岩相识别的神经网络计算  

Facies Recognition Using the Neural Networks

作　　者：王硕儒[1] 范德江[1] 汪丙柱 

机构地区：[1]青岛海洋大学

出　　处：《沉积学报》1996年第4期154-160,共7页Acta Sedimentologica Sinica

摘　　要：模式识别在定量研究岩相模式中是一种有效的方法，神经网络是模式识别的一种新方法，和其它模式识别方法不同之处在于它能模拟人脑并行处理信息的模式和神经系统的可塑性。经传统岩相分析，鄂中拗陷二叠系碳酸盐岩确认为碳酸盐岩台坪相，它包含五个亚相，即局限台洼、台坡Ｂ、台洼、台坡Ａ和台滩相。它们在岩石的颜色、矿物组分、古生物种属、结构等特征上都有不同的差异。应用ＢＰ神经网络，特别是与模糊模式识别的结合，对拗陷区的岩相识别是成功的，令人鼓舞，两种方法各自的正确判对率约为７５％，而综合两种方法的判对结果。Pattern recognition methods are a powerful means in studying facies quantitatively. The neural network is a new method and has many improvements such as parallel processing and plasticity imitating the human brain compared with other pattern recognition methods. Permian carbonate rocks in the central Hubei basin have been identified as carbonate platform facies after traditional facies analysis. It includes five subfacies, i.e. sub-facies of districted depression, slope B, depression, slope A and shallow out. They are different from each other in rock color, mineral components, paleobiology components, rock structures and so on. The application of pattern recognition on the basis of the BP neural network for Permian carbonate rock facies studies, particularly comparison with fuzzy pattern recognition is very successful and inspiring. The correct identification ratios of fuzzy sets and the neural network are both about 75%. And the corret identification ratio of the combination of fuzzy sets and neural networks is 100%.

关 键 词：岩相 模式识别 神经网络 BP算法 碳酸盐岩 

分 类 号：P588.245[天文地球—岩石学]