Parallel and Sequential explicit analysis

[dudchenkoav]

Hello STKO Team!

I am sorry to trouble you but I have not found any information or suggestions for use of explicit analysis in opensees in parallel mode. Could you please clarify whether it is possible or not and what analysis settings are optimal in that case?
Could you also provide some information (unfortunately there is no information on it in https://opensees.berkeley.edu/) on the following transient integrators and what ones can be used in transient analysis:
HHT_TP
HHTExplicit_TP
HHTGeneralizedExplicit_TP
HHTGeneralizedExplicit_TP
KRAlphaExplicit_TP
NewmarkExplicit
What analysis settings for them can be adopted so that the solution could converged and be correct.


Best Regards,
Aleksandr

[Clarabella]

Dear dudchenkoav,
we are preparing a document for explicit integration. As soon as possible we will send it to you. Meanwhile please attend the webinar next Thursday, Massimo will talk about explicit integration. If you have specific questions he will be happy to answer.
Regards

[dudchenkoav]

Dear Clarabella,

thank you for your reply. I will be waiting for the information from you as it is quite important to understand appropriate settings for explicit analysis in STKO and OpenSee.

Best Regards,
Aleksandr

[dudchenkoav]

Dear Clarabella,

I am sorry to trouble you. Could you please provide me with the information on the status of the document on explicit integration in STKO?

Best Regards,
Aleksandr

[STKO Team]

Dear Aleksandr,

we are still working on it. If you want to start right now with explicit dynamics, probably the best way is to prepare your model and share it here.
We will give you the basic information you need to run it with explicit dynamics.
In this way, you will get the information you need while we proceed with the documentation

[dudchenkoav]

Dear Clarabella,

thank you for your reply. I will be waiting for the document. The model itself can be foung by the following link:
https://drive.google.com/open?id=1HK-Rt ... p=drive_fs
If you give me an advice on how to set an explicit parallel analysis I would really appreciate that.

Best Regards,
Aleksandr.

[STKO Team]

Dear Aleksandr,

The first thing that I noticed is that you are using ExplicitDifference. That is an explicit integrator based on the Leap-Frog method (by prof. Lu Xing Zheng, see his works here: http://www.luxinzheng.net/english.htm).

We are still studying his method, however, from what I know, it is supposed to work with a diagonal matrix, so it should not be used with rayleigh damping. Instead, it should be used with modal damping.

I think the best way is to start with the default explicit integrator, which is the Central Difference.
And I would start with a smaller model (let's say a portion of interest of your model), so it will be easier (faster) for you to test all the possibilities. Keep in mind that with explicit dynamics, the time step size is of the order of micro-seconds..

If it is fine for you, I will set up a modified version of your model, considering only the portion of you model near the bottom-left corner, where the collapse is triggered.

Another question:
Are you planning to use element removal?
If that is the case, the CentralDifference is not going to work (not because of the theory, but because it cannot handle a change in the domain).
In that case you can use the HHTExplicit_TP with alpha = 1

[dudchenkoav]

Dear STKO Team,

To be honest, I started with the explicit method as I thought it was the standard one due to lack of the documentation. So,the approach you suggested would be greate for me to understand how it works in OpenSee.
If it is possible, could you please also set the model with HHTExplicit_TP with alpha = 1 method?

Best Regards,
Aleksandr

[STKO Team]

Dear Aleksandr,

Here is a small portion of your model, with a coarser mesh, that we believe can be sufficiently small to run fast tests.

FrameConcentrPanel_v1-7SPexpl_reduced.zip

(455.17 KiB) Downloaded 208 times

I kindly ask you to clean it up, in order to make the test easier:

• remove materials/conditions that became "unused" after the removal of portions of the original model
• I've seen that you applied masses at the corner nodes of your grid. While the internal nodes of beams and columns will have no mass. This is fine in implicit analysis, but not in explicit ones... every free dof should have an associated mass.
  I suggest you to replace those lumped masses with edge-masses, so that STKO will distirbute them on all nodes of an element.
• Since this is not the final "real" model, you can even make it simpler, by assuming the same mass and force on all columns... It will be easier for me when I will explain some key points about rotational masses...

When you are done with this "simplification" post your model, so it will be easier for me to explain the keypoints of the explicit analysis

[dudchenkoav]

Dear STKO Team,

here is the corrected model attached. As you told me I did the following:
-added distributed masses in element properties for coulomns and beams instead of lumped;
-removed all unused properties and loads;
-changed loads and recalculated damping.
Please consider these changes.

Best Regards,
Aleksandr