An adaptive multi-fidelity approach for design optimization of mesostructure-structure systems

Abstract

Metamaterials are engineered structural materials with special mechanical properties (e.g., negative Poisson’s ratio) that are not found in nature materials. The properties of the metamaterials can be tailored by designing the cellular structure at the mesoscale. Additively manufactured metamaterial structures provide new opportunities for the development of next-generation functional lightweight vehicle components. This paper proposes a general method for the design optimization of the metamaterial infilled structure component under transient, dynamic loads (i.e., when gradient information is not available). We propose to integrate the addition correction-based multi-fidelity approach with the Probabilistic Multi-Phase Sampling Strategy, which maximizes the information gain of a limited number of sample points. The propose method continuously improves the predictability of the multi-fidelity surrogate model during the iterative optimal search process. This method is demonstrated on two benchmark problems: optimization of a 2D cellular metamaterial structure under static tensile loads and optimization of a pseudo 3D cellular metamaterial structure under transient dynamic loads. In both cases, the proposed method finds the optimal design with fewer number of expensive, high-fidelity design evaluations than the traditional design optimization methods.