Roller bearing envelope analysis method based on local mean decomposition (LMD) and kurtogram was proposed. In this method

the original roller bearing fault vibration signal was decomposed into a set of frequency components and preliminary reduced noises adaptively by LMD

and the parameters of band-pass filter in envelope analysis were provided objectively by kurtogram

thus the envelope analysis accuracy for roller bearing can be enhanced. The analytical results from roller bearing simulation signals and experiment signals indicate that the natural vibration components of roller bearing can be separated by LMD adaptively and the parameters of band-pass filter in envelope analysis can be determined by kurtogram automatically in low signal-to-noise ratio. Compared with traditional envelope analysis

the proposed method can extract the fault features of roller bearing more clearly and accurately.