Skip to content

Latest commit

 

History

History
53 lines (46 loc) · 3.04 KB

TODO.md

File metadata and controls

53 lines (46 loc) · 3.04 KB
Raw

Things to be done

  • Sort the content of READMEs
  • Setup a docker image of DeaLAMMPS based on existing dealii existing ones
  • Profile Deal.ii + I/O
  • Write a Python wrapper around the c++ functions found in dealammps.cc or single_md.cc compiled as a shared library
    • the Python wrapper would only execute the higher-level functions such as do_timestep (and the other ones in the HMMProblem class)
    • complete separation of the different models and synchronisation relying on MUSCLE
      • base the work on pre-separated models in version standalone_md
    • see the following slides for tutorials on interface Python and C++
  • Outputs management
    • store for each iteration (timestep), the bare minimum of information to restart the simulation (see restarting process)
    • prepare executable to output visualization files (VTK/XML) from restart checkpoints, and lammps scripts to compute time averaged variables (either global or local)
    • build database (SQL, PostgreSQL, ...) from checkpoints or visualization files
  • Improve restarting process
    • serialize global simulation state: continuum (triangulation, nodes data, cell data), atomistic (atoms position, topology)
    • finite element cell number independence (location-based? else?)
  • Improve homogenization procedure
    • enhanced stress homogenization, stiffness from fluctuations (Luding, S.)
    • MercuryDPM (Luding, S.), LIMEpy (Leither, K., ARL)
  • Adaptative mesh refinement
    • based on deal.ii capabilities
    • transfer mother cell features (strain, atomic model) to child cells
    • handle cell renumbering
  • Avoid one fe_problem.h per FE configuration (mesh+BC)
  • Separate strain checking and spline comparison from the FE_Problem class
  • Random combination of MD replicas associated to each FE quadrature point (or at least a different initial velocity)
  • Upscale local information from MD simulation to the FE problem for spatial visualization
    • bond count
    • dissipated energy via thermoset (but hard to compute)
  • Write documentation
  • Pass common features from all forks to master
  • N-scales extension
    • apply periodic boundary conditions on the mesoscale finite element model

Summary of work:

Setting up the Finite Element simulation

Includes solving quasi-static or dynamic equilibrium of continuum mechanics, solve equilibrium incrementally, generate or import a mesh from gmsh, assign heterogenous materials properties.

Constitutive behaviour (strain/stress relation)

Includes linear relation, MD simulation based relation, or statistically infered relation

FE/MD Coupling

Includes passing down a macroscale strain, and transfering up an homogenized stress and/or stiffness tensor

MD jobs scheduler

Includes splitting the processors adequately in between MD jobs, and Pilotjob.

Database of mechanical states

Includes storing stress/strain space trajectories

Error quantification

Includes investigating the error at each scale and the propagation between the two.