关键词: 往复机械, 信号降噪, 特征提取, 过零率分析, 总体平均经验模式分解, 能量矩

Abstract: According to the instantaneous nonlinear and non-stationary characteristics of the vibration signals from reciprocating machine with fault，a novel adaptive denoising method based on Ensemble Empirical Mode Decomposition（EEMD） and zero-crossing detection is proposed and combined with energy moment and Support Vector Machine（SVM） to apply in fault diagnosis.With the method of EEMD，the non-stationary vibration signals are adaptively decomposed into a finite number of Intrinsic Mode Function（IMF），which can alleviate model mixing that may appear in conventional EMD method.It calculates the zero-crossing ratio of every IMF components and compares them to the predetermined threshold value，the IMF components which are satisfied for request of threshold value are obtained.That means make zero-crossing rate serve as the criterion to separate desirable components from jamming ones.So the denoised signal is obtained through reconstructing desirable IMF components.Otherwise，the energy moments of desirable IMF components are extracted as the input vector of Binary Tree Support Vector Machine（BTSVM） to realize the fault diagnosis of diesel engine，which validates the effectiveness of the method.

Key words: reciprocating machinery, signal denoising, fault feature extraction, zero-crossing rate, Ensemble Empirical Mode Decomposition（EEMD）, energy moment