CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   Free convection (http://www.cfd-online.com/Forums/main/106074-free-convection.html)

ebrattr August 17, 2012 11:29

Free convection
 
I have had a big problem trying to simulate free convection over a vertical plate. One of my problems is i want to control the pressure gradient, I mean one of the unknown variables is the pressure p. But, in free convection or at least, in my model I don't take into account the x-momentum equation. In fact, my equations are the following

\frac{\partial u}{\partial x}+\frac{\partial v}{\partial y}=0
u \frac{\partial u}{\partial x} + v\frac{\partial u}{\partial y}=\zeta \beta g(T-T_{\infty})+\nu \frac{\partial^2 u}{\partial y^2}
u\frac{\partial T}{\partial x}+ v\frac{\partial T}{\partial y} = \alpha \frac{\partial^2 T}{\partial y^2}

With this boundary conditions

u(x,0)=u_w(x)
u(x,\infty) = 0
v(x,0) = 0
T(x,0) = T_w(x)
T(x,\infty) = T_{\infty}

This is the problem,
http://s7.postimage.org/xobd7x255/image.png
Link: http://s7.postimage.org/xobd7x255/image.png

I take F_x = -\zeta\rho g and \frac{\partial p}{\partial x} = -\zeta \rho_{\infty}g

And I compute \rho as, \rho = \rho_{\infty}+\left(\frac{\partial \rho}{\partial T} \right)_P (T-T_{\infty}). Replacing, \beta=-\frac{1}{\rho} \left(\frac{\partial \rho}{\partial T}\right)_P.

I will initially define:
T_w(x), \, u_w(x).

Finally,
\zeta \in \{-1,0,1\}.

In fact, is it possible to compute this in a COMSOL? Mi unknown are only u,v and T not P. Because I assume the pressure gradient as I explained before.

leflix August 17, 2012 14:30

Quote:

Originally Posted by ebrattr (Post 377515)
I have had a big problem trying to simulate free convection over a vertical plate. One of my problems is i want to control the pressure gradient, I mean one of the unknown variables is the pressure p. But, in free convection or at least, in my model I don't take into account the x-momentum equation.

Hi ebrattr,

No but in the model you have presented here you have taken the x-momentum equation into account.
I'm afraid your model is completely rubbish.
Free convection is a kind of flow which is vertically dominated as it is induced by gravity.
So taking x momentum equation does not make sense. If you assume that your flow is unidirectional then choose rather the y momentum because the buoyancy force which drives such flow is oriented along y axis.
I even do not undertand how you can have your buoyancy term in the x momentum or I have missed something.:confused:
f= rho*g with vector g=-9.81y where here f g and y are vectors.


Quote:

In fact, is it possible to compute this in a COMSOL? Mi unknown are only u,v and T not P. Because I assume the pressure gradient as I explained before.
If COMSOL is available for you, why do you want to use a simplified model and solve it with COMSOL ? Solve the full 2D Navier-Stokes equations as COMSOL is capable to do it.
Then you could verify a posteriory if your assumptions were right or not especially concerning the pressure gradient, but I guess it won't ;)

ebrattr August 17, 2012 14:38

I turned over the axes, look at my axes. I will follow you advice. But I want to compute also a solution for those equations because I want to learn about it.

What is the simplest way to compute it ? With a good solution (exact) it doesnt matter if the convergence takes to much for the algorithm jaja

leflix August 17, 2012 20:33

Quote:

Originally Posted by ebrattr (Post 377542)
I turned over the axes, look at my axes.

Don't do such stuff when you submit a problem for which you look for a solution ....fluid mechanic is already arduous and strenuous enough that changing axes brings more difficulties than help. You confused us (at least me) for no reasons.
Better keep the usual notations well mastered by every one here


Quote:

What is the simplest way to compute it ? With a good solution (exact) it doesnt matter if the convergence takes to much for the algorithm jaja
I would say try finite difference or finite volume method
It should be easy if you can extract the v componnent from integrating the continuity equation.

ebrattr August 18, 2012 01:00

Quote:

Originally Posted by leflix (Post 377574)
I would say try finite difference or finite volume method
It should be easy if you can extract the v componnent from integrating the continuity equation.

Dont you have any reference guide ?

leflix August 18, 2012 04:28

Quote:

Originally Posted by ebrattr (Post 377587)
Dont you have any reference guide ?


You have to learn howto discetize a partial derivative equation based on finite difference, finite volume or finite element. Just choose your method.

If you want to use finite volume books of Versteek or Peric are great
http://www.amazon.com/An-Introductio.../dp/0131274988
http://www.amazon.com/Computational-...keywords=peric

for finite differences or finite element I don't have particular books in mind
but you can find many lecture notes on internet.

once you discretized your equations, in one time step solve:
your momentum equation, then use continuity equation to obtain the second velocity component, then solve heat equation finaly advance time step..

ebrattr August 20, 2012 16:05

Efficiency
 
I solved the problem in MATLAB.

Now I'm going to compare my results with COMSOL results. But I see this weird behaviour and I don't know why ?

http://s10.postimage.org/5q2xjuz2v/Temperature.png
Imagen: http://s10.postimage.org/5q2xjuz2v/Temperature.png

http://s12.postimage.org/mmjk7giq3/velocity.png
Imagen: http://s12.postimage.org/mmjk7giq3/velocity.png

Here is my COMSOL file: http://www.mediafire.com/?fsdkoqa1uumm3cc

Thanks !


All times are GMT -4. The time now is 13:01.