Computational Fluid Dynamics and Machine Learning for B-2 Spirit Flying-Wing Aerodynamic Prediction

Abstract

The study includes a hybrid of high-fidelity computational fluid dynamics (CFD) and artificial neural networks (ANNs) for the investigation and prediction of the aerodynamic performance of the B-2 Spirit stealth flying-wing aircraft. The analysis of the three-dimensional geometry was performed by reconstructing it and simulating it using Reynolds-averaged Navier–Stokes (RANS) equations in conjunction with a [Formula: see text] turbulence model to accurately capture boundary-layer behavior and flow separation over a variety of subsonic conditions. A comprehensive mesh independence study confirmed the numerical reliability of the CFD results, and the aerodynamic characteristics were characterized as a function of angle of attack and freestream velocity. Because of the high computational cost of running CFD simulations multiple times, an optimized ANN model was developed from CFD generated data. The ANN’s mean square error was 0.0011, and the [Formula: see text] value was 0.995, demonstrating a high level of accuracy in predicting values based on CFD data.