Additive cascaded model identification of linear and nonlinear dynamics of piezoelectric actuator

The modeling of the piezoelectric actuator is very important for the high-accuracy positioning control. However, the complex dynamics, e.g., creep, vibration and hysteresis, make the modeling very difficult. Due to the unknown model uncertainties, it is necessary to identify the model from the input-output data. In this study, an additive cascaded model is proposed for modeling the piezoelectric actuator. The model is called N2L (Nonlinear-Linear+Linear) Hammerstein model because it consists of one static nonlinear block and two dynamic linear blocks. The hysteresis is described by the static nonlinearity. The creep and vibration are described by slow and fast linear dynamic systems respectively. A two-stage incremental learning approach is proposed to identify the model from the experimental data. The effectiveness of the proposed additive cascaded modeling approach is verified by the experiments on typical piezoelectric actuators.