This paper discusses the fracture response in a collision scenario together with the discretization of the model, and it compares various approaches suggested by current design rules and proposed by researchers using a validation approach with different model scales; including material tests, indentation experiments and simulations of an actual full-scale collision event. Calibration of the fracture criteria is performed based only on data that can safely be assumed to be available in a design situation, namely, those from a uniaxial tensile test. This approach reveals the behavior of each criterion as it would be in a design situation and avoids the possibility of calibration toward a known solution. The robustness of the fracture criteria, i.e., their ability to simulate fracture given varying stress states and mesh sizes, are investigated based on simulations of experiments. The statistical variations for the various criteria are shown.