Exploring Induced Heterogeneity in Elastic Discrete Mechanical Models

Abstract

Mesoscale discrete lattice models offer a direct way to incorporate the heterogeneous microstructure of concrete and other geomaterials efficiently, using vector-based constitutive laws with homogeneous material parameters. These models exhibit stress oscillations, which, if deemed non-physical, can be suppressed using methods such as auxiliary stress projection or deviatoric-volumetric decomposition to produce homogeneous elastic stress fields. This study examines the elastic behavior of the homogenized models with controlled heterogeneity introduced via spatial randomization of material parameters, with an emphasis on the replication of the oscillations in the non-homogenized discrete model. Simulations with varying degrees of spatial correlation under different macroscopic loading conditions reveal that the original stress oscillations are best replicated with spatially independent randomization. However, none of the techniques fully reproduce the original oscillations.

Mesoscale discrete lattice models offer a direct way to incorporate the heterogeneous microstructure of concrete and other geomaterials efficiently, using vector-based constitutive laws with homogeneous material parameters. These models exhibit stress oscillations, which, if deemed non-physical, can be suppressed using methods such as auxiliary stress projection or deviatoric-volumetric decomposition to produce homogeneous elastic stress fields. This study examines the elastic behavior of the homogenized models with controlled heterogeneity introduced via spatial randomization of material parameters, with an emphasis on the replication of the oscillations in the non-homogenized discrete model. Simulations with varying degrees of spatial correlation under different macroscopic loading conditions reveal that the original stress oscillations are best replicated with spatially independent randomization. However, none of the techniques fully reproduce the original oscillations.