help with equalDOF command

jwieser · Post by **jwieser** » Thu May 14, 2009 1:42 pm

I am trying to model an interior column of a multiple story building in a 3D model. There are four beams connected to column at each floor level, all of which are pinned.

I have defined 5 nodes at this location and connected them with the equalDOF command. The master node is defined as node1 and four different equalDOF constrains are used to constrain node1 to node2, node1 to node3, node1 to node4, and node1 to node5.

Is this the proper way of doing this or is there a problem with the order in which this is done? also is the transformation constraint handler ok for this definition.

If anyone has any input it would be helpful

jwieser

silvia · Post by **silvia** » Thu May 14, 2009 4:41 pm

if everything is pinned everywhere, why don't you just use trusses?
or just remove the rotational dofs...

jwieser · Post by **jwieser** » Fri May 15, 2009 8:05 am

This is part of a 3-story building with 6 bays in one plan direction and 4 bays in the other plan direction. The moment frames are on the exterior frames in each direction but the interior frame have simple beam column connections.

When you say remove the rotational degree of freedom do you mean within the equalDOF command? That is what I have done.

My issue is mainly with the periods that I am getting. They don't make sense when I implement end releases using the equalDOF command.

I would use zeroLength elements with low modulus material in the rotational DOFs but this greatly increases computational effort and time for just a simple connection.

Any suggestions would help.

Thank you

silvia · Post by **silvia** » Fri May 15, 2009 8:24 am

only use the equaldof for the constrained dofs at the nodes.

jwieser · Post by **jwieser** » Fri May 15, 2009 8:40 am

Thank you, That's what I have done. I think the constraints handler may be an issue. I understand that the Penalty and Lagrange methods are not the best for transient analysis. Thus that leaves the transformation method, however if I want to assign rigid diaphragms at the floor level I then have constraints in series, which I understand may be an issue.

Do you have any suggestions as to how to avoid constraints in series. Or a different way of modeling this type of floor system.

Thank you for you help.

silvia · Post by **silvia** » Fri May 15, 2009 9:02 am

i'll have frank look at it.

Post by **fmk** » Fri May 15, 2009 9:38 am

if the node at the column element is that master and the nodes at the beams are slaves you will have no problem using Transformation providing you have no other constraints, e.g. rigid diaphragm acting.

I don't like rigid diaphragm with anything but elastic beam elements becauase of the high and unrealistic axial forces they tend to generate in the nonlinear beams. I suggest using quad elements or Penalty method if this is the case. But have a look at the forces yourself to see what can happen.

the limit on the use of the Transformation constraint handler is that one node cannot be both a slave and a master.

as for ZeroLength elements, adding them should not greatly increase the computation time if you choose a sparse solver, e.g. UmfPack (or Mumps in the soon to be released version 2.1.1). If it does prove to be send me the model so i can have a look at what the bottleneck is.

Also as these joints are not truly pinned in real life it is probably a better modelling assumption to model with ZeroLength.