分析了经验模式分解的滤波性能；针对旋转机械振动信号信噪比低及非平稳特性，应用经验模式分解对振动信号进行降噪处理，突出了有用振动信号，为旋转机械在线监测和故障诊断提供了有效的依据；仿真实验及真实数据分析表明，经验模式分解在振动信号降噪处理中是有效的，为振动信号分析提供了新的方法。

The filtering property of empirical mode decomposition is analyzed in the paper. Aimed at the low signal/noise ratio and non stationary feature of vibration signal of cyclic machine, EMD is introduced to the noise reduction of vibration signal and the useful signal is given prominence efficiently, which offers the more efficient foundation to monitor on line and fault diagnosis of cyclic machine. By the simulation and application, it shows that EMD is very useful in reducing noise and provides new means of vibration signal analyzing.