A machine learning method for transition prediction in hypersonic flows over a cone with angles of attack

Abstract

The prediction of the transition location (TL) in three-dimensional (3D) hypersonic boundary layers is of great importance in hypersonic engineering. In the present work, a method using machine learning techniques is presented for the prediction of TLs based on experiment data over a Mach 6.5 inclined cone. A mapping function is directly constructed between TLs and the circumferential angle θ by neural networks (NNs). The results show that the present NN predicts well for both interpolations of both the angle of attack (AOA) and unit Reynolds number Re0 and extrapolation of only Re0 whereas errors increase for the extrapolation of a higher AOA. This work sheds new light on the fast prediction of TLs in hypersonic complex 3D boundary layers.