The safe and efficient operation of nuclear power plants critically depends on the timely and accurate determination of fault procedure entry conditions—specific states or parameters that, when met, necessitate the initiation of particular fault procedures. Traditional methods require operators to manually assess complex data under time constraints, which can lead to delays and human errors. This paper proposes an integrated approach leveraging Artificial Intelligence (AI) technologies, specifically Large Language Models (LLMs) and Knowledge Graphs (KGs), to accelerate and enhance this determination process. In our work, the LLM serves as an intermediary between the operators and the KG. It understands the operators’ questions posed in natural language and translates them into Cypher queries, which are used to interact with the KG. The KG stores comprehensive information on existing fault procedures, including symptoms and entry conditions, represented through interconnected nodes and relationships. Upon receiving an operator’s query, the LLM processes and converts it into a Cypher query to retrieve relevant data such as fault types and corresponding fault symptom from the KG. After obtaining the results, the LLM translates the information back into understandable language for the operator. This bidirectional translation enables seamless communication between the operator's intent and the system's knowledge base. The integration of LLMs and KGs enhances speed and efficiency by significantly reducing the time required for determination of entry conditions.