Abstract
Moving-window micromechanics techniques are emerging as a useful tool for characterizing random composites, i.e., those with spatially random microstructures, in terms of local material property fields. These material property fields represent an approximate constitutive model for local material response. In this work, elastic property fields for randomly generated composite materials are estimated using a moving-window technique in combination with three micromechanical models, Mori-Tanaka, the generalized method of cells, and a brute-force finite element analysis. The first two methods make predictions of local properties based on small areas of the full sample. The brute-force finite element method performs its initial analysis on the full microstructure. Using the finite element results as the baseline solution, the fields developed from the methods are compared. Results illustrate the differences between the three methods and highlight the strong influence of window size on the analysis. Two criteria are suggested for further study of appropriate window size: A comparison of the correlation areas of the fields, which suggests an upper bound, and the analysis of a single fiber problem with known elasticity solution, which examines a potential lower bound. All of the techniques are demonstrated by application to numerically generated composite microstructures.