多频信号经验模态分解的理论研究及应用

Theoretical research on empirical mode decomposition of multi-frequency signal and its application

对多频信号的经验模式分解进行了理论分析 ,分析表明分解结果取决于各分量的频率和幅值。得到了一个正确分解的充分条件 ,仿真试验证明了这一充分条件。将经验模式分解用于分析一个包含故障轴承的实际齿轮箱表面振动信号 ,分析结果表明在经验模式分解得到的固有模态函数中 ,轴承故障特征信息比原始信号相比得到显著增强。

Characteristic frequencies of vibration signal are important information of fault diagnosis. Weak characteristic vibration is often flooded by other strong vibration. A new method, empirical mode decomposition, is applied to decomposition of mixed vibration signals from different sources. The empirical mode decomposition of multiple frequency signals is theoretically investigated. It is found that the output of empirical mode decomposition is decided by amplitude and frequency of different components. A sufficient condition of proper decomposition is deduced. Numerical results of signal with different frequency signals are given to testify the correctness of the sufficient condition. The empirical mode decomposition is applied to vibration data measured from the case of a real gearbox consisting a fault bearing. The results show that characteristic information of faulted bearing is much more distinct in the intrinsic mode function from empirical mode decomposition than in the original data.