Skills agent for finite element analysis of earth-rock dam heightening and capacity expansion based on memory enhancement

Abstract

The application discloses a kind of based on memory enhancement new and old earth-rock dam deformation mechanism finite element analysis skills intelligent agent system, comprising: external input module, for receiving the input information of user;Intelligent agent decision module, with large language model as core, for input information is carried out semantic analysis and task understanding, according to task requirement corresponding skills are selected to process;Finite element tool module, for supporting intelligent agent by calling corresponding skills, realize the automatic conversion from natural language input to numerical analysis result;Result storage module, for storing various files generated in finite element modeling process.The application adopts skills modularization organization analysis process, strong expansibility, applicable to new and old earth-rock dam deformation mechanism analysis and other complex engineering scenarios;Memory enhancement mechanism is based on simultaneously, historical cases are combined to assist decision-making, effectively improve the rationality of intelligent agent parameter generation, modeling stability and analysis result accuracy.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and numerical simulation, and in particular to a Skills agent for finite element analysis of increasing the height and capacity of earth-rock dams based on memory enhancement. Background Technology

[0002] With the continuous expansion of water conservancy projects and the increasing demand for upgrading the functions of existing reservoirs, the heightening and expansion of earth-rock dams has become an important means to improve water resource regulation capabilities and ensure flood control safety. During this process, due to significant differences in material properties, construction dates, and stress history between the old and new dam bodies, they are prone to deformation inconsistencies under long-term loads and complex environmental conditions, leading to stress concentration, abnormal seepage, and structural safety hazards.

[0003] As a core numerical tool in geotechnical engineering for describing the response and multi-physics coupling behavior of complex structures, the finite element method can simulate the mechanical response process of earth-rock dams under multiple loads such as water pressure, self-weight, and earthquakes. It can accurately depict the spatiotemporal evolution of deformation field, stress field, and seepage field inside the dam body, providing important support for revealing the interaction mechanism of the interface between the old and new dam bodies and the path of disaster evolution, and providing theoretical basis for the design optimization and safety assessment of heightening and expansion projects.

[0004] With the rapid development of artificial intelligence technology, its application in engineering numerical simulation is constantly deepening, bringing new implementation methods to finite element modeling and analysis. Large language models possess the ability to semantically understand and parse parameters for complex engineering problems, reducing the reliance on human experience in traditional finite element modeling to a certain extent. Based on this, agent technology, as a task-oriented intelligent framework, parses user natural language input based on the ReAct pattern, connecting parameter configuration, model construction, and numerical solution, realizing an automated processing flow from "natural language description" to "finite element analysis results." Simultaneously, skills, as an important way to structurally encapsulate agent capabilities, modularize trigger rules, execution strategies, and external resources, enabling agents to perform multi-functional collaborative calls in complex tasks, thereby enhancing their executability in the finite element modeling and analysis process.

[0005] While agents based on large language models exhibit a degree of automation in finite element analysis (FEM), their accuracy remains insufficient when dealing with complex engineering problems. On one hand, FEM modeling involves numerous physical parameters with engineering experience characteristics. Parameters generated solely through general semantic understanding often fail to fully meet the requirements of rationality and accuracy in numerical calculations, leading to model results that do not accurately reflect reality. On the other hand, existing agents typically lack effective mechanisms for utilizing historical modeling experience, making it difficult to reuse existing analysis results in similar engineering scenarios, thus limiting their applicability to complex problems. In contrast, validated typical modeling cases in engineering practice contain rich empirical knowledge and reasonable parameter configurations, playing a crucial role in ensuring the computability and engineering effectiveness of FEM. Therefore, transforming human experience into structured, searchable case resources and introducing corresponding memory mechanisms, enabling agents to refer to historical case information during modeling, is of great significance for improving parameter generation quality, enhancing the stability of the analysis process, and increasing the engineering reliability of the results.

[0006] In existing technologies, Skills agents and their memory enhancement methods have been explored in some fields, but their application in the finite element analysis of deformation mechanisms of old and new earth-rock dams remains relatively limited. Currently, no published patents have been found to apply agent technology systems to the finite element analysis of deformation mechanisms of old and new earth-rock dams, especially in the area of ​​automated modeling and analysis under complex multi-field coupling and engineering experience constraints, where a significant gap remains. Existing research mainly focuses on enhancing the general capabilities of agents. Chinese patent (application publication number: CN120688545A) "A Long-Term Memory Modeling Method for Agents Based on Memory Networks" improves information storage and retrieval to some extent by constructing a memory network to model the long-term memory capabilities of agents. However, this type of method is mainly geared towards general task scenarios, focusing on the construction and optimization of the memory structure itself, and has not yet combined with the finite element modeling requirements of specific engineering problems. It lacks dedicated designs for complex physical processes, engineering experience constraints, and parameter rationality requirements, making it difficult to directly apply to high-precision engineering calculation scenarios such as the analysis of deformation mechanisms of old and new earth-rock dams. Summary of the Invention

[0007] To address the shortcomings of existing technologies in finite element analysis of complex engineering projects, such as insufficient accuracy in parameter generation, lack of engineering experience support, and limited automation in the modeling and analysis process, this invention proposes a Skills-based intelligent agent for finite element analysis of earth-rock dam height enhancement and expansion. This invention constructs an intelligent agent system integrating a large language model, Skills, and finite element tools, including an external input module, an intelligent agent decision-making module, a finite element tool module, and a result storage module.

[0008] The external input module is used to receive user input information;

[0009] The intelligent agent decision-making module, with a large language model as its core, is used to perform semantic parsing and task understanding on the input information and match corresponding skills according to task requirements; wherein, the skills are used to structurally encapsulate the finite element analysis process, enabling the finite element modeling task to realize module combination and call.

[0010] The finite element tool module is used to support intelligent agents to complete the finite element modeling and analysis process by calling the corresponding skills, including mesh generation, finite element model generation, finite element solution and result visualization, realizing the automatic conversion from natural language input to numerical analysis results;

[0011] The result storage module is used to store various files generated during the finite element modeling process, realizing structured storage of process data and final results.

[0012] The beneficial effects of this invention are:

[0013] This invention proposes a Skills agent for finite element analysis of earth-rock dam height increase and capacity expansion based on memory enhancement. This agent automates the process from natural language description to finite element analysis results. By modularizing the analysis process using Skills, the system achieves good scalability and can be applied to complex engineering scenarios such as deformation mechanism analysis of new and old earth-rock dams. Based on the memory enhancement mechanism, it enables explicit expression and effective utilization of engineering experience, allowing the agent to combine historical cases for decision support, improving its performance in terms of parameter generation rationality, modeling process stability, and analysis result accuracy. Attached Figure Description

[0014] Figure 1 The intelligent agent system architecture diagram provided by this invention;

[0015] Figure 2 The intelligent agent reasoning flowchart provided by this invention;

[0016] Figure 3 This invention provides a flowchart of user-assisted intelligent agent modeling and analysis.

[0017] Figure 4 The present invention provides a flowchart for generating parameters from natural language using skills;

[0018] Figure 5 A schematic diagram of the memory enhancement mechanism provided by the present invention;

[0019] Figure 6 This is a schematic diagram of the case selection provided by the present invention;

[0020] Figure 7 The RAG retrieval flowchart provided by this invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] like Figure 1 As shown, this invention provides a Skills agent for finite element analysis of earth-rock dam height increase and capacity expansion based on memory enhancement, including an external input module, an agent decision-making module, a finite element tool module, and a result storage module.

[0023] The external input module is used to receive engineering problems described by users in natural language.

[0024] The intelligent agent decision-making module uses a large language model as its core to perform semantic parsing and task understanding of the input information, and selects the appropriate skills for processing according to the task requirements. The skills are used to encapsulate the finite element analysis process in a structured way, including functions such as parameter generation, mesh generation, model construction, model solving and result visualization, so that complex modeling tasks can be called by module combination.

[0025] The finite element tool module comprises four units: mesh generation, finite element model generation, finite element solver, and result visualization. The agent, by invoking corresponding skills, sequentially completes the mesh generation, finite element model generation, finite element solution, and result visualization processes, achieving automated conversion from natural language input to numerical analysis results. Specifically, the mesh generation unit generates a discrete finite element mesh from a .json file and outputs a .msh file; the finite element model generation unit constructs the input data file required for analysis based on the .json and .msh files and outputs a .data file; the finite element solution unit performs numerical calculations and outputs a .vtk format result file for post-processing and visualization analysis; and the result visualization unit displays the .vtk file as a cloud map for user viewing.

[0026] The result storage module is used to store process files generated during the modeling process. These files include .json files, .msh files, DATA files, .vtk files, and a case library, achieving structured storage of process data and final results. Figure 4As shown, the case library stores verified and reasonable cases in a structured format, including tags, user input, and structured parameter lists.

[0027] The contents of the process files are detailed in Table 1 below:

[0028] Table 1. Process document contents

[0029] Parameter.json Analysis settings, geometry, material properties, boundary conditions, loading conditions, time series, preset crack locations, mesh node information, initial state. Stored in key-value pairs in a structured manner Model.msh Mesh information includes physical groups, nodes, and cells. —— Inputfile The assembly of the finite element model includes analysis settings, mesh information, material properties, boundary conditions, load cases, and time series. Covers all the information needed by the finite element solver Results.vtk The results include displacement, stress-strain, pore pressure, and effective stress. Visualization can be performed using commonly used visualization tools on the market.

[0030] Furthermore, such as Figure 2 As shown in the figure, this diagram illustrates the reasoning process of a memory-enhanced agent in the finite element modeling and analysis task. The specific steps are as follows:

[0031] First, the user inputs a description of the engineering problem in natural language. The input includes, but is not limited to, dam structure information, material parameters, working conditions, and analysis objectives. The agent performs preliminary reasoning and semantic parsing on the input information, extracts key elements related to finite element modeling, and forms a structured task representation.

[0032] Secondly, after completing the initial understanding, the agent calls the memory enhancement module to vectorize the input content and perform a similarity-based retrieval operation in the case library to obtain historical modeling cases related to the current engineering problem. The retrieval results include information such as tags, user input, and structured data lists, which are used to provide reference for the subsequent reasoning process.

[0033] Next, the agent enters the deep reasoning stage. Based on the user input information and the retrieved case information, it refines and decomposes the modeling task, generating specific execution plans. During this process, the agent searches for and matches corresponding skill modules according to task requirements. These skills encapsulate key operational steps in the finite element analysis process, including parameter generation, mesh generation, finite element model construction, and solution visualization.

[0034] Subsequently, after determining the required skills, the agent invokes the corresponding tool modules in a predetermined order to complete the finite element modeling and analysis process, including mesh generation, finite element model construction, and numerical solution, and generates execution results. The agent observes and analyzes the results, extracting key response information to evaluate the rationality of the calculation.

[0035] Finally, in the results evaluation phase, the system provides feedback on the analysis results based on preset standards or user subjective judgment. When the final result meets the requirements, the process ends; when the process result meets the requirements, the agent continues to execute the next stage task; when the result does not meet the requirements, the user can provide correction suggestions or supplementary information, and the agent re-executes the reasoning process based on the feedback, including re-searching cases, adjusting parameters, and re-calling relevant skills, thus forming a closed-loop iterative optimization process.

[0036] Furthermore, such as Figure 3 As shown in the figure, this diagram illustrates the interactive process of a user completing finite element modeling and analysis using a memory-enhanced agent. The specific steps are as follows:

[0037] First, users input relevant information about the problem to be analyzed using natural language, based on engineering requirements. This includes the dam structure, material zoning, physical parameter ranges, boundary conditions, and analysis objectives.

[0038] Secondly, after receiving the input information, the agent performs semantic analysis, reasoning and execution step by step, and generates various results, including parameter files, mesh partitioning results, and calculation results. The agent then feeds back the generated parameter files (such as .data files), mesh settings (such as .msh files), and solution results to the user.

[0039] Subsequently, users evaluate the feedback results, including whether the parameter values ​​conform to engineering experience, whether the mesh generation is reasonable, and whether the boundary conditions are complete. During the user review process, adjustments can be made to the model based on engineering experience or computational needs, and feedback on modifications can be provided to the agent, such as parameter correction, mesh refinement, or adjustment of the analysis step size. Simultaneously, users can also evaluate the model's numerical convergence, computational stability, and the reasonableness of the results based on existing calculations. If non-convergence or abnormal results are found, further correction instructions can be proposed.

[0040] Next, the agent updates the modeling scheme based on user feedback, regenerates the parameter file and mesh settings, and submits it to the user for confirmation again. After the parameters and model settings meet the requirements, the agent calls the finite element solver to complete the numerical calculation and outputs the result file (such as a .vtk file) and the corresponding visualization analysis results.

[0041] Finally, the user further evaluates the calculation results, including whether the dam deformation characteristics, stress distribution, and seepage response meet the engineering expectations. If the results still do not meet the requirements, the model is further corrected through the feedback mechanism; if the results meet the requirements, the modeling and analysis process is completed.

[0042] Furthermore, such as Figure 4As shown in the figure, this illustrates the specific process of generating finite element parameters from natural language input using the skills mechanism. The user inputs their modeling requirements in natural language, the agent performs semantic parsing of the input, and determines whether to trigger the corresponding skill module based on the task type. When the input is identified as involving a finite element modeling task, the agent selects the natural language-parameter conversion skill to perform subsequent operations.

[0043] Furthermore, in this embodiment, the natural language-parameter conversion skill is organized in a hierarchical structure, including the following three levels:

[0044] (1) Level 1: Control and Description Layer

[0045] In this embodiment, this layer is used to define the basic attributes and triggering mechanism of the skill, including the skill name, function description, triggering conditions, and execution method after triggering. The triggering conditions are used to determine whether the user input meets the requirements for calling the skill; when the conditions are met, the specific processing flow of the skill is initiated according to the preset execution method, thereby realizing the call control from natural language to parameter generation task.

[0046] (2) Level 2: Execution process layer

[0047] In this embodiment, this layer is used to define the specific execution steps of the skill and is the core process for parameter generation. The execution process includes the following steps:

[0048] Step 1: Parse the user input and extract key modeling information;

[0049] Step 2: Search for similar cases in the case library based on the input content;

[0050] Step 3: Combine user input and search results to construct mixed prompt information to constrain the generation process;

[0051] Step 4: Generate the parameters required for finite element analysis based on the prompts.

[0052] Step 5: Organize the generated results into a structured format and output them, such as a JSON file or a solver input file.

[0053] (3) Level 3: External Resource Layer

[0054] In this embodiment, this layer provides experience support for skill execution and mainly consists of a case library. The case library stores historical modeling cases in a structured format. Each case includes a case number, tag information, user input examples, and corresponding parameter results. The parameter results are typically represented in a structured format, including geometric dimensions, material parameters, boundary conditions, and calculation control parameters. This embodiment achieves categorized storage and retrieval of different analysis types (such as seepage analysis and deformation analysis) through case tagging management.

[0055] Furthermore, such as Figure 5 As shown, this invention presents a memory enhancement mechanism based on a case library, used to store, retrieve, and dynamically update historical experience during finite element modeling and analysis, thereby assisting the agent in parameter generation and model construction. The specific steps are as follows:

[0056] First, after receiving the modeling task input by the user, the agent performs semantic parsing on the input content and converts it into a vectorized representation. Based on this representation, a similarity search is performed in the case library to retrieve historical cases related to the current task. The case library stores typical modeling cases in a structured format. Each case includes information such as an engineering scenario description, parameter files, and calculation results, and is accompanied by corresponding tags to identify the analysis type and applicable scope.

[0057] Next, after retrieving relevant cases, the agent uses the key parameter information and modeling strategies in the cases as a reference to participate in the parameter generation process of the current task, thereby constraining and guiding the generated results to better meet engineering experience and numerical calculation requirements.

[0058] Subsequently, after completing the finite element analysis, the system evaluates the calculation results. This evaluation can be based on preset indicators, including but not limited to numerical convergence, computational stability, result rationality, and engineering effectiveness. When the calculation results meet the evaluation criteria, the relevant information of the current modeling task (including user input, parameter files, and calculation results) is structured and stored as a new case in the case library, expanding memory capacity. When the calculation results do not meet the evaluation criteria, the case is not added to the library, and the parameter generation process can be adjusted based on the evaluation results, such as re-retrieving cases, correcting parameters, or optimizing model settings to improve the rationality of subsequent calculation results.

[0059] Therefore, this embodiment uses a closed-loop mechanism of "case retrieval - auxiliary generation - result evaluation - memory update" to enable the content in the case library to be continuously accumulated and optimized as it is used, thereby gradually enhancing the adaptability and engineering reliability of the intelligent agent in complex finite element modeling tasks.

[0060] Furthermore, such as Figure 6As shown in the figure, this embodiment provides a case screening mechanism for the case library construction process, which is used to evaluate and screen the finite element modeling results generated by the intelligent agent to ensure the engineering applicability of the cases included in the library.

[0061] Furthermore, after the intelligent agent system completes finite element modeling and numerical calculation, the generated parameter file, mesh settings, finite element model, and calculation results are used as candidate cases input into the screening mechanism. The screening mechanism judges candidate cases based on preset evaluation indicators, and the evaluation content includes, but is not limited to, the following aspects:

[0062] (1) Computability judgment: used to judge whether the generated finite element model can successfully complete numerical calculation, including whether the model can run normally and whether the solution process is interrupted.

[0063] (2) Numerical convergence judgment: used to evaluate whether the iterative solution converges during the calculation process, including whether the residuals meet the convergence conditions and whether there is divergence in the calculation process.

[0064] (3) Result rationality judgment: used to judge whether the calculation results conform to basic physical laws and engineering cognition, such as whether the deformation distribution is continuous and whether the stress response is abnormal.

[0065] (4) Engineering applicability judgment: used to evaluate whether the modeling parameters and calculation results have engineering significance, including whether the parameter values ​​are within a reasonable range and whether the model settings are in line with the actual engineering scenario.

[0066] (5) Generation stability judgment: used to evaluate whether the model generation and calculation results are consistent and stable under similar input conditions.

[0067] Furthermore, when a candidate case meets all of the above evaluation criteria, it is determined to be a valid case, and after being structured, it is stored in the case library; when a candidate case fails to meet the above evaluation criteria, it is not stored in the library, and feedback can be given to the modeling process based on the evaluation results, such as adjusting parameter settings, regenerating the model, or re-calling relevant skills, in order to improve the quality of subsequent results.

[0068] Furthermore, such as Figure 7 As shown, this embodiment employs a case retrieval mechanism based on retrieval enhancement generation technology to effectively utilize historical modeling experience. The specific steps are as follows:

[0069] (1) All historical cases are pre-vectorized and a vector index structure is established in the case library. Each case in the case library includes information such as user input description, parameter file, model configuration and calculation results, and is also labeled to identify the case type.

[0070] (2) The agent receives the user's input of natural language modeling requirements, performs semantic parsing on the input content, and extracts key information related to finite element modeling. Based on this, the parsed information is converted into a vectorized representation, which serves as the query vector for subsequent retrieval.

[0071] (3) Perform similarity retrieval in the case library based on the query vector, and obtain several candidate cases most relevant to the current task by calculating the similarity between vectors. The similarity measure can be cosine similarity or other vector distance measures.

[0072] (4) After obtaining candidate cases, the search results are filtered and sorted. Priority is given to selecting the cases that match the current task in terms of analysis type, structural features and working conditions, and their key parameter information and modeling strategies are extracted.

[0073] (5) The retrieved case information is integrated with the user's original input to construct enhanced prompt information, including system prompt words and few-sample examples, to guide the subsequent parameter generation process. The enhanced prompt information retains user needs while incorporating empirical constraints from historical cases.

[0074] (6) Based on the constructed enhanced prompt information, call the large language model to generate structured parameter results and output the parameter file for finite element modeling.

[0075] The above description is merely a description of preferred embodiments of this application and is not intended to limit the scope of this application in any way. Any changes or modifications made by those skilled in the art based on the above-disclosed technical content should be considered as equivalent and valid embodiments and fall within the scope of protection of the technical solution of this application.

Claims

1. A skill agent for finite element analysis of increasing the height and capacity of earth-rock dams based on memory enhancement, characterized in that, It includes an external input module, an intelligent agent decision-making module, a finite element tool module, and a result storage module; The external input module is used to receive user input information; The intelligent agent decision-making module, with a large language model as its core, is used to perform semantic parsing and task understanding on the input information and match corresponding skills according to task requirements; wherein, the skills are used to structurally encapsulate the finite element analysis process, enabling the finite element modeling task to realize module combination and call. The finite element tool module is used to support intelligent agents to complete the finite element modeling and analysis process by calling the corresponding skills, including mesh generation, finite element model generation, finite element solution and result visualization, realizing the automatic conversion from natural language input to numerical analysis results; The result storage module is used to store various files generated during the finite element modeling process, realizing structured storage of process data and final results.

2. The Skills agent for finite element analysis of earth-rock dam height increase and capacity expansion based on memory enhancement as described in claim 1, characterized in that, The finite element tool module includes mesh generation units, finite element model generation units, finite element solver units, and result visualization units; The mesh generation unit is used to generate a finite element discrete mesh based on the .json file and output a .msh file; The finite element model generation unit is used to construct the input data file required for analysis based on the .json and .msh files and output the .data file; The finite element solving unit is used to perform numerical calculations and output .vtk format result files; The result visualization unit is used to display the .vtk file to the user in the form of a cloud map.

3. The Skills agent for finite element analysis of earth-rock dam height increase and expansion based on memory enhancement as described in claim 1 or 2, characterized in that, The results storage module stores various types of files, including .json files, .msh files, DATA files, .vtk files, and a case library; wherein, the case library stores verified and reasonable cases in a structured format.

4. The Skills agent for finite element analysis of earth-rock dam height increase and capacity expansion based on memory enhancement as described in claim 3, characterized in that, In the finite element modeling and analysis task, the agent performs the following inference: The intelligent agent performs preliminary reasoning and semantic parsing on the engineering problem description input by the user, extracts relevant elements for finite element modeling, and forms a structured task representation; After the agent completes its initial understanding, it calls the memory enhancement module to vectorize the input information and then retrieves historical modeling cases related to the current engineering problem through similarity retrieval in the case library. After acquiring historical modeling cases, the agent enters the deep reasoning stage, searching for and matching the corresponding skills according to the task requirements; After determining the required skills, the intelligent agent calls the corresponding tool modules in sequence to complete finite element modeling and analysis, generates results, observes and analyzes them, and extracts response information to evaluate the rationality of the calculation. During the results evaluation phase, the analysis results are based on preset standards or user judgment feedback: if the results meet the standards, the process ends or proceeds to the next phase. If the target is not met, users can provide additional suggestions for improvement, and the intelligent agent will then reason again based on these suggestions, forming a closed-loop iterative optimization.

5. The Skills agent for finite element analysis of earth-rock dam height increase and expansion based on memory enhancement as described in claim 4, characterized in that, The user completes the finite element modeling and analysis through a memory-enhanced agent via the following interactive steps: Users input information about the problem to be analyzed using natural language, based on the project requirements; After receiving input information, the intelligent agent performs semantic parsing, reasoning, and execution to generate results, which are then simultaneously fed back to the user. Users can evaluate the feedback results, adjust the model based on experience or needs, and provide feedback on modifications; they can also evaluate the calculation results and propose correction instructions when problems are found. The agent updates the modeling scheme based on user feedback, regenerates the parameter file and mesh settings, and submits them for confirmation. Once the parameters and model settings meet the requirements, call the finite element solver to complete the numerical calculation and output the result file and visualization analysis results. Users evaluate the results; if the results meet the requirements, the finite element modeling and analysis are completed; otherwise, the model is further corrected through a feedback mechanism.

6. The Skills agent for finite element analysis of earth-rock dam height increase and expansion based on memory enhancement as described in claim 5, characterized in that, The intelligent agent performs semantic parsing on the input information. If it identifies that the input information relates to the finite element modeling, it selects the natural language-parameter transformation skill to perform subsequent operations. This skill employs a hierarchical structure and includes: The control and description layer is used to define the basic attributes and triggering mechanism of the skill, including name, function description, triggering conditions and execution method. The triggering conditions determine whether to call the skill. If the conditions are met, the processing flow is started according to the preset execution method, realizing the call control from natural language to parameter generation task. The execution flow layer is used to define the specific execution steps of the skill and to generate parameters. The external resource layer, with a case library at its core, provides experience support for skill execution. The case library stores historical modeling cases in a structured manner. Each case includes a number, tag, user input example, and structured parameter results. The case library achieves classification, storage, and retrieval of cases of different analysis types through tag-based management.

7. The Skills agent for finite element analysis of earth-rock dam height increase and capacity expansion based on memory enhancement as described in claim 6, characterized in that, The execution steps of the execution process layer are as follows: Parse user input information and extract modeling information; Search for similar cases in the case library based on user input; Combine user input with search results to create hybrid prompts; Generate the parameters required for finite element analysis based on the prompts; Organize the generated results into a structured format and output them.

8. The Skills agent for finite element analysis of earth-rock dam height increase and capacity expansion based on memory enhancement as described in claim 7, characterized in that, The case library is based on a memory enhancement mechanism. By storing, retrieving, and dynamically updating historical experience in finite element modeling and analysis, it assists the agent in parameter generation and model construction. The specific steps are as follows: After receiving the modeling task, the agent first performs semantic parsing and quantization on the input content, and then retrieves similar historical cases from the case library based on the vector. After the agent retrieves similar historical cases, it uses the parameter information and modeling strategies in those cases as a reference to constrain and guide the generation of parameters for the current task. After the finite element analysis is completed, the calculation results are evaluated: if they meet the standards, the current modeling task information is structured and stored in the case library to expand memory; If the criteria are not met, the data entry will not be performed, and the parameter generation process can be adjusted based on the evaluation results.

9. The Skills agent for finite element analysis of earth-rock dam height increase and expansion based on memory enhancement as described in claim 8, characterized in that, After the intelligent agent completes the finite element modeling and numerical calculation, it inputs the generated parameter file, mesh settings, finite element model and calculation results as candidate cases into the screening mechanism. The screening mechanism judges the candidate cases based on preset evaluation indicators, including judgment of computability, numerical convergence, result rationality, engineering applicability and generation stability.

10. The Skills agent for finite element analysis of earth-rock dam height increase and capacity expansion based on memory enhancement as described in claim 6, characterized in that, To effectively utilize the historical modeling cases, a case retrieval mechanism based on retrieval enhancement generation technology is adopted, and the execution steps are as follows: All historical cases are pre-vectorized and encoded in the case library, and a vector index structure is established. After receiving the user's natural language modeling request, the intelligent agent first performs semantic parsing to extract relevant information on finite element modeling, and then converts it into a vectorized representation as the query vector for subsequent retrieval. Based on the query vector, a similarity retrieval is performed in the case library to calculate the similarity between vectors and obtain the candidate cases most relevant to the current task. After obtaining candidate cases, the best cases are first filtered and sorted by matching degree, and then their parameter information and modeling strategies are extracted. The retrieved case information is combined with the user's original input to construct enhanced prompts; Based on the enhanced prompts, the large language model is invoked to generate structured parameters, and a parameter file for finite element modeling is output.

Images