The OpenSees Community

mhscott wrote: ↑Fri Nov 27, 2020 7:21 am Thanks! I ran into similar issues recently with the OpenSees.exe compilation. I recorded the process, but haven't edited the video yet.

But here's a video on compiling openseespy.pyd https://youtu.be/l5-vJDZR_hA

Thanks Michael! I will look into the video.

While I find you here conveniently, can I ask these two questions ?:

1. Has there been any attempt to generate static or dynamics libraries of OpenSees classes (maybe without TCL + OpenSees Command Line classes)? It seems like I will be heading to C++ environment for programming rather than TCL or Python.

2. Is there a way to get rid of the spams in the forum? Looks ugly and not encouraging to new users.

Thanks again.

Hello,

There are couple of things I would like to get opinion on and provide my experience on while compiling tcl OPS with Windows 10 and Visual Studio 2019. This is as of today, which I cloned the OPS repo from Github just couple of days ago. So, some of the comments here may not be valid for the upcoming changes in the Github repo.

1. Is there a specific or suggested order of builds of the libraries and OpenSees.exe?

Based on the error I encountered, I ended up building the projects according to the following order:

- actor
- material
- element
- OpenSees

When I tried to build OpenSees directly, I found out that errors occur mostly in the above projects, so instead of directly going for OpenSees, I followed the above order of builds.

2. The _DLL issue similar to this one . What is the Solution Configuration option is used while compiling: Debug, Debug.DLL Build, Build.DLL? Do these options have smtg to do with the Runtime Library option in VC2019 (i.e., /MT, /MTd, /MD, /MDd). I understand that some variables related with Rayleigh Damping are defined based on the value of _DLL, which prevents successful build if correct options are not selected. I also 'more or less' understand the Run Time Lib options; how DLLs are called dynamically or integrated into the build statically depending on our selection. My question is what is the usual/suggested way of selecting these?

The way I solved the issue I had, I have built -actor, - material, - element with "Build" Solution Configuration. For OpenSees, both Build and Debug worked but Debug.DLL and Build.DLL configurations did not work. But the funny thing, the repo came with the Debug.DLL is selected by default. Also, OpenSees Project Runtime Lib option came with /MTd is selected but other projects have /MD selected. Is this normal?

3. There is an issue with the relative path of include folders if the build is done directly running VC from the search or start of Windows. Almost all the folder paths are defined with '..\..\..\' (except tcl folder, which is defined with the absolute path of C:\Program Files\tcl) This requires the starting folder of VC be in one of the project folders. So, what I did was I created a shortcut of VC2019, where I erased the "Start in" folder and placed this shortcut into the OpenSees project folder. This helped VC2019 find the include paths correctly. There were also two header files that were not found. I included the "..\..\..\SRC\element\PFEMElement" and 1..\..\..\OTHER\CSPARSE" folders to include those header files.

Of course I could have modified the include paths manually or ask VC2019 shortcut to start in one of the project folders, but these did not seem to be useful solutions. For tcl, I changed the path to C:\tcl, which I have on my computer.

What is the usual/proper way of starting VC2019 to get the paths correctly?

4. The ifconsol.lib files. I downloaded these files from here and placed under "..\OpenSees\Win64\lib" folder. Is this usual/suggested way of keeping these libs?

BTW, I got the OpenSees.exe finally Now I am gonna head for the OpenSeesPy and probably go for Code::Blocks with msys2 + mingWG.

We have identified the error and verified it with the frame example (thanks to Ufuk Yazgan). The error was the assignment of translational mass value to the rotational degree of freedom by mistake. If the model is a 2D model where 3 DOFs are defined a given point, then the a lumped mass should include two translational mass and one rotational mass.

We have posted the frame example to the same link: https://1drv.ms/f/s!Aqx30_D8oGfG5GJJerOAn4-i8TXi
Interested readers may have a look the files. We were able to get almost same results with SAP and OpenSees for a zero-damping time-history analysis. We have verified the stiffness matrices as well.

It is unfortunate that we cannot extract the mass matrix easily, while stiffness matrix can be extracted using PrintA command.

We have been encountering an issue with the mass and probably with the mass matrix.

We have considered a simple weightless linear cantilever column, where a lumped mass is attached at the top, as a MWE.

We hand-calculate the period of the example and show that it is same as the period obtained from a SAP2000 analysis. However, OpenSees give a different value:
Hand Calculated: 0.46171793 sec
SAP2000: 0.461718 sec
OpenSees: 0.51982 sec

The OpenSees and SAP files are located at https://1drv.ms/f/s!Aqx30_D8oGfG5GJJerOAn4-i8TXi (hope the system allows access)

We are guessing that the stiffness matrix is estimated correctly, but the mass (or the mass matrix) is multiplied with a factor of approximately 1.13 in OpenSees.

We have not checked the stiffness for the cantilever example, but we have checked the stiffness matrix of a single bay/single story frame example in OpenSees and found that the stiffness matrix is identical to that of obtained from SAP2000. Again, for the frame example, we can only get the same periods if the mass values are scaled down with the factor of approximately 1.13. We can post the frame example as well if needed.

So, did anybody verified the mass matrix in OpenSees, at least for a linear problem?