Improved decentralized structural identification with output-only measurements

Abstract

This paper proposes an improved decentralized structural identification approach with output-only measurements. The improved approach can be used for system identification of both linear and nonlinear structures. A large-scale structure is divided into a number of smaller zones according to its finite element configuration. Each zone is dynamically tested in sequence with its own set of sensor placement. The external excitation forces in each zone are identified using the Kalman filter technique. Structural parameters of the whole structure are divided into several subsets and then updated by using the Newton-SOR method. Both the external excitations and structural parameters are iteratively updated until a defined convergence criterion is met. The proposed technique is then applied to two numerical examples: a six floor building and a planar truss structure. The nonlinear system parameters of the building are correctly identified. The unknown excitation force, damage location, and damage severity in the plane truss structure are successfully identified. The effect of measurement noise on the identified results is also studied. An eight floor shear type structure is finally tested in the laboratory. The experimental results further verify the effectiveness and efficiency of the proposed technique in damage identification using output-only measurements.