System identification and control of magnetic bearing systems

This thesis presents investigations aimed at obtaining a system model for the stabilisation of an Active Magnetic Bearing (AMB) System. Furthermore, the study reported here set out to design both conventional and advanced controllers based on the system model. This research report demonstrates that the literature on AMBs shows that AMBs are making their mark in the industry; they are increasingly being used in applications including jet engines, compressors, pumps and flywheel systems. In this study, it has also been observed that the basic design of AMBs is an arrangement of electromagnets placed equidistant in a ring round a rotor. The point of departure for this study is that AMBs are highly nonlinear and inherently unstable. Hence, the need for an automatic control to keep the system stabilized. The first step of the research was to determine the transfer function of the MBC 500 magnetic bearing system both analytically and experimentally. An analytical model has been derived based on principle of physics. As the AMB system under analysis is inherently unstable, it was necessary to identify the model using a closed-loop system identification. Frequency response data has been collected using the two-step closed loop system identification. As there are resonant modes in the MBC 500 magnetic bearing system, the system identification approach has identified the corresponding resonant frequencies. Subsequent to obtaining the model, a conventional controller was designed to stabilise the AMB system. Two notch filters were designed to deal with the magnitude and phase fluctuations around the two dominant resonant frequencies. The designed conventional controller and notch filters have been implemented using MATLAB, SIMULINK and dSPACE DS1102 digital signal processing (DSP) card. Both the step response and robustness tests have demonstrated the effectiveness of the conventional controller and notch filters designed. A significant conclusion has been drawn when designing the conventional controller. It was found that a controller that had a large positive phase angle had a negative effect on the system. This finding was very significant because it restricted the controller specifications and yielded an optimum lead angle for the conventional controller. The advanced PD-like Fuzzy Logic Controller (FLC) has also been designed for AMB system stabilisation. The designed FLC can deal with the magnitude and phase fluctuations around resonant frequencies without using notch filters. The performance of the designed FLC has been evaluated via simulation. Simulation results show that the FLC designed leads to a reliable system performance. Comparison studies of the FLC performances with two different sets of rules, two different inference methods, different membership functions, different t-norm and s-norm operations, and different defuzzification were investigated. To further improve system performance, scaling factors were tuned. Again, simulations showed highly promising results. Comparative studies between the conventional and advanced fuzzy control methods were also carried out. Advantages and disadvantages of both approaches have been summarised. The thesis has also suggested further research work in the control of AMBs