Actuator fault and abnormal operation diagnoses for auto-balancing two-wheeled cart control

This paper employs a PI observer to detect abnormal information in an auto-balancing two-wheeled cart caused by actuator faults and steering load torques. A fault diagnosis scheme which contains the strategies of both fault-detection and fault-evaluation is proposed in this paper. Analysis and design of the PI observer are then discussed in detail. A statistical threshold that has the benefits of improving decision-making reliability is investigated for diagnosing a possible abnormal operation and/or a serious actuator malfunction. Illustrated examples are given for evaluating the performance and feasibility of the presented fault diagnosis strategy.