OpenSees freezing during analysis

[gor]

I have a relatively simple model and I am running a parametric study on it with a number of ground motions, so i have a simple for loop type analysis set up.

There is no problem running the model in pushovers or dynamic analysis, but sometimes when i start the loop, the analysis seems to freeze at a certain point. I mean freezing in the sense that it seems to be stuck on a particular run for an abnormally large amount of time. There are no errors in the analysis, such as non-convergence as I have numerous puts flags to say so. Currently I have one run stuck on i=14, where there is nothing special or different about the ground motion 14. In addition, the model has already run this ground motion 14 at a both a lower and higher scale factor before (I'm doing IDA). Previously, the same model had failed to start running at the very first analysis and just seemed to be stuck for a long time (few hours) on a single ground motion. However, if i hit ctrl+c it moves onto the next run no problem.

I have seen this on both Windows and Mac versions of OpenSees, where I start the model by calling it from the command or terminal window. I have a feeling this could be a computer programming issue rather than a structural analysis issue. Some thought/comments or even solutions(!) would be greatly appreciated.

[fmk]

are you using forceBeamColumn elements in the model? and just because a simulation gets through the higher and lower scaled motion, does not necessarily mean it will get through something in between the two.

it is possible you have some other issue, have you tried just running the motion with the scale factor sequentially to see how it runs.

[gor]

Hi Frank, Yes I have forceBeamColumn elements in the model.

Yes, but if I close the program and run that specific intensity again it has no problems whatsoever. For example, I closed OpenSees, rebooted the PC and started the analysis again and it has gotten past the last point where it got caught without any problems.

Since I was getting no convergence errors in the analysis, I figured it may have something to do with not having a wipe command where needed so things eventually clogged up. Has all the signs of a Heisenbug in any case, so I wanted to ask to see if any one has seen this before.

[fmk]

if it goes fast with same elements sequentially it is not internal iterations in force beam element which was one possibility .. could be a memory leak that is causing the application to start page faulting (you could use the system monitor to look at this), though if this was to happen the subsequent analysis should also be slow .. another possibility is writing to the same file

[gor]

Sounds like it may be the last one,

I have the following type loop set up

for i from 1 to # of ground motions
open a text file for ground motion i to print scale factors to
for j from 1 to # of scale factors
do some stuff to determine an appropriate scale factor
print the current scale factor to text file
run the analysis for current i and j
end
close text file
end

I'll try appending the scale factors to an array inside the j forloop and then maybe print just once to the text file for each i