An improved Kalman filter with dummy measurement for identification of structural load and unknown parameters

Abstract

In this paper, a three-stage Kalman filter is proposed for identifying both the load and the unknown parameters in structures. To suppress the load drift caused by using only acceleration measurements, the dummy measurement is introduced. Given the poor accuracy and stability in dealing with a multitude of unknowns, the local multiple-iteration method is proposed to enhance the robustness of the Kalman filter. In addition, by means of an index reflecting the changes in the identified values, a three-step strategy is proposed to track the time-varying parameters. The results demonstrate that the local multiple-iteration method significantly improves the accuracy of the Kalman filter and reduces the difficulty of parameter assignment. The introduction of the dummy measurement can suppress the load drift but may increase the identified errors of the unknown parameters. Discussions on the cases of low sampling frequencies and few measurement points illustrate that the former has a smaller impact on the results than the latter. The results of a long response duration case indicate that the identified errors of the damping may increase with the response duration. Additionally, the results of an abrupt damage case illustrate that the three-step strategy can automatically and accurately track the time-varying parameters. Finally, an experimental example is conducted, and the results further demonstrate the applicability and feasibility of the proposed method for practical structures.