SimScale CAE Forum

CFD-Diffuser Simulation errors


#1

Hi everyone

I have to run a CFD simulation of a diffuser that will be used in a Formula student car.
However, I was advised to begin with using k-omega or k-epislon models and then use LES model for more precision.
I always seem to be getting errors with I use k-omega in steady state (I am required to use k-omega in transient as well!).

My next question would be - how do I go about setting my simulation for transient models?

Any advice or help would be great.


#3

Hi @hmattoo, is this the project?
https://www.simscale.com/projects/hmattoo/diff_smaller/

Kind regards,
Darren


#4

Hi Darren

Its these 2, essentially they both are similar with minor difference but it would be great to know how to set the simulation parameters up.

https://www.simscale.com/projects/hmattoo/diff_throat_3ch/
https://www.simscale.com/projects/hmattoo/diif_throat/

Looking forward to hearing from you.
Warm regards

H. Mattoo


#5

Hi H. Mattoo,

So a few thinks to think about here:

  1. The mesh size is large, biggest at 16 million. For a steady state simulation this is definitely do-able, however starting small is always best to get the correct setup and then improve the mesh with a setup you know works. If you are however moving up to a LES then I would say restrict yourself to maybe 2 million cells and aim for less.

  2. Running more processors with large mesh sizes, at 16 million cells I would be surprised if you didn’t get an error unless working on 16 or 32 cores because the amount of ram allocated to your run is also scales with core usage.

  3. boundary cells, need to be monitored for y+ distance to ensure the wall model is modelling correctly.

I also wondered about the bounding box size I would have personally made it larger but pressumed there must be some reason as to why it is that size. Otherwise your setup seems sound. If you are still having troubles I would be more than happy to help.

Kind regards,
Darren


#6

Hi Darren

Thanks for the info.
Could you tell me how can I get more information with regards to point 1 and 3. As I usually use ‘wind tunnel’ mode for meshing and I don’t see any option apart from meshing quality to control the meshing size.
Is there any other mode that enables me to control the meshing quality?

How much larger do you suggest the tunnel box to be?

Thank you very much!


#7

Hi H. Mattoo,

So for the more advance options you will need to use parametric meshing, this will give you the power to precisely control refinements. For the y+ value you can get an approximation on the boundary size (in parametric meshing turn relative layer sizing off) and then when you run your first run with Y+ result control enable you will get a more accurate idea of what your wall Y value should be.

in terms of bounding box size, it needs to be large enough as to not effect your results, I think the rule of thumb was approximately 5 times the largest length in each direction, except the streamwise which needs to be even longer. Expect of course the lower boundary representing the road.

Kind regards,
Darren