Simulation processing method and device based on large language model agent

Abstract

The embodiment of the invention provides a simulation processing method and device based on a large language model agent, and relates to the technical field of industrial simulation, and the method comprises the steps: obtaining the demand information inputted by a user, calling a large language model connected to a simulation process arrangement system through a model interface to carry out the intention recognition of the demand information, and obtaining the intention recognition result; obtaining an intention type corresponding to the demand information; if the intention type belongs to a simulation operation class, calling a simulation process management tool to perform simulation parameter setting according to demand information stored in a long and short memory module to obtain a corresponding simulation task, and executing the simulation task to generate a corresponding simulation result; according to the method, the file type which is input by the user and aims at the simulation result is obtained, the display template corresponding to the file type is determined, the simulation result is input into the display template, and the data file corresponding to the simulation result is obtained, so that the efficiency and accuracy of simulation process arrangement are improved, and the error rate is reduced.

Description

Technical Field

[0001] The present invention relates to the field of industrial simulation technology, and in particular to a simulation processing method based on a large language model agent, a simulation processing device based on a large language model agent, an electronic device, and a computer-readable storage medium. Background Art

[0002] In the field of industrial simulation, process orchestration is the core link to achieve complex simulation tasks. When orchestrating the corresponding simulation process, users often need to manually write scripts or set them through complex interfaces, which not only consumes a lot of time and energy of users, but also is difficult to cope with dynamic and changing requirements. For example, when users need to frequently modify the relevant content involved in the simulation task, they often need to complete it manually, increasing the operation complexity, resulting in low execution efficiency and easy errors of the simulation task. In addition, the result display method in the simulation process is too single and inflexible, and cannot dynamically adjust the display method according to user needs, restricting the application value of the simulation results. Summary of the Invention

[0003] Embodiments of the present invention provide a simulation processing method, device, electronic device, and computer-readable storage medium based on a large language model agent to solve or partially solve the problems of low efficiency and easy errors in the process of orchestrating the simulation process, and too single and inflexible result display.

[0004] Embodiments of the present invention disclose a simulation processing method based on a large language model agent, which is applied to a simulation process orchestration system. The simulation process orchestration system at least includes a model interface, a long short-term memory module, and a simulation process management tool. The method includes:

[0005] Obtain the requirement information input by the user, and call the large language model accessed to the simulation process orchestration system through the model interface to perform intent recognition on the requirement information to obtain the intent type corresponding to the requirement information;

[0006] If the intent type belongs to the simulation operation category, call the simulation process management tool to perform simulation parameter setting according to the requirement information stored in the long short-term memory module, obtain the corresponding simulation task, and execute the simulation task to generate the corresponding simulation result;

[0007] Obtain the file type of the user input for the simulation result, determine the display template corresponding to the file type, and input the simulation result into the display template to obtain the data file corresponding to the simulation result.

[0008] In some feasible implementation manners, the simulation process management tool at least includes tool nodes and a flow chart; wherein, specific operations are encapsulated in the tool nodes, and each of the tool nodes is used to execute one of the operations of setting parameters, running a simulation, and generating a report. The flow chart is a tool for defining and managing a state machine, and the flow chart is composed of a plurality of the work nodes and the conversion logic between the work nodes.

[0009] In some feasible implementation manners, when the simulation process management tool is called to set simulation parameters according to the requirement information stored in the long short-term memory module to obtain a corresponding simulation task, it includes:

[0010] Extracting the requirement information from the long short-term memory module;

[0011] Extracting target key information from the requirement information, where the target key information is information related to simulation testing;

[0012] Calling the tool node to set simulation parameters matching the target key information;

[0013] Calling the flow chart to set a simulation process corresponding to the simulation parameters and generate a test task corresponding to the simulation process.

[0014] In some feasible implementation manners, the simulation process choreography system further includes simulation tool functions. The simulation tool functions at least include a parameter setting function, a simulation execution function, and a result generation function. When the simulation task is executed to generate a corresponding simulation result, it includes:

[0015] Calling the parameter setting function to pass the simulation parameters into the simulation software;

[0016] Calling the simulation execution function to control the simulation software to execute the simulation process according to the simulation parameters;

[0017] In response to the completion of the execution of the simulation process, calling the result generation function to generate a simulation result corresponding to the simulation process.

[0018] In some feasible implementation manners, the simulation process choreography system further includes a simulation manager; wherein, the simulation manager is used for the setting and management of simulation types, performance, architectures, and models, and supports the setting of simulation parameters in multiple rounds of conversations. The method further includes:

[0019] Obtaining an adjustment instruction for the test parameters input by the user;

[0020] Calling the large language model through the model interface to identify the adjustment instruction to obtain a simulation adjustment parameter corresponding to the adjustment instruction;

[0021] Adjust the simulation parameters to target simulation parameters that match the simulation adjustment parameters.

[0022] In some feasible implementation manners, the simulation parameters at least include a first simulation type, a first performance mode, and first model parameters, and the simulation adjustment parameters at least include a type adjustment parameter, a performance adjustment parameter, and a model adjustment parameter. Adjusting the simulation parameters to target simulation parameters that match the simulation adjustment parameters includes:

[0023] Adjust the first simulation type to a second simulation type that matches the type adjustment parameter;

[0024] Adjust the first performance mode to a second performance mode that matches the performance adjustment parameter;

[0025] Adjust the first model parameters to second model parameters that match the model adjustment parameter.

[0026] In some feasible implementation manners, it further includes:

[0027] If the intention type belongs to the knowledge acquisition category, obtain a knowledge Q&A result corresponding to the requirement information.

[0028] In some feasible implementation manners, obtaining a knowledge Q&A result corresponding to the requirement information includes:

[0029] Call the large language model through the model interface to generate a knowledge Q&A result corresponding to the requirement information; and / or,

[0030] Call a preset knowledge extraction function to extract a knowledge Q&A result corresponding to the requirement information from a preset model construction document; and / or,

[0031] Call a preset experience extraction function to extract a knowledge Q&A result corresponding to the requirement information from a preset experience document.

[0032] An embodiment of the present invention also discloses a simulation processing device based on a large language model agent, which is applied to a simulation process orchestration system. The simulation process orchestration system at least includes a model interface, a long short-term memory module, and a simulation process management tool. The device includes:

[0033] An intention recognition module, configured to obtain requirement information input by a user, and call the large language model accessed to the simulation process orchestration system through the model interface to perform intention recognition on the requirement information, and obtain an intention type corresponding to the requirement information;

[0034] A simulation processing module, which is configured to, if the intention type belongs to the simulation operation category, call the simulation process management tool to set simulation parameters according to the requirement information stored in the long short-term memory module, obtain a corresponding simulation task, and execute the simulation task to generate a corresponding simulation result;

[0035] A result presentation module, which is configured to obtain the file type of the user input for the simulation result, determine a display template corresponding to the file type, and input the simulation result into the display template to obtain a data file corresponding to the simulation result.

[0036] In some feasible implementation manners, the simulation process management tool at least includes tool nodes and a flow chart; wherein, specific operations are encapsulated in the tool nodes, and each of the tool nodes is used to execute one of the operations of setting parameters, running a simulation, and generating a report. The flow chart is a tool for defining and managing a state machine, and the flow chart is composed of multiple working nodes and the conversion logic between the working nodes.

[0037] In some feasible implementation manners, the simulation processing module is specifically configured to:

[0038] Extract the requirement information from the long short-term memory module;

[0039] Extract target key information from the requirement information, where the target key information is information related to simulation testing;

[0040] Call the tool node to set simulation parameters matching the target key information;

[0041] Call the flow chart to set a simulation process corresponding to the simulation parameters and generate a test task corresponding to the simulation process.

[0042] In some feasible implementation manners, the simulation process orchestration system further includes simulation tool functions, and the simulation tool functions at least include a parameter setting function, a simulation execution function, and a result generation function. The simulation processing module is specifically configured to:

[0043] Call the parameter setting function to pass the simulation parameters into the simulation software;

[0044] Call the simulation execution function to control the simulation software to execute the simulation process according to the simulation parameters;

[0045] In response to the completion of the execution of the simulation process, call the result generation function to generate a simulation result corresponding to the simulation process.

[0046] In some feasible implementation manners, the simulation process orchestration system further includes a simulation manager; wherein, the simulation manager is used for setting and managing simulation types, performance, architectures, and models, and supports setting simulation parameters in multiple rounds of conversations. The device further includes:

[0047] An instruction acquisition module, configured to acquire an adjustment instruction for the test parameters input by the user;

[0048] An identification module, configured to call the large language model through the model interface to identify the adjustment instruction, and obtain a simulation adjustment parameter corresponding to the adjustment instruction;

[0049] An adjustment module, configured to adjust the simulation parameters to target simulation parameters matching the simulation adjustment parameters.

[0050] In some feasible implementation manners, the simulation parameters at least include a first simulation type, a first performance mode, and first model parameters, and the simulation adjustment parameters at least include a type adjustment parameter, a performance adjustment parameter, and a model adjustment parameter. The adjustment module is specifically configured to:

[0051] Adjust the first simulation type to a second simulation type matching the type adjustment parameter;

[0052] Adjust the first performance mode to a second performance mode matching the performance adjustment parameter;

[0053] Adjust the first model parameters to second model parameters matching the model adjustment parameter.

[0054] In some feasible implementation manners, it further includes:

[0055] A knowledge feedback module, configured to obtain a knowledge Q&A result corresponding to the requirement information if the intention type belongs to the knowledge acquisition category.

[0056] In some feasible implementation manners, the knowledge feedback module is specifically configured to:

[0057] Call the large language model through the model interface to generate a knowledge Q&A result corresponding to the requirement information; and / or,

[0058] Call a preset knowledge extraction function to extract a knowledge Q&A result corresponding to the requirement information from a preset model construction document; and / or,

[0059] Call a preset experience extraction function to extract a knowledge Q&A result corresponding to the requirement information from a preset experience document.

[0060] An embodiment of the present invention also discloses an electronic device, including a processor, a communication interface, a memory, and a communication bus. Among them, the processor, the communication interface, and the memory complete communication with each other through the communication bus;

[0061] The memory is used to store computer programs;

[0062] When the processor is used to execute the program stored on the memory, the method described in the embodiment of the present invention is implemented.

[0063] An embodiment of the present invention also discloses a computer-readable storage medium, on which instructions are stored. When executed by one or more processors, the processors are caused to execute the method described in the embodiment of the present invention.

[0064] The embodiment of the present invention has the following advantages:

[0065] In the embodiment of the present invention, by constructing a simulation process orchestration system for simulation processing, the system at least includes a model interface, a long short-term memory module, and a simulation process management tool. By obtaining the requirement information input by the user, the large language model accessed to the simulation process orchestration system is called through the model interface to perform intention recognition on the requirement information, and the intention type corresponding to the requirement information is obtained. Then, if the intention type belongs to the simulation operation category, the simulation process management tool is called to set simulation parameters according to the requirement information stored in the long short-term memory module, obtain the corresponding simulation task, and execute the simulation task to generate the corresponding simulation result. Then, the file type of the user input for the simulation result is obtained, the display template corresponding to the file type is determined, and the simulation result is input into the display template to obtain the data file corresponding to the simulation result. Thus, with the large language model as the "foundation" of the system, the user intention is deeply understood and automatically recognized, and the simulation parameter setting, simulation task execution, and simulation result output are completed according to the user intention, helping the user to complete complex simulation tasks more efficiently, improving the efficiency and accuracy of simulation process orchestration, reducing the error rate, and enriching the diversity and flexibility of simulation result display according to the corresponding display template, and effectively meeting the user's needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 is a flowchart of the steps of a simulation processing method based on a large language model agent provided in an embodiment of the present invention;

[0067] Figure 2 is a structural block diagram of a simulation processing device based on a large language model agent provided in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0068] To make the above objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0069] As an example, in the field of industrial simulation, process orchestration is the core link for implementing complex simulation tasks, which involves the automated management and optimization of multiple steps such as setting simulation parameters, the execution order of simulation tasks, and the processing and display of simulation results. However, with the increasing complexity and diversity of simulation tasks, related process orchestration methods have gradually revealed problems such as low efficiency, insufficient flexibility, and difficulty in meeting the diverse needs of users. For example, in the related process orchestration process, it relies on users to manually write scripts or configure through complex interfaces. This method not only consumes time and effort but also is difficult to cope with dynamically changing simulation requirements. Moreover, the related process orchestration process often requires users to complete manually, resulting in low simulation efficiency and easy errors. In addition, the result display method in the simulation process is too single and inflexible, unable to dynamically adjust the display method according to user needs, which limits the application value of simulation results.

[0070] In response to this, in the embodiments of the present invention, a corresponding simulation process orchestration system is constructed to meet different needs of users. The simulation process orchestration system includes a model interface, a long short-term memory module, and a simulation process management tool. The model interface is used to access the large language model, taking the large language model as the "foundation" of the simulation process orchestration system to enhance the in-depth understanding and automatic recognition ability of user questions, and at the same time improve the intelligent level of simulation process orchestration. At the same time, the long short-term memory module enhances the system's understanding ability of multi-round conversation content, enabling the simulation process orchestration system to dynamically plan the execution order of the simulation process by combining context information and historical data, etc., to ensure the efficient execution of simulation tasks. In addition, to meet the diverse needs of users, the simulation process orchestration system can also implement multiple simulation result display templates, and users can select the corresponding display method to present the simulation results according to their own needs. Under the condition of meeting user needs, the application value of simulation results is effectively improved.

[0071] Refer to Figure 1 , which shows the step flow chart of a simulation processing method based on a large language model agent provided in the embodiments of the present invention, applied to a simulation process orchestration system. The simulation process orchestration system at least includes a model interface, a long short-term memory module, and a simulation process management tool, and specifically may include the following steps:

[0072] Step 101, obtain the requirement information input by the user, and call the large language model accessed through the model interface in the simulation process orchestration system to perform intent recognition on the requirement information to obtain the intent type corresponding to the requirement information;

[0073] For a simulation process orchestration system (hereinafter referred to as the system), it may include a model interface, a long short-term memory module, and a simulation process management tool. Among them, the model interface is used to access a large language model and use the large language model as the "foundation" to build corresponding "agents". Through these "agents", the natural language input by the user is recognized for its intent, classified based on the result of the intent recognition, and then the corresponding process orchestration is executed according to the classification result; the long short-term memory module can be used to manage the context information of the conversation, which may include short-term memory and long-term memory, etc., respectively used to store the temporary information of the current conversation and the persistent information across conversations. For example, the short-term memory can be used to store the context information of the current conversation, including user input, system response, tool call results, etc., to maintain conversation coherence, support complex tasks, and reduce user repeated input, etc.; the long-term memory is used to store the persistent information across conversations, including user preferences, historical simulation results, common configurations, etc., to provide personalized services, historical records and backtracking, and cross-conversation task support, etc.; and the simulation process management tool can be the core component of the simulation process orchestration system, responsible for coordinating and controlling the execution of simulation tasks. It helps users complete simulation tasks efficiently by integrating functions such as parameter setting, simulation execution, and result generation.

[0074] It should be noted that for the system, it can be a system constructed with the one-dimensional simulation process of industrial simulation software as the application scenario, used for intelligent orchestration of simulation tasks in this scenario. Specifically, during the process of orchestrating the simulation process, the system can, according to the natural language input by the user, call the corresponding large language model through the model interface to recognize the intent of the user input content, and judge whether to execute the corresponding simulation process according to the intent recognition, so as to execute the corresponding operations through the simulation process management tool. At the same time, during the above process, the corresponding context information can be managed through the long short-term memory module to enhance the system's ability to understand the content of multi-round conversations.

[0075] In the embodiment of the present invention, for the demand information input by the user, the system can call the large language model accessed by the simulation process orchestration system through the model interface to recognize the intent of the demand information, so as to obtain the intent type corresponding to the demand information, and further execute the corresponding process orchestration according to different intent types.

[0076] Among them, the requirement information can be the natural language input by the user in the system to express the user's requirements. For example, "Set the temperature parameter to 300K and run the simulation", "Adjust the pressure parameter to 100Pa and run the simulation again", "Generate a simulation report including temperature distribution and pressure distribution in word format", "Set the performance mode to high-precision mode", "I want to know what the acceleration per 100 kilometers means? What factors is it related to in the a model?" etc. Based on the requirement information input by the user, the system can call the large language model for intent recognition and judge the intent type corresponding to the requirement information based on the result of the intent recognition.

[0077] Optionally, the intent type can be used to characterize the purpose or intent expressed behind the requirement information input by the user. By identifying the intent type of the user, the system can understand the user's requirements and take corresponding actions. Among them, the intent type can include simulation operation types and knowledge acquisition types, etc. The simulation operation type represents that the user hopes to execute specific simulation tasks, such as setting parameters, running simulations, and generating reports; the knowledge acquisition type represents that the user hopes to obtain information or knowledge, etc., such as querying simulation parameters, understanding simulation types, and obtaining help documents, etc. The present invention does not limit this.

[0078] In addition, during the process of the user "chatting" with the system, the long short-term memory module can record the context information in multiple rounds of conversations. For example, the short-term memory in the long short-term memory module can be implemented through the InputStoreMemory class of LangChain and can record the context information of the current conversation in real time. The short-term memory saves the user input and the model output through the save_context method and loads the context information through the load_memory_variables method; the long-term memory can be implemented through an external vector store (such as a vector database). Using the Embedding similarity calculation of the knowledge base, the agent can recall historical information related to the current task from the external storage. The introduction of long-term memory enables the agent to call historical data or knowledge in complex tasks, improving the ability to deeply understand and process tasks.

[0079] Optionally, the system can also, based on the long short-term memory module, combine prompt engineering to achieve in-depth understanding of the user's natural language input and dynamic scheduling of the simulation process. In some feasible implementation manners, the system can generate prompt information related to the user's Query, combine the context information of the short-term memory and the historical data of the long-term memory, dynamically plan the execution order of the simulation process, and at the same time combine the simulation process management tool to achieve dynamic scheduling of the simulation process. Furthermore, the system can call the corresponding tool functions according to the user's intent to complete simulation parameter setting, simulation execution, and result generation.

[0080] Step 102, if the intention type belongs to the simulation operation category, call the simulation process management tool to set simulation parameters according to the requirement information stored in the long short-term memory module, obtain the corresponding simulation task, and execute the simulation task to generate the corresponding simulation result;

[0081] When the system identifies that the intention type belongs to the simulation operation category, the system can call the simulation process management tool to set simulation parameters according to the requirement information stored in the long short-term memory module, obtain the corresponding simulation task, and then call the corresponding simulation software to execute the simulation task to generate the corresponding simulation result, thus realizing the full-process automation from intention recognition to simulation task execution, effectively helping users complete complex simulation tasks more efficiently, improving the efficiency and accuracy of simulation process scheduling, and reducing the error rate.

[0082] Optionally, the simulation process management tool includes at least a tool node and a flow chart; among them, the tool node encapsulates a specific operation, and each tool node is used to execute one of the operations of setting parameters, running the simulation, and generating a report. The flow chart is a tool for defining and managing the state machine, and the flow chart consists of multiple working nodes and the conversion logic between the working nodes. For example, the tool node can be ToolNode, and the flow chart can be StateGraph, etc. Through the two, tool call and process control can be realized, and dynamic simulation process scheduling is supported.

[0083] In the case where the system identifies that the user's intention belongs to the simulation operation category, the system can extract the requirement information from the long short-term memory module, and extract the target key information from the requirement information. The target key information is the information related to the simulation test. Then, call the tool node to set the simulation parameters matching the target key information, and then call the flow chart to set the simulation process corresponding to the simulation parameters, and generate the test task corresponding to the simulation process.

[0084] Among them, the simulation process orchestration system also includes simulation tool functions. The simulation tool functions include at least a parameter setting function, a simulation execution function, and a result generation function. Different simulation tool functions can be used to execute different operations. Optionally, after the system generates the corresponding simulation task, it can first call the parameter setting function to pass the simulation parameters into the simulation software, then call the simulation execution function to control the simulation software to execute the simulation process according to the simulation parameters, and then, in the case where the simulation process is executed, in response to the completion of the simulation process, call the result generation function to generate the simulation result corresponding to the simulation process.

[0085] In some examples, the system can extract corresponding requirement information and target key information through a long short-term memory module. For example, when the user's intention is recognized as a simulation operation type, the system can extract requirement information related to the current task from the long short-term memory module. For example, if the user previously entered: "Set the temperature parameter to 300K.", the system extracts the requirement information from the short-term memory: {"temperature": 300K}; Another example is that the system can extract target key information related to simulation testing from the requirement information, including simulation parameters, performance settings, and model configurations, etc. For example, when the requirement information is {"temperature": 300K, "pressure": 100Pa}, the target key information extracted by the system is {"temperature": 300K}, etc. Thus, using the large language model as the "base" of the system, it can deeply understand and automatically identify the user's intention, and complete the setting of simulation parameters, the execution of simulation tasks, and the output of simulation results according to the user's intention, helping the user to complete complex simulation tasks more efficiently, improving the efficiency and accuracy of simulation process orchestration, and reducing the error rate.

[0086] Furthermore, the tool node can be a basic unit in the simulation management tool, used to encapsulate specific operations. Each working node is responsible for executing a specific task, and the flowchart is a tool for defining and managing the state machine, consisting of multiple working nodes and the transition logic between the working nodes. The flowchart manages the execution order and conditional jumps of simulation tasks in the form of a state machine. For example, setting parameters, running simulations, and generating reports, etc. Then, based on the extracted target key information, the system can call the corresponding working nodes to set simulation parameters, execute the simulation process, and generate simulation reports, etc., and then define the execution order of each working node through the flowchart, so that the system can execute the corresponding working nodes in sequence according to the flowchart.

[0087] In one example, assume that the requirement information input by the user is: Set the temperature parameter to 300K and the pressure parameter to 100Pa. Then the system can extract the corresponding target key information from the requirement information: {"temperature": 300K, "pressure": 100Pa}. Then call the tool node to set the simulation parameters: Call the parameter setting tool node to set the temperature parameter to 300K and the pressure parameter to 100Pa; and call the flowchart to set the simulation process, and execute the tool nodes in sequence according to the flowchart: Set parameters - Run simulation - Generate report. Then, generate the following test tasks:

[0088] Verify whether the temperature parameter is set to 300K;

[0089] Verify whether the pressure parameter is set to 100Pa;

[0090] Verify whether the simulation results contain the expected temperature distribution and pressure distribution data;

[0091] Verify whether the simulation report contains charts of temperature distribution and pressure distribution.

[0092] Finally, execute the generated test tasks in sequence to ensure that each step of the simulation process is correctly completed, so as to complete the setting of simulation parameters, the execution of simulation tasks, and the output of simulation results according to the user's intention, helping the user to complete complex simulation tasks more efficiently, improving the efficiency and accuracy of simulation process orchestration, and reducing the error rate.

[0093] Optionally, for processes such as parameter setting and process generation during the simulation test, the system can also be generated through corresponding models. For example, corresponding simulation parameters can be selected through a parameter selection model, and corresponding simulation processes can be generated through corresponding simulation models, etc. The present invention does not limit this.

[0094] In some feasible implementation manners, the system may further include a simulation manager, which is used for the setting and management of simulation types, performance, architectures, and models, and supports the setting of simulation parameters in multiple rounds of conversations. Then, during the process of the system setting simulation parameters, executing the simulation process, and generating a simulation report through the simulation process management tool, if the user wants to adjust the simulation task, the user can input the corresponding adjustment instruction. Thus, the system can obtain the adjustment instruction for the test parameters input by the user, and identify the adjustment instruction through the model interface by invoking the large language model to obtain the simulation adjustment parameters corresponding to the adjustment instruction, and then adjust the simulation parameters to the target simulation parameters matching the simulation adjustment parameters. Therefore, the system can flexibly adjust the relevant content involved in the simulation orchestration process according to the dynamically input adjustment instruction of the user, without the need for the user to manually adjust, which not only improves the flexibility and efficiency of the simulation process orchestration, but also can reduce the workload of the user and reduce the errors caused by manual operations.

[0095] Among them, the simulation parameters at least include the first simulation type, the first performance mode, and the first model parameters, and the simulation adjustment parameters at least include the type adjustment parameters, the performance adjustment parameters, and the model adjustment parameters. Then, the process of parameter adjustment may include at least one of the following: adjusting the first simulation type to the second simulation type matching the type adjustment parameters; adjusting the first performance mode to the second performance mode matching the performance adjustment parameters; adjusting the first model parameters to the second model parameters matching the model adjustment parameters.

[0096] For example, the user can input instructions for adjusting the numerical values of simulation parameters (such as "adjust the temperature parameter to 350K", etc.), or input instructions for adjusting the simulation process (such as "optimize the grid resolution before running the simulation", etc.), or input instructions for adjusting the performance mode (such as "set the performance mode to high-precision mode", etc.), or input instructions for adjusting the simulation type (such as "set the simulation type to hydrodynamics", etc.), etc. The present invention is not limited thereto.

[0097] Step 103: Obtain the file type of the simulation result input by the user, determine the display template corresponding to the file type, and input the simulation result into the display template to obtain a data file corresponding to the simulation result.

[0098] After the system completes the corresponding simulation task through the simulation software, it can output the corresponding simulation result. During the process of outputting the simulation result, when the user wants to present the simulation result in a corresponding manner, the user can input the file type of the simulation result in the system, and then the system can determine the display template corresponding to the file type, and then input the simulation result into the display template to obtain a data file corresponding to the simulation result, thus enriching the diversity and flexibility of the simulation result display and effectively meeting the user's needs.

[0099] Optionally, the file type may include text type, chart type, image type, etc. For example, the user can choose ways such as charts, PDF format, Word documents, etc. to display the simulation result according to their own needs, thus enriching the diversity and flexibility of the simulation result display and effectively meeting the user's needs.

[0100] In the above process, the intention type is taken as an example of the simulation operation type for illustrative purposes. If the intention type belongs to the knowledge acquisition type, the system can obtain the knowledge Q&A result corresponding to the demand information. In some feasible implementation manners, the system can call a large language model through a model interface to generate the knowledge Q&A result corresponding to the demand information; and / or call a preset knowledge extraction function to extract the knowledge Q&A result corresponding to the demand information from a preset model construction document; and / or call a preset experience extraction function to extract the knowledge Q&A result corresponding to the demand information from a preset experience document.

[0101] Through the above process, the simulation process orchestration system can use the large language model as the basis of the intelligent agent to achieve in-depth understanding and automatic recognition of user intentions, and call the API (Application Programming Interface) of the simulation software according to the user intentions to complete the setting of simulation parameters, the execution of simulation tasks, and various template displays of simulation results, thereby significantly improving the intelligence level of the simulation software, helping users complete complex simulation tasks more efficiently, and promoting the further development of industrial simulation technology.

[0102] It should be noted that the embodiments of the present invention include but are not limited to the above examples. It can be understood that those skilled in the art can also make settings according to actual needs under the guidance of the ideas of the embodiments of the present invention, and the present invention does not limit this.

[0103] In the embodiments of the present invention, by constructing a simulation process orchestration system for simulation processing, the system at least includes a model interface, a long short-term memory module, and a simulation process management tool. By obtaining the requirement information input by the user, the large language model accessed through the model interface is called to perform intention recognition on the requirement information to obtain the intention type corresponding to the requirement information. Then, if the intention type belongs to the simulation operation category, the simulation process management tool is called to set the simulation parameters according to the requirement information stored in the long short-term memory module, obtain the corresponding simulation task, and execute the simulation task to generate the corresponding simulation result. Then, the file type of the user input for the simulation result is obtained, the display template corresponding to the file type is determined, and the simulation result is input into the display template to obtain the data file corresponding to the simulation result, so as to use the large language model as the "basis" of the system to deeply understand and automatically recognize the user intentions, and complete the setting of simulation parameters, the execution of simulation tasks, and the output of simulation results according to the user intentions, helping users complete complex simulation tasks more efficiently, improving the efficiency and accuracy of simulation process orchestration, reducing the error rate, and enriching the diversity and flexibility of simulation result displays according to the corresponding display templates, and being able to effectively meet the user needs.

[0104] To enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the following provides exemplary descriptions through corresponding examples:

[0105] The present invention provides an intelligent simulation process orchestration method based on a large language model (Large Language Model, LLM), which realizes in-depth understanding of user intentions and dynamic simulation process scheduling through the long short-term memory (Memory) module and prompt engineering (Prompt Engineering) in the LangChain framework. The following are the specific implementation manners of the present invention:

[0106] 1. System Architecture and Module Design

[0107] This invention uses the LangChain framework as the core technology and combines the large language models provided by corresponding AI (Artificial Intelligence) to build an agent system for realizing the automated orchestration of the simulation process. The system mainly consists of the following modules:

[0108] Large Language Model Interface: Uses the large language model provided by AI that supports natural language processing and intent recognition as the core inference engine.

[0109] Long Short-Term Memory Module: Implements short-term and long-term memory through the BaseMemory interface of LangChain, which is used to store and manage the dialogue context information.

[0110] Tool Nodes and Flowcharts: Realize tool calls and process control through LangChain's ToolNode (a node used to call specific tools or operations, tool node) and StateGraph (a tool used to define and manage state machines, state graph), and support dynamic simulation process scheduling. Through the state machine, the call order of tool nodes can be defined, and unconditional jumps are supported to achieve dynamic process scheduling.

[0111] Simulation Manager: Responsible for the setting and management of simulation types, performance, architecture, and models, and supports the setting of simulation parameters in multi-round dialogues.

[0112] Simulation Tool Functions: Include SetParameterValue (set simulation parameters), SimulateModel (used to execute simulation tasks), and generate_report (used to generate simulation result reports), which are used to call the simulation software API to achieve parameter setting, simulation execution, and result generation.

[0113] 2. Implementation of the Long Short-Term Memory Module

[0114] To enhance the agent's ability to understand the content of multi-round dialogues, this invention introduces a long short-term memory module, and the specific implementation is as follows:

[0115] Short-Term Memory: Implemented through the InputStoreMemory class of LangChain, which can record the context information of the current dialogue in real time. The short-term memory saves the user input and the model output through the save_context method and loads the context information through the load_memory_variables method.

[0116] Long-term Memory: Implemented through external vector storage (such as a vector database), using the Embedding similarity calculation of the knowledge base, the agent can recall historical information related to the current task from external storage. The introduction of long-term memory enables the agent to call historical data or knowledge in complex tasks, enhancing the in-depth understanding and processing ability of tasks.

[0117] 3. Prompt Engineering and Dynamic Process Scheduling

[0118] The present invention combines the Prompt Engineering of LangChain with the long short-term memory module to achieve in-depth understanding of the user's natural language input and dynamic simulation process scheduling. The specific implementation is as follows:

[0119] Prompt Generation: Through the Prompt engine of LangChain, prompt information related to the user's Query (user question) is generated. Combining the context information of short-term memory and the historical data of long-term memory, the agent can dynamically plan the execution order of the simulation process.

[0120] Flowchart and Tool Invocation: The dynamic scheduling of the simulation process is achieved through the StateGraph and ToolNode of LangChain. The flowchart defines multiple nodes (such as scheduling decision nodes and tool nodes, etc.), and the jump between nodes is realized through conditional edges and ordinary edges. The agent calls the corresponding tool functions according to the user's intention to complete the simulation parameter setting, simulation execution, and result generation.

[0121] 4. Implementation of the Simulation Manager

[0122] The SimulationManager is responsible for the setting and management of simulation types, performance, architecture, and models. The specific implementation is as follows:

[0123] Simulation Type and Performance Setting: The agent can dynamically set the simulation type (such as power performance simulation, economic performance simulation) and simulation performance (such as 0-100 km / h acceleration, maximum speed) according to the user input.

[0124] Vehicle Architecture and Model Setting: The agent can dynamically set the vehicle architecture and simulation models and other simulation types according to the user input.

[0125] Simulation Parameter Setting: The agent can dynamically set the simulation parameters according to the user input and achieve parameter setting through the Prompt Engineering and tool invocation of LangChain.

[0126] 5. Implementation of Simulation Tool Functions

[0127] The present invention defines multiple simulation tool functions for calling the simulation software API to implement simulation parameter setting, simulation execution, and result generation.

[0128] 6. User Intent Recognition and Process Classification

[0129] Intent Recognition: Through the large language model, the intelligent agent can judge whether the user's intention is "setting simulation parameters" or "acquiring knowledge" based on the user input. For example, when the user inputs "I want to conduct a dynamic performance simulation", the intelligent agent judges that the intention is "setting simulation parameters".

[0130] Process Classification: According to the result of intent recognition, the intelligent agent classifies the user's requirements into "setting simulation parameters" or "acquiring knowledge", and executes the corresponding process. For example, if the user's intention is "setting simulation parameters", the intelligent agent calls the simulation manager to set the simulation parameters; if the user's intention is "acquiring knowledge", the intelligent agent calls the large language model to generate knowledge Q&A results.

[0131] In the above process, with the large language model as the base of the intelligent agent, in-depth understanding and automatic recognition of the user's intention are achieved, and according to the user's intention, the API of the simulation software is called to complete the setting of simulation parameters, the execution of simulation tasks, and the display of various templates of simulation results. The implementation of the present invention can significantly improve the intelligent level of the simulation software, help users complete complex simulation tasks more efficiently, and promote the further development of industrial simulation technology.

[0132] It should be noted that for the method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the embodiments of the present invention are not limited by the described action sequence, because according to the embodiments of the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily essential for the embodiments of the present invention.

[0133] Refer to Figure 2 , which shows a structural block diagram of a simulation processing device based on a large language model intelligent agent provided in an embodiment of the present invention, applied to a simulation process orchestration system. The simulation process orchestration system at least includes a model interface, a long short-term memory module, and a simulation process management tool, and specifically may include the following modules:

[0134] An intent recognition module 201, configured to obtain the requirement information input by the user, and call the large language model accessed to the simulation process orchestration system through the model interface to perform intent recognition on the requirement information, and obtain an intent type corresponding to the requirement information;

[0135] The simulation processing module 202 is configured to, if the intention type belongs to the simulation operation category, call the simulation process management tool to set simulation parameters according to the requirement information stored in the long short-term memory module, obtain a corresponding simulation task, and execute the simulation task to generate a corresponding simulation result;

[0136] The result presentation module 203 is configured to obtain the file type of the user input for the simulation result, determine a display template corresponding to the file type, and input the simulation result into the display template to obtain a data file corresponding to the simulation result.

[0137] In some feasible implementation manners, the simulation process management tool at least includes tool nodes and a flow chart; wherein, specific operations are encapsulated in the tool nodes, and each of the tool nodes is used to execute one of the operations of setting parameters, running a simulation, and generating a report. The flow chart is a tool for defining and managing a state machine, and the flow chart is composed of multiple working nodes and the conversion logic between the working nodes.

[0138] In some feasible implementation manners, the simulation processing module 202 is specifically configured to:

[0139] Extract the requirement information from the long short-term memory module;

[0140] Extract target key information from the requirement information, where the target key information is information related to simulation testing;

[0141] Call the tool node to set simulation parameters matching the target key information;

[0142] Call the flow chart to set a simulation process corresponding to the simulation parameters and generate a test task corresponding to the simulation process.

[0143] In some feasible implementation manners, the simulation process orchestration system further includes simulation tool functions. The simulation tool functions at least include a parameter setting function, a simulation execution function, and a result generation function. The simulation processing module 202 is specifically configured to:

[0144] Call the parameter setting function to pass the simulation parameters into the simulation software;

[0145] Call the simulation execution function to control the simulation software to execute the simulation process according to the simulation parameters;

[0146] In response to the completion of the execution of the simulation process, call the result generation function to generate a simulation result corresponding to the simulation process.

[0147] In some feasible implementation manners, the simulation process orchestration system further includes a simulation manager; wherein, the simulation manager is used for setting and managing simulation types, performance, architectures, and models, and supports the setting of simulation parameters in multiple rounds of conversations. The device further includes:

[0148] An instruction acquisition module, configured to acquire an adjustment instruction for the test parameters input by the user;

[0149] An identification module, configured to call the large language model through the model interface to identify the adjustment instruction, and obtain a simulation adjustment parameter corresponding to the adjustment instruction;

[0150] An adjustment module, configured to adjust the simulation parameter to a target simulation parameter matching the simulation adjustment parameter.

[0151] In some feasible implementation manners, the simulation parameter at least includes a first simulation type, a first performance mode, and a first model parameter, the simulation adjustment parameter at least includes a type adjustment parameter, a performance adjustment parameter, and a model adjustment parameter, and the adjustment module is specifically configured to:

[0152] Adjust the first simulation type to a second simulation type matching the type adjustment parameter;

[0153] Adjust the first performance mode to a second performance mode matching the performance adjustment parameter;

[0154] Adjust the first model parameter to a second model parameter matching the model adjustment parameter.

[0155] In some feasible implementation manners, it further includes:

[0156] A knowledge feedback module, configured to obtain a knowledge Q&A result corresponding to the demand information if the intention type belongs to the knowledge acquisition category.

[0157] In some feasible implementation manners, the knowledge feedback module is specifically configured to:

[0158] Call the large language model through the model interface to generate a knowledge Q&A result corresponding to the demand information; and / or,

[0159] Call a preset knowledge extraction function to extract a knowledge Q&A result corresponding to the demand information from a preset model construction document; and / or,

[0160] Call a preset experience extraction function to extract a knowledge Q&A result corresponding to the demand information from a preset experience document.

[0161] For the device embodiments, since they are basically similar to the method embodiments, they are described relatively simply. For related parts, please refer to the descriptions in the method embodiments.

[0162] In addition, an embodiment of the present invention further provides an electronic device, including: a processor, a memory, and a computer program stored on the memory and executable on the processor. When the computer program is executed by the processor, it implements each process of the above-mentioned method embodiment for simulation processing based on a large language model agent, and can achieve the same technical effects. To avoid repetition, it will not be elaborated here.

[0163] An embodiment of the present invention further provides a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is executed by the processor, it implements each process of the above-mentioned method embodiment for simulation processing based on a large language model agent, and can achieve the same technical effects. To avoid repetition, it will not be elaborated here. Among them, the computer-readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc, etc.

[0164] Each embodiment in this specification is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. For the same or similar parts among the embodiments, reference can be made to each other.

[0165] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a device, or a computer program product. Therefore, the embodiments of the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the embodiments of the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, EEPROMs, Flash, and eMMC, etc.) containing computer-usable program codes.

[0166] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the method, terminal device (system), and computer program product according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal devices generate for implementing in the process Figure 1 each process or multiple processes and / or blocksFigure 1 means for the functions specified in one or more boxes.

[0167] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing terminal device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufactured article including an instruction means that implements the functions specified in one Figure 1 process or more processes and / or boxes Figure 1 means for the functions specified in one or more boxes.

[0168] These computer program instructions may also be loaded onto a computer or other programmable data processing terminal device, such that a series of operation steps are performed on the computer or other programmable terminal device to produce a computer-implemented process, and thus the instructions executed on the computer or other programmable terminal device provide steps for implementing the functions specified in one Figure 1 process or more processes and / or boxes Figure 1 means for the functions specified in one or more boxes.

[0169] Although the preferred embodiments of the embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concepts. Therefore, the appended claims are intended to be construed as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present invention.

[0170] Finally, it should also be noted that in this text, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or terminal device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or terminal device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or terminal device comprising the element.

[0171] The above has introduced in detail a simulation processing method based on a large language model agent and a simulation processing device based on a large language model agent. In this article, specific examples are used to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A simulation processing method based on large language model agents, characterized in that, Applied to a simulation process orchestration system, the simulation process orchestration system at least includes a model interface, a long short-term memory module, and a simulation process management tool. The method includes: Obtain the requirement information input by the user, and call the large language model accessed through the model interface to perform intent recognition on the requirement information to obtain the intent type corresponding to the requirement information; If the intent type belongs to the simulation operation category, call the simulation process management tool to perform simulation parameter setting according to the requirement information stored in the long short-term memory module, obtain the corresponding simulation task, and execute the simulation task to generate the corresponding simulation result; Obtain the file type of the user input for the simulation result, determine the display template corresponding to the file type, and input the simulation result into the display template to obtain the data file corresponding to the simulation result.

2. The method according to claim 1, wherein The simulation process management tool at least includes tool nodes and a flowchart; wherein, specific operations are encapsulated in the tool nodes, and each tool node is used to perform one of the operations of setting parameters, running simulations, and generating reports. The flowchart is a tool for defining and managing state machines, and the flowchart consists of multiple working nodes and the conversion logic between the working nodes.

3. The method according to claim 2, characterized in that, The calling the simulation process management tool to perform simulation parameter setting according to the requirement information stored in the long short-term memory module to obtain the corresponding simulation task includes: Extract the requirement information from the long short-term memory module; Extract the target key information from the requirement information, and the target key information is the information related to simulation testing; Call the tool node to set the simulation parameters matching the target key information; Call the flowchart to set the simulation process corresponding to the simulation parameters and generate the test task corresponding to the simulation process.

4. The method according to claim 3, characterized in that The simulation process orchestration system further includes simulation tool functions. The simulation tool functions at least include a parameter setting function, a simulation execution function, and a result generation function. The executing the simulation task to generate the corresponding simulation result includes: Call the parameter setting function to pass the simulation parameters into the simulation software; Call the simulation execution function to control the simulation software to execute the simulation process according to the simulation parameters; In response to the completion of the execution of the simulation process, call the result generation function to generate the simulation result corresponding to the simulation process.

5. The method according to any one of claims 1 to 4, characterized in that The simulation process orchestration system further includes a simulation manager; wherein, the simulation manager is used for the setting and management of simulation types, performance, architectures, and models, and supports the setting of simulation parameters in multi-round conversations. The method further includes: Obtain the adjustment instruction input by the user for the test parameters; Call the large language model through the model interface to recognize the adjustment instruction to obtain the simulation adjustment parameters corresponding to the adjustment instruction; Adjust the simulation parameters to the target simulation parameters matching the simulation adjustment parameters.

6. The method according to claim 5, characterized in that, The simulation parameters at least include a first simulation type, a first performance mode, and first model parameters. The simulation adjustment parameters at least include a type adjustment parameter, a performance adjustment parameter, and a model adjustment parameter. Adjusting the simulation parameters to target simulation parameters matching the simulation adjustment parameters includes: Adjusting the first simulation type to a second simulation type matching the type adjustment parameter; Adjusting the first performance mode to a second performance mode matching the performance adjustment parameter; Adjusting the first model parameters to second model parameters matching the model adjustment parameter.

7. The method according to claim 1, characterized in that, It further includes: If the intent type belongs to the knowledge acquisition category, obtaining a knowledge Q&A result corresponding to the demand information.

8. The method according to claim 7, wherein The obtaining of the knowledge Q&A result corresponding to the demand information includes: Invoking the large language model accessed through the model interface to generate a knowledge Q&A result corresponding to the demand information; and / or, Invoking a preset knowledge extraction function to extract a knowledge Q&A result corresponding to the demand information from a preset model construction document; and / or, Invoking a preset experience extraction function to extract a knowledge Q&A result corresponding to the demand information from a preset experience document.

9. A simulation processing device based on a large language model agent, characterized in that, Applied to a simulation process orchestration system, the simulation process orchestration system at least includes a model interface, a long short-term memory module, and a simulation process management tool. The device includes: An intent recognition module, configured to obtain demand information input by a user, and perform intent recognition on the demand information by invoking the large language model accessed through the model interface to obtain an intent type corresponding to the demand information; A simulation processing module, configured to, if the intent type belongs to the simulation operation category, invoke the simulation process management tool to set simulation parameters according to the demand information stored in the long short-term memory module, obtain a corresponding simulation task, and execute the simulation task to generate a corresponding simulation result; A result presentation module, configured to obtain the file type of the user input for the simulation result, determine a display template corresponding to the file type, and input the simulation result into the display template to obtain a data file corresponding to the simulation result.

10. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus. Among them, the processor, the communication interface, and the memory complete communication with each other through the communication bus; The memory is used for storing a computer program; When the processor executes the program stored on the memory, it implements the method according to any one of claims 1-8.

11. A computer-readable storage medium, on which instructions are stored. When the instructions are executed by one or more processors, the processors are caused to execute the method according to any one of claims 1-8.

Images