Error minimization of the back-propagation algorithm applied to a motor maintenance problem

The aim of this paper is to treat the preventive maintenance problem in squirrel-cage induction motors. Neural networks are applied to the problem taking advantage of their well-known classification and pattern recognition abilities. The focus of the study is on the learning process, which is by the back-propagation algorithm, discussing techniques that may improve its convergence, reducing oscillations and the CPU time required. Computational results seem to indicate that these techniques are quite effective, and promising in this application field.