Abstract:

In order to on-line process engine cylinder head vibration signals of mobile equipment, a de-noising method with gradual changing threshold based on lifting wavelet package transform was proposed. The lifting wavelet package transform based on interpolating subdivision was introduced. The engine cylinder head vibration signal’s time-frequency characteristics were analyzed. To fit in with the features of signal and noise, a strategy of gradual changing threshold selection and quantization was presented. The evaluation criterion of de-noising effect based on separability measurement of feature classes was introduced, and the quantitative evaluation of de-noising effect for the real vibration signals was realized.From the application example,using the strategy of gradual changing threshold selection and quantization,the noise is deleted and the useful information is reserved well in the whole frequency band.Compared with other methods,it owns better de-noising effect and universality.

Key words: lifting wavelet package transform, vibration signal, threshold, de-noising, feature extraction, fault diagnosis

摘要：

为了实现移动装备发动机缸盖振动信号的在线处理，提出一种基于提升小波包变换的渐变式阈值降噪方法。分析了移动装备发动机缸盖振动信号的时频特性，介绍了插补细分提升小波包变换算法。提出一种渐变式阈值选择与量化策略，以适应信号与噪声特点。引入基于类可分离性测量的降噪效果评价标准，实现了复杂实测信号降噪效果的量化评价。应用实例表明，渐变式阈值与量化策略，能在全频带内很好地消除噪声并保留有用信息，具有更好的降噪效果。

关键词: 提升小波包变换, 振动信号, 阈值, 降噪, 特征提取, 故障诊断