Neural network hysteresis modeling with an improved Preisach model for piezoelectric actuators

摘要

Purpose - Widely used in micro-position devices and vibration control, the piezoelectric actuator exhibits strong hysteresis effects, which can cause inaccuracy and oscillations, even lead to instability. If the hysteretic effects can be predicted, a controller can be designed to correct for these effects. This paper aims to present a neural network hysteresis model with an improved Preisach model to predict the output of piezoelectric actuator. Design/methodology/approach - The improved Preisach model is given: A wiping-out memory sequence is defined that is along a single axis only and at the same time the ascending and the descending extreme points are separated. The extended area variable is calculated according to wiping-out memory sequence. The relationship between the two inputs (the extended area variable and variable rate of input signal) and the hysteresis output is implemented with a neural network to approximate the hysteresis model for the piezoelectric actuators. Findings - Some experiments are carried out with a piezoelectric ceramic (PST150/7/40 VS12) and the results show the neural network hysteresis model can reliably predict the hysteretic behaviours in piezoelectric actuators. Originality/value - The improved Preisach model is a simple model that is implemented by a neural network to reliably predict the hysteretic output in piezoelectric actuators.