View Issue Details
ID | Project | Category | View Status | Date Submitted | Last Update |
---|---|---|---|---|---|
0003027 | OpenFOAM | Bug | public | 2018-08-03 20:48 | 2018-08-08 16:19 |
Reporter | Darko.Radenkovic | Assigned To | will | ||
Priority | high | Severity | major | Reproducibility | always |
Status | closed | Resolution | no change required | ||
Platform | Ubuntu 16.04 | OS | Other | OS Version | (please specify) |
Product Version | dev | ||||
Summary | 0003027: Particles are not introduced in turb. channel flow with cyclic boundary conditions, when simulation is run on multiple cores | ||||
Description | When solver icoUncoupledKinematicParcelFoam is used for particle tracking in turbulent channel flow, where channel flow is taken from a tutorial, case channel395, with imposed periodic boundary conditions in x and z direction, parallelization leads that particles are not introduced in domain. On single core, solver works as expected. When further PIMPLE procedure is combined with evolving of kinematic cloud, or kinematic colliding cloud, again, parallelization leads that particles are not introduced in domain. On single core, everything works perfectly. Solver pimpleFoam in parallel mode again works normally. But above mentioned combination, does not work in parallel mode. | ||||
Steps To Reproduce | 1. In case channel395, particles are introduced, and for example, solver icoUncoupledKinematicParcelFoam is used. When solver is started on single core, everything is normal. When solver works on multiple cores, particles are not introduced in domain. Case is decomposed using simple method, on 6 cores. | ||||
Additional Information | Problem persists also in OpenFOAM version 1706. Case is attached, with log file as a result of work in parallel mode, and *.png figure of work on single core. | ||||
Tags | No tags attached. | ||||
|
|
|
I discovered that this is not a bug; there should be a line in decomposeParDict: preservePatches (inout1_half1 inout2_half1); (for example in above attachment) and everything works fine. |
|
Yes, preserve patches is necessary in this case as without it the cyclics get converted into processor-cyclics by the decomposition. This means that the injection model now points to a patch that no longer exists, and nothing gets injected. It's not ideal, but that's how it is at the moment. |
Date Modified | Username | Field | Change |
---|---|---|---|
2018-08-03 20:48 | Darko.Radenkovic | New Issue | |
2018-08-03 20:48 | Darko.Radenkovic | File Added: channel395PlusKinematicCloud.zip | |
2018-08-03 20:48 | Darko.Radenkovic | Tag Attached: cyclic | |
2018-08-05 20:23 | Darko.Radenkovic | Note Added: 0009905 | |
2018-08-05 20:33 | Darko.Radenkovic | Tag Detached: cyclic | |
2018-08-08 16:19 | will | Assigned To | => will |
2018-08-08 16:19 | will | Status | new => closed |
2018-08-08 16:19 | will | Resolution | open => no change required |
2018-08-08 16:19 | will | Note Added: 0009926 |