Dual-Type Structural Response Reconstruction Based on Moving-Window Kalman Filter with Unknown Measurement Noise

Abstract

Various measurements are now available for structural health monitoring (SHM) due to the fast development of sensory systems. Utilization of multitype measurements including local and global information for SHM has typically outperformed that of solo type measurements. However, the limited number of sensors for measurements hampers the effectiveness of SHM. Thus, response reconstruction at the locations of interest in which sensors are unavailable with limited measurements has drawn significant research attention. The Kalman filter (KF) is a powerful tool to estimate optimally the unknown state vector of a structure that has numerous applications in civil engineering. One main concern for KF is that it requires good estimates of the noise covariance information, which is generally difficult to determine. Therefore, this paper investigates the dual-type responses reconstruction by using the moving-window Kalman filter (MWKF) with unknown measurement noise covariance (MNC). The weighted average of the MNC was first evaluated by utilizing the moving-window estimation technique. Then the dual-type of measurements including strains and displacements were fused together to reconstruct the structural responses at unmeasured locations. Numerical and experimental investigations were conducted to verify the effectiveness and feasibility of the MWKF in dual-type response reconstruction. The results indicate that the MNC can be well estimated and the reconstructed responses agree well with the real or measured responses.