On equilibration and sparse factorization of matrices arising in finite element solutions of partial differential equations (Q4944253)

Properties of LU factorization of large sparse square matrices are studied. The authors revive the technique of matrix equilibration, i.e. the method of scaling matrix coefficients in order to improve the quality of the factorization. They examine the effects of equilibrating large and sparse matrices that arise from the discretization and linearization of a class of coupled systems of nonlinear elliptic partial differential equations. Specifically, they test equilibration of matrices arising in the finite element solution of some models of material processing systems. The authors present the practical evidence that scaling is an indispensable companion for sparse factorization algorithms when applied to realistic problems of industrial interest. They conclude that equilibration with the \(\infty\)-norm is superior than equilibration with the 2-norm.