View Issue Details
ID | Project | Category | View Status | Date Submitted | Last Update |
---|---|---|---|---|---|
0001340 | OpenFOAM | Bug | public | 2014-07-09 08:01 | 2016-02-03 21:31 |
Reporter | Assigned To | henry | |||
Priority | low | Severity | feature | Reproducibility | sometimes |
Status | closed | Resolution | fixed | ||
Platform | all | ||||
Summary | 0001340: turbulent length scale limiter | ||||
Description | I've had some convergence issues for one of my steady state cases that just crashes very abruptly and I've looked at the turbulent viscosity and noticed that it becomes really high sometimes due to epsilon being bound to 'small'. Im using kEpsilon btw. I've been playing around with adding a turbulent length scale limiter and that seems to help especially in the start-up phase of the simulation. I think it would be nice to have this option for all turbulence models though, also it seems like its more reasonable to limit epsilon by using some length scale than just by small. | ||||
Steps To Reproduce | I have just added a ltMax variable to the kEpsilonCoeffs and removed the bound-call after the epsilon equation and then added this after the k equation. bound(k_, kMin_); scalar Cmu34 = pow(Cmu_.value(), 3.0/4.0); forAll(epsilon_, celli) { scalar epsilonMin = Cmu34*pow(k_[celli], 1.5)/ltMax_.value(); if (epsilon_[celli] < epsilonMin) { epsilon_[celli] = epsilonMin; } } | ||||
Tags | No tags attached. | ||||
|
How should ltMax be set? Maybe a core value could be set which is then limited by the distance to the wall but what should the core value be set to? Presumably it should be some function of the Reynolds number or is this unnecessarily complicated? Should epsilon still be limited by the neigbour values (as done by bound) before the length scale limiter is applied? |
|
Ah, I didnt know bound worked that way, I think I'll remove the comment then. It sounds too complicated. I just set its value to 0.1 m, which is a typical length scale for my type of applications. I've kept track on the max value of the length scale during the simulation and once the residuals are down a bit its no longer used. I guess the laplacian goes bonkus when your length scale is a couple of kilometers. Anyways, I just wanted to raise the idea :) |
|
fvOptions are fully supported in the turbulence models in OpenFOAM-dev and so it would be straight forward to create an fvOption to apply such limiters to the turbulence properties. |
Date Modified | Username | Field | Change |
---|---|---|---|
2014-07-09 08:01 |
|
New Issue | |
2014-07-09 09:50 | henry | Note Added: 0003159 | |
2014-07-09 11:28 |
|
Note Added: 0003160 | |
2016-02-03 21:31 | henry | Note Added: 0005907 | |
2016-02-03 21:31 | henry | Status | new => closed |
2016-02-03 21:31 | henry | Assigned To | => henry |
2016-02-03 21:31 | henry | Resolution | open => fixed |