On-board Identification and Control Performance Verification of an MR Damper Incorporated with Structure

Abstract

An experimental study is conducted with the intention of identifying the dynamics of a magnetorheological (MR) damper incorporated with a steel beam structure and evaluating the efficiency of the formulated forward and inverse dynamic models for online application of semiactive vibration control. A hybrid modeling method by synthesizing NARX (non-linear autoregressive with exogenous inputs) model and neural network technique within a Bayesian framework is developed to figure out the forward and inverse dynamics of the MR damper from real-time experimental data. The forward dynamic model has the ability to predict the one-step-ahead damper force, while the inverse one is able to estimate the current/voltage command for the MR damper. To evaluate the implementability of the formulated models for online application, semiactive control experiments are performed on the beam structure incorporating the MR damper. The forward model is integrated into a LQR-based semiactive controller to estimate the instant damper force limitation for the semiactive constraint decision, and the inverse model is synthesized to determine current command for the MR damper to track the desirable control force, which compensates for the intrinsic hysteresis effect of the MR damper. The experimental results demonstrate considerably accurate prediction capability and practicability of the Bayesian NARX network models for the MR damper, and the effectiveness of the semiactive controller accommodating the forward and inverse dynamics of the MR damper for structural vibration mitigation.