Identification of Unbalance Voltage in Three Phase Induction Motor Based on Vibration Signature Using Neural Networks

Abstract

Supply unbalanced voltages on induction motors can be fatal if not handled immediately. Early identification is needed to reduce losses. This paper proposes a method for detecting an unbalanced supply voltage on a three phase induction motor. By using Back Propagation Neural Network as a classifier, the distortion of voltage imbalance in the induction motor can be detected based on the vibration. The vibration signal is recorded using an accelerometer 3 axis. The recorded vibration signal is then processed through several stages. The first phase of the signal is decomposed using a 3-level wavelet to obtain an approximation signal and a detailed signal. The next stage transforms the first detail signal into a signal with the frequency domain using FFT. The next stage is to calculate features based on level 3 detail signal and FFT signal. The value of the calculating features is then extracted using Principal Component Analysis (PCA). The extraction results of this feature are further classified using BPNN to identify the voltage unbalance fault. Using the BPNN architecture with 3 hidden layers and 75 neurons, the recognition rate of error identification is 78.77% in MSE 6.1x10-5.