Characterization, synthetic generation, and statistical equivalence of composite microstructures

Abstract

Mechanical behavior and reliability of composites are driven significantly by microstructural variability. Such variability can be present in the form of both morphological and constituent property variability. To understand the effect of this variability on macroscopic mechanical behavior, many statistically equivalent microstructures must be evaluated. This requires the ability to generate such microstructures. In this work, morphological variability was quantified by image analysis of actual microstructures. To reproduce this variability, a methodology was developed in which random microstructures are generated and subsequently adjusted to simultaneously match both short- and long-range statistics of actual microstructures. Synthetic microstructures were generated at a length scale of 70 µm, corresponding to the length scale at which fiber volume fractions of adjacent microstructures are uncorrelated. The utility of this methodology was also demonstrated for larger microstructures containing defects such as alignment fibers, voids and resin seams.