Initial state analyis

[miran]

Hi,

I have some problems running a pile - soil model. The model closely follows the model from the 22nd webinar on SSI. The initial state analysis works seemingly fine, but a error message concerning the inversion of the flexibility matrix occurs during the subsequent analysis. I cannot see any lack of constraint that are rigid body motions. Also, when I check the results in the post-processor, the domain is only partly recorded (added?) for the initial state analysis. Clearly I am getting something wrong. I have constructed similar models before, but never with this type of subdomains.

Any suggestions on what may cause this is very much appreciated. The model is added.

pile_soil.zip

(1.7 MiB) Downloaded 162 times

[STKO Team]

but an error message concerning the inversion of the flexibility matrix occurs during the subsequent analysis.

The error is due to the Force Beam-Column. It internally iteratively tries to find the deformation which is compatible with the assumed force field. To do so, it needs to invert the section flexibility matrix. When you have brittle materials (in your case either the shear model or the confined concrete that at a certain point drops to zero...), this may happen.

However, in your case, you have other issues before that:

• The contact at the bottom should be a BeamEndContact... while you used the same element as per the side contact.
• The DisplacementControl is not fine for OpenSeesMP. You should either use ParallelDisplacementControl or LoadControl with imposed displacement (which is a valid alternative only when you have a load on a single point as in this case)

After these changes, you will see that you will have a lot of convergence issues due to contact, and some compenetration.

However, we recently added a new contact element that has a much better convergence rate. You can refer to webinar #24.
In this working file, I used that element. In this case, the mesh-nodes of the pile are aligned with the mesh-nodes of the soil

pile_soil.zip

(1.7 MiB) Downloaded 174 times

[miran]

Of course, thank you! I have watched webinar #24 and have been playing around with the element. Very impressive! The model also works better with the displacement-based beam.

Both the displacement/force- based beams and the new contact element are somewhat new to me. I can find a lot of literature/videos on displacement versus force-based beams, but not so much regarding the contact element.

Have you written a paper or some form of documentation that describes the algorithms of the new contact element in more details? I am very curious on these types of implicit/explicit procedures because I am currently working on a macro-element for pile groups that will probably have some similar features.

[STKO Team]

I wrote this model but still working on publishing a paper on it.

Anyways, the contact condition follows the standard Coulomb friction law, for frictional contact. It has a faster convergence rate (even if you don't select IMPL-EX algorithm) because we formulated the plastic corrector in an alternative way, which leads to the same result, but avoiding internal iterative loops for the return mapping (as I said, I still need to publish about this feature).

Even if the convergence rate is faster, it is still implicit, so you may experience convergence issues in some problems. If this happens, you can switch on the IMPLEX algorithm, which makes the problem step-wise linear (but in this case you need to pay attention to the time-step size)
Everything you need to know about IMPLEX is in this reference:
https://www.researchgate.net/profile/Al ... oblems.pdf

Here instead you can find the documentation for the element in OpenSees:
https://opensees.github.io/OpenSeesDocu ... mplex.html

[miran]

Ok, I will read the paper. Thank you again for great support!

[STKO Team]

You're welcome