Feature Extraction Method Analysis And Dynamic Simulation Of Aeroengine Bearing

Aeroengine is the power plant of aircraft. And bearing of main shaft is not only the key component of aeroengine, but also the weak component. Using effective methods to monitor bearing and realize fault early warning is of great significance to ensure the safe of aircraft.This paper was aiming at helping aeroengine to carry out feature extraction and diagnosis of weak fault signal, at the same time, a three-dimensional finite element model was set up and dynamic analysis was carried out.First, feature extraction method based on noise suppression was studied. CEEMD was combined with correlation theory to analyze the fault signal under a variety of speed. And NA-MEMD was used to analyze the signal collected in multiple locations. Then a method based on ITD and PE was put forward and the signal under low speed was analyzed. For the shape of time domain waveform, mathematical morphology was studied, and the influences of structural elements and morphological operator were analyzed.Then, pattern recognition method based on fractal theory was studied. Morphology and correlation dimension was used to pattern recognition of bearing. A method based on box dimension is put forward. And use box dimension as characteristic parameters to recognize bearing fault after combining with distance function and neural network. Then generalized dimension was used to describe the fractal characteristics under multiple measures.Finally, simulation of bearing based on the dynamics was carried out. Finite element model of normal and typical fault were established. Motion state and stress distribution of bearing was analyzed. The characteristics of simulation signal under various speed is recognized.The results show that the feature extraction methods based on noise suppression can effectively reduce the noise of the original signal and pattern recognition method based on fractal theory can identify typical bearing fault after combining with distance function and neural network. Maximum stress is located in the contact surface and increasing with the degree of fault. Stress of each component was reducing from roller to inner, outer and roller cage. Signal extracted from simulation model can identify the fault types by FFT.