Research On Recognition And Separation Technology Of Aero-engine Multi-mixed Weak Vibration Signals

The research of weak vibration signal processing methods is an active topic all the time and it iscrucial for the vibration monitoring and fault diagnosis of the equipments. The weak early faults ofaero-engine often mix with multiple vibration sources and interference signals, so it is very difficultfor weak vibration signals of aero-engine early faults to effectively recognize and separate. At present,the vibration signal processing methods are rarely applied to the recognition and separation ofaero-engine weak vibration signals. Especially the usage of blind source separation (BSS) technologyin aero-engine multi-mixed vibration signals is still in the initial stage. Therefore, the further study ofrecognition and separation about aero-engine multi-mixed weak vibration signals has importanttheoretical meaning and practical engineering value.For the weak characteristics of aero-engine early faults and the multi-mixed characteristics ofvibration signals, in this article, a further study on the existing weak vibration signal processingmethods and BSS algorithms was done and some new methods were proposed through the analysis ofthe interference characteristics removal, weak feature extraction and weak vibration source separation.The main research contents of this paper include:(1) The basic theoretical knowledge of BSS was summarized in details. Three kinds of typicalblind source separation algorithms were studied systematically and the BSS algorithm which suit forthe aero-engine vibration signals was identified.(2) A joint interference removal method based on median-singular value decomposition wasproposed. The impulse noise could be removed by the median filter, the random noise could beinhibited by singular value decomposition (SVD) and the improved energy difference spectrum can beused to determine the de-nosing order of SVD. The median filter and improved SVD de-noisingmethod were combined to reduce the impulse noise and random noise, and then the useful signalsincluding the weak fault signals were extracted effectively. Through the simulation and theengineering application in aero-engine test vibration signals, the joint method was verified to beeffective. The joint method was applied to the empirical mode decomposition (EMD) to solve theproblem of mode mixing and end effect.(3) A new kind of weak vibration signal extracting method based on independent componentanalysis (ICA) was proposed. Using the good separation characteristic of ICA, the simulation analysisof extracting weak vibration signals from the strong noise or large signal interference was done respectively and the method was verified to be validity through the experiment. Combining EMD andICA method, the prior knowledge of noises was obtained; Combining exploring phase and correlationanalysis method, the initial phase was estimated.(4) For the multi-mixed characteristic of aero-engine weak vibration signals and the noiseinterference to BSS algorithms, a new BSS method based on the multiple time-delay autocorrelationde-noising was proposed. Combining the time-delay autocorrelation and BSS method, theindependent source signals were separated out from the mixing signals after the time-delayautocorrelation de-noising. Through the simulation analysis, the experiment data analysis of rubbingfault rotor and the application analysis of aero-engine test vibration signals, the BSS method based onthe multiple time-delay autocorrelation de-noising was verified to be effective and applied.