A Novel Online Sliding Blind Deconvolution Algorithm for Satellite Microvibration Source Separation in Time-Varying Environment
Xin Luo, Zhousuo Zhang · IEEE Transactions on Instrumentation and Measurement · 2024
Online identification of satellite micro-vibration sources can provide a basis for online suppression of satellite micro-vibration, which has significant application value for improving the positioning accuracy and resolution of satellite. The complex time-varying environments in orbit, high real-time requirements and numerous parameters to be estimated pose great challenges to the adaptability, efficiency and accuracy of online source identification algorithms. Aiming at the challenges for online satellite source identification, a novel online sliding blind deconvolution algorithm is proposed. Firstly, blind deconvolution is converted online to the instantaneous blind source separation (BSS) in the selected single frequency bin by sliding discrete Fourier transform (SDFT), which greatly reduces the computation and improves the efficiency of online separation. Secondly, my previous work, namely adaptive step size EASI algorithm based on nonlinear correlation (NC-EASI algorithm), is extended to the complex-domain to achieve the online separation. Finally, the separated signals in time-domain are obtained by performing the inverse SDFT. The effectiveness of the proposed algorithm under time-varying environment is validated by several numerical simulations and the excitation experiments of aluminum honeycomb panel cabin structure. Compared with the comparison algorithms, the separation accuracy of simulation and experiment in time-varying environments has been improved by an average of 4.295dB and 6.64dB, respectively, and the convergence time has been shortened by an average of 44.42% and 65.11%. The proposed algorithm can provide an effective means for the online source separation of complex mechanical systems.