Identification of State Space Models Using a Stochastic Realization Algorithm with Application to Modal Identification of Flexible Aerospace Structures

Abstract

Abstract Accurate Control of Large Space Structures (LSS) for pointing and tracking applications requires accurate knowledge of the vibrational characteristics of the LSS. Since it is not practical to duplicate the space environment during ground vibration tests, accurate models of LSS can only be obtained via System Identification algorithms applied to real data collected from the LSS after deployment in orbit. A similar problem arises in Aircraft Flutter Testing, which is done to verify that the flight envelop does not contain any aeroelastic or aeroservoelastic instabilities. In this paper, we show that System Identification car be done with high accuracy using State Space Models and a Stochastic Realization Algorithm (SRA) It is shown that the Stochastic Realization Algorithm (SRA) outperforms other available techniques and produces excellent results for the AFAL Flexible GRID test data and the X-29 Aircraft Flutter data under natural turbulence conditions. SRA also produces robust estimates of frequencies, dampings and mode shapes for repeated frequency modes of an elastic membrane even under conditions of 300% multiplicative noise. SRA identifies a multi-input multi-output stochastic state space model for the data using a noniterative technique based on Singular Value Decomposition of a Hankel matrix.