Onda
Marine/Ocean
- Jun 28, 2008
- 189
Running a 80k shell elements model in MSC Nastran 2007 on Windows X64 with SOL101 I've noticed the following problem:
I've first attempted to run the model with new shell formulation (CQUADR) using the Nastran system (370)=5 and Nastran didn't converge for an negative diagonal ratio.
The grid point showing the problem is part of a T-Branch zone. Changing the zone the same problem is moved away.
I've run the model with old formulation (CQUAD4) and the run was good (with K6ROOT=0).
I've then changed the PARAM SNORM to the value of 60 (default=20) and the run of first model (the one with CQUADR) was good. (No negative ratio).
The model is only build by shells. (quad & tria), there are some RBE3 and RBE2.
In the model there aren’t any CELAS, CBUSH, CGAP, CBEAMS or other elements.
RBE3 are 1to6 on dependent grid, 1to3 on independent grid.
RBE2 are 1to3 on independent grid.
Both kind of RBE have enough grid points to do not rotate on a line.
Materials are sandwich composites, so I've used pcomp cards.
I don't have zone with different stiffness very close together.
I've also performed a free-free check ( for reference) and results are corrects: 6 free body motion and seventh frequency 25hz.
the G-Set fail on DOF 5&6 all others sets (N-Set F-Set A-Set) pass the test.
The mesh is very good and there are only 15 elements that don't pass the standard Nastran test.
I’ve tried to change the geometry of T-branch elements and this seem to affect the problem, but I can’t change all stiffeners of my model to an incorrect geometry!
My questions are:
-1 Why there is so much difference in the stiffness matrix between the old and new formulation to cause negative ratio?
-2 It is correct to run with SNORM=60. or this number is way to high? Does this parameter influence the solution in any other way?
-3 There is any other way to have the model with CQUADR to run? (except with PARAM BAILOUT=-1 that isn’t acceptable!)
-4 Does anyone has noticed the same problem?
Thanks
I've first attempted to run the model with new shell formulation (CQUADR) using the Nastran system (370)=5 and Nastran didn't converge for an negative diagonal ratio.
The grid point showing the problem is part of a T-Branch zone. Changing the zone the same problem is moved away.
I've run the model with old formulation (CQUAD4) and the run was good (with K6ROOT=0).
I've then changed the PARAM SNORM to the value of 60 (default=20) and the run of first model (the one with CQUADR) was good. (No negative ratio).
The model is only build by shells. (quad & tria), there are some RBE3 and RBE2.
In the model there aren’t any CELAS, CBUSH, CGAP, CBEAMS or other elements.
RBE3 are 1to6 on dependent grid, 1to3 on independent grid.
RBE2 are 1to3 on independent grid.
Both kind of RBE have enough grid points to do not rotate on a line.
Materials are sandwich composites, so I've used pcomp cards.
I don't have zone with different stiffness very close together.
I've also performed a free-free check ( for reference) and results are corrects: 6 free body motion and seventh frequency 25hz.
the G-Set fail on DOF 5&6 all others sets (N-Set F-Set A-Set) pass the test.
The mesh is very good and there are only 15 elements that don't pass the standard Nastran test.
I’ve tried to change the geometry of T-branch elements and this seem to affect the problem, but I can’t change all stiffeners of my model to an incorrect geometry!
My questions are:
-1 Why there is so much difference in the stiffness matrix between the old and new formulation to cause negative ratio?
-2 It is correct to run with SNORM=60. or this number is way to high? Does this parameter influence the solution in any other way?
-3 There is any other way to have the model with CQUADR to run? (except with PARAM BAILOUT=-1 that isn’t acceptable!)
-4 Does anyone has noticed the same problem?
Thanks
