Study of imperfections in the cubic mesh of the truss-like discrete element method

Abstract

In the truss-like discrete element method (DEM), masses are lumped at nodal points and interconnected by means of one-dimensional elements with arbitrary constitutive relations. In previous studies of nonhomogeneous concrete cubic samples subjected to nominally uniaxial tension, it was verified that numerical predictions of fracture using discrete element method models are feasible and yield results that present good correlation with the experimental evidence so far available, including the prediction of size and strain rate effects. In the discrete element method approach, material failure under compression is assumed to occur by indirect tension. In previous simulations of samples subjected to uniaxial compression, it was verified that the response is satisfactorily modeled up to the peak load, when a sudden collapse usually occurs, characteristic of fragile behavior. On the other hand, experimental stress versus displacement curves observed in small specimens subjected to compression typically present a softening branch, in part due to sliding with friction of the fractured parts of the specimens. A second deficiency of discrete element method models with a perfectly cubic mesh is that the best correlations with experimental results are obtained with material parameters that differ in tension and compression. This paper examines another cause of the fragile behavior of discrete element method predictions of the response of concrete elements subjected to nominally uniaxial compression, namely the regularity of the perfect cubic mesh, which is unable to capture nonlinear stability effects in the material at a microscale. It is shown herein that the introduction of small perturbations of the discrete element method regular mesh significantly improves the predicting capability of the model and in addition allows adopting a unique set of material properties, which are independent of the applied loading.