›› 2012, Vol. 32 ›› Issue (1): 27-35.doi: 10.3780/j.issn.1000-758X.2012.01.005

Previous Articles     Next Articles

High Precision Nonlinear FDD Design Using RBF with the Ability to Learn

 WANG  Zheng1, LI  Yan-Jun1, SUN  Xiao-Wei2   

  1. (1SchoolofAstronautics,NorthwesternPolytechnicalUniversity,Xi′an710068)
  • Received:2011-04-18 Revised:2012-02-25 Online:2012-02-25 Published:2012-02-25

Abstract: On-orbit spacecraft fault detection and diagnosis (FDD) problem nonlinear model always needs to deal with. It′s also important to improve the accuracy ofthe methodology as far as possible. Radial basis function (RBF) neural network was firstly applied for compensation of modeling error in order to increase the accuracy of the scheme, a Lyapunov function was used to prove its property. The residual and threshold and fault detection time were got through nonlinear observer. Fault signal was reconstructed by using RBF network, and accordingly a learning FDD strategy was presented. Validations were respectively designed for characteristics of residual threshold, detection time and the ability of RBF for reconstruction. The simulation results show the effectiveness of the proposed algorithm.

Key words: Fault detection and diagnosis, Radial basis function neural network, Simulation, Spacecraft