Active Modeling and Control for Shape Memory Alloy Actuators

Daohui Zhang,Bi Zhang,Jianda Han,Xiaoguang Li,Xingang Zhao

doi:10.1109/access.2019.2936256

Abstract

In this paper, an active modeling and control scheme is developed for Shape Memory Alloy (SMA) actuators to eliminate the negative influences caused by the uncertainties in its dynamics. First, a nonlinear SMA dynamic model based on Liang model and the empirical models is built and linearized, and all the uncertainties due to time-varying parameters, external disturbances, as well as the linearization, are considered as model error of the linearized model. Secondly, an active modeling based on Kalman filter is constructed to estimate the model error in real time, which intends to improve the model accuracy actively. Finally, an active modeling based control method is proposed to compensate the model error in order to improve control performance of SMA actuators. Experiments are conducted on a one degree-of-freedom (DOF) testbed actuated by a SMA wire. The experimental results of the active model error estimation, and the control performance with and without the active model based compensation are presented and compared to demonstrate the improvements of the proposed scheme.

Highlights

The shape memory alloy (SMA) wire can decrease in actuating length through a resistive heating from an electrical current due to a crystalline phase transformation from martensite to austenite [1]
There are two ways to improve the performance of a Shape Memory Alloy (SMA) actuator, namely: 1) trying to construct a more accurate model that meets the SMA dynamics
We proposed to use the state feedback control (SFC), i.e., v = −K e where K = k1 k2 k3 is designed such that Ac − BcK is Hurwitz; e = z1 − rd z2 − rd z3 − rd T is the error vector; rd is the reference trajectory; and z1, z2, and z3 are the states of the reference model

Summary

INTRODUCTION

The shape memory alloy (SMA) wire can decrease in actuating length through a resistive heating from an electrical current due to a crystalline phase transformation from martensite (the low temperature phase) to austenite (the high temperature phase) [1]. The main difficulty of this NN-based way is how to acquire an ‘enough’ dataset to train the neural network offline so that the model could present good performance in all of the working conditions of SMA actuators. Fuzzy logic [28] and neural network [25], [29] enhanced SMC were further proposed to avoid the frequent mode switches, which might excite the resonance vibrations of SMA actuators Adaptive control algorithms, such as output feedback direct adaptive control [30], model reference active control [31], and robust indirect adaptive control [32], etc., were proposed to control SMA actuators. In order to demonstrate the effectiveness of the proposed scheme, control experiments with and without active model based compensation were conducted and compared on a one-DOF testbed actuated by a SMA wire.

PRE-KNOWLEDGE

EXPERIMENTAL STUDIES

Findings

CONCLUSION