Dual-loop integration framework for model-based system design and reliability analysis using Bayesian networks

Integrating model-based systems engineering (MBSE) and reliability analysis is challenging due to the lack of life-cycle reliability consideration in system design, difficulty in tracking design changes in reliability modeling, and limited quantitative analysis support in system modeling. To address these issues, this paper proposes a dual-loop modeling framework integrating system design and reliability analysis based on models and Bayesian networks. This framework supports a systematic and iterative design process, facilitates quantitative analysis throughout the system life cycle, and enables upstream integration of system-related and mapped reliability parameters from dynamically evolving design architectures at both modeling method and framework levels. Specifically, we propose a novel integrated modeling approach based on existing theories and methods, detailing the methodological, procedural, and linguistic foundations. Furthermore, a mapping method based on SysML activity diagrams basic types is proposed to connect system models with dynamic Bayesian networks (DBN) for quantitative analysis. Finally, the effectiveness of the proposed method is validated through its application to a reaction wheel actuator and a remote sensing satellite project. • A novel dual-loop modeling framework that integrates system design and reliability analysis, incorporating knowledge engineering, XMI (eXtensible Markup Language for Interchange) middleware, MBSE meta-models, and dynamic Bayesian theory. • A rich system modeling approach that achieves high coupling between system design and reliability (e.g., FMEA ﹨& FTA), based on mature axiomatic design methods and the Magicgrid system engineering modeling theory, combined with zero-trust principles and duality theory. • A mapping method based on the SysML (System Modeling Language) activity graph connects the system model with the Dynamic Bayesian Network (DBN) to achieve quantitative analysis and an instance verification.