均值优化经验模态分解及其在转子故障诊断中的应用

经验模态分解（Empirical mode decomposition，EMD）作为一种自适应的信号分解方法已经被广泛应用于诸多工程领域。为了提高EMD的分解性能，分别考虑从不同权值均值曲线的迭代筛分结果中选择正交性最小以及从每层内禀模态函数迭代结果中选择最优以保证整体分解最优，发展了两种均值优化经验模态分解（Mean-optimized empirical mode decomposition，MOEMD）算法。通过仿真信号分析，将MOEMD方法与EMD等现有信号分解方法进行了对比，结果表明，MOEMD方法在分解性能和分解精度方面比EMD等方法有显著提高。最后，将MOEMD方法应用于转子碰摩故障信号分析，并与EMD进行了对比分析，结果表明，MOEMD方法不仅能够有效地识别转子碰摩故障，而且识别效果优于EMD方法。

Mean-optimized Empirical Mode Decomposition and Its Application in Rotor Fault Diagnosis

Empirical mode decomposition (EMD), as an adaptive signal decomposition method, has been widely applied to many engineering fields. Two kinds of mean optimized empirical mode decomposition (MOEMD) methods are developed to improve the decomposition performance of EMD. The first method is based on selecting decomposition results with the smallest orthogonality that obtained by different weighted mean curve iteration sifting as the optimal one and the second is selecting the optimal intrinsic mode function in each layer from the results obtained by different weighted mean curve iteration sifting to ensure an overall optimal decomposition result. The two MOEMD algorithms are compared with EMD and other existing signal decomposing methods by analyzing simulation signals and the comparison results indicate that the proposed methods have much more significantly improvement in decomposition performance and precision than EMD and other methods. Finally, the proposed methods are applied to the rotor rubbing fault signal analysis by comparing with EMD and the results show that MOEMD method could effectively identify the rotor rubbing fault and get much better recognition effect than EMD.