中文摘要：

针对发电机组碰摩故障特征的提取问题，构造了一种提取该类信号时域特征的改进第2代小波算法。该算法以第2代小波为基础，构造第2代小波等效滤波器，分析其存在的频带交叠问题；采用基于数据的优化方法，设计能够自适应匹配每层分解信号特征的预测器和更新器；对分解得到的每层逼近信号和细节信号采用傅里叶变换方法，消除每个分解频带中其它频带的无关频率成分，以突出相应频带信号的时域特征。仿真信号和工程振动信号分析表明，改进第2代小波算法的信号特征提取效果优于第2代小波方法，较理想地提取出了汽轮发电机组发生径向碰摩故障时的时域故障特征。

英文摘要：

To extract impact-rub fault feature of generation unit, an improved second generation wavelet algorithm is presented to extract fault feature in time domain. This algorithm is based on second generation wavelet, and the equivalent filters are constructed to analyze frequency band aliasing causes. Using data-based optimization algorithm, the prediction operator and update operator which lock on to the dominant signal structure at every level are designed. Applying Fourier transform method, the high-order polynomial signal structure in approximation signal is eliminated and leaving only the low-order signal structure, the low-order polynomial signal structure in detail signal is removed and preserving only the high-order polynomial signal structure, and the signal characteristics in time domain at the corresponding frequency band are well revealed. Simulative and engineering results confirm the advantage of improved second generation wavelet algorithm over second generation wavelet for feature extraction, and the typical features of impact-rub of turbo-generation unit in time domain are desirably extracted by improved second generation wavelet algorithm.