Development and Application of a Method for Real Time Motor Fault Detection
Byung Gun Joung, Wo Jae Lee, Aihua Huang, John W. Sutherland
Abstract
Predictive maintenance (PdM) has been widely used in manufacturing to reduce maintenance cost and unexpected downtime. A common element within manufacturing equipment/machines is a motor. This paper aims to detect motor faults by collecting and analyzing vibration data with wireless sensors. A cloud-based motor condition monitoring system is also built to detect motor faults by analyzing the data. An Artificial Intelligence (AI) model is trained using the collected vibration data, and principal component analysis (PCA) is utilized to detect abnormal behaviors of the motor. Hostelling’s T2 statistics and squared prediction error (SPE) statistics are then applied to clarify criterions for abnormal operations of the motor.