Unscented Kalman filter with unknown input and weighted global iteration for health assessment of large structural systems

A novel concept denoted as unscented Kalman filter with unknown input and weighted global iterations (UKF-UIWGI) to assess health of large structural systems is proposed. It incorporates the basic features of UKF to identify systems in the presence of severe nonlinearity but then added a few desirable features to increase its implementation potential. Because the information on excitation and unknown initial state vector is necessary to implement any UKF-based approach, a substructure concept is introduced to generate them. The traditional UKF concept attempts to identify small structural systems using large duration of response histories in one global iteration. Because it fails to assess the defective states of large structures in most cases, a weighted multiple global iterations procedure with objective functions using short duration responses is introduced. The superiority of UKF-UIWGI over the traditional UKF is demonstrated with the help of several illustrative examples using single and multiple substructures in identifying both defect-free and defective states. With the help of the same examples, the superiority of the proposed method over the extended Kalman filter-based method developed by the team earlier is conclusively documented. With the help of parametric studies, it is documented that the proposed method is robust, accurate, and stable. The study confirms that UKF-UI-WGI can identify large structural system using only limited response information measured at a small part of a structure without using any excitation information. The concept significantly advances the state-of-the-art in UKF-based nonlinear system identification and considerably improves its implementation potential. Copyright (C) 2015 John Wiley & Sons, Ltd.