Since 1978
Published in Sarov (Arzamas-16), Nizhegorodskaya oblast

RUSSIAN FEDERAL
NUCLEAR CENTER -
ALL-RUSSIAN RESEARCH INSTITUTE
OF EXPERIMENTAL PHYSICS
 
 Русский |  English
ABOUT EDITORIAL BOARD PUBLICATION ETHICS RULES FOR AUTHORS AUTHORS ARCHIVE MOST RECENT ISSUE IN NEXT ISSUE PAPER OF THE YEAR




THE INCOMPLETE FACTORIZATION METHOD FOR ITERATIVELY SOLVING SYSTEMS OF DIFFERENCE EQUATIONS AND ADAPTATION OF THIS METHOD TO NON-M-MATRICES

V. P. Ginkin, K. G. Chernov, Yu. G. Bartenev, Yu. A. Bondarenko, R. M. Shagaliev, E. B. Shchanikova
VANT. Ser.: Mat. Mod. Fiz. Proc 2009. Вып.3. С. 3-17.

New effective preconditioners - DIF and PIF - of the stabilized method of biadjoint gradients have been developed to solve systems of 2D and 3D elliptic finite-difference equations with asymmetric poorly conditioned M-matrices of coefficients. To solve problems with non-M-matrices, we offer a stable and effective method of eliminating positive non-diagonal elements from a preconditioned matrix with simultaneously increasing its diagonal elements by the summarized value of all the elements eliminated from a given matrix row.
The rate of convergence of the methods has been tested using both Dirichlet and von Neumann test problems for diffusion and diffusion- convection equations and real 9- and 27-diagonal positively defined non-M-matrices.

Keywords: systems of linear equations, iterative methods, method of conjugate gradients, BICGSTAB method, preconditioner, incomplete factorization method, two-dimensional problems, three-dimensional problems, 9-diagonal matrices, 27-diagonal matrices, nonsymmetric matrices, non-M-matrices, paralleling.

THE INCOMPLETE FACTORIZATION METHOD FOR ITERATIVELY SOLVING SYSTEMS OF DIFFERENCE EQUATIONS AND ADAPTATION OF THIS METHOD TO NON-M-MATRICES

V. P. Ginkin, K. G. Chernov, Yu. G. Bartenev, Yu. A. Bondarenko, R. M. Shagaliev, E. B. Shchanikova
VANT. Ser.: Mat. Mod. Fiz. Proc 2009. Вып.3. С. 3-17.

New effective preconditioners - DIF and PIF - of the stabilized method of biadjoint gradients have been developed to solve systems of 2D and 3D elliptic finite-difference equations with asymmetric poorly conditioned M-matrices of coefficients. To solve problems with non-M-matrices, we offer a stable and effective method of eliminating positive non-diagonal elements from a preconditioned matrix with simultaneously increasing its diagonal elements by the summarized value of all the elements eliminated from a given matrix row.
The rate of convergence of the methods has been tested using both Dirichlet and von Neumann test problems for diffusion and diffusion- convection equations and real 9- and 27-diagonal positively defined non-M-matrices.

Keywords: systems of linear equations, iterative methods, method of conjugate gradients, BICGSTAB method, preconditioner, incomplete factorization method, two-dimensional problems, three-dimensional problems, 9-diagonal matrices, 27-diagonal matrices, nonsymmetric matrices, non-M-matrices, paralleling.








[ Back ]


 
 
 
© FSUE "RFNC-VNIIEF", 2000-2024