Modelling dynamic tensile failure of quasi-brittle materials using stress-enhanced nonlocal models

The development of realistic numerical tools to efficiently model the response of concrete structures subjected to close-in detonations and high velocity impacts has been one of the major quests in defense research. Under these loading conditions, quasi-brittle materials undergo a multitude of failure (damage) mechanisms. Dynamic tensile failure (e.g. spalling), characterized by a significant strength increase associated with loading rate, has revealed to be particularly challenging to represent. This phenomenon has been modeled by means of continuous damage mechanics in the last decades. To minimize pathological mesh sensitivity, a nonlocal formulation is generally considered. Nevertheless, these models fail to properly represent damage initiation and growth around discontinuities, such as notches, damage areas and free boundaries. These inconsistencies are the consequence of using a fixed interaction domain (characteristic length) in the nonlocal formulation. In spite of limited experimental knowledge about the definition of the characteristic length parameter, there is now consensus that this quantity is not constant. In this contribution, an enhanced nonlocal model, where the interaction domain of any gauss point contracts or expands according to the stress-state of its neighbors, is used. This formulation was coupled to the well-known Mazars damage model and implemented within the framework of LS-DYNA using a fully explicit computation scheme. Two sets of numerical studies are presented in this paper. One shows the applicability and limitations of the implemented explicit algorithm to compute nonlocal quantities. The other shows that with this stress-enhanced regularization model it is possible to represent damage initiation and growth more realistically and correct the inconsistencies emerging from the traditional nonlocal formulations.
Structural Engineering
Civil Engineering and Geosciences