file:///home/matthieu/Documents/dotblocks/dotb_case_study/results/diffusion_2D/T.gif

dotb_case_study

Centralized implementation of time marching methods dedicated to equations of type :

∂y(x, t)/∂t = F (y, x, t)

where y is a tensor field and F a generic function.

Three solvers :

• Euler explicit
• Adams Bashforth
• Crank Nicolson

have been implemented to run on three study cases :

• rabbit population dynamics
• ballistic equation
• 2D diffusion problem

Installation :

Pull the git repo : https://github.com/Matthieudese78/dotb_case_study.git

In the root directory (dotb_case_study) do :

pip install .

Running input.yaml scripts

Three defaults input scripts are given in the root directory. Run them with :

python main.py input.yaml

Or as a module from outside the root directory :

first run

python setup.py install

then

python -m dotb_case_study.main input_file.yaml

Three example of configuration .yaml files are given in the root repository. Each one can be run with the three solvers.

In the absence of an input file as argument, a default 2D diffusion case will be launched.

Non-declarative usage :

Alternatively, a demo script "demo_diffusion.py" is available in the root directory for non-declarative data setting.

Results

Figures are saved in the save_dir directory given in the input.yaml.

Features

• A refined mesh class Mesh enables to refine the mesh locally at the borders and corners of the domain to avoid convergence difficulties.

• For the 2D diffusion case, it is possible to smoothly interpolate the boundary conditions at the corners. The parameter : 'interpolation_coeff' represents the percentage of the edges length on which the interpolation takes place.

Architecture :

Three classes of solvers (module solver.py)

• EulerExplicit
• AdamsBashforth
• CrankNicolson

They take one of the three following equation right hand side classes as argument (module second_member.py):

• Ballistic
• Rabbit
• Diffusion

They are assembled in the build_model.py module in the function create_model. The mapping Solver - Study Case happens through a dictionary defined in the same module.

Add a new solver / study case :

This way, adding a new study case consists in two steps :

• crate a class in second_member.py that contains three methods :
  • the right hand side of the equation F
  • the newton raphson for the Crank Nicolson (form of the residual derivative etc..)
  • the boundary conditions function (identity if none)
• Add the class name in the solver-to-case mapping dictionary in build_model.py

Conversely, adding a new solver consists in two steps :

• Code the general structure of the solver by adding a class in the solver.py module,
• Add the class name in the solver-to-case mapping dictionary in build_model.py

For larger solver or case study catalogues, it could be considered to create parent classes Solver and CaseStudy to fix and control the structure of the children classes.

Tests

To do...

Documentation

To do...

Demo jupyter notebook

To do...