explicit momentum exchange

Submitted by cupes on Wed, 04/05/2017 - 15:53

Hi there,

I am a newbie of CFDEM. I want to try simulating a newtonian fluid with very large viscosity or HB fluid. I tried to use implicit momentum exchange provided by the latest release of CFDEM. However, if the viscosity is very large, the simulation will fail even the time step is already very small. So after I read the paper Blais et al 2015 on Journal of Computational Physics, I tried to modify the PISO solver to consider the explicit momentum exchange. Now it seems that I get the code working. I modified the cfdemSolverPiso.C (add "if" to separate the implicit and explicit momentum exchange).

The problem is that the simulation of ErgunTestMPI cannot complete, the dragtota will become super large, like:

***********************************
Time = 0.007

Courant Number mean: 1148.9 max: 124662

Coupling...
Starting up LIGGGHTS
Executing command: 'run 100 '
run 100
Setting up run at Wed Apr 5 21:27:31 2017

Memory usage per processor = 11.7595 Mbytes
Step Atoms KinEng rke Volume centerOf dragtota dragtota dragtota
10600 10000 0.019058966 0.00057649428 5.076e-05 0.0082919729 -169.94831 -67.352717 2185.5666
10610 10000 695.88922 0.25643382 0.033235707 0.0094573001 -169.94831 -67.352717 2185.5666
************************************

Another embarrassing problem is, I cannot use the modified PISO solver to run ErgunTestMPI using implicit momentum exchange! The simulation was also stuck at certain time, like 0.014s.

Any advice would be very appreciated! Thanks a million!

Chao

zbtlfd | Fri, 11/17/2017 - 13:42

Hey Chao,

did you get a solution for your problem?

Thanks
Lukas

paul | Wed, 01/17/2018 - 22:34

I can imagine the very loose coupling interval being the problem here, as high viscosity would imply a very small relaxation time, meaning that forces are momentarily very large until the velocities of fluid and particle equilibrate. This scenario needs a very low coupling interval. Use the couplingInterval forceModel to check for this or look up the formula for the relaxation time in literature.

Ermek Asylbekov | Tue, 03/27/2018 - 11:17

Hi Chao,

your Courant Number is way to large (mean: 1148.9 max: 124662). For a stable Simulation it should <1 (around 0.1 or 0.2). You'll have to set yout CFD timestep way lower. A smaller coupling interval (as suggested by paul) is also usefull when stabilising your simulation.

Regards,
Ermek