基于PhaseNet的地震信号自动处理方法准确性分析

利用一种深度学习地震信号检测方法（PhaseNet模型）检测甘肃省岷县MS6.6地震序列，根据地震震相关联技术和绝对定位Hypoinverse方法构建AI检测目录，对比人工分析地震目录和震相报告，分析AI自动处理地震事件的误差范围。结果表明，AI技术能够实现85.5%的人工工作量，定位误差在20 km范围内呈正态分布；在低震级地震事件检测中，AI方法也发挥了显著能力。相比于人工处理方法，AI方法稳定性好、不依赖个人经验、分析速度快，在大震后的地震目录快速产出中能发挥关键作用，可提高地震分析能力、节省人力成本。

Accuracy Analysis of Automatic Seismic Signal Processing Method Based on PhaseNet

We use a phasenet model for deep learning detection of seismic signals to detect the MS6.6 earthquake sequence in Minxian, Gansu province. According to the seismic correlation technology and absolute positioning hypoinverse method, we construct the AI detection directory. Compared with the manual analysis of seismic directory and seismic phase report, we analyze the error range of AI automatic processing of seismic events. The results show that AI technology realizes 85.5% manual work, and the positioning error is normally distributed within 20 km. AI method has played a significant role in the detection of low magnitude earthquake events. Compared with manual processing, the AI method has good stability, does not depend on personal experience level, and has fast analysis speed. It can play a key role in the rapid output of earthquake catalog after a large earthquake, quickly improve the ability of earthquake analysis and save human labor.