Robust sensor fault estimation for induction motors via augmented observer and GA optimisation technique

Induction motors have been extensively employed in industrial automation systems owing to their inexpensiveness and ruggedness. Current sensors of induction machines would have faults or malfunctions due to the age, which may lead to wrong commands of the controller, causing system performance degradation and even dangerous situations. Therefore, it is motivated to detect the current sensor faults at the early stage so that necessary actions can be taken to avoid further damage of the induction machines and serious situations. In this study, an augmented observer is designed to simultaneously estimate system states, and current sensor faults. In order to attenuate the effects from the modelling error and environment disturbances/noises, a genetic algorithm is employed to design observer gain by minimizing the estimation error against modelling errors and environmental disturbances/noises. The real-data of the induction motor collected by experiment is utilized to validate the proposed methods, which has demonstrated the efficiency of the proposed sensor fault diagnosis approaches. The proposed methods have great potential to improve the reliability of the real-time operation of the induction motor drive systems.