地震地磁观测与研究2025，Vol.46Issue(5)：28-36,9.DOI:10.3969/j.issn.1003-3246.2025.05.004
内蒙古及邻区地震事件的卷积神经网络分类研究
Convolutional neural network-based classification of earthquake events in Nei Mongol and its adjacent areas
摘要
Abstract
Taking 4 596 natural earthquake event data and 1 175 anthropogenic earthquake event data recorded by the Nei Mongol Seismic Network from 2010 to 2025 as research samples,this study conducts event classification research based on a lightweight Convolutional Neural Network(CNN)architecture.Through short-time fourier transform(STFT),seismic signals are converted into time-frequency images,and the Hamming window is adopted for window function optimization,where the window length is set to 2 seconds and the overlap ratio is 0.5.On this basis,a CNN classification model containing 3 groups of convolutional layers is constructed.The experimental results show that the model exhibits excellent classification performance,with an accuracy of 96.557%and a precision of 96.328%.Meanwhile,the model consumes less time in training and possesses the characteristic of efficient computation,making it more suitable for real-time seismic monitoring scenarios.It can realize a accurate and rapid distinction between natural and anthropogenic earthquake events in Nei Mongol and its adjacent areas,providing technical support for regional seismic monitoring and early warning work.关键词
卷积神经网络(CNN)/地震分类识别/非天然地震/人工智能Key words
Convolutional Neural Network(CNN)/seismic classification and identification/anthropogenic earthquakes/artificial intelligence引用本文复制引用
堵伟鹏,杨红樱,丁修毅,谭园梦,吴桐,王慧,邹鑫慈,乌兰,周金玲,张帆..内蒙古及邻区地震事件的卷积神经网络分类研究[J].地震地磁观测与研究,2025,46(5):28-36,9.基金项目
2023年度内蒙古自治区地震局局长基金(项目编号:2023QN18) （项目编号:2023QN18）
