On Fri, 3 Oct 2008, Adam C Powell IV wrote: > On Fri, 2008-10-03 at 10:50 -0500, Roy Stogner wrote: >> On Thu, 2 Oct 2008, Nasser Mohieddin Abukhdeir wrote: >> >>> Moving back to John's point about the previously mentioned >>> Predictor/Corrector algorithm (assuming we compute everything >>> correctly), I'm assuming that idea is trashed. >> >> The idea that we can use the first Newton step as an explicit solution >> is wrong, but using a forward Euler solve for the predictor and >> following it up by a backward Euler Newton solve for the corrector >> should still work just fine. > > Are you sure about that?
No, I'm not. It will depend on your problem, and I've only got the half-remembered anecdotal experience of others to suggest that this might be useful on common PDEs. > It seems that if delta t is significantly larger than the forward > Euler stability criterion, this would make an awful predictor. In > the past I've used a linear projection of the previous two timesteps > into the new time as the predictor, and saved a couple of Newton > iterations vs. using the previous timestep alone, which is more > stable than a forward Euler step, but doesn't always work well > either... I'm hoping that we can write the PC solver to be agnostic to the solvers used for each phase - that way if it turns out that Forward Euler -> Backward Euler is a disaster, we can switch to Extrapolation -> Backward Euler (or Adams-Bashforth -> Crank-Nicholson, which IIRC is what I've heard success stories about) or whatever without extra work. --- Roy ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ Libmesh-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/libmesh-devel
