Hysteresis Compensation Control Algorithm for the Giant Magnetostrictive Actuators

When the giant magnetostrictive actuator (GMA) is used in the control system, the hysteresis of the GMA is particularly significant and causes undesired effect. Therefore, no matter what model of the actuator is employed, the efficient procedures for hysteresis and nonlinear compensation are required. Here a hysteresis compensation control strategy is proposed. To reduce model mismatch effect and unmodeled dynamics, a model estimator using a dynamic recurrent neural network (DRNN) is first constructed to obtain dynamic hysteresis characteristics of the GMA. Then a feedforward compensator using another DRNN learns and obtains inverse hysteresis characteristics of the GMA according to dynamic characteristics from the estimator. The system has been approximately linearized through compensation. Tracking the errors for random multisine wave reference signals, simulation results validate the excellent performances of the proposed strategy