.png){kind=icon}
.png){kind=icon}
Return
A knowledge system for aeroengine: an LLM-based graph construction framework and a query-context-aware graph transformer model for knowledge reasoning
.png){kind=icon}
PRE
AI
DOI：10.1016/j.aei.2026.104650.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Aeroengine design involves a vast amount of knowledge that is often difficult to share and reuse effectively, leading to redundant and inefficient design efforts. To address this challenge, we propose an Intelligent Knowledge System for Aeroengine (IKSA), which addresses two core challenges: knowledge graph (KG) construction and reasoning. We present an automated KG construction framework based on large language models (LLMs) that extracts knowledge from multi-source, heterogeneous data into a standardized knowledge graph. Second, we propose a novel knowledge graph reasoning model named QCAT-Net, which integrates a Query-Context-Aware message-passing network with a linear-complexity graph Transformer to effectively infer answers to design-related queries. By integrating the KG construction pipeline with QCAT-Net, we establish the complete IKSA system to assist designers in querying and understanding aeroengine knowledge. We construct an aeroengine knowledge dataset using IKSA and evaluate the performance of the proposed reasoning model. Our method achieves superior results compared to existing baseline approaches on this dataset. Ablation studies further demonstrate the effectiveness of both the attention mechanism and the query-context-aware message-passing module. Case studies highlight the practical value of IKSA in real-world design queries, providing a systematic solution for intelligent knowledge management in aeroengine development.
Keywords:
Aeroengine design
Knowledge graph construction
Large language models
Query-context-aware reasoning
Graph Transformer
Journal
IF:
9.9
Papers:
4.0K
Citations:
1.7W