中文摘要：

水轮机轴系摆度信号蕴含着丰富的水轮机状态信息，对水轮机监测有必要提取摆度信号的特征参量。信号的经验模态分解方法可以理解为以信号极值特征尺度为参量的时空滤波过程，此过程充分保留了摆度信号的非线性非平稳特征，有效分离了信号与水轮机运行背景噪声，并在EMD分解基础上对摆度信号进行经典的快速傅里叶变换。通过结合EMD和啉法，得到了与以往单一FFT不同的分析效果，大大提高了水轮机摆度监测的准确性。为水轮发电机组在线监测系统提供了一种新的分析方法。

英文摘要：

Hydroturbine shaft swing signal contains a wealth of state information, it is necessary to extract the signal parameters of swing characteristics to monitor the turbine. The empirical mode decomposition method of signal can be understood as a signal characteristic scale of the extreme spatial and temporal filtering process parameters. This process retains the full non-linear non-stationary characteristics of swing signal, effectively separates the signal from background noise of turbine operation, and processes the classic fast Fourier transform on swing signal according to the EMD decomposition. Significant effect has been obtained through a combination of EMD and FFT method comparing to previous single FFT analysis, which has greatly improved the accuracy of turbine swing monitoring. A new analysis method has been provided effectively for on-line monitoring system of the hydro-generating unit.