system Mumps in OpenSeesSP and push over

[alirezacivil]

I am doing gravity analysis and push over analysis in succession using OpenSeesSP.
When I use system Mumps in OpenSeesSP just for the gravity analysis it works. But it does not work for the push over analysis after performing gravity analysis.
What is the problem?

[fmk]

can you send the script.

[alirezacivil]

Hi Frank,

Thanks for your response.

First I perform gravity analysis with these scripts:
constraints Penalty 1.e25 1.e25
numberer RCM;
system Mumps;
test NormDispIncr 1.0e-6 20;
algorithm KrylovNewton ;
integrator LoadControl 0.1;
analysis Static;
analyze 10;

Then I carry out push over analysis with this code:
constraints Transformation;
numberer RCM;
system Mumps;
algorithm KrylovNewton ;
integrator DisplacementControl $IDctrlNode $IDctrlDOF $Dincr ;
analysis Static ;

In general, I am modelling a 12 storey building in 3D. The beam/column elements are elastic elements with concentrated hinges using zero length elements and shear walls are modelled using displacement beam column elements.
Because the code is too long, I cannot send it here. But, I can email you if it helps.

Also, I have another question, I saw somewhere you where suggesting to use:
system Mumps -ICNTL14 80;
What is it exactly, what does each number means?

Kind Regards,
Alireza

[fmk]

for now try without issuing the second system, numberer, algorithm, analsysis command .. as going from one static to another you do not need to redo these commands.

[alirezacivil]

Thanks.
It is ok now.

Can you please explain the "system Mumps -ICNTL14 80"? what is 80 means? I did not find it in wiki.

[fmk]

Mumps assumes a default amount of fill in during factorization and sets aside memory for it .. in a structural analysis problem if you have a number of MP constraints say for the floors this greatly increases the amount of fill in and the default is not going to be enough .. the switch just allocates more memory, 80 says it expects 80% fill in in the factorization.

[alirezacivil]

Hi Frank,
Thanks for your response.

I used -ICNTL 90. I got this error:
Work array too small; use -ICNTL14 option, the default is -ICNTL 20 make 20 larger.

The memory you are talking about is RAM in computer?
Should I enhance the RAM or I should use better CPU.
My CPU has 8 cores and I am using 7 cores to run the model. I kept one of them for windows to work on it.
What is your recommendation?

Thanks

[fmk]

the memory is not RAM, it is memory the solver puts aside for fill in during the matrix factorization phase, 90% says assume 90% fill in .. which is incredibly large for a sparse matrix, might as well make a full matrix.