EAGER: Extended Particles in Turbulent Flow: A Grand Computational Challenge

Johns Hopkins University, Baltimore MD

Investigators

Abstract

1258398 PI: Prosperetti This project is concerned with the development of a flexible and accurate numerical method, named Physalis, for the simulation of particulate flows. The work to date has shown that this method has significant advantages over the other methods used at present such as the lattice-Boltzmann method and the immersed-boundary methods in that the method permits the exact satisfaction of the no-slip condition on each particle and that each particle is represented in a spectrally convergent fashion. This permits simulations involving 1000 particles with sufficiently high accuracy for particle Reynolds numbers up to about 100. The objective of the work proposed here is to port the existing code to GPU-based computers with the goal of attaining speed up of at least two orders of magnitude. The project will explore novel approaches to the parallelization of the code and memory management to achieve this speed up. The code thus developed will be made available to the research community through the PIs web. Natural processes and contemporary technology offer a wealth of instances in which solid particles are transported by fluid flows: marine sediments, gravity currents, fluidized bed combustors, catalytic oil cracking and chemical looping combustors are just a few examples among many. A reliable theoretical understanding of these systems is essential for a variety of reasons ranging from the modeling of environmental flows to carbon sequestration, to the development of new or scaled-up designs without expensive preliminary experimentation and many others. Sophisticated numerical methods and state-of-the-art computational hardware to be employed in this project will be useful in advancing the understanding of particulate systems.