基于自适应神经网络的岩性识别动态建模

Dynamic Modeling of Lithologic Identification Based on Self-adapted Neural Network

针对研究对象定量研究复杂度高的地质工作,应用计算机进行定量化和信息化的研究,需要建立一定的数学模型,然而,传统的数学方法难以得到精确的数学模型,神经网络作为一非线性建摸方法,具有良好的自组织和自适应性等功能,可以逼近任意的非线性函数(映射)。本文提出利用神经网络的自组织、自学习、自适应功能实现数学模型的实时建立的方法,并在反传神经算法前馈神经网络(BP)模型引入了自适应动量因子α,使得网络计算量小,收敛速度快。最后将该模型应用到某地岩性识别动态建模中,取得了较好的效果。

In geological research work,the construction of the mathematic models of the research objects with complex characters is a more and more difficult job.Especially,the traditional mathematic methods sometimes can not model our research object;To deal with this situation,a selfstudied and selfadapted neural network based on BP algorithm with a momentum factor α is put forward to realize the models building at run time.At last a test of this method has been used in the lithology identification; a good effect has been achieved with a rapid constringency.