Model modularity

When using the SUND toolbox, it is possible to combine models into a single simulation. Furthermore, the outputs from some models can be used as input in other models. These are defined in the INPUTS header and in the OUTPUTS header.

In the example below, the model m1 has an internal variable named plasma_ins which it provides as an output insulin_plasma to a "shared pool" of variables that other models can use. The model m2 needs ins as an input, which it can take from the "shared pool". Below is a graphical example:

graph LR
    %% Add text above the boxes
    subgraph "Model 1"
        m1["plasma_ins"]
    end
    subgraph "Shared variables"
        shared["insulin_plasma"]
    end
    subgraph "Model 2"
        m2["ins"]
    end

    %% Define the variable renaming flow
    m1 -->|plasma_ins=insulin_plasma| shared
    shared -->|insulin_plasma=ins| m2

Below is an example from the corresponding model files:

########## NAME
m1
...
########## OUTPUTS
insulin_plasma = plasma_ins

########## NAME
m2
...
########## INPUTS
ins = insulin_plasma

Now, if the models are put together in a simulation, the plasma insulin provided by m1 will be used by m2 automatically:

m1 = sund.load_model('m1')
m2 = sund.load_model('m2')

sim = sund.Simulation(time_unit = 'm', models = [m1, m2])

Using default values for inputs:

It is also possible to define default values that the model should take if the input is not supplied from another model or activity. This is done by using @ value. So, e.g. to set ins to 0 if no input is given for m2:

########## NAME
m2
...
########## INPUTS
ins = insulin_plasma @ 0

Inputs from other sources than models

In many cases, the inputs needed to model might not be available from other models. For example, if simulating an experiment with time-dependent stimulations. To avoid having to create an extra model for these stimulation inputs, you can instead use an activity object which only provides static time-dependent outputs to the shared pool.