Neural-Network Surrogate Model for Flow Stability Analysis Based on Parabolized Stability Equations

Use of artificial neural networks (ANN) for effectively modeling the Tollmien-Schlichting (TS) instabilities in transitional flow simulations has been investigated. This ANN surrogate model will enable an automated and robust framework for predicting the laminar-to-turbulent transition of boundary layers based on parabolized stability equations (PSE) and the N-factor envelope. Our goal is to couple the PSE-based transition model with the Spalart-Allmaras (SA) turbulence model and others used in aerodynamic design and optimization. Our results show that the proposed surrogate model can accurately predict the streamwise growth rate of perturbations for Falkner-Skan-Cooke (FSC) boundary layers. The N-factor computed based on the outputs of the ANN surrogate model is also sufficiently accurate to form a reliable N-factor envelope. We verify and validate our surrogate model on airfoil boundary layers by comparing it with available computational and experimental data for the NACA 0012 and NLF(1)-0416.