This paper describes the use of neural networks based on self-organising maps in order to diagnose the health conditions of induction motors in trains operating daily services around Madrid, Spain. This kind of neural networks is used for the creation of models able to characterise the normal behaviour of the electrical motors of the train. These models will allow for the on-line detection as soon as possible of any anomaly that could evolve into a failure. The models formulated use non-intrusive measurements taken from different points of the train. They are based on the measurement of electrical currents and axial and radial vibrations on the electrical motor. This is part of an expert system existing at a higher level named the Intelligent System for Predictive Maintenance Applied to Trains (ISMAPT) which monitors and diagnoses some components of the above mentioned trains.
Palabras clave: Electrical train, electrical motor, diagnosis on-line, predictive maintenance, neural network, self-organising map, expert system
3rd IEEE International Symposium on Diagnostics for Electrical Machines Power Electronics and Drives. ISBN 88-9000645-0-1. Record IEEE SDEMPED 2001, Grado, Gorizia (Italia). 01 septiembre 2001
Fecha de publicación: septiembre 2001.
M.A. Sanz-Bobi, J. Besada, R. Palacios, A. Muñoz, R. García-Escudero, M. Pérez, Á.L. Matesanz, Diagnosis of the electrical motors of a train using self-organised maps, 3rd IEEE International Symposium on Diagnostics for Electrical Machines Power Electronics and Drives. ISBN 88-9000645-0-1. Record IEEE SDEMPED 2001. ISBN: 88-9000645-0-1, Grado, Italia, 01-03 Septiembre 2001