ABSTRACT

Active Magnetic Bearings (AMBs) endure rotating member without any frictional loss. Therefore compared to the journal and hydrodynamic bearings, AMBs have benefits of reduced frictional losses as a result they can support high speed rotor. Stable working operation of AMB is principally depends on the working condition of its position sensors, actuators and controller. Therefore sensors and actuators are the important working elements of AMB system. Fault or failure in any one the sensor or actuator of AMB system can results in undesired rotor dynamics behaviour. Hence to ensure the safe operation and authentic performance of AMB system, fault detection and diagnosis (FDD) of sensors and actuators is very much essential. Principal component analysis (PCA) is a model free and robust statistical technique which can detect and diagnose the faults in engineering systems with high accuracy. Therefore, in the present work, PCA based FDD methodology is employed to detection and diagnosis of AMB sensors and actuators fault. Simulations have been carried out to diagnose the sensor and actuator faults of AMB system. Q-statistic or square prediction error is used for bias and noise faults.

Keywords: Active Magnetic Bearing (AMB), Fault detection and diagnosis (FDD), Principal component analysis (PCA), Sensors and actuators faults.