基于轴承仿真信号的小波阈值及阈值函数的优化
2020-10-16
作者：杨铮，王伟兵，李仁华，霍迎科，贾永坡
单位：河北钢铁集团邯郸分公司
摘要：针对传统小波阈值法在轴承信号去噪效果不佳的难题，对阈值和阈值函数采用层层递进的方法进行降噪研究。阈值函数方面，在软阈值函数基础上提出了指数阈值函数，在指数阈值函数的基础上结合硬阈值函数，提出了带有收缩性指数的阈值函数。阈值方面，给出了估噪阈值，在估噪阈值的基础上，结合噪声的小波系数，随着分解层的增大而减小并引入调节量，给出了分层估噪阈值，表征去噪效果优劣指标为SNR&RMSE。通过选取估噪阈值、分层估噪阈值、软阈值函数、指数阈值函数和收缩指数阈值函数进行去噪试验，结果表明，收缩指数阈值函数和分层估噪阈值去噪效果较好。从阈值函数选取上看，收缩指数阈值函数去噪效果最好（32.3118 & 0.0141），其次是指数阈值函数（30.4301 & 0.0175），最后是软阈值函数（23.5500 & 0.0385）；从阈值选取上看，分层估噪阈值去噪效果最好（33.9360 & 0.0117），其次是估噪阈值（32.3118 & 0.0141）。该方法去噪效果好，可从噪声中提取有用信息，可用于对大型轴承运转的有效诊断。
关键词：去噪；阈值函数；阈值；轴承信号
中图分类号：TG702;TH133.3；TN911.7文献标志码：ADOI：10.3969/j.issn.1000-7008.2020.03.020
Improvement on Threshold and Threshold Function of Waveleton Based on
Bearing Simulation Signal
Yang Zheng，Wang Weibing，Li Renhua，Huo Yingke，Jia Yongpo
Abstract:According to the poor denoising effect of the traditional wavelet method in bearing signal，improved wavelet threshold and improved wavelet threshold function are proposed.In the threshold function，an exponential threshold function is proposed on the basis of the soft threshold function.On the basis of threshold，the estimated value of horizontal noise is introduced on the basis of fixed threshold，and the threshold of noise estimation is given.On the basis of estimating the threshold of noise，the amount of the noise is combined with the increase of the decomposition layer to reduce this property，and the thresholding is given.By selecting the fixed threshold，the threshold value of the noise estimation and the thresholding，the soft threshold function，the exponential threshold function and the contraction index threshold function are used to carry out the denoising experiment，the SNR & RMSE is used to characterize the effect of bearing signal denoising.The experimental results show that the threshold function of the contraction exponent，the threshold denoising threshold，is the best (33.9360 & 0.0117).From the threshold function selection，the shrinkage index threshold function has the best denoising effect (32.3118 & 0.0141)，followed by the exponential threshold function (30.4301 & 0.0175)，and the last is the soft threshold function (23.5500 & 0.0385).From the threshold selection，hierarchical noise threshold has the best denoising effect (33.9360 & 0.0117)，followed by noise threshold (32.3118 & 0.0141).This method has better removal effect，can extract useful information from noise，lay a foundation for effective diagnosis of large bearing operation.
Keywords:denoising；threshold function；threshold；bearing signal