In the industrial technological field, running equipment or processes usually is monitored through automatic diagnosis systems. Within several Technologies for implementing such systems, the artificial neuronal networks are the most successful and widely spread. The data signals coming from the equipments or processes under supervision are interpreted by the neuronal networks so as to diagnose the presence of any fault. In this work three models of artificial neural networks and two methods of training are analyzed so as to establish, based on real experiences, the best combination of the neuronal model and the training method for recognizing in an efficient way the ambiguous patterns of faults.
Presentado en el I Workshop Procesamiento de señales y Sistemas de Tiempo Real (WPSTR)
Red de Universidades con Carreras en Informática (RedUNCI)