摘要
针对EEMD算法在以往选取有效固有模态函数(IMF)时存在误判现象,提出了一种将EEMD与云相似度理论相结合的去噪方法。通过构造连续声发射信号的仿真实验,并与互相关系数法选取固有模态函数作了比较,以信噪比和均方误差作为指标,验证了该方法能更好地提高信噪比;利用改进EEMD算法对红砂岩声发射信号进行频率特征提取,结果表明:声发射信号分解的固有模态函数前三个IMF分量的云相似值较大,与其它分量的差值范围为0.346~0.906,经过多组实验统计分析选取阈值为0.655;通过功率谱分析,将0~25 kHz对应于IMF4、IMF5、IMF6、IMF7、IMF8低频分量,25~150 kHz对应于IMF1、IMF2、IMF3高频分量,且红砂岩破裂过程的声发射信号的有效频段为25~150 kHz。
In view of the fact that the ensenble empirical mode decomposition(EEMD)algorithm selects the effective intrinsic mode function(IMF)in the past,the phenomenon of misjudgment was considered.A de-noising method combining EEMD with cloud similarity theory was proposed.First,through the simulation experiment in which we constructed continuous acoustic emission signals and used the correlation coefficient method to select the intrinsic mode function compared with the signal-to-noise ratio and mean square error as the index,we verified that this method can improve the signal-to-noise ratio.Second,when we used the improved EEMD algorithm for frequency characteristics of the red sand rock acoustic emission,the signal extraction results showed that the acoustic emission signal was a decomposed intrinsic mode function.The first three IMF components have high cloud similarity value,with difference range of 0.346-0.906 from other components.The experimental statistical analysis to select the threshold is 0.655.Finally,we analyzed the power spectrum,0-25 kHz,corresponding to IMF4,IMF5,IMF6,IMF7,IMF8 low frequency components;25-150 kHz,corresponding to IMF1,IMF2,IMF3 high frequency components,and the red sandstone rupture of the effective frequency of the acoustic emission signals for 25-150 kHz.
作者
戴聪聪
程铁栋
宗路
罗小燕
DAI Congcong;CHENG Tiedong;ZONG Lu;LUO Xiaoyan(School of Mechanical and Electrical Engineering,Jiangxi University of Science and Technology,Ganzhou 341000,China)
出处
《振动与冲击》
EI
CSCD
北大核心
2018年第16期118-123,共6页
Journal of Vibration and Shock
基金
国家自然科学基金(51464017)
江西省教育厅科技重点项目(GJJ150618)
关键词
声发射信号
EEMD
云相似度
模拟仿真
功率谱
acoustic emission signal
EEMD
cloud similarity
simulation
power spectrum
