Study Of Technology Of Feature Extraction Of Gearbox Fault

As a necessary part of connecting and force transmission in mechanical equipment, gearbox system play a comprehensive role, with the development of modern industry. When the gear system works, periodic impulse impact force will produce owing to the faults of gear abruption or corrosive pitting, etc. So phenomena of modulation produced at this moment and spectrum of the vibration signal has symmetrical modulation side-bands besides the meshing frequency or natural frequency. Extracted modulation information and its strength and modulation frequency order can be analyzed which can be used to judge the fault parts. It is a compressively used in mechanical fault diagnosis.In order to effectively extract the feature vector of fault in the signals of gear and bear vibration and achieve the purposes of identifying and eliminating the faults, according to the mechanism and characteristics of gearbox vibration, the following researches on the basis of the wavelet analysis and cyclostationary demodulation:1) The development of technology of fault diagnosis for gearbox, the mechanism and characteristics of gear and bear vibration are discussed.2) The fundamental theory of wavelet analysis and its application are introduced on the application of fault diagnosis for gearbox. Then by means of the wavelet analysis and wavelet packet transform, the fault feature vectors of fault are successfully extracted and this effective method is employed to identify the fault pattern.3) A power spectrum analysis method based on the phase of complex- wavelet-transform, which derives fault information from gear vibration signal by phase, is presented.4) A new conclusion is discussed that cyclic autocorrelation function possesses demodulation capability. The superiority of the demodulation is verified by the simulation.5) A new diagnosis and analysis method were proposed, which was wavelet packet-cyclostationary demodulation, based on the organic combination of the noise elimination ability of wavelet packet transform and cyclostationary demodulation.Experimental results show these proposed methods, which can protrude the fault character of the gearbox, are practical in gearbox local fault diagnosis and of important meaning.