A large language model and knowledge graph collaborative generative framework for aircraft manufacturing system design in MBSE

Aircraft manufacturing system design has traditionally relied on expert experience and problem-specific procedures, resulting in limited automation, poor reusability, and insufficient life-cycle consistency and traceability. Although model-based systems engineering (MBSE) provides a formal means to improve consistency and traceability, constructing MBSE models requires manual effort. To address the above questions, a large language model (LLM) and knowledge graph (KG) collaborative generative framework is proposed in this work to support aircraft manufacturing system design in the MBSE paradigm. This framework aims to integrate domain Q&A, automated plan generation, MBSE model transformation, simulation-based verification, and feedback-based regeneration to enhance the automation in aircraft manufacturing system design. It combines the semantic understanding and reasoning capabilities of LLMs with the structured knowledge representation of KGs, achieving end-to-end generation from natural language design scenarios to design plans. By transforming generated design plans into formal MBSE models and simulation models, the framework supports semantic consistency and traceability through the design life cycle. Finally, a feedback-based regeneration mechanism is proposed to improve the previously generated design plans, forming a closed-loop design process. Experiments on the aircraft fuselage joint case show that the proposed framework achieves the highest Micro-F1 score in domain Q&A and generates feasible plans across multiple design scenarios.