Hysteresis modeling of piezo actuator using neural networks

Hysteresis and nonlinearity of piezo actuator are the major factors affecting the motion accuracy in controlling a micro system. The prediction accuracy of classical Preisach model could be improved only by mass experiments for measuring hysteresis. A model based on BP neural networks was proposed to improve the prediction accuracy. The stoke of piezo actuator has relation to current exciting voltage and historical extrema according to ‘wiping-out’ property of Preisach model. A model based on neural networks was established. The input of the model is current exciting voltage, historical voltage at nearest turning point and its corresponding stroke and the output is piezo actuator's stroke. Results of simulation and experiments show that the proposed hysteresis model can exactly describe and predict the hysteresis of piezo actuator in comparison with traditional bilinear interpolation and has the better generalization ability.