interFoam: Solve gives better results than MULES
Hi everyone,
I have been toying with interFoam lately. I have observed that solve method gives better results than MULES::explicitSolve. When I switched off the U and p equations and added a source term to alpha evolution equation, I got better analytical results using solve than with MULES. Has anyone seen effect like this before? Thanks, Hrushi |
You are not solving for U and p, and are comparing with analytical results, so I am assuming your case is steady state? Would you be able to apply your solver on, e.g., the dam break case?
|
Hi Bernhard,
Sorry for a delayed reply. I am out of town and little access to my results. I will try to get some pictures before eve and put them online. I am solving for a transient problem and not steady state one. Its a modified alpha equation with source term. I first tried it on damBreak case, but water column did not break. In fact, I got some unphysical results in the dam itself. My UEqn also does not seem to be working. Since then, I am trying to debug by trying various options. One of them was to remove U and p equations and see how the alpha equation is solved. I am trying different solutions for solve on randomly initialized alpha field. But to my surprise, mules gave a very uniformly distributed alpha while solve method gave me a distinction between different phases, which is the typical solution with extra source term. Not sure why my UEqn fails and nor do I have any idea why alpha equation should give me unphysical answer for same source term. Thanks, Hrushi |
1 Attachment(s)
Hi Bernhard,
Following is my solve output. I could not get the MULES output but I can describe that MULES gives much diffuse interface than solve. Does it give you an idea? Thanks Hrushi |
Hi, probably solve give you less diffusive solutions, but in some cases mass conservation could not be guaranteed, MULES is more robust in this sense.
Regards. |
Hi Santiago,
Thanks for your reply. I have read part of your thesis where you have explained MULES solver. But I found that if we decrease the time step by one order of magnitude, we get better results and at the same time, it has as good conservation as MULES. Any idea why? Regards, Hrushi |
Hi, MULES as other FCT methods make a blending between a first order flux and and a high order flux which is in somewhat fixed. Using TVD methods with solver allows you to use different limiters to obtain a solution as sharp as you want (within the possible range of solution given by the limiters).
The FCT version coded in MULES is iterative, which is a difference with the original Zalesak method, going up in iterations recovers the original method, with less iterations it is less diffusive, but at the same time more unstable I guess, you could try to change the number of iterations in the MULESLimiter method (they are hardcoded). Regards |
All times are GMT -4. The time now is 02:59. |