
Re: Stiffness matrix diagonal coefficient is zero at equation: 20407
Anthony Botting May 5, 2011 3:09 PM (in response to Daniel Hernly)Hi Daniel:
That "zero in the main diagonal" just means the solver cannot invert the stiffness matrix, so it gets stuck. Well, that's the math explanation, but it translates to physically mean there is a component loose in the assembly (they call that a "rigid body mode").
Try checking for bonded contact between the part you removed and other parts. There could be a gap and the nodes did not "bond" across the junction.
..or the part is just not restrained enough
...or the math got messed up when the solver generated the stiffness matrix. You could try a different mesh density (and even try the other mesher  not sure which one you used first, but try the other one) and see if it solves.
There is also a "Soft spring" option in the properties of the study. That will insert a very small value of stiffness to prevent the zero in the main diagonal, which is not supposed to affect the answer very much. There is also an "inertial effect" switch in there that is designed to offset any imbalance between applied loads, reactions, and inertia (happens sometimes in the nth digit on dynamic problems)  you can switch that on, too.
If it only runs with those switches on, make sure to plot displacements, undeformed, and observe colors. The redcolored areas should reveal components that are 'loose'.
Hope that helps a little.
Tony

Re: Stiffness matrix diagonal coefficient is zero at equation: 20407
Daniel Hernly May 6, 2011 9:13 AM (in response to Anthony Botting)Tony,
I had talked with another individual about this problem and they suggested the same thing you did. They also suggested that I change the solver to FFEPlus. I did that first and when I ran the results it worked correctly.
Thanks for your help.
Daniel

Re: Stiffness matrix diagonal coefficient is zero at equation: 20407
Anthony Botting May 6, 2011 9:49 AM (in response to Daniel Hernly)Thanks Daniel for letting me know. It appears the math may have gotten mixed up when the Direct Sparse solver created the stiffness matrix. I would like to know, do the deflections (using the FFEPlus solver) appear reasonable? (i.e., nothing is loose?) The only reason I ask is to try and determine if there's something amiss with the Direct Sparse solver that would make it worthy to turnin to the developers to fix. Thank you.
Tony

