We at the Turbulence Lab of Yonsei University study turbulence mainly using numerical approaches. From the fundamental aspect, we are particularly interested in the interaction between turbulence and laden particles small or of finite size in various kinds of turbulence. We are also applying various deep learning algorithms to turbulence problems such as prediction and control of turbulence, development of the subgrid-scale model of LES and characterization of turbulent heat transfer. For the application side, we are developing real-time prediction algorithms of pollutant dispersion in the outdoor and indoor environments based on GPU parallelization.
Turbulence and Particle-laden Turbulence
Using direct numerical simulation of particle-laden turbulence, we are investigating various aspects of the interaction between particles and turbulence such as finite-time blowup or multifractal nature of particle concentration, the behavior of settling nonspherical particles, and settling particles or rising bubbles in Rayleigh-Benard convection.
Deep Learning of Turbulence
Various deep learning algorithms including GANs and reinforcement learning have been applied to the prediction and control of turbulence, prediction of turbulent heat transfer, turbulent inflow field generation, super-resolution reconstruction, development of control schemes for drag reduction, and subgrid-scale modeling of LES.
Computational Fluid Dynamics
For accurate and efficient simulations of the interaction between turbulence and finite-sized particles, a novel immersed boundary method which can be applied to nonuniform grids has been developed. A new formulation of the evolution equation of a nonspherical particle is proposed.
For the optimal design of a gas turbine, turbulent flow and heat transfer in the coolant passage are simulated. For the real-time prediction of pollutant dispersion in complex outdoor or indoor environments, new algorithms based on LES and IBM have been developed.