Effect of Representative Volume Element Size for Creating Statistically Equivalent Composite Microstructures

Abstract

Micromechanical models of fiber reinforced composites often need to have fiber arrangements which are equivalent to manufacture samples to capture local features such as fiber clusters or matrix pockets which may impact failure. These features can vary in size, requiring that models must be large enough to capture features which may be most important. In this study, different size windows of larger composite scans were sampled, and descriptors were measured which characterize fiber and matrix dense regions. Statistically equivalent microstructures to these windows were generated and simulated for transverse strength and stiffness using a reduced order micromechanical model. Results showed that certain statistics could be captured by smaller window sizes compared to larger, and that some statistics could not be captured by the largest window size. Also shown is how transverse strength and stiffness varied depending on microstructure size.