Incorrect result when using fixed number of iterations command

[luckypants]

Dear STKO Team：

I want to do the hybrid simulation, but implicit integration methods are not suitable for hybrid simulation. Therefore, I try to analyze seven story nonlinear model by using explicit integration methods or implicit integration method with fixed number of iterations.

After the analysis completed, the result was not correct. Although the top displacement and acceleration was little different, the plastic hinge result was very different. Then, I have tried to use other integration method but the plastic hinge result was still unreasonable.

How could I solve the problem?

The attachment is my model (scd file) and TCL file (Because there is no NewmarkHSFixedNumIter, HHTHSFixedNumIter and CollocationHSFixedNumIter in STKO, I change my model into TCL file to analyze).

Any advice and helps that you provide are highly appreciated.

Best Regards

[Clarabella]

Dear Luckypants,
did you calculate the critical time step for the explicit? Explicit solutions require very small time steps.
The best documentation can be found in Abaqus:
https://abaqus-docs.mit.edu/2017/Englis ... ynamic.htm

I believe your time steps need to be probably reduced.

Regards

[luckypants]

Dear Clarabella：

Thank you for your advice.

I change the time step of the ground motion 0.02 sec to 0.001953 sec, but the result is still wrong.

How could I solve the problem?

The following is the analyses command.

constraints Transformation
numberer RCM
system BandGeneral
test FixedNumIter 10
algorithm KrylovNewton
integrator NewmarkHSFixedNumIter 0.5 0.25
analysis Transient

Best Regards

[Clarabella]

I suggest to compare the results with a simple cantilever calculating the minimum time step starting from the formula I gave you.
Then we can complicate the model step by step

[luckypants]

Dear Clarabella：

Thank you for your advice.

I change my model to the simple cantilever column. The time step of the ground motion is 0.005 sec.

I use the same analyses command. The result is wrong.

The attachment is my model (scd file) and TCL file.

Best Regards

[Clarabella]

Please post the .scd of both files for comparison. You can run explicit also from STKO

[luckypants]

Dear Clarabella：

I completed the analysis in STKO. Here are the result.

Then, the attachments are my models (.scd).

Regards

[STKO Team]

If you take your FixedNumIter example, reduce the time step by 10 times, you will see that the results are still wrong.
But if you switch to Newmark instead of NewmarkHSFixedNumIter it will work.

The problem is in the integrator.
The NewmarkHSFixedNumIter does a very weird update of the solution. It compute the new solutiuon as a lagrange interpolation of the previous n solutions.
Probably this is fine for certain kind of probelms, but not this one, where you may have many strong discontionuities in the rotations due to plastic hinge.
I think this approach is good with smooth problems only

[luckypants]

Dear STKO Team：

Thank you for your reply.

But I want to do hybrid simulation, I should use explicit method or implicit integration method with fixed number of iterations. Then, Newmark method is implicit method.

So, is it possible to use explicit method or implicit integration method with fixed number of iterations to solve this problem?

Best Regards

[STKO Team]

Then you should directly move to explicit dynamics.
However, you will find many issues when running explicit dynamics for the first time...

The main issue is the critical time step. Theoretically, it should be less than 2/max_freq. However computing the maximum frequency requires the computation of all the eigenvalues of your system, and if the model is large it will take a lot of time.

Otherwise, you can use simplified formulas, for each kind of element. You can find information on the abaqus theory manual for explicit dynamics.

In general, for structural problems, the critical time step is of the order of micro-seconds!
In your model, with 26 seconds, it would require more that 10 millions time steps. I lowered down the stiff material and the initial stiffness of your Plastic hinge by 1 order of magnitude (it will not change the results) but will allow you to perform just 1 Million time steps.

Another problem is that for explicit dynamics, you need to assign a mass for every DOF. So you cannot just put some translational mass on the top of the column. Therefore I added also the column density as translational (rho * A) and rotational (rho * I) on the column. I assumed a concrete mass density.. You can change it.

Another problem is that, if you use the HingedBeam element in STKO, it will create the internal nodes on the fly, however those nodes will not have any mass.
To solve this problem I changed your model by manually splitting your column and creating 2 small pieces (1 mm each one). Then I assigned the zero length to those small pieces that would represent the zero length created by the HingedBeam.
The results are the same:

zoom.png (3.55 KiB) Viewed 5881 times

Here is your model for explcit dynamics. Try to play with it and understand how it works.

plastic hinge explicit 2.zip

(116.72 KiB) Downloaded 414 times