On Bayesian active vibration control of structures subjected to moving inertial loads

Abstract

This study introduces a novel Bayesian framework for online and real-time vibration control of beam type structures, which represent a comprehensive control system associated with input-state algorithms. Control design systems typically require knowledge of system states, which in structures are displacements and velocities at some degrees of freedom. Currently, full-field measurements of displacements and velocities in large structural systems are not feasible. Also, properties of the moving inertial loads as key parameters in control designs are assumed known; however, in practice, measuring their characteristics is a challenging issue. As a remedy, an observer is required to estimate quantities of interest needed in control algorithms via indirect measurement of structural response. To estimate the entire state of the system in the absence of input information, often the input needs to be estimated together with the unknown states. Three state of the art methods for estimation of input and states of a linear state-space system, namely augmented Kalman filter (AKF), dual Kalman filter (DKF), and joint input state estimation algorithm (JIS), were adopted to assess controller performance in the absence of direct displacement measurements in comparison to conventional Kalman filter (KF). The linear quadratic Gaussian (LQG) framework, in association with these three filters, presents a novel control algorithm. Various sensor networks featuring pure accelerations and a combination of strain gauges and accelerometers with different noise levels were considered. Extensive numerical examples and comparison with the passive control system proved the viability of input-state estimation as an integral part of active control algorithms in structural systems subjected to moving inertial loads with different speeds.