Sensitivity of stiffness condition assessment results to unscented Kalman filter initial parameters
Seismic response measurements from small and frequent earthquakes(e.g.aftershocks)and analysis of these records may reveal deficiencies in stiffness which may be due to large main shocks or poor construction quality.If condition is assessed,corresponding measures can be taken to prevent failure/undesired response during future big earthquakes.However,application of system identification methods poses various challenges in real world.Common problem which prevents application of such methods to full-scale bridges is the lack of robustness and the data.In this study,first a robust stochastic approximation scheme is proposed to optimize the process and measurement noise covariance matrices in unscented Kalman filter(UKF).Advantages of the method,UKF-RM,over its conventional counterpart are shown comparatively first utilizing a simple single-degree-of-freedom(SDOF)system.Numerical simulation is conducted on a 3 DOF system which represents a bridge girder to investigate the stiffness estimation performance in terms of initial process and measurement noise covariance matrices,and with added various measurement noise levels.Furthermore,seismic acceleration response measurements obtained from a real bridge during earthquakes using a wireless sensor array is utilized to validate the applicability of the method to existing bridges.Sensitivity of the stiffness estimations of the rubber bearings is investigated in terms of change in initial process and measurement noise covariance matrices.Results from both simulations and experiment indicate that the proposed method accurately estimate stiffness parameters of the lateral resisting systems of a bridge with a fast convergence rate and track ability in an objective manner.
system identification stiffness estimation wireless sensor unscented Kalman filter (UKF) Robbins Monro (RM)
M.Kuleli T.Nagayama
Department of Civil Engineering,The University of Tokyo,Tokyo,Japan
国际会议
The 7th World Conference on Structural Control and Monitoring(7WCSCM)(第七届结构控制与监测世界大会)
青岛
英文
1773-1782
2018-07-22（万方平台首次上网日期，不代表论文的发表时间）