UpdateMaterialStage in OpenSeesSP

[brag006]

Hi Frank,

Any luck with this issue. The problem I am finding is that in the updatematerialstage command after it searches the first subdomain, it doesn't go on to the second subdomain.
Basically it searches the domain for an element which uses the material to be updated. As soon as it finds that element it updates the material. Unfortunately when the domain is partitioned the loop looks through the first subdomain and then fails to move on to the second domain. How can I force it to search through all subdomains?

[fmk]

no. and not sure what code you are looking at , the code is in the SRC/material/nD/soil directory for the updateMaterialStage. it adds a new parameter to the domain and then updates it. the code for this is in SRC/domain/domain/PartictionedDomain.cpp .. if you look each subdomain is asked to add and update before the domain on P0.

the problem lies in the recvSelf of the PDMY02. in fact there are quite a few errors in the recvSelf, and i have not got them all fixed yet! .. still have a problem with loadStage set to 1, stages 0 and 2 are now working.

[brag006]

Any luck with this yet?

[fmk]

yes. should be fixed. i preseumed it was the same problem in thread:

http://opensees.berkeley.edu/community/ ... =7&t=60879

[brag006]

Hi Frank,

I seem to be still having the same problem. I recompiled opensees with the latest code but I still get the following error:

WARNING: MaterialStageParameter::setDomain() - no effect with material tag 2

I am running this example (http://opensees.berkeley.edu/wiki/index ... _Analysis))

[brag006]

In some case I am getting the following error:

StaticDomainDecompositionAnalysis::recvSelf - failed to get the Solver
Fatal error in MPI_Recv: Message truncated, error stack:
MPI_Recv(184).....................: MPI_Recv(buf=0x3baabb0, count=4, MPI_INT, src=0, tag=0, MPI_COMM_WORLD, status=0x7fff44472a10) failed
MPIDI_CH3U_Receive_data_found(131): Message from rank 0 and tag 0 truncated; 24 bytes received but buffer size is 16


Btw did you fix both the PDMY and PDMY02?

[fmk]

1. the updateMaterialStage::setDomain could still occur if the material does not exist on all domains.

2. i will need the example for the second error.

3. The PDMY and PDMY02 ar both in SRC/actor/objectBroker/FEM_ObjectBrokerAllClasses.cpp

[brag006]

I am running the site response analysis geotechnical example (http://opensees.berkeley.edu/wiki/index ... _Analysis)) and I get the following error:
StaticDomainDecompositionAnalysis::recvSelf - failed to get the Solver
Fatal error in MPI_Recv: Message truncated, error stack:
MPI_Recv(184).....................: MPI_Recv(buf=0x208abb0, count=4, MPI_INT, src=0, tag=0, MPI_COMM_WORLD, status=0x7fff6f78fe20) failed
MPIDI_CH3U_Receive_data_found(131): Message from rank 0 and tag 0 truncated; 24 bytes received but buffer size is 16


Also I ran the effective site response analysis example (http://opensees.berkeley.edu/wiki/index ... oil_Column) and I also get the same error:

StaticDomainDecompositionAnalysis::recvSelf - failed to get the Solver
Fatal error in MPI_Recv: Message truncated, error stack:
MPI_Recv(184).....................: MPI_Recv(buf=0x33b4bb0, count=4, MPI_INT, src=0, tag=0, MPI_COMM_WORLD, status=0x7fff5a149bf0) failed
MPIDI_CH3U_Receive_data_found(131): Message from rank 0 and tag 0 truncated; 24 bytes received but buffer size is 16

[brag006]

When I change the solver to Mumps I get the following error:

WARNING: MaterialStageParameter::setDomain() - no effect with material tag 1
WARNING: MaterialStageParameter::setDomain() - no effect with material tag 2
...
...

For all the materials.

[brag006]

I think this is working now. Initially I didn't understand your answer correctly.

[Ozgun]

hi brag006,

Can you briefly explain the way you achieved to solve the problem you stated here?
I am running site response analysis.
Depending on the size of the problem, I am having the warning "WARNING: MaterialStageParameter::setDomain() - no effect with material tag 1". An this warning is for all of the elements I have in the model.

The interesting thing on using PDMYM with Mumps is that, even though I have this warning in some of the models,the results are okay. However, if i use Mumps for gravity analysis and SuperLU for transient, I am having a little bit distorted results with the same warning still staying there. If i use SuperLU in both gravity and transient analysis, even though the analysis are complete, I am ending up with rigid body motion (basically no relative acceleration).

In case I use Dafalias Manzari model, totally reverse observations are occuring. Using Mumps does not work at all. SuperLU ends up with rigid body motion, which is weird because I do not have any problem in sequential analysis.

Thanks for your response in advance,

Ozgun

[mo_zarrin]

URGENT ATTENTION IS REQUIRED FROM DR. FRANK:

I am going to simulate the seismic response of a 3D model of Jacket Type Offshore Platform with its 3d soil environment. According to its long piles (4 Piles), my model will contain approximately 100000 to 200000 solid 8 node SSPBrickUp elements. This is a huge model which requires at least 15 Gb of Ram and takes about 2 months of running time in a capable computer, which makes it even impossible to conduct.Therefore, the only way of running it is taking the advantages of Parallel Computing.
At the moment, I am in the verification phase of the project using OpenSeesSP. Unfortunately, every thing is running without any error except the UpdateMaterialStage command in OpenSeesSP.
I am going to use PDMY and PIMY material for soil constitutive behavior, BeamContact3D for interface between soil and pile and NonlinearBeamColumn elements for pile and structural members.

I have provided for you 3 examples of single pile model created with OpenseesPl software (Please find the attached rar file).

https://www.dropbox.com/s/cu378ozansmqa ... P.rar?dl=0

All these models works well in sequential run. The first of them is a small model (smodel0) which does not work at all using OpenSeesSP parallel. This is because of updateMaterials command. The second model (smodel1) is the same as the first one with the disabled "updateMaterials" command. This model works in parallel run which means that a small model works well in parallel version. The only thing that does not work in this model is "updateParameter" command.
The third of them (lmodel) is relatively a larger model of single pile which shows an UpdateMaterialStage error after the elastic gravity analysis.
WARNING: MaterialStageParameter::setDomain() - no effect with material tag ....
It seems that this command has a problem with domain decomposition when the model becomes large. It should be noted that this large model contains the materials,elements and analysis commands similar to the previous models and the only difference is the size of the model.
I have tried every thing to solve the problem, trying different solvers(SuperLU, Mumpus -INCTL 80, ....), different numbrer, different constraints and so on. Nothing could solve the problem.
Summary: 1- updateMaterials does not work. 2-UpdateMaterialStage error prevents the interpreter keep going. 3-updateParameter does not work, but, the analysis is going on.
All these 3 commands are useful and should be fixed to be working in OpenSeesSP, anyhow, among these, UpdateMaterialStage error is crucial and with this error the geotechnical simulations can't be done with OpenSeesSP.

Dear Frank, Please solve this issue asp.

[sreyadhar]

Can I UpdateMaterialStage from -0 to -1 for PIMY material in OpesSeesSP? As when I'm updating in conventional OpenSees.exe its working, but in OpenSeesSP.exe its not!

[fmk]

is it a warning message or does it just not work. if you can repeat it for a very small example (and i mean small 6 ele or less)