I have the attached model. and I got stuck with this error for sol101
*** SYSTEM FATAL MESSAGE 3000 (SITDELC)
ITERATIVE SOLUTION FAILED DUE TO FAILURE OF PRECONDITIONER TO FACTOR.
THIS ERROR CAN RESULT IF THE STRUCTURE IS NOT RESTRAINED SUFFICIENTLY TO PREVENT
RIGID BODY MOTION OR IF INTERNAL MECHANISMS EXIST.
1 * * * END OF JOB
Solved! Go to Solution.
you don't say which solution you are running ! if SOL101 then 1st think to check is if you have some free parts in your model i.e. have you got things properly constrained or attached together!
Run normal modes (SOL103). If you get first near zero eigenvalues like 1e-3...1e-6 then you model is badly constrained. You can animate modes and see what parts are moving like rigid body.
Merci, en effet cela ma permet de voir les élements qui avait un mouvement rigide. cepandant méme aprés avoir supprimé ces élements le probléme persiste toujours.
qu'est ce que vous en pensez ?!
Can you post .dat file here, for those who use Femap instead of NX. Run SOL101 and search for *dat file in project directory.
Yes , this is exactly what I asked for...
Wow, 1,2M of elements. I don’t launch it, but for the first look:
1) There is one node that not connected to other structure. Constrain it by all six DOF`s or remove it.
2) RBE2 element connected to solids by two nodes. nodes of solid elements don’t have rotational DOF`s so in your connection model can rotate about line between these two nodes.
Add to spider at least one node not collinear with two existing nodes. Or constrain cylindrical surfaces of holes in lugs.
3) There are 319 Glue contact pairs they also can produce errors if you mess with contact (select wrong surface or search distance). .
Repair problems (1) and (2) and run Normal Modes. If you get additional zero modes - check animated view and fix/constrain free parts.
thank you very much for your détailed answer, i did run sol 103 and you were right about the rotation. but even when i did all what you said i still keep getting the same erreur message ?
and acctually i didn't really understand what you meant by "Add to spider at least one node not collinear with two existing nodes" !
Check two animations. When all dependent nodes of RBE2 are located on line you get pivot, first mode is zero, cube rotate about nodes without deformation. Move one node out of line and you get reasonable first mode.
Select RBE2 nodes as shown on the picture below. This is only Solid element feature you can conect plate or beam element with just one node to RBE element.
In addition, node, it is a rule of thumb - when you use linear tetra elements use at least 3-4 elements per thickness or enable midside nodes. Linear tetrahedral element have only one interpolation point and therefore they overestimate stiffness of structure. Therefore, you can get big error in deflection, stress and mode values.
Here is pictures from simple test - bending of cantilever beam. L = 1m, square cross section 1x1cm. Load applied at end. Last column in table is difference between FEA and theory. When I use 1 tetra element per beam cross section the difference is ~70%!!!
P.S. Why you don’t use Plate elements instead of solid? Your structure have many thin walls. With plate elements, you can dramatically reduce model size and therefore computational time.