An augmented physics informed neural network approach for blunt-body dynamics

Abstract

This paper presents an ansatz-informed approach to modeling the dynamics of blunt-body entry vehicles by combining physics-based modeling with machine-learning techniques. The main focus is developing an augmented physics-informed neural network (PINN) to simulate the vehicle's behavior during atmospheric entry. The proposed PINN architecture is carefully crafted to incorporate existing knowledge about the fundamental physical principles governing the system. This integration enhances the accuracy and applicability of the model compared to conventional data-driven approaches. The proposed methodology is evaluated across various scenarios and demonstrates precise predictions of the vehicle's trajectory and essential parameters.