New decentralized control technique based on substructure and LQG approaches

Abstract

Decentralized control is more suitable for structural control of large-scale structures. In this paper, a new decentralized control technique is proposed based on the linear quadratic Gaussian (LQG) and substructure approaches. A large-scale structure is divided into a set of smaller substructures. Each substructure is controlled by its own local controller with interaction forces at substructural interfaces, which are considered as “unknown external inputs” to the substructure concerned. An algorithm of recursive least squares estimation for the unknown excitation is proposed. A numerical example of the decentralized control of a tall building is studied to illustrate the new proposed algorithm. Simulation results show that the proposed decentralized control provides satisfactory control performance when comparing with the conventional centralized LQG control algorithm and is viable for the future structural control of large-scale structures.

Keywords

substructures / decentralized control / linear quadratic Gaussian (LQG) / Kalman filter / unknown input / least-squares estimation