Analysis Error (Convergence problem)

Post Reply
luckypants
Posts: 42
Joined: Wed Jun 17, 2020 6:07 pm

Analysis Error (Convergence problem)

Post by luckypants » Tue Sep 15, 2020 12:13 pm

Dear STKO Team:

There are some problem with my model.

I want to set the plastic hinge at the location of 0.95 and 0.05 of the beam(The location both ends of the beam equals 0 to 1.), because the failure usually happens at that location not at both ends of the beam. Then, I analyze it through pushover analysis and time history analysis.

After I finish my analysis, the result of pushover analysis is almost the same as the one in ETABS(I set plastic hinge at the same location in ETABS).

But when I try to analyze it through time history analysis, the result is not convergence.

The model seems to be worked only in pushover analysis.

The following is my model(pushover analysis and time history analysis). How could I solve this problem?

Any advice and helps that you provide are highly appreciated.

Best Regards
Attachments
ETABS.jpg
ETABS.jpg (15.06 KiB) Viewed 343 times
error.jpg
error.jpg (323.22 KiB) Viewed 345 times
pushover.jpg
pushover.jpg (26.69 KiB) Viewed 345 times
dynamic.rar
(95.16 KiB) Downloaded 13 times
pushover.rar
(85.9 KiB) Downloaded 10 times

STKO Team
Posts: 325
Joined: Tue Oct 29, 2019 8:45 am

Re: Analysis Error (Convergence problem)

Post by STKO Team » Wed Sep 16, 2020 9:25 am

In the static case I told you to use the NormUnbalance because testing the displacement was not enough with that rigid problem.
However the secon part of the analysis is not that rigid al all (when it goes into plastic regime). Now in the static it was still working. However here in the dynamic case probably the sudden change from very stiff to perfectly plastic, generates some very high inertia forces... And the Norm of the Unbalance becomes to high with respect to your tolerance.
You could use some damping, but since you are doing a comparison it is not fair.

So I simply decided to test for EnergyNorm. with a very small tolerance = 1.0E-14. Keep in mind that the energy norm is
dot_product(dU, Residual)... so it should be very small becasue it multiplies each term of the displacement increment vector with the respective term in the residual vector.

I also added the monitor. It is a new feature that allows you to monitor in real time some responses.
monitor.png
monitor.png (246.72 KiB) Viewed 336 times

luckypants
Posts: 42
Joined: Wed Jun 17, 2020 6:07 pm

Re: Analysis Error (Convergence problem)

Post by luckypants » Wed Sep 16, 2020 3:22 pm

Dear STKO Team:

Thank you for your help. I appreciate your effort to revise my model.

Displacement time-history response of the model in STKO is almost the same as the one in ETABS.

Then, I want to ask the question.

1. You told me to use the test normUnbalance, when I had a stiff problem. But in this case the test normUnbalance seemed not to work. You use the test Energy increment. My question is why not you used the test Energy increment first in the previous case? If there were some shortcomings in the test Energy increment?

2. I don't really understand why do you choose a very small tolerance = 1.0E-14? How do you decide this value? Could I choose 1.0E-18 or 1.0E-20?

3. Is it convenient for you to upload the model you revised ?

Best Regards
Attachments
displacement.jpg
displacement.jpg (44.48 KiB) Viewed 327 times

STKO Team
Posts: 325
Joined: Tue Oct 29, 2019 8:45 am

Re: Analysis Error (Convergence problem)

Post by STKO Team » Thu Sep 17, 2020 9:09 am

Is it convenient for you to upload the model you revised ?
Oh Yes, I forgot to attach it to the previous post.
plastic hinge.zip
(134.93 KiB) Downloaded 15 times
You told me to use the test normUnbalance, when I had a stiff problem. But in this case the test normUnbalance seemed not to work. You use the test Energy increment. My question is why not you used the test Energy increment first in the previous case? If there were some shortcomings in the test Energy increment?
I told you to use normUnbalance because it was easier to explain... you cannot check displacements -> check forces... I already knew that some problems could arise, since after yielding, your problem is not that stiff anymore...
I will explain it more in detail in the webinar. I did not tell you to use directly the energy test, because often peaople do not know how it works and they choose wrong values.
Could I choose 1.0E-18 or 1.0E-20?
No! those values are too small, it's very hard for a floating point (from a computational point of view) to be that small due to round-off errors.
Why 1e-14?
a signfican value for displacement is 1e-11, that was the order of magnitude of your yield rotation.. and 1e-3 was an acceptable tolerance for forces... so, since the energy is the inner product of the solution vector and the residual vector... I used 1.0e-14

luckypants
Posts: 42
Joined: Wed Jun 17, 2020 6:07 pm

Re: Analysis Error (Convergence problem)

Post by luckypants » Thu Sep 17, 2020 2:58 pm

Dear STKO Team:

Thany you very much. :D :D

Best Regards

luckypants
Posts: 42
Joined: Wed Jun 17, 2020 6:07 pm

Re: Analysis Error (Convergence problem)

Post by luckypants » Sun Sep 20, 2020 8:02 am

Dear STKO Team:

There is a new question I want to ask.

How do you know the value 1e-3 was an acceptable tolerance for forces?

How do you decide the value?

Best Regards

Clarabella
Posts: 33
Joined: Wed Jul 08, 2020 3:50 pm

Re: Analysis Error (Convergence problem)

Post by Clarabella » Mon Sep 21, 2020 8:43 am

Please check the webinar of September 17.

Post Reply