Qualitative analysis of fiber composite microstructure: Influence of boundary conditions

The principal objective of this work is to classify the influence of different boundary conditions that can be applied when computing various statistical descriptors for generally random microstructures. Although applicable to most statistical descriptors, attention is limited to two-point probability functions. Binary images of random fibrous composites are assumed in the present study. Here, the periodic, mirror and no (plain) boundary conditions are addressed when processing the binary images. Also, the minimum number of samples (Representative Volume Elements—RVEs) in the ensemble that still provide statistically representative results is sought. While the available results promote the use of periodic boundary conditions, particularly from the computational point of view regardless of the RVE size, the number of samples required is largely affected by their size. If available, the most efficient procedure results from the use of a large binary image of a real microstructure combined with periodic boundary conditions.