Deep- Hammerstein - Model- Based Predictive Control for Piezoelectric Actuators in Trajectory Tracking Applications

With the high motion resolution and rapid response ability, piezoelectric actuators (PEAs) have been used for many precision systems. However, the inherent hysteresis nonlinearity is very difficult to deal with, due to the amplitude-, frequency-, and memory-dependent characteristics, which severely limits the tracking accuracy of the PEA. In this paper, we first propose a deep neural network based Hammerstein (dHam) model to predict the displacement of the PEA, which consists of a hysteresis block and a linear block in series. Experimental results are presented to demonstrate the satisfactory performance of the dHam modeling method. Based on the trained dHam model, a nonlinear model predictive control (NMPC) approach is proposed to address the trajectory tracking problem under constraints. The tracking results demonstrates the efficiency and efficacy of the NMPC approach.