Dear all,
The sand hypoplastic model has been used to simulate the interaction between soil and pile. However, there is a bad convergence in the process of calculation. When I try to set 0.5m displacement loading on the top of pile, the calculating will end with very small increment. After checking the messages files, I find there is the key tip, as following:
Warning:
THE SYSTEM MATRIX HAS 7 NEGATIVE EIGENVALUES.
NEGATIVE EIGENVALUES MEAN THAT THE SYSTEM MATRIX IS NOT
POSITIVE DEFINITE:
FOR EXAMPLE, A BIFURCATION (BUCKLING) LOAD MAY HAVE BEEN EXCEEDED.
Could anyone kindly help me solve this problem? Thanks a lot!
Yours
Joe
Dear Joe, did you try to increase “pt” value? (parameter 2). It often helps. Regards David
Dear Prof. David
Thanks for your kind suggestions. However, sorry that it still doesn’t work.
Does the set of the interface between the pile and soil matter a lot? I just use *surface to surface contact. If I should use a hypoplastic interface instead of that? And there is another question like, in step modulus, what kinds of Matrix storage should I choose, unsymmetric or symmetric or use the default?
Sorry to take up your precious time, but your advice is really valuable.
Yours
Joe
Dear Joe, unsymmetric matrix. Yes, interface is important, please check this: https://web.natur.cuni.cz/uhigug/masin/plaxumat/node11.html. Although written for Plaxis, comments are valid for Abaqus as well. Mohr-Coulomb interface should be fine, but if parameter change does not help, you may try solid elements interface, either Mohr-Coulomb, or hypoplastic. There is also hypoplastic interface routine available at SoilModels. Regards David
Hey Joe,
You do not give too many details. It is difficult to say what could be the reason. I can still give some maybe useful tips.
Have you had a closer look on all the information in the message file? For example, I see you have some problem with maximum penetration vs applied displacement increment. Note that those tolerances on errors can be errors can be relaxed although cushion is needed when doing so [generally for beginners in modelling it is not recommended]. Where is the node soil 1.333? If at the top adding more cohesion could help as suggested by prof. Masin.
Based on my experience, usual interface definition is enough to get analysis converged.
Have you looked at the deformed mesh? Something strange maybe there?
Hope you can resolve your problem
Piotr
Dear Joe,
I think the main problem is interaction in your model in most of abaqus models interaction play a pivotal role in convergence.