Comments

• Victor

  July 2018

  The change is that version 8 and earlier calculated thermal stress using a single average temperature over the whole element but version 9 interpolates the temperature at each integration point which is more accurate. That means non-uniform temperature fields within elements were hidden in version 8 but are apparent in version 9.
  
  This model will give incorrect results in all versions because the midside nodes without temperatures default to the reference temperature - ie no thermal stress. To correct it, use the same tri6 elements in the thermal model as the static model.
• waluyo

  July 2018

  Thank you Victor,
  I rerun the heat transfer after changing to tri6 elements (from tri3), then creating a static model with thermal stress. Yes, it works and showing the same displacement contour as shown in previous one. Perhaps this converts to a better and more accurate result as you have explained.

  result-mecway9-tri6.png 90.3K
• Victor

  July 2018

  Though it looks similar, this new displacement is about double what it was before. The maximum is now 1.3e-8 m compared to 6.7e-9 m without the midside nodes included. That's because without temperatures on the midside nodes, half the nodes of each element were contributing a 0 to the average temperature change so the thermal strain was about half what it should have been.