A geotechnical engineering finite element modeling and optimization method and system based on a large language model and a storage medium

Abstract

This invention discloses a method, system, and storage medium for finite element modeling and optimization of geotechnical engineering based on a large language model. By acquiring task integration information of the target geotechnical engineering project to be constructed, engineering data information and model configuration attribute information are sequentially identified and generated. A structured simulation plan is determined, finite element software interface code is generated, and baseline finite element model construction and numerical calculation are performed. Parameter optimization and structural configuration selection are also performed to obtain an optimized finite element model corresponding to the target geotechnical engineering project. The method and system provided by this invention automatically construct and optimize finite element models by executing finite element software interface code step by step and transferring the completed model state to subsequent steps. This reduces manual work in model construction, code debugging, or scheme optimization, and improves model executability, process traceability, design space search efficiency, and reduces costs.

Description

Technical Field

[0001] This application relates to the field of geotechnical engineering design technology, and in particular to a method, system and storage medium for finite element modeling and optimization of geotechnical engineering based on a large language model. Background Technology

[0002] In geotechnical engineering design and safety evaluation, finite element analysis (FEM) is a core numerical simulation method. Its basic process involves: establishing a numerical model that accurately reflects the stratum distribution, support structure, and construction procedures; and solving this model to calculate key response indicators such as stratum deformation, structural internal forces, and overall stability. Because this method can comprehensively consider complex geological conditions, the impact of phased construction, and soil-structure interactions, it has been widely used in engineering practice.

[0003] However, to establish a reliable finite element model, it is essential to rationally determine the soil constitutive relations, model boundary conditions, and simulation strategies for each construction stage. After completing the numerical calculations, it is also necessary to comprehensively evaluate the deformation, internal forces, and stability responses at each construction stage to identify the controlling conditions and determine the engineering rationality of the calculation results. If design optimization is involved, it is necessary to repeatedly adjust the structural dimensions, component layout, or construction parameters, and to calculate and evaluate each candidate scheme one by one to ultimately select the scheme that meets the requirements of safety, applicability, and economy.

[0004] Currently, with the continuous improvement of computer performance, the computational cost of numerical solutions has been significantly reduced. Correspondingly, the manpower invested in model building, calculation result verification, and analysis and evaluation has become the main bottleneck restricting design efficiency and limiting the full exploration of design space.

[0005] In recent years, Large Language Models (LLMs) have demonstrated powerful capabilities in natural language understanding, task decomposition, code generation, tool invocation, and contextual reasoning. Theoretically, they can understand engineering tasks and modeling requirements, autonomously organize finite element modeling processes, generate interface code for interaction with finite element software, and provide feedback on parameter or structural scheme adjustments based on calculation results. This provides a potential technological foundation for overcoming the aforementioned human resource bottlenecks and achieving automated modeling and design optimization in finite element analysis.

[0006] However, the interface methods and solution commands of specialized finite element method (FEB) software are highly specialized and constitute a relatively small proportion of the publicly available training corpora for large language models. This makes it highly susceptible to creating fictitious interface methods, misusing parameters, or referencing incorrect objects when generating corresponding code. More importantly, without interactive execution with actual software and feedback, these errors are difficult to identify and correct in a timely manner, severely limiting the reliability and practicality of large language models in this field.

[0007] Therefore, the existing technology needs further improvement. Summary of the Invention

[0008] In view of the shortcomings of the prior art, the purpose of this invention is to provide users with a method, system and storage medium for geotechnical engineering finite element modeling and optimization based on a large language model, overcoming the defects of existing technologies where professional finite element software code generated by large language models is difficult to execute reliably and lacks consistency between different modeling steps.

[0009] The technical solution adopted by this invention to solve the technical problem is as follows: In a first aspect, the present invention provides a finite element modeling and optimization method for geotechnical engineering based on a large language model, wherein: Obtain task integration information for the target geotechnical engineering project to be constructed, and use a large language model to identify the task integration information to generate engineering data information and model configuration attribute information; The structured simulation plan is determined based on the engineering data and model configuration attribute information; The structured simulation plan generates finite element software interface code, and performs benchmark finite element model construction and numerical calculation. The calculation results of the numerical calculation are compared with the preset standard results to obtain the conformity evaluation results. Based on the constructed benchmark finite element model and the compliance evaluation results, parameter optimization and structural configuration screening are performed to obtain the optimized finite element model corresponding to the target geotechnical engineering.

[0010] Optionally, the step of obtaining task integration information of the target geotechnical engineering project to be constructed, and using a large language model to identify the task integration information to generate engineering data information and model configuration attribute information includes: The system receives task integration information for the target geotechnical engineering project and uses a large language model to identify engineering data information within the task integration information. The task integration information includes one or more of the following: textual information related to the target geotechnical engineering project, engineering drawings, geological profiles, and design documents. The engineering data information includes one or more of the following: engineering data information, soil layer distribution information, material information, support system layout, load information, and construction information. A complete information check is performed on the engineering data information to obtain the engineering data information after the complete information check. According to the preset model configuration rules, the model configuration attribute information is derived from the engineering data information after the complete information check; the model configuration attribute information includes: the model boundary range, the simulated position of the retaining wall in the model space, and the set elevation of the supporting structure.

[0011] Optionally, the step of determining the structured simulation plan based on the engineering data information and model configuration attribute information includes: The engineering data and model configuration attribute information are used to identify modeling items and distinguish between modeling items and non-modeling items. The modeling items include one or more of the following: constitutive model identifier, material parameter values ​​and their data sources, drainage condition identifier, groundwater treatment method identifier, boundary condition definition, mesh generation control parameters, structural unit type definition, construction stage sequence, monitoring point coordinates, and result extraction item list. Modeling items are constructed according to the preset modeling and planning process, and corresponding planning data are retrieved based on the modeling items; Based on the planning data, various modeling decisions are determined, and a structured simulation plan is generated after verifying that the information of each modeling decision is correct.

[0012] Optionally, the steps of generating finite element software interface code based on the structured simulation plan and performing baseline finite element model construction and numerical calculation include: Obtain the structured simulation plan and model configuration attribute information, determine the current simulation step and the prerequisites for each simulation step based on the structured simulation plan and model configuration attribute information, and read the model state summary formed by the current simulation step. According to the current simulation steps, the corresponding implementation rules are loaded from the pre-built practice rule library; wherein, the practice rule library consists of implementation rules saved according to the modeling tasks, and the modeling tasks include: geometry construction, constitutive model setting, soil-structure interaction, construction stage setting, and result extraction; Based on the verified code patterns retrieved from the reference code library that match the current simulation steps; The current simulation step, prerequisites, model state summary, implementation rules, reference code pattern, and interface information are combined into a code generation context to generate finite element software interface code for completing the current simulation step. Execute the finite element software interface code, save the finite element software interface code corresponding to the current step, record the model state summary formed in this step, and use the model state summary as the code generation context for the next simulation step, until all modeling steps are completed. Execute each calculation stage specified in the structured simulation plan, determine whether each calculation stage has met the preset convergence requirements, and extract the calculation results specified in the structured simulation plan.

[0013] Optionally, the step of executing the finite element software interface code and saving the finite element software interface code corresponding to the current step further includes: Collect logs and exception tracking information after the finite element software interface code is executed in the current simulation step; Determine whether the current simulation step was executed successfully based on the verification requirements corresponding to the current simulation step in the structured simulation plan; If the current simulation step fails, then the interface documentation, technical reference information and implementation examples corresponding to the finite element software interface code are diagnosed, and the updated code after diagnosis is regenerated and executed.

[0014] Optionally, the step of performing parameter optimization and structural configuration screening based on the constructed benchmark finite element model and the compliance evaluation results to obtain an optimized finite element model corresponding to the target geotechnical engineering includes: The system acquires the constructed baseline finite element model, its execution context, and evaluation results, and acquires and loads parameter information for optimizing the workflow; wherein the parameter information includes parameter dependencies, parameter adjustment priorities, and search space hierarchical operations. By analyzing the finite element model, the continuous design variables, feasible parameter boundaries, and pre-defined dependencies between variables are determined. The baseline scheme is determined with minimizing the project cost as the optimization objective and pre-defined constraints. Under the condition of the structural configuration of the benchmark scheme, a surrogate model is established based on the preset compliance evaluation results, and candidate parameter combinations are proposed within the preset parameter space; The candidate parameter combination is input into the constructed finite element model to obtain the candidate model, and then executed to extract the finite element response corresponding to the constraint conditions output by the candidate model. Obtain the finite element response corresponding to each candidate model, and determine the optimal design finite element scheme under the current structural configuration; Based on the optimal design finite element scheme, determine whether there are multiple structural design variables; Based on multiple structural design variables, the geotechnical finite element model corresponding to the optimal mechanism configuration is determined.

[0015] Optionally, the step of determining whether there are multiple structural design variables based on the optimal finite element design scheme includes: Analyze the finite element response and safety margin of the optimal design finite element scheme under the current structural configuration, determine whether there are structural design variables, and propose multiple candidate structural configurations under the constraint conditions. The optimization results under each candidate structural configuration are calculated, and based on the optimized structure under each candidate structural configuration, the optimal structural configuration and its continuous design parameters, engineering cost, cost change relative to the benchmark scheme, convergence history and evaluation results are determined.

[0016] Optionally, the method further includes: All data generated during the construction and optimization of the finite element model are stored in the shared space. The data generated during the construction and optimization of the finite element model includes one or more of the following: engineering data information and model configuration attribute information, structured simulation plan, model status summary, model access identifier, code generation results, finite element calculation results and compliance evaluation results, finite element model, candidate parameters, finite element response, and optimization history.

[0017] Secondly, this invention provides a finite element modeling and optimization system for geotechnical engineering based on a large language model, comprising: The task clarification agent module is used to acquire task integration information of the target geotechnical engineering to be constructed, and to use a large language model to identify the task integration information to generate engineering data information and model configuration attribute information. The simulation planning intelligent agent module is used to determine the structured simulation plan based on the engineering data information and model configuration attribute information; The code generation intelligent agent module is used to generate finite element software interface code according to the structured simulation plan, and to perform benchmark finite element model construction and numerical calculation. The calculation results of the numerical calculation are compared with preset standard results to obtain compliance evaluation results. The design optimization intelligent agent module is used to perform parameter optimization and structural configuration screening based on the constructed benchmark finite element model and the compliance evaluation results, so as to obtain the optimized finite element model corresponding to the target geotechnical engineering.

[0018] Thirdly, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a computer, the computer is used to execute the aforementioned geotechnical engineering finite element modeling and optimization method based on a large language model.

[0019] Beneficial effects: This invention discloses a method, system, and storage medium for finite element modeling and optimization in geotechnical engineering based on a large language model. By executing finite element software interface code step by step and passing the completed model state to subsequent steps, it can realize continuous automated processing of finite element task definition, simulation planning, model generation, calculation verification, result evaluation, and design optimization in geotechnical engineering. This reduces the manual work required for model construction, code debugging, result processing, and scheme trial calculation, thereby improving the efficiency of finite element analysis and design optimization. Furthermore, it improves the executability of the model, process traceability, design space search efficiency, and reduces costs. Attached Figure Description

[0020] Figure 1This is a flowchart illustrating the steps of the finite element modeling and optimization method for geotechnical engineering based on a large language model in an embodiment of the present invention. Figure 2 This is a flowchart illustrating the processing of the task clarification agent in an embodiment of the present invention. Figure 3 This is a flowchart illustrating the processing of the simulation planning agent in an embodiment of the present invention; Figure 4 This is a flowchart illustrating the code generation, execution verification, and error correction process of the code generation intelligent agent in this embodiment of the invention. Figure 5 This is a flowchart of the finite element coupling optimization process for designing and optimizing the intelligent agent in an embodiment of the present invention; Figure 6 This is a structural principle block diagram of the geotechnical engineering finite element modeling and optimization system based on a large language model in an embodiment of the present invention; Figure 7 This is a schematic diagram illustrating the principle and structure of the finite element modeling and optimization method for geotechnical engineering in this embodiment of the invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0022] To overcome the problems of excessive manual workload required for existing finite element modeling, calculation verification, and result analysis in geotechnical engineering, unreliable execution of professional finite element software code generated by large language models, insufficient consistency between different modeling steps, and the inability of existing finite element optimization to simultaneously search for continuous design parameters and discrete structural configurations, this invention discloses a method, system, and storage medium for finite element modeling and optimization in geotechnical engineering based on a large language model. The method involves acquiring task integration information of the target geotechnical engineering project to be constructed, identifying the task integration information using a large language model to obtain engineering data information and model configuration attribute information, determining a structured simulation plan based on the engineering data information and model configuration attribute information, generating finite element software interface code based on the structured simulation plan, and performing benchmark finite element model construction and numerical calculation. The calculation results of the numerical calculation are compared with preset standard results to obtain a conformity evaluation result. Finally, parameter optimization and structural configuration screening are performed based on the constructed benchmark finite element model and the conformity evaluation result to obtain an optimized finite element model corresponding to the target geotechnical engineering project. The method provided by this invention sequentially performs task identification, simulation planning, code generation, and design optimization, thereby realizing the automatic construction of finite element models, reducing manpower, and the method is highly executable, traceable, and easy to implement, providing technical support for geotechnical engineering design.

[0023] The embodiments disclosed in this invention will now be described in more detail with reference to the accompanying drawings.

[0024] Firstly, this invention provides a finite element modeling and optimization method for geotechnical engineering based on a large language model, such as... Figure 1 As shown, it includes: Step S1: Obtain the task integration information of the target geotechnical engineering project to be constructed, and use the large language model to identify the task integration information to generate engineering data information and model configuration attribute information.

[0025] This step is used to obtain engineering data and finite element analysis tasks corresponding to the target geotechnical project, form task integration information, and use a large language model to analyze the obtained task integration information to obtain engineering data information and model configuration attribute information.

[0026] The engineering data may contain detailed information related to the target geotechnical project, including survey results, process records, construction and acceptance data. Therefore, based on this engineering data, accurate finite element design of the geotechnical project can be performed.

[0027] Once the relevant engineering data is obtained, finite element modeling can be initiated based on the assigned finite element analysis task. To achieve more accurate task confirmation, this step utilizes a large language model to identify the received task integration information, revealing the engineering data and model configuration attribute information contained within. The engineering data includes: geometric dimensions, soil layer distribution, material information, support system layout, load conditions, construction stage, and drawing annotations. The model configuration attribute information includes: derived model boundaries, retaining wall location, and support elevation, among other key model configuration attributes.

[0028] In detail, this step includes: Step S11: Receive the task integration information of the target geotechnical engineering project, and use a large language model to identify the engineering data information in the task integration information; wherein, the task integration information includes: one or more of the following: text information, engineering drawings, site profiles and design documents related to the target geotechnical engineering project; the engineering data information includes: one or more of the following: engineering data information, soil layer distribution information, material information, support system layout, load information and construction information.

[0029] The system receives natural language descriptions, engineering drawings, geological profiles, and design documents. Utilizing the multimodal capabilities of the large language model, it identifies geometric dimensions, soil layer distribution, material information, support system layout, load conditions, construction stages, and engineering information as indicated on the drawings. In practice, following the task scope rules, the current processing is limited to the engineering problem definition stage, collecting only information describing the engineering physical problem. This step does not make finite element modeling decisions; constitutive model selection, parameter determination, mesh generation, and boundary condition settings are completed in subsequent steps.

[0030] Step S12: Perform a complete information check based on the engineering data information to obtain the engineering data information after the complete information check.

[0031] The system checks geometric conditions, material descriptions, load conditions, construction stages, units, and analysis objectives according to information integrity requirements. If missing information exists, the human-computer interaction unit raises supplementary questions for the missing items; for example, if the engineering data does not specify the required output engineering response, it asks the engineer to determine the analysis objective. Upon receiving supplementary information, the integrity check is repeated until all required information is complete. The analysis objectives include retaining wall displacement, ground settlement, and safety factor evaluation.

[0032] Step S13: According to the preset model configuration rules, derive the model configuration attribute information from the engineering data information after the complete information check; the model configuration attribute information includes: the model boundary range, the simulated position of the retaining wall in the model space, and the set elevation of the supporting structure.

[0033] After all required information is filled in, key model configuration attributes such as model boundary range, simulated location of retaining structure, and layout of support system are derived according to model configuration rules.

[0034] This embodiment uses the engineering data information of geotechnical engineering to derive model configuration attribute information as an example for illustration. The engineering data that has passed the integrity check rule verification is as follows: the excavation depth of the foundation pit is 20.5 meters, the net bottom dimensions are 80 meters × 60 meters, the underground continuous wall retaining structure is adopted, the wall thickness is 1.2 meters, and three internal support systems are set up.

[0035] First, the pit bottom elevation of -16.0 meters and the ground elevation of 4.5 meters were read, and the excavation depth of 20.5 meters was calculated. Based on the preset model boundary rules, the finite element analysis range was automatically determined by combining the excavation depth and engineering dimensions. The horizontal range was expanded outwards according to the principle of not less than three times the excavation depth, determining the horizontal range of the model to be (-61.5, -61.5) to (141.5, 121.5). The vertical bottom boundary was set approximately twice the excavation depth below the pit bottom, i.e., -57.0 meters, and the top boundary was set at the natural ground elevation of 4.5 meters to ensure that the boundaries were outside the main stress disturbance zone.

[0036] Secondly, the layout data of the diaphragm wall centerline along the foundation pit boundary was extracted, with a wall thickness of 1.2 meters. Based on the finite element configuration rules, the wall's mid-surface was used as the reference position for the structural elements, and a discrete model of the wall was automatically established. Combining the top elevation and design depth of the diaphragm wall, the top elevation was derived to be 5.0 meters and the bottom elevation to be -38.0 meters. Concrete material properties were then assigned, forming a discretized simulation scheme for the retaining structure.

[0037] Furthermore, the system automatically generates an initial three-tiered support scheme based on the support system layout requirements. First, the system determines the support design elevations to be 2.0 meters, -5.0 meters, and -12.0 meters based on the pit depth and construction stage. Then, it verifies the scheme by considering the main structure floor slab layout, construction clearance requirements, and support spacing constraints. If a conflict is found between the support components and the main structure, or if the support spacing exceeds the preset range, the system automatically adjusts the relevant support positions and re-verifies the overall support system layout. If necessary, the system can generate supplementary support components according to preset rules and simultaneously determine their cross-sectional parameters and stiffness properties.

[0038] Through the above process, the model boundary range, discrete locations of the retaining structure, and layout information of the support structure are sequentially derived from the engineering data. The model configuration is automatically checked and corrected, generating the model configuration data required for finite element preprocessing. The results are then output as a structured JSON model configuration file for subsequent simulation planning, code generation, and design optimization steps. This step mainly generates two deliverables: engineering data information for engineers to read and structured JSON model configuration information. The engineering data information summarizes engineering geometry, soil layer distribution, support structure layout, load conditions, construction stages, and analysis objectives; the structured JSON model configuration saves the derived geometric attributes required by the downstream intelligent agent.

[0039] Furthermore, this step may also include: inputting the identified engineering data and model configuration attribute information into the interactive interface for display. The engineer reads and manually verifies the information; after engineer approval, the two results are written to the shared storage space for other components to read and use.

[0040] Combination Figure 2As shown, in one implementation, this step can be achieved using a pre-defined task clarification agent. The input to this agent is task integration information including text descriptions, engineering drawings, geological profiles, and other engineering data. The agent uses a large model to identify the received information and determine its completeness. If complete, it generates a task description, guides the configuration of key model parameters, and outputs the aforementioned results. Otherwise, it retrieves the missing information through a human-computer interaction interface by asking targeted questions. The implementation of this step relies not only on the agent's own information processing but also on the engineering knowledge embedded within the agent, including task scope constraints, information completeness rules, and model configuration conventions.

[0041] Step S2: Determine the structured simulation plan based on the engineering data information and model configuration attribute information.

[0042] This step, building upon step S1 above, determines the structured simulation plan. The structured simulation plan is a pre-defined, full-lifecycle technical solution for the numerical simulation analysis of the target geotechnical engineering problem to be solved in this step. It features clearly defined data interfaces, modular workflow arrangement, and quantifiable result standards. This technical solution is used to construct the simulation model, specifically including modeling assumptions, execution steps, task dependencies, and verification requirements.

[0043] In detail, the step of determining the structured simulation plan based on the engineering data information and model configuration attribute information includes: Step S21: Identify and distinguish modeling items from non-modeling items in the engineering data information and model configuration attribute information. The modeling items include one or more of the following: constitutive model identifier, material parameter values ​​and their data sources, drainage condition identifier, groundwater treatment method identifier, boundary condition definition, mesh generation control parameters, structural unit type definition, construction stage sequence, monitoring point coordinates, and result extraction item list. For example, for a foundation pit project with an excavation depth of 20.5 meters and a bottom dimension of 80 meters × 60 meters, the system determines the finite element analysis range according to the model configuration rules, obtaining the horizontal and vertical boundaries of the model. Since the above boundary parameters directly affect the mesh partitioning area, boundary condition settings, and subsequent calculation accuracy, they are identified as modeling items and included in the model boundary definition subtask. Simultaneously, the discretization location of the diaphragm wall is identified as a structural generation modeling item, and the support system layout information is identified as a construction stage modeling item, both serving as the basic input for generating the structured simulation plan.

[0044] This step reads the confirmed engineering data and model attribute configuration information from the shared storage area, identifies the modeling items that need to be determined, and distinguishes between items that can be automatically determined based on engineering data and modeling rules, and items that require engineer confirmation. The modeling items include constitutive model, material parameter values ​​and sources, drainage conditions, groundwater treatment methods, boundary conditions, mesh generation, structural unit types, construction stages, monitoring points, and result extraction items.

[0045] Step S22: Construct modeling items according to the preset modeling planning process, and retrieve corresponding planning data based on the modeling items.

[0046] Modeling tasks are organized according to the standard PLAXIS modeling workflow in the preset planning protocol. The workflow corresponding to the modeling tasks covers model initialization, soil layer definition, material allocation, structural element generation, excavation stage setting, finite element calculation, result extraction and reporting, monitoring point location, and the final strength reduction stage for safety evaluation.

[0047] During the simulation plan development process, relevant planning data is dynamically retrieved for the modeling items identified in step S21. This planning data includes technical documents in areas such as coordinate conventions, boundary condition handling, structural element definitions, and mesh generation, as well as annotated simulation planning examples that match the current engineering problem. When supplementary modeling guidance is needed, semantic searches are performed on the indexed PLAXIS documents.

[0048] For example, regarding the determination of model boundary conditions, after identifying the current project as a foundation pit excavation analysis, the system first determines the finite element model boundary range based on project data and model configuration attributes. Taking a foundation pit project with an excavation depth of 20.5 meters and a bottom net dimension of 80 meters × 60 meters as an example, the system determines the horizontal boundary range of the model to be (-61.5, -61.5) to (141.5, 121.5) and the vertical boundary range to be 4.5 meters to -57.0 meters above the natural ground level, based on preset model boundary rules. Subsequently, the system searches for planning data and modeling examples using keywords such as "foundation pit boundary conditions," "model truncation boundary," "standard fixed boundary," and "finite element boundary treatment." Based on the search results and the current model boundary range, the system determines that the model's side boundaries adopt horizontal constraints and the model's bottom boundaries adopt fixed constraints, and writes the corresponding boundary condition definitions into the structured simulation plan. For special working conditions such as dynamic analysis and seepage analysis, the system can further search for boundary treatment rules corresponding to the analysis type and generate boundary condition modeling decisions corresponding to the analysis type, which are then incorporated into the structured simulation plan.

[0049] Step S23: Determine various modeling decisions based on the planning data, and generate a structured simulation plan after verifying that the information of each modeling decision is correct.

[0050] Modeling decisions are determined based on engineering data and planning knowledge. When both indoor experimental data and empirical formulas can be used to determine the same parameter, indoor experimental data is preferred. For modeling decisions that cannot be automatically determined, clarification questions are posed to the engineer through the human-computer interaction unit, and the relevant modeling decisions are re-determined based on the engineer's supplementary information until no unresolved modeling issues remain.

[0051] For example, in a foundation pit project, the engineering data provided information on soil layer distribution, diaphragm wall dimensions, and support system layout, but did not specify the type of soil constitutive model. The simulation planning agent, by retrieving planning data, discovered that the Mohr-Coulomb model or the Hardening Soil model could be used to simulate medium-dense sand. When the engineering data only provides conventional geotechnical parameters and lacks the parameters required for an advanced constitutive model, the system preferentially recommends the Mohr-Coulomb model. If the engineer further provides triaxial test data and loading path information, the system automatically adjusts its modeling decision, selecting the Hardening Soil model and supplementing the corresponding parameter configurations, thus completing the determination of the constitutive model selection.

[0052] Before formulating the simulation plan, the interrelationships between modeling decisions, such as constitutive model selection, groundwater treatment, mesh generation, construction sequence, and monitoring point configuration, are comprehensively considered, and the geometric information is checked according to the preset modeling conventions. Once all relevant modeling decisions are finalized and the geometric information conforms to the preset modeling conventions, the simulation plan generation phase begins.

[0053] Once the modeling decisions are determined, the finite element modeling process is decomposed into sub-tasks that can be executed sequentially and verified independently, generating a structured simulation plan. For example, in the scenario of foundation pit engineering modeling, after determining the adoption of the Hardening Soil constitutive model, diaphragm wall retaining structure, three-layer support system, and phased excavation scheme, the system decomposes the overall modeling task into multiple independently verifiable sub-tasks according to execution dependencies. These include: (1) Model initialization subtask, used to determine the unit system, coordinate system and model boundary; (2) Soil layer modeling subtask, used to establish soil layer structure and configure soil material parameters; (3) Enclosure structure modeling subtask, used to create underground continuous walls and their material properties; (4) Support structure modeling subtask, used to create support components and connection relationships; (5) Modeling subtasks during the construction phase, used to define the excavation sequence, support activation sequence, and load application rules; (6) Calculation and result extraction subtask, used to perform finite element solution and extract results such as displacement, internal force and safety factor.

[0054] The system further automatically generates a structured simulation plan based on the dependencies between the aforementioned subtasks. Specifically, the soil modeling subtask depends on the model initialization subtask, the retaining structure modeling subtask depends on the soil modeling subtask, and the construction phase modeling subtask depends on the retaining structure and support structure modeling subtasks. This ultimately results in a structured simulation plan that includes modeling assumptions, execution steps, task dependencies, verification rules, and result extraction requirements.

[0055] Additionally, in this step, the structured simulation plan can be presented to the engineer via a human-computer interaction interface, and confirmation or modification feedback can be received. After the engineer's approval, the simulation planning agent writes the approved structured simulation plan in ordered JSON format to a storage space where information can be shared.

[0056] Combination Figure 3 As shown, this step is based on the constitutive model, parameter sources, computational domain, boundary conditions, groundwater treatment, mesh, structural elements, construction stages, monitoring points, and result extraction methods obtained in step S1 above, as well as the resulting modeling assumptions and simulation plan. In one implementation, this step is implemented using a simulation planning agent, whose input is engineering data information (i.e.,... Figure 3 The task description (including information such as geometry, materials, loads, units, and construction stages) and model attribute configuration information (i.e., model configuration JSON) shown are used by the simulation planning agent to make finite element modeling decisions based on the received information. During decision identification, the agent reviews the decisions according to task scope constraints, defining only the engineering problem. Undetermined or ambiguous modeling decisions are resolved through a human-computer interaction interface. Finally, a simulation plan is generated and output. This step requires not only the simulation planning agent to perform the corresponding data processing steps, but also to store simulation planning knowledge, including: planning protocols, technical reference materials, planning examples, and retrieval tools.

[0057] Step S3: Generate finite element software interface code according to the structured simulation plan, and perform benchmark finite element model construction and numerical calculation. Compare the calculation results of the numerical calculation with the preset standard results to obtain the compliance evaluation results.

[0058] This step generates finite element software interface code step by step according to the structured simulation plan obtained in step S2 above. When generating the code for the current step, the model state summary formed in the previous step is read, and the implementation rules, reference code, and finite element software interface data corresponding to the current step are loaded. The generated code is executed through the script interface of the finite element software, and the success of execution is determined based on the execution log, exception tracking information, and the verification requirements of the current step. If execution fails, relevant implementation knowledge is retrieved based on the execution diagnosis, the affected code is regenerated, and executed again. If execution is successful, the verified code is saved and the model state summary is recorded as the context for generating code in the next step. After the finite element calculation is completed, the calculation results specified in the structured simulation plan are extracted. The code generation agent compares the calculation results with the control requirements in the applicable design standards or engineering requirements to form a compliance evaluation result.

[0059] In detail, the steps of generating finite element software interface code based on the structured simulation plan and performing baseline finite element model construction and numerical calculation include: Step S31: Obtain the structured simulation plan and model configuration attribute information, determine the current simulation step and the prerequisites for each simulation step based on the structured simulation plan and model configuration attribute information, and read the model state summary formed by the current simulation step.

[0060] The system reads the approved structured simulation plan and structured model configuration from the shared storage space, determines the current simulation step and its prerequisites, and, if any steps have been successfully executed, simultaneously reads the model state summary generated by those steps. For example, in a foundation pit project, the structured simulation plan includes multiple simulation steps such as model initialization, soil layer modeling, material definition, retaining structure modeling, support structure modeling, construction stage setting, and result extraction. Specifically, the soil layer modeling step depends on the model initialization step; the material definition step depends on the soil objects generated in the soil layer modeling step; the retaining structure modeling step depends on the material objects generated in the material definition step; the support structure modeling step depends on the structural objects generated in the retaining structure modeling step; the construction stage setting step depends on the completion of all structural objects; and the result extraction step depends on the completion of finite element calculations.

[0061] When the system executes the retaining structure modeling step, it automatically identifies its prerequisites as model initialization, soil layer modeling, and material definition steps, and reads the model state summary generated by these steps from the shared storage space. The model state summary includes created soil layer objects, material objects, object names, object reference relationships, and spatial coordinate information. The code generation agent generates corresponding finite element software interface code based on this model state summary, ensuring that the newly generated code can directly reference existing objects without recreating related model entities. The model state summary records established geometric objects, material objects, structural objects, construction stages, and their object references, ensuring that the code generation for the current step is based on the model state already formed in the finite element software.

[0062] Step S32: Load the corresponding implementation rules from the pre-built practice rule library according to the current simulation step; wherein, the practice rule library is the implementation rules saved according to the modeling task, and the modeling task includes: geometry construction, constitutive model setting, soil-structure interaction, construction stage setting and result extraction.

[0063] According to the implementation steps in the current simulation plan, the corresponding implementation rules are loaded from the practice rule base. The practice rule base stores implementation rules according to modeling tasks such as geometric construction, constitutive model setting, soil-structure interaction, construction stage setting, and result extraction; each code generation only loads the implementation rules related to the current simulation step.

[0064] Step S33: Retrieve a verified code pattern from the reference code library that matches the current simulation step.

[0065] Retrieve verified code patterns from the reference code library that match the current simulation steps; when additional specific interface call guidance is needed, perform semantic searches on finite element software interface documents and technical reference materials.

[0066] Step S34: Combine the current simulation step, prerequisites, model state summary, implementation rules, reference code pattern, and interface information into a code generation context to generate finite element software interface code for completing the current simulation step.

[0067] The current simulation step, prerequisites, model state summary, implementation rules, reference code patterns, and interface information are combined into a code generation context to generate finite element software interface code for completing the current simulation step. The code generation agent uses a structured model configuration and the modeling assumptions, engineering parameters, execution order, and task dependencies recorded in the simulation plan. When the current step requires converting engineering parameters into finite element software input, the agent calculates component coordinates, equivalent parameters, or construction stage settings based on the aforementioned information.

[0068] Step S35: Execute the finite element software interface code, save the finite element software interface code corresponding to the current step, record the model state summary formed in this step, and use the model state summary as the code generation context for the next simulation step, until all modeling steps are completed.

[0069] In detail, the step of executing the finite element software interface code and saving the finite element software interface code corresponding to the current step further includes: Step S351: Collect logs and exception tracking information after the finite element software interface code is executed in the current simulation step.

[0070] Step S352: Determine whether the current simulation step has been successfully executed based on the verification requirements corresponding to the current simulation step in the structured simulation plan.

[0071] The execution and correction process in the above steps continues until the current step is successfully executed or the preset maximum number of corrections is reached. If the execution is still unsuccessful after reaching the preset maximum number of corrections, the execution diagnosis and correction history is displayed to the engineer through the human-computer interaction interface, and the engineer's processing opinions are received. The current step is then corrected and executed again based on the processing opinions.

[0072] Step S353: If the current simulation step fails, perform a diagnosis on the interface document, technical reference information and implementation example corresponding to the finite element software interface code, regenerate the updated code after the diagnosis, and execute it.

[0073] After the current step is executed successfully, the verified code for this step is saved, a model state summary generated in this step is recorded, and this model state summary is used as the code generation context for the next simulation step. This step selects the next simulation step sequentially according to the structured simulation plan and repeats the above steps until all modeling steps are completed.

[0074] The code for the current step is executed via the script interface of the finite element software. Execution logs and exception tracking information are collected, and the success of the current step is determined based on the verification requirements corresponding to the current step in the structured simulation plan. If the current step fails, the code generation agent retrieves relevant finite element software interface documents, technical references, and implementation examples based on the execution diagnostics, regenerates the affected code, and executes it again.

[0075] Step S36: Execute each calculation stage specified in the structured simulation plan, determine whether each calculation stage has met the preset convergence requirements, and extract the calculation results specified in the structured simulation plan.

[0076] Once all modeling steps are completed, the code-generating agent calls the finite element software to execute each calculation stage specified in the structured simulation plan, and checks whether each calculation stage has completed the calculation and met the convergence requirements. If a calculation stage fails, the corresponding execution diagnostics are read, and corrections are made according to the method proposed in step S35; if the calculation is still not completed after reaching the preset number of corrections, an engineer is requested to handle the issue through the human-computer interaction interface.

[0077] After all required calculations are completed, the calculation results specified in the structured simulation plan are extracted through the output interface of the finite element software. These calculation results include at least one of the following: safety factor, surface settlement, support structure displacement, and internal forces of the structure.

[0078] Based on applicable geotechnical engineering design standards or engineering requirements, the calculation results output by the finite element software are compared with the corresponding control requirements to form a compliance evaluation result. The compliance evaluation result includes evaluation indicators, finite element calculation values, corresponding control requirements, and a conclusion on whether the requirements are met. The code generation agent writes the finite element calculation results, the compliance evaluation result, and the model access identifier corresponding to the verified finite element model into the shared state for the design optimization agent to read.

[0079] like Figure 4 As shown, in one implementation, this step can be achieved using a code generation agent. The input information for this agent is a simulation plan, which includes modeling assumptions, execution order, parameters, code templates, task dependencies, and prior model context. The code generation agent processes the received simulation plan, generates the code for the current step based on the plan, context, and retrieved knowledge, automatically runs the script, and collects execution logs. It verifies the execution results, analyzes the execution logs and anomaly tracking, and determines whether the execution was successful or failed. The agent can also retrieve documents or relevant examples, regenerate affected code, perform automatic error correction, and finally output and store the baseline simulation model and execution summary generated in this step. The code generation agent also needs to possess corresponding code generation knowledge, including a practice rule base, a reference script library, and Plaxis documentation related to the geotechnical finite element software PLAXIS.

[0080] Step S4: Perform parameter optimization and structural configuration screening based on the constructed benchmark finite element model and the compliance evaluation results to obtain the optimized finite element model corresponding to the target geotechnical engineering.

[0081] This step begins by reading the validated benchmark finite element model, its execution context, and compliance evaluation results to identify optimization variables, feasible parameter boundaries, and candidate parameterization locations, thereby determining an optimization problem with the goal of minimizing engineering costs and constraints of safety and normal use requirements.

[0082] Secondly, after the engineer confirms the optimization variables, parameter boundaries, and related assumptions, the Bayesian optimization module is called to generate candidate parameter combinations. The candidate parameters are written into the finite element model and calculations are performed. Based on the compliance of the finite element response evaluation constraints, the project cost is calculated through the parameterized cost model, and the evaluation results are fed back to the Bayesian optimization module until the convergence conditions of continuous parameter optimization are met, thus obtaining the optimal solution under this structural configuration.

[0083] This step searches for continuous design parameters through Bayesian optimization and explores different structural configurations based on finite element calculation results and safety margins. This expands the design optimization from parameter adjustment under fixed configurations to a joint search of discrete structural configurations and continuous design parameters, improving the efficiency of design space exploration and the economy of engineering solutions.

[0084] Furthermore, after the continuous parameter optimization converges, the design optimization agent determines whether there are still structural design variables with engineering significance based on the finite element response and safety margin of the current optimal feasible solution. If the support components can be further reduced or reconfigured while satisfying the constraints, the design optimization agent proposes candidate structural configurations.

[0085] Finally, the finite element model and optimization space are updated, the surrogate model is reinitialized, and continuous parameter optimization is performed again under the new structural configuration. If none of the sampled schemes under the new structural configuration meet the constraints, the previous feasible structural configuration is restored and the structural configuration exploration is terminated. This achieves the joint search of structural configuration and continuous design parameters.

[0086] In detail, the steps of performing parameter optimization and structural configuration screening based on the constructed benchmark finite element model and the conformity evaluation results to obtain the optimized finite element model corresponding to the target geotechnical engineering project include: Step S41: Obtain the constructed benchmark finite element model, the execution context and evaluation results of the benchmark finite element model, and obtain and load parameter information for optimizing the workflow; wherein, the parameter information includes parameter dependencies, parameter adjustment priorities and search space hierarchical operations.

[0087] Based on the baseline finite element model constructed in step S3 above, its execution context, and compliance evaluation results, and by loading the optimization workflow, parameter adjustment priority, and search space management rules, the baseline finite element model is analyzed to identify optimization variables, feasible parameter boundaries, and candidate parameterization locations.

[0088] Step S42: By analyzing the finite element model, determine the continuous design variables, feasible parameter boundaries, and preset dependencies between variables, and determine the benchmark scheme with minimizing the project cost as the optimization objective and preset constraints.

[0089] By analyzing the baseline finite element model and combining engineering constraints and optimization knowledge, the continuous design variables, feasible parameter boundaries, and pre-defined dependencies between variables are determined. When the continuous design variables change, the dependent variables are updated synchronously according to the pre-defined dependencies.

[0090] Step S43: Under the condition of the structural configuration of the benchmark scheme, establish a surrogate model based on the preset compliance evaluation results, and propose candidate parameter combinations within the preset parameter space.

[0091] The optimization objectives and constraints are determined based on the design requirements. The optimization objective is to reduce project costs, and the constraints include safety and normal operation requirements specified in applicable design standards or engineering requirements. A human-computer interaction interface is used to display continuous design variables, parameter boundaries, variable dependencies, optimization objectives, and constraints, and to receive confirmation or modification feedback from engineers.

[0092] The benchmark scheme is evaluated, and its design parameters, engineering cost, finite element response, and constraint margin are recorded. The parametric cost model calculates the quantities of the corresponding components based on the design parameters, determines the direct engineering cost in conjunction with the unit price of the project, and adjusts it according to the indirect costs of the project; the constraint evaluation is performed in accordance with applicable design standards or engineering requirements.

[0093] Under the condition that the current structural configuration remains unchanged, the Bayesian optimization module establishes a surrogate model based on the existing evaluation results and uses the acquisition function to propose candidate parameter combinations in the preset parameter space, so as to search for low-cost feasible solutions within a limited number of finite element calculations.

[0094] Step S44: Input the candidate parameter combination into the constructed finite element model to obtain the candidate model, and execute it to extract the finite element response output by the candidate model corresponding to the constraint conditions.

[0095] The candidate parameter combination is written into the candidate parameterization position determined in step S41. The relevant model input is updated synchronously according to the variable dependency relationship. The script interface of the finite element software is called to execute the candidate model and extract the finite element response corresponding to the constraint conditions. The parameterized cost model calculates the engineering cost of the candidate scheme at the same time.

[0096] Step S45: Obtain the finite element response corresponding to each candidate model and determine the optimal design finite element scheme under the current structural configuration.

[0097] The engineering cost and constraint evaluation results of candidate solutions are fed back to the Bayesian optimization module. Candidate solutions that satisfy all constraints are recorded as feasible samples; those that do not satisfy constraints are penalized according to the degree of constraint violation. The Bayesian optimization module updates the surrogate model based on feasible and infeasible samples.

[0098] Record the optimal feasible solution and its engineering cost under the current structural configuration. When the relative improvement of the optimal feasible cost in three consecutive feasible iterations is no greater than 1%, the continuous parameter optimization under the current structural configuration is considered to have converged, and the optimal design finite element scheme under the current structural configuration is obtained. If the convergence condition is not met, the Bayesian optimization module establishes a surrogate model based on the existing evaluation results under the condition that the current structural configuration remains unchanged, and uses the acquisition function to propose candidate parameter combinations in the preset parameter space to search for low-cost feasible solutions within a limited number of finite element calculations.

[0099] Step S46: Based on the optimal design finite element scheme, determine whether there are multiple structural design variables.

[0100] Once the optimal finite element design scheme corresponding to the current structural framework is obtained, it is determined whether there are multiple structural framework schemes and which structural framework scheme is the optimal scheme.

[0101] In detail, the step of determining whether there are multiple structural design variables based on the optimal design finite element scheme includes: Step S461: Analyze the finite element response and safety margin of the optimal design finite element scheme under the current structural configuration. Determine whether there are structural design variables, and propose multiple candidate structural configurations while satisfying the constraints.

[0102] After continuous parameter optimization of the new structural configuration, the design optimization agent determines whether the structural configuration can form a feasible solution that satisfies all constraints. If a feasible solution exists, the optimization result of the structural configuration is recorded, and the agent continues to determine whether there are other explorable structural configurations. If all sampled solutions under the new structural configuration do not satisfy the constraints, the previous feasible structural configuration is restored and the structural configuration exploration is terminated.

[0103] Step S462: Calculate the optimization results under each candidate structural configuration, and determine the optimal structural configuration and its continuous design parameters, engineering cost, cost change relative to the benchmark scheme, convergence history and evaluation results based on the optimized structure under each candidate structural configuration.

[0104] The system summarizes the optimization results for each structural configuration, outputting the optimal structural configuration and its continuous design parameters, project cost, cost change relative to the benchmark scheme, convergence history, structural configuration exploration results, and compliance evaluation results. The optimization results are displayed using a human-computer interaction interface, and engineers' confirmation opinions on the final design scheme are received.

[0105] Step S47: Based on multiple structural design variables, determine the geotechnical finite element model corresponding to the optimal mechanism configuration.

[0106] Once the optimal structural design variables are determined, the optimal design result is obtained, and the optimal configuration, continuous design parameters, engineering cost, structural configuration exploration structure, and evaluation results are output.

[0107] Combination Figure 5 As shown, in practical implementation, the data processing flow in step S4 can be implemented using a design optimization agent. The inputs to this agent include a benchmark simulation model, compliance summary, benchmark performance indicators, and project constraints. The design optimization agent sequentially performs benchmark analysis, displays information through an interactive interface and receives confirmation or adjustment information from engineers, evaluates the benchmark scheme corresponding to the benchmark simulation model, executes a Bayesian optimization loop, performs finite element analysis, conducts cost and compliance evaluation, performs optimization updates and convergence checks, generates an optimization summary, and outputs the optimal design scheme. This design optimization agent stores code generation-related knowledge, including optimization protocols, operational knowledge, and procedural knowledge.

[0108] Furthermore, the method also includes: All data generated during the construction and optimization of the finite element model are stored in the shared space. The data generated during the construction and optimization of the finite element model includes one or more of the following: engineering data information and model configuration attribute information, structured simulation plan, model status summary, model access identifier, code generation results, finite element calculation results and compliance evaluation results, finite element model, candidate parameters, finite element response, and optimization history.

[0109] In this step, a preset data model is used to save the structured results generated by each agent, including task description, structured model configuration, simulation plan, code generation results, model state summary, model access identifier, finite element calculation results, compliance evaluation results, optimization records, and agent state information. Each type of result is written to the state field corresponding to its processing stage, and each agent only accesses the designated field corresponding to its responsibility.

[0110] Secondly, this invention provides a finite element modeling and optimization system for geotechnical engineering based on a large language model, such as... Figure 6 As shown, it includes: The task clarification agent module 100 is used to acquire task integration information of the target geotechnical engineering to be constructed, and to use a large language model to identify the task integration information to generate engineering data information and model configuration attribute information; its function is as described in step S1.

[0111] The simulation planning intelligent agent module 200 is used to determine a structured simulation plan based on the engineering data information and model configuration attribute information; its function is as described in step S2.

[0112] The code generation intelligent agent module 300 is used to generate finite element software interface code according to the structured simulation plan, and to perform benchmark finite element model construction and numerical calculation. The calculation results of the numerical calculation are compared with preset standard results to obtain compliance evaluation results; its function is as described in step S3.

[0113] The design optimization intelligent agent module 400 is used to perform parameter optimization and structural configuration screening based on the constructed benchmark finite element model and the compliance evaluation results, so as to obtain the optimized finite element model corresponding to the target geotechnical engineering. Its function is as described in step S4.

[0114] This system also includes a human-computer interaction unit, which performs different functions at different stages, specifically including: During the task clarification phase, the human-computer interaction unit is used to display identified engineering information, missing items, task descriptions, and structured model configurations. When task information is incomplete, the human-computer interaction unit receives supplementary engineering data from engineers; after the task descriptions and structured model configurations are finalized, it receives confirmation or modification comments from engineers.

[0115] During the simulation planning phase, the human-computer interaction unit is used to display the constitutive model, parameter sources, computational domain, boundary conditions, groundwater treatment, mesh, structural elements, construction stages, monitoring points, and result extraction methods, as well as the resulting modeling assumptions and simulation plan. After engineer confirmation or modification, the simulation planning agent uses the approved simulation plan as the basis for execution during the code generation phase.

[0116] During the code generation phase, the human-computer interaction unit is used to automatically handle errors that can be corrected based on execution feedback within a preset number of corrections. If the code generation agent still fails to execute successfully after reaching the preset number of corrections, the human-computer interaction unit displays the failure steps, execution diagnosis, retrieved interface information and correction history, and receives interface calls, object references or modeling strategy processing opinions provided by engineers.

[0117] Before design optimization begins, the human-computer interaction unit (HCI) displays the optimization objectives, constraints, continuous design variables, parameter boundaries, and variable dependencies, and receives confirmation or modification feedback from engineers. During continuous parameter optimization, the HCI displays the current optimal parameters, finite element response, engineering cost, and convergence history. When the design optimization agent proposes structural configuration adjustment schemes, the HCI displays candidate structural configurations, presents the finite element calculation results, control indicators, and safety margins upon which the configuration is based; engineers can confirm or modify optimization settings, terminate structural configuration exploration, and confirm the final design scheme.

[0118] Combination Figure 7As shown, in practical implementation, this system includes a human-computer interaction unit, a shared information unit, and intelligent agents composed of various language models with different functions. The human-computer interaction unit receives information acquired and generated by each intelligent agent and feeds the received information back to the operator through a display interface, receiving confirmation and correction information from the operator. The shared information unit receives information output by each intelligent agent and shares and stores this information to facilitate information interaction between the agents, thereby automatically realizing the framework and design optimization of the finite element model.

[0119] The method and system provided by this invention can achieve continuous automated processing of finite element task definition, simulation planning, model generation, calculation verification, result evaluation, and design optimization in geotechnical engineering. This reduces manual work required for model construction, code debugging, result processing, and trial calculations, thereby improving the efficiency of finite element analysis and design optimization. Furthermore, by generating and executing finite element software interface code step-by-step, correcting errors based on execution feedback, and transferring the successfully established model state to subsequent steps, the executability of complex finite element model code is improved, and the impact of object referencing, parameter passing, and inconsistencies in construction stages on model reliability is reduced.

[0120] Furthermore, the method and system of this invention record task descriptions, simulation plans, code generation results, model status, calculation results, and optimization processes in a structured shared state record. Engineers are introduced to review the process at nodes such as task confirmation, plan approval, handling of continuous failures, confirmation of optimization settings, and confirmation of the final solution. This enables the automatic finite element modeling and design optimization process to be inspected and traced, thereby improving the transparency and interpretability of the system.

[0121] Thirdly, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a computer, the computer is used to execute the aforementioned geotechnical engineering finite element modeling and optimization method based on a large language model.

[0122] This invention discloses a method, system, and storage medium for finite element modeling and optimization of geotechnical engineering based on a large language model. The method involves acquiring task integration information of the target geotechnical engineering project to be constructed, and using a large language model to identify this information to generate engineering data and model configuration attribute information. A structured simulation plan is determined based on the engineering data and model configuration attribute information. Finite element software interface code is generated based on the structured simulation plan, and a baseline finite element model is constructed and numerical calculations are performed. The calculation results of the numerical calculations are compared with preset standard results to obtain a conformity evaluation result. Parameter optimization and structural configuration screening are performed based on the constructed baseline finite element model and the conformity evaluation result to obtain an optimized finite element model corresponding to the target geotechnical engineering project. The method and system provided by this invention automatically construct and optimize finite element models by executing finite element software interface code step by step and transferring the completed model state to subsequent steps. This reduces manual workload in model construction, code debugging, or scheme optimization, and improves model executability, process traceability, design space search efficiency, and reduces costs.

[0123] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0124] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0125] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0126] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0127] The embodiments described above are merely examples of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application.

Claims

1. A finite element modeling and optimization method for geotechnical engineering based on a large language model, characterized in that, include: Obtain task integration information for the target geotechnical engineering project to be constructed, and use a large language model to identify the task integration information to generate engineering data information and model configuration attribute information; The structured simulation plan is determined based on the engineering data and model configuration attribute information; The structured simulation plan generates finite element software interface code, and performs benchmark finite element model construction and numerical calculation. The calculation results of the numerical calculation are compared with the preset standard results to obtain the conformity evaluation results. Based on the constructed benchmark finite element model and the compliance evaluation results, parameter optimization and structural configuration screening are performed to obtain the optimized finite element model corresponding to the target geotechnical engineering.

2. The finite element modeling and optimization method for geotechnical engineering based on a large language model according to claim 1, characterized in that, The steps of acquiring task integration information of the target geotechnical engineering project to be constructed, and using a large language model to identify the task integration information to generate engineering data information and model configuration attribute information include: The system receives task integration information for the target geotechnical engineering project and uses a large language model to identify engineering data information within the task integration information. The task integration information includes one or more of the following: textual information related to the target geotechnical engineering project, engineering drawings, geological profiles, and design documents. The engineering data information includes one or more of the following: engineering data information, soil layer distribution information, material information, support system layout, load information, and construction information. A complete information check is performed on the engineering data information to obtain the engineering data information after the complete information check. According to the preset model configuration rules, model configuration attribute information is derived from the engineering data information after complete information check; the model configuration attribute information includes: model boundary range, simulated position of the retaining wall in the model space, and set elevation of the supporting structure.

3. The finite element modeling and optimization method for geotechnical engineering based on a large language model according to claim 1, characterized in that, The step of determining the structured simulation plan based on the engineering data information and model configuration attribute information includes: The engineering data and model configuration attribute information are used to identify modeling items and distinguish between modeling items and non-modeling items. The modeling items include one or more of the following: constitutive model identifier, material parameter values ​​and their data sources, drainage condition identifier, groundwater treatment method identifier, boundary condition definition, mesh generation control parameters, structural unit type definition, construction stage sequence, monitoring point coordinates, and result extraction item list. Modeling items are constructed according to the preset modeling and planning process, and corresponding planning data are retrieved based on the modeling items; Based on the planning data, various modeling decisions are determined, and a structured simulation plan is generated after verifying that the information of each modeling decision is correct.

4. The finite element modeling and optimization method for geotechnical engineering based on a large language model according to claim 1, characterized in that, The steps of generating finite element software interface code according to the structured simulation plan and performing baseline finite element model construction and numerical calculation include: Obtain the structured simulation plan and model configuration attribute information, determine the current simulation step and the prerequisites for each simulation step based on the structured simulation plan and model configuration attribute information, and read the model state summary formed by the current simulation step. According to the current simulation steps, the corresponding implementation rules are loaded from the pre-built practice rule library; wherein, the practice rule library consists of implementation rules saved according to the modeling tasks, and the modeling tasks include: geometry construction, constitutive model setting, soil-structure interaction, construction stage setting, and result extraction; Based on the verified code patterns retrieved from the reference code library that match the current simulation steps; The current simulation step, prerequisites, model state summary, implementation rules, reference code pattern, and interface information are combined into a code generation context to generate finite element software interface code for completing the current simulation step. Execute the finite element software interface code, save the finite element software interface code corresponding to the current step, record the model state summary formed in this step, and use the model state summary as the code generation context for the next simulation step, until all modeling steps are completed. Execute each calculation stage specified in the structured simulation plan, determine whether each calculation stage has met the preset convergence requirements, and extract the calculation results specified in the structured simulation plan.

5. The finite element modeling and optimization method for geotechnical engineering based on a large language model according to claim 4, characterized in that, The step of executing the finite element software interface code and saving the finite element software interface code corresponding to the current step further includes: Collect logs and exception tracking information after the finite element software interface code is executed in the current simulation step; Determine whether the current simulation step was executed successfully based on the verification requirements corresponding to the current simulation step in the structured simulation plan; If the current simulation step fails, then the interface documentation, technical reference information and implementation examples corresponding to the finite element software interface code are diagnosed, and the updated code after diagnosis is regenerated and executed.

6. The finite element modeling and optimization method for geotechnical engineering based on a large language model according to claim 1, characterized in that, The steps of performing parameter optimization and structural configuration screening based on the constructed benchmark finite element model and the conformity evaluation results to obtain the optimized finite element model corresponding to the target geotechnical engineering project include: The system acquires the constructed baseline finite element model, its execution context, and evaluation results, and acquires and loads parameter information for optimizing the workflow; wherein the parameter information includes parameter dependencies, parameter adjustment priorities, and search space hierarchical operations. By analyzing the finite element model, the continuous design variables, feasible parameter boundaries, and pre-defined dependencies between variables are determined. The baseline scheme is determined with minimizing the project cost as the optimization objective and pre-defined constraints. Under the condition of the structural configuration of the benchmark scheme, a surrogate model is established based on the preset compliance evaluation results, and candidate parameter combinations are proposed within the preset parameter space; The candidate parameter combination is input into the constructed finite element model to obtain the candidate model, and then executed to extract the finite element response corresponding to the constraint conditions output by the candidate model. Obtain the finite element response corresponding to each candidate model, and determine the optimal design finite element scheme under the current structural configuration; Based on the optimal design finite element scheme, determine whether there are multiple structural design variables; Based on multiple structural design variables, the geotechnical finite element model corresponding to the optimal mechanism configuration is determined.

7. The finite element modeling and optimization method for geotechnical engineering based on a large language model according to claim 6, characterized in that, The step of determining whether there are multiple structural design variables based on the optimal finite element design scheme includes: Analyze the finite element response and safety margin of the optimal design finite element scheme under the current structural configuration, determine whether there are structural design variables, and propose multiple candidate structural configurations under the constraint conditions. The optimization results under each candidate structural configuration are calculated, and based on the optimized structure under each candidate structural configuration, the optimal structural configuration and its continuous design parameters, engineering cost, cost change relative to the benchmark scheme, convergence history and evaluation results are determined.

8. The finite element modeling and optimization method for geotechnical engineering based on a large language model according to claim 1, characterized in that, The method further includes: All data generated during the construction and optimization of the finite element model are stored in the shared space. The data generated during the construction and optimization of the finite element model includes one or more of the following: engineering data information and model configuration attribute information, structured simulation plan, model status summary, model access identifier, code generation results, finite element calculation results and compliance evaluation results, finite element model, candidate parameters, finite element response, and optimization history.

9. A finite element modeling and optimization system for geotechnical engineering based on a large language model, characterized in that, include: The task clarification agent module is used to acquire task integration information of the target geotechnical engineering to be constructed, and to use a large language model to identify the task integration information to generate engineering data information and model configuration attribute information. The simulation planning intelligent agent module is used to determine the structured simulation plan based on the engineering data information and model configuration attribute information; The code generation intelligent agent module is used to generate finite element software interface code according to the structured simulation plan, and to perform benchmark finite element model construction and numerical calculation. The calculation results of the numerical calculation are compared with preset standard results to obtain compliance evaluation results. The design optimization intelligent agent module is used to perform parameter optimization and structural configuration screening based on the constructed benchmark finite element model and the compliance evaluation results, so as to obtain the optimized finite element model corresponding to the target geotechnical engineering.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a computer, enables the computer to perform the geotechnical engineering finite element modeling and optimization method based on a large language model as described in any one of claims 1-8.

Images