American Nuclear Society

Home / Store / Journals / Electronic Articles / Nuclear Science and Engineering / Volume 93 / Number 1

Parallel Processing Algorithms for the Finite Difference Solution to the Navier-Stokes Equations

Joseph M. Doster, Matt B. Richards

Nuclear Science and Engineering

Volume 93 / Number 1 / May 1983 / Pages 69-77


Member Price:$27.00
Member Savings:$3.00

Numerical solutions involving finite difference representations of the equations governing fluid flow, heat conduction, and diffusion processes (including neutron diffusion) usually consist of solving large sparse matrix equations. These matrix equations can be recast into M smaller coupled matrix equations amenable to solution by using M multiple computer processors operating in parallel. A special form of the fluids equations commonly used in nuclear reactor thermal-hydraulic analysis, i.e., one-dimensional flow in closed loop geometry is emphasized. Parallel algorithms for solving these equations are developed and evaluated in terms of computational speed against conventional solutions on a serial machine. Timing studies are performed to assess the efficiency of these methods and to determine the optimum number of parallel processors for these applications.

Questions or comments about the site? Contact the ANS Webmaster.