Prediction of Multiple Aerodynamic Coefficients of Missiles using CNN

This paper proposes a deep-learning-based methodology for predicting the aerodynamic coefficients of various missile nose shapes under extreme flow conditions. In the case of missile development, when the design shape changes frequently, efficient prediction of aerodynamic coefficients can save time and cost. Using Missile DATCOM, the first step of the procedure generates a low-cost, low-fidelity database. Then, we propose a neural network architecture consisting of two CNN layers that predicts normal force, pitching moment, and axial force coefficients for a given nose shape at various angles of attack and Mach numbers. Finally, we demonstrate that the prediction accuracy of the network can be improved by optimizing with a linear combination of multiple loss functions. The efficacy of the proposed strategy is demonstrated through a test case study.