Data-driven simulations for turbulence inside and over porous media

Turbulence in porous media is a controversial issue. However, high velocity fluid flow through large porous media can lead to turbulent flow within the pores. Several applications exist wherein the pore Reynolds numbers can be so large that the unsteady inertial effects become important giving rise to transitional and/or turbulent flow. Examples include, but are not limited to, pebble-bed high temperature nuclear reactor, gas turbine cooling and packed bed catalysis. The project is aimed to understand the turbulence inside and over porous media via HPC-aided direct numerical simulation (DNS) and data-driven simulation methods. DNS is enabled by arbitrary high-order highly-scalable spectral/hp element solver. Artificial neural-network (ANN) based super-resolution, LSTM based flow prediction and causality analysis belong to the plan of the project.