Abstract

Impact sensitivity (literally, how hard do you need to hit a material to cause it to initiate) is a critically important safety metric and performance indicator for explosives. It’s a challenging property to measure experimentally, however, which has fuelled the need for physical models to understand the link between structure and material property. While predicting property from structure is valuable, perhaps an even greater prize is having the ability to run the process in reverse, i.e. to predict structures likely to present with a desired property. In this talk I will outline, (i) how we can link impact sensitivity to structure through a first-principles physical model, and (ii) how we can broaden out the pool of structures studied through a supervised machine learning study, using features derived from the physical model. The output from this work are tools that can guide the discovery, design and synthesis of safer explosives.