A GPU Computing Platform for Modeling Fluid-sheared Granular Beds

Abstract

The onset of sediment transport in response to hydrodynamic driving has broad applications across industry and the geological sciences. Accurately modeling this phenomenon requires a fundamental understanding of the static shear strength of granular beds, as well as the dynamics of mobilized grains, both of which depend crucially on particle size and shape. Many previous computational studies of the onset of sediment transport have modeled granular beds using spherical grains with weak size polydispersity. However, gravel-bedded rivers are typically composed of grains with a wide range of shapes and sizes. Modeling a broad range of grain shapes and sizes requires the development of new simulation algorithms that can be deployed on parallel computing platforms. In general, parallel computing allows many processors (e.g. CPUs) on a single node to work together (for instance, computing forces on grains in far-separated regions of a granular bed simultaneously in the discrete element method (DEM) approach), which makes calculations of large systems much more efficient. Additionally, the development of highly parallelized graphics processing units (GPUs), originally designed for 3D computer graphics rendering, now provide even faster performance than parallelized CPUs. Our research group has developed new models to accurately calculate the forces between non-spherical grains that can be implemented efficiently into DEM simulations of sediment transport. To study sediment transport in laboratory-scale beds containing realistic grains, we are requesting a GPU computing platform with 8 servers each containing 4 Nvidia GPUs. This proposal summarizes our previous and ongoing ARO-funded work on fluid-driven granular beds, describes the new algorithms and models that we have developed to calculate forces between nonspherical grains, details the equipment we are proposing to purchase, and explains how our research will be accelerated by the acquisition of the GPU platform.

Document Details

Document Type
DoD Grant Award
Publication Date
Feb 14, 2019
Source ID
W911NF1810252

Entities

People

  • Corey O Hern

Organizations

  • Army Contracting Command
  • United States Army
  • Yale University

Tags

Readers

  • Computational Fluid Dynamics (CFD)
  • Parallel and Distributed Computing.
  • Powder metallurgy of Titanium alloys.