A Large Language Model as an Assistant for Software- Independent Structural Analytical Model Review and Editing

Abstract

While structural analysis software facilitates the modeling of complex structures, it requires significant effort to master these tools. With advancements in large language models (LLMs), this study explores the potential of leveraging an LLM as a software-independent assistant for reviewing and editing structural analytical models. The proposed framework was tested using GPT-40 and an analytical model from Midas Gen, focusing on tasks such as model information extraction and editing. The results demonstrated the frameworks potential, achieving an accuracy of 91% when a system prompt With model data ontology information was used, compared to a 20% accuracy without the use of any prompt engineering.