基于U-Net的矿山微震初至拾取研究

初至到时拾取是微震数据处理中基础而又重要的环节，关系到整个微震监测系统的精度与可靠性。因此，为了解决传统初至拾取方法存在拾取效率低和拾取精度差的问题，引入深度学习方法，构建U型神经网络(U-Net)来预测三分量矿山微震数据P波、S波和噪声的概率分布，并根据概率峰值提取其初至到达时间。实验结果表明：本文算法的拾取结果准确度高且误差范围较小，与AR pick (auto regression pick)算法相比，其拾取结果具有明显的优越性。所构建的初至拾取模型可用于矿山动力灾害微震监测，解决微震监测的瓶颈问题，为矿山安全生产风险智能监测预警提供强力技术支撑。

Arrival picking of mine microseismic events using U_Net

Arrival picking is a basic and important link in microseismic data processing, which is related to the accuracy and reliability of the whole microseismic monitoring system. Therefore, in order to solve the problems of low arrival picking efficiency and poor arrival picking precision in the traditional methods, the paper introduces the depth learning method, the U network (U-Net) is constructed to predict the probability distribution of P wave, S wave and noise in three-component mine microseismic data, and the arrival time is extracted according to the peak of probability. Experimental results show that the proposed algorithm has high accuracy and small error range, and it has obvious advantages over AR pick (auto regression pick) algorithm. The arrival picking model can be used in microseismic monitoring of the mine dynamic disaster, solve the bottleneck problem of microseismic monitoring, and provide strong technical support for mine safety production risk intelligent monitoring and early warning.