The safety management in the infrastructure industry is governed by a variety of regulatory documents and management codes. However, the current regulation compliance checking often relies on manual or exhaustive matching methods, which can result in issues such as inconsistent rating scales, artificial manipulation, and time-consuming procedures. Therefore, this study presents a novel approach that integrates knowledge graph (KG) and large language model (LLM) to facilitate knowledge mining and application in safety management regulatory texts. First, a semantic expression framework is established to represent the domain knowledge of regulatory texts. Subsequently, considering the requirements of rule interpretation, knowledge extraction is carried out on the regulatory texts to obtain KG. Then, based on the prompt-tuning, the KG is leveraged to enhance LLM’s inference of regulatory documents and management codes. Finally, a safety management workflow based on automatic rule interpretation is proposed and successfully applied to a hydropower project. The results demonstrate that this method can effectively enhance the interpretability of safety management codes and reduce reliance on domain experts. This study substantiates the effectiveness of integrating KG and LLM for knowledge mining and application. This work can provide a practical reference for enhancing automated and intelligent safety management in the infrastructure industry.