Surrogate Models to Accelerate High-Fidelity Physics Based Simulation

This topic focuses on developing fast-running, computationally efficient, surrogate models to be used in place of slow running, computationally expensive, high-fidelity physics-based models inside of end-to-end missile defense simulation frameworks. Recent advances in physics informed machine learning (PIML) have made possible the substantial speed improvement of physics-based models. Corvid Technologies and Auburn University propose to leverage Corvid’s existing PIML-based fast-running penetration model, along with their extensive capabilities and expertise in modeling hyper velocity impact events, in order to develop a state-of-the-art fast-running surrogate model for the generation of post-intercept debris (PID). This tool, called Advanced VIDEO, will have the capability to generate predictions of debris states (size, shape, material, velocity, temperature, etc.) with comparable fidelity to those produced using hydro-structural solvers on a time scale of seconds. Approved for Public Release | 22-MDA-11339 (13 Dec 22)