石油与天然气地质 ›› 1999, Vol. 20 ›› Issue (1): 81-84,93.doi: 10.11743/ogg19990119

• 论文 • 上一篇    下一篇

前馈神经网络的岩性识别方法

卢新卫, 金章东   

  1. 南京大学地球科学系, 南京 210093
  • 收稿日期:1998-10-22 出版日期:1999-03-25 发布日期:2012-01-18

METHOD OF LITHOLOGIC RECOGNIZING BASED ON FEEDFORWARD NEURAL NETWORK

Lu Xinwei, Jin Zhangdong   

  1. Department of Earth Sciences, Nanjing University, Nanjing
  • Received:1998-10-22 Online:1999-03-25 Published:2012-01-18

PDF (PC)

991

被引次数

63

摘要：

测井资料的地质解释是测井过程中十分重要的环节。岩芯资料少,测井资料较多及测井参数分布的模糊性,是岩性识别中的困难所在。在引入前馈神经网络方法的基础上,以取芯井岩芯与测井参数的对应关系作为识别模式,经过向识别模式学习获得模式识别智能知识,从而利用这些智能知识去识别未取芯井的测井岩性。通过对胜利油田永一地区沙河街组四段测井岩性的计算机判识,正确判别率达100%.应用结果表明,神经网络方法性能良好,具有极好的应用前景。

关键词: 神经网络, BP训练算法, 模式识别, 测井岩性

Abstract:

Geological explanation of logging data is an important link of logging work.When there are scanty core data and more log data,feedforward neural network is a very effective method to recognize the logging lithology.This paper introduces an identifying method of logging lithology by the pattern recognition approach using feedbach nueral network.The approach firstly regards the relationship between drilling cores and logging parameters as an identifying pattern so as to obtain intelligent knowledge from the identifying pattern,and finally applies the knowledge to identify the logging lithology of well.The method was used in lithologic recognition of the 4th member of Shahejie Formation in Shengli Oilfields,the result is satisfactory and shows good application prospects.

Key words: neural network, BP algorithm, pattern recognition, logging lithology