Decreasing computational effort in structural optimization using a multifidelity metamodel

High fidelity models are becoming increasingly common in engineering optimization. The computation burden is often caused by expensive analysis and simulation processes in order to reach a comparable level of accuracy as physical testing data. The metamodels are initially developed as surrogates of the expensive simulation process in order to improve the overall computation efficiency. This work presents a new multilevel optimization approach for multidisciplinary structural design optimization based on multi fidelity modeling to decrease computational effort. Such method is a composition of a statistical estimating method and a metaheuristic algorithm. A low fidelity analysis response determines if the high fidelity analysis should be done or not. As a result, most of unnecessary high fidelity calculation will be omitted. The empirical results show the new algorithm causes a significant decrease in computational load as well as increase in convergence rate. Keywords: Multi level optimization; Metamodeling; Harmony search algorithm; Inverse distance weighting model.