Strain Hardening Simulation on GPU

We can run a tensile test simulation of a single-crystal Cu using the following commands (running ExaDiS on GPU). GPU allows us to run the simulation more efficiently and hence reach a greater strain to see the strain-hardening behavior more clearly. Execute the following commands to enter the directory containing the example.

cd ${OPENDIS_DIR}
cd examples/10_strain_hardening/

Before running the simulation, we shall edit the test_strain_hardening_exadis.py file to change the constructor of sim from maxstep=100 to max_step=10000. Execute the following commands to run the simulation.

export OMP_NUM_THREADS=8
python3 test_strain_hardening_exadis.py

Simulation Behavior

The simulation creates a folder called output_fcc_Cu_15um_1e3 to store the results files. On MC3.stanford.edu (gpu-ampere), it takes about 13.6 hours to run 10000 steps of the simulation. The simulation will write a data file to the output folder for every 100 steps. The stress_strain_dens.dat file stores certain essential information of the tensile test — it contains 5 columns corresponding to step, strain, stress (Pa), dislocation density (m^-2) and wall-clock time (sec), respectively.

The final dislocation configuration (config.10000.data) after 10000 steps is shown below.

The predicted stress-strain curve is shown below.

Here is how the total dislocation density changes with strain. The increase of dislocation density (i.e. dislocation multiplication) with strain is a key mechanism for strain-hardening.