Learning Forum

[ganesh2516822]

Continuation of

I'm trying to find out the thrust generated by an air bearing that I've designed. For that I need to know the pressure drop across the porous graphite. Below figure shows the inlet portion of the graphite section (porous media).

https://forum.ansys.com/uploads/560/VDGDDF4QKL2S.png

The below figure shows the outlet portion. The outlet is a thin layer of non-porous material (B). This is to mimic the load on top of the bearing (A) and B is the clearance/lift from the bearing.

https://forum.ansys.com/uploads/531/M1TAC34PMSMC.png

I need to know the pressure distribution on underside of A when 60psi compressed air is supplied at the inlet. The values that I've used is attached below

https://forum.ansys.com/uploads/305/FKDCJF3GRCQV.txt

The solution does not converge. I've also tried doing this with a pressure based solver. The solution doesn't converge or produce any satisfactory results. Is this because I'm not providing any velocity/mass flux ?

[Rob]

Pseudo transient isn't transient, it's a steady state approach that uses a time scale or pseudo time step. It confuses most people, and isn't the best introduction to the solver!
Looking at the mesh I suspect that whilst you have a decent cell quality you've not resolved the flow channel. Aim for around 10 cells across the gaps, and see how it behaves. You may also need to extend the outlet, but see how you get on with a more refined mesh first.

[ganesh2516822]

Thanks for the response, Rob
I've refined the mesh to yield the maximum amount of cell nos, per flow channel (further refinement shall make my system run out of RAM). After 200 iterations the following residuals is what I've got (took so many hours due to large no. of elements.
The continuity residuals almost reached the value of 1e50 (Usually the solver stops the iteration if there was such a big divergence, I wonder why it didn't this time). The updated mesh...

And by the way while doing the analysis this time, a warning and an info popped up in fluent console, "Info: Interface zone 14 & 15 penetrate each other" and "Warning: Failed to correct face handedness 9 out of 116 left handed faces on the sliding interface zone 5". Since it was just an info and a warning I didn't research upon it.
Regarding the pseudo transient solver, my problem is in fact a steady state one but the solver automatically picked pseudo transient for me, it it OK to switch back to steady state ?

[Rob]

Pseudo-transient is steady, and is a part of the Pressure Based Coupled Solver. It's the default solver, and is good for most applications.
Looking at the model, the cell size across the interface looks to be very different. That'll cause the overlaps due to the faceting of the curved interface so needs to be corrected.
Roughly, you'll need 2-3GB RAM per million cells, and you'll see solver speed up down to around 50-100k cells per partition. However, Student has a limit of 512k cells and 4 solver cores. If your PC only has 4 cores, try and run Fluent on 3. Do NOT rely on hyperthreading.

[ganesh2516822]

If I understand this correctly, the meshes at both the sides of the interface needs to be comparable in size, am I right ?
And Is this the curved interface that you've mentioned in your reply ?

[Rob]

Correct. The above might work if the blue zone is solid, but you'll need to resolve the curves better. Check the interface options as you can increase the tolerance.

[ganesh2516822]

Hey Rob I've tried to match the element size in each consecutive zones as such But the plot follows more or less similar trend