Experiment study of trust-region method based instantaneous optimal semi-active control for MR dampers

A trust-region based instantaneous optimal semi-active control algorithm is proposed and verified through a control experiment on a cantilever beam installed with MR damper. Rather than taking the active optimal control force as target control force, an instantaneous performance index is first set up for semi-active control system in terms of the characteristics of MR damper. Then, the maximum and minimum control forces for the next time step are obtained from MR damper forward model and considered as constraints for the semi-active control system. Finally, by the application of precise integration method, the semi-active control problem is described as a typical single target optimization problem subjected to a bound constraint and trust-region method is utilized to find the optimal semi-active control force within the bound constraint at each time step. Control experiment is conducted through dSPACE control platform. The dynamic behavior of the MR damper is accurately formulated by NARX-based approach and neural network technique. During experiment, both low and high level's random excitations are used. The performances of two passive control algorithms, clipped-optimal control algorithm and the proposed control algorithm are studied and compared. The experiment results show that time step has great influence on the performance of the proposed controller and should be carefully designed. The comparison results also demonstrate the better control efforts achieved by the proposed controller, especially its superiority on suppressing root-mean-square responses over other controllers.