No label defined (Q3715486)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 3944357
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | No label defined |
scientific article; zbMATH DE number 3944357 |
Statements
1984
0 references
numerical solutions of singular perturbation problems
0 references
Mixed Defect Correction
0 references
Mixed Defect Correction Process
0 references
local mode analysis
0 references
boundary analysis
0 references
0.7865022420883179
0 references
0.7838307023048401
0 references
0.7810423970222473
0 references
0.7798013687133789
0 references
[For the entire collection see Zbl 0545.00019.] The author presents a complete study for numerical solutions of singular perturbation problems which can be modelized by: (1) \(-\epsilon \Delta u+\vec a\cdot \vec Vu=f\) with a method he calls Mixed Defect Correction. The difficulties to represent simultaneously with good accuracy the smooth parts and the boundary or interior-layers of this kind of solutions are well-known. Various possibilities of improvements are still studied by many authors. The purpose of this work here for problems (1) is to obtain a method in which no information is used about the flow direction \(\vec a\) and where a high order of accuracy is still available in the smooth parts of the solution. On one hand a direct application of D. C. P. (Defect Correction Principle) cannot satisfy this condition but on the other hand with \(D\subset P\) one has the possibility to improve the order of accuracy of a stable low order discretization by means of an accurate but instable higher other method. Taking into account this idea the author has elaborated an MDCP (Mixed Defect Correction Process) in which in each iteration step an improved approximation is obtained by consideration of two (instead of one) discretizations of the given problem. Let be the linear problem: \(Lu=f\) (with boundary conditions) with two associate discrete problems \(L_ h^{\alpha}u_ h^{\alpha}=f_ h^{\alpha}\), \(\alpha =1,2\) and \(\tilde G_ h^{\alpha}\) the approximate universe operators of \(L_ h^{\alpha}\) \((\alpha =1,2)\). The MDCP method consists in applying the iterative process: \(u_ i\mapsto u_{i+1}\) defined by \[ u_{i+1/2}=u_ i- \tilde G^ 1_ h(L^ 1_ hu_ i-f^ 1_ h);\quad u_{i+1}=u_{i+1/2}-\tilde G^ 2_ h(L^ 2_ hu_{i+1/2}-f^ 2_ h). \] After an introduction where the difficulties of numerical solutions for singular perturbation problems are analyzed the author sums up the local mode analysis and D.C.P.. Then he studies MDCP with local mode analysis and asserts the convergence of the iterative process. The boundary analysis of the MDCP solutions is also given and the paper ends with various examples and numerical results which show clearly the interest of MDCP. Proofs are not given in this item, they can be found in previous preliminary papers by the author on the same subject [''Theory and applications of singular perturbations'', Lect. Notes Math. 942, 192- 206 (1982; Zbl 0504.65064) and ''Multigrid methods'', Lect. Notes Math. 960, 485-501 (1982; Zbl 0505.65047), and ''Defect correction for the solution of a singular perturbation problem'' (with \textit{P. M. de Zeeuw}) Scientific Computing, 113-118 (1983)].
0 references