Actuation Force Modeling and Identification for Piezoelectric Actuator With Neural Network Compensator and Nonlinear Hysteresis

A high-accuracy actuation force model of piezoelectric actuator (PEA) is very important for the high-frequency positioning control of the nanopositioning robot because the force measurement is often unavailable. In this article, an actuation force modeling and identification approach with neural network compensator of PEA with nonlinear hysteresis is proposed to realize the accurate driving force prediction, which is called neural-network-based stiffness compensation model (NNSCM). The model inputs are analog voltage signal and actual driving displacement, and the model output is actuation force. The internal creep, vibration, and hysteresis nonlinearity of the PEA are identified simultaneously based on a creep-vibration-hysteresis-dynamics (CVHD) model to calculate the nonload driving displacement from the input voltage signal on the actuator. The actuator stiffness model is compensated by a neural-network-based nonlinear model to reduce the effects of parameter and system uncertainties. The corresponding NNSCM identification algorithm is developed for accurate force prediction. Experiments on the PEA are conducted to verify the effectiveness of the proposed modeling approach and demonstrate a better model accuracy against other methods.