Approach and Application of Semi-Blind Source Separation for Aero-Engine Vibration Signals Using ICA-R

Abstract

In this paper, the approach and application of semi-blind source separation (SBSS) in aero-engine vibration signal is studied. Firstly, the features of aero-engine vibration signal and difficulties for blind source separation (BSS) are summarized, and the SBSS incorporated the available prior knowledge is match to the goal of signal processing. Then, the ICA with reference (ICA-R) algorithm based on classical FastICA is introduced, with Newton iteration and gradient descent iteration approach to obtain optimal solution. The unique parameters in ICA-R for aero-engine vibration signal are also provided. Finally, the efficacy and the accuracy of the ICA-R algorithm are verified by numerical simulations and real engine vibration signals. The approach of SBSS in this paper perfectly suited to handle aero-engine vibration source separation and it lead to efficient implementation in fault diagnosis.