reducing model DOF's

aemmons · Post by **aemmons** » Thu Jul 21, 2005 10:09 am

I am using the equalDOF command to reduce the number of DOF in my system. I was using an eigen analysis to compare frequencies and mode shapes of the reduced model to that of the full model.

When I reduce 2 of the horizontal DOF's on each floor of the structure, I got good eigen/mode shape results as compared to the full model. Also good time history data.

When I reduce all the DOF's on each floor, I had exactly the same eigen/mode shape results as the partially reduced model. Seemed good ( i need as few DOF's as possible due to real-time considerations )

However, the displacement time histories between the partially and fully reduced models are very different, with the fully reduced model yielding bad results.

also I got different results if I:

equalDOF 22 23 1 2 3

....VS....

equalDOF 22 23 1
equalDOF 22 23 2
equalDOF 22 23 3

is there a more effective tool for model reduction???
any ideas???

Post by **fmk** » Thu Jul 21, 2005 10:21 am

the similarity of eigenvalues depends on the model and the stiffnesses of the elements .. typically equalDOF is just used to tie to adjacent nodes together .. to reduce dof in your model you would use the rigidDiahragm.

if you want smaller no.s of dof suggest you do your modelling differently.

aemmons · Post by **aemmons** » Thu Jul 21, 2005 10:28 am

SO,

the equalDOF cannot be used to "condense out" degrees of freedom.

this still does not explain why I get different results when I put all three DOF in a single command line VS three seperate lines.

Post by **fmk** » Thu Jul 21, 2005 12:41 pm

eualDOF does what it says it will do .. it makes the displ at two nodes equal .. you can use it to condense out DOF if you want .. that is a modelling decision that you make ..

as for not giving equal reults .. i missed that .. so here:
you are probably using transformation constraint handler for the analysis .. Penalty would give the result you are looking for .. the transformation constraint handler can only handle 1 at multi-point constraint at a node; in the second example you have 3 so it is iignoring 2 of them .. which ones ignored you can tell from looking at the displ at the nodes.