Quantification of uncertain macroscopic material properties resulting from variations of microstructure morphology based on statistically similar volume elements: application to dual-phase steel microstructures

A method to quantify uncertain macroscopic material properties resulting from variations of a material’s microstructure morphology is proposed. Basis is the computational homogenization of virtual experiments as part of a Monte-Carlo simulation to obtain the associated uncertain macroscopic material properties. A new general approach is presented to construct a set of artificial microstructures, which exhibits a statistically similar variation of the morphology as the real material’s microstructure. The individual artificial microstructures are directly constructed in a way that a lower discretization effort is required compared to real microstructures. The costs to perform the computational homogenization for all considered SSVEs are reduced by an adapted form of the Finite Cell concept and by applying the multilevel Monte-Carlo method. As an illustrative example, the proposed method is applied to a real Dual-Phase steel microstructure.