lid driven cavity 2d with the MRT model

Palabos General
#1

Hi everybody,

I simulated the 2D cavity with the BGK and the MRT model implemented in Palabos. The BGK model works very well but the MRT model doesn’t.
The results in the beginning are as expected, but the vortex wears off and then a vortex in the other direction develops.
I implemented the MRT model with the following lines:

#define DESCRIPTOR MRTD2Q9Descriptor
typedef MRTdynamics<T,DESCRIPTOR> BackgroundDynamics;

MultiBlockLattice2D<T,Descriptor> lattice (
parameters.getNx(), parameters.getNy(),
new MRTdynamics<T,DESCRIPTOR>(parameters.getOmega()) );

my parameters are:

(T) 0.01 – uMax
(T) 200. – RE
100 – N

  1. – lx
  2. – ly

maxT was very high
Everything else is from the cavity2d case.

Did anyone encounter similar problems? Or knows if and what I did wrong.

Thank you
Konstantin

P.S.: I’m sorry for posting twice. I noticed to late that the other post is in the LBM theory part and I haven’t yet figured out how to delete the post.

#2

Hello,

the relaxation times may be not “optimal”. (At least they are not those given in Lallemand’s paper.)

Maybe you should have a look at the MRTD2Q9descriptor and tune them. In the next release this should be corrected. Nevertheless I never noticed that the MRT gave worse results than the BGK model. Normally they are more or less equivalent.

Orestis

#3

Hello,

Thank you very much for your answer.

I tried to tune the MRTD2Q9 descriptor. I changed all relaxation times
to the omega of the simulation with the same parameters (1.976284585).
To my knowledge, this should give EXACTLY the same results as the simulation with the BGK
model.

The simulation with the BGK model becomes stationary after some
thousands of iterations. However, with the MRT model (and a constant
relaxation time of 1.976284585) the vortex rises, then decreases and
re rises with a contrary sense of rotation.
The x-velocity gradient is getting huge over time.

I changed the file mrtLattices.hh as follows:
(lines 66 to 70)
template
const T MRTD2Q9DescriptorBase::S[MRTD2Q9DescriptorBase::q_] =
{ (T)1.976284585, (T)1.976284585, (T)1.976284585, (T)1.976284585,
(T)1.976284585, (T)1.976284585, (T)1.976284585, (T)1.976284585,
(T)1.976284585 };
// s7=s8 to have a shear viscosity nu
// and the bulk viscosity depends on s2.

Are these changes right or do I have to change more to change the model?

Again thanks for your help.
Konstantin

#4

Dear malaspin
I hope every thing is okay with you.
I ergently need the Lallamond paper, as I’m at home, this paper is not accessible for me.
Could you please send me this article if you have it?
Thanks in advace
H. Nemati
malaspin Wrote:

Hello,

the relaxation times may be not “optimal”. (At
least they are not those given in Lallemand’s
paper.)

Maybe you should have a look at the
MRTD2Q9descriptor and tune them. In the next
release this should be corrected. Nevertheless I
never noticed that the MRT gave worse results than
the BGK model. Normally they are more or less
equivalent.

Orestis