asdeasoft.net

Dear STKO Team,

I was studying the model from Massiomos' last Webinar. I've changed the mesh as he described during the webinar. But when I ran I got the error like below. What am I doing wrong? Attached are the error and the model.

P.S. I sent an email on Monday, I wonder if you've received it.

Thank you
Oguzhan

Hello STKO Team,

I solved the problem, I forgot to add selection sets for monitor soil base and top...

Now It's running. You can ignore the post. But I'm still waiting for your response to my email.

Have a great weekend.
Oğuzhan

Dear Oguzhan,
we are glad you found a solution to your problem.
I will get information on the status of your email request.
Have a good day.
Francesca

Dear Francesca,

Thank you for your reply.

About the model, I run it and I got the Warnings as below. I did as the instruction led me. And I checked -ICNTL14 box in dynamic analysis, left it as default for ICNTL14 Value as 200.

I successfully ran the analysis. Actually, computation time was less than before. (when I ran the Massimo's model with finer meshes in the middle of the model without checking ICNTL14 it took 21 hours, my model with ICNTL14 Value as 200 took 13 hours)

But I don't know what I am doing with this -ICNTL14 value.

Could you please explain what it means? Thank you

Regards,
Oğuzhan

ICTL14 attribute controls the "percentage increase in the estimated working space"
If it's equal to 20 for example, It means you are telling the solver that it can increase the allocated space for the analysis by maximum of 20 times, it is a way to avoid OpenSees saturating your compute memory. How large is your model?

Sometimes the MUMPS solver requires more memory than what it initially estimates.
the ICNTL value allows you to increase the amount of memory MUMPs can allocate

marafini.f wrote: ↑

Mon Nov 22, 2021 10:21 am

ICTL14 attribute controls the "percentage increase in the estimated working space"
If it's equal to 20 for example, It means you are telling the solver that it can increase the allocated space for the analysis by maximum of 20 times, it is a way to avoid OpenSees saturating your compute memory. How large is your model?

I see, It's 200 meters, cubic 3D-solid model. So it is about memory.

STKO Team wrote: ↑

Mon Nov 22, 2021 10:55 am

Sometimes the MUMPS solver requires more memory than what it initially estimates.
the ICNTL value allows you to increase the amount of memory MUMPs can allocate

So, it doesn't affect the result?
Computation time decreased almost in half when using 200 as the ICNTL value.
Is it normal?

So, it doesn't affect the result?

No, it doesn't. It's just about solving the linear system of equations. If it fails, it will give an error (no result), it won't lead to a wrong result.

Computation time decreased almost in half when using 200 as the ICNTL value.
Is it normal?

Yes, because when MUMPS fails in solving the system of equations, it will result in a non-convergent step, so OpenSees will try to cut the time step and try again.

Anyway, it typically happens when the model is highly non-linear, so the solver needs to do a lot of PIVOTING in the factorization of the system, but our example about the 3D model should be elastic... what model are you referring to?

Yes, it is actually elastic, but I'm not using the one you uploaded. I remodelled the old model from the previous webinar as I was watching Massimo's last webinar. I built finer meshes in the center like a + symbol as Massimo described.

And I thought the warning was because of me meshing finer so it was computationally cumbersome.

I attached the model. By the way, thank you very much for your replies.