| The next generation of extreme ultraviolet(EUV)lithography tools will adopt magnetic levitation guidance technology to achieve long-travel transmission of the mask in vacuum.The levitation guidance technology of the reluctance actuator is considered to be one of the best candidates.However,the inherent non-linearities of reluctance actuators,the mismatch disturbance of gap measurements,the non-orthogonality of horizontal traction forces,etc.,limit the application of the reluctance actuator.This dissertation focuses on solving above-mentioned problems from three aspects: actuator structure optimization,nonlinear modeling and compensation,and robust nonlinear control algorithms.The main contributions and conclusions are as follows: 1)The multi-objective optimization function of the reluctance actuator is established,and the influence of magnetic saturation is eliminated by adding optimization constraints.The impact of force unidirectionality,gap dependence,magnetic saturation characteristics,magnetic flux leakage and fringe flux on the thrust of the reluctance actuator are analyzed.Combining the thrust objective function,mass objective function and power objective function are,and using the magnetic saturation as a constraint,the objective function of multi-objective optimization for reluctance actuator is finally constructed with regard to mass,power consumption and thrust.2)A global optimization algorithm that can be used for identifying the nonlinear actuator model and a multi-objective optimization algorithm that can be used for actuator structure optimization were proposed.A dynamic mutual learning(DML)strategy with both exploration and exploitation capabilities was designed.DML has been embedded into the differential evolution(DE)algorithm,and a new DMLDE algorithm was formed,which has strong global optimization capability for identifying nonlinear model parameters.Combining the DMLDE algorithm with the Pareto frontier idea,a multi-objective DMLDE(MO-DMLDE)algorithm was proposed.Numerical experiments and actual applications have verified the powerful search capabilities of the DMLDE algorithm and the MO-DMLDE algorithm in global optimization and multi-objective optimization,respectively.3)A dynamic separation model(DSM)for leakage flux and fringing flux and a Two-stage (TS)model for eddy current and hysteresis were proposed.The DSM model is divided into two parts,including current-magnetic flux density(I-B)and magnetic flux density-force(B-F),which can be used for the compensation of leakage flux and fringing flux by combining an adaptive neural network in the control process.The TS model treats eddy as the dynamic fluctuation part of hysteresis,and build a direct inverse rate-dependent hysteresis model to compensate for both eddy and hysteresis in reluctance actuators.Simulation and experimental results has verified the effectiveness of the DSM model and the TS model.4)A robust nonlinear control method was proposed for the levitation ascend/descend process and the levitation guidance process.The method uses the DSM model to build a feedback linearized sliding mode control architecture,and adopts an adaptive neural network to deal with model uncertainties of leakage flux and fringing flux in the B-F relationship.In order to compensate for the mismatch disturbance of the suspension gap measurement and the non-orthogonal disturbance of the horizontal traction force,two new neural networks are used to further improve the robust nonlinear control method in the levitation guidance process.By constructing the Lyapunov function,the stabilities of the proposed robust nonlinear control methods in the maglev control were proved.5)Three sets of levitation ascend/descend experiments with stepped trajectories and two sets of levitation guidance experiments with different traction trajectories were designed.Results show that the proposed robust nonlinear control methods have high-performance for the control of the reluctance actuator in a maglev system. |
PDF Full Text Request