单分量微地震信号的P波S波到时自动拾取方法

 单分量传感器依然是目前矿山中应用最广泛的传感器。针对矿山每天采集到的大量信号,提出了一种针对单分量微地震信号的P 波S 波到时拾取算法。长短时窗均值比法(STA/LTA)可以用来检测信号的变化趋势,结合改进的特征函数对信号的几何特征予以加强,进而可以提高震相识别的精度。随着长短时窗的滑动,信号的变化趋势表现在走势图上,再利用震相到时判别阈值可以初步定位到P 波S 波的到时区间。初步确定到时区间后,分别在P 波S 波的到时区间内应用AR-AIC 准则,可以计算出信号发生突变的位置,这两个点即对应着P 波S 波的到时。结合时差阈值判别和时频图谱,对拾取的结果再次做检验。经过两次判别,有效地提高了该算法的拾取精度。通过对1000 多条单分量微地震信号进行验证,自动拾取能拾取到71%的P 波到时和79%的S 波到时,相同条件下,自动拾取的结果不低于手动拾取的精度,满足矿山应用的要求。

An automatic P-and S-wave onset-picking designed for single-component microseismic signals

The geophone can daily detect a large amount of microseismic signals and it is time consuming for manual picking the arrival times of P-S waves. We propose an automatic P-and S-wave onset picking method based on single-component geophone, which is widely used in mines. Firstly, we filter the original signal with a Butterworth bandpass filter in order to increase the SNR. The mine microseismic signal frequency ranges from 10 Hz to 2000 Hz, so we choose the band-pass filter interval to be 3-2400 Hz. Secondly, we need to locate the P-S wave phases in the signal. An improved characteristic function can strengthen the change of frequency and amplitude. We define two neighboring windows of different sizes, with the short time window lying in the end of the long time window. Then we compute the average value of the characteristic function within the window, which is also called STA/LTA (short-term average value/long-term average value). With the sliding of the long-short time window, we can get the positions of the P-S wave onset windows. The ratio of STA/LTA will have two peaks because the arrivals of P wave and S wave. Through a preset threshold, we can locate two time windows including P wave onset time and S wave onset time, respectively. In the last step, Akaike's information criterion(AR-AIC criterion) is used in the two time windows, the global minimum of AIC in the window is the arriving time. We have tested this method with one thousand single-component microseismic signals, the result has shown that 71% of P-wave arrival time picking is correct and 79% of S-wave picking is correct.