In this paper a non-stationary stochastic embedding technique is used for identification and modeling of a simply-support flexible beam between two collocated points in the presence of measurement noise in the system. In comparison with standard identification methods which give only a model for the true system, this procedure tries to delivers a tight bound on uncertainties of the identified model of the beam and for this purpose it is required to introduce some basis functions. In this application knowing some prior information about the resonant nature of the system, second order basis functions are used for identification of nominal model using least-square estimation method, and integrated random-walk process is used to model uncertainties. The identified model as well as its accompanied uncertainty region can be used as the input of robust controller design methods like Minimax LQG, H2 and H∞ controllers.