LLM-Driven Automotive Wiring Harness Design Optimization: A Case Study Focused on Cost and Time Waste Elimination

The fuses identification in power distribution boxes, which demands gathering and synthesizing information from diverse sources, represents a significant time consumption for engineers. Furthermore, the inherently repetitive nature of this manual task renders it susceptible to inaccuracies. To address this limitation, this paper examines the application of Large Language Models (LLMs) in the form of chat-bots for analyzing and optimizing vehicular Electrical Distribution Systems (EDS). The research investigates the capabilities of such a system to process complex EDS data, using Vehicle Manual Owner as a study case, with the goal of identifying optimization opportunities and improving project efficiency. The results of the application of Retriever Augmented Generation (RAG) enhanced the model’s ability to handle domain-specific data and function as a specialist assistant for Power Distribution Boxes. Experiments suggest this automated approach can generate valuable insights, such as identifying fuse component locations, specific fuse identifiers, amperage ratings, and the connectors associated with particular modules.