自组织特征映射神经网络在岩性识别中的应用

传统的岩性识别技术主要基于统计学理论，如贝叶斯方法、回归方法等，近年来人工神经网络方法如反向传播算法（ Ｂａｃｋ － Ｐｒｏｐａｇａｔｉｏｎ ， Ｂ－ Ｐ） 也应用于岩性识别，取得了一定的效果。用 Ｋｏｈｏｎｅｎ 提出的自组织特征映射神经网络对测井数据进行岩性识别，该方法具有较强的自组织性、自适应性，有较高的容错能力。与 Ｂ－ Ｐ 算法相比较，计算量小，收效速度快，且不需要已知的先验信息而自动确定分类类别。结果表明与统计方法、岩性录井分析结果一致。

APPLICATION OF SELF-ORGANIZING FEATURE MAP NEURAL NETWORK IN LITHOLOGICAL IDENTIFICATION

The classical lithological recognition technology is based on stastics theory,such as Bayesian method,regressive method et al. This paper introduces the self-organizing map neural network(SOM_,presented by T. Kohonen,to recognize the lithology based on the log data.The method has many advantages especially in self-adaptation,self-organization and high fault tolerance capacity.Comparing with B-P algorithm,the method has small amount of calculation and fast convergence rate,and can automatically decide the kind of pattern without prior information.It has the same result with statistical method and logging interpretation.