AirReq: Automated Requirements Smell Detection and Elimination for Commercial Aircraft Systems
Tianyi Wang, Weiru Wang, Yilong Yang · 2025
System requirements quality is critical in commercial aircraft systems engineering, directly impacting subsequent processes. However, issues like requirements writing smells and low usability pose significant challenges. This paper proposes an automated method for requirements smell detection and elimination tailored for the commercial aircraft domain. Using Large Language Models (LLMs), our approach integrates Prompt Engineering and Retrieval-Augmented Generation (RAG). We identify 12 requirements smell features and propose a workflow for their detection and iterative elimination. Experimental results show our detection method achieves quality comparable to human experts but with significant gains in speed and reduced effort. Moreover, our semi-automated elimination method, enhanced by external knowledge bases, improves efficiency and the quality of refined requirements over manual processes. This research offers a robust solution for enhancing requirements quality in complex aerospace projects.