Trying to remove nodes

motalaat · Post by **motalaat** » Sun Jul 17, 2005 3:59 pm

Hello, I have a 2D propped cantilever model. Nothing fancy - two orthogonal beanWithHinge elements set up between 3 nodes. Two nodes are constrained in all DOF and one is free. Within the analysis, I try to use

Code: Select all

remove element; remove node; remove loadPattern

to simulate element loss. The program crashes without even going on to the analysis phase!!!

If I limit myself to removing only the element and loadPattern, the program willl not crash, but of course if I issue an analysis command later on it will not converge because one node will be unconnected to any elements. Any idea as to what might be wrong with the

Code: Select all

remove node

command? Might it be that the node to be removed has constraints and those - too - have to be removed. Will that be

Code: Select all

remove constraint

or what? And what would be the syntax for it?

silvia · Post by **silvia** » Sun Jul 17, 2005 6:16 pm

i have had problems with this, too. and, ultimately i do think there is a problem with opensees, but i don't remember.
however, the order in which you remove things is important, you can't remove nodes before you remove a load pattern, the load-pattern removal should be done first, then the elements, then the nodes -- top to bottom kind of thing.
hope this helps, and let me know if you still have problems.
tomorrow i'll try to remember to find my e-mail to frank about this.
silvia

motalaat · Post by **motalaat** » Sun Jul 17, 2005 7:44 pm

Thanks, Silvia, for a very prompt reply on a Sunday

I had tried this. I have also dug into the code and found that there is a "remove sp" command, which supposedly removes constraints and "fix" commands". I tried to understand the syntax for that from the coding and I think I got it properly ow to identify an sp-constraint tag. Anyway, when I tried doing that, it didn't help, and didn't do nothing worse, either. So, I can remove elements, loadPatterns, possibly constraints, but the moment I uncomment the line saying "remove node 3", and source in the file, it crashes wihtout even going into the analysis phase. Seems to me there is some memory leak left there in the implementation somewhere.

Cheers,

M.Talaat

silvia · Post by **silvia** » Sun Jul 17, 2005 9:05 pm

let's see what Frank has to say on this one!!

Post by **fmk** » Tue Jul 19, 2005 9:05 am

post the simple script and i will have a look at it.

motalaat · Post by **motalaat** » Wed Jul 20, 2005 12:08 am

Here is the code. Ideally, the "dummy" element 3 should be commented out, and the two commented-out lines near the very end should be uncommented. However, the program crashes before performing the analysis command. Adding the "dummy" element and commenting out the node removal lines was a way to test that the model is otherwise fine and running. However, the dummy element should eventually go and nodes that have no elements attached to them after an element fails should be removed to avoid a singularity.

Code: Select all

model BasicBuilder -ndm 2 -ndf 3

# Model definition
#define nodes and constraints
node 1    0.0   0.0
node 2    0.0  100.0
node 3    100. 100.0
fix 1 1 1 1
fix 3 1 1 1
mass 2 1. 0. 0.

#define materials and sections
uniaxialMaterial Steel01 1 5000.0 1750000.0 0.01
section Uniaxial 1 1 Mz

#define elements
geomTransf Linear 1
element beamWithHinges 1 1 2 1 10 1 10 21000.0 100.0 833.33 1 
element beamWithHinges 2 2 3 1 10 1 10 21000.0 100.0 833.33 1 
element beamWithHinges 3 3 1 1 10 1 10 21000.0 100.0 833.33 1  ; #dummy

# Define recorders
recorder Node -file Node.out -time -node 2 -dof 1 3 disp
recorder Element -file ele.out -time -ele 1 force
recorder plot Node.out Node-2-Dir-1 0 0 350 350 -columns 3 1 -columns 2 1

# Define loads and analysis
pattern Plain 1 Linear {
sp 2 1 10.0
}

#static analysis prior to sudden removal of element 2
constraints Penalty 1.e12 1.e12
numberer RCM
system BandSPD
#system SparseGeneral -piv
test NormDispIncr 1.0e-12 100
algorithm Newton
integrator LoadControl 0.01
analysis Static
analyze 10

print model.out

# now delete element 2 and run an undamped dynamic analysis from its deformed position
set dt 0.1;
integrator Newmark 0.5 0.25
analysis Transient

remove loadPattern 1
puts "pattern removed"
remove sp 3 1 
remove sp 3 2
remove sp 3 3
puts "constraints removed"
remove element 2
puts "element removed"
#remove node 3
#puts "node removed"

analyze 100 $dt

wipe

Post by **fmk** » Wed Jul 20, 2005 3:09 pm

for now you will have to do a wipeAnalysis before you remove nodes .. it will be fixed in next release .. or you can get it now if you use cvs.