Local Transition Model for Crossflow Instability in High-Speed Boundary-Layers

A local transition model is presented for predicting transition due to crossflow instability in high-speed boundary-layer flows. The model is based on an amplification factor transport equation whose production term is modeled with a neural network. The neural network was trained using a comprehensive database built from linear stability theory analyses of Falkner-Skan-Cooke profiles for 90 boundary-layer flows, encompassing various pressure gradients, wall temperatures, Mach numbers, and sweep angles. The model performance is evaluated for the HiFiRE-5 geometry at zero angle of attack for the freestream conditions of the Boeing/AFOSR Mach 6 Quiet Tunnel at Purdue University. The experimental transition front is matched for a critical amplification factor of 3.2. The results obtained with the new local transition model are compared to experimental measurements. Gridconvergence of the transition front is demonstrated. The present analysis also provides insight into the dependence of the amplification factor production on the neural network inputs and the dependence of the transition prediction on the diffusion term in the modified intermittency equation.