Abstract
Design space exploration at early and system-level phases often relies on low-fidelity, qualitative evaluations due to the prohibitive cost of high-precision methods. This research proposes an approach to bridge this gap by utilizing AI and modeling tools and leveraging existing datasets and detail-level analysis to enhance system-level decision-making. Creating a platform of datasets is suggested by shared features for a product model hierarchy that mirrors the functional decomposition of the product architecture. This is achieved by integrating function-mean (FM) as an early design modeling tool with an assembly of datasets acquired from Finite Element (FE) as a higher fidelity concept evaluation tool. This is an effort to link the early functional space to the later response space. It is shown that training prediction models on datasets from low-level product architecture, which are more viable to obtain, enables design concept evaluation at the system-level. Moreover, by introducing a modular and radical change into product architecture, which represents new technologies that require performance evaluation, the built dataset platform is investigated. Finally, a metric to evaluate the platform's success is suggested to increase reliability. As lower levels of product architecture often experience more frequent changes than system-wide modifications. This method can be effective in scenarios where component-level innovations require rapid performance evaluations. The implications of such a dataset platform can reduce the concept evaluation cost in early phases by accelerating the testing of innovations and ideas in product architecture.
