集合经验模态分解的稳健滤波方法研究

A method of robust filtering based on EEMD

为了消除野值和噪声信号对观测数据的影响,给出一种基于集合经验分解的具有稳健性的滤波算法:首先用滑动中值滤波算法剔除原始数据中的野值,然后采用集合经验模态分解算法,抑制数据中的噪声。数值仿真和实际工程应用表明,该方法不仅能剔除野值,抑制信号中的噪声,提高信噪比,还能够有效消除模态混叠问题,将被测信号中不同的频率成分独立分解在不同的固有模态函数中,从而得到更清晰的时频分布,有利于实际数据处理中的信号分析和故障诊断。

In order to eliminate outliers and noise impact on observed data, a robust filtering algorithm based on ensemble empirical mode decomposition （EEMD） was proposed. Firstly, with the sliding median filtering the outliers in raw data were removed. Then, EEMD algorithm was used to suppress noise in the data. Numerical simulations and actual engineering applications showed that the method can not only eliminate outliers, suppress noise and improve signal to noise ratio, but also effectively solve the problem of modes aliasing, independently decompose a measured signal into different IMFs to get a clearer time-frequency distribution, so it is useful for signal analysis and fault diagnosis in actual data processing.