View Issue Details
| ID | Project | Category | View Status | Date Submitted | Last Update |
|---|---|---|---|---|---|
| 0001924 | OpenFOAM | Bug | public | 2015-11-24 15:11 | 2015-11-25 08:18 |
| Reporter | Timm Severin | Assigned To | henry | ||
| Priority | low | Severity | trivial | Reproducibility | always |
| Status | closed | Resolution | suspended | ||
| Platform | GNU/Linux | OS | Ubuntu | OS Version | 14.10 |
| Summary | 0001924: LaunderSharmaKE crashes when epsilon = 0 | ||||
| Description | While initialising the LaunderSharmaKE turbulence model, the solver crashes when epsilon is set to zero at walls. However, zero is the value that should be set for those models. Of course the problem can be avoided by setting it to e.g. 1e-08, but that might not be obvious, and the error (floating point exception) is not that informative. I suggest modifying src/turbulenceModels/incompressible/RAS/LaunderSharmaKE/LaunderSharmaKE.C by adding VSMALL to epsilon in a number of equations. | ||||
| Steps To Reproduce | E.g. in tutorial incompressible/boundaryFoam/boundaryLaunderSharma, set epsilon at a wall to 0, run the case. | ||||
| Tags | No tags attached. | ||||
|
|
Please check OpenFOAM-3.0.? and/or OpenFOAM-dev as I believe this issue is resolved in the new turbulence model framework. |
|
|
I just tried the tutorial case above with built 3.0.x-e48f65e34d9e, there it doesn't work. I'm updating my 3.0.x right now, and might give -dev a try tomorrow. Apart from that VSMALL seems not to work, ROOTVSMALL looks more promising so far. If the issue has not been solved with the current version, I might be able to provide a patch. Will come back to this tomorrow. |
|
|
You are right, models based on epsilon-tilda should cope with epsilon being 0 at the wall whereas those based on epsilon need not as epsilon should not be 0 at the wall. In order to handle the division by epsilon-tilda correctly the source-term expessions should not evaluate the BCs but doing this elegantly will require some more supporting framework in GeometricField and DimensionedInternalField which is on the list of things to do but not very high priority. In the meantime it is necessary to set epsilon-tilda to a small but non-zero value at the wall as is currently done in the tutorials. |
|
|
Shouldn't it still evaluate for BCs, where epsilon != 0? E.g. slip boundaries, or maybe (here I'm not sure) moving walls? However, I take it that this problem is to be solved later, so I assume there is no point in supplying a patch where the division is "fixed". |
|
|
The BCs need to be evaluated but the sources of the equations do not need to be evaluated for the boundaries as the information is not used. |
| Date Modified | Username | Field | Change |
|---|---|---|---|
| 2015-11-24 15:11 | Timm Severin | New Issue | |
| 2015-11-24 16:14 | henry | Note Added: 0005668 | |
| 2015-11-24 16:22 | Timm Severin | Note Added: 0005669 | |
| 2015-11-24 16:51 | henry | Note Added: 0005671 | |
| 2015-11-25 08:09 | Timm Severin | Note Added: 0005675 | |
| 2015-11-25 08:16 | henry | Note Added: 0005676 | |
| 2015-11-25 08:18 | henry | Status | new => closed |
| 2015-11-25 08:18 | henry | Assigned To | => henry |
| 2015-11-25 08:18 | henry | Resolution | open => suspended |
