Hi

I am running a simulation of a 3d diffuser of a 4 degree angle.

The geometry has no sharp angles, the inlet BC is a velocity profile with suitable turbulence parameters. The outlet is area averaged static pressure of 0 relative to atmospheric pressure

I have run this with both SST and Reynolds stress models and i allways get flow separation when it should be attached.

Is thee any kind of setting in CFX (i am fairly new) that i may have missed

(have also tried total pressure at inlet with mass flow boundary at outlet and still separation)

Thanks bob

Hi,

How far are you modelling upstream and downstream of the diffuser? What Reynolds number and Mach number?

Glenn Horrocks

It's just the diffuser, no upstream or downstream modeling. The inlet initially has a very shallow expansion angle so is practically a pipe for the first 1/5th of the length then blends smoothly towards the outlet.

Re= 30,000 M=0.1

bob

Hi,

You should model further upstream and downstream. If you don't know what is upstream make sure your inlet velocity and turbulence profile is correct. You will definitely need a domain or a pipe or something for the flow to go into downstream of the diffuser.

Glenn Horrocks

My experience with this kind of calculation is that the inlet b.c. has to be specified very carefully. If the flow isn't fully developped there is a probably a radial component to the inlet velocity. This would bring energy towards the boundary layer and give it robustness.

If you want to make a quick check, run the k-epsilon turbulence model with wall functions. In this case the wall function is likely to inaccurately rise the near-wall velocity and the flow will then stay attached longer. If it does so, it means that the k-epsilon sort of "hides" the fact that your inlet b.c. is not well defined and that you should have a radial velocity towards the wall in the inlet plane.

Then you'll know that you have to model further upstream, as Glenn pointed out.

Regards,

Felix