Research Of Rate-dependent Hammerstein Model And Neural Network Sliding Mode Control Method For Piezoceramic Actuated Micro-positioning Stage

Piezoceramic actuated micro-positioning stage is widely used in the fields of aerospace industry,microelectronics,precision optics,and ultra-precision manufacturing due to its many advantages such as fast frequency response,high displacement resolution,and large mechanical force.In micro-positioning stage,piezoelectric actuators are used as the core driving elements and flexible hinges are used as the transmission mechanism of piezoceramic actuated micro-positioning stages.However,the inherent hysteresis nonlinearities exist in piezoelectric actuators which has blocked the further engineering applications in micro-nano drive system.In this paper a high-precision hysteresis nonlinear modeling method and a realtime tracking control scheme are designed to eliminate the hysteresis nonlinearity of piezoceramic actuated micro-positioning stage.According to formal structure of piezoceramic actuated micro-positioning stage,it is decomposed into the series form of static hysteresis and dynamic linearity.The Duhem hysteresis model is used to describe the static hysteresis characteristics and a linear transfer function is used to characterize the frequency-dependent characteristics of piezoceramic actuated micro-positioning stage,then a high-precision Hammerstein rate-dependent hysteretic nonlinear model is established.A step-by-step identification method is designed to identify the model parameters,the Particle Swarm Optimization(PSO)and improved PSO with adaptive disturbances are used to identify Duhem model parameters,the MATLAB identify toolbox and frequency response method are used to identify the linear transmission function.Without loss of generality,the obtained transfer function is compared with the frequency response of the actual system.The experimental results preferably validate the excellent modeling ability of the Hammerstein rate-dependent hysteresis model.The feedforward controller is designed to suppress the hysteresis nonlinearity of the entire positioning system,and experimental results show that the output curve exhibits an approximate linear characteristic.Because feedforward control lacks anti-jamming capability,a compound control scheme based on RBF self-tuning PID parameters is designed to further improve the control accuracy.This compound control method does not require human experience to adjust PID parameters and the experimental results demonstrate that the compound controller effectively improves the control accuracy of piezoceramic actuated micro-positioning stage.Finally,a neural network sliding mode controller without hysteresis compensation is designed to further improve the robustness of piezoceramic actuated micro-positioning stage.Based on the second-order linear dynamic transfer function,a hysteretic characteristic decomposition method is proposed to characterize the hysteresis nonlinearity of piezoelectric stage.The hysteresis is decomposed into a linear function and a bounded nonlinear perturbation related to the input voltage.A neural network sliding mode controller is designed based on the hysteresis decomposition.The sliding mode controller eliminates bounded nonlinear disturbances and the RBF neural network adaptively approaches the modeling uncertainty.The global stability of the control system is solved by Lyapunov method.The experimental results also show that the designed neural network sliding mode controller can effectively eliminate the hysteresis nonlinearity of piezoceramic actuated micro-positioning stage.