Blind Source Separation of Gearbox Fault Signals under Impulse Noise

Abstract

Aiming at the shortage of blind source separation (BSS) in processing gearbox fault signals under α stable distribution impulse noise, a new BSS algorithm based on fractional lower order (FLO) S time-frequency distribution was proposed in this paper. In this algorithm fault signals of gearbox were prewhitened in FLO subspace, then S time frequency transform for prewhitened lower-order signals were performed, finally the source signals were restored via joint approximate diagonalization. Simulation results show that the proposed algorithm can effectively restrain impulse noise influence, the BSS effect is good and robust.