中文摘要：

针对小波包分解广泛存在频率折叠、频带重叠与频带错位缺陷，分析了其产生的根源，并以此提出了一种改进的冗余小波包分解算法。该算法通过交换偶数位置子带小波包分解后的两子带顺序来消除频带错位缺陷，通过引入两算子依据傅里叶变换滤波原理分别从频域滤去低、高频子带理论频率范围外的频率成分来避免频带重叠缺陷。分别使用仿真信号与某直升机中减速器疲劳试验的故障数据对该算法进行了仿真验证与试验验证。分析结果表明：相对于Mallat小波包算法和通常的冗余小波包算法，改进的冗余小波包分解算法确实成功消除了频率折叠、频带重叠和频带错位等三类缺陷，因此该算法能更有效地提取淹没在强噪声和其他强干扰背景下微弱故障特征，具有一定的工程应用价值。

英文摘要：

Aimed at the problems of the frequency aliasing, frequency band overlap and derangement in wavelet packet, the reason to cause the problems is researched and an improved decomposition algorithm of redundant wavelet packet is proposed. The problem of frequency band derangement is overcome by exchanging the order of sub-nodes originated from father nodes which are sequenced in even serial numbers, and the problem of frequency band overlap is avoided by introducing two operators which can eliminate respectively the frequencies outside the pass bands of low frequency and high frequency sub-band. The simulated fault signals and endurance test datasets of a helicopter intermediate gearbox are respectively collected to certificate the improved decomposition algorithm. The simulation and analysis of the gearbox fault dataset shows that, compared with the Mallat wavelet packet algorithm and traditional redundant wavelet packet algorithm, the improved redundant wavelet packet can more expediently and effectively detects the fault feature submerged the strong interference context, owing to the elimination of frequency aliasing, frequency band overlap and derangement. So, it has much application value in fault diagnosis.