Comments

• disla

  May 28

  There is a mixture of First and Second order elements.
  
  There is a line element in between the model too.
  You can find it with the tool Select elements by ----> and then Line 2 Line 3. Then erase it.
  
  What a cool shape.
  
• Wjasiew

  May 30

  Hi,accordingly to your suggestions I removed Line2 element (single only).Nothing has changed!Next I changed Hex8 elements to Hex20 in the sense that all 9 segment lines are identical now.As displayed in animation 8 segment lines are moving as expected but 9th segment line does not move at all!.Modified project file is attached.Any idea??.Regards

  A-S-S1Q50x9-90degP0,8.liml 277.8K
• disla

  May 30

  You still have a mixture of First and Second order elements. Merge nodes after homogenization. Then it will work.
  
  Your model looks difficult to me. Your spring is thin and has significant presence of Shear Mode III . . I would try to do some validation first to find which could be the best element to capture that and how much refinement do I need for the area.
  
  
  
  
• cwharpe

  May 30 edited May 30

  We used to call these things "flexures". Your *8.liml file is (still) a mix of linear with 2nd order elements. Does it not solve and displace correctly if all elements are made 2nd order?
  
  Going the other way, something weird happens if you try and convert all elements to linear. Not only the "hex8...deprecated" warning, but the results differ from full 2nd order case by 1/100.
  
  EDIT: When the model was flattened to shells and cleaned up, linear analysis behaved as expected and took just a couple extra resolution increases to equate with quadratic. Extruding the shells back into hexes, still got the "hex8" error. Dunno. Maybe the linear wedges are causing that??
• Wjasiew

  May 30

  Hi Everybody,I owe you some explanation.I am so resistant against linear to quadratic elements replacement because I'm using free Mecway version and I was trying to overcome the node limit I'm touching with these models.
  I was surprised that the 9-line model was partially (9th line) out of order while the 8-line model is simulated quite correctly (both animations included). The structures of both models are identical.Is the mixture of hex8 and hex20 really the reason of this selective fault?It is rather hard to believe for me.
  Thank you all for your kind activity.
  
  
  
  
• Victor

  May 30

  @cwharpe Mecway has a new formulation for hex8 (incompatible modes). The old formulation (standard) is over-stiff in bending so it's hopeless for this type of problem. The new one not as bad still somewhat over-stiff for high aspect ratios like this. The warning appears if you set it to the old formulation (Analysis settings -> Incompatible hex8 elements turned off).
  
  @Wjasiew One half of the 9th line is only hex8 so it's relatively rigid compared to all the others. You should use shells and/or diagonal symmetry to keep under the 1000 node limit.
  
• cwharpe

  May 31 edited May 31

  @Victor: Aaah. I noted the poor results in comparison and see why the warning was added. Thanks.
  
  @Wjasiew: [Hex8 with Hex20] It's generally advised to avoid discontinuities in a mesh (try View Cracks). Been where you are with the node limits. Makes it really tough to do a convergence study on your solution. A good number of us who like Mecway's user interface have found the licensed version a fantastic purchase value.