Good Day all,
Has anyone run into the following message while running an analysis on an assembly with one component containing hyperelastic properties?
*** Decomp. Singular Stiffness Matrix. Recovering...
Thanks for any ensight you can provide.
Don Anderson
Begin Displacement and Stress Calculation
Tue Oct 29, 2013 08:48:18
Begin Reaction Calculation
Tue Oct 29, 2013 08:48:20
Begin Time Step 9 of 15: 6.00000e-01
Tue Oct 29, 2013 08:48:20
*** Decomp. Singular Stiffness Matrix. Recovering...
Iteration Residual norm
------------- -------------
1 1 Tue Oct 29, 2013 08:48:40
Load Factor: 0.566667
*** Decomp. Singular Stiffness Matrix. Recovering...
Iteration Residual norm
------------- -------------
1 0.0560233 Tue Oct 29, 2013 08:49:04
Load Factor: 0.55
*** Decomp. Singular Stiffness Matrix. Recovering...
Iteration Residual norm
------------- -------------
1 0.0480005 Tue Oct 29, 2013 08:49:27
Load Factor: 0.541667
*** Decomp. Singular Stiffness Matrix. Recovering...
Iteration Residual norm
------------- -------------
1 0.0469847 Tue Oct 29, 2013 08:49:49
Solved! Go to Solution.
I haven't seen it in Creo / ProM, but in general FEA, I believe a singular stiffness matrix occurs when either an element becomes inverted ("negative volume"), or there is rigid body motion (insufficient constraints). Assuming it runs ok with a linear material (constraints ok), try larger minimum element angles for better quality elements; the default of 5 deg can be upped to 15-20 deg and often 25 deg, depending on the geometry. The cost is slower meshing and many more elements. I think hyperelastic models with strain softening make elements more likely to invert.
I haven't seen it in Creo / ProM, but in general FEA, I believe a singular stiffness matrix occurs when either an element becomes inverted ("negative volume"), or there is rigid body motion (insufficient constraints). Assuming it runs ok with a linear material (constraints ok), try larger minimum element angles for better quality elements; the default of 5 deg can be upped to 15-20 deg and often 25 deg, depending on the geometry. The cost is slower meshing and many more elements. I think hyperelastic models with strain softening make elements more likely to invert.
David,
By using a larger minimum element angle setting I was able to fix the issue I was having.
The issue was related to a very thin piece of material being streched..
Local autogem settings can also be used to improve the mesh in the local area of the thin member.
Thanks for the insight David and Sylvain,
Don Anderson
Great to hear it worked!
Here is my experience with Singular Stiffness Matrix :
http://communities.ptc.com/message/205627#205627
As you will see, I solved it by reducing the mesh element's size and adding a displacement constraint.