基于小波包的经验模态分解法的研究及应用

Wavelet Packet Analysis-based Empirical Mode Decomposition and Its Application

针对多频信号经验模态分解中的模态混成现象,提出了一种经验模态分解与小波包分解相结合的新方法。经验模态分解方法基于信号的局部特征时间尺度,把复杂的信号函数分解为有限的固有模态函数之和。利用最小Shannon熵准则,对出现混成模态的固有模态函数进行小波包分解,并根据小波包分解后各频带信号的频率分布特征,选择能量比重较大的频带信号进行重构,将重构信号作为新的固有模态函数分量。仿真信号和齿轮箱故障实测数据表明,新方法能将不同的频率成分提取出来,从而提高了经验模态分解的分解能力。

In allusion to the phenomena of mode mixing in empirical mode decomposition （EMD） of multi-frequency signal, a novel method combining EMD and wavelet packet decomposition was introduced. The methodology developed herein decomposed the original times series data into intrinsic mode functions, using the empirical mode decomposition. The wavelet packet decomposition based on Shannon entropy was applied into the selected intrinsic mode function （IMF） in which the mixing mode existed. According to the frequency distribution characteristics of components from wavelet packet decomposition and the components ＇ energy proportion, the dominating components were extracted, then restructured . The restructured signal was noted as sub-IMF. Finally, the examples of emulated signal experimentation and gear box fault diagnosis were given and the result shows this method is effective.