The OpenSees Community

Hi Everyone,

I have a relatively large model that is taking approximately 12 days to run on a cluster. Unfortunately I'm not actually utlising the full capabilities of the cluster because I don't know how to run this large model using multiple processors.

Basically I successfully use OpenSeesMP for running parametric studies, but I want to run one long EQ motion for one model. I don't know how to speed up that process so that each analysis step is being done faster.

I've tried using OpenSeesSP but that is giving me all kinds of errors. I don't know if it's because of the way it was compiled on the cluster or due to some other reason.

So again I'm looking at ways of improving the speed at which each step is completed.

Any guidance would be much much appreciated.

Thanks for that, your work is much appreciated.

Do you have any advice on temporary solutions? Time is of the essence and this is a really big obstacle for me.

Any suggestions would be much appreciated.

Regards,

Any suggestions on solving this problem? It seems to be a weird and I can't find the source of the problem

Hi,

I am running an analysis using OpenSeesMP on a cluster and using the Mumps solver. I am getting the following error:

WARNING MumpsParallelSolver::solve(void)- Error -9 returned in substitution dmumps()
Work array too small; use -ICNTL14 option, the default is -ICNTL 20 make 20 larger

The interesting thing is I only get that error when I run more the 4 tasks. I have set the system command as follows:

system Mumps -ICNTL14 100

I tried numbers smaller than 100 but with no luck.

I can't seem why it fails only when i run more than 4 tasks in parallel.

Also is it possible to fix the element so it does work in a transient analysis? If so, any guidance with that would be much appreciated.

Thank for the prompt reply professor. Do you have any recommendations for the kind of interface elements I could use in my model of a wharf. I am developing a 2D model including the soil domain (SSPQuad) and the wharf (dispBeamColumn). How can I capture the response at the interace?

Does anyone have any experience with using the BeamContact2D element in a transient analysis? I have been having significant issues trying to get it to work at the interface between piles and surrounding soil. I have been having convergence issues even though I have tried a wide variety of changes to the analysis parameters including the solution algorithm, the size of the time step, and the integrator used. I tried having 1 element only and that worked, and I kept increasing the number of elements added and it works until approximately 5 elements. After which it no longer converges.

Any assistance would be much appreciated.

I only have the problem with a 778 triangular elements out of 9443 total elements. I thought it might be possible to create the triangular nodes and elements in a ndm 2 ndf 2 system and then constraint them using equalDOF in the 1 and 2 direction to duplicate nodes in the ndm 3 ndf 6 system which will have the shell elements in. Do you think that has a chance of working? I am also not sure if it makes much sense. I only need it to calculate the natural frequency of the structure.

I tried that but I am getting singularity issues. Here's the error when I assign Mumps for sys_of_eqns

WARNING MumpsParallelSolver::solve(void)- Error -10 returned in substitution dmumps()
cause: Matrix is Singular Numerically

and when using BandGeneral:

WARNING BandGenLinLapackSolver::solve() -LAPACK routine returned 1

Any other ideas that I could try to make it work in OpenSees.

Regards

Hi,

I have a gate that has already been meshed using another structural analysis software and contains quadrilateral and triangular plate elements. In converting the input files from this software to OpenSees I have found that there appears to be no triangular shell elements. Are there any elements that I am unaware of that could assist me in modelling these triangular plates because it is not practical for me to re-mesh the gate using only quad-shell elements.

Regards,

Hi,

Just checking what the purpose of this output in NodeRecorder.cpp on line 804

opserr << "NodeRecorder::initialize - numNodes: " << numNodes << endln;

It's coming up in one of my analyses and not sure if it is an 'error' or not

I am not sure if this is an error or not, but I am getting this message when I run a large model on OpenSeesSP on a cluster using several processors:

You just increased the maxndoms: 5 4


Does this have any meaning? It comes when I run 8 processes for example but not when I run 10.

I am facing convergence problems and I developed some code to try out different soultion algorithms and in most cases that works fine. However for a small number of simulations I am still not converging even after decreasing the timestep. In some cases changing the convergence test resulted in convergence. So I am wondering what is the guidance in terms of choosing the most appropriate test command? Or does it work like the solution algorithm, the idea being to try what works.

I only get memory corruption problems when I use mumps. I changed to umfpack and no longer memory problems. Any thoughts on why could that be? I tried using the -intcl14 option but no luck.

I am running a medium size model through approximately 100 eq motions and using OpenSeesMP to run multiple processes. I used code similar to the example provided with the source code. After many days of having random errors related to memory (glibc errors) it appears I have a race condition. I read something related to mutex which can be used to lock code while it is being executed. Can you please help me with how to implement that in OpenSees?