CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Running, Solving & CFD (http://www.cfd-online.com/Forums/openfoam-solving/)
-   -   Laminar field as initial state for turbulent two phase pipe flow (http://www.cfd-online.com/Forums/openfoam-solving/66507-laminar-field-initial-state-turbulent-two-phase-pipe-flow.html)

kjetil July 17, 2009 04:51

Laminar field as initial state for turbulent two phase pipe flow
 
2 Attachment(s)
Hello,
I am setting up a two-phase 40m pipe case, and have so far been successful using the laminar model.

I will also have to do a turbulent case, and this is where it gets interesting - say difficult. Not awfully challenging per se, having a working laminar case. So I have been trying to tune the fvSolutions and PISO settings adding correctors to pressure, and using smaller and smaller delta T's, but still - the simulation keeps exploding at about 1.01 seconds every run.

I tried to use a an exisiting field for alpha and the phase velocities Ua and Ub - using the result from laminar case @ 1.6 seconds. It seems to run nicely, but then suddenly the Courant number increases from 0.37 to 81 in two iterations, and it fails again.

Is this a good way to do it - using a laminar field as initial condition? And should it be chosen closer to 0 or could it be later? Should tune my fvSolutions differently? The laminar case is steady and stabile after about 30 seconds. I attach two photos from the laminar case - the one showing the entire pipe visualize how far the gas has reached by the 1.6 seconds.

Any suggestions to this?

mahaputra July 18, 2009 05:17

Quote:

Originally Posted by kjetil (Post 223014)
Hello,
I am setting up a two-phase 40m pipe case, and have so far been successful using the laminar model.

I will also have to do a turbulent case, and this is where it gets interesting - say difficult. Not awfully challenging per se, having a working laminar case. So I have been trying to tune the fvSolutions and PISO settings adding correctors to pressure, and using smaller and smaller delta T's, but still - the simulation keeps exploding at about 1.01 seconds every run.

I tried to use a an exisiting field for alpha and the phase velocities Ua and Ub - using the result from laminar case @ 1.6 seconds. It seems to run nicely, but then suddenly the Courant number increases from 0.37 to 81 in two iterations, and it fails again.

Is this a good way to do it - using a laminar field as initial condition? And should it be chosen closer to 0 or could it be later? Should tune my fvSolutions differently? The laminar case is steady and stabile after about 30 seconds. I attach two photos from the laminar case - the one showing the entire pipe visualize how far the gas has reached by the 1.6 seconds.

Any suggestions to this?

Hei Kjetil




maybethis thesis report useful for you.Bay, M. O. (2008). "Development of Transient One-Dimensional Solver for Severe Slugging Simulation". M.Sc. Thesis, Aalborg University Esbjerg.


Hilsen,



Nugroho Adi
Stavanger

kjetil July 21, 2009 05:26

Thanks, Nugroho Adi
that was an interesting paper for sure - and I believe it may become useful to me later. In this case though, I think my problem is quite simpler than the case in that paper - and the author doesn't seem to use an initial laminar field either :o

kjetil July 21, 2009 09:15

I have found a solution to this, and as courtesy to other users that have (or will have) a similar problem, I'll post the quick solution:

When it comes to slowly increasing Courant numbers, members on this forum have been referring to boundary conditions, and to verify that they are set properly. Though, this was not a solution to me - my case (just a pipe w/two phase flow) is also fairly simple.

Until this point I had not been able to use a turbulence model starting from T=0 without crashing in an early stage. As mentioned - the laminar worked well. However, first thing I did was to refine my mesh - so the cells are distributed more evenly.

Then - the major change I did was to alter the Divergence schemes in fvSchemes from limitedLinear or limitedLinearV to upwind. Now I am also being told that using upwind (which is a lower order scheme) initially is common. I changed back to the limitedLinear or limitedLinearV after a while (I had just commented them out when adding the upwind alternative) - the simulation is still running so I can't say 100% if it's successful - but so far it hasn't crashed again.


All times are GMT -4. The time now is 20:33.