding diagnosis result is generated. Then, the fault type of the device is determined according to the diagnosis result, and the corresponding state change is generated through the large language model and outputs natural language that is easy for users to understand, thereby improving the interpretability of the diagnosis result. At the same time, by combining the advantages of Monte Carlo tree search and large prediction model, the potential failure mode of the device can be quickly and accurately identified while having good interpretability, thereby accurately diagnosing the fault of the device in actual operation.
[0074] Figure 5 FIG. 5 is a block diagram of an electronic device 500 according to an exemplary embodiment. Figure 5 As shown, the electronic device 500 may include: a processor 501 , a memory 502 . The electronic device 500 may also include one or more of a multimedia component 503 , an input / output (I / O) interface 504 , and a communication component 505 .
[0075] The processor 501 is used to control the overall operation of the electronic device 500 to complete all or part of the steps in the above-mentioned device operation fault diagnosis method. The memory 502 is used to store various types of data to support the operation of the electronic device 500. These data may include instructions for any application or method used to operate on the electronic device 500, and application-related data, such as contact data, sent and received messages, pictures, audio, video, etc. The memory 502 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read-Only Memory, referred to as EPROM), programmable read-only memory (Programmable Read-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic memory, flash memory, magnetic disk or optical disk. The multimedia component 503 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signal may be further stored in the memory 502 or sent through the communication component 505. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 504 provides an interface between the processor 501 and other interface modules, and the other interface modules may be keyboards, mice, buttons, etc. These buttons may be virtual buttons or physical buttons. The communication component 505 is used for wired or wireless communication between the electronic device 500 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, 4G, NB-IOT, eMTC, or other 5G, etc., or a combination of one or more of them, is not limited here. Therefore, the corresponding communication component 505 may include: Wi-Fi module, Bluetooth module, NFC module, etc.
[0076] In an exemplary embodiment, the electronic device 500 can be implemented by one or more application specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), controllers, microcontrollers, microprocessors or other electronic components to execute the above-mentioned device operation fault diagnosis method.
[0077] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, and when the program instructions are executed by a processor, the steps of the above-mentioned device operation fault diagnosis method are implemented. For example, the computer-readable storage medium can be the above-mentioned memory 502 including program instructions, and the above-mentioned program instructions can be executed by the processor 501 of the electronic device 500 to complete the above-mentioned device operation fault diagnosis method.
[0078] In another exemplary embodiment, a computer program product is further provided. The computer program product includes a computer program executable by a programmable device, and the computer program has a code portion for executing the above-mentioned device operation fault diagnosis method when executed by the programmable device.
[0079] The preferred embodiments of the present disclosure are described in detail above in conjunction with the accompanying drawings; however, the present disclosure is not limited to the specific details in the above embodiments. Within the technical concept of the present disclosure, a variety of simple modifications can be made to the technical solution of the present disclosure, and these simple modifications all fall within the protection scope of the present disclosure.
[0080] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present disclosure will not further describe various possible combinations.
[0081] In addition, various embodiments of the present disclosure may be arbitrarily combined, and as long as they do not violate the concept of the present disclosure, they should also be regarded as the contents disclosed by the present disclosure.
Claims
1. A method for diagnosing equipment operation faults, characterized in that: include: Get device status information; The state information is input into a fault diagnosis model to obtain a diagnosis result output by the fault diagnosis model, wherein the diagnosis result includes the probability of the device corresponding to different fault types, and the fault diagnosis model is constructed based on Monte Carlo tree search and a large language model.
2. The equipment operation fault diagnosis method according to claim 1, characterized in that: The construction process of the fault diagnosis model includes: Constructing a search space corresponding to the device operation failure problem, wherein each node in the search space represents a state of the device, and each edge in the search space represents a transition action from one state to another state; According to the search space and the historical state information of the device, simulating an operation logic tree of the device in a preset state by Monte Carlo tree search, the operation logic tree including at least one of the operation state, fault type and operation process of the device; The large language model is used to simulate each operation logic in the operation logic tree and generate corresponding simulation diagnosis results to obtain a fault diagnosis model.
3. The equipment operation fault diagnosis method according to claim 2, characterized in that: The step of simulating the operation logic tree of the device in a preset state by Monte Carlo tree search according to the search space and the historical state information of the device includes: Constructing an initial operation logic tree according to the historical status information of the device; According to the search space, the decision path of the Monte Carlo tree search is iteratively expanded to update the initial running logic tree until the number of iterations reaches a threshold or a preset iteration stop condition is met, thereby obtaining the running logic tree.
4. The equipment operation fault diagnosis method according to claim 3, characterized in that: The iteratively expanding the decision path of the Monte Carlo tree search according to the search space comprises: For each iteration process, determining a target node from the initial running logic tree by selecting an algorithm; According to the target node, randomly select a new node from the search space, and add the new node to the initial running logic tree; According to the new node, random simulation is performed through a large language model until a preset depth is reached to obtain a simulated diagnosis result; According to the simulation diagnosis result, the node information on the path of the initial operation logic tree during the current expansion process is reversely updated, and the node information includes the reward value and the number of visits of the node.
5. The equipment operation fault diagnosis method according to claim 4, characterized in that: Determining the target node from the initial running logic tree by selecting an algorithm includes: Starting from the root node in the initial operation logic tree, for each node, determine the average reward, number of selections, and total number of visits corresponding to the node; Substituting the preset exploration parameter, the average reward, the number of selections and the total number of visits into the selection calculation formula to obtain the probability of the corresponding node; Recursively select child nodes until a leaf node in the initial running logic tree is reached, and the node with the highest probability is used as the target node; Wherein, the selection calculation formula includes: , UCT(s, a) represents the probability of the transition action a corresponding to the node in state s, Q(s, a) represents the average reward of the transition action a in state s, N(s, a) represents the number of times the transition action a is selected in state s, N(s) represents the number of times state s is visited, and C represents the exploration parameter.
6. The equipment operation fault diagnosis method according to any one of claims 1 to 5, characterized in that: The diagnostic process of the fault diagnosis model includes: Determining a target decision path through the Monte Carlo tree search according to the state information; The target decision path is executed through the large language model, and a diagnosis result is generated.
7. A device for diagnosing equipment operation faults, characterized in that: include: An acquisition module, configured to acquire status information of a device; The execution module is configured to input the state information into a fault diagnosis model to obtain a diagnosis result output by the fault diagnosis model, wherein the diagnosis result includes the probability of the device corresponding to different fault types, and the fault diagnosis model is constructed based on Monte Carlo tree search and a large language model.
8. The equipment operation fault diagnosis device according to claim 7, characterized in that: It also includes a construction module, wherein the construction module is configured to construct a search space corresponding to the device operation failure problem, each node in the search space represents a state of the device, and each edge in the search space represents a transition action from one state to another state; According to the search space and the historical state information of the device, simulating an operation logic tree of the device in a preset state by Monte Carlo tree search, the operation logic tree including at least one of the operation state, fault type and operation process of the device; The large language model is used to simulate each operation logic in the operation logic tree and generate corresponding simulation diagnosis results to obtain a fault diagnosis model.
9. The equipment operation fault diagnosis device according to claim 8, characterized in that: The building module is configured to build an initial operation logic tree according to the historical status information of the device; According to the search space, the decision path of the Monte Carlo tree search is iteratively expanded to update the initial running logic tree until the number of iterations reaches a threshold or a preset iteration stop condition is met, thereby obtaining the running logic tree.
10. The equipment operation fault diagnosis device according to claim 9, characterized in that: The building module is configured to determine the target node from the initial running logic tree by selecting an algorithm for each iteration process; According to the target node, randomly select a new node from the search space, and add the new node to the initial running logic tree; According to the new node, random simulation is performed through a large language model until a preset depth is reached to obtain a simulated diagnosis result; According to the simulation diagnosis result, the node information on the path of the initial operation logic tree during the current expansion process is reversely updated, and the node information includes the reward value and the number of visits of the node.
11. The equipment operation fault diagnosis device according to claim 10, characterized in that: The construction module is configured to start from the root node in the initial operation logic tree, and for each node, determine the average reward, the number of times selected, and the total number of times visited corresponding to the node; Substituting the preset exploration parameter, the average reward, the number of selections and the total number of visits into the selection calculation formula to obtain the probability of the corresponding node; Recursively select child nodes until a leaf node in the initial running logic tree is reached, and the node with the highest probability is used as the target node; Wherein, the selection calculation formula includes: , UCT(s, a) represents the probability of the transition action a corresponding to the node in state s, Q(s, a) represents the average reward of the transition action a in state s, N(s, a) represents the number of times the transition action a is selected in state s, N(s) represents the number of times state s is visited, and C represents the exploration parameter.
12. The equipment operation fault diagnosis device according to any one of claims 7 to 11, characterized in that: The fault diagnosis model is configured to determine a target decision path through the Monte Carlo tree search according to the state information; The target decision path is executed through the large language model, and a diagnosis result is generated.
13. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.
14. An electronic device, characterized in that: include: a memory having a computer program stored thereon; A processor, configured to execute the computer program in the memory to implement the method according to any one of claims 1 to 6.
15. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.
Citation Information
Cited By
Intelligent fault diagnosis method for numerical control machine tool
CN119472509A
Intelligent fault diagnosis method for numerical control machine tool
CN119472509B
