Real time health monitoring of industrial machine using multiclass support vector machine

Most of the failures in the industrial systems are due to motor faults which can be catastrophic and cause major downtimes. Hence, continuous health monitoring, precise fault detection and advance failure warning for motors are pivotal and cost-effective. The identification of motor faults requires sophisticated signal processing techniques for quick fault detection and isolation. This paper presents a real time health monitoring technique for induction motor using pattern recognition method. The proposed fault detection and isolation scheme comprises three stages: data acquisition, feature extraction and multiclass support vector machine classifier. This paper investigates single and multiple faults in single-phase induction motor including bearing fault, load fault and their combination. The testbed consists of 1/2 hp, 220V squirrel cage induction motor with load, vibration sensor, current sensor, data acquisition system and controller. Two features standard deviation and average value are computed for each sensor's data. Multiclass support vector machine classifier is implemented using a low-cost arduino controller for fault detection and isolation. The performance analysis of the classifier with real time sensor's data is presented which shows superior capabilities of developed method.