中文摘要：

针对水轮机尾水管动态特征信息提取问题,本文利用具有近似平移不变性、完全重构性、以及能量集中优点的二元树复小波变换作为信息提取的工具,并利用信息熵能够反映系统信号中短暂的异常信号的特点,将二元树复小波分解系数和信息熵相结合,取复合信息的特征熵作为故障模式识别的特征矢量。以水轮机尾水管压力脉动信号为例,运用此方法进行了尾水管动态特征信息的提取。试验表明基于二元树复小波特征熵的特征提取法是故障特征提取的有效方法,为流体机械故障诊断开创了新思路。

英文摘要：

This paper adopts a technique of dual-tree wavelet transformation as information extracting tool to study the dynamic behaviors of turbine draft tube. This technique has advantages of approximate shift invariance, complete reconstruction and better energy focus. As information entropy reflects abnormal characteristics of short-term system signal, combination of wavelet coefficients and information entropy provides a device for recognition of fault pattern with characteristic entropy used as recognition feature vector. This new method was applied to the monitored signal of pressure fluctuation in a draft tube. Results show that it is effective in extracting fault information. Thus, it opens up a new direction for fault diagnosis of hydraulic machinery.