Algoritmo de detección de fallas para motores trifásicos tipo jaula de ardilla basado en el análisis de componente principales

Abstract

This research aims to detect faults in three-phase squirrel cage motors using a fault detection algorithm based on Principal Component Analysis in Matlab. The method is based on forced training through normalized historical data of all measurable or recorded variables in a squirrel cage induction motor. The study was performed by PLECS-Matlab co-simulation, establishing control limits through the ratio of the variances of the principal components and Hotelling’s T² distribution, proposing an ellipsoid threshold separating the normal operating regime from the fault regime. For data logging in fault operation, a fault injection system was developed on the electrical diagram of the squirrel cage induction motor model, proposing three characteristics faults: eccentricity, stator harmonics, and rotor bar breakage. The algorithm was validated with a 10% and 90% failure rate, managing to detect 100% of failures from its first rates. Furthermore, the same results were obtained with measured in the field.