基于多输入卷积神经网络的天然地震和爆破事件识别

Discrimination of earthquake and quarry blast based on multi-input convolutional neural network

天然地震和爆破事件识别是地震监测预警的重要内容.近年来,快速发展的深度学习算法以其强大的数据特征挖掘和图像识别能力,能够较快并准确地约束地震事件属性.利用多输入卷积神经网络算法构建天然地震和爆破事件自动分类网络模型,其中输入信息包括多台站地震波形和单台站的地震时频数据,使得卷积神经网络同时获取事件的波形、频谱和极性特征.根据美国犹他州2012年记录到的天然地震和采石场爆破的观测资料,构建深度学习的训练数据集并进行模型训练,并据此判断2013—2016年间已知的天然地震和爆破事件.结果表明,多输入卷积神经网络具有较高的识别精度,识别率高达97%.

The discrimination of natural earthquakes and quarry blasts is an important part of earthquake monitoring and early warning.In recent years,the fast-developing deep learning algorithm with its powerful data feature extraction and image recognition capabilities can quickly and accurately constrain the classifications of seismic events.The multi-input convolutional neural network is used to construct the automatic classification network of natural earthquake and quarry blasts.The input layer includes the waveform data of multiple stations and the spectrogram data of a single station,so that the multi-input convolutional neural network can learn the waveform,spectrogram,and polarity characteristics of the event at the same time.Earthquakes and quarry blasts recorded in Utah,the United States in 2012 are used to construct the training data set,and the known natural earthquakes and quarry blasts from 2013 to 2016 in the area are utilized to test the trained network model.The results show that the multi-input convolutional neural network has high recognition accuracy,and the discrimination accuracy is as high as 97%.