Knowledge-Based Question Answering System for Hydraulic Hoist Using Large Language Models

The hydraulic hoist is an essential electromechanical equipment widely used in water conservancy projects. Due to its system complexity and the diversity of failure modes, traditional knowledge management is inefficient, and it is difficult to support the demand for timely and accurate knowledge application during the operation and maintenance process of the equipment. This paper develops a knowledge-based question answering (QA) system for hydraulic hoists using the Large Language Model (LLM) combined with the Retrieval Augmented Generation (RAG) and LangChain methods. Firstly, a high-quality knowledge base is constructed with unstructured text data such as hydraulic hoist design data, failure cases, and maintenance manuals. Secondly, database retrieval technology is utilized to achieve graph retrieval and vector retrieval. Finally, combining the LLM and RAG methods, the answers are generated by reasoning between the retrieved relevant text segments and the user's questions. To verify the application effect of the QA system, this study employed a dataset comprising common failure types of hydraulic hoists and conducted a systematic QA evaluation and knowledge reasoning process test based on typical failure scenarios. The results show that the system can effectively deal with complex problem scenarios, provide fast and accurate solutions, and enhance the interpretability of the results through additional information, providing a valuable reference for the intelligent maintenance of electromechanical equipment in water conservancy projects.