1989 Fiscal Year Final Research Report Summary
Development of general-purpose Navier-Stokes code for parallel computers
| Project/Area Number |
63850041
|
|---|
| Research Category |
Grant-in-Aid for Developmental Scientific Research
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
Fluid engineering
|
|---|
| Research Institution | Kyoto Institute of Technology |
|---|
Principal Investigator |
SATOFUKA Nobuyuki KIT, Eng. and Design, Professor, 工芸学部, 教授 (30027891)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
NISHIDA Hidetoshi KIT, Eng. and Design, Assistant, 工学部, 助手 (40164561)
MORINISHI Koji KIT, Eng. and Design, Associated Professor, 工芸学部, 助教授 (20174443)
TOKUNAGA Hiroshi KIT, Eng. and Design, Associated Professor, 工芸学部, 助教授 (10027906)
|
|---|
| Project Period (FY) |
1988 – 1989
|
| Keywords | Parallel computer / Navier-Stokes equations / Rational Runge-Kutta scheme / Group explicit method / Finite difference method / Domain decomposition method |
|---|
| Research Abstract |
The purpose of this research is to develop a general-purpose Navier-Stokes code for parallel computers. In this research, four subjects, i.e., devising a numerical scheme which is compatible with parallel computers, improvement of turbulence model, establishment of grid generation methods, and programming and evaluation of the Navier-Stokes code, were carried out in priority. Two numerical schemes were investigated. One is a method of lines approach, the other is a group explicit (G.E.) method. The method of lines approach is based on a central finite difference approximation for discretizing spatial derivatives and a rational Runge-Kutta (RRK) scheme for subsequent time integration of resulting system of ordinary differential equations in time. The G.E. method is based on the use of asymmetric finite difference approximation to groups of 2 adjacent points successively on the grid which yielding a set of implicit equations which can be easily converted to explicit form. The domain decomposition technique is incorporated into these methods for parallel computing. The speedup of these methods on parallel computer are 3.51 in the case of 4 processors. The experimented data in the case of minimum inter-processor communication overhead comes very close to linear speedup and gives high efficiency of 0.91. These results were presented at 11th International Conference on Numerical Methods in Fluid Dynamics(1988) and International Symposium on Computational Fluid Dynamics-Nagoya(1989). The method of lines approach combined with the domain decomposition method seems to be valuable approach to the general-purpose Navier-Stokes code on parallel computers.
|
