LCD Denoise and Full Vector Mutual Information in Application of Gear Fault Diagnosis Under Different Working Conditions

Different conditions will change the amplitude and frequency modulation characteristics of gear vibration signals and then lead the miscalculation of the gear working state. In view of this, a combination of full vector mutual information and local characteristic-scale decomposition(LCD)method is put forward. First, the vibration signal of different working conditions is decomposed by use of LCD, to obtain the Intrinsic scale component (ISC)whose instantaneous frequency with physical significance. At the same time, it eliminates the interference caused by frequency modulation and modal aliasing effect. Then, calculate of the cross correlation coefficient between ISC and the original signal and achieve noise reduction by the principle of maximizing cross correlation coefficient. At last, the sum of absolute vector mutual information was calculated between the sample under different working condition and the denoised ISC as the characteristics to classify by use of support vector machine(SVM). After the 100 groups signal recognition test in the different conditions, its showed that the method can effectively distinguish weak gear fault characteristics under different working conditions, as well as reduce dependence on the man's subjective experience.