大規模言語モデルを用いた汎用設計審査の可能性

抄録

This study explores the potential application of Large Language Model (LLM) in the field of mechanical engineering design review. The research highlights how LLM can effectively complement traditional design review processes by improving the accuracy of issue detection using well-crafted prompts. While the results demonstrate the potential of LLM to identify design flaws effectively, particularly in scenarios where hazards are clearly linked, challenges remain when the models are required to address more complex designs involving specialized terminology, materials, and processes. The study suggests that by refining prompts or specifying review perspectives, LLM can improve their accuracy, though further work is needed to develop methods that enable comprehensive design reviews without pre-defined focal points.