Parallel Computation of Compressible Flows for Implicit Schemes Without Sub-iteration and Modification of the Implicit Solver
- 259 Downloads
Parallel computation for fluid flow problems has become very useful in the end of the last century and will certainly be very popular in the 21st century for handling large scale problems. Implicit schemes are efficient for obtaining steady state solutions and for treating slowly unsteady flows. However, implicit schemes are not inherently parallel and their parallelization usually requires addtional iteration or modification of the implicit solvers. Thus the parallelization of implicit schemes is accessible only to experts specialized in parallization. In [9, 12], a parallel method based on grid overlapping was proposed for inviscid flow computation. This method requires no additional iterations at each time step, and nor modification of the implicit solver. In this paper we analyze and use this method for viscous flow computation. We conclude that optimal parallel efficiency corresponds to an overlapping width determined by CFL,and the results are similar to inviscid problems for a sufficiently large mesh Reynolds number.
KeywordsComputational Fluid Dynamics Interface Condition Single Domain Mesh Point Implicit Scheme
Unable to display preview. Download preview PDF.
- T.F. Chan and D. Goovearts, “Schwartz=Schur, overlapping versus non-overlapping domain decomposition”, Technical Reports, CAM 88–21, UCLA, 1988.Google Scholar
- D. E. Keyes, “Domain decomposition: a bridge between nature and parallel computers”, ICASE Report No. 92–44 (1992).Google Scholar
- A. Lerat, “Multidimensional centered schemes of the Lax-Wendroff type”, CFD Review 1995,J. Wiley, pp.124–140.Google Scholar
- D. Roose and R.V. Driessche, “Parallel computers and parallel algorithms for CFD: an introduction’, AGARD-R.-807, 1995,pp. 1–23.Google Scholar