Implementing the Direct Simulation Monte Carlo (DSMC) Algorithm using Field-Programmable Gate Arrays (FPGA) for Predicting Behaviours of Rarefied Flows

Abstract

The Direct Simulation Monte Carlo (DSMC) method has proved to be a very useful computational tool, particularly for predicting the behaviours of rarefied flows in small-scale or high-altitude fluid environments, where the continuum approximation that is assumed for most Navier-Stokes solvers does not apply, by simulating the effects of molecular-scale collisions. While the benefits of the method are well established, DSMC is known to be computationally intensive due to the Monte-Carlo nature of the simulation requiring sequential calculation of many random numbers. In this project we propose to implement the DSMC algorithm in hardware on a field-programmable gate array (FPGA) to determine the feasibility and relative advantages in terms of massive parallelization and power reduction that can be realised compared to the typical implementation on a CPU or GPU architecture. In this project we will investigate which parts of the DSMC algorithm can be advantageously implemented on a modern FPGA architecture, and how the algorithm can be modified to best suit the advantages and disadvantages of the FPGA. We will compare the power consumption and the speed of solution on the FPGA with a standard CPU-based implementation of the algorithm.

Document Details

Document Type
DoD Grant Award
Publication Date
Feb 16, 2024
Source ID
FA23862314106

Entities

People

  • David Petty

Organizations

  • Air Force Office of Scientific Research
  • United States Air Force
  • University of New South Wales

Tags

Readers

  • Computational Fluid Dynamics (CFD)
  • Parallel and Distributed Computing.
  • Systems Analysis and Design