CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Running, Solving & CFD (http://www.cfd-online.com/Forums/openfoam-solving/)
-   -   Unrealistic behavior in twoPhaseEulerFoam (http://www.cfd-online.com/Forums/openfoam-solving/97048-unrealistic-behavior-twophaseeulerfoam.html)

voingiappone February 7, 2012 22:17

Unrealistic behavior in twoPhaseEulerFoam
 
2 Attachment(s)
Howdy!

As the title says I am experiencing an extremely unrealistic behavior in my simulation which runs twophaseeulerfoam.

I am injecting air in an air lift reactor with a flow of 200 ml/min (pretty slow). The outlet is an inletoutlet boundary and the rest of the walls are... well... walls.
The real thing has 4 nozzles in the bottom plate but the mesh (two dimensional) only has 2 which are, of course linear (not holes anymore).
The problems I have are mostly two:

- The gas hold-up is completely wrong. Just injecting air for 1 second makes the free surface of the fluid on the top rise 3 or 4 cm.... in the real thing it just moves a little (still have to make proper measurements). Moreover if I run the 3D case using interFoam I have a completely logic rise (with calculation deriving from the postpro to back-up the thing). However the 2D interFoam too tends to make wrong estimations of the hold-up. Weird....

- The turbulence of the liquid-gas interface. This is really something I would not expect by the solver, expecially because the turbulence model is off.... When the gas phase reaches the interface everything starts to break up and to revolt like it was bubbled with tons of air which is definitely not the case. Weird^2.

I attach an image of what I am saying so you can visualize...
I am quite confused. At least...:confused: There surely is something I am missing, hence the questions:

- If I have 4 nozzles with an inlet U of 1.06 m/s in a 3D mesh which give good results, why if I make a 2D mesh with only 2 nozzles with the same velocity I get this hold-up discrepancy? If U is used to calculate back the flow, the resulting flow (m3/s/m) shoud be halved, right? I don't get the point.

- If in the 3D the interface is so steady and "realistic", why in the 2D it gets that messy?

I ask for the help of somebody more experienced than me in this solver, meanwhile I will try to look more thoroughly through the config files.

Thanks in advance!

Luca

voingiappone February 16, 2012 00:58

It seems like nobody knows the reason why I get these results... BTW I did some modifications. Here is what I did.

Having in the real reactor 4 nozzles and not an uniform bottom surface injecting the gas, it is difficult to realize a real 2D model of the reactor. I chose to leave two nozzles and in a first try I adopted half of the total gas flow for the inlet boundary condition. That's an error... because the distribution of the gas seems augmented by this assumption. I decided then to change approach: I caculated the surfacial average of injected gas and used that for the selected section I chose in the 2D. This lowered the flow and the interface is no longer the messy thing I posted above.

I have another problem now... which it seems to be way more complicated to avoid. I have a perfectly reversed fluid flow in the column when compared to the logic..
The gas rises in the draft tube but the liquid rises in the downcomers! It should create a circular flow rising in the draf tube and descending in the downcomers but I achieve the exact opposite...

Can somebody explain me why? With high gas flows I have the right pattern but with low flows it reverses... Somehow the ascending gas "pushes" the liquid sidewards and it starts to rise in the downcomers.

I tried to initialize the fluid fields to "help" the startup but despite an initial logical behavior, after 1 second in the simulation the flow reverses....

HELP:confused:

robbirobocop February 23, 2012 05:48

Well, I simulate(d) multiphase flow in a steamdrum and from my experience with interFoam, bubbleFoam and twoPhaseEulerFoam a 2D case might be more instable and result in less physical results as well.

Whereas interFoam is the most stable (at least in my opinion) and the results are pretty good, the problem here is the drag that is too high. Thus, steam pushes away water...

Therefore, I checked bubbleFoam and twoPhaseEulerFoam as well Luca.
And the 2D case is more instable than the 3D case...
Why is that? A 2D "test" case always has other assumptions and probably the adjustment from the 3D to the 2D case you made, can result in other problems.

In your example the results are the problems. Would it be possible for you to upload / post some of your adjustments? That is schemes, solution and your boundary conditions. I guess it is hard to help you out of the blue without any further information...

alberto February 24, 2012 04:15

Quote:

Originally Posted by voingiappone (Post 343238)
Howdy!

As the title says I am experiencing an extremely unrealistic behavior in my simulation which runs twophaseeulerfoam.

I am injecting air in an air lift reactor with a flow of 200 ml/min (pretty slow). The outlet is an inletoutlet boundary and the rest of the walls are... well... walls.
The real thing has 4 nozzles in the bottom plate but the mesh (two dimensional) only has 2 which are, of course linear (not holes anymore).
The problems I have are mostly two:

- The gas hold-up is completely wrong. Just injecting air for 1 second makes the free surface of the fluid on the top rise 3 or 4 cm.... in the real thing it just moves a little (still have to make proper measurements). Moreover if I run the 3D case using interFoam I have a completely logic rise (with calculation deriving from the postpro to back-up the thing). However the 2D interFoam too tends to make wrong estimations of the hold-up. Weird....

The incorrect prediction of the expansion is typical of 2D simulations of multiphase flows, and it is not limited to your case.

If you want a realistic representation of your system, you have to run a 3D case, since the effect of inlets and walls will be significantly different (think to the flow pattern induced by four inlets, for example), and also the structures of the flow are not 2D.

Quote:

- The turbulence of the liquid-gas interface. This is really something I would not expect by the solver, expecially because the turbulence model is off.... When the gas phase reaches the interface everything starts to break up and to revolt like it was bubbled with tons of air which is definitely not the case. Weird^2.
Is the solution numerically stable, and are the residuals of the equations converging at each time step? Could you post a picture of your mesh?

Quote:

- If I have 4 nozzles with an inlet U of 1.06 m/s in a 3D mesh which give good results, why if I make a 2D mesh with only 2 nozzles with the same velocity I get this hold-up discrepancy? If U is used to calculate back the flow, the resulting flow (m3/s/m) shoud be halved, right? I don't get the point.
No, at the inlets you should specify the velocity that gives you the correct mass-flow rate of the phase.

Quote:

- If in the 3D the interface is so steady and "realistic", why in the 2D it gets that messy?
I suspect some numerical problem in your 2D case. It would be useful to see the mesh, and maybe the fvSchemes (if they are different from those used in 3D).

Best,

mwaqas December 3, 2014 15:01

Hello everyone

I am sorry for posting my question here, but it is somehow related to the discussion.
I am working with fluidizedBed/RAS (twoPhaseEuler) model. I am using 2.3.0. my simulation is working fine. Pressure drop and bed expansion ration is correct. but I am facing a problem in mass conservation. When I apply following command in terminal window

patchIntegrate phi inlet/outlet

the inlet and outlet mass should remain conserve, but outlet mass is 5-6% higher than the inlet mass. One probable reason may be the solid particle elutriation, but it not the case, as no particle is moving outside the domain because of particle boundary condition. also the following command shows zero particle's flow rate in/out of the domain.

patchIntegrate phi.particles inlet/outlet

Anyone has idea what could be possible error

alberto December 3, 2014 19:03

This was addressed in 2.3.x with the fully conservative form of the equations. However, the price to pay is less stability / smaller time-steps.

Best,

mwaqas December 5, 2014 16:17

Hello Alberto

Thank you very much for your reply. I have also installed OF2.3.x and ran my case. I have following queries.

  1. How can we check mass balance in OF2.3.x (air at inlet and outlet) as the "patchIntegrate phi inlet/outlet" is not working in 2.3.x
  2. I ran my case in 2.3.x, transferred it to 2.3.0 and there checked air mass balance (patchIntegrate phi inlet/outlet). Again it is giving some extra mass flow rate from out (5-6%). But there is no elutriation of particles with (checked by following command patchIntegrate phi.particles inlet/outlet), then why it is giving extra flow rate
Regards

mwaqas December 5, 2014 17:33

Hello
I am sorry for another quick comment, I want to correct little bit my previous comment

patchIntegrate phi inlet/outlet is working for 2.3.x also but outlet flow rate is 5-6% more than the inlet flow rate. And there is no particle's entrainment.

PS: I have downloaded and installed 2.3.x from openfoam.org through git protocol

Regards


All times are GMT -4. The time now is 09:22.