A new hybrid algorithm - Navier-Stokes as a DSMC filter

Abstract

Direct Simulation Monte Carlo (DSMC) and NavierStokes (NS) solutions are combined to investigate a coupled approach to solving mixed flow problems. DSMC, a statistical particle simulation technique used primarily for nonequilibrium high-Knudsen-number flows, is extremely computer intensive. On the other hand NS, a continuum fluid approximation valid for low Knudsen number (Kn), is much less expensive to implement. In the proposed hybrid approach, NS and DSMC codes have been used on the same computational domain to compute the properties of a test problem for two different values of Kn (0.07 and 0.14). We first discuss how each code was modified to accept data from the other, both with respect to internal flow variables and boundary conditions. In this process, DSMC may be viewed as providing the physically correct values of local boundary conditions and transport coefficients for the NS equations. The DSMC solution averaged over very few samplings, rather than without averagings, provided input to NS to produce the filtered solution field, which is very similar to DSMC converged solution. Longer use of the NS filter did not show large improvements in the solution. The total computational time for the DSMC computation was reduced by factor of two by using the NS code as a filter in this study. However, the reduction of computational time would be more significant for three-dimensional problems. The conclusion is that the long period in which statistics must be gathered in order to achieve a good DSMC solution can be virtually eliminated by a short pass through the NS computations.