摘要
提纯旋转机械设备故障振动信号中的冲击特征,可以有效地实现相关故障的诊断。利用S变换适合于处理冲击特征信号的特点,提出基于S变换谱阈值去噪的冲击特征提取方法。先将信号进行S变换,得到其时频谱。考虑到此S变换谱为一复数矩阵,故而根据谱系数的模值大小进行阈值去噪。去噪过程中分别采用了基本的硬阈值函数和软阈值函数。对于最优阈值的估计,以所提出的改进风险函数为评价标准,利用步长迭代算法在零到系数最大模值的区间内获取。最后将去噪后的时频谱进行S逆变换,重构得到时域冲击特征。仿真信号与滚动轴承故障振动信号的处理结果表明,利用所获取的最优阈值,S变换谱阈值去噪方法能够从噪声混合信号中提取出冲击特征,从而实现相关故障的诊断。
Extracting impact features from a rotating machinery fault vibration signal is significant to some fault diagnoses.Based on the fact that S-tansformation is suitable for processing signals with impact features,an effective impact feature extracting method with the threshold denoising for S-transformation spectrum was proposed.Firstly,the time-frequency spectrum of a time-domain signal was obtained using S-transformation.As the spectrum coefficients were complex numbers,the spectrum threshold denoising was performed according to the magnitude of coefficients.During denoising,both the hard and the soft threshold functions were used,respectively.Most importantly,the optimal threshold estimation was obtained between zero and the maximum modulus of the coefficients with the step iterative algorithm,and evaluated with the introduced modified risk function.Finally,the inverse S-transformation was carried out on the denoised time-frequency spectrum to reconstruct the time-domain impact features.The processing results of simulated signals and rolling bearing fault vibration signals showed that using the acquired optimal threshold,S-transformation spectrum threshold denoising method can be used to successfully extract impact features from noise mixed signals,so as to realize the related fault dignosis.
出处
《振动与冲击》
EI
CSCD
北大核心
2014年第21期44-50,共7页
Journal of Vibration and Shock
基金
国家自然科学基金项目(51275453)
关键词
故障诊断
冲击特征
S变换
步长迭代算法
改进风险函数
最优阈值估计
阈值去噪
fault dignosis
impact feature
S-transformation
step iterative algorithm
modified risk function
optimal threshold estimation
threshold denoising
作者简介
郭远晶 男，博士生，1987年生
通信作者 魏燕定 男，教授，博士生导师，1970年生
