This is a small repository with some examples of the
incidence_analysis contributed package in Pyomo.
The incidence_analysis code can be found
here.
incidence_analysis implements (using NetworkX wherever possible)
three algorithms that are useful for debugging equation-oriented
chemical process models:
- Maximum matching
- Block triangularization
- Dulmage-Mendelsohn partition
This repo has three main parts:
- "example1", an example using the Dulmage-Mendelsohn partition to debug degrees of freedom in a chemical looping model from IDAES
- "example2", an example using a block triangularization (strongly connected components) to solve a subsystem of the same chemical looping model for initialization
- "tutorial", a walkthrough of the three graph algorithms implemented applied to a small system representing the thermodynamics of a porous solid particle
These examples and tutorial depend on the following Python packages:
idaes-pse, pyomo, scipy, networkx, matplotlib.
I am not 100% sure this is an exhaustive list of dependencies.
These packages (and any others I've forgotten) should be easily
installable with pip.
The tutorial also relies on the PyNumero ASL library, for getting Jacobians,
which can be obtained via pyomo build-extensions.
To run these examples and tutorial, you need to download the code in this repository, probably by running:
git clone https://github.com/robbybp/incidence_examples.git
You then need to "install" this code by running:
python setup.py develop
You can check that you have the required dependencies and have
"installed" this code properly by running pytest (this may require
a pip install pytest if you don't have pytest already, although
pytest is not the only test runner you could use here).
The examples and tutorial are provided as scripts as well as Jupyter
notebooks (.ipynb files).
To open a Jupyter notebook (say to follow along during a demonstration),
navigate to the directory of the notebook you want to open and run:
jupyter notebook filename.ipynb