Research On Multi-mode Drive Method Of Ultrasonic Motor Based On Deep Koopman Theory

With the rapid development of ultra-precision machining and manufacturing,aerospace,biomedical science,optical precision engineering,semiconductor lithography and detection systems,there is an urgent need for precision positioning technology with high velocity and high accuracy.In terms of compact size,simple structure,high power density,self-locking with power off as well as zero positioning noise,the ultrasonic motor can achieve high-precision positioning from micron to nanometer.Facing the major scientific instrument and equipment development project “Ultrasonic-driven cross-scale high-velocity and high-stability 3D scanning technology and system”,by breaking through the multi-mode drive technology of the ultrasonic motors and deeply studying the compensation control strategy,it is expected to achieve the effective combination of high-velocity and high-precision positioning,which has important theoretical significance and engineering value.In this thesis,a composite modal linear ultrasonic motor motion system is applied as the experimental object.To achieve the targets,a multi-mode drive method for the ultrasonic motor based on deep Koopman theory is proposed.The key technologies are studied to address the hysteresis,temperature drift,and nonlinear output characteristics in AC/DC mode.In order to achieve an effective balance between high-velocity and highprecision positioning,a multi-mode drive method based on fuzzy optimization control is proposed by time-sharing excitation of the AC resonant state and the DC non-resonant state of the ultrasonic motor.Considering the non-ideal factors of motor motion system operation,dielectric loss,mechanical loss,piezoelectric loss,and temperature dependent parameters are introduced,the hysteresis model that qualitatively reflects frequency-dependent hysteresis and the temperaturedependent model that quantitatively reflects temperature drift are established.This provides theoretical basis for the compensation and control strategies.In order to deal with the problem of non-dispersion and dispersion hysteresis of the system in DC mode,a memory-related global linearization hysteresis prediction model is constructed based on Koopman theory.By means of deep training,the identification of high-precision hysteresis prediction model is completed.On this basis,a stair-like incremental linear model predictive control algorithm is designed,which converts the objective function optimization problem into a constrained quadratic linear programming problem.This method provides a basis for global linearization modeling and compensation strategies for improving piezoelectric hysteresis characteristics.Aiming at the temperature drift and nonlinear output characteristics of system in AC mode,a drive frequency-amplitude composite control strategy is proposed.Firstly,a temperature-dependent linear parameter-varying model is built,and a very strictly passive controller is designed to compensate temperature drift by tracking the optimal operating frequency.Subsequently,a global linearization Koopman velocity prediction model is formed,and an adaptive stair-like incremental model predictive control algorithm is proposed to effectively improve the nonlinear output characteristic by adjusting the amplitude.The above strategies are applied as the internal and external loop of the composite control strategy,which can simultaneously compensate for the temperature drif and nonlinear output characteristics of the system.Finally,a displacement-velocity dual closed-loop control strategy is proposed.Experiments have shown that the system can achieve accurate velocity and displacement tracking under different loads and voltage amplitudes.On the basis of tackling key technical issues,an ultrasonic motor motion system experimental platform is established to verify the hysteresis compensation effect,temperature drift and nonlinear output characteristics control effect.Besides,the performance of the motion platform employing the proposed multimode drive method is tested.The experimental results show that the system can achieve a velocity of more than hundred millimeter per second in the rapid positioning stage,and can achieve a nano level bidirectional repeatable positioning accuracy in the precise positioning stage.