Multidisciplinary robust design optimization considering parameter and metamodeling uncertainties

Abstract

Multidisciplinary robust design optimization (MRDO) is a useful tool to improve the stability of the performance of complex engineering systems involving uncertainty. However, the majority of existing MRDO studies only consider the parameter uncertainty. Metamodeling uncertainty, defined as the discrepancy between the computer model and metamodel at un-sampled locations, is often overlooked in MRDO. To solve the multidisciplinary problems under parameter and metamodeling uncertainties, this paper proposes a new framework called MRDO under parameter and metamodeling uncertainties (MRDO-UPM). The collaboration model is used to select the samples which satisfy coupled state equations. The selected samples are employed to construct the Gaussian process metamodels of the objective, constraint, and multidisciplinary coupled functions. Monte Carlo simulation is adopted to quantify the compound impact of parameter and metamodeling uncertainties. The MRDO-UPM framework is employed to explore the optimum. The proposed framework is verified through a numerical example, and the design of a speed reducer and a liquid cooling battery thermal management system.