Predicting terminal ballistics using an iterative application of an artificial neural network

The need to accurately predict the terminal ballistics of kinetic energy projectiles in Vulnerability/Lethality models is of paramount importance to the U.S. Department of Defense. In previous work, an artificial neural network was trained on a set of 1,877 data points to predict perforation, residual velocity, and residual mass of a kinetic energy projectile impacting a target consisting of a single homogeneous-monolithic-metallic element. The capability of that neural network was extended to handle targets consisting of multiple homogeneous-monolithic-metallic elements by iteratively applying the artificial neural network on each element in the target and using the predicted residual properties as inputs for the next element. The performance of that approach is analyzed against a phenomenological model that is currently used by the U.S. Department of Defense for modeling the terminal ballistics of kinetic energy projectiles. For completeness, the performance of the neural network against single element targets is also revisited since more data are now available.