Time domain estimation of excitation forces for mechanical structures by Kalman filter

(weighted coefficients for stabilization of estimation)

Abstract

This paper is concerned with the estimation of non-stationary input using Kalman filter theory. It is important to understand the excitation input for vibration control. In reality, it may be difficult to directry measure the excitation input, so we will work on the inverse problem of indirectly estimating the excitation input from the vibration response. There are two types of input estimation in the frequency domain and in the time domain. In this research, the latter method is adopted. This method has the advantage of being able to estimate nonstationary inputs compared to the another method. In this study, the impulse input is estimated as a nonstationary input. The input is estimated using an expansion system in which the input is included in the state quantity. In the simulation, the excitation input is applied to the mechanical structure with 5 degrees of freedom, and the excitation input is estimated from the vibration response. In this study, we aim to improve the accuracy of input estimation by time updating some covariance matrices which are adjustment parameters of Kalman filter.