Substructuring 1

[Luca97]

Hello, I'm trying to create a superelements, but I guess that I'm doing something wrong. That's the sequence of command that I'm using now:



/PREP7
SETTOL,0.00000001
NSEL,s,loc,z,0,294.3225 !Here I select what I need in
KSEL,S,LOC,Z,0,294.3225 !the structure, and I copy the
FTRAN !forces from KP to Nodes.


/FILNAME,SUBSTR
/SOL
ANTYPE,SUBSTR !That's are the general options
NLGEOM,OFF !for the solver
NSUBST,20,100,1
EQSLV,FRONTAL
AUTOTS,1

SEOPT,SUBSTR,1,1,1,RESOLVE
SAVE,MODEL

At the end I run a "SOLVE", ansys says to me a warning that is pointless, because it's referred to the fact that my materials are nonlinear and the program is running a linear analysis.
The problem is that I do not have other errors message, but it doesn't generate in my working directory the GEN.SUB file that I need to run the nonlinear analysis of the whole structure later.

Of course I'm using that sequence because I'm using a "Top-Down" substructuring, so I've the whole model and I've selected the part that I must transform in a superelement.

Any ideas?

Thanks,

Best regards.

 

[Transient1]

I've never seriously done substructuring....but it looks like you haven't selected any master degree of freedom nodes. Have you read the substructuring chapter in the advanced analysis help book and look at the Verification Manual examples?

 

[Luca97]

Yes, that was one of the problem. At the end i've reached my purposed and i runned the substructuring analysis. Now my only only problem is that I've to connect the superelement with the non linear part... I'll figure out something...

 

[Stringmaker]

To interface your parts use contact and make contact make nodes of the contact elements master DOF's as well.

 

[Luca97]

Are suggesting me to utilize a contact element to connect my non linear part of the model to the superelement? Because I don't know if is it a real good idea, my model is quite big (300k DOF) and I've used the SE to reduce the number od DOF and the computational time. I was thinking to use CPINTF, because I've a contact planar surface, do you think that is a bad idea?

 

[cbrn]

I think it's a better idea to use full-formulated contacts, as suggested by Ansys Manual. In fact, couplings would not capture any non-linear behaviour. 300000 dofs is really not a threshold to worry about, in general. If you find "eternal" solution times, I think it's because you have convergence problems rather than mem occupancy problems...

 

[Luca97]

Yes, that's possible... I was guessing that 300000 dof was a big number, but I've seen that Ansys's Manual talk about 10 milions DOF problems.
I've tried to run a complete analysis, it takes me about one day, and at the end the solution didn't converge, because it has said to me the one element has to much displacemente (it's very weird because this element should be connected with another element, but it's like that it has a "free end", that it should'nt have, because it's connected). So if the problem is the convergence usually what do you do to solve the problem?

Sorry if I ask stupid questions, but it's the first time that I use Ansys, it's a very big software, and I've read the manual, but it is so big that I cannot read everything and maybe I missed something of important.

Thank you again.

 

[cbrn]

Hi,
no, there are no stupid questions at all !
I think there might be some problem with the DOFs of at least one master node, if you "condensed" parts into superelement matrices.
I've never done a serious substructuring until now for lack of time (I would need it for a very, very complex system which, for the moment, we analyze "in pieces" making assuptions for the mutual influences), but I suppose most of these problems could disappear by re-formulating the substructures using the contacts. I mean, first of all setup the assembly with all the contacts needed, then "substructure" it.
The drawback of all this, is that contacts have many parameters to adjust, which can heavily influence the convergence (or trivially the "quality" of the solution, if and when it converges).
I'd suggest to let you be helped by your Ansys Support for a complicated matter like that, especially if you are somewhat new to the program.

 

[Luca97]

Well, I didn't explain myself very well I guess...
Now I'm trying to run the analysis without the superelement, because I need to compare the results after, between the whole model and the model with the superelement, to assure me that my results have good accuracy. But the solution doesn't converge, at least, it doesn't converge with manual setup of the options:

/SOL
!General solution options:
ANTYPE,0
NLGEOM,OFF
NSUBST,20,100,1

!Set the solution solver
EQSLV,SPARSE
!MSAVE,1
!PRECISION,1
AUTOTS,1

!Nonlinear options:
LNSRCH,AUTO
NEQIT,20
NCNV,2
TREF,20

!Results:
OUTPR,ALL,3


This kind of sequence doesn't work, after one day of calculations it doesn't converge.

So now I'm trying with SOLCONTROL,ON

I will see, tonigth I should have the results... Because I've tried to run all the CHECK command that Ansys has, and I didn't find out any errors.