Development of an LLM-Driven Abaqus MCP-Integrated Finite Element Analysis Automation System with Hallucination Suppression and Performance Improvement

Eun Hyun Kim; Jae Hyuk Lim

doi:10.7734/COSEIK.2026.39.2.123

All Issue

2026 Vol.39, Issue 2 Preview Page Next Page

Research Paper

Development of an LLM-Driven Abaqus MCP-Integrated Finite Element Analysis Automation System with Hallucination Suppression and Performance Improvement LLM 기반 Abaqus MCP 연동 유한요소해석 자동화 시스템 개발과 환각 억제 및 성능 향상에 관한 연구

30 April 2026. pp. 123-135

PDF XML

Abstract

This paper proposes an agentic automation system in which a large language model (LLM)-based agent supports an automated finite element analysis (FEA) workflow, spanning from problem definition to final result derivation, using the commercial software Abaqus. To implement this system, an integrated pipeline was established connecting an LLM agent (Claude Code) and Abaqus via the Model Context Protocol (MCP). Existing LLM-based automation has often been limited to mere code execution, leading to discrepancies with the user’s design intent or hallucination issues where parameters are arbitrarily assumed. To address these limitations, this study introduces a ‘Logical Constraint Mechanism’ that formalizes parameters and enforces user confirmation at the prompt input stage, thereby effectively mitigating hallucinations and improving the intent reproducibility of the analysis. Furthermore, a diagnostic document-based ‘Self-correction loop’ was implemented to enable the agent to identify and rectify errors occurring during the analysis process. Performance validation through representative static and modal analyses, as well as a highly complex wind turbine blade case, quantitatively demonstrated that as experiential data accumulated, the number of error recovery loops decreased from five to zero, and the total execution time was reduced by 90%. This confirms a significant optimization in execution performance based on empirical knowledge. Ultimately, this study presents a concrete technical framework for utilizing LLMs beyond simple code generation tools, toward reliable engineering automation.

Keywords

large language model

finite element analysis

Abaqus

model context protocol

automation

hallucination suppression

본 논문에서는 대규모 언어모델(LLM) 기반 에이전트를 활용하여 상용 유한요소해석(FEA) 소프트웨어인 Abaqus의 해석 절차를 자동화하는 에이전틱(Agentic) 시스템을 제안한다. 이를 위해 LLM 에이전트(Claude Code)와 Abaqus를 모델 컨텍스트 프로토콜(MCP) 규격으로 연동하는 통합 파이프라인을 구축하였다. 기존의 LLM 기반 자동화는 단순한 코드 실행 수준에 머물러 사용자의 설계 의도와 불일치하거나 모델이 임의로 파라미터를 설정하는 환각(Hallucination) 현상 등의 한계가 존재하였다. 이를 해결하기 위해 본 연구는 프롬프트 입력 단계에서 핵심 파라미터를 정형화하고 사용자 확인 절차를 포함하는 논리적 제약 메커니즘(Logical Constraint Mechanism)을 도입하여 환각을 효과적으로 억제하고 해석 결과의 의도 재현성을 향상시켰다. 또한 해석 과정에서 발생하는 오류를 진단 문서 기반으로 식별하고 수정하는 자기 수정 루프(Self-correction loop)를 구현하였다. 대표적인 정적 및 모달 해석 사례와 고난도 풍력 블레이드 사례를 통해 시스템을 검증한 결과, 경험적 지식 데이터가 축적됨에 따라 오류 복구 루프 횟수는 5회에서 0회로 감소하였고 전체 소요 시간은 90% 단축되었다. 이러한 결과는 제안한 프레임워크가 경험 기반의 수행 효율 향상에 효과적임을 보여준다. 본 연구는 LLM을 단순한 코드 생성 도구를 넘어 신뢰성 있는 공학 자동화 체계로 확장하기 위한 구체적인 기술적 프레임워크를 제시한다.

키워드

대규모 언어모델(LLM)

유한요소해석(FEA)

아바쿠스(Abaqus)

모델 컨텍스트 프로토콜(MCP)

자동화

환각 억제

References

Alansari, A., Luqman, H. (2025) A Comprehensive Survey of Hallucination in Large Language Models: Causes, Detection, and Mitigation, arXiv preprint, arXiv:2510.06265.

Anthropic (2024) Introducing the Model Context Protocol, Anthropic News, Nov 25, 2024.

Anthropic (2026) Claude 4.6: The Next Frontier in Coding and Software Engineering Agents.

Baker, C., Rafferty, K., Price, M. (2025) Large Language Models in Mechanical Engineering: A Scoping Review of Applications, Challenges, and Future Directions, Big Data & Cogn. Comput., 9(12), p.305.

10.3390/bdcc9120305

Bathe, K.-J. (1996) Finite Element Procedures, Prentice Hall, New Jersey.

Dassault Systèmes (2018) Introduction to Abaqus Scripting, SIMULIA Documentation.

Google (2026) Gemini 3.1: Advancing Long-Context Reasoning for Complex Engineering Tasks.

Heo, S., Na, S. (2025) Ready for Departure: Factors to Adopt Large Language Model (LLM)-based Artificial Intelligence (AI) Technology in the Architecture, Engineering and Construction (AEC) Industry, Results Eng., 25, 104325.

10.1016/j.rineng.2025.104325

Huang, L., Yu, W., Ma, W., Zhong, W., Feng, Z., Wang, H., Chen, Q., Peng, W., Feng, X., Qin, B., Liu, T. (2023) A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions, ACM Trans. Inf. Syst., 42(2), pp.1~55

10.1145/3703155

Kim, E.H., Park, J.H., Seong, A.J., Oh, H.U., Kang, S.H., Lim, J.H. (2025) Development of LLM Driven Automatic Component Recommendation System for Small Satellites, J. Aerosp. Syst. Eng., 19(6), pp.58~66.

Lee, J., Kim, J.Y., Ryu, S. (2025) IM-Chat: A Multi Agent LLM based Framework for Knowledge Transfer in Injection Molding Industry, J. Manuf. Syst., 83, pp.839~854.

10.1016/j.jmsy.2025.11.007

Model Context Protocol (2025) Specification, Version 2025-03-26, Model Context Protocol Documentation.

Mustapha, K.B. (2025) A Survey of Emerging Applications of Large Language Models for Problems in Mechanics, Product Design, and Manufacturing, Adv. Eng. Inform., 64, 103066.

10.1016/j.aei.2024.103066

Oberkampf, W.L., Trucano, T.G. (2002) Verification and Validation in Computational Fluid Dynamics, Prog. Aerosp. Sci., 38(3), pp.209~272.

10.1016/S0376-0421(02)00005-2

OpenAI (2026) GPT-5.2 Thinking: Enhancing reasoning and code synthesis performance.

Park, D., Moon, H., Ryu, S. (2026) A Self Correcting Multi Agent Framework for Language based Physics Simulation and Explanation, npj Artificial Intelligence, 2, 10.

10.1038/s44387-025-00057-z

SWE-bench (2026) SWE-bench Verified: Real-World Software Engineering Performance Leaderboard.

Wilberg, F. (2025) Large Language Model based Knowledge Creation Verified by Knowledge Graphs and News Articles, TU Wien.

Zhang, L., He, S., Zhang, C., Kang, Y., Li, B., Xie, C., Wang, J., Wang, M., Huang, Y., Fu, S., Nallipogu, E., Lin, Q., Dang, Y., Rajmohan, S., Zhang, D. (2025) SWE-Bench Goes Live, arXiv preprint, arXiv:2505.23419.

Information

Publisher :Computational Structural Engineering Institute of Korea
Publisher(Ko) :한국전산구조공학회
Journal Title :Journal of the Computational Structural Engineering Institute of Korea
Journal Title(Ko) :한국전산구조공학회 논문집
Volume : 39
No :2
Pages :123-135
Received Date : 2026-03-06
Revised Date : 2026-03-27
Accepted Date : 2026-03-30
DOI :https://doi.org/10.7734/COSEIK.2026.39.2.123

Journal of the Computational Structural Engineering Institute of Korea ISSN:1229-3059(Print) 2287-2302(Online) 한국전산구조공학회 논문집

All Issue