基于改进CEEMD的超声检测信号自适应降噪  

Self-adaptive noise denoising for ultrasonic detection signal based on improved CEEMD

作　　者：孙灵芳 徐曼菲[2] 朴亨[2] 李霞[2] 

机构地区：[1]吉林省节能与测控技术工程实验室,吉林吉林132012 [2]东北电力大学自动化工程学院,吉林吉林132012

出　　处：《振动与冲击》2017年第20期225-232,共8页Journal of Vibration and Shock

基　　金：国家自然科学基金(51176028);吉林省科技发展计划项目(20140204030SF)

摘　　要：针对超声时域检测污垢数据的非平稳性和模态混叠难以实现污垢特征分离的问题,对于采用功率谱密度判定噪声区间的CEEMD分解,进而直接舍弃高频分量容易造成有效信息损失的不足,以及传统小波降噪易造成重构信号的变形等缺陷,提出基于CEEMD自相关函数的自适应软阈值降噪,引入模态自相关特性曲线判定含有噪声成分较多的IMF分量,并结合小波自适应软阈值方法拾取噪声分量中的高频有用信号。仿真分析和实验研究表明:基于CEEMD和自相关的自适应降噪方法优于传统小波阈值和单纯的CEEMD,且能很好的解决模态混叠问题,提取出污垢特征信号,对超声检测信号的处理具有重要意义。The paper aims to detect the non-stationary of fouling data and modal aliasing which may make it difficult to realize dirt characteristic separation of ultrasonic in the time domain. As for deficiencies such as insufficient effective information loss caused by directly application of power spectral density in determination of CEEMD decomposition of noise interval as well as deformation of reconstructed signal caused by traditional wavelet denoising, self-adaptive soft threshold noise reduction of autocorrelation function based on CEEMD as well as modal correlated characteristic curve were introduced to determine the IMF component with higher noise contribution. Besides, the method of wavelet self-adaptive soft-threshold value was also applied to collect useful high-frequency signal in noise component. According to the results of simulated analysis and experimental research, the self-adaptive noise reduction method based on CEEMD and autocorrelation is more effective than traditional wavelet threshold and pure CEEMD. It can better solve the problem of modal aliasing and extract dirt characteristic signal, which is of great importance to the processing of ultrasonic detection signal.

关 键 词：超声检测 完备总体经验模式分解 自相关函数 自适应 降噪 

分 类 号：TB53[理学—物理]