Parallelization of a finite volumes discretization for anisotropic diffusion problems using an improved Schur complement technique

Hassan Belhadj; Samir Khallouq; Mohamed Rhoudaf

doi:10.3934/dcdss.2020260

Article Contents

2021, Volume 14, Issue 7: 2075-2099. Doi: 10.3934/dcdss.2020260

This issue Previous Article Fractional and fractal advection-dispersion model Next Article An analysis of tuberculosis model with exponential decay law operator

Parallelization of a finite volumes discretization for anisotropic diffusion problems using an improved Schur complement technique

1.
Abdelmalek Essaadi University, Faculty of Sciences and Techniques of Tangier, Laboratory of Mathematics and Applications, Department of Mathematics, BP. 416, Tangier 90000, Morocco
2.
Moulay Ismail University of Meknes Faculty of Sciences and Techniques of Errachidia Department of Mathematics MSISI Laboratory, AM2CSI Group BP. 509, Boutalamine Errachidia 57000, Morocco
3.
Moulay Ismail University of Meknes, Faculty of Sciences of Meknes, Department of Mathematics and Informatics, BP. 11201 Zitoune Meknes, Morocco

^* Corresponding author: Samir Khallouq
^* Corresponding author: Samir Khallouq

Received: February 28, 2019

Revised: May 31, 2019

Early access: September 2020

Published: July 2021

Abstract Full Text(HTML) Figure(11) / Table(6) Related Papers Cited by

Abstract

We present in this paper a new algorithm combining a finite volume method with an improved Schur complement technique to solve $ 2D $ anisotropic diffusion problems on general meshes. After having proved the convergence of the finite volume method, we have given a description of the proposed algorithm in the case of two nonoverlapping subdomains. Several numerical tests are achieved which illustrate the theoretical results of convergence of the finite volume method and show the advantages of the proposed algorithm.

Keywords:

Mathematics Subject Classification: Primary: 76R50, 65N08, 65N55; Secondary: 65Y05.

Citation:

Full Text(HTML)

Figure 1. DDFV mesh

Download: Full-size image PowerPoint slide

Figure 2. Diamond cell $ \mathcal{D}_{\sigma,\sigma^{*}} $

Download: Full-size image PowerPoint slide

Figure 3. A DDFV mesh $ \mathcal{T} $ of the whole domain $ \Omega $

Download: Full-size image PowerPoint slide

Figure 4. The compatible meshes $ \mathcal{T}_{1} $ and $ \mathcal{T}_{2} $, of the whole domain $ \Omega $, corresponding to the DDFV mesh $ \mathcal{T} $ of figure 3

Download: Full-size image PowerPoint slide

Figure 5. Two independent DDFV meshes $ \mathcal{T}_{1} $ and $ \mathcal{T}_{2} $ for both subdomains $ \Omega_{i} $

Download: Full-size image PowerPoint slide

Figure 6. The compatible meshes corresponding to two independent DDFV meshes $ \mathcal{T}_{1} $ and $ \mathcal{T}_{2} $ of figure 5

Download: Full-size image PowerPoint slide

Figure 7. Primal mesh $ Mesh_{1} $ (left) and $ Mesh_{2} $ (right)

Download: Full-size image PowerPoint slide

Figure 8. Analytic solution (left) and DDFV solution (right) for the test case $ 1 $ and $ h = 0.0118 $

Download: Full-size image PowerPoint slide

Figure 9. Analytic solution (left) and DDFV solution (right) for the test case $ 2 $ and $ h = 0.0207 $

Download: Full-size image PowerPoint slide

Figure 10. Primal mesh $ Mesh_{2,1} $ (left) and $ Mesh_{1,3} $ (right)

Download: Full-size image PowerPoint slide

Figure 11. CPU time to solve the test case $ 2 $ for different numbers of diamonds $ \mathcal{D} $

Download: Full-size image PowerPoint slide

Table 1. $ L^{\infty}(\Omega) $ and $ L^{2}(\Omega) $ errors by DDFV, for test case $ 1 $ and for different values of $ h $

Mesh	Nbr of $ \mathcal{D} $	$ h $	$ e_{h,L^{\infty}} $	$ EOC_{L^{\infty}} $	$ e_{h,L^{2}} $	$ EOC_{L^{2}} $
$ Mesh_{1} $	$ 488 $	$ 0.1184 $	$ 0.0037 $	-	$ 5.3348 $E-$ 04 $	-
$ Mesh_{2} $	$ 1912 $	$ 0.0645 $	$ 8.5460 $E-$ 04 $	$ 2.4127 $	$ 1.3259 $E-$ 04 $	$ 2.2920 $
$ Mesh_{3} $	$ 7568 $	$ 0.0364 $	$ 2.5019 $E-$ 04 $	$ 2.1472 $	$ 3.3329 $E-$ 05 $	$ 2.4136 $
$ Mesh_{4} $	$ 30112 $	$ 0.0207 $	$ 8.2489 $E-$ 05 $	$ 1.9658 $	$ 8.3355 $E-$ 06 $	$ 2.4554 $
Average				$ 2.1752 $		$ 2.3870 $

| Show Table

DownLoad: CSV

Table 2. $ L^{\infty}(\Omega) $ and $ L^{2}(\Omega) $ errors by DDFV, for test case $ 2 $ and for different values of $ h $

Mesh	Nbr of $ \mathcal{D} $	$ h $	$ e_{h,L^{\infty}} $	$ EOC_{L^{\infty}} $	$ e_{h,L^{2}} $	$ EOC_{L^{2}} $
$ Mesh_{1} $	$ 488 $	$ 0.1184 $	$ 0.2402 $	-	$ 0.0452 $	-
$ Mesh_{2} $	$ 1912 $	$ 0.0645 $	$ 0.0423 $	$ 2.8592 $	$ 0.0106 $	$ 2.3876 $
$ Mesh_{3} $	$ 7568 $	$ 0.0364 $	$ 0.0132 $	$ 2.0356 $	$ 0.0028 $	$ 2.3269 $
$ Mesh_{4} $	$ 30112 $	$ 0.0207 $	$ 0.0048 $	$ 1.7922 $	$ 7.3969 $E-$ 04 $	$ 2.3584 $
Average				$ 2.2290 $		$ 2.3576 $

| Show Table

DownLoad: CSV

Table 3. $ L^{\infty}(\Omega) $ and $ L^{2}(\Omega) $ errors by DDFV-SC, for the test case $ 1 $ and for different mesh $ Mesh_{i,i} $ with $ i = 1,\ldots,4 $

Mesh	Nbr of $ \mathcal{D} $ in $ \Omega_{1}\cup \Omega_{2} $	$ h $	$ e_{h_{1,2},L^{\infty}} $	$ EOC_{L^{\infty}} $	$ e_{h_{1,2},L^{2}} $	$ EOC_{L^{2}} $
$ Mesh_{1,1} $	$ 546 $	$ 0.1191 $	$ 0.0020 $	$ - $	$ 5.4853 $E-$ 04 $	$ - $
$ Mesh_{2,2} $	$ 2124 $	$ 0.0626 $	$ 5.6512 $E-$ 04 $	$ 1.9650 $	$ 1.3729 $E-$ 04 $	$ 2.1535 $
$ Mesh_{3,3} $	$ 8376 $	$ 0.0354 $	$ 2.0197 $E-$ 04 $	$ 1.8050 $	$ 3.4357 $E-$ 05 $	$ 2.4301 $
$ Mesh_{4,4} $	$ 33264 $	$ 0.0200 $	$ 7.0126 $E-$ 05 $	$ 1.8527 $	$ 8.5752 $E-$ 06 $	$ 2.4308 $
Average				$ 1.8742 $		$ 2.3381 $

| Show Table

DownLoad: CSV

Table 4. $ L^{\infty}(\Omega) $ and $ L^{2}(\Omega) $ errors by DDFV-SC, for the test case $ 2 $ and for different mesh $ Mesh_{i,i} $ with $ i = 1,\ldots,4 $

Mesh	Nbr of $ \mathcal{D} $ in $ \Omega_{1}\cup \Omega_{2} $	$ h $	$ e_{h_{1,2},L^{\infty}} $	$ EOC_{L^{\infty}} $	$ e_{h_{1,2},L^{2}} $	$ EOC_{L^{2}} $
$ Mesh_{1,1} $	$ 546 $	$ 0.1191 $	$ 0.3694 $	$ - $	$ 0.0980 $	$ - $
$ Mesh_{2,2} $	$ 2124 $	$ 0.0626 $	$ 0.0733 $	$ 2.5145 $	$ 0.0152 $	$ 2.8975 $
$ Mesh_{3,3} $	$ 8376 $	$ 0.0354 $	$ 0.0191 $	$ 2.3592 $	$ 0.0036 $	$ 2.5267 $
$ Mesh_{4,4} $	$ 33264 $	$ 0.0200 $	$ 0.0053 $	$ 2.2452 $	$ 8.6492 $E-$ 04 $	$ 2.4976 $
Average				$ 2.3730 $		$ 2.6406 $

| Show Table

DownLoad: CSV

Table 5. $ L^{\infty}(\Omega) $ and $ L^{2}(\Omega) $ errors by DDFV-SC, for the test case $ 3 $ and for different mesh $ Mesh_{i,j} $ with $ i,j = 1,\ldots,4 $

Mesh	Nbr of $ \mathcal{D} $ in $ \Omega_{1} $	Nbr of $ \mathcal{D} $ in $ \Omega_{2} $	$ h_{1} $	$ h_{2} $	$ e_{h_{1,2},L^{\infty}} $	$ e_{h_{1,2},L^{2}} $
$ Mesh_{1,1} $	$ 273 $	$ 273 $	$ 0.1191 $	$ 0.1187 $	$ 0.0238 $	$ 0.0068 $
$ Mesh_{2,1} $	$ 1062 $	$ 283 $	$ 0.0626 $	$ 0.1187 $	$ 0.0195 $	$ 0.0053 $
$ Mesh_{2,2} $	$ 1062 $	$ 1062 $	$ 0.0626 $	$ 0.0621 $	$ 0.0129 $	$ 0.0035 $
$ Mesh_{3,2} $	$ 4188 $	$ 1082 $	$ 0.0354 $	$ 0.0621 $	$ 0.0103 $	$ 0.0028 $
$ Mesh_{3,3} $	$ 4188 $	$ 4188 $	$ 0.0354 $	$ 0.0347 $	$ 0.0066 $	$ 0.0018 $
$ Mesh_{4,3} $	$ 16632 $	$ 4228 $	$ 0.0200 $	$ 0.0347 $	$ 0.0053 $	$ 0.0014 $
$ Mesh_{4,4} $	$ 16632 $	$ 16632 $	$ 0.0200 $	$ 0.0191 $	$ 0.0033 $	$ 9.1142 $E-$ 04 $

| Show Table

DownLoad: CSV

Table 6. CPU time to solve the test case $ 1 $ for different numbers of diamonds $ \mathcal{D} $

Nbr of $ \mathcal{D} $ in $ \Omega $	CPU time for DDFV (in seconds)	CPU time for DDFV-SC (in seconds)
$ 8376 $	$ 18.149494 $	$ 24.458821 $
$ 20860 $	$ 94.936105 $	$ 59.901873 $
$ 33264 $	$ 418.468599 $	$ 229.306277 $
$ 82920 $	$ 2899.437600 $	$ 1285.708950 $
$ 132576 $	$ 18312.136003 $	$ 3947.530428 $
$ 330960 $	$ 126004.456142 $	$ 7801.259200 $

| Show Table

DownLoad: CSV

Related Papers

Cited by

References

[1]	L. Agelas and R. Masson, Convergence of the finite volume MPFA O scheme for heterogeneous anisotropic diffusion problems on general meshes, C. R. Acad. Sci. Paris. Ser. I, 346 (2008), 1007-1012. doi: 10.1016/j.crma.2008.07.015.
[2]	B. Andreianov, F. Boyer and F. Hubert, Discrete duality finite volume schemes for Leray-Lions type elliptic problems on general $2D$ meshes, Num. Meth. PDE., 23 (2007), 145-195. doi: 10.1002/num.20170.
[3]	H. Belhadj, M. Fihri, S. Khallouq and N. Nagid, Optimal number of schur subdomains: Application to semi-implicit finite volume discretization of semilinear reaction diffusion problem, Discrete Contin. Dyn. Syst. Ser. S, 11 (2018), 21-34. doi: 10.3934/dcdss.2018002.
[4]	P. E. Bjørstad and O. B. Widlund, Iterative methods for the solution of elliptic problems on regions partitioned into substructures, SIAM J. Numer. Anal., 23 (1986), 1097-1120. doi: 10.1137/0723075.
[5]	F. Boyer, F. Hubert and S. Krell, Non-overlapping Schwarz algorithm for solving two-dimensional m-DDFV schemes, IMA J. Numer. Anal., 30 (2010), 1062-1100. doi: 10.1093/imanum/drp001.
[6]	S. C. Brenner, The condition number of the Schur complement in domain decomposition, Numer. Math., 83 (1999), 187-203. doi: 10.1007/s002110050446.
[7]	T. F. Chan, E. Weinan and J. Sun, Domain decomposition interface preconditioners for fourth-order elliptic problems, Appl. Numer. Math., 8 (1991), 317-331. doi: 10.1016/0168-9274(91)90072-8.
[8]	Y. Coudière, J.-P. Vila and P. Villedieu, Convergence rate of a finite volume scheme for a two dimensional convection-diffusion problem, ESAIM:M2AN, 33 (1999), 493-516. doi: 10.1051/m2an:1999149.
[9]	K. Domelevo and P. Omnès, A finite volume method for the Laplace equation on almost arbitrary two-dimensional grids, ESAIM:M2AN, 39 (2005), 1203-1249. doi: 10.1051/m2an:2005047.
[10]	R. Eymard, T. Gallouët and R. Herbin, Convergence of finite volume schemes for semilinear convection diffusion equations, Numer. Math., 82 (1999), 91-116. doi: 10.1007/s002110050412.
[11]	R. Eymard, T. Gallouët and R. Herbin, Discretization of heterogeneous and anisotropic diffusion problems on general nonconforming meshes SUSHI: A scheme using stabilization and hybrid interfaces, IMA J. Numer. Anal., 30 (2010), 1009-1043. doi: 10.1093/imanum/drn084.
[12]	R. Eymard, T. Gallouët and R. Herbin, Finite volume methods, in Handbook of Numerical Analysis (eds. P.G. Ciarlet and J.L. Lions), Elsevier, 7 (2000), 713–1020.
[13]	R. Herbin and F. Hubert, Benchmark on discretization schemes for anisotropic diffusion problems on general grids, in ISTE. Finite Volumes for Complex Applications V (eds. R. Eymard and J.-M. Hérard), Wiley, 5 (2008), 659–692.
[14]	F. Hermeline, A finite volume method for the approximation of diffusion operators on distorted meshes, J. Comput. Phys., 160 (2000), 481-499. doi: 10.1006/jcph.2000.6466.
[15]	L. Mansfield, On the conjugate gradient solution of the Schur complement system obtained from domain decomposition, SIAM J. Numer. Anal., 27 (1990), 1612-1620. doi: 10.1137/0727094.
[16]	A. Quarteroni and A. Valli, Domain Decomposition Methods for Partial Differential Equations, Oxford University Press, 1999.