Ejector troubles

Questions and remarks about Code_Saturne usage
Forum rules
Please read the forum usage recommendations before posting.
Post Reply
tpa
Posts: 5
Joined: Fri Mar 16, 2012 11:35 pm

Ejector troubles

Post by tpa » Sun Apr 14, 2019 12:14 am

Hello honorable users of Code Saturne.

I am trying to model a air/air ejector at RT but I am facing instabilities.

I use tretrahedal mesh (Salomé) have one 3mm nozzle as inlet with 100 m/s, one suction inlet with 1 m/s and one outlet. All other boundaries are smooth walls.

After some iterations/time steps the system gets unstable and at some point the solutions ceases to converge. This behaviour is for steady/unsteady approach as well as for various turbulence models.

So either the model is bad. One concern I have is that the simulation needs to be done in the compressible module (which I still do not master) or something else is wrong.

The development/decline of pressure and velocity is illustrated by short video sequences to be found in this folder: https://my.pcloud.com/publink/show?code ... EiIHsxovjy (hope it works) as well as the mesh and xml file.

Any input will be appreciated. Thanks.
Attachments
setup.xml
(7.49 KiB) Downloaded 12 times

Yvan Fournier
Posts: 2625
Joined: Mon Feb 20, 2012 3:25 pm

Re: Ejector troubles

Post by Yvan Fournier » Sun Apr 14, 2019 12:41 am

Hello,

The steady algorithm is not very robust, so you may have better luck with the unsteady scheme, or illlas an intermediate solution, the pseudo-steady using a local time step.

You flow conditions could probably benefit from th compressiblt model, but I do not know it well either. I suspect your instability is related to the mesh quality, os o detemine if this is the cause or not, you could try the same mesh with a slower flow. In case of mesh robustness issues, you will find recommendations on the best practice guidelines on the website (a bit old, but mostly still relevant) and on recent posts in the forum.. Basically, you can check the CFL is not too high and adjust the time step, try the least-squares gradient reconstruction with extendedn neighborhoor, and/or test using upwind on variables (which increases solution diffusivity but also stability).

Regards,

Yvan

tpa
Posts: 5
Joined: Fri Mar 16, 2012 11:35 pm

Re: Ejector troubles

Post by tpa » Mon Apr 15, 2019 9:50 am

Thank you very much for the proposals. I have now made mesh a bit coarser and have begun the timestep/Courant factor approach. I will post back when I have some usable results. The choice of 100m/s in nozzle was to keep mach number reasonably below unity. Once I get this to work I intend to see if I can get the compressible model running for supersonic nozzle velocity. Best regards.

Post Reply