Software tutorial/Integration of ODEs

From Process Model Formulation and Solution: 3E4
Jump to navigation Jump to search

In our course we have learned several ways of integrating a single equation \(\displaystyle \frac{dy(t)}{dt} = f(t, y) \) with a given initial condition \(y(t=0)=y_0\); and we have shown that a system of \(n\) first order ODE's can be integrated: \[ \displaystyle \frac{d{\bf y}(t)}{dt} = {\bf f}(t, {\bf y}) \]given a set of \(n\) initial conditions \({\bf y}(t=0) = {\bf y}_0\), i.e. \( {\bf y}_0\) is an \(n \times 1\) vector. The course notes covered Euler's method, Heun's method and Runge-Kutta methods.

However, we only coded Euler's method (because it was simple!), but not the others. These other methods have been (reliably) coded in software packages and sophisticated error correction tools built into their algorithms. You should always use these toolbox functions to integration your differential equation models. In this section we will show how to use MATLAB and Python's built-in functions to integrate:

  • a single differential equation
  • a system of differential and algebraic equations.

We will only look at initial value problems (IVPs) in this tutorial.

MATLAB

MATLAB's has several ODE solvers available. You can read up more about the implementation details in this technical document.

Python

Like MATLAB, several integrators are available in Python. The integrator I will use in this tutorial is one of the most recent additions to SciPy and is the VODE integrator developed at Lawrence Livermore National Laboratories in 1988. It is a good general-purpose solver for both stiff and non-stiff systems.

MATLAB Python