基于卷积神经网络的天然地震和非天然地震识别

CLASSIFICATION AND RECOGNITION OF NATURAL EARTHQUAKES AND BLASTS BASED ON CONVOLUTIONAL NEURALNETWORKS

为实现天然地震和非天然地震分类识别,利用云南省及周边地区范围内测震台站所记录的302次天然地震和308次非天然地震事件(爆破、塌陷、强夯土等)为神经网络模型测试集、训练集数据,设计了VGG19卷积神经网络模型对天然地震和非天然地震进行分类识别。结果表明:VGG19对训练集与测试集数据的识别准确率达92%以上;天然地震的识别准确率为96%以上,非天然地震的识别准确率约为98%。通过实验说明,VGG19神经网络模型对天然地震和非天然地震识别具有实用意义。

In order to quickly and efficiently classify and identify natural and non natural earthquakes,this pa-per uses 302 natural earthquakes and 308 non natural earthquake events such as blasting,collapse,and dy-namic compaction recorded by seismic stations in Yunnan Province and surrounding areas as the test and train-ing data of the neural network model.A VGG19 convolutional neural network model is designed to identify and classify natural and non natural earthquakes.The results show that the recognition accuracy of VGG19 on the training and testing sets is over 92%,and the recognition accuracy of natural earthquakes is about 96%;The recognition accuracy of non natural earthquakes is about 98%.Through experiments,it has been demonstrated that the VGG19 neural network model has practical significance for identifying natural and non natural earth-quakes.