Boundary force vector

[ateshian]

Hi Maryrose,

Thanks, I still see distinct values of the traction components on my computer.

I still can't figure out why you are getting different results, so I would suggest you try the following:

(1) Using my test model, please try to apply only one non-zero traction component at a time (i.e., (0.5,0,0), (0,0.25,0) and (0,0,-0.1)). In each case the deformation of the block should be consistent with the applied traction (leaning along x, leaning along y, compressing along z, with magnitudes proportional to the applied traction).

(2) Please repeat these analyses with the FEBio 2.9 executable downloaded from the febio.org website as well as with your compiled version and compare the results.

Please let me know what you find for both (1) and (2). (In my case, the FEBio 2.9.1 executable downloaded from febio.org gives me the expected behavior for (1).) If you notice differences between your compiled version and the downloaded executable, we can focus on your method of compiling the code (unless you made other changes to the code?).

Best,

Gerard

[maryrose]

Hi Gerard,

I'm getting the same issue when I apply apply only one non-zero traction component i.e. for (0.5, 0, 0), my output is (0.5, 0.5, 0.5) and for (0, 0.25, 0) and (0, 0, -0.1), my output is (0, 0, 0).

When I run the same .feb file with the FEBio 2.9.1 executable I get slightly different results (.log and .xplt files) between the FEBio executable case and through the source code. I have no other changes to the code.

Regards,
Maryrose

[ateshian]

Hi Maryrose,

Would you please email me the plotfiles generated when you run these six problems (the three cases using your executable and the same three using the downloaded executable)? My email is ateshian@columbia.edu.

Also, would you please tell me which operating system you use and which compiler?

Thanks,

Gerard

[maryrose]

Hi Gerard,

I've emailed on the plot files and the operating system and compiler information.

Many Thanks,
Maryrose

[ateshian]

Hi Maryrose,

I got the files and I confirmed that the results are different when using your executable versus the one on febio.org, consistent with the differences you've noted in the traction vector.

At this point I have to assume that the issue must be with the compilation of the source code. Since you are using linux, I will have to hand this issue over to Michael Herron. I have emailed him and asked him to review this post. Hopefully Michael can find a reasonable explanation for what's happening.

Best,

Gerard

[maryrose]

Hi Gerard,

Thanks for your help.

Regards,
Maryrose

[mherron]

Maryrose,

Unfortunately, I also can't seem to replicate this problem on my system. I've rebuild FEBio2 with your FETractinLoad.cpp file with both icpc and g++, and I've used both your test files, and Gerard's. I'm not sure what could be causing this problem. I've asked our lead developer for his input on the issue, but while we are waiting for his response I'd like you try something for me.

I've just committed some files to our git repository that should help to automate the build system a bit more by using CMake. Update your local copy, make a new directory in the root of your copy called "cbuild" (this just helps keeps things cleaner), then in the cbuild directory run

ccmake ..

It should hopefully find your libraries automatically and I'm hoping that this build will be free of the problem that you're having. If you're unfamiliar with CMake, let me know and I can give you some more explicit instructions.

Thanks,

Michael Herron

[maryrose]

Hi Michael,

When I run "ccmake .." I get an error when turning on USE_MKL. It says the MKLROOT environment variable is not set. Previously I updated the lnx64d.mk to link to the libraries.

Many Thanks,
Maryrose

[mherron]

Maryrose,

The CMake script relies on the Intel compiler variables being set in your terminal. There should be a script called compilervars.sh somewhere in your intel installation. Mine is located in /opt/intel/bin/compilervars.sh. Before running cmake, you should run

source /opt/intel/bin/compilervars.sh intel64

in your terminal. This will set up all of the necessary environment variables. If you compile often using the intel compiler, I would recommend that you add that line to your .bashrc file so that it will be run every time that you open a terminal.

Also, just so that you are aware, the lnx64.mk makefile in our repository also depends on these environment variables, and so running this script would allow you to use the makefiles in our repository without any manual editing of paths. I thought that our build guide mentioned this, but apparently it's only mentioned at the top of the file build/Makefile.

Let me know if you have any other problems,

Michael Herron

[maryrose]

Thanks Michael, that worked. The traction vector is outputting as expected now.

Many Thanks,
Maryrose