On Sat, Jul 17, 2010 at 5:06 AM, Benny Malengier <benny.maleng...@gmail.com> wrote: > Hi, > > I'd like to know the result of the stokesCavity.py test of some other > people. Do they also fail?
Not for me at least. > When I run the flow/stokesCavity.py, the test fails. Adding at the bottom of > that file: > > Test values in the last cell. >>>> print pressure.getGlobalValue()[...,-1], 145.233883763 > > I obtain the result : > False > False > False It's failing for you. I'd don't think the test is very good though. I might try and improve the test method when I include your version of the example. > Are the RHS the values from Dolphyn? Is this test supposed to fail? I don't believe the test is supposed to fail, but I used Dolfyn originally to compare the results and provide the comparison numbers. A lot has changed since then though. > Without knowing what this test does, there is little conclusion I can draw. > I'll post the Rhie-Chow correction when I tested some more with it. It > defenitely removes the pressure oscillation in my work, so it seems ok. Good. As I said I need to think of a better way to compare the solution. > I also wonder about the part of the solver after: > ## update the pressure using the corrected value but hold one cell fixed > then code follows to keep cell[0] at pressure 0. Is this really needed? That might depend on which solver is used. It is probable that an iterative solver might wander, while a direct solver does not have that issue (or visa-versa). > Not > doing it results in > sweep: 299 , x residual: 0.0196690174713 , y residual 0.0177051252932 , p > residual: 1.70403715572e-07 , continuity: 3.94729816867e-06 > 151.910461226 145.233883763 > 0.202533960386 0.24964673696 > -0.21920158527 -0.164498041783 > which is hardly different from the original. I would prefer Imposing a > pressure value via a boundary condition on the pressure correction equation, > this seems to work great if one sets eg a boundary at a specific pressure to > a fixed value. Okay, if it works that is much cleaner. -- Daniel Wheeler
