|December 15, 2015, 06:55||
Possibly serious MRF implementation issue
Join Date: Jun 2015
Posts: 24Rep Power: 3
I'm trying to validate Openfoam's numerical machinery for simulating mechanically agitated / stirred tanks. This includes the use of MRF and the Herschel-Bulkley non-newtonian model. However the results are not satisfactory. So have been trying to see where the “error” stems from. So for now have reduced the problem to case close to Taylor Couette flow with a Newtonian fluid inside. Basically a shaft rotating within a fixed outer cylinder. The shaft is split into three parts. The middle part is included in the MRF zone, while the upper and lower parts are retained in the stationary zone. Their motion is modeled by the “rotatingWallVelocity” boundary condition. The top and lower walls of the cylindrical tank are modeled as slip boundary condition. The U velocity mag at a longitudinal slice along the axis of the cylinder is shown in UmagSliceCouette.png. We can observe some unphysical flow features in the cylinder mid-section whose cells are included in the MRF region.
Now further simplifying the problem, a case of still liquid in a tank with the MRF region rotating at some angular speed is considered. The result should be of course zero for velocity field, but as the result in the figure shows this is not the case at all (see UmagSliceNoShaft.png and UmagGlobalNoShaft.png). These deviation well perhaps is rather negligible for high Re applications such as Wind turbine simulations, but for stirred tank reactors of high viscous fluids which mainly operate in the laminar regime, this can result in visible discrepancies.
I also tried the case of periodic BCs for the top and bottom to reduce their influence on the results by using cyclicAMI, but a similar observation is made (see UmagCyclicNoShaftCylic.png and UmagGlobalNoShaftCylic.png).
I highly doubt it if this is due to the generated mesh? I used snappyHexMesh to generate the MRF zone. I was suggested that describing the zone, through a similar procedure as AMI (using baffles) would perhaps yield better results. I also tried this approach but again I observed similar results as for the earlier cases obtained through sHM-created MRF zone (see UmagSliceNoShaftAMI.png and UmagGlobalnoShaftAMI.png).
I've included the case files in the following weTransfer link (
I'm using OpenFOAM 2.4.x, but I think implementation wise, same holds for OpenFOAM 3.0.x.
I very much hope some one with experience would give their thoughts on this, perhaps suggesting means to fix this, or perhaps this is something quite trivial which I'm not seeing!
Last edited by Ali Blues; December 15, 2015 at 07:21. Reason: adding clarity
|December 16, 2015, 07:04||
Join Date: Mar 2009
Posts: 64Rep Power: 9
This is not a bug in OpenFOAM, but a limitation of MRF. A similar issue was discussed in this thread:
Should an empty MRF zone really induce this much cross-flow?
|bug, discontinuity, mrf, openfoam|
|Thread||Thread Starter||Forum||Replies||Last Post|
|Compiling issue regarding to Solid Particle injector's(solidParticle) implementation||liguifan||OpenFOAM Programming & Development||4||May 25, 2016 11:30|
|SA turbulence model implementation issue||jaiswal||Main CFD Forum||8||February 10, 2014 23:42|
|simpleFoam serious mass balance issue||fivos||OpenFOAM Running, Solving & CFD||2||November 6, 2011 09:21|
|implementation issue for kOmega SAS model||khurram||OpenFOAM||0||February 8, 2011 04:57|
|Implementation Issue||CFDN00b||Main CFD Forum||0||September 3, 2009 04:28|