Research On Blind Soure Separation Of Muti-mixed Vibration Signal And Its Experiment

Blind source separation (BSS) is a new signal processing method based on neural network. BSS have an active functioning impelled on not only the signal processing researching but also the development of neural network's theory. BSS extended to other non-stationary signals and fault diagnosis is only the beginning. This paper deals with the research on blind source separation's method that is known. Have more excellent computation efficiency through mend the BSS algorithm based on Jacobi optimization in this paper. At the same time, this paper research on post-nonlinear BSS, Post-nonlinear is a weak nonlinerity.Nonlinear BSS have more good robust and separation performance through research the MISEP algorithm.In order to prove BSS's practicability and applicability, this paper finished the experiment of muti-mixtures vibration signal on the combustor of some aero-engine. This experiment not only builds a foundation for BSS's real application but also provide a new method for fault diagnosis.This paper uses the theory of design on parameters to achieve the software package of BSS and the stator vane's vibration and fatigue analysis software, which realize the process of stator vane's vibration and fatigue use a key to control.The formulas and improving algorithm of BSS in this paper, which have reference value for new researcher.The software package of BSS and the stator vane's vibration and fatigue analysis software as the analysis tools can be operate more easily.