Haecheon Choi is a professor of mechanical engineering at Seoul National University, South Korea since 1993. He obtained his BS (1985) and MS (1987) degrees from Seoul National University, and a Ph.D degree (1992) from Stanford University. His research areas are turbulence, flow control, CFD, and bio-mimetic engineering. He is a member of The National Academy of Engineering of Korea and The Korean Academy of Science and Technology. He is a Fellow of American Physical Society and serves as an Associate Editor of Journal of Fluid Mechanics.

Inthis talk, we present subgrid-scale (SGS) models based on machine learning (ML) for large eddy simulation of turbulent flow over/inside a complex geometry. Previous ML-based SGS models have shown slight better performances than traditional physics-based SGS models in both a priori and a posteriori tests. However, ML-based SGS models have been tested mainly for trained flows, so there remain questions about generalization of those ML-based SGS models to complex flows. To extend the applicability of ML-based SGS model to untrained flows at various Reynolds numbers, we present neural network modification, recursive process, and dynamic approach. The ML-based SGS model developed performs well for flows over a circular cylinder and airfoil, and backward-facing step flow.