Montazeri A., Esmaeil Sabzali H., Poshtan J.

In this paper, two different methods for robust identification of a lightly damped flexible beam are applied and compared. The simply-support flexible beam which is considered in this study is assumed to be out of steal with predefined specifications. The identification experiment has been simulated using a “Finite Element” model of the beam. Here, the first two modes are considered as the in-bandwidth modes which are aimed to be modelled and controlled, and the last two modes are assumed as the uncertainty of the identified model. The first approach is by using NSSE algorithm described in section 2, with both random walk and integrated random walk processes. The second method used for robust identification of this lightly damped model is SM approach using non-stochastic but bounded assumptions on the amplitude of the noise and uncertainty of the model. The estimated model and its corresponding estimated uncertainty band are computed using ellipsoid set-membership method.NSSE method with random walk process and SM methods deliver high amplitude uncertainty bands at the in-bandwidth modes while the band for uncertainty modes is tight. This identified model will reduce the performance of the controller for the first two modes. Otherwise, the identified model using NSSE with integrated random walk process shows tight fitting at the first two modes, while the amplitude of the third mode is not covered totally by the uncertainty band. In comparison with the previous identified model this model will result in better performance for the damping of the first two modes, while the stability of the controller is somewhat questionable because of uncertainties. However, since there is always a trade-off between stability and robustness of the controllers, the identified model in using NSSE with integrated random-walk shows good compromise between these two goals.