[Zzmon]

Hi all,

I am simulating the stator of a small axial turbine for a turbocharger. I am using CFX but, just for comparison, I tried to simulate it with Fluent: the results are very different!

- different pressure distribution
- different max Mach number
etc.

I attach some images for you, can you please explain the computational/physical reasons because this happens?

P.s. = the boundary conditions are obviously the same, the models too (in Fluent is Reynolds stress 7 eqn while in CFX is the SSG Reynolds stress).

Thank you in advance!

Mach CFX.jpg

Mach Fluent.jpg

static pressure CFX.jpg

static pressure Fluent.jpg

[MB72]

Your post is very interesting for me.
Looking at the results I have one question:
Are you sure your fluid model is the same? I've noticed in the past that CFX and FLUENT doesn't have a unified material database.
For example: Do you use the same viscosity model for CFX and FLUENT?

It is just an Idea. Please let me know

[Zzmon]

Thank you for your response MB72!

Well, I've never thought about this thing. The material is the same (Air Ideal gas), but there are very little differences in values as thermal conductivity and viscosity.

There are very slight differences, in your opinion can they justify such a variation in the results?

If yes, should I "trust" the CFX air ideal gas or the Fluent one? (very strange question, I know).

Thanks again, very much!

[MB72]

Following my little experience in internal fluid dynamics the most underrated simulation parameters are:
-Description of fluid viscosity (change with temperature and pressure).
-surface roughness expressed in "sandgrain" a strange measuring unit related to the moody diagram.
-model of fluid viscous heating.
Working on CFX with these parameters, you will experience significant difference on fluid temperature and therefore Mach Number.

Hold on. I will provide some useful information regarding viscosity and "sandgrain" definition.

Comparison between FLUENT and CFX: I use CFX (FLUENT only for confirmation). I think FLUENT is a little bit "scientific" with huge documentation. Regarding the materials database (if I remember well) has several model already implemented in comparison to CFX. However CFX is much more quick when you have to run to get a resonable result.

However you are lucky. If you use a typical wing section for compressor, you can check the results with an impressive amount of data recorded and published by researcher (aircraft/energy gas turbine field)

[Zzmon]

Yes, I am sure you are right and these are precious suggestions. For now, I am simulating with smooth wall option so no problem with sgr but I know that it's an unusual unit.

Unfortunately I am also designing the blade so I think I can't refer to literature too much. But I am always looking for research work that can help me.

Thanks

[ghorrocks]

Comparing results between CFD codes is not a simple thing.

You have to be sure the material properties are the same. But you also have to make sure you are applying the same boundary conditions in both cases. Also any physics models you have implemented. And a further consideration is that you can only compare fully converged simulations - and by converged I mean converged for mesh density, convergence criteria, time step size and any other tunable parameters.

As you can see this is not a straight forward thing and simply comparing results using default options on a guessed mesh is meaningless.

[Thomas MADELEINE]

your domains also look different:
you have a quite large space for the Inlet with CFX that you don't have in Fluent
I would recommend to enlarge your Inlet's domain in Fluent to the size of the one in CFX
and enlarge both at Outlet region to get rid of averaged value at BC.