Aeroengine Rotor Fault Diagnosis Based On Improved EWT Feature Fusion

With the continuous development of science and technology and the improvement of people’s living standards,the requirement for the safety and stability of aircraft is increasing.As the power source of aircraft,aeroengine is composed of various complex rotating machinery,which undertakes the important task of ensuring the smooth flight of aircraft.As the core of aero-engine,the rotor system of aero-engine is prone to failure and flight accident caused by complex vibration under high temperature and high pressure.Therefore,the research on the early fault diagnosis method of aeroengine rotor system is of great significance for ensuring the safety and smooth flight of aircraft.Due to the characteristics of vibration signal of aeroengine rotor fault early is very weak and drowned in the strong noise environment,it is difficult to extract the fault feature information.In view of this problem,the research contents of this paper include:(1)The common fault types of aeroengine rotor are introduced and analyzed in detail.The theory and limitations of empirical wavelet transform(EWT)are analyzed.An improved empirical wavelet transform method is proposed and its advantages over the original empirical wavelet transform are proved.(2)A hybrid domain high-dimensional fault feature set based on time-domain and frequency-domain feature set,autoregressive(AR)model coefficient feature set and power spectral entropy feature set is proposed to reflect the fault characteristics more comprehensively in many aspects.The improved empirical wavelet transform is combined with hybrid domain feature set,and the manifold learning algorithm of linear local tangent space arrangement(LLTSA)is combined to reconstruct the high-dimensional fault feature set in low-dimensional manifold to achieve fault feature fusion reduction.A feature extraction method based on improved EWT hybrid domain feature set-LLTSA is proposed.(3)Support Vector Machine(SVM)is used to classify the feature extraction results for fault diagnosis of aero-engine rotor.The grid search algorithm is used to optimize the parameters of support vector machine,and the fault diagnosis model of improved EWT hybrid domain feature set-LLTSA-SVM is established.Experiments show that the fault diagnosis accuracy of the model reaches 98.36%,which meets the requirements of fault diagnosis.The validity of the fault diagnosis model is verified.Compared with the fault diagnosis model of EMD hybrid domain feature set-LLTSA-SVM and the fault diagnosis model of data direct processing hybrid domain feature set-LLTSA-SVM,the superiority of the model is verified.(4)GUIDE based on MATLAB platform has developed the graphical user interface of fault diagnosis system,which makes the whole process of fault diagnosis more intuitive and more convenient to operate.