Finite element (FE) simulations of structures and materials are becoming increasingly accurate, but also more computationally expensive as a collateral result. This development occurs in parallel with a growing demand for data-driven design. To reconcile the two, a robust and data-efficient optimization method called Bayesian optimization (BO) has been previously established as a technique to optimize expensive objective functions. The mesh width of an FE model can be exploited to evaluate an objective at a lower or higher fidelity (cost & accuracy) level, which is the domain of multi-fidelity BO (MFBO) applications. However, BO and MFBO are usually not directly compared in the literature. Moreover, sampling quality and assessing design parameter sensitivity are often underrepresented parts of data-driven design. This paper combines global sensitivity analysis and (MF) BO into a novel, efficient Bayesian data-driven framework. We compare the performance of BO with that of MFBO by maximizing the energy absorption (EA) problem of spinodoid cellular structures. The findings show that similar or better designs are suggested by MFBO with 16% fewer expensive objective evaluations compared to BO when maximizing the EA. The results, which are made open-source, serve to support the utility of multi-fidelity techniques across expensive data-driven design problems.
Guo et al. (2026) studied this question.