摘要
在岩石声发射源时差定位的研究中,信号到达时间是重要信息。岩石声发射信号复杂,含有大量脉冲干扰与随机噪声,到达时间可读性差。针对以上问题,首先对原始声发射信号进行中值滤波和奇异值分解,消除部分脉冲干扰与随机噪声;其次进行小波包分解软阈值消噪,保留信号的主要成分,提高信噪比,增强到达时间的可读性;最后结合信号与噪声的时间序列模型(AutoRegressive,AR模型),第1次计算赤池信息量准则的K值(Akaike Information Criterion,AIC(K)值),获得到达时间窗口,在该窗口内第2次计算AIC(K)值,实现了到达时间的自动识别,避免了在整个信号序列下计算AR模型的阶数与次数。
Signal arrival time is vital information in the research of rock acoustic emission source TDOA location. Rock acoustic emission signal is complex and contains a lot of pulse interferences and random noises. Its arrival time is of poor readability. The method of traditional manual identification is time-consuming,for instance,the threshold method presents low accuracy,and AR-AIC method's precision would decrease if SNR is low. To solve the problems mentioned above,firstly,part of the pulse interferences and random noises in original acoustic emission signals were eliminated by median filter and singular value decomposition. Secondly,the main components of the signals were retained to improve SNR and enhance the readability of arrival time by wavelet packet decomposition soft threshold de-noising. Finally,arrival time was identified in the gained time window through combining with the AR model of signals and noises by calculating AIC( K) twice.
出处
《煤炭学报》
EI
CAS
CSCD
北大核心
2015年第S1期100-106,共7页
Journal of China Coal Society
基金
中央高校基本科研业务费资助项目(CDJZR12248801)
长江学者和创新团队发展计划资助项目(IRT13043)
关键词
声发射源
小波包分解
信号消噪
AR模型
到达时间
acoustic emission source
wavelet packet decomposition
signal de-noising processing
AR model
arrival time