On some pressure segregation methods of fractional-step type for the finite element approximation of incompressible flow problems (Q2372350): Difference between revisions

The decoupling of pressure from velocity in numerical approximations of incompressible Navier-Stokes equations has been traditionally attempted using different approaches. Perhaps the best known is the use of fractional step methods, although there are other techniques based on the solution of pressure Poisson equation and predictor-multicorrector algorithms. The interest in this type of methods is due to their computational efficiency, since only scaled equations need to be solved. The objective of this paper is to discuss several ways to implement first- and second-order time integration schemes for incompressible Navier-Stokes equations that allow to decouple the calculation of velocities and pressure. The first family is the classical fractional step methods. The approach here is to present the splitting at the pure algebraic level rather than at the space continuous level. The algebraic viewpoint obviates the discussion on pressure boundary conditions. The second family of methods is the one based on a particular pressure Poisson equation presented in [\textit{J. L. Guermond} and \textit{J. Shen}, J. Comput. Phys. 192, No. 1, 262--276 (2003; Zbl 1032.76529)]. The authors' contribution herein is to compare the method based on this approach with fractional step methods. The authors show the equivalence between this momentum-pressure Poisson equation approach and classical splitting methods. The third and last family of methods is of predictor-multicorrector type starting from fractional step methods. The authors propose an iterative scheme the goal of which is to converge to the solution of the monolithic problem. An important part of this paper is devoted to the stability analysis of fractional step methods. Four methods are considered. In the first two, first-order time integration is used, with a first- and second-order splitting errors. For the third and fourth methods the splitting error is of second order, the same as the time integration error. Two of these four stability results are already presented in [\textit{R. Codina}, J. Comput. Phys. 170, No. 1, 112--140 (2001; Zbl 1002.76063)], whereas the other two are new. Numerical examples are presented and illustrated using tables and graphs.