中文摘要：

针对实际工程信号易受噪声干扰导致提取的故障特征不明显的问题，将小波改进阈值方法和经验模态分解（EmpiricalModeDecomposition，EMD）相结合，提出一种基于小波改进阈值的经验模态分解去噪方法，并应用到旋转机械故障特征提取中。首先，为了克服传统小波阈值方法在阈值函数的连续性以及重构误差等方面的不足之处，研究小波改进阈值方法并利用其进行振动信号预处理，减少随机噪声对振动信号的干扰，同时减少EMD分解过程中的分解层数以及其边缘效应对有用信号分解质量的影响。在实际应用中，由于振动信号中}昆有多种不同性质的噪声，预消噪处理常常不足以消除全部噪声的干扰，因此有必要用EMD相关度方法适当地消噪后处理，提高故障特征提取的准确度，研究为旋转机械故障进一步识别诊断提供了重要的参考。

英文摘要：

Actual engineering signals are easily interfered by noises, which leads to invisible fault features. Thus, a de- noising method based on improved wavelet threshold and EMD （Empirical Mode Decomposition） is proposed when they are combined, and applied in rotating machinery fault feature exaction. First, compared with the traditional wavelet threshold methods, the improved wavelet threshold method overcomes some shortcomings of the threshold function continuity and reconstruction error, and it was applied to reduce the random noise. At the same time, it could reduce the decomposition layers and end effect of the EMD. In reality, however, there are different kinds of noises in practical applications, which is uncomfortably to reduce all the effects. Therefore, EMD was performed properly on de-noising post-processing to increase the accuracy of fault feature exaction. The research provides an important reference for the further fault diagnosis of rotating machinery.