Department of Mathematics Applications and Methods for Artificial Intelligence, Faculty of Applied Mathematics, Silesian University of Technology, Gliwice, Poland
The magnetic spring placed on vibration generator consists of top and bottom fixed magnets with a third magnet levitated between them. Nonlinear dynamic analysis appears in the magnetic flux density of the magnetic spring activated by the vibration generator. This paper is focused on fast data-driven and accurate model for the prediction of magnetic flux density using a deep neural network (DNN) in the form of a Long Short-Term Memory (LSTM). As the input and training data for LSTM are used: supply voltage of the vibration generator, its frequency and measured magnetic flux density. The magnetic flux density measurements of the magnetic spring have been recorded by three Hall effect sensors. The prediction of magnetic flux density in magnetic spring has given accurate results and good applicability for better characterization of the device in energy harvesting system.
