Fault classification using support vectors for unmanned helicopters

Abstract

This paper develops a fault classification algorithm using support vectors for monitoring operation anomalies and early fault detection in unmanned helicopters. In general, a fault classifier identifies the operating state of the system which can be further monitored by developing a necessary control action. Hence, to achieve the fault detection, the data between normal operating condition and faulty conditions is monitored. In this paper, a two-class classification method is developed using support vector data description for identifying the motor faults in unmanned helicopter system. The algorithm is developed by observing the data of pitch and yaw motors of a 2DoF helicopter. Further, the action of SVDDs during unknown faults is observed by developing a residual generator which improves the functioning of classifier. The results depicted 98.2% training accuracy and efficient results when tested with a faulty condition.