Modelling a Magnetic Levitation System in MATLAB, Design & Development of Mag Vehicle, Girder Bridge
Maglev Modeling
Using NARX (Nonlinear AutoRegressive with eXternal input) neural network that can model a magnet levitation dynamical system.
The Problem: Model a Magnetic Levitation System
Here, we attempt to build a neural network that can predict the dynamic behavior of a magnet levitated using a control current.
The system is characterized by the magnet's position and a control current, both of which determine where the magnet will be an instant later.
This is an example of a time series problem, where past values of a feedback time series (the magnet position) and an external input series (the control current) are used to predict future values of the feedback series.
Why Neural Networks?
Neural networks are very good at time series problems. A neural network with enough elements (called neurons) can model dynamic systems with arbitrary accuracy. They are particularly well suited for addressing non-linear dynamic problems. Neural networks are a good candidate for solving this problem.
The network will be designed by using recordings of an actual levitated magnet's position responding to a control current.