[enayath]

Hello,

I am simulating natural convection in a closed domain. My BC is 370K and 280 K. I run the case in steady laminar mood. when I check the temperature distribution inside the fluid domain, it is 450K!!!
I set the initial temperature at 325K.I am pretty sure with the quality of my mesh...there si no source term

Does anybody know how come I have such a this problem?

Thank you in advance!

[GM_XIII]

It may happen that temperature goes up the BC value at the early stage of the iteration, try to limit it to the BC value and see if the number of limited cells drops down while the case is converging

[enayath]

GM_XIII,

I run the case with second order for Momentum and Energy...when I switch to first order, every thing looks alright...

In second order simulation, i run the case for 200,000 iterations but there were fluctuations.

Hooman

[GM_XIII]

You mean fluctuations in residuals or in max temperature?

[enayath]

In max temperature

[kad]

Fluctuations are characteristc for 2nd order discretization methods when gradients aren't resolved good enough by the grid. Maybe you can try to resolve your grid finer especially where large gradients occur in the flow and temperature field.

[GM_XIII]

Originally Posted by enayath

In max temperature

Which model are you using for air? Have you tried underrelaxing the solution of the energy equation or density? Also check what kad said, recently i read in a topic here to use adaptative grid adaptation based on temperature gradient, but be careful not to adapt too many cells in one time as it can unstabilize your run.

[enayath]

David,

I am sure about the quality of grid...I created several cases with different mesh....all resulted in fluctuations...

I use steady laminar model with water...
I played with URF...it was helpful regards to fluctuations but caused a new problem...the residual of continuity went up to 10e3!!

When I run the case with second order transientو every thing is alright


Hooman

[kad]

Do you use double precision solver?

[enayath]

yes I do.
yes I do.
yes I do.

[CFD-fellow]

Using second order discretization almost cause convergency problem, which one of its solutions is to switch to second order after complete convergence of first order. But I suppose you have done this.
High order schemes better capture physical changes. So if your physic is inherently time dependent and you are modeling it as steady state, its easier to converge with first order. According to your information that you said with transient run is ok, I think your convergency problem with your second order discretization is because of your physic.