CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   Main CFD Forum (https://www.cfd-online.com/Forums/main/)
-   -   Buoyancy driven flow (https://www.cfd-online.com/Forums/main/98585-buoyancy-driven-flow.html)

 chivaee March 14, 2012 09:07

Buoyancy driven flow

Hi everybody,

I'm going to solve a simple flow over flat plate and I'm going to have heat equation and buoyancy effects in my flow. My domain size is 1500(height)x6000x6000 [m] and my first grid center is located 0.03 away from the ground. I have periodic lateral BC, Symmetry top BC and no-slip wall BC.

I can simulate the flow when I have a neutral case but with temperature it diverges very fast. Do you have any idea what could be wrong?

Note that with this size Grashof number becomes enormous, but I don't know if that's the problem.

I appreciate all your suggestions :)

 sgrshukla March 14, 2012 14:07

 mettler March 14, 2012 14:26

it sounds like a grid problem. Especially with buoyancy driven flow. Most of the interaction will be close to the wall, so you need a good grid in that region. I am assuming your plate is vertical and not horizontal.

 FMDenaro March 15, 2012 08:02

maybe, it can be a case of numerical instability due to the high Gr number...

 chivaee March 15, 2012 10:03

Quote:
 Originally Posted by sgrshukla (Post 349473) How about first trying with reducing your URF and making your solution more stable.
I see your point, but I've tried with URFs of down to 0.15 with no success ...

 chivaee March 15, 2012 10:11

Quote:
 Originally Posted by mettler (Post 349475) it sounds like a grid problem. Especially with buoyancy driven flow. Most of the interaction will be close to the wall, so you need a good grid in that region. I am assuming your plate is vertical and not horizontal.
Hmmmm it could be. I've used a simple tanh function to stretch in the wall normal direction and uniform in the other two directions ... the plate is horizontal, in fact its an attempt to model atmospheric boundary layer simulation with 1.5 km height ... my issue is that the stretching rate is more than 30% but it looks quite smooth.

Now, even the temperature equation works fine but only if I disregard the gravitation.

 chivaee March 15, 2012 10:13

I might also take a look at the block communications (it's a multi-block structured code I'm using) to see if something goes strange ...

 chivaee March 15, 2012 10:15

Quote:
 Originally Posted by FMDenaro (Post 349591) maybe, it can be a case of numerical instability due to the high Gr number...
mmm, yeah, meybe, since it scales up as L^3 (L being characteristic length scale) ... but isn't it what normally happens? I mean I'm just simulating a small part of the real atmosphere with temperature stratification ...

 chivaee March 19, 2012 11:39

problem partly solved, not sure if it's true ...

Hi again,
regarding the issue with buoyancy driven flow, I tried to decouple temperature and momentum equations completely i.e. after the solution for pressure and velocities converge, I solve the heat equation and most importantly "add the buoyancy contributions".

Seems it solved the issue ...

I know it's a general practice to treat the heat equation like this (quasi-steady) but not sure if it's true when you have buoyancy? since buoyant force affects the flow field directly and might make it inaccurate ...

 FMDenaro March 19, 2012 11:56

Quote:
 Originally Posted by chivaee (Post 350231) Hi again, regarding the issue with buoyancy driven flow, I tried to decouple temperature and momentum equations completely i.e. after the solution for pressure and velocities converge, I solve the heat equation and most importantly "add the buoyancy contributions". Seems it solved the issue ... I know it's a general practice to treat the heat equation like this (quasi-steady) but not sure if it's true when you have buoyancy? since buoyant force affects the flow field directly and might make it inaccurate ...
are you searching for the steady solution? then your method can be quite good, otherwise you need more accuracy in time, for example doing a predictor-corrector integration

 happy April 19, 2012 04:46

show me by CFX

Quote:
 Originally Posted by chivaee (Post 350231) Hi again, regarding the issue with buoyancy driven flow, I tried to decouple temperature and momentum equations completely i.e. after the solution for pressure and velocities converge, I solve the heat equation and most importantly "add the buoyancy contributions". Seems it solved the issue ... I know it's a general practice to treat the heat equation like this (quasi-steady) but not sure if it's true when you have buoyancy? since buoyant force affects the flow field directly and might make it inaccurate ...
could you please show me how you could do it by CFX:o?
My problem is the same as yours. I have fluid flow with heat transfer and the buoyancy takes place through my model.
Regards

 happy May 14, 2012 23:56

Quote:
 Originally Posted by happy (Post 355555) could you please show me how you could do it by CFX:o? My problem is the same as yours. I have fluid flow with heat transfer and the buoyancy takes place through my model. Regards
finally, I got it!!