我们已经研制了利用向后传播神经网络由测井数据自动确定岩性的计算机程序。与常规的串行计算机不同,神经网络是一种由节点(有时称为神经元、神经极或单元)和节点间的连接组成的计算系统。神经计算试图仿效哺乳动物的脑功能,从而模拟思维过程。神经网络法与以前的模式识别法的不同之点在于它向范例“学习”的能力。与常规的统计模式识别法不同,这种新方法不需要复杂的数学运算和大量的统计数据。本文讨论了神经网络在地质学中模式识别问题方面的应用:依据测井数据确定岩性。神经网络由选定的测井数据确定岩性(石灰岩、白云岩、砂岩、页岩、砂质及白云质灰岩、砂质白云岩和泥质砂岩)所需的时间仅为一个有经验的测井分析家所需的一小部分。 We have developed a computer program that uses back-propagation neural networks to automatically determine lithology from log data. Unlike conventional serial computers, neural networks are computational systems that consist of nodes (sometimes referred to as neurons, neurons, or cells) and connections between nodes. Nerve computation attempts to emulate the brain function of mammals, thus simulating the process of thinking. The difference between the neural network method and the previous pattern recognition method lies in its ability to “learn” from the examples. Unlike conventional statistical pattern recognition, this new approach does not require complex mathematical operations and a large amount of statistical data. This paper discusses the application of neural networks in pattern recognition in geology: lithology determination based on log data. Neural Networks The time required to determine lithology (limestone, dolomite, sandstone, shale, sandy and dolomitic limestone, sandy dolomite, and argillaceous sandstone) from selected log data is only an empirical A small part of what logging analysts need.