Design-Variable Hypernetworks for Flowfield Emulation and Shape Optimization of Compressor Airfoils

What is it about?

Deep-learning-based flow emulators are adapted to predict the flow field around parametrically-defined airfoils, and then used for design optimization in an industrial design setting. Under transonic conditions the emulator-driven optimization achieves the same optimal design as a high-fidelity solver in a reduced number of iterations at a fraction of the online computational cost, while providing similarly-performing designs at off-nominal conditions.

Featured Image