A subspace fitting method for structure modal identification in time domain
Abstract
In order to obtain good estimates of the mechanical system of the parameters, an original identification method will be presented in this paper. A standard modal analysis is performed by using an autoregressive model. Our main contribution is to link the prediction error approach of standard system identification (based on an autoregressive data model) to a subspace based realization algorithm. By merging these two different approaches, a method which uses the advantages of both is obtained. The key idea is to use the non-iterative realization algorithm to obtain an order estimate of a state-space model. The state-space model order so obtained defines a subspace which is used to reduce the set of modal parameters. An evaluation of the method is presented in the experimental case of a free-end beam, showing satisfactory results.