Research On Related Theory And Technology Of Fuzzy Control For AMB

An active magnetic bearing (AMB), which levitates a rotor by magnetic forces, is a type of novel bearings and has many merits over conventional mechanical bearings, such as no contact, frictionless, high-speed operation and so on. However, it also has a lot of shortages which is the highly nonlinear and inherently open-loop unstable electromechanical dynamics, and its performance depends on the control system to a large extent. This dissertation emphasizes on fuzzy modeling and fuzzy control for AMB and main researches are as followings.The calibration principle of PID parameters is given on the basis of analysis of PID and improved PID algorithms; the simulating method is discussed by using the Simulink toolbox and S-function of Matlab; the improved ITAE performance index is used to evaluate the system performance and using the simplex method optimizes PID parameters. A proportional-fuzzy-PID multi-model controller is used to control AMB, which is achieved by synthesizing the merits of the proportional control, fuzzy control and PID control. When the error is big, a proportional controller is adopted; when the error is medium, a fuzzy controller is adopted; when the error is small, a PID controller is adopted. The simulation results show that the rotor of the AMB can levitate quickly with minor overshoot and zero steady-state error by using the proposed multi-model controller. But it is difficult to determine switching thresholds.A fitted-modifying-factor fuzzy controller is proposed to control AMB. In order to attain the minimum of the improved ITAE performance index, the simplex method is used to optimize the fuzzy controller with four modifying factors. Without the quantization of input and output variables, a third-order polynomial is used to approximate the former four modifying factors in the fitted-modifying-factor fuzzy controller. Simulation results show that, using the proposed fuzzy controller with a fitted modifying factor, the dynamics and steady-state performance are improved efficiently, and satisfying response performance can be attained.A method of establishing the T-S fuzzy model and the inverse T-S fuzzy model of AMB is proposed. The main difference of two models is that they have different input and output parameters. In the T-S fuzzy model, the control current and the rotor's displacement are defined as input parameters, and the electromagnetic force is defined as output parameter. In the inverse T-S model, the electromagnetic force and the rotor's displacement are defined as input parameters, and the control current is defined as output parameter. This method synthesizes fuzzy cluster analysis, least square method, and remnant error analysis effectively. Fuzzy cluster analysis method is used to determine the premise structure and parameters; least square method is employed to determine the consequence parameters; at last the amount of fuzzy rules is further determined by using remnant error analysis. The simulation results demonstrate that both of the T-S fuzzy model and the inverse T-S fuzzy model can reach very high fitting-precision.An inverse controller for AMB is proposed based on the inverse T-S fuzzy model of the nonlinear force function. The inverse T-S fuzzy model is cascaded with the AMB system to constitute a pseudo-linear plant. A linear PID controller is designed by using linear system theory. The PID controller and the inverse T-S model form the inverse controller. The simulation results show that the proposed method is effective. By using the proposed inverse controller, the AMB system has better dynamic performance and stronger disturbance rejection ability, and the operating range is extended.By using the idea of parallel distributed compensators, a fuzzy control method based on the T-S fuzzy model is proposed for AMB control. State feedback controllers are designed by using pole placement method of linear system theory according to the linear sub-systems described by the consequent part. Take the weighted sum of the outputs of state feedback controllers as the final output and the weights are the degrees of membership. The simulation results verify the effectiveness of the proposed parallel distributed compensator.The hardware platform based on TMS320LF2407A and software design idea are introduced. And then the key problems are discussed. The various degrees of freedom at radial axis affect one another when the center is measured for a suspended rotor. So the measurement can not be finished at one time. An iterative method with a successive -approximation is presented based on the center measurement method of single degree of freedom.