Vibro-Acoustic Signal Based Fault Diagnosis of Electric Motor Using Artificial Neural Network

Abstract

This work presents a model for fault detection of an electrical motor using vibration and noise signals. Most of the motor faults generate specific patterns in the motor noise and vibration that can be captured and analyzed for diagnosis. Early detection of motor faults will save the motor from sequent deteriorations into a lot of severe conditions, and thus can save lot of maintenance costs. An accelerometer was used to capture decently accurate information of the motor-vibration and Microphone was used to capture decently accurate information of the noise. Features were extracted in time and frequency domain using which an Artificial Neural Network (ANN) called Multi-Layer Perceptron (MLP) was trained to learn different motor conditions such as healthy and faulty. This study shows that mistreatment easy options and ANN structure will effectively and with efficiency classify differing types of motor faults. The use of low-cost sensors has made this method very attractive to wide range of applications where a cost-effective solution is desired.