An optimised pseudo-inverse algorithm (OPIA) for multi-input multi-output modal parameter identification

A multi-input multi-output (MIMO) algorithm for modal parameter identification is developed based on linear prediction theory and its numerical efficiency compared to that of the well-known Eigensystem realisation algorithm ERA. The problem of determining the correct system's order is analysed through perturbation theory of singular values of oversized Hankel-block matrices. The popular technique of using overdetermined models to mitigate the effects of external noise on the quality of the estimated parameters is justified here by showing analytically that the system's singular values are enhanced when oversized Hankel matrices are used. Numerical experiments illustrating the performance of the proposed algorithm on both simulated and real mechanical systems, are included. (C) 1996 Academic Press Limited