岩石破裂声发射压缩波到时拾取方法及其优化改进研究

 岩石破裂过程中的微震及声发射监测技术已广泛应用于岩石工程领域。P波到时自动拾取是进行岩石破裂源定位和矩张量反演等声发射(AE)技术研究的基础与关键，为提高到时拾取精度，分析Allen拾取法、Baer-Kradolfer改进拾取法、高阶统计量拾取法以及AR-AIC拾取法对模拟正弦信号振幅、频率与相位突变识别的敏感性。基于加拿大原子能公司地下实验室(URL)的隧道密封试验现场监测数据，对比分析几种拾取方法对不同信噪比(SNR)水平的信号拾取结果，研究表明。Allen拾取法、Baer-Kradolfer拾取法和高阶统计量拾取法具有更广的信噪比识别范围，特别对低信噪比水平的信号具有较强识别能力，进而提出对AR-AIC拾取法的改进思路。利用改进AR-AIC拾取法对真实声发射信号进行到时拾取，得到影响拾取精度的关键因素以及适用于工程尺度的局部范围的声发射信号的合理参量，进而成功对信噪比水平小于10的声发射信号进行自动到时拾取，研究认为高阶统计量法中的峰度拾取法是应用改进AR-AIC法初拾阶段的最优方法，其自动拾取与人工识别结果时差小于5 ?s的准确率为94%，表明提出的改进AR-AIC拾取法在实际应用中，特别是对低信噪比水平信号进行到时拾取具有良好的适用性。

Methods of P-onset picking of acoustic emission compression waves and optimized improvement

Microseismic and acoustic emission(AE) monitoring during rock fracture process have been widely applied in the area of rock engineering. Automated P-onset picking is a fundamental and key link of location and moment tensor inversion in AE technique. The identification sensitivities regarding the jumping of the amplitude，frequency and phase of the analog signal waveform with the Allen picker，the Baer-Kradolfer picker，the Higher Order Statistic picker and the AR-AIC picker were analyzed for improving the picking accuracy. A comparison of P-onset picking for different signal to noise(SNR) levels by different methods was carried out based on the field monitoring data from the Tunnel Sealing Experiment in Atomic Energy of Canada Limited?s Underground Research Laboratory(URL). The result showed that the Allen picker，the Baer-Kradolfer picker and the Higher Order Statistic picker had a wider scope of SNR identification and better identification ability especially for AE signals at low SNR level. Thus，an improvement for the AR-AIC picker was proposed. The key factors affecting the picking accuracy were discussed and the reasonable parameters applied to AE signals were gained when picking P-onset of real signal waveforms using the improved AR-AIC picker. The P-onset was successfully picked when using the improved AR-AIC picker on AE signals with SNR level below 10，and it indicated that the Kurtosis method was the best picker during the preliminary P-onset detection stage. Comparisons of the results of automatic and manual identification showed that the accuracy rate reached 94% with a time lag less than 5 ?s. The improved AR-AIC picker shows good feasibility in practical application especially for AE signals in low SNR level.