Research On Identification Method Of Typical Coupling Fault Feature Of Aero Engine

Rubbing refers to the insufficient clearance and static parts between continuous intermittent contact behavior of the rotating machinery in the rotating process of rotating parts.In order to improve the performance and efficiency,aero engine makes the radial clearance between rotor and stator become smaller and smaller which leads to rotor and stator rub fault mechanism and system response induced by rub touch are so complex.Moreover,once the rubbing fault occurred,has not been detected,will bring very serious consequences for machinery.Therefore,to study the coupling rubbing fault has important significance.This paper studies the rotor experimental platform based on single span rotor bearing coupling system on the rubbing fault,carries on the simple analysis of the mechanism,presents the feature extraction method based on EMD and BP network and BP_Adaboost feature recognition method.Fault diagnosis usually pursuits stable fault recognition rates, the point is to be able to extract the key features of the fault signal and effective classification identification methods.The EMD method is based on local characteristics of data in time domain. It can depose the complex data into limited data which is usually a very small number of intrinsic mode functions(IMFs). This kind of method has the advantages of strong adaptability and high work efficiency, what's more it can intuitively show overlapping complex intrinsic mode signal which makes the nonlinear variational problem particularly effective. This paper discusses the EMD decomposition theory in detail and uses the EMD decomposition to extract the fault characteristics of the coupled vibration signal.Classification and identification methods of fault diagnosis select the BP neural network and structured BP_Adaboost classifier, this two methods are minutely introduced in basic theories and characteristics. On the basis of rub and impact fault diagnosis analysis, this paper selects the energy of signal frequency in different frequency bands as characteristic value of each failure mode, uses characteristic value by BP neural network and BP AdaBoost model to pattern classification. Through comparative analysis on the experimental results of this twomethods, it comes to the fault diagnosis. The experimental results show that the fault feature recognition method based on BP_Adaboost model has a high accuracy, and it can be used in many fields because of its strong fault tolerance.