Multiclassificação de falhas em motores de indução trifásicos utilizando um único transformador de corrente e redes neurais artificiais

Abstract

The element most used to convert electrical energy into mechanical is the phase induction motor, which is indispensable in industrial production processes. This equipment is constantly subject of research to identify defects in order to reduce maintenance costs, minimize unscheduled process stoppages, reduce costs and mainly increase availability. Hence, This work proposes the study, development and implementation of a system for detecting and classify stator short-circuit faults, broken rotor bars and bearing faults in induction motors by monitoring the stator current signals in the time domain by using a current transformer. More specifically, this work consider the usage of Artificial Neural Networks of Mult Layer Perceptron type, using as the inputs the current signals measured by the current transformer, to proper identify and classify induction motor faults. The database used for the development has been obtained through the experiments performed with motors of 0,736 kW and 0,736 kW on a test bench. The system is validated with both machines working with load torque variation and voltage unbalance. Still, the proposal is embedded in dedicated hardware based on a digital signals processor and experimental tests are performed with the system running in real time.