Diagnosis of Electric Motor Using Acoustic noise Based on CNN

Abstract

It is very important to diagnose the mechanical failure of the motor in the industry. The conventional method is difficult to include both the various motors and the driving environment. In order to solve these problems, researches have been actively conducted to apply deep learning to fault diagnosis. Many studies focus on vibration data. Noise data can be collected easily and inexpensively compared to vibration data. In this paper, vibration data and performance were compared using noise data as training data from the diagnosis of motor failure. In the first experiment, vibration and noise data were collected in the same experimental environment. The collected data were compared through the same algorithm. In the second experiment, we compared the performance of the three deep learning models. As a result, both the noise and vibration data showed high accuracy and sufficient fault diagnosis was possible considering the vulnerability of the disturbance.