Modeling And Control Method Of Output Force Hysteresis Effect Of Piezoelectric Actuator

Piezoelectric ceramic actuators are widely used in high-precision positioning systems due to its useful characteristics of rapid response,small size,high working bandwidth.However,the hysteresis of piezoelectric ceramic actuators increases as the frequency of input voltage increases,which will seriously affect the control system's accuracy when the operating frequency change in an extensive range.In this thesis,the piezoelectric actuator is consists of a piezoelectric ceramic and flexure hinges.An experimental platform was established to test and analyze the hysteresis characteristics,including the memory,wiping-out,congruency,and frequency-dependent.At a fixed frequency,the output force hysteresis model based on the Preisach model can accurately predict the output force of the piezoelectric actuator.To overcome the complex dynamic hysteresis,a novel modeling method based on the GRU(Gated Recurrent Unit)was proposed.In this modeling method,the GRU unit process information about the memory characteristic of hysteresis,and two full-connected layers process information about the frequency-dependent characteristic of hysteresis.The model established by the GRU modeling method accurately forecasted the output force under different frequencies and waveforms of the input voltages and has been verified to have good generalization ability for frequency-dependent hysteresis.Based on the inverse hysteresis control method,a composite control algorithm combining the feedforward control of the inverse model of the Preisach model and the PID control algorithm is proposed.At a fixed frequency,the composite control algorithm can weaken the hysteresis.The experimental results showed that the composite control algorithm effectively decreased the effect of the piezoelectric actuator's hysteresis under fixed frequency conditions.For complex dynamic hysteresis,an inverse hysteresis model was built through the same structure as the GRU modeling method.By comparing with the feedforward control method of the Preisach model,the proposed output force control method based on GRU recurrent neural network effectively weakens the dynamic hysteresis of the piezoelectric actuator's output force.Finally,the feasibility experiment analysis of piezoelectric drive probe detection is carried out.