Time-varying frequency parameter identification of space solar power satellites based on an improved recursive subspace algorithm and optimal sensor placement

Abstract

This study focuses on the time-varying modal frequency identification of space solar power satellites. Accordingly, an improved recursive subspace identification algorithm based on the variable forgetting factor is proposed to enhance the tracking adaptability of the original method in time-varying systems. Considering the changes in the structural configuration induced by the rotation of the solar concentrator mirror, a time-varying attitude-vibration coupling dynamic model of a space solar power satellite with an integrated symmetrical concentrator configuration is developed based on the modal synthesis technology and the substructure method. Moreover, based on the projection subspace theory, the time-varying forgetting factor is determined by minimizing the mean square deviation of the system estimator. Subsequently, the variable forgetting factor subspace algorithm is adopted to recursively identify the time-varying pseudo modal frequency parameters. A finite element model of the space solar power satellite is established via numerical simulations, and several optimal sensor placement methods are applied to determine the positions of the vibration sensors. Based on the output signals obtained from the optimal sensor placement results, the time-varying frequency parameters of the system are obtained using the proposed algorithm. The computational results reveal that a decrease in the average relative error by 4.2% compared with that obtained via the conventional recursive subspace method is observed when the signal-to-noise-ratio of the measured output is selected as 15 dB. In addition, the results also reveal that an average relative error reduction of approximately 0.4%–1.4%, over two other methods with a fixed forgetting factor, can be achieved using the proposed algorithm when the rotation speed of the solar concentrator mirror is increased. The findings of this study confirm that the proposed algorithm demonstrates a high noise-immune ability and tracking performance for fast time-varying systems.