Surrogate-based optimization of functionally graded plates under thermomechanical loadin

Abstract

Efficient designs for a Functionally Graded Plates (FGP) can be defined via structural optimization. Usually, bio-inspired algorithms are employed in order to carry out the optimization process. However, when analyses are very time-demanding, the process may be too costly, since hundreds or even thousands of analyses may be required. In this work, Sequential Approximate Optimization is employed to provide a more efficient approach. This paper focuses on the maximization of buckling temperature of a ceramic-metal FGP. B-Splines are used to define a continuous material gradation along the thickness direction. Effective material properties are evaluated by the rule of mixtures. The Particle Swarm Optimization (PSO) is applied for structural optimization and Isogeometric Analysis (IGA) is employed to evaluate the structural responses. Then, Sequential Approximate Optimization (SAO) is carried out to reduce the computational cost using Kriging to fit an approximate response surface. A comparison between conventional optimization and SAO is performed, and results show that SAO achieved the optimum design much earlier the conventional approach, requiring fewer high-fidelity evaluations.