Store model "parameters" in a dictionary in Project object

[ma-sadeghi]

There are two types of model parameters: those that could have a spatial distribution, and those that are constant across the entire domain. For instance, the "exchange current density", j0, is typically constant, whereas "temperature" (in a mass transfer simulation) could vary pore by pore.

Normally, we store the dictionary key of that parameter in the pore-scale model, so the values can be dynamically changed without the need to re-store the values in the model. Ideally, we don't want to do that for constant parameters, as they take up Np by 1 ndarray whereas they can simply be stored as float.

Currently, we pass the exact value of that parameter to the pore-scale model, which will then be hard-coded in the model. So, every time we want to "change" that parameter, let's say for a parametric study for instance, we need to re-instantiate the model with the new value, which is not convenient.

Instead, we can simply store such parameters in the Project object, and pass a reference to the pore-scale model. This makes it much easier to keep track of parameters and is less error-prone.

[jgostick]

I see a lot of value in this, since it is a nice way to address 'constants' as well, which was considered in #717, but ignored due to the challenges. We'd need to think about the implementation a bit though.

[jgostick]

You could actually do this with the code as it is now, just write a new model that accepts "parameters", and then looks up the values from the Project, assuming you've manually added the 'parameters' attribute as a dictionary. A sample model would look like:

proj.parameters = {}  # Add your own parameters attr
proj.parameters['exchange_current_density'] = 1e-6  # Populate it

...

# Define custom model
def model(target, ... , i_0='exchange_current_density'):
    ...
    val = a*b*target.project.parameters[i_0]  # Use param in model
    return val

We could / should officially support this by adding an empty params dict to each project. We could then include this dict when saving / loading projects from file.