Adaptive parameter calibration of UKF towards optimal model updating in real-time hybrid simulation

Abstract

(Real-time) Hybrid simulation with model updating (HSMU/RTHSMU) can correct the numerical substructure in an online manner based on the measured information of corresponding physical substructure, which enables more economical and efficient seismic performance assessment of structures. Unscented Kalman Filter (UKF) is the widely used model updating method in HSMU/RTHSMU so far, but its performance is often affected by its parameters and currently there exists no general guidance. This study proposes an adaptive calibration method for the UKF parameters in RTHSMU/HSMU. Two objective functions are constructed according to the loading characteristics of the physical substructure, and Kriging is used to approximate the response surface of corresponding objective function. Efficient Global Optimization and optimal Latin hypercube design are integrated to estimate the optimal parameters to minimize objective function. A two-story steel moment resisting frame with self-centering viscous dampers is selected as prototype structure, and two series of experimental evaluations are conducted to verify the efficacy of proposed method. Independent of whether PS is non-reloadable or reloadable, the results demonstrate that the proposed method facilitates cost-effective calibration of initial UKF parameters within RTHSMU. The calibrated UKF parameters significantly reduce errors associated with parametric and model uncertainties and exhibit robustness across various ground motions, thereby supporting optimal model updating in both HS and RTHS applications.