Reliability-based structural optimization using a multiscale material model

In this paper, we present a methodology for reliability-based design optimization (RBDO) of structural components using a multiscale material model. This methodology focuses on the quantification of uncertainties in a multiscale material model, the influence of uncertainties on constitutive relationships at different strain rates, and the integration of such models into RBDO of structural components subject to multiple failure criteria. Metamodeling approximation is found to be a necessary component of this methodology when considering the noise and nonlinearity in failure indices of interest and the computational cost of RBDO with a multiscale material model. The application of this methodology is demonstrated in design optimization of an A356 cast aluminum alloy component. The accuracy and performance of several parametric and non-parametric metamodels are compared while investigating the effect of multiscale modeling on optimum design. The results show that Gaussian process provides the best metamodel for damage index. Moreover, the damage index in the design based on von Mises failure criterion is found to be twice that in the design based on the multiscale material model, although both designs have the same weight and maximum stress values.