To enhance skill transfer efficiency in the manufacturing industry, this study develops a platform to support the automated generation of System Architecture Maps (SAMs) using Large Language Models (LLMs). Our proposed method integrates the dynamic process description of IDEF0 with the static design method of Axiomatic Design’s Zigzagging, enabling consistent information extraction by the LLM. Applied to an egg-cooking system case study, a comparison between a human’s subjective model and an LLM-generated model revealed differences in how knowledge was abstracted and specified. Furthermore, an analysis using a design sequencing system showed that the LLM-generated model creates more uniformly distributed design groups. While this research confirms the potential for automating SAM creation, challenges remain in handling knowledge that requires complex reasoning.
KIKUCHI et al. (Wed,) studied this question.