Dynamic crack initiation toughness : experiments and peridynamic modeling.

This is a dissertation on research conducted studying the dynamic crack initiation toughness of a 4340 steel. Researchers have been conducting experimental testing of dynamic crack initiation toughness, K{sub Ic}, for many years, using many experimental techniques with vastly different trends in the results when reporting K{sub Ic} as a function of loading rate. The dissertation describes a novel experimental technique for measuring K{sub Ic} in metals using the Kolsky bar. The method borrows from improvements made in recent years in traditional Kolsky bar testing by using pulse shaping techniques to ensure a constant loading rate applied to the sample before crack initiation. Dynamic crack initiation measurements were reported on a 4340 steel at two different loading rates. The steel was shown to exhibit a rate dependence, with the recorded values of K{sub Ic} being much higher at the higher loading rate. Using the knowledge of this rate dependence as a motivation in attempting to model the fracture events, a viscoplastic constitutive model was implemented into a peridynamic computational mechanics code. Peridynamics is a newly developed theory in solid mechanics that replaces the classical partial differential equations of motion with integral-differential equations which do not require the existence of spatialmore » derivatives in the displacement field. This allows for the straightforward modeling of unguided crack initiation and growth. To date, peridynamic implementations have used severely restricted constitutive models. This research represents the first implementation of a complex material model and its validation. After showing results comparing deformations to experimental Taylor anvil impact for the viscoplastic material model, a novel failure criterion is introduced to model the dynamic crack initiation toughness experiments. The failure model is based on an energy criterion and uses the K{sub Ic} values recorded experimentally as an input. The failure model is then validated against one class of problems showing good agreement with experimental results.« less