Design research on LLM-based intelligent information systems for radiation equipment safety management

This study proposes a next-generation radiation equipment management system that integrates Quick Response (QR) code-based digital tagging with Retrieval-Augmented Generation (RAG)-based generative Artificial Intelligence (AI). The proposed system addresses management gaps by enabling real-time equipment tracking and immediate transfer recording through QR code-based digital tagging with unique identifiers. A simple QR code scan connects to the central database, allowing instant verification of equipment history and inspection status, while simultaneously providing customized safety management information, generated in real-time by RAG-based AI. We used an intelligent document processing technique, where AI initially analyzed radiation equipment manuals and converted them into structured question-and-answer formats. Experiments were conducted using a dataset derived from the manuals of ten distinct radiation measurement devices, with a QWEN3 8B large language model serving as the core of the generative system. In our quantitative evaluation tailored for radiation-safety QA—measuring Factual Consistency, Safety Score, and Hallucination Rate—the RAG approach substantially reduced hallucination rates compared to the baseline model. The RAG system maintained stable performance despite temperature parameter variations, confirming its viability for field deployment. This study proposes a practical and effective approach to simultaneously enhance the efficiency and safety of radiation equipment management.