Ảnh hưởng giá trị khởi đầu của phép lọc Kalman tới kết quả dự báo chuyển dịch ngang công trình

Kalman filter is mainly used for processing data and predicting horizontal movements in the deformation monitoring of engineered structures. The techique consists of two main steps: estimation of the current state variables and updating observation of next measurement. It allows one to have the optimal estimation of the current movement as well as the estimation of movements in the next cycles. The determination of initial state vector and covariance matrix is crucial procedure of this algorithm as it impacts the final results of the filter. This study focuses on the determination of initial state vector and covariance matrix, and the results show that the initial state vector and covariance matrix only influence on the accuracy of prediction in the several initial cycles, but after the cycle of six the error of predicted movements is just under 10% of their values.