Application of an Iterative Solution Scheme to MARS

The recent version of MARS code (MARS 3.1) has been enhanced over the previous version of MARS for several important models for the last couple of years. The most prominent enhancement that distinguishes MARS versions 3.X from the previous MARS versions 2.X is the fully multi-dimensional thermal-hydraulic modeling capability in the recent MARS code. This makes it possible to examine detailed behaviors where multi-dimensional effects may occur. This enhancement, however, makes the problem size too large to be able to efficiently handle it by using the default direct sparse matrix solver. To effectively solve the wide-banded system of linear equations for a multi-dimensional component, the latest version of RELAP5 replaced its default solver with the Border Profiled Lower Upper (BPLU) matrix solver which is designed to solve 'nearly-banded' coefficient matrices directly. However, direct matrix solvers usually cause the problems as the matrix size becomes larger because operation count for this method is proportional to third power of the matrix size and the truncation error during the forward and backward sweeps increases. The new iterative sparse matrix solver has been implemented to the latest version of MARS to efficiently solve sparse linear systems. The accompanying advantage of using an iterative solver is that this solver is easy to take advantage of a parallel machine. This paper will briefly describe the iterative solver and the evaluation results