WCSE

An Aviation Process Knowledge Question-Answering System Based on Knowledge Graph and Large Language Model

Abstract— Discrete manufacturing enterprises like aviation face significant challenges in managing fragmented process knowledge. This paper presents an innovative question-answering system that synergizes knowledge graphs with large language models to overcome these limitations. Our approach systematically processes multi-source aviation data through dual pathways: constructing a structured aviation process knowledge graph (AeroKG) while simultaneously building a vector database for unstructured documents using LangChain and FAISS. The integrated system leverages retrieval-augmented generation to enhance the DeepSeek-v3 LLM, achieving a 0.91 F1-score - a 15% improvement over conventional methods. Experimental validation confirms the system’s robust performance in supporting critical manufacturing operations, from process design to parameter optimization. The implemented web interface demonstrates practical viability for real-world industrial knowledge management applications. 
Index Terms— Aviation Process QA, Knowledge Graph, LLM 

Qian Zhao, Rui Zuo, Ruiqiang Lv, Tingting Du
AVIC Manufacturing Technology Institute, CHINA
Wenhao Xu
Beihang University, CHINA

[Download]

Cite: Qian Zhao, Wenhao Xu, Rui Zuo, Ruiqiang Lv, Tingting Du, "An Aviation Process Knowledge Question-Answering System Based on Knowledge Graph and Large Language Model", 2025 the 15th International Workshop on Computer Science and Engineering (WCSE 2025), pp. 15-20, Jeju Island, South Korea, June 28-30, 2025.