摘要
提出了最优尺度小波包络流形算法,以反映信号冲击特性的指标为原则选择尺度带。由于多尺度下的包络信号中的噪声是随机的,而故障特征是确定的,使用非线性的流形方法结合多个尺度下的含故障信息的包络信号,可以抑制带内噪声。首先,对故障信号进行连续小波变换,得到全尺度下的小波包络信号;然后,通过最优尺度指标选择蕴含故障信息的最优尺度带;最后,对最优尺度带内的小波包络进行流形学习,提取本征结构,消除带内噪声,并进行FFT谱分析,实现故障识别。利用轴承试验信号分析对比了5种最优尺度指标的效果并验证了基于最优小波尺度的包络流形算法的有效性。
A method called envelope manifold of optimal-wavelet-scale (OEM) is introduced in this paper to solve these problems. Selecting scale band with indicators which can describe the impulsiveness of signal. The noise of envelope signal at each scale is occurs randomly, however the information of fault is certain, so non-linear method- manifold can combine fault information of multi-scale and lead to noise suppression. First. the continuous wavelet transform is employed to obtain the wavelet envelopes at all scales; Second, optimal scale indicators are employed to determine the optimal scale band containing the information of fault impacts. Third, the manifold learning algorithm is applied on the wavelet envelope of the selected optimal scale band to extract fractual structure and reduce the in-band noise. Finally, the power spectrum of the envelope manifold is calculated to identify the faults. A total of five optimal scale indicators are employed to select the optimal scale band and their effectiveness is verified and compared through the experimental ease study.
出处
《机电一体化》
2016年第7期61-67,共7页
Mechatronics
基金
安徽大学大学生科研训练计划项目
安徽省教育厅重点项目(KJ2013A010)
关键词
故障诊断
小波变换
流形学习
最优小波尺度
局部切空间排列
复morlet小波
fault diagnosis wavelet transform manifold learning optimal wavelet scale local tangent space alignment complex morlet wavelet