摘要
采用典型的逆向传播学习神经网络,并根据学习样本较少的特点引入函数扩展模型,对输入各参量进行函数扩充,识别六种碳酸盐岩相,如潮坪、台坡等。神经网络经过样本学习后,能正确识别学习的样本,对测试的32组剖面样本也取得了85%的识别率。由于人工神经网络无需建立数学模型,学习过程通过自动调节神经元之间的连接权值完成,在选取有代表性的训练样本情况下,人工神经网络可以作为一种常用的模式判别方法。
Artificial neural networks (ANNS)have been studied in recent years and applied to problems on pattern recognition in many fields. ANNs are a complex system consisting of a large number of simple processing units like human being's nerves by interacting with each other to be able to carry out highly nonlinear mapping. The backpropagation learning algorithm and the function-link model which needs only few training samples are introduced in this ANNs to recognize six kinds of carbonate lithofacies,such as tidal flat and platform ramp,etc. After learning,ANNs can correctly recognize and classify the trained samples and get up to 85% justification rate of the 32 tested samples. Since ANNs needn't establish mathematical model,instead,automatically adjust their connected weight values,ANNs can be popularly used as a common method for pattem recognition.
出处
《石油学报》
EI
CAS
CSCD
北大核心
1996年第4期50-54,共5页
Acta Petrolei Sinica
基金
国家"863"高科技研究项目
关键词
岩相
海相
碳酸盐相
模式识别
神经网络
artificial neural networks(ANNs)
lithofacies
marine facies
carbonate lithofacies
pattem recognition