Graph approach to the constitutive modelling of heterogeneous solids (Q1097731)

In this paper we show how a graph representation may be used to analyze the constitutive laws of granular-type materials with the inclusion of microscale randomness. Specifically, we derive the constitutive coefficients for linear elastic interactions between grains of arbitrary polygonal convex shapes. The new approach which applies at the scale of a single cell of the dual graph, gives explicit dependence of the fourth- order elasticity tensor on the cell's mechanical and geometrical properties, and thus leads rigorously to a statistical continuum approximation. Next, we show how a graph model of a heterogeneous spatially random material with elastic-inelastic edge interactions leads to a concept of a statistical family of failure (or yield) surfaces in place of a single failure surface of deterministic models. The problem of determination of an effective failure surface is reduced to the stochastic problem of percolation of inelastic edges on the graph.