Dynamic Formulation and Execution of MDAO Workflows for Architecture Optimization
Sparsh Garg, Raúl García Sánchez, Jasper Bussemaker, Luca Boggero, Björn Nagel · 2024
One of the critical aspects to consider in early design phases of complex system is the system architecture, as it has major influence on final performance of the product. System Architecture Optimization (SAO) is a technique where the architecting process is formulated as a numerical optimization problem, which allows more extensive and less biased design space exploration. To evaluate the performance of architecture alternatives at the system-level, Multidisciplinary Design Analysis and Optimization (MDAO) is seen as a key enabling technology. Application of MDAO for architecture optimization is currently challenged by the observation that different architectures require different MDAO problem formulations, for example in discipline selection, data connections, and execution order. This work proposes a set of capabilities that MDAO formulation platforms should provide in order to support such dynamic MDAO workflow formulations. Special focus is placed to enable such capabilities in the context of collaborative MDAO, a specialization of MDAO where emphasis is on collaborative and cross-organizational definition, integration, and execution of MDAO workflows using central data formats. The proposed capabilities are implemented in MDAx, an existing MDAO workflow modeling tool developed for formulating collaborative MDAO workflows. The research is demonstrated by an open-source launch vehicle architecture design problem, showing that a dynamic MDAO workflow can be generated that modifies its execution behavior automatically for each architecture to be evaluated.