%0 Journal Article %T Finding Reliable Identification Results for Nonlinear SDOF Systems, based on Sensitivity Analyses %J Journal of Seismology and Earthquake Engineering %I International Institute of Earthquake Engineering and Seismology %Z 1735-1669 %A Mohammadi, Majid %A Asghari, Mohammad Ali %D 2021 %\ 08/01/2021 %V 23 %N 3 %P 25-38 %! Finding Reliable Identification Results for Nonlinear SDOF Systems, based on Sensitivity Analyses %K Esimability %K Unscented Kalman filter %K Sensitivity analysis %K post-earthquake assessment %R 10.48303/jsee.2023.563534.1030 %X Post-earthquake assessment of buildings is one of the fundamental questions that needs to be answered immediately after a strong seismic event. Taking a building out of service after an earthquake can have a financial impact even greater than the earthquake itself. Up to now, damaged buildings are categorized in three groups, with red, yellow, or green labels, by engineering judgment based on visual screening. To have a more accurate method, the response of the buildings in aftershocks can be focused on new vibration-based system identification methods. But determining the system parameters is still a challenging subject; involving parameters with low identifiability, or correlated parameters can potentially influence the results in model updating problems. In this paper, a sensitivity matrix-based method is introduced to prioritize parameter estimability. The matrix-based process is capable of quickly determining the correlation between different parameters. Moreover, this method provides an explicit criterion for determining the optimal number of identifiable parameters. To indicate the efficiency of the method, a nonlinear Single-Degree of Freedom (SDOF) system has been simulated. Multiple model updating procedures have been carried out on the selected system, using the Unscented Kalman Filter (UKF). The result shows that system identification based on the sensitivity analysis outcome improves the quality of the identification precision. Additionally, this method decreases identification time by 35 percent that this amount can be crucial for updating large-scaled models. %U http://www.jsee.ir/article_701974_314fca061b678aa10684621495d7813c.pdf