Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

COMSOL - convection in a room

  1. Nov 11, 2009 #1
    thanks in advance for any help you can throw my way!

    I am in my masters year of an Acoustical Engineering course in the UK. We have been set a group design project to with essentially setting up a forced convection cycle in a room in order to redistribute the heat from a radiator and make it more efficient. To model this situation we are (trying) to use COMSOL multiphysics. At the moment we are trying to do it in 2D, with the aim of getting a 3D model in the future.

    We have never used COMSOL before, or done anything to do with heat transfer so we're kinda stuck!

    The model is very simple; a side view of a rectangular room, with a radiator on one side. We are having trouble making COMSOL create a convection current in the room.

    Does anyone have any models that are similar that they could send me to have a look at please?

    Secondly, does anyone know of any useful reading (textbooks, websites etc) that would cover this sort of situation? Most of the books in our library detail convection onto a wall etc but not within a room or cavity...

    Thank you very much for any help!
  2. jcsd
  3. Nov 11, 2009 #2


    User Avatar
    Science Advisor

    I'm not sure what type of analysis you're doing, but this seems like a fluid dynamic problem. If you're studying forced convection, you'll do the following.

    Set up your 2D room, with whatever obstacles, boundaries, etc. Place a heat generation source. Then create your "convection". You can place an inflow boundary somewhere and call it a "fan". Give it a pressure or velocity boundary condition.

    Before trying to guess, tell us exactly how you're proceeding with your analysis.
  4. Nov 12, 2009 #3
    Hi, thanks for the reply.

    At the moment we are trying to create natural convection. We are using the general heat transfer and weakly compressible navier stoke modules.
    In comsol we have a rectangle of air (subdomain settings loaded from materials library).
    The boundary settings for navier stokes are all set to wall/no slip. In general heat transfer the walls are set to insulation/symmetry with the exception of one section which is temperature and that is set to 20 above room temperature.

    We found in other forums that because we have a closed system the reference pressure is undefined so we have set the pressure at a point in the top corner.

    When we try and run a transient solution it does not converge, but it does calculate for t=0.
    If we run a stationary solution the answer returned is the one expected - constant temperature across the subdomain.

    The error most likely lies with the volume force Fy in the navier stokes subdomain settings. We have found in the comsol tutorials several different methods for modelling this as the density changes, and we are unsure as to which one is the most appropriate.

    I hope this information is useful, thanks very much

  5. Nov 12, 2009 #4


    User Avatar
    Science Advisor

    One problem is that your boundary conditions are not right. If you have all insulated walls and impose a temperature, what do you expect your final solution to look like? Well it's going to be all the temp that you defined.

    A better way to do this is to give you walls a heat transfer boundary condition if possible. Assume an outside ambient temperature and give it a small convection coefficient. You absolutely need a way to remove heat from the system.

    Then, rather than imposing a temperature, give it heat generation. This is a better posed problem that what you were giving it before.

    However, note that natural convection is a very small thing to see. You will not see huge convection currents or anything like that. The solution should be a pretty gradual decline in temperature from the source to the boundaries. I do however think you'll need to include bouyant terms to get any effect though.
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook