摘要
为了更加有效地去除噪声对被测信号的干扰，分析了传统小波阈值估计方法的局限性，提出了一种基于样本熵的最优小波包阈值估计去噪算法。该方法利用样本熵作为信息价值函数以确定最优小波包，且以样本熵为判据，对不同的分解层数设置不同的阈值，选取使得去噪后得到的噪声估计信号样本熵值最大的阈值作为最优阈值。对仿真信号进行分析证明了该方法的有效性，将该方法应用于滚动轴承振动信号去噪分析且与其他阈值方法相对比，结果表明该方法去噪后的信号较其他方法而言频谱中的干扰频率更少且滚动轴承的基频以及故障频率更为突出，去噪效果更好，是一种更为优越的去噪算法。
In order to remove the interference of noise to measured signal more effectively, the limitation of traditional wavelet threshold estimation method was analyzed, and an optimal wavelet packet threshold estimation denoising algorithm based on sample entropy was proposed. The method take sample entropy as information value function and criterion to determine the optimal wavelet packet, and set different threshold to different decomposition layer, choose best threshold which makes the sample entropy of noise estimation signal largest. The analysis of simulation signal proved the effectiveness of this method. Apply the method to rolling bearing vibration signal denoising analysis and compared with other threshold methods, the results show that the method removed noise better and restored signal characteristic frequency, it＇ s a greater denoising method.
出处
《机械设计与研究》
CSCD
北大核心
2018年第1期39-42,共4页
Machine Design And Research
基金
国家重大科学仪器设备开发专项资助项目(2013YQ13042902)
关键词
样本熵
最优小波包
阈值估计
去噪
sample entropy
optimal wavelet packet
threshold estimation
denoising
