Fault Diagnosis and System Maintenance Based on Large Language Models and Knowledge Graphs

The continuous development of Large Language Models (LLMs) has driven a new round of technological change, and for the specialized fields, LLMs have been criticized by their "hallucination". To this end, we propose a framework for applying LLMs to specialized fields. Specifically, the main work includes constructing Knowledge Graphs (KGs) and fine-tuning datasets for industrial fields fine-tuning and deploying LLMs. As well as combining multimodal data to construct more detailed prompts for inference, and realizing multi-hop inference with the help of KGs when necessary. This study takes Computer Numerical Control (CNC) machine tools as an example, by using this framework, effective fault prediction can be performed and reliable maintenance recommendations can be provided to effectively avoid and minimize the losses caused by equipment faults, and improve productivity and efficiency.