一种EMD改进方法及其在旋转机械故障诊断中的应用

针对经验模态分解(Empirical Mode Decomposition, EMD)中存在的端点效应问题，提出一种波形特征匹配延拓与余弦窗函数相结合的改进方法。首先对信号进行波形特征匹配延拓，实现延拓数据与原信号交界处的光滑过渡，避免边界处瞬时频率的跳跃；其次针对该延拓方法存在延拓误差的问题，对信号加余弦窗处理，将延拓误差控制在信号两端，使其无法(或以较慢速度)向数据内部发展，保证信号有效数据的正确分解，提高信号的分解精度，实现EMD算法的改进。通过仿真分析和不对中故障诊断实例研究表明，该方法能较好地抑制EMD端点效应，实现旋转机械故障的有效诊断。

An improved method of EMD and its applications in rotating machinery fault diagnosis

For the end effect of empirical mode decomposition(EMD), a novel integrated method that combines waveform feature matching extension and cosine window function is proposed. Firstly, waveform feature matching extension achieves a smooth transition at the junction of the extension and the original signal, and avoids the instantaneous frequency jump at the boundary. Secondly, for extension error exists in the extension method, the signal is processed by cosine window function. Thus the error is controlled at both ends so that it can not (or at a slower speed) spread to the internal signal. This ensures the correct decomposition of the effective data, improves the decomposition accuracy and achieves the EMD algorithm improvement. Experimental results and misalignment fault diagnosis example show that the improved method can inhibit end effect effectively.