Application of an Artificial Intelligence Agent with RAG Technique to Reduce Diagnostic Time in Corrective Maintenance: A Case Study of an Air Conditioning Manufacturing Plant

Abstract

This research aims to enhance the efficiency of the corrective maintenance process in the air conditioning manufacturing industry, focusing on reducing waste in the diagnostic phase (Mean Time to Diagnose: MTTD). Based on a case study of the MC-01 machine, which experienced a maximum accumulated downtime of 362.81 hours per year, root cause analysis revealed that 36.58% of the total repair time delay was attributed to limitations in retrieving historical repair data from the Computerized Maintenance Management System (CMMS). The existing system lacked a standardized structure, making data difficult to reuse. This research developed an intelligent maintenance assistant system via the LINE Official Account application by integrating an Artificial Intelligence (AI) Agent with the Retrieval-Augmented Generation (RAG) technique and a Large Language Model (LLM) through the n8n automation system, acting as a mobile knowledge base for on-site technicians. The performance evaluation demonstrated that the system accurately supports decision-making and provides precise repair guidelines, resulting in a statistically significant reduction in MTTD by 46.77% at a 0.05 significance level. Furthermore, this system serves as a crucial tool that enables maintenance technicians of all levels to perform their tasks rapidly and efficiently in accordance with established standards.