Page 1 of 1

Equal DOF incoherency in Eigenvalue Analysis

Posted: 19 Oct 2010, 19:37
by Andre
Hi, I'm working on a model where I have pairs of nodes connected with equal DOF constraints (x+y+rx+ry+rz), in order to unite originally separated buildings.

When I run an eigenvalue analysis the deformed shape of the first 3 modes is very reasonable and one can see the expected effect of those constraints.

However, from the 4th mode on, I realise that the rotations of those pairs are not equal. In fact, the deformed shape shows almost independent behaviour of the buildings.

First I thought this could be just some sort of graphical bug. But then I checked the step output table and the rotations that are supposed to be equal on those pairs of nodes are actually different, confirming that it's not just a graphical problem. I've checked the constraints very carefully and I'm sure that they are connecting the right nodes. In fact, all the translations are equal. Only the rotations are different and only from the 4th mode on (vertical modes).

I apologise for the extension of the text. Thanks a lot once again for this forum.

Re: Equal DOF incoherency in Eigenvalue Analysis

Posted: 20 Oct 2010, 11:49
by seismosoft
André, hi.

Please try to reproduce the apparent inconsistency in a small model featuring only a few elements, and then send it to us, if the problem does still present itself.

Thanks,

Seismosoft Support

Re: Equal DOF incoherency in Eigenvalue Analysis

Posted: 30 Oct 2010, 04:35
by seismosoft
Dear André,

You can readily eliminate the apparent inconsistency by either increasing the Rigid Link Weights of the Penalty Functions or by using Lagrange Multipliers for the modelling of the constraints.

Seismosoft Support

PS: on a different note, we found your model to be excessively discretised (e.g. 14 elements to model a 4 metres beam) which not only renders the model very heavy and thus very slow to run, but may also lead to numerical difficulties and/or errors.

Re: Equal DOF incoherency in Eigenvalue Analysis

Posted: 06 Nov 2010, 20:39
by Andre
Thank you for your advice, it worked.

About the slab elements that lead to excessively discretised beams, how can I reproduce the orthotropic behaviour of the slab (flexure stiffness in only one direction) without them? In models without the the concrete walls used to retrofit, the flexure stiffness of these slabs is very important because there are no beams in this direction. Actually, if I remove them the first mode of vibration is in Y direction and the second in X direction, when it's supposed to be the other way around.
I'm thinking of replacing them by other elastic frame elements connecting only column nodes, with an EI that leads to the same horizontal displacements (Y direction) of a similar model with shell elements. However that would affect the moments at the columns. Is there a way?

Sorry for the amount of questions so far and thank you for your kind attention.

Re: Equal DOF incoherency in Eigenvalue Analysis

Posted: 09 Nov 2010, 12:35
by seismosoft
André,

We see your point, and understand your modelling needs. Modelling of slab flexural stiffness is complex indeed, and it might well be that the approach you had in place is the most correct one; you will find some past research work in the literature on this subject, but not always with definitive conclusions.

Hence, do feel free to proceed in the manner you had originally devised, trying perhaps to use elastic rather than inelastic beam elements whenever possible (i.e. whenever you are sure inelasticity will not develop), since the latter may lead to erroneous results (or numerical instabilities) when their length is small (with respect to the section dimensions).

Seismosoft Support