Based On EEMD And SVM Gearbox Fault Diagnosis And Identification

As a necessary component of mechanical equipment,gearbox is one of the important tools to transmit mechanical movement and power.Gearbox can make its internal parts failure and will delay the production efficiency heavily under the condition of high load and high speed working space for a long time.Therefore,it is particularly important for gearbox to achieve the fault diagnosis accurately and effectively.Through the Ensemble Empirical Mode Decomposition(EEMD)combined with Hilbert-Huang Transform(HHT)on fault information extraction and analysis,gearbox fault diagnosis is implemented.At the same time,by using the EEMD combined with Support Vector Machine(SVM),the gearbox achieves fault identification.Both of the methods achieve very good results,the gearbox fault diagnosis and identification are implemented.In this paper,the end effect and mode mixing of Empirical Mode Decomposition(EMD)are restrained effectively and the effective noise reduction and the vibration signal characteristic frequency decomposition are implemented through the analysis of the vibration signal of gear box and processing non-stationary and nonlinear properties of gearbox vibration signal by EEMD method.The actual fault is simulated by using the Gearbox Dynamics Simulator.The time-domain waveform,power spectrum chart,EEMD decomposition waveform,the Hilbert marginal spectrum and other time-frequency analysis methods are used to realize the preliminary diagnosis of the gearbox fault effectively.A series of Intrinsic Mode Function(IMF)is got based on decomposing vibration signal by EEMD method.The energy entropy of the IMF components,which has high correlation with the original signal and can reflect the characteristic information of the original signal,is selected by using the correlation analysis and is input to the SVM as the feature vector.Because of the advantage of SVM that is not dependent on a large number of sample size,each 40 sets of samples under three different conditions are selected as the training samples.The inherent law under three kinds of working conditions is learned through SVM,and the test is carried out by using 10 sets of samples as the test samples.According to three different types of gearbox working states,the highly identification rate is achieved so as to realize the identification of gearbox fault.