CGS Unit Coefficients

Submitted by NTT1508 on Wed, 11/09/2016 - 09:09

Hi Joself and Alice,

I believe I have gone through this forum carefully before asking this question. My modeling is in CGS unit. The major problem is now the CGS unit in the controldict file. I found in CFDEM/.../etc the file controldict_cgs_3.0.x but in fact this file does not contain the CGSCoeffs. Only SI and USCSC are seen here. Can you clarify whether this file will work if I copy it and replace the one in OpenFOAM/.../etc ? And can you provide the CGSCoeffs ? I dont think the controldict will work for CGS case without adding CGSCoeffts on it.

Thank you,

Regards,

Forums:

NTT1508 | Wed, 11/09/2016 - 23:04

Hi Josef,

This is great, thank you. I have some more questions: (1) Do I need to modify the controldict file in OpenFOAM/.../etc OR just use the controldict file in the CFD/system of CFDEM ? (2) I found two controldict files; they are controlDict and controlDict.foam in the CFD/system. What are they different ?

Many thanks,

Regards,

alice's picture

alice | Thu, 11/10/2016 - 07:54

Good morning,
I would suggest doing all the changes locally in your case forder, just as in the case suggested by Josef. Basically there is only one controlDict (the one without extension). The file controlDict.foam is only used for loading the case in paraview . The command paraFoam is not necessary when it exists. Usually it is created at some point during the case setup by copying the original controlDict, and it does not really matter, what it contains. (In many cases it is just an old version of the controlDict)
Best regards,
Alice

NTT1508 | Tue, 01/10/2017 - 09:29

Dear Josef and Alice,

I have spent a time to play with CGS Unit using the ErgunTestMPI-cgs tutorial. I noted the original rho is very small here (0.01 g/cm3, possibly for air). To simulate an interaction between fluid-particle, I increase rho to 1.0 g/cm3 (a normal density of water) while nu is 1.5 cm2/s. However the simulation cannot implemented well. The Courant number is increasing to very high value (10 to some thousands) and the simulation cannot go further. I try to reduce input velocity to a very small value, e.g., 0.1 to 0.01 cm/s, but particles are still moving harsh. This is not understandable because when decreasing fluid velocity, the simulation should be more stable. Can you advise what I have made a mistake here ? What else should I look into ?

Thank you,

Regards,

Nathan,

j-kerbl's picture

j-kerbl | Tue, 01/10/2017 - 14:53

Hi Nathan,

with higher fluid densities, the drag forces will be higher, therefore the particles accelerate faster. You need to decrease the timestep until you can resolve the particle relaxation time. This is the time a particle needs to speed up from 0 to fluid flow. Otherwise this is strongly overshot and you end up with an unstable result.
Also please try to keep the Courant Nr down, a reduced time-step will help here as well.

Regards,
Josef

NTT1508 | Wed, 01/11/2017 - 01:52

Thank Josef,

Yes, reducing time step made the results much more acceptable. I am comparing SI and CGS for the same case, will report result soon.

In addition, do you know how to solve the error: Atom sorting has bin size = 0.0. I followed the instruction by adding a command to atom_modify: sort Nfreq binsize. It helped the computation pass with no more error, but the result is very strange.

Regards,

NTT1508 | Fri, 01/13/2017 - 07:04

I have fully investigated the results obtained from CFDEM using different units CGS and SI. I set few particles (10) under a flow which is actually modified from ErgunMPITest. There no significant difference between these two cases. However when one increases the size of simulation, for example 10000 particles, the difference becomes more apparent (but still acceptable for me).

Regards,