中文摘要：

经验模态分解广泛应用于机械故障诊断，但其模态混淆问题影响了方法的有效性。从机械故障诊断的角度提出了EMD模态混淆的基本概念，根据表现形式将模态混淆定义为“向下”、“交叉”和“向上”3种基本类型，通过深入研究经验模态分解的模态混淆问题，发现产生模态混淆的原因主要有2类：一类是由方法基本原理所导致，有模态向下混淆和模态交叉混淆2种表现形式，称为工类模态混淆问题；一类是由筛分算法缺陷所导致，有模态向上混淆1种表现形式，称为Ⅱ类模态混淆问题。根据模态混淆产生的机理，提出了针对性的解决方案：对于工类模态混淆问题需要引进辅助手段加以解决，例如异常排除法、信号滤波法和辅助信号加入法；对于Ⅱ类模态混淆问题，完善了本征模态函数的定义，改进了筛分算法最后进行了仿真，结果表明所提解决方案有效。

英文摘要：

Empirical mode decomposition （EMD） is widely used in mechanical fault diagnosis, but its inevi- table mode mixing in EMD exerts an influence on effectiveness. Mode mixing in EMD is studied, and two kinds of causes leading to mode mixing are found. One caused by the basic principle of EMD is called type -I mode mixing, such as mode down mixing and mode cross mixing, the other caused by faults of the sifting process is called type-II mode mixing, such as mode upward mixing. The improvement schemes are proposed according to the different mode mixings. For type-I mode mixing , some supplemental measures such as abnormal event elimination, signal filter and supplemental signal added method are necessarily in- troduce in problem solving. For type -II mode mixing, the definition of intrinsic mode function is perfec- ted, and the sifting process is improved. The simulating results show that the proposed schemes are effective.