CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   VOF necessary mesh and timestep (http://www.cfd-online.com/Forums/fluent/115895-vof-necessary-mesh-timestep.html)

Andrey.M. April 9, 2013 02:02

VOF necessary mesh and timestep
 
Hello, dear community of cfd-online!

Since VOF model is of much computational cost, there are 2 important issues:
How fine domain should be meshed and what timestep should be used.
If we have a 2D case of straight channel and to describe cross-section we use say 20 cells in the core and 6 cells per boundary layer to solve homogeneous flow. We want to solve a flow with gas and liquid phases, which volume fraction is 10%. Does it mean that for same accuracy we have to use 10x more cels to describe cross section?
If we have a comparatively large timestep which results in Courant number equal to 30 and each timestep converges, is it still a good solution? What timesteps and Courant numbers could be applied to get a relevant solution with not much time spent on calculation?

Thank you in advance for your answers!

Jabba April 9, 2013 10:02

are you trying to simulate a stratified flow in a channel?
or the gas phase is diluted in the liquid one? if you are trying to simulate small bubbles or droplets, i guess you should use other multiphase model

mesh resolution of VOF calculations should be fine enough to capture the interface between the fluids, when the length scale of the interface is much larger than mesh size (as in stratified flow)

in a 2D domain, i would use a variable time-step with courant number of 2, as fluent manual suggests

with courant number near 30, you may get qualitatively good results as well, but lower courant numbers will provide more meaningful results



Quote:

Originally Posted by Andrey.M. (Post 419287)
Hello, dear community of cfd-online!

Since VOF model is of much computational cost, there are 2 important issues:
How fine domain should be meshed and what timestep should be used.
If we have a 2D case of straight channel and to describe cross-section we use say 20 cells in the core and 6 cells per boundary layer to solve homogeneous flow. We want to solve a flow with gas and liquid phases, which volume fraction is 10%. Does it mean that for same accuracy we have to use 10x more cels to describe cross section?
If we have a comparatively large timestep which results in Courant number equal to 30 and each timestep converges, is it still a good solution? What timesteps and Courant numbers could be applied to get a relevant solution with not much time spent on calculation?

Thank you in advance for your answers!


Andrey.M. April 10, 2013 00:45

Are there any practical advices for mesh? Any criteria.
If I get solution but surface of the liquid in gas is not as clear as it should be, are there any approaches to estimate how far my solution from reliable one?

Jabba April 10, 2013 09:41

it's hard to tell a specific criteria for VOF calculations
along with the usual ones (skewness and ortogonality), one has to considerer the mesh size which obviously will dictate how thin the interface will be captured

but it will also have an impact on solution time... since you are running 2D calculations, it will be easier to evaluate the influence of mesh size
I would proceed this way: estipulate three mesh sizes and run transient simulations with variable time stepping during some time, keeping courant number equal to 2

then compare results from them: instantaneous hold-up, mean hold-up, position of liquid or gas front

take a look at this presentation for additional information
http://pt.scribd.com/doc/62308472/Be...otive-Industry

regards



Quote:

Originally Posted by Andrey.M. (Post 419536)
Are there any practical advices for mesh? Any criteria.
If I get solution but surface of the liquid in gas is not as clear as it should be, are there any approaches to estimate how far my solution from reliable one?



All times are GMT -4. The time now is 00:55.