Large Language Model Empowered by Domain-Specific Knowledge Base for Industrial Equipment Operation and Maintenance

Industrial equipment operations and maintenance (IEOM) refers to ensuring the normal, safe, and reliable operation of industrial facilities, which covers condition monitoring, equipment maintenance, troubleshooting, repair and maintenance, and system optimization. Currently, the advancements in artificial intelligence have greatly improved the efficiency and effectiveness of IEOM. However, its robustness and generalization in practical applications still need to be improved. Large language models (LLMs) like ChatGPT have recently made breakthrough progress, demonstrating highly intelligent language comprehension capabilities. Therefore, they are expected to drive a new round of transformation in IEOM, promoting the automation and intelligence of the entire IEOM process. However, when using LLMs for practical industrial applications, existing LLMs have fatal limitations as they severely lack domain-specific expertise. This makes it difficult for LLMs to handle technical issues in the industrial field. To this end, this study explores a new solution: LLMs empowered by domain-specific knowledge base (LLM-DSKB). This paper provides a detailed introduction to the core components and implementation details of LLM-DSKB, including the knowledge base, text embedding, vectorized retrieval, etc. The performance of LLM-DSKB is analyzed using real industrial cases, and the results demonstrate that LLM-DSKB can provide more accurate, specific, and industrially relevant results compared to traditional LLMs. This solution will drive the application of LLMs in the industrial field, significantly enhancing the efficiency, effectiveness, and quality of IEOM.