Improvements in the kalman filter concept for structural health assessment

Health assessment of existing structural systems has developed multi-disciplinary research interests in the recent past. The research team at the University of Arizona is in the process of developing several finite elements (FE)-based system identification (SI) procedures that provide information on the location(s) of defect(s) and their severity by tracking the signature embedded in the dynamic responses. The procedures do not require information on input excitation. To address the issue related to measured noise-contaminated acceleration time-histories at limited DDOFs in large structural systems, the team developed integrated procedures by improving the extended Kalman Filter (EKF) concept. However, during the development phase, the team needed to address several fundamental challenges. Appropriate offline signal processing schemes, including filtering, baseline removal, and mitigation of amplitude and phase-shift errors, etc. needed to be introduced to address the issues of non-convergence. They are discussed in this paper.