Modified Adaptive Local Iterative Filtering Method And Its Application To Gearbox Fault Diagnosis

The gearbox fault diagnosis method based on the vibration characteristics is widely used for its convenience,reliability and real time.But the vibration signal of gearbox fault is nonlinear,nonstationary and modulated,which results in the difficulty in extracting the vibration characteristics of gearbox fault,so appropriate signal processing method is needed for processing these complicated signals.These common signal processing methods such as Complementary Ensemble Empirical Mode Decomposition(CEEMD),Modified Ensemble Empirical mode decomposition(CEALIF)and Adaptive Local Iterative Filtering(ALIF)are continuously modified,but there still exist some disadvantages in these methods.An adaptive local iterative filtering method modified by sifting through permutation entropy which is based on ALIF is proposed,and using this signal processing method to realize the diagnosis of gearbox fault by combining 1 Dimensional Convolutional Neural Networks(1D CNN).The main research contents are as follows:(1)The theory of the ALIF method is researched.By comparing the decomposition results using EEMD,LCD and LMD methods of the simulation signals and gearbox failure experimental signals,the accuracy and its ability to resistant to the noise interference on decomposition are proved.(2)The procedure of Complementary Ensemble Adaptive Local Iterative Filtering(CEALIF)is researched.By comparing the decomposition results using ALIF,CEEMD methods of the simulation signals and gearbox failure experimental signals,proved its advantages on suppressing the model mixing problem and other aspects.(3)Based on ALIF,a new signal decomposition method named ALIF-PE is proposed,which uses permutation entropy to sift abnormal components.The practicality of using permutation entropy to sift abnormal components in decomposition results is proved.By comparing the decomposition results of the simulation signals and gearbox failure experimental signals using CEALIF,CEEMD and CEEMDAN methods,the advantages of ALIF-PE are proved.(4)A gearbox fault diagnosis method based on ALIF-PE and 1D CNN is proposed,which uses ALIF-PE method to decompose the gearbox fault signal firstly,and then input those components with the largest correlation coefficient into 1D CNN to complete the training procedure.Its high ratio of fault recognition and advantages comparing with other diagnosis methods are proved.