改进的卷积型小波包分解及在故障诊断中的应用

Improved Convolution-Type Wavelet Packet Decomposition with Applications to Fault Diagnosis

针对卷积型小波包分解存在频带错位与频带重叠缺陷,提出了一种改进的卷积型小波包分解算法。该算法通过交换偶数位置节点小波包分解后的两节点顺序来消除频带错位缺陷,引入两算子分别从频域除去低、高频子带理想通带范围外的频率成分以消除频带重叠缺陷。由构造的故障信号进行仿真实验,并使用某直升机中减速器疲劳实验的故障数据进行了验证。结果表明:由于消除了卷积型小波包和内积型小波包分解算法中广泛存在的频率折叠、频带重叠和频带错位缺陷,改造的卷积型小波包分解算法能更方便、更有效地提取隐藏在强噪声和其他强干扰背景下的故障特征信息,从而为机械故障的诊断提供了一种强有力的分析手段。

Aiming at frequency band derangement and overlap in convolution-type wavelet packet, an improved one is proposed. Exchanging the order of sub-nodes originated from father nodes sequenced in even serial numbers, the frequency band derangement is removed. Introducing two operators eliminating the frequencies outside the pass bands of low and high frequency sub-bands, the frequency band overlap is forced to be avoided. The simulated fault signals and endurance test datasets of a helicopter intermediate gearbox are respectively collected to certificate the improved convolution-type wavelet packet. The results show that the improved convolution-type wavelet packet enables to expediently and effectively detect the fault features submerged in strong interference owing to the elimination of frequency aliasing, frequency band overlap and derange- ment.