On May 31, 2012, at 5:08 AM, Matej Svejda wrote: > Thanks, I now defined h as a variable that automatically updates and > the speed is much better now. I'm still getting unexact results... > Besides sweeping (which we established doesn't help in my case) and > using smaller timesteps/higher grid resolution is there any way to > improve the accuracy?
As I asked on May 21, 2012, at 11:08 AM: > It is difficult to be sure, but the green curve (Mathematica solution?) seems > to have a small slope at alpha=1, where as the blue curve (FiPy solution?) > looks like it has zero slope. Could there be a difference in the boundary > conditions? FiPy assumes zero flux on all boundaries unless specified > otherwise. Are you sure you are solving both problems with the same boundary conditions? The discrepancies you are seeing are not generally characteristic of poor solution (whether due to excessive timesteps, grid resolution, or inadequate sweeping for non-linearity). Your solution is smooth and basically looks "OK", it just doesn't agree numerically. Poor solutions tend not to look right at all. _______________________________________________ fipy mailing list [email protected] http://www.ctcms.nist.gov/fipy [ NIST internal ONLY: https://email.nist.gov/mailman/listinfo/fipy ]
