For those who're interested, a test using different solvers yield the following results.
First sign of leakage occurs at time step: 529 - default solver 529 - linearPCG 1597 - linearLU >2000 - linearCGS >2000 - linearGMRES Jonathan has also helpfully informed me of the difference between np.trapz and direct cell volume calculation. The former indeed returns a reduced value for the reason suggested. Thanks again for the kind assistance. Cheers, Yun On Fri, Jul 6, 2012 at 7:31 AM, Daniel Wheeler <[email protected]>wrote: > On Tue, Jul 3, 2012 at 1:34 AM, Yun Tao <[email protected]> wrote: > >> Hm. Reducing the mesh range (l=50. -> l=30.) does seem to introduce >> leaking even when the initial cell volume = 1 (timestep 485). Why would the >> domain matter when the BCs are zero-flux? > > > The cell volumes averages do change, but very slowly. This could be due > round off in the solver. The np.trapz seems to be very different from the > cell average. Is that calculation taking into account that the values are > cell centered so the end points may not be weighted correctly? > > -- > Daniel Wheeler > > _______________________________________________ > fipy mailing list > [email protected] > http://www.ctcms.nist.gov/fipy > [ NIST internal ONLY: https://email.nist.gov/mailman/listinfo/fipy ] > > -- Graduate Group of Ecology Doctoral Candidate Department of Environmental Science and Policy Center for Population Biology University of California, Davis
_______________________________________________ fipy mailing list [email protected] http://www.ctcms.nist.gov/fipy [ NIST internal ONLY: https://email.nist.gov/mailman/listinfo/fipy ]
