AnsweredAssumed Answered

Temperature goal plot is erratic.

Question asked by Juan Castellanos-Smith on Jan 13, 2015
Latest reply on Jan 15, 2015 by Jared Conway

Like • Show 0 Likes0
Comment • 10

I am trying to see what temperature a semiconductor reaches inside of a burn-in socket. I have downward air impinging on a finned heat sink atop a burn-in socket inside which a semiconductor is generating heat. I also have horizontal air travelling in a direction that is transverse to the fins and exits through a pressure opening at atmospheric pressure. I have a persistent "vortex crosses pressure opening" warning. I have tried enlarging the domain so that the vortex moves/disappears but have had no change in the temperature plot. I have seen this warning in past simulations, but have not seen this type of curve behavior. What could be responsible for this and how can I fix it to get a smoother temperature plot?

Thanks,

Juan

No one else has this question

Outcomes

10 Replies

Amit Katz Jan 13, 2015 10:32 AM
Mark Correct
Correct Answer
Your inlet/out value on the vortex is very small, I only get concerned about those warnings if (1) the boundary is close to my area of interest and (2) the value is near 1.0

As for the erratic plot, it could be the result of a coarse mesh or large time steps which are causing some instability.
Like • Show 1 Like1
Actions
- Juan Castellanos-Smith @ Amit Katz on Jan 13, 2015 11:43 AM
  Mark Correct
  Correct Answer
  Thanks for your insight. I am seeing some improvement by specifying an initial mesh of Level 8, but I'm still seeing some jumps in the curve. I don't understand why the number of solid cells (36,027) for the 2912 product is so small compared to the solid cells (254,076) for simulation of the 2397 product (which is similar). In addition, the 2397 had a mesh of Level 4 that is more coarse than the 2912 and yet the curve is significantly smoother. Both have the same size time step. Any thoughts?
  
  2912
  
  2397
  Like • Show 0 Likes0
  Actions
  - Amit Katz @ Juan Castellanos-Smith on Jan 13, 2015 11:58 AM
    Mark Correct
    Correct Answer
    Take a look at the warnings in those plots. The software is automatically choosing a time step for you because the one you specified is too large. I'm not sure how the automatic time step is determined. In the case of the smooth plot, the time step is about 10x smaller than the coarse plot.
    Like • Show 1 Like1
    Actions
    - Juan Castellanos-Smith @ Amit Katz on Jan 13, 2015 2:39 PM
      Mark Correct
      Correct Answer
      Well, I've tried changing the manual time step to something smaller, but the calculation time becomes extremely long (in excess of 4 hours). I also suspect that the time step doesn't stay that small throughout the simulation since the finer plot was expected to finish in 2 hours. As for the solid mesh, do you know how to make the number of solid cells of the coarse plot similar to that of the finer plot?
      Like • Show 0 Likes0
      Actions
      - Amit Katz @ Juan Castellanos-Smith on Jan 13, 2015 2:44 PM
        Mark Correct
        Correct Answer
        If you want finer detail on a particular solid, I would recommend using a local initial mesh. Do not use automatic settings, I would recommend that you specifically set the refinement level for partial and solid cells in that volume until the resolution is to your liking.
        I try to avoid using automatic meshing whenever possible, because the mesh really is a huge factor in the quality of your analysis. An extra 30 minutes spent on refining your mesh could save you hours of work and rework down the line.
        Like • Show 1 Like1
        Actions
        Juan Castellanos-Smith @ Amit Katz on Jan 15, 2015 2:49 PM
        Mark Correct
        Correct Answer
        Thanks for the suggestion. I tried the manual local mesh, but I couldn't get remove the jumps in the temperature curve.
        Like • Show 0 Likes0
        Actions
      - Jared Conway @ Juan Castellanos-Smith on Jan 15, 2015 2:24 PM
        Mark Correct
        Correct Answer
        2hrs for a transient nonlinear seems not bad
        
        i'm guessing you have a setup issue and also a mesh issue
        
        would recommend setting up something with a known solution first (block with temperature applied that heats up and cools down) see how long that takes and then scale up
        
        for your model, start dead simple and add complexity as you need and get comfortable with the process
        Like • Show 1 Like1
        Actions
        Juan Castellanos-Smith @ Jared Conway on Jan 15, 2015 3:00 PM
        Mark Correct
        Correct Answer
        Yeah 2 hours is somewhat reasonable for us, but 4 hours is way too long. The thing is, even when I specify a 0.5 sec manual time step, I get a 5 hour calculation time. But as you suggest, I should reduce the complexity of the assembly. Thanks for the feedback.
        Like • Show 1 Like1
        Actions
        Jared Conway @ Juan Castellanos-Smith on Jan 15, 2015 5:29 PM
        Mark Correct
        Correct Answer
        the world is transient and nonlinear
        sometimes you have to make assumptions to help reduce the time to solve
        
        even 5 hrs seems fine for a transient simulation
        Like • Show 0 Likes0
        Actions
    - Jared Conway @ Amit Katz on Jan 15, 2015 2:23 PM
      Mark Correct
      Correct Answer
      autotime step is based on flow through the smallest cell i believe (in the swx kb if i remember correctly)
      Like • Show 0 Likes0
      Actions

Temperature goal plot is erratic.

Outcomes

Related Content

Recommended Content