Εκτίμηση της δυναμικής συμπεριφοράς μη γραμμικών κατασκευών με χρήση φίλτρου KALMAN (UKF)

Abstract

This thesis project discusses the dynamic behavior estimation of nonlinear structures using the Unscented Kalman Filter (UKF). The Filter combined with typical methodologies of numerical analysis provides a very useful tool, through which we can predict the behavior of the entire system, having as input data measurements from just few points. Initially, the theoretical background of Unscented Kalman Filter (UKF) is presented. Afterwards the Filter in applied to numerical analysis of constructions. The first "construction", wherein the methodology is applied, is a nonlinear mass-spring chain-like model. Both state and parameters estimation are performed. Moreover a statistical analysis is performed, in order to determine the way that the choice of measurement points, measurement noise and modeling error affect the accuracy of estimations. Similarly a vehicle-like frame is analyzed. The aim of studying this structure is to determine the behavior of the methodology in a more realistic case. Finally, the possibility of a "connection" between the Filter and a finite element solver is examined. Such a "connection" is desirable because it enables us to apply the Unscented Kalman Filter (UKF) to large-scale, realistic problems.