神经网络自动道编辑的改进与应用

详细信息 本馆镜像全文    |  推荐本文 | | 获取馆网全文

英文篇名：Improvement and Application of Automated Trace Editing with Neural Network 中文刊名：吐哈油气 英文刊名：Tuha Oil & Gas 作者：周振晓 ; 倪凤田 ; 刘建红 ; 徐少波 英文作者：Zhou Zhenxiao ; Ni Fengtian ; Liu Jianhong ; Xu Shaobo (Research Center for Geophysical Exploration Technology ; Bureau of Geophysical Prospecting ; PetroChina ; Zhuozhou 072750 ; Hebei) 中文关键词：神经网络 ; 互相关 ; 废道识别 英文关键词：neural network ; cross correlation ; abandoned channel identification 出版日期：2012-03-15 机构：中国石油东方地球物理公司物探技术研究中心; 年：2012 期：01 出版单位：吐哈油气

摘要

针对地震单炮的地球物理特征,将单炮划分为三个区域,利用各个区域的信号特征作为神经网络的输入神经元,通过网络的组织学习能力,来预测废道的类型,之后通过相关性等道间属性来对预测的废道进行二次判别。该方法提高了废道识别的准确率,有效的降低了网络的学习时间,具有更明显的地球物理特征。实际资料应用证明该方法具有良好的适应性。
According to the geophysical characteristics,seismic shot record is divided into three regions.Signal characteristics of each region were used as input neurons of neural network.The type of abandoned channel is predicted by self-organized learning capability of neural network,and then the secondary discrimination for the abandoned channel can be conducted by trace attributions such as correlation.The method improved the accuracy of abandoned channel identification,reduced the learning time of the network effectively and had more obvious geophysical characteristics.The practical application shows the method has a good adaptability.

引文

[1]Anderson R G,MCMechan GA.Automatic editingofnoise seismic data.Geophysical Prospecting,1989,37:875-892.
    [2]张学工,李衍达.用人工神经网络实现地震记录中的废道自动切除.地球物理学报,1992,35(5):637-643.
    [3]McCormach MD,Zaucha DE,Dushek DW.First-break re-fraction event pickingand seismic data tracingeditingusing neural networks.Geophysics,1983,58(1):67-78.
    [4]周竹生,周熙襄.一种用于地震道自动编辑的人工神经网络方法,石油地球物理勘探,1994,29(2):147-155.