bubble flow with pulsating inlet in a large-scale Domain
 

Hello Guys,

first of all you have to excuse for my english, as this isn't my mother tongue.
I am quite new to the world of CFD and i have to do some cfd studies for my internship report on an large scale tank with serveral gas inlets at the bottom, as well as an overlaying water stream entering the tank from the middle of on side and exits on top of the other side.

The aim of the study is then to compare different ways for running the Spargerlines. For example i have to compare one case where all Sparger operate at the same time, with one where only one line is active for a certain time (and switched off after e.g.20s).
So this would be a pulse stream inlet.

Therefore i created a 2.5x6x3m tank with 3x8 lines of Spargers (each Sparger created as 25x25cm surfaces) at the bottom. And rectangular Bodies on both sides through whitch the water enters and exits the domain (see attached figure). Then i run the simulation without any gas entering the domain to achieve a steady solution for the waterflow which can be used as ini file for the transient cases (monodisperse bubble stream).

Right now i already face a lot of problems, for example the finest structured mesh i can create without exaggerating the number of cells has a extension of 2.5cm in each direction which leads to a total number of more than 2.8 million cells. Since I have the opportunity to perform the simulation on a cluster, this number of grid cells is still feasible.
But I know that these are quite big cells, so i wonder if it's still possible to achieve a correct solution?

Another problem is the pulse stream, because since now the solver always stops after a few steps with the error: fatal overflow in linear solver.
The question is if its possible at all to do a converging simulation with that kind of inlets?

Maybe anybody have any suggestion to that? Any help would be really appreciated. My Settings are listed below.
Thanks in advice.

Settings
disperse Phase: Air at 25°C (3mm bubbles)
continuis Phase: Water
bouancy on
Heat Transfer: Isothermal
turbulence: SST


Fluid Pair
Surface tension: 0.072 [N m^-1]
Drag: Ishii Zuber
Lift: Tomiyama
Virtual mass force: Coefficient 0.5
Turbulent viscosity: implemented through expressions
Wall Lubrication Force: Hosokawa (implemented in customized solver)
Turbulent Dispersion Force: Favre Averaged Drag Force

Boundary
Top: Degassing
3 Interfaces: Default Fluid Fluid Interface
WaterOut: static pressure 0
WaterIn: normal velocity (0.016 m/s)
Gasinlet: Bulk massflowrate for 2 cases: 1,8 & 4,4 mm/s (for each case the rate will be 3times larger if only one line is used)
others: Wall (free slip for Gas)

Overflow error: This is an FAQ: http://www.cfd-online.com/Wiki/Ansys...do_about_it.3F

The configuration you have drawn can be modelled in CFX. There will be either a setup error or poor numerical stability causing the divergence.

Where are the interfaces in your simulation?

I do not understand your question "But I know that these are quite big cells, so i wonder if it's still possible to achieve a correct solution?" Can you say it again please?

Thanks for your reply Glenn,
the 3 interfaces are shown in the attached figure. The one at the bottom containing only 1 cell in z-direction.
I had to create these, because it was not possible to build a structed hexa mesh without them. (I use Ansys Meshing)

I thought that the simulation might diverge either beacuse of the big grid cell size (until now there is also no boundary layer created, so all cells are equal in Dimensions) or the pulsating inlet because this could produce numerical instability.

There is no need for those interfaces. You can mesh this body in high-quality hex meshes as a single part with no interfaces. You want to avoid bodies only 1 element thick if possible - so when you fix this and remove the interfaces it might improve your convergence.

To mesh this as a single body in ANSYS meshing you have a few options - the easiests is to use a multi-body part to split it up (but the resulting mesh will be a single part).

Ok i understand that these interfaces might be a problem.
But can you give me just a short explanation how i have to do this? I tried a few things but none of them was working to produce a nice hex-mesh.

I did it that way: Import my created geometry as one single part with surface bodies (which represents all the inlets at the bottom) from solidworks in Designmodeler.
Then i created the inlet faces from surface bodies with boolean (imprint). Now i have one volume with 24 faces on the bottom. In order to split it to get a multi-body part i used "slice by plane" option in DM. The resulting 4 Volumes are now formed to one part by clicking RMB and select "form new part".

If i now open this in ansys meshing, the 4 volumes should be treated as one part. I am right with that?

But its not working in that way so maybe you can give me an advice what i'm doing wrong.

Have a look at the meshing tutorials on the ANSYS Customer page for examples of how to do this. It is easy when you know how.