基于响应曲线聚类的时间域电磁法地电结构类型划分研究  

Classification of geoelectric structures using time-domainelectromagneticmethod based on response curve clustering

作　　者：施明智 曹辉[1] 陈柏午 SHI Mingzhi;CAO Hui;CHEN Baiwu(College of Geophysics,Chengdu University of Technology,Chengdu 610059,China;School of Artificial Intelligence,Nanning College for Vocational Technology,Nanning 530008,China)

机构地区：[1]成都理工大学地球物理学院,成都610059 [2]南宁职业技术学院人工智能学院,南宁530008

出　　处：《成都理工大学学报（自然科学版）》2024年第3期522-530,共9页Journal of Chengdu University of Technology: Science & Technology Edition

基　　金：国家自然科学基金(41974090)。

摘　　要：为了充分利用时间域电磁法正演响应曲线携带的特征信息实现地电结构类型快速预测,通过将曲线的分布特性和形态特性相结合构造了曲线距离,采用相关系数作为权重参数衡量曲线形态相似程度,实现了时间域电磁法正演响应曲线聚类。应用效果表明,依据该方法得到的聚类结果能实现正演响应样本的标签化,从而通过多分类任务神经网络完成地电结构的类型划分。In this study,we aimed to make full use of the characteristic information carried by the forward response curve in the time-domain electromagnetic method to quickly and accurately predict the types of geoelectric structures.We developed the curve distance by combining the characteristics of distribution and morphology of the curve,and used the correlation coefficient as the weighting parameter to measure its degree of similarity.This enabled us to cluster the forward response curve of the time-domain electromagnetic method.The results of its application showed that the outcomes of clustering obtained by this method could be used to label the forward response samples,and to subsequently classify geoelectric structures by using a multi-classification neural network.

关 键 词：时间域电磁法 曲线聚类 地电结构 

分 类 号：P319.2[天文地球—固体地球物理学]