Hi Jed. I have a similar situation here. I am intentionally testing an iterative method. It requires me to do some handling (for example copying the current solution vector out) right after a nonlinear iteration is finished, but before anything happening of the next nonlinear iteration. Is there any way to do so?
Best, Ling On Thu, Nov 21, 2013 at 5:41 AM, Jed Brown <[email protected]> wrote: > TAY wee-beng <[email protected]> writes: > > > Hi Jed, > > > > I need a solve a Poisson equation, which is part of my finite volume CFD > > code. > > > > The RHS terms are calculated from the velocity. > > > > In the standard mtd, the RHS terms are calculated from the velocity, > > which then solves the Poisson eqn to get the pressure. > > > > Now, to improve the accuracy of the pressure obtained, a paper suggested > > obtaining the pressure from the velocity in 1 iteration. Then use the > > new pressure for the momentum eqn to get a more accurate intermediate > > velocity. The RHS terms are recomputed again using the new velocity, > > which is then used to calculate the new pressure. This cycle goes on > > until the pressure converges to a certain value. > > This is a crude iterative method for solving a coupled problem. I > recommend that you solve the coupled problem. See the section of the > user's manual on solving block problems, and the examples in PETSc that > solve Stokes-type problems. >
