Parametric FEM for computational homogenization of heterogeneous materials with random voids

Abstract Computational homogenization is a powerful tool which allows to obtain homogenized properties of materials on the macroscale from the simulation of the underlying microstructure. The response of the microstructure is, however, strongly affected by variations in the microstructure geometry. The effect of geometry variations is even stronger in cases when the material exhibits plastic deformations. In this work we study a model of a steel alloy with arbitrary distributed elliptic voids. We model one single unit cell of the material containing one single void. The geometry of the void is not precisely known and is modeled as a variable orientation of an ellipse. Large deformations applied to the unit cell necessitate a finite elasto‐plastic material model. Since the geometry variation is parameterized, we can utilize the method recently developed for stochastic problems but also applicable to all types of parametric problems — the isoparametric stochastic local FEM (SL‐FEM). It is an ideal tool for problems with only a few parameters but strongly nonlinear dependency of the displacement fields on parameters. Simulations demonstrate a strong effect of parameter variation on the plastic strains and, thus, substantiate the use of the parametric computational homogenization approach.