Random models versus periodic models for fibre reinforced composites

In the computational analysis of composite materials usually periodic models are employed. These models assume that the material has a deterministic and ordered distribution of fibres. This assumption leads to lower computational costs. However, real composites show random distribution of fibres. Nowadays it is computationally affordable to model composite materials as random composites. But is it worthy such a modeling? This work tries to answer this question by comparing stress and strain distributions between a periodic and a random model for a carbon reinforced polymer. The comparison shows that periodic models may be used for the simulation of effective properties but random models must be considered for the simulation of local phenomena such local damage or matrix cracking.