Validation of large scale simulations of dynamic fracture

A novel integrated approach is developed for a systematic validation of large-scale finite element simulations on dynamic crack propagations along a weak plane [1]. A set of well-controlled experimental scheme is specifically designed to provide accurate input data for the numerical simulations as well as to provide metrics for quantitative comparisons between experimental and numerical results. Dynamic photoelasticity with high-speed photography is used to capture experimental records of dynamic crack propagations along a weak plane and to provide the crack propagation history. In the dynamic experiments, a modified Hopkinson bar setup with a notch-face loading configuration is used to obtain controlled loading conditions for the dynamic fracture problem. Also an inverse-problem approach of cohesive zone model is employed to obtain a realistic cohesive law, i.e. a traction-separation law, of the weak plane, from independently measured crack-tip deformation fields using speckle interferometry technique. The experimentally collected data, the loading conditions and the cohesive law, are considered as input for the finite element simulations [2]. We employ finite-deformation cohesive elements to account for crack initiation and growth in bulk finite-element discretizations of the experimental sample. As it is well know, the cohesive elements introduce an additional material-dependent length-scale into the finite element model. The demand of accurately resolving this length-scale by the finite-element discretization, as required for truly mesh-independent results, may often lead to discretizations containing several millions of elements. We therefore resort to massively parallel computing. A comparison of the metrics from the numerical simulations with those from the experimental measurements is performed to validate the large-scale simulations. The numerical results show good agreements with the experimental results, leading to a successful validation of the large sca