Predicting the mechanical behavior of two-phase materials with cellular automata (Q677052)

The method for predicting of overall and local mechanical behavior of two-phase materials is derived from both a self-consistent scheme and a model of cellular automata. First, an aggregate of equivolumic grains of two phases is submitted to plane strain deformation. These phases are assumed to be isotropic, incompressible, and linearly viscous or, alternatively, linearly elastic and incompressible. At the first step of localization, the strain rate associated with any cell of the automation is calculated by using Eshelby's results and by considering in turn each grain as an inclusion embedded in its neighborhood. At the second step, an averaging procedure is applied to derive the overall deviatoric stress. Then, when the grains are assumed to be equiaxed, three possible choices of the average of viscosity in the neighborhood are proposed, namely: the arithmetic and harmonic averages of the viscosities associated with the adjacent cells, and an average calculated from a self-consistent model extended over the neighborhood. The obtained analytical results are compared with classical approaches to the homogenization problem. So, the viscosity predicted by the cell model falls between the Hashin-Shtrikman bounds. The three above variants of viscosity are most contrasting in the case when the hard phase is nondeformable or the soft phase has a zero viscosity, but is still incompressible. In consideration of the overall mechanical behavior and local morphology changes of an aggregate of two phases submitted to plane strain compression, the analytical approach is replaced by the cellular automation one. The numerical results show that in the case of uniformly random aggregates, a \(30\times 30\) cell array is large enough to simulate an infinite structure. Significant morphological hardening or softening is shown to occur with increasing strain, depending on the initially equiaxed or elongated grain shapes. Finally, it is demonstrated that the model allows to calculate the distributions of local strains and to predict the development of strain inhomogeneities.