Abstract:To extract the fault characteristics of rolling bearings of the railway vehicles, whose fault signal is usually modulated to high frequency with lots of noise, this paper presents a method combining EMD( Empirical Mode Decomposition)and adaptive generalized morphological filtering based on LMS(least mean square). It first uses EMD to acquire high frequency signal and separates low frequency interference and noise. Then it uses the LMS morphological filtering and closed operation method to demodulate forms. At last, the fault characteristics are extracted through frequency analysis. Simulation experiments and practical examples have proved that this method can effectively eliminate lots of noise and low frequency interference to extract fault characteristics of rolling bearings. This method has good practical value with high calculation speed and operation convenience.