usermat_2scale_lsdyna

Basics to implement user-defined material routines in LS-Dyna with Python and C++ and realize two scale simulation schemes in LS-DYNA.

External dependencies

Dependencies below are included in the external_packages directory. They may be included as submodules but for the sake of reproducablity a frozen version of their corresponding repositories is included here.

• ttb: Tensor Toolbox for Modern Fortran (ttb), hosted at https://github.com/adtzlr/ttb

• forpy: A library for Fortran-Python interoperability, hosted at https://github.com/ylikx/forpy

• ezh5: Easy HDF5 C++ Library, hosted at https://github.com/mileschen360/ezh5

  Test cases of these packages are included in external_packages/test_*.sh

Licensing

See the license file for the project license and the licenses of the included dependencies.

Citations

Shadi Alameddin, Felix Fritzen. LS-DYNA two-scale homogenization extension. Version 1.0.0 (2021).

@software{alameddin2021,
  author       = {Shadi Alameddin, Felix Fritzen},
  title        = {LS-DYNA two-scale homogenization extension},
  month        = Aug,
  year         = 2021,
  version      = {v1.0.0},
  url          = {https://gitlab.com/shadialameddin/dae_umat_2scale_lsdyna}
}

Compiling the base version on Linux (Ubuntu 20.04.2 LTS)

• Start by obtaining ls-dyna_smp_d_R12_0_0_x64_redhat65_ifort160.tgz usermat package/object version of LS-DYNA from your local distributor of LS-DYNA. The directory where ls-dyna_smp_d_R12_0_0_x64_redhat65_ifort160.tgz is placed will be referred to as dyna_umat_directory

• Install Intel® C and Fortran compilers that are included in the Intel® oneAPI HPC Toolkit which is available for free. Examples below have been compiled using ifort, icc version 2021.1 and Python 3.7.9 & 3.8.5. In addition, some tests use GNU Fortran 9.3.0.

• Obtain the new interface files from GitLab and position it in dyna_umat_directory via

cd dyna_umat_directory
git clone git@gitlab.com:shadialameddin/dae_umat_2scale_lsdyna.git lsdyna_added_files_ref

• Using a terminal, navigate to lsdyna_added_files_ref and create a symbolic link to the main setup file via

cd lsdyna_added_files_ref
./create_links.sh
cd ..

• Edit set_env.sh to point to the correct intel_dir and add python examples within lsdyna_added_files_ref to PYTHONPATH.

• Edit setup_dyna.sh to point to the absolute path of dae_umat_2scale_lsdyna.

• Now everything is in place to compile the usermat package, to do that, from dyna_umat_directory run ./setup_dyna.sh. The new executable will be placed in dyna_umat_directory/lsdyna_object_version/lsdynaumat

  Note

  In case of issues while running from a new shell, setting the correct environmental variables via running source set_env should fix the problem.

Test cases:

• External packages: test cases of these packages are included in external_packages/test_*.sh
  • test_ttb.sh
  • test_forpy.sh
  • test_ezh5.sh
• Mixed language programming
  • test_call_cpp.sh: compiles and runs a Fortran function that calls a C++ one
  • test_call_py.sh: compiles and runs a Fortran function that calls a Python one via C++

Examples:

• examples/two_scale/analytical_mat_parameter

  Temperature dependent material parameters are considered here and given as lambda functions in material_parameters.py.

  To get an idea about the implementation check:

  • umat_elastic_44_14.F90
  • material_parameters.py
  • umat.py

• examples/two_scale/homogeneous_single_track

  Here, an RVE effective response under different load temperatures is assumed to exist and stored in a tabulated format in an HDF5 file. Linear interpolation is used to evaluate effective properties at current temperature given the stored response at one higher and one lower temperatures.

  To get an idea about the implementation check:

  • umat_elastic_44_14.F90
  • umat.py
  • rve_elastic.py

• Discontinued

  • examples/two_scale/3d_rve
  • examples/two_scale/2d_rve

  Here, a thermo-mechanical response of a representative volume element is computed using LS-DYNA.