Piezoelectric Hysteresis Modeling Under a Variable Frequency Based on a Committee Machine Approach.

Area of Science:

• Mechatronics and Control Systems
• Nonlinear Dynamics
• Data-Driven Modeling

Background:

• Piezoelectric actuators exhibit complex hysteresis, challenging to model with time-varying inputs.
• Existing dynamic physical models inadequately capture hysteresis in industrial mechatronic devices.
• Accurate hysteresis modeling is crucial for precise micro-positioning and active control.

Purpose of the Study:

• To propose a novel hybrid data-driven model for appraising and compensating piezoelectric actuator hysteresis.
• To integrate Bouc-Wen and backlash hysteresis formulations for enhanced nonlinear effect modeling.
• To improve the accuracy of piezoelectric actuator models under time-dependent input amplitudes and frequencies.

Main Methods:

• Developed a hybrid model combining Bouc-Wen and backlash hysteresis formulations.
• Utilized genetic algorithm (GA) and particle swarm optimization (PSO) for parameter identification.
• Employed a committee machine approach with frequency clusters for parameter aggregation.
• Simulated and experimentally validated the piezoelectric actuator performance.

Main Results:

• Achieved a 23.54% error reduction using the committee machine approach compared to a complete model.
• The Bouc-Wen sub-model, tuned with GA, yielded a root mean square error (RMSE) of 0.42 µm.
• Maximum absolute error (MAE) was approximately 0.86 µm within the 150-250 Hz frequency domain.

Conclusions:

• The proposed hybrid data-driven model effectively appraises and compensates for piezoelectric actuator hysteresis.
• The committee machine approach offers superior performance in reducing modeling errors.
• Accurate hysteresis compensation enhances the precision of piezoelectric actuators in control applications.