Identification of large space structures - A factorization approach

Abstract

Reliable techniques for on-orbit identification of the vibrational dynamics of large space structures are important chiefly for two reasons. First, the complexity of such vehicles makes accurate finite-element modeling difficult; second, extremely limited ground testing possible for these structures prevents correction of modeling errors before flight. This paper describes a new algorithm that uses in-flight vibration data to estimate directly a factored form for the positive definite symmetric mass matrix of the structure. Applications of matrix factorization methods to, for instance, Kalman filtering and least-squares problems suggest greatly improved accuracy over methods that estimate the mass matrix itself. The properties of the new identification algorithm are analyzed in detail here and illustrated by means of examples.