Towards Using Functional Decomposition and Ensembles of Surrogate Models for Technology Selection in System Level Design

Technology selection as part of the complex system design process involves multiple teams focusing on different subsystems, which in turn brings on interoperability challenges, extends design evaluations, and complicates design cycles. Utilizing function-mean modeling and the ensemble of surrogate model techniques, the paper reveals how low-level input parameters in the design can be instrumental in predicting higher-level performance outcomes. Applying the suggested framework to a case study from the space industry, surrogates trained on system level and component levels in a flow management system are shown to be generalizable. Exploring the methods for aggregating surrogates into ensembles shows how such an assembly can be built and utilized for navigating through the early fuzzy phases of design. © 2024 Proceedings of the NordDesign 2024 Conference