Inverse model–based iterative learning control on hysteresis in giant magnetostrictive actuator

Abstract

This article presents a new iterative learning control algorithm based on inverse model to decrease the hysteresis-caused tracking error of a giant magnetostrictive actuator. The first iteration input is calculated by the inverse model of the giant magnetostrictive actuator system according to the desired output. Compared to a standard iterative learning control algorithm—in which the first iteration input usually is proportional to the desired output—the proposed algorithm converges more rapidly. Performance of the approach is demonstrated both theoretically and experimentally. The experimental results show that the inverse model–based iterative learning control converges about twice as fast as standard iterative learning control and reduces the hysteresis-caused error of giant magnetostrictive actuator to 0.5% of the total displacement range, which is comparable to the noise level of sensor measurement. Two sets of model parameters were identified by 0–1.5 and 0–5 V major hysteresis loop, respectively, and evaluated. The best convergence rate is obtained with the former case.