Subspace-based identification of a nonlinear spacecraft in the time and frequency domains

• The identification of a real-life spacecraft with non-smooth nonlinearities is achieved.
• Nonlinear subspace methods in the time and frequency domains are exploited to this end.
• Their respective merits in terms of accuracy and efficiency are compared.
• The use of stabilisation diagram in the presence of nonlinearity is also introduced.
• The variability of the identification results due to noise is finally studied.

The objective of the present paper is to address the identification of a strongly nonlinear satellite structure. To this end, two nonlinear subspace identification methods formulated in the time and frequency domains are exploited, referred to as the TNSI and FNSI methods, respectively. The modal parameters of the underlying linear structure and the coefficients of the nonlinearities will be estimated by these two approaches based on periodic random measurements. Their respective merits will also be discussed in terms of both accuracy and computational efficiency and the use of stabilisation diagrams in nonlinear system identification will be introduced. The application of interest is the SmallSat spacecraft developed by EADS-Astrium, which possesses an impact-type nonlinear device consisting of eight mechanical stops limiting the motion of an inertia wheel mounted on an elastomeric interface. This application is challenging for several reasons including the non-smooth nature of the nonlinearities, high modal density and high non-proportional damping.