An LLM-Powered Pipeline for Automated Extraction, Analysis, and Validation of Requirements in Systems Engineering

Abstract This research paper discusses the development and capabilities of an LLM-powered pipeline for engineering requirements validation. The pipeline tests each requirement for completeness, missing information, and testability. Requirements that pass these tests are further checked for inconsistency and conflict. Cleaned requirements are then categorized and converted into SysML graph objects using Gaphor, with support for SysML 1.6 and Python 3.13.2. The pipeline uses RAG to connect external schemas, performs topic modeling to discover stakeholder priorities, and extracts requirements from MME transcripts. Preliminary results show promise in identifying missing information and inconsistencies. This work may be of help in integrating LLMs into requirements engineering processes.