Research On Fault Diagnosis Of Rotor Crack Based On Time-frequency Characteristics Of Acoustic Emission Signal

Rotor crack is one of the commonest failure in the fans,steam turbines, generators,compressors and other large rotating machineries.Research and application of fault diagnosis of rotor crack technology can accurately determine the causes,location, severity and its development trend.It has great significance to avoid serious accidents and huge economic losses.It is a bottleneck problem which is hard to identify the fault characteristics from strong background noise in the fault diagnosis of rotor crack with traditional vibration method.Acoustic emission(AE)signals has several advantages,such as characteristic frequency is significantly, failure forecast time is earlier, detection device is simple and ability of anti-noise is Powerful,etc., Using it in feature extraction of early rotor crack can effectively improve the accuracy of rotor fault diagnosis.In this paper,acoustic emission signal of rotor crack was tooked as a research object,theoretical analysis and experimental study were combined,fault feature extr- action and diagnostic methods of rotor crack was studied.Details are as follows:1.Taking fracture mechanics and plate wave theory as an entry point,study the acoustic emission mechanism,rules and characteristics of crack propagation.Geted the relationship between the acoustic emission signal and the crack extension, established a acoustic emission model of rotor crack.2.Denmark SWAES full-waveform acoustic emission monitor and BeiJing ShengHua's analysis software are used as analysis tools,America SpectraQuest'rotor machinery fault simulator is used as operating environment of cracked rotor, Carried out acoustic emission detection experiments under different conditions of rotor crack fault and studied the character- istics of AE signals caused by rotor crack fault.3.Wavelet packet analysis method,short time fourier transform(STFT)analysis method,pseudo-Wigner-Vill distribution ( PWVD ) analysis method,Hilbert-Huang transform(HHT)analysis method were used in feature extraction of AE signal of rotor crack.Studies have shown that these time-frequency analysis method can effectively extract the AE signal characteristics and the hidden characteristics information.4. Based on analyzing of the classification principles of support vector machine and the methods how to select nuclear function and its parameters, C-SVM algorithm model was used and the RBF kernel was Selected as a kernel function,"one to many" multi-class classification method was adopted to establish a rotor crack multi-fault classifier.Different crack depths of the rotor were recognized effectively.The feasibility that acoustic emission technology is used to fault diagnosis of rotor cracks was further verified.This study can not only improve the efficiencies and accuracies of incipient faults prediction and diagnosis of rotor crack,but also favorable to promote the waveforms analysis techniques of AE signals of rotor crack.The study provides a new idea for using acoustic emission technology in on-line detection and fault diagnosis of rotor crack fault.