VFEAgent: A Multimodal Agent Framework for End-to-End Automated Finite Element Analysis

VFEAgent：用于端到端自动化有限元分析的多模态智能体框架

Abstract: Finite Element Analysis (FEA) serves as the cornerstone of modern engineering design. However, its workflow is inherently complex and relies heavily on domain expertise. Although recent efforts have integrated Large Language Models (LLMs) into FEA, existing approaches face limitations in handling multimodal inputs and executing complex tasks.

摘要： 有限元分析（FEA）是现代工程设计的基石。然而，其工作流程本质上非常复杂，且高度依赖领域专业知识。尽管近期的研究尝试将大语言模型（LLMs）集成到 FEA 中，但现有方法在处理多模态输入和执行复杂任务方面仍面临局限性。

To address these limitations, we propose VFEAgent, an end-to-end multi-agent system designed to automate FEA modeling and simulation directly from input images and problem descriptions. Our methodology integrates two core components: (1) a multimodal vision-language multi-agent pipeline that employs ReAct-driven reasoning to extract structured FEA specifications from heterogeneous inputs and (2) a verification-first code synthesis framework, incorporating robust self-debugging and fallback mechanisms to ensure executability and physical validity.

为了解决这些局限性，我们提出了 VFEAgent，这是一个端到端的多智能体系统，旨在直接根据输入的图像和问题描述实现 FEA 建模与仿真的自动化。我们的方法集成了两个核心组件：（1）一个多模态视觉-语言多智能体流水线，利用 ReAct 驱动的推理从异构输入中提取结构化的 FEA 规范；（2）一个“验证优先”的代码合成框架，结合了强大的自我调试和回退机制，以确保代码的可执行性和物理有效性。

We systematically evaluated the system across various engineering mechanics scenarios. The results demonstrate that VFEAgent achieves a high success rate in generating complete and physically valid simulations, outperforming LLM-based baseline methods in reliability and correctness. These findings validate the feasibility of automating the complete FEA workflow, highlighting the framework’s potential to liberate engineers from tedious manual analysis.

我们在各种工程力学场景中对该系统进行了系统性评估。结果表明，VFEAgent 在生成完整且物理有效的仿真方面实现了高成功率，在可靠性和正确性上均优于基于 LLM 的基准方法。这些发现验证了自动化完整 FEA 工作流程的可行性，凸显了该框架在将工程师从繁琐的手动分析中解放出来的巨大潜力。

Paper Details:

Authors: Jiachen Zhang (1 and 2), Junyi Lao (1), Chenghao Liu (1), Siyuan Liu (1), Shixin Wu (1), Linsen Zhang (1), Boyu Wang (1), Songfang Huang (1)
Affiliations: (1) Peking University, (2) China Agricultural University
arXiv ID: 2605.28978
Submission Date: 27 May 2026

论文详情：

作者： Jiachen Zhang (1, 2), Junyi Lao (1), Chenghao Liu (1), Siyuan Liu (1), Shixin Wu (1), Linsen Zhang (1), Boyu Wang (1), Songfang Huang (1)
机构： (1) 北京大学, (2) 中国农业大学
arXiv ID: 2605.28978
提交日期： 2026年5月27日