近震P波震相自动识别方法研究

发布时间：2018-05-29 16:34

本文选题：近震 + P波震相　；参考：《中国地震局地球物理研究所》2015年硕士论文

【摘要】：地震震相的检测和识别是地震学研究中的重要课题,它是地震定位、地震预警及地球深部结构等研究的基础。地震震相的自动识别可大大提高地震速报和地震预警的速度,为震后应急救援赢得宝贵时间。总结了目前较常用的震相自动识别方法,对比分析了三大类方法即单特征法(包括能量分析、偏振分析、高阶统计量、分形分维、赤池信息量和频谱分析等)、多特征法(包括全波震相分析、相关法及人工神经网络等)和综合分析方法的优缺点,并指出了震相自动识别的发展方向。在前人研究的基础上,提出了一种震相自动识别新方法,即小波包-峰度AIC(Akaike Information Criteria,赤池信息准则)方法。该方法由三部分组成：(1)利用加权STA/LTA (short term average/long term average,长短时窗平均比)方法自动检测出有效的地震事件并拾取P波初至的粗略到时；(2)对粗略拾取的到时前后各推3秒的时间窗内的信号进行小波包三尺度分解重构；(3)分别计算三个尺度重构信号的峰度AIC曲线并进行叠加,将叠加的AIC曲线的最小值作为最终拾取到的精细P波初至到时。为检验新方法效果,将其应用于模拟事件的理论地震记录,模拟事件参照云南地区实际震例设计。对理论地震记录加入不同信噪比的高斯白噪声和实际地震噪声,以由射线追踪技术得到的到时为标准,对比了加权STA/LTA法、峰度AIC法和本文方法识别P波的效果。结果表明本文方法具有更强的抗噪能力,P波识别的精度更高。以云南地区722个近震垂直向记录为例,考察了滤波方法、信噪比及初至清晰度对震相识别精度的影响。结果表明：FIR最佳频带滤波方法在提高信噪比及P波识别的精度上优势更突出；相对信噪比的影响,P波识别的精度受初至清晰度的影响更大。以人工拾取的震相到时为标准,与加权STA/LTA、峰度AIC两种方法相比,本文方法效果更好。对比了人工与自动拾取的P波走时曲线,进一步验证了本文方法的可靠性。
[Abstract]:Seismic phase detection and identification is an important subject in seismology. It is the basis of earthquake location, earthquake warning and deep structure of the earth. The automatic identification of seismic phases can greatly improve the speed of earthquake rapid reporting and earthquake warning, and win valuable time for post-earthquake emergency rescue. In this paper, the methods of automatic seismic phase identification are summarized, and three kinds of methods, I. e., single feature method (including energy analysis, polarization analysis, high-order statistics, fractal dimension), are compared and analyzed. The advantages and disadvantages of red pool information and spectrum analysis, multi-feature method (including full-wave phase analysis, correlation method and artificial neural network) and synthetic analysis method are discussed. The development direction of automatic seismic phase recognition is pointed out. On the basis of previous studies, a new method of automatic seismic phase recognition is proposed, that is, wavelet packet kurtosis AIC(Akaike Information criteria (red cell information criterion). This method consists of three parts: 1) using weighted STA/LTA / short term average/long term average ratio) method to automatically detect effective seismic events and pick up the rough arrival of P wave. The signal in the time window is reconstructed by wavelet packet three-scale decomposition. (3) the kurtosis AIC curves of the three scale reconstructed signals are calculated and superposed, respectively. The minimum value of the superimposed AIC curve is regarded as the first arrival time of the finer P wave which is finally picked up. In order to test the effect of the new method, it is applied to the theoretical seismic records of the simulated events, and the simulated events are designed according to the actual earthquake examples in Yunnan region. The white Gao Si noise with different signal-to-noise ratio and the actual seismic noise are added to the theoretical seismic records. Using the arrival time obtained from the ray tracing technique as the standard, the effects of weighted STA/LTA method, kurtosis AIC method and the present method on P wave identification are compared. The results show that the proposed method has stronger anti-noise ability and higher accuracy of P wave recognition. Taking 722 vertical seismic records in Yunnan as an example, the effects of filtering method, signal-to-noise ratio (SNR) and initial resolution on the accuracy of seismic phase identification are investigated. The results show that the ratio Fir optimal band filtering method has more advantages in improving the signal-to-noise ratio and the accuracy of P wave recognition, and the relative signal-to-noise ratio affects the accuracy of P wave recognition more greatly due to the initial clarity. According to the criterion of phase arrival, the proposed method is more effective than the weighted STA-LTA and kurtosis AIC methods. The P wave travel time curves of manual and automatic pick-up are compared, and the reliability of this method is further verified.
【学位授予单位】：中国地震局地球物理研究所
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P315.6

【参考文献】