Experimental evaluation of unscented Kalman filter for model updating of inaccurately modeled physical substructure in real time hybrid simulation

Real-time hybrid simulation (RTHS) provides an effective experimental technique to evaluate the seismic performance of large- or full-scale structures in size limited laboratories. Critical parts of the structure are taken as physical substructures (PS) to be experimentally tested in the laboratory, while the rest are computationally simulated as numerical substructures (NS). When the number of PS in the structure exceeds the laboratory capacity, traditional RTHS has to perform numerical modeling on selected PSs, which inevitably presents challenges especially when PS is difficult for numerical modeling. RTHS with model updating (RTHS-MU) utilizes measured responses of PS for on-line calibration of similar parts in NS. There are few existing RTHS-MU studies, lack of analysis of its impact on effectiveness, and are mainly limited to PS that is easy to accurately model. In this study, a detailed experimental evaluation of the RTHS-MU with UKF is performed on a highly nonlinear self-centering viscous damper (SC-VD). RHTS-MU results are compared with reference cases with all critical parts tested in laboratory. Numerical simulations are performed to verify the effectiveness of parameter identification and its application. Despite highly nonlinear behavior of SC-VDs and inaccurate numerical model in RTHS-MU, experimental results show that UKF still enables accurate response replication. The effectiveness of RTHS-MU and identification parameters however could be reduced due to parameter redundancy, performance difference and model error of the PS.