Two story frame analysis(convergence problem)

[luckypants]

Dear STKO Team：

There is the convergence problem in my model. I try to build the two story frame with plastic hinges.
(The ground motion I input is 0921Chi Chi_TCU075, and I change its intensity to 1.5g.)

I have tried to use norm unbalance test, norm displacement increment test and energy increment test. It still not work.

The following is my model. How could I solve the problem?

Any advice and helps that you provide are highly appreciated.

Best Regards

[STKO Team]

I simply increased the maxDim of the Krylov-Newton to be equal to the maximum number of iterations. It means that the Krylov-Newton will update the tangent matrix only once for each step.

Basically it seams that the tangent matrix is too unstable in this dynamic problem.
I also increased the unifrom excitation of 1.5 just to push it into a more nonlinear range...

nonlinear.zip

(163.96 KiB) Downloaded 255 times

Senza nome.png (73.19 KiB) Viewed 4624 times

[luckypants]

Dear STKO Team：

There is one question I want to ask. When I check the analysis process, there are some errors in some steps(In the following picture).

Although it analyze successfully, there is some step which is not converge.

Could it influence the result or it is doesn't matter?

Thank you very much.

Best Regards

[STKO Team]

It goes on even if some step do not reach convergence because you are using the Adaptive Time Step we use in STKO.
Basically, if you use the Fixed Time Step (as usual in OpenSees), once a time step does not converge, the analysis stops.
With the adaptive time step, it first tries to solve the current time step with the current time increment. If it does not converge, STKO reduced the time step, and tries again.
So don't worry, it does not influence the results.

[luckypants]

Dear STKO Team：

So if I want to use the Fixed Time Step, is it possible to complete the analysis?

I am a beginner in OpenSees, and I am interested in how other people solve this kind of problem. Because they don't use STKO, they use Fixed Time Step in OpenSees. Once a time step does not converge, the analysis stops.

Do you have any ideas?

Thank you very much.

Best Regards

[STKO Team]

Well, if you want to use the fixed time step, start with a timestep of your choice. If it does not converge, the analysis will stop. So you will try again with a smaller time step.

[luckypants]

Dear STKO Team：

There is a question I forget to ask.

1. I check the default value of the maxDim is 3. So if I use the default value, the Krylov-Newton will update the tangent matrix 3/100 for each step. Is it right?

Because you increased the maxDim of the Krylov-Newton to be equal to the maximum number of iterations, and the Krylov-Newton will update the tangent matrix only once for each step. (It means 100/100=1)

2. What is your reason for changing the maxDim of the Krylov-Newton?

Thank you very much.

Best Regards

[STKO Team]

If the maxDim is 3 , it will update the tangent every 3 steps, so if your maximum number of iterations is 100, it will update the tangent 100/3 times...
I set maxDim =maxIter, so that it will update the tangent only once per step, more specifically at the beginning of step, which is the safest part to compute a stable tangent because it is close to. the previously converged point. Why did I do that? Because I noticed that updating the tangent too frequently was leading to divergence.

[luckypants]

Dear STKO Team：

I see. Thank you very much.

Best Regards