利用神经网络的LM算法确定碳酸盐岩声波孔隙度

确定碳酸盐岩声波孔隙度是测井解释中的一个难题，传统方法是利用平均时差公式经过适当校正或使用声波、中子、密度等两种以上的测井资料求取，在具体使用中误差较大且很不方便。为此，基于Levenberg—Marquardt算法，提出一种确定碳酸盐岩声波孔隙度的神经网络方法，主要步骤包括：样本信息的预处理、网络结构的设计、采用LM算法的网络学习训练、碳酸盐岩声波孔隙度的确定。仿真实验和比较分析表明，该方法快速稳定，其结果与真实值吻合程度高。

Determining carbonate formation acoustic porosity by neural network based on LM algorithm

It is a difficult problem to determine the carbonate formation porosity in oil logging interpretation. The traditional methods are the time-average equations derived or acquired based on two of acoustic, neutron and density logging date , but these methods are inconvenient in concrete application. The method of artificial neural network(ANN) has the characteristics of high self-study, self-adaptation and interference resistance. A network structure with single-hidden-layer is adopted. The application of experimentation shows the Levenberg-Marquardt algorithm determine the carbonate formation porosity not only possesses the merits of algorithm stability, but also has a superior truth value to that of traditional methods.