Numerical homogenization ofN-component composites including stochastic interface defects (Q4943441)

For composite materials, defects appearing on boundaries between composite constituents play a dominant role in the evolution of non-linear behaviour and strength of the composite structure. The present paper addresses the homogenization of a random periodic fiber \(N\)-component constituent with stochastic interface defects. Employing a probabilistic averaging method, it is shown that the former can be replaced by a \(2(N-1)\)-component composite with deterministic geometry and stochastically defined elasticity tensor components. For such a model, the homogenization is carried out using the (probabilistic) effective modulus method. The criterion for comparing the properties of real and homogenized materials is based on the assumption that the strain energies stored in both media are approximately equal, i.e. in the sense of weak convergence when the scaling parameter tends to zero. The computational implementation uses the Monte Carlo simulation technique and the so-called MCCEFF system designed for numerical homogenization of stochastic periodic fiber-like composites with or without stochastic interface defects. Numerical experiments reveal, among others, that the results of composite homogenization with stochastic interface defects are sensitive to parameters defining these defects. This sensitivity is significantly stronger than the sensitivity to randomness of Young's modulus components. The concepts introduced are well motivated and explained, and numerous references (76) are given.