Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

[Turbulent flow + Heat transfer] Impacting Jet in Low-Re Comsol 4.1 model

Please login with a confirmed email address before reporting spam

Hello all !


I am trying to validate my model by comparing a simple axisymetric 2D impacting flow of air on a surface.
The geometry is simple: a cylindrical (diameter d=26.5e-3[m]) jet of air (debit Speed Ud=13.46[m/s]) is impacting a surface at distance H=0.2833[m]. Flow temp is Tj=20°C Wall temp is Tp=30°C

The input flow profile is either
- a uniform flow with speed Ud and standard length scales It=0.05 and Lt=0.01
- an established flow profile Ue, ke, epe
- a non-established flow profile Ueb, keb, epeb

The model is very capricious, I'm on it since 2 monthes and it seems to converge quite rarely...

1) First, I tried to get the turbulent low-Re flow physic to work. The first file shows my first approach, with a thin wall at the side of the jet. In that case, the model converge to a solution where the flow in the dead zone is homogenous, and the shape of the turbulent viscosity seems correct. But if I try to refine the mesh, then it doesn't converge anymore, and the 'thin wall' solution is not really practical either.

2) The second file shows a model which converges to a solution where on can see a (toric) rotation structure in the flow of the dead zone. As a result, the shape of the turbulent viscosity looks quite bad. Here the mesh is quite dense, and the model does not converge anymore if I try to enlarge it in the zones that aren't that interesting. You will see a lot of unused meshes and functions that come from many tries I made to make this damn *thing* converge. Many times, the rotation structure goes mad on itself and creates a wirlpool with growing turbulent energy that makes the model diverge.

3) Based on that 2nd model, I tried a multiphysics model adding a heat transfer physics (ht+spf) or using non-isothermal Low-Re turbulent flow (nitf). Whatever I try (mesh, BC, etc.), there is no convergence.

So could anyone help me to understand:
- what makes my first model converge only with large mesh and diverge if I refine the mesh?
- where do those rotational structures come from? Are they realistic physics, or artefacts? Maybe the incoming flow from the outputs BC adds turbulent energy to the system? But changing BC outputs to open frontiers and specifying k=0 and ep=0 does not seem to help.
- Is there anything wrong in my model (in Discretization or Stabilization prameters for instance)
- And finally, are my solver configuration parameters correct?

Thanks a lot in advance!


(All parameters are taken from: "Etude comparative de modeles à bas nombre de Reynolds dans la prédiction d'un écoulement à point de stagnation" (in French) R. Hadef, B. Leduc. 5, 2002, Int. Comm. Heat Mass Transfer, Vol. 29, pp. 683-695.)


3 Replies Last Post 2012/11/14 1:25 GMT-5
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2011/11/28 2:27 GMT-5
Hi

I'm no CFD expert, but in general, when you get a solution, i.e. as your case 1, you should analyse the gradients and ensure that you have enough elements to correctly resolve these (for me this is along the jet lines and along the top surface hit), while to the right lower part, not much happens so there the mesh could be kept coarser.

Then with turbulence, the mesh ratio to turbulent phenomena must be respected, else your meshing will be the main result driver, and a solution should be mesh independent to be used for any conclusive analysis

Adding HT does not make life easier, as you must also check the diffusive and transport thermal behaviour and gradients, and their sizes w.r.t. the mesh

--
Good luck
Ivar
Hi I'm no CFD expert, but in general, when you get a solution, i.e. as your case 1, you should analyse the gradients and ensure that you have enough elements to correctly resolve these (for me this is along the jet lines and along the top surface hit), while to the right lower part, not much happens so there the mesh could be kept coarser. Then with turbulence, the mesh ratio to turbulent phenomena must be respected, else your meshing will be the main result driver, and a solution should be mesh independent to be used for any conclusive analysis Adding HT does not make life easier, as you must also check the diffusive and transport thermal behaviour and gradients, and their sizes w.r.t. the mesh -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2012/11/13 22:34 GMT-5
Hi Ivar, what do you mean by "Analyze the gradients" and how to analyze that?

--
Doing what you like is freedom, Liking what you do is happiness
Hi Ivar, what do you mean by "Analyze the gradients" and how to analyze that? -- Doing what you like is freedom, Liking what you do is happiness

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2012/11/14 1:25 GMT-5
Hi

COMSOL gives you the gradients of your scalar fields as variables with the suffix x,y,z or r such as Tx = dT/dx, Ty=dT/dy etc so you can plot these. Or just look at the slope of your solutions / initial conditions, as the gradient is an expression of the local slope angle of your functions.

Then be sure your meshing and way you interpret these values are correctly linked

--
Good luck
Ivar
Hi COMSOL gives you the gradients of your scalar fields as variables with the suffix x,y,z or r such as Tx = dT/dx, Ty=dT/dy etc so you can plot these. Or just look at the slope of your solutions / initial conditions, as the gradient is an expression of the local slope angle of your functions. Then be sure your meshing and way you interpret these values are correctly linked -- Good luck Ivar

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.