A non-local damage approach compatible with dynamic explicit simulations and parallel computing

Abstract

Abstract In the automotive industry, crack prediction is an important step of the design: its accuracy is crucial to avoid additional development costs and delays. However, its simulation is not always reliable yet which could be explained by the use of too simple fracture criteria. A possible solution could be the improvement of the fracture behavior prediction through the use of coupled damage models. Unlike the fracture criteria, damage models consider the loss of resistance on the elements behavior, which gives a better definition of the strain localization and crack path. However, due to stress softening, the problem becomes ill posed, generating damage localization on a single row of elements. The results are then dependent on the mesh size and the mesh orientation. To obtain mesh independent results, a possible solution is to resort to regularization methods, but only a few of them are compatible with dynamic explicit simulations, especially for ductile failure. This paper proposes to extend the implicit second gradient non-local regularization approach to crash simulations. This is achieved by modifying the second gradient equation to ensure its robustness for dynamic explicit simulations. This extended second gradient approach is implemented by enriching under-integrated continuum elements so as to naturally preserve the parallel computing ability. A comparison between simulations and experimental results obtained with specimens machined in a dual-phase steel sheet is realized to validate the proposed approach. Numerical results obtained with different mesh sizes and mesh orientations illustrate the mesh independence and are in very good agreement with the experiments in terms of both load-displacement curves and crack path.