AN IMPROVED SUBSPACE METHOD AND ITS APPLICATION TO PARAMETER IDENTIFICATION OF TIME-VARYING STRUCTURES

This paper develops a subspace tracking algorithm for parameter identification of time-varying structures. In subspace method, eigen-analysis is conducted and signal subspace is obtained through matrix decomposition. Firstly, the matrix to be tracked is transformed into a form suitable for on-line tracking. The new data information is put into a matrix with fixed dimension. The signal subspace is updated by singular value decomposition (SVD) of the matrix. Consequently, the SVD of an increasing Hankel matrix is avoided in the algorithm and therefore the computational effort is reduced. The algorithm is applied to a manipulator. The natural frequencies of the manipulator is controlled by a time-varying force. The tracking ability is reconciled with accuracy by introducing a forgetting factor. On-line identification simulation results demonstrate the ability of the algorithm to track time-varying parameters.