Predictor and steplength selection in continuation methods for the Navier-Stokes equations (Q1178550)

The paper deals with continuation methods of the predictor-corrector type for solving the nonlinear equations arising from discretizations of the stationary Navier-Stokes equations. The equations are written in terms of the primitive variables, and Galerkin-type methods are employed, but the results derived may be extended to other formulations as well. The Reynolds number is used as the continuation parameter. Three different predictor methods are considered, and Newton's method is used as the corrector iteration. It is shown that whenever the Navier-Stokes equations possess a unique solution and when one is not in the neighborhood of singular points, then the step size used in predicting an initial guess for the Newton iteration may be chosen largely independently of the predictor employed. The paper ends with some numerical examples illustrating the validity of the theoretical considerations.