[bobmalaria]

Hi everyone,

I model a surface deposition model in 2D. Because I have a moving box (see picture) in my domain I don't want to use a symmetry axis, even though the model is y-axis symmetric.




While my model converges nicely when I apply y-symmetry, it won't converge when I use the full model. I think that is due to reversed flow in the pressure in and outlet as it often occurs in the first iterations which in turn causes a wrong split of the inlet flow.

Any thoughts on how I could get equal pressures on both sides of the model? I need an equal flow split of the inlet but often the flow fully goes in one direction. (see picture 2 for illustration of my problem)



Thank you very much

[-mAx-]

are you sure your model is symetric if your box is moving?
Despite reversed flow warning, let iterate

[bobmalaria]

Hi max,

in an initial state I left the box stationary, basically to check my model. I expected to see a 50/50 split of the flow at least in the stationary case.

When the box starts moving of course I expect to see a change of the velaocity vectors. But with the initial soution beeing not correct I saw no point in proceeding any further in the moment (I should have said this in the first place, sorry).

I checked the solution several times and after initialization it seems to be better, but with ongoing iteration I will see this tilt towards one direction.

If I change the pressure inlets to wall, which turns the domain into a -20Pa low pressure region, I get the expected flow split. But I need this opening to atmosphere since this is how the real process works...
This finding and the fact that it works when I use a y-symmetry are what made me to assume a problem with the pressure.

Thanky for helping

[-mAx-]

Originally Posted by bobmalaria

in an initial state I left the box stationary, basically to check my model. I expected to see a 50/50 split of the flow at least in the stationary case.

you mean you compute at first your model without Moving Mesh, in steady state?
And you don't get a converged symmetrical result?

[bobmalaria]

hi,

exactly. I thought at first I do a steady state simulation and later I initialize from this solution and change to the transient solver method.

The model converges, but looking at the velocity vectors, I can still see the tilt towards one direction. Also the reversed flow never disappears and occurs always at somewhat between 5 and 21 faces on the pressure-outlets.
The geometry was created and meshed in Gambit and during design I used the y-symmetry and one side of the model is just a copy of the other so dimensions must be exactly the same. I use 12 cells to cover the 3mm gap between box and the above coating head, do you think this could be a problem?

[-mAx-]

In Gambit as you copied the flipped domain, did you check that the 2 domains are connected (along the y-axis)?
in gambit you can check it by trying to move one domain. If you can , then the 2 domains aren't connected (else it returns you you cannot since another entity is connected)

[bobmalaria]

hi,

when i created the flipped bit in gambit i refelected just the vertexes and connected them with edges and made faces from thos edges. So those are conneted.

However, because this box is supposed to move at some point I have an interface in the model. But I defined the grid interfaces and checked the mesh in fluent. The flow across the interface also seems not to be the problem. I am really puzzled why in a fully symmetric geometry the solution is not symmetric.

Thanks again for helping, I appreciate any input

[-mAx-]

at the other hand, a flow in a symmetrical geometry isn't necessarly symmetric... (cf von Karmann at specific Re)
Is your steady state solution fully converged?
Try to monitor the massflow at your inlet, and check if you get a converged solution