Backstepping Dynamic Surface Control of an SMA Actuator Based on Adaptive Neural Network

Abstract

Shape memory alloy(SMA) actuators have the characteristics of high force-to-mass ratio, high energy density, and lightweight, leading to broad perspective applications in electromechanical systems. Due to the hysteretic nonlinear characteristic of SMA during phase transition, the traditional linear control method can not achieve the precise trajectory tracking control of SMA actuators. In this paper, we propose a backstepping dynamic surface control method based on an adaptive neural network. First, we establish a third-order nonlinear model with the internal dynamics of the SMA actuator. Secondly, we design the nonlinear controller using the backstepping dynamic surface method. Finally, the nonlinear function and parameter of the system are estimated using the designed radial basis function neural network(RBFNN) and adaptive law. This paper solves the problem that the controller depends on the SMA mathematical model. The controller has the characteristics of model-free, fast response, high precision, strong robustness, and low complexity. Compared with PID control and iterative learning control(ILC), the proposed control strategy has the advantages of high precision, rapid response, and fast anti-disturbance performance.