Elastoplastic Mesoscale Homogenization of Composite Materials

Mesoscale homogenization provides a computationally efficient way of capturing some degree of local variation in the behavior of a composite microstructure. In this work, techniques are explored in which the local two-phase microstructure is homogenized using the moving-window generalized method of cells (GMC) technique. Both elastic and plastic material behavior is investigated using GMC-generated anisotropic stress-strain curves. An optimization procedure is used to define Hill's yield criterion parameters which best fit the GMC-generated data. Two perfectly plastic models are developed based on the GMC results; these are called the subcell initial yield model and the matrix average yield model. A technique is also developed which incorporates hardening behavior. Different windowing techniques are investigated: an overlapping windowing technique which requires more computational time, and a nonoverlapping technique which requires less computational time. It is found that the matrix average model using small nonoverlapping windows is the best technique in the cases studied, combining accuracy and computational efficiency.