Checking the validity of Opensees

[carlos]

Hello,

I would suggest to stablish a procedure for testing the accuracy of Opensees taking in account that:

Is a still developing program.
Softwares performing second order analysis can have big differences between them (see http://scholar.lib.vt.edu/theses/available/etd-04302001-120116/unrestricted/AMSThesis.pdf).


I have seen some testing models for testing on the developer package, but I think that are intended to test individual functions of Opensees and I have not found any documentation about these tests.

Some testing models could be included on the package, both for users and developers, so the users could have an idea about the accuracy or problems of the different versions on different conditions. It would also show a guideline for improvements on the software.

Examples could be extracted from the technical literature and compared with results obtained from commercial and tested softwares.


Regards,

[silvia]

Great idea, thank you!

I am currently working on a comparison of different materials and elements for RC structures. However, I am doing a comparison between the materials, not a validation, which is typically done by the element/material developers in their release, even though this does need to be done more rigorously.

It is difficult to compare opensees results to those obtained using other programs, as this is very time-consuming and costly. This can also be misleading, as it is difficult to judge which model is more accurate in predicting the response of "real" systems.

We are, indeed, working on a more rigorous protocol for testing OpenSees components.
_________________
Silvia Mazzoni, PhD
OpenSees-NEESit Simulation User Support Manager
University of California, Berkeley

[uyazgan]

This is definitely needed.

We have made some studies in this direction. We analyzed a single cantilever RC column in OpenSees, Ruaumoko, SeismoStruct, DRAIN-2DX and Rechenbrett-2D. We compared both the simulated response under static-cyclic and dynamic loading.

The computed maximum displacements were found to be inagreement with each other. However, differences could be seen in some other response quantities like residual displacements.

The results are available in:
Yazgan U. and A.Dazio, 2006. [i]"Comparison of different finite-element modeling approaches in terms of estimating the residual displacements of RC structures"[/i]. Proc. of the 8NCEE April 18-22, 2006, San Francisco.

[silvia]

interesting.
remember, though, I can give you different results within the same program, too, depending on what elements and materials I use.
i'll look over your paper,
thank you
_________________
Silvia Mazzoni, PhD
OpenSees-NEESit Simulation User Support Manager
University of California, Berkeley

[Boris]

Hello there,

This is fairly mature field (verification and validation).

It is generally not a good idea to compare two implementations directly, but rather go for verification against closed form solutions (math issue) and do validation against careful, simple (unit) lab tests (physics issue). Take a look at the excellent paper (somewhat long):

@inproceedings
{Oberkampf2002,
author = { William L. Oberkampf and
Timothy G. Trucano and
Charles Hirsch },
title = { Verification, Validation and Predictive Capability in
Computational Engineering and Physics },
booktitle = { Proceedings of the Foundations for Verification and
Validation on the 21st Century Workshop },
year = { 2002 },

editor = { },
volume = { },
series = { },
pages = { 1-74 },
address = { Laurel, Maryland },
month = { October 22-23 },
organization = { Johns Hopkins University / Applied Physics Laboratory },
publisher = { },
note = { },
napomena = { local CM988 ; 24Sept2002 },
}


There should be a copy (PDF) at

http://www.usacm.org/vnvcsm/

Best regards, Boris

[silvia]

thank you for the excellent reference!
_________________
Silvia Mazzoni, PhD
OpenSees-NEESit Simulation User Support Manager
University of California, Berkeley

[canim_ken]

hi there.

i have examined a 3-D linear 8storey model with SSI effect and a TMD attached. in both SAP2000 V12 and opensees and i got the same results in both of the softwares. the errors in eigen value problem were almost zer0 (about 0.0001%!).
i did time history anlaysis under bi-directional ground motions and compared the results. they were the same. i can send you my models and resluts if you want.
_________________
Hadi Kenarangi