CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   FLUENT (https://www.cfd-online.com/Forums/fluent/)
-   -   Natural horizontal convection (https://www.cfd-online.com/Forums/fluent/69448-natural-horizontal-convection.html)

 aneres85 October 23, 2009 05:23

Natural horizontal convection

Hello everyone. I'm trying to model a 2D case of natural convection in a rectangular box, where the bottom side is divided into two part, on the left side the heat flux is imposed while the right side is at a certain temperature. I use the boussinesq model and the fluid is liquid-water. I use the k-epsilon model for turbulence, but this one doesn't ensure convergence. Can you help me?

Thank you!

 avuttide October 28, 2009 09:41

re:natural convection

hi,

First of all try to think if the problem is laminar or turbulent, that means try to calculate the Grashof Number of the flat plate and check if the value is in the turbulent or laminar range.

If it is turbulent, I suggest that "RNG K-E" is much more helpful than "standard K-E", due to the possibility of mark the full buoyancy effect, that normally appears when you switch on the boussinesq model.

hope it helps

 aneres85 October 29, 2009 09:48

Thank you for your answer. My problem is turbulent: this is one of the first thing that I have verified.

I'using a fine mesh in the boundary layer and a coarse one in the central zone of the box.
My problem, now, is to understand what is the best way to consider turbulence in this case of natural convection. I'm using the "RNG k-epsilon" ma results are not of physical interest. I think that the problem is in the setting of some variables. Can you help me about this problem?
Thank you!

Run simulation for sometime to get an approximate value for turbulence parameters.
Initialize the parameters accordingly which will ensure convergence in my opinion.
PLS Report about result if it works (for me it worked some time back)

 avuttide November 2, 2009 13:02

re

hi

can I have some further information??
first of all are you simulating transient or steady state??
If it is a transient, is the time step size short enough??
have you check the report fluxes at the end of your simulation to see if
the balance of mass and energy ar ok??
what is really going wrong in your simulation??

 aneres85 November 5, 2009 06:37

It's a steady-state turbulent analysis. The balance of mass is ok because the the box is closed, while the balance of flux is not ok.
The residual for the continuity equation has a strange trend, because firstly it decreases, then it increases to a big value, although the mesh grid is very fine. This fact can't ensure the convergence of the problem.
I think that this problem is due to the under-relaxation factors. At this moment, I'm trying to find the better factor in order to have a good convergence in the residuals and the balance of the fluxes. Can you suggest me a solution for this problem?

Thank u at all!

 avuttide November 5, 2009 09:42

re

Hi

so I have simulated first an Un-steady state, this is also suggest by the manual, then I have used the steady-state as a benchmark.
I have follow what is written in the manual about the under relax factor,
for the discretization of the eqq. I have used both Presto! and BFW for the pressure, and all the as 2nd order.

What you have to check is the energy equation, the buoyancy takes a long long time to develop in the steady state, probably because the terms of the velocity are much more stronger then term of the buoyancy, so u have to let the natural convection develop. But this is a personal idea, not a scientific suggest!!
I think that the correct solution is the unsteady simulation, so try to
see first the results of a transient simulation

hope it helps

 lyes43 February 28, 2011 08:17

how can i set grshof number in Fluent ? what should I do to have a Gr+ 2.5 E+7 if I have ro =1 viscosity =1/Re velocity inlet =1 d=1 Re=5000 ??

 All times are GMT -4. The time now is 14:06.