Detection of Rotor Eccentricity Faults in a Closed-Loop Drive-Connected Induction Motor Using an Artificial Neural Network

A new method for the detection of rotor eccentricity faults in a closed-loop drive-connected induction motor is reported in this paper. Unlike a line-fed electric motor, the eccentricity-related fault signals exist in the current as well as the voltage of a drive-connected motor. Meanwhile, since the speed and therefore the mechanical load can change widely in variable speed applications, the amplitudes of the fault signals will vary accordingly. An artificial neural network is used in the detection to learn the complex relationship between the eccentricity-related harmonic amplitudes and the operating conditions. The neural network can estimate a threshold corresponding to an operating condition, which can then be used to predict the motor condition. The neural network is trained and tested with data collected on drive-connected 4-pole, 7.5 Hp, three-phase induction motors. The experimental results validate that the detection method is feasible over the whole range of operating conditions of the experimental motors.