LES and DNS of Multiphase Flows in Industrial Devices: Application of High-Performance Computing

Abstract

High fidelity solutions of turbulent flow equations are obtained by large eddy simulation (LES) and direct numerical simulation (DNS). These techniques are devoted for resolving most of the energy-carrying scales in a turbulent flow. Grid resolution in LES or DNS is determined by the lengths of the finest scale of motion which is to be directly simulated. In multiphase flows, further refinement in the grid topology is required to capture the bubble or droplet front and also to resolve the small structures that are created in the wake zone of the bubble/droplet. Owing to the grid size and finer timescales, the computational complexities in LES or DNS are extremely high, and parallel computing resources are often deployed. The present chapter reviews computational efforts involved in turbulent multiphase flow simulation in industrial devices. Several high-performance computing (HPC) strategies like distributed computing using message passing interface (MPI), general purpose graphics processing unit (GPGPU) accelerated computing using CUDA and their hybridizations are also reviewed. Estimations of the computational requirement for simulation of large industrial devices are presented, and potential use of modern computational science and hardware are critically assessed.