CFD Online Discussion Forums

CFD Online Discussion Forums (
-   FLUENT (
-   -   buoyancy driven flow + unsteady (

co2 May 3, 2004 15:53

buoyancy driven flow + unsteady
hi folks,

I request some help from you in understanding the rules given on fluent online help for calculation of time step for buoyancy driven flows.

11.5.5 Solution Strategies for Buoyancy-Driven Flows tells us the formula for calculation of time step size for unsteady buoyancy driven flow. For my case the time step size comes out to be 0.74 sec! -- Now the problem is I need to simulate flow over 30 days and I certainly need a bigger time step size -- CAn you help me in understanding how I can increase time step size and still get convergence in each time step?

thanks,. CO2

Evan Rosenbaum May 4, 2004 12:10

Re: buoyancy driven flow + unsteady
The formula in the online help is a guideline, not a requirement. Transient evaluations have a maximum number of iterations per time step. If your time steps are converging before maxing out the iterations, you're *probably* OK. We often start with small time steps at first, then increase them gradually as the number of iterations per time stepstarts to drop.

co2 May 4, 2004 23:40

Re: buoyancy driven flow + unsteady
Dear Evan Rosenbaum:

Thanks a lot for your reply.

I am finding that with 1 sec time step size, my solution is converging in 3-4 iterations -- Thus I believe I should be increasing the time step... Thanks for that tip!

I also have made an interesting observation, but I have not been able to understand why it happens -- I was unable to get my unsteady model to converge if I just start the model as unsteady. Instead if I run a steady model first with proper BC's at t=0 conditions, then I get good convergence for unsteady iterations later .. I would imagine, if I use very small time step size of say 0.01 sec to start with, even with wrong initialization I should be able to get my unsteady model to converge without running steady state first .. can you suggest me some ways of achieving that ? -- I was just curious.

I have one more question -- For buoyancy driven flows, what type of density formulation works for you ? Is idea gas ok although my flow is incompresible.. Or do I need to use boussenisq? Or density as function of temperature ?

Have you tried buoyancy driven flows with species transport? -- I am unable to get convergence for that case -- I wonder why ! -- especially when I get convergence for just air, why would things mess up when I start using mixture template (species transport with mixture of co2 + o2 + n2)

looking forward to learning more about dealing with buoyancy driven flows using fluent -- I bet it is tough to run long unsteady simulations for buoyancy driven flows ... You just need a lot of experience -- Is it fair to say that fluent is not capable of handling buoyancy driven flows well? -- I have been struggling a lot getting convergence .. Please share any of your experiences or special tips.

again thank you very much and looking forward to posts by experienced people!

best regards, co2

Evan Rosenbaum May 5, 2004 09:11

Re: buoyancy driven flow + unsteady
1. Increase your step size. 2. The steady-state initial condition gives you well developed flow and temperature gradients. This is a stable starting point. Starting from an initialized solution should work as well, but your time step might have to be really small. You'll probably also have to reduce the underrelaxation on momentum as well. 3. All kind of density formulations will work. We have used ideal gas, Boussinesq, and density versus temperature. 4. No species transport here. 5. Most codes struggle with buoyancy flows. The driving forces are so small that the even small errors can significantly affect the numeric solution. We have had lots of difficulties when trying to do buoyancy driven systems with both liquid and gas circulations, some of which never worked and had to be abandoned.

co2 May 6, 2004 11:37

Re: buoyancy driven flow + unsteady
Many thanks for your post. That certainly helps.

One of my concerns is meshing -- I am kind of sure that my mesh is not the best and that is part of the problem (perhaps a big part! )

see, the top part of my 2D axisymmetric model is conical frustum like (there is a vent at the top which is pressure outlet) -- so you can imagine it is hard to fit a quad mesh there. So I use pave scheme there -- Any thought on a better meshing style there?

evan, what under relax would you suggest for pressure and momentum ? i have heard that they need to add up to 1.

All times are GMT -4. The time now is 10:16.