The OpenSees Community

Hello,

I think that there is an error in the main function of the quickmain project.

In version 2.2.1 there is reported one change in code:

2. redo of how tags were cimplemented for MP and SP constraints

I think that since that version, the SP constrains definition in the main function should be:

// create the single-point constraint objects using constructor:
// and then add them to the domain

// SP_Constraint::SP_Constraint(int node, int ndof, int clasTag)

SP_Constraint *sp1 = new SP_Constraint(1, 0, 0.0);
SP_Constraint *sp2 = new SP_Constraint(1, 1, 0.0);
SP_Constraint *sp3 = new SP_Constraint(2, 0, 0.0);
SP_Constraint *sp4 = new SP_Constraint(2, 1, 0.0);
SP_Constraint *sp5 = new SP_Constraint(3, 0, 0.0);
SP_Constraint *sp6 = new SP_Constraint(3, 1, 0.0);

/*

// SP_Constraint(tag, nodeTag, dofID, value)

SP_Constraint *sp1 = new SP_Constraint(1, 1, 0, 0.0);
SP_Constraint *sp2 = new SP_Constraint(2, 1, 1, 0.0);
SP_Constraint *sp3 = new SP_Constraint(3, 2, 0, 0.0);
SP_Constraint *sp4 = new SP_Constraint(4, 2, 1, 0.0);
SP_Constraint *sp5 = new SP_Constraint(5, 3, 0, 0.0);
SP_Constraint *sp6 = new SP_Constraint(6, 3, 1, 0.0);

*/

Regards,

Hi,

I have experienced some issues during downloading and compiling Opensees V2.3.2.

1. The link in the developers area os for the 2.3.0 version. At least this is the name of the zip file.
2. Then I have downloaded the 2.3.2 version frm SVN. It works OK.
3. Using Visual C++ express I tried to open the Opensees2010.sln but it says that projects . vcxproj are missing.
4. It is possible to open Opensees2005.sln and compile all projects without any problem, excepting Quicmain.
5. As it was reported in the forum, some errors related with _ops_getdoubleinput appeared.
6. Adding win32functions.cpp to the project and the related references to header files from ..\OpenSees\DEVELOPER\core the project compiles successfully.
7. Running the quickmain project an error appears during the execution. After some analysis it seems that the problem is caused due to an obsolete definition of the SP_Constraints in main.cpp. So modifying the following lines the project can be compiled and used OK
// create the single-point constraint objects using constructor:
// SP_Constraint(tag, nodeTag, dofID, value)
// and then add them to the domain

//NEW constructor
//SP_Constraint::SP_Constraint(int node, int ndof, double value, bool ISconstant)
/*
SP_Constraint *sp1 = new SP_Constraint(1, 1, 0, 0.0);
SP_Constraint *sp2 = new SP_Constraint(2, 1, 1, 0.0);
SP_Constraint *sp3 = new SP_Constraint(3, 2, 0, 0.0);
SP_Constraint *sp4 = new SP_Constraint(4, 2, 1, 0.0);
SP_Constraint *sp5 = new SP_Constraint(5, 3, 0, 0.0);
SP_Constraint *sp6 = new SP_Constraint(6, 3, 1, 0.0);
*/
SP_Constraint *sp1 = new SP_Constraint( 1, 0, 0.0,false);
SP_Constraint *sp2 = new SP_Constraint( 1, 1, 0.0,false);
SP_Constraint *sp3 = new SP_Constraint( 2, 0, 0.0,false);
SP_Constraint *sp4 = new SP_Constraint( 2, 1, 0.0,false);
SP_Constraint *sp5 = new SP_Constraint( 3, 0, 0.0,false);
SP_Constraint *sp6 = new SP_Constraint( 3, 1, 0.0,false);

Hi Aizen,

The element 1 (beam between 1 and 2) is able to rotate free in their axis if you only fix the translations DOF, so this is in my opinion the cause of the error. This is similar to modelizate a simply supported beam fixing the 1, 2 and 3 DOF. Usually this fails. This only happens in your model as the nodes 1,2,3 and 4,5,6 are aligned. If they would make a triangle, I guess it will run properly (I have do not try it).

Depending of your model I would suggest you to use:

equalDOF 1 4 1 2 3 4;

I think that other possibility is to add a Zerolentgh element for this DOF with a small stiffness.

I can not able to answer to your other questions.

Regards,

Hello,

I'm not familiar with Uniformexcitation loads and they do not appear in the post, but I think that there are some wrong things in your code:

Your sample is uniaxial but you are using a 2D model
I think the Y degree it is not fixed (the zerolength is only for X degree).

You could try with
equalDOF 1 2 2
or even:
equalDOF 1 2 2 3

to fix the Y degree of freedom. Another posibility is to include a stiffness for the Y degree. If there are no loads in this direction the displacements and forces should be close to zero.

element zeroLength 1 1 2 -mat 1 1 1 -dir 1 2 3;

Although may be there is not important, you are asking displacements for 6 DOF when the model only has 3

Regards,

Hello,

if there is a non linear problem, I would suggest you to apply the load in several steps changing for example

integrator LoadControl 0.1

analyze 10

The other possibility is that the system it is not stable.

For me the other question is a little bit surprissing as it should stop the analysis as the element is not created. I would check carefully the model and results as also the warning (maximum 1e-5) it is not related with the test you have used:

test NormDispIncr 1. e-4 25 5;

Hope this helps

Hello,

In my opinion the most required feature is to have a pre-processor and a post-processor. In order to integrate the OPS framework with geometric packages in an easier way, I think that some features are needed:

Possibility of obtaining Rigid body shapes in eigen analysis. Now it is quite difficult to check models with joints (zerolengths elements) when they are created with manual routines and checking these shapes would be helpfull to see which elements are not connected.

Add some additional examples for developers as the example1 of the EXAMPLES folder. Perhaps the typical TCL examples could be translated. I think they are very usefull for non C++ skilled users.

Add some solutions and projects. I would suggest solutions for Windows 64 and a project similar to quickmain in order to create a DLL that could be called from the pre-processor.

I think that these features would help to new users to understand the source code of OPS.

Regards

Hi Frank,

I can imagine from your answer that you are preparing a new version for the Parallel Workshop.

1. Is this new feature of adding external solvers related with Parallel computing or is an indepentent one.?
2. Will you include also additional solvers or we must develop them?.
3. I'm specially would know if this improvement could be used to perform dynamic analysis on free structures, it means, rigid body modes.


I think that it is an interesting option to check complex models. In my opinion, actually one typical error is not to fix properly the model and then it is quite difficult to check the script to see what it is lack.

Regards,

Hi,

I would recommend you to have a look to OpenCascade:
www.opencascade.org

I think it is quite similar to VTK, it is free (even for commercial uses) and it also supports TCL. Of course there are also some "minor problems" as there is not any official support.

There is another related code, Salome, focused on scientific applications (including meshing) although in this case Windows it is not supported.

Regards,

Hello Silvia,

Are the presentations of this workshop available for downloading?

Regards,

Carlos

Hello,

I have installed MPICH2 on my Core duo labtop running with Windows XP and I also have been checking the presentations of the workshops. I have some doubts about how to run a script with parallel computing.

I'm able to run Opensees.exe an existing script on both processors using:
mpiexec -np 2 opensees.exe example.tcl

I can see on the Windows taks that the two proccessors of the computer are running (at 100%).
If the command is:
mpiexec -np 1 opensees.exe example.tcl

Only 1 processors runs. I must say that both runs spent about 1 minute to complete the analysis.

The command:
mpiexec -np 1 openseesSP.exe example.tcl

runs on the computer, BUT
mpiexec -np 2 openseesSP.exe example.tcl

Fails to do it.

In the last case I have change the sentence system UmfPack with system Mumps -ICNTL14 100 as Frank recommends on this forum but still there are some errors.

The first error is something like:

FEM_ObjectbrokerAllClasses:getCrdTransf3d -- No CrdTransf3d type exist for class tag 6


Questions:

1. There are three (not two) executables availables on the web page:
OpenseesSP.exe
OpenseesMP.exe
I understand the difference between the SP and MP, but what is inside opensees.exe?

2. Do you think that the Parallel version is worthy for running on Labtop computers with two processors, usually not very big models.

3. I have read on one presentation of the workshop that there are some problems with the eigen solver on the OpenseesSP.exe version. Do these problems appear also on the OpenseesMP version?.

4. By the way, Can we expect to have an eigen solver that solve indeterminated systems (rigid body movement). This could help to check models and see if there is one part that is not fixed or linked to others.

5. Are you going to prepare an "example1\main.cpp" for parallel computing ?

Thanks to the Openses team because this seems a big step in the development of Opensees and the "2.0 version" reflects this.

Hello Silvia,

I hope you have enjoyed your long holidays, in the european style.

Could you please tell me if the presentations of the last workshop will be available?

Regards,

Hello,

You can also download the file for the Unix version (http://opensees.berkeley.edu/OpenSees/c ... 7.4.tar.gz). I did it and I descompresed it with Winrar.

As it is stated on point Nº. 4 of the source code distribution chapter, it contains the same code that the Windows version.

(The source code is distributed as either a Unix-style tar archive compressed by GNU zip (gz) or as a PC-style ZIP archive. (The only difference is the type of archive; the same source code files are inside the archive. )

I have compiled the source code with Visual Studio 2003 without any problem.

Regards,

Hello,

I have read on the change log of the V1.7.4 version that has been added some code for building a parallel version.

Unfortunately, I have checked the source code (the link of the windows version is not rigth. I have use the link for the unix version) and I have not found any instructions or solution provided for this type of version.

Have you already check the possibility of creating a parallel version of Opensees for Windows that you indicated on a previous post?. Is there any instruction available for it?

Regards,

Hello,

I have a new laptop. It has an Intel Core 2 CP T7200 @ 2. GHz processor with 1 Gb RAM.

Opensees V1.7.3 runs only at 50% of the capacity of the computer.

I heared and read something something about parallel version of OPS but I'm not sure if this applies to this case.

Could you give me some advices to compile Opensees in this case?.

By the way, Frank, Could you tell us something about the new version of Opensees and what are the points in which you are working now?.

Will it include a new eigen solver capable for zero values, ie structures with no restrictions?

Regards

Hello Frank,

I not really interested on modify the code, only on using it. I have seen that exist the file corotational.lib and that the binary file Opensees.exe works with the Corotational transformation.

I think that my real problem is that I'm not skilled on C++ and I do not know how to compile the example (The same command appears on the file TclGeomTransfCommand.cpp and it runs).

I`m trying to have access during the different steps to the existing forces on the model (in the same way as with the TCL command nodedisp) and modify the model according them. If it is not possible to compile the example I think that I even could modify some of the existing files on the code source to gain access to the Corotational Transformation. But this solution would not be very fancy and will difficult futures updates.


Could you help me please?.

Regards,