Neural network based iterative learning control for magnetic shape memory alloy actuator with iteration-dependent uncertainties

The magnetic shape memory alloy based actuator (MSMA-BA) is an indispensable component mechanism for high-precision positioning systems as it possesses the advantages of high precision, low energy consumption, and large stroke. However, hysteresis is an intrinsic property of MSMA material, which seriously affects the positioning accuracy of MSMA-BA. In this study, we propose a multi meta-model approach incorporating the nonlinear auto-regressive moving average with exogenous inputs (NARMAX) and Bouc-Wen (BW) models to describe the complex dynamic hysteresis of MSMA-BA. In particular, the BW model is introduced into the NARMAX model as an exogenous variable function, and a wavelet neural network (WNN) is adopted to construct the nonlinear function of the multi meta-model. In addition, iterative learning control is combined with a WNN to improve its convergence speed. A two-valued function is employed in the controller design process, so as to make use of history iteration information in updating control input. The main contribution of this study is the convergence analysis of the proposed iteration learning controller with iteration-dependent uncertainties (non-strict repetition of the initial state and varying iteration length). The experiments conducted on the MSMA-BA illustrate the validity of the proposed control scheme.