1. Limited time only! Sign up for a free 30min personal tutor trial with Chegg Tutors
    Dismiss Notice
Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Transient Thermal Analysis in Ansys (APDL)

  1. Jun 19, 2011 #1
    Hi,

    I have to model a transient thermal problem in Ansys. I just need an outline of how I should proceed.

    The Problem:
    I am trying to model the cooling of a solar collector reciever tube from 150 degree celsius to ambient.

    My attempts have involved the following:
    1. Initial condition of 423 at the inner dia of the circle
    2. Convection Boundary condition (film coefficient of 25 and bulk temp of 298) at the outer dia of the circle

    It is not displaying the results properly (or I am reading them wrong). It shows the entire system at 298 apparently all the time.

    Are these conditions sufficient to constitute a solution in Ansys? What else am i doing wrong or I need to do?

    Thanks in advance.
     
  2. jcsd
  3. Jun 21, 2011 #2
    Not sure about ANSYS specifically, but are you sure the results are unrealistic? Is the time between steps smaller than your estimated time to reach equilibrium?

    Have you specified heat capacity of the tube? With that zero it'll probably reach equilibrium instantly.

    Try putting a different bulk temperature somewhere and see that it produces a temperature gradient between then. That should confirm that it's at least doing something! And suggest that it's cooled too quickly.
     
  4. Jun 21, 2011 #3

    Mech_Engineer

    User Avatar
    Science Advisor
    Gold Member

    Did you set an initial condition for the starting temperature of the parts? Default starting temperature is 298 or so, this might be your problem...
     
Know someone interested in this topic? Share this thread via Reddit, Google+, Twitter, or Facebook




Similar Discussions: Transient Thermal Analysis in Ansys (APDL)
  1. Ansys coupled analysis (Replies: 3)

Loading...