Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

Free convection and conduction heat transfer

Please login with a confirmed email address before reporting spam

Hi,

I am trying to build a model to simulate the heat transfer in a vertical pipe filled with liquid helium. Heat transfer will be due to two mechanisms, namely thermal conduction and free convection.

I have built a model with COMSOL Multiphysics 3.5a (Fluid Thermal Interaction -> Fluid-Thermal Incompressible Flow -> Steady State Analysis). I have attached the mph file.

My model consists of two subdomains, both of which represent the liquid helium volume. The two subdomains share a common boundary. One of the two subdomains has a constant heat dissipation per unit volume. The other subdomain has a boundary with a constant temperature. I expect the heat generated in the first subdomain to be conducted to the fixed-temperature boundary. What I am interested in is the temperature distribution in the two subdomains.

All is well if free convection is neglected, and only conduction is taken into account. The results make sense.

Heat transfer should be much better if free convection is included, however. I did that by applying a force of g*rho_He (T) (where g is the gravitational acceleration, and rho_He(T) the density of liquid helium as a function of temperature) to each volume unit of the helium subdomains. In addition, I defined the heat capacity of liquid helium (cp_He(T)).

The velocity field calculated with the Navier-Stokes equations (u and v) should then be used in the convection & conduction calculations, but it seems to be not, and I don't know why, at least the temperature profile is completely unaffected by changing g from 9.81 to -9.81 and 0. I did enter u and v as the x and y component of the fluid velocity in the convection & conduction subdomain setting window.

What puzzles me is that when I do the postprocessing, fairly high velocities are yielded (which is expected), but those values do not seem to be used in the convection calculations. If I enter much lower values manually (constant values instead of u and v in the subdomain setting window), the calculation works, i.e. convection occurs (or at least there is a very distinct influence on the temperature distribution in the subdomain).

Does anybody know what's wrong? Maybe I am making a very obvious mistake?

Best Regards,
Patrick


2 Replies Last Post 2 nov 2009, 12:32 GMT-5
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2 nov 2009, 02:48 GMT-5
Hi Patric

I have tried to run your case, and there is a couple of things puzzling me, first you have u,v,T and P as variables but the pressure is (and remain) at 0 or going negative. I would select a point on the model and give it an initial pressure > 0. With no pressure I do not feel it's physical, what do you think ?

Then, for me, you model does not really loop around in the solver, I need to change the settings of the solver for that, first in the General tab under Settings I use 100 as maximum number of iterations (but later I have noticed it also works with 50). And more important, in the Stationry tab I would use "automatic" and not linear, as I fear this is rather "higly" non-linear (I'm never 100% sure which of these settings are overriding the other and mostly COMSOL is rather good to adapt its solver settings).

By running like that I get a pressure and temperature values that changes over the volume, but I have problems getting it to converge, it seems rather mesh dependent too.

I'm starting to wounder if the issue is not so non-linear and turbulent that it really will not converge

Hope this helps a little on the way
Good luck
Ivar
Hi Patric I have tried to run your case, and there is a couple of things puzzling me, first you have u,v,T and P as variables but the pressure is (and remain) at 0 or going negative. I would select a point on the model and give it an initial pressure > 0. With no pressure I do not feel it's physical, what do you think ? Then, for me, you model does not really loop around in the solver, I need to change the settings of the solver for that, first in the General tab under Settings I use 100 as maximum number of iterations (but later I have noticed it also works with 50). And more important, in the Stationry tab I would use "automatic" and not linear, as I fear this is rather "higly" non-linear (I'm never 100% sure which of these settings are overriding the other and mostly COMSOL is rather good to adapt its solver settings). By running like that I get a pressure and temperature values that changes over the volume, but I have problems getting it to converge, it seems rather mesh dependent too. I'm starting to wounder if the issue is not so non-linear and turbulent that it really will not converge Hope this helps a little on the way Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2 nov 2009, 12:32 GMT-5
Ditto- I looked at the model and first I was suprised how quickly it found a "solution" but that is because of the linear setting which is incorrect. This is a highly non linear problem.

As for whether to use a laminar or turbulent model, I suggest you calculate the Grashoff number to see if you need a turbulence model. I also suggest you use weakly incompressible NS mode (not incompressible), and set the pressure at one poitn as Ivar suggests. Further also note that a steady solution might not exist and you may have to solve it as transient for a very long time to study the nature of the solution.

good luck
- Ozgur
Ditto- I looked at the model and first I was suprised how quickly it found a "solution" but that is because of the linear setting which is incorrect. This is a highly non linear problem. As for whether to use a laminar or turbulent model, I suggest you calculate the Grashoff number to see if you need a turbulence model. I also suggest you use weakly incompressible NS mode (not incompressible), and set the pressure at one poitn as Ivar suggests. Further also note that a steady solution might not exist and you may have to solve it as transient for a very long time to study the nature of the solution. good luck - Ozgur

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.