A NOVEL UNSCENTED KALMAN FILTER WITH UNKNOWN INPUT

The unscented Kalman filter (UKF) has been applied as an effective method for the identification of nonlinear systems. However, the conventional UKF method requires the information of external excitation (input), which causes a limitation to the applications of UKF. So far, there have been very few studies on UKF with unknown excitations. In this paper, a novel UKF with unknown input (UKF-UI) is proposed. The analytical recursive solutions are proposed based on the procedures of conventional UKF. The unknown input is identified by minimizing the error of predicted measurement errors with the solution of a nonlinear equation. Moreover, the data fusion of partially measured acceleration and displacement responses is applied to eliminate the drift problem in identification results. The numerical examples for the identification of nonlinear hysteric systems and excitation are used to verify the proposed UKF-UI approach. The computational results show that proposed UKF-UI method can effectively identify the nonlinear system and unknown input using partial measurements of system responses.