Faults in engineering systems are difficult to avoid and may result in serious consequences. Effective fault detection and diagnosis (FDD) can improve system reliability and avoid expensive maintenance. FDD is especially important for some special applications, such as Navy ships operating in hostile environments. So far, FDD for nonlinear systems has not been fully explored. There is still a big gap between FDD theories and applications. This dissertation makes an effort to fill the gap by developing an integrated FDD system structure and a series of algorithms for FDD of Permanent Magnet Synchronous Motors (PMSM).
A fault model is proposed for the stator winding turn-to-turn fault of PMSM. The model provides a good compromise between computational complexity and model accuracy and is versatile for both the healthy and the fault condition. Simulation studies demonstrate a good correspondence with both the theoretical analysis and the experimental observations in the existing literature. The model is especially important for the design of model based FDD algorithms.
Based on the fault model, a series of algorithms are proposed for the fault detection and diagnosis of PMSM. Since the reliability of sensors is the basis of FDD and control systems, a nonlinear parity relation based algorithm is proposed for sensor fault detection. The algorithm can successfully detect single faults in the currents and speed sensors. To track the parameter variations, which are symptoms of system internal changes and faults, an adaptive synchronization based parameter estimation algorithm is proposed. Simulation and experimental studies demonstrate that the algorithm can estimate not only constant parameters but also slowly time varying and abruptly changing parameters in a fast manner. Besides the detection of PMSM internal faults, the algorithm can also provide accurate parameters for the sensor fault detection algorithm. Based on the fault data, a Particle Swarm Optimization based fault diagnosis approach is proposed to find the fault location and severity information of a stator winding turn-to-turn fault. Finally, the proposed algorithms are integrated into a general FDD system structure. The integration provides a FDD system with enhanced robustness to system parametric uncertainties, as shown by extensive simulation studies.
