Dynamic Lagrangian - wiggles on unstructured grid
 

Dear CFDOnline / OpenFoam Community,

right now I am trying to setup a Dynamic Lagrangian LES Simulation, but I have big issues getting rid of wiggles during the calculations. It seems that it is a time step problem, since those wiggles should occur when high and low wave number transport deviates. But even decreasing to very short timesteps does not help. The domain is a open channel flow, with a hybrid grid (close to the wall structured and then unstructured). This doesn't seem to be the problem since close to the wall everything is fine. Right now I am trying to use small aspect ratios, to make sure that the calculated filter width isn't the problem. What is your experience with aspect ratios? Any further suggestions? The creation of wiggles seems to occur at certain regions, but I can not identify any grid problems (nevertheless I tried different algorithms and smoothing steps)


Thank You,
Soeren


ChannelFoam (2.11)
Courant Number: < 0.1
Max Aspect: 150
Re: 10^3

Hi Soeren,

Which filter do you use?
If it is the anisotropicFilter, have a look at the sources of OF 2.2.x cause in the 2.1.x there is a bug in the computation of the anisotropic coefficient. You can simply copy the sources of 2.2.x to the 2.1.1 installation and recompile (that part of) OF, no need to do a full update to version 2.2.x.

Similar for the dynLagrangian model, there is a bug in the computation of nuSgs. This is also corrected in 2.2.x. I don't know how backward compatible the turbulence models are, but this is a fix you can easily apply yourself without having to copy the whole file.

I'm not claiming this will be the solution of your problem, but you might want to start from corrected sources :-).

Cheers,

Lieven

Hi Lieven,

thanks for your fast reply. I am already preparing a new solver for OpenFoam 2.2, but first I wanted to resolve the problems with wiggles.

I think the errors in the Sgs model are minor, when I remember it correctly it was just some small factor. If there are no further suggestions I will definitely go for the new version.

cheers,

Soeren

Dear Soeren,

Why would you be interested in using a rather advanced Sgs model if you do not care whether it works properly or not?
The 'small factor' as you refer to it makes a difference of > 40% in the prediction of nut. So to my humble opinion, this is quite a significant error.
Since wiggles can be suppressed by viscosity, this can already help you in the good direction. Also, check the mesh quality, make sure there are no abrupt size changes (< 1.05) from one cell to an other.

Kind regards,

Lieven

You're right I'll start transferring... (I did not expect that the error is that large)...

If anyone else has suggestions, please share the information.

Ok, it seems that something in the channelFoam solver was the Problem. With pisoFoam and constant pressure gradient everything works perfectly.

Dear all,

I would like to use dynLagrangian with pisoFoam.

I just don't know how to set the initial condition and boundary condition for flm and fmm.

In my case study, I have a turbulent inlet, an outlet, a symmetry plane and some walls.

Regards,

vut