A fractional step method on a special mesh for the resolution of multidimensional evolutionary convection-diffusion problems

Abstract

In this paper we consider numerical schemes for multidimensional evolutionary convection-diffusion problems, where the approximation properties are uniform in the diffusion parameter. In order to obtain an efficient method, to provide good approximations with independence of the size of the diffusion parameter, we have developed a numerical method which combines a finite difference spatial discretization on a special mesh and a fractional step method for the time variable. The special mesh allows a correct approximation of the solution in the boundary layers, while the fractional steps permits a low computational cost algorithm. Some numerical examples confirming the expected behavior of the method are shown.

Robust Numerical Methods for Singularly Perturbed Differential Equations

“…To justify this statement, we follow Clavero et al [CJLS98,Appendix A] in sketching a proof of an S-decomposition of the solution u of (4.1). More details can be found in [Shi92b,Shi], and in fact the spatial domains Ω considered there are n-dimensional with n ≥ 2.…”

A fractional step ELLAM approach to high-dimensional convection–diffusion problems with forward particle tracking

“…The sequential solution of the splitting system of (8), (9) and (10), (11) is called to converge to the exact solution of equation (5) (see [19,24]) if it holds that lim Dt!0 kcðx; y; t nþ1 Þ Àĉðx; y; t nþ1 Þk L 2 ðXÞ ¼ 0 ð12Þ…”

An efficient S-DDM iterative approach for compressible contamination fluid flows in porous media

“…Domain decomposition methods allow the reduction of the sizes of problems by decomposing domain into smaller ones on which the problems can be solved by multiple computers in parallel (see [9,13,18,23,24,29], etc). The splitting technique is another attractive and popular technique to reduce high-dimensional problems to a series of one-dimensional problems at each time step for saving the memory and CPU time (see [11,30,31] and some recent works [4,5,7,8,12,19,20,28], etc). Since non-overlapping methods have low computation and communication cost for each time step, they are preferable for large scale problems on massively parallel machines.…”