Strategies for Efficient Particle Resolution in the Direct Simulation Monte Carlo Method

Abstract

The accuracy and computational cost of a direct simulation Monte Carlo simulation are directly related to the number of particles per cell. Optimal computational efficiency is achieved when the minimum number of particles needed for accurate resolution is used in each cell. Particle count is shown to scale proportionally with the inverse of gas density. This indicates that high density regions will tend to have few particles while low density regions are over resolved. Three methods of controlling the distribution of particles are presented—direct variation of particle weights, variation of time steps, and grid manipulation. A combination of time step variation and grid manipulation is indicated to be the most effective strategy. A sample plume expansion problem is used to demonstrate these strategies. Computational savings of up to an order of magnitude are observed.