【摘要】 应用人工神经网络对铀矿测井解释中岩性识别和孔隙度预测等问题进行了研究。采用了一种改进的 BP算法 ,其方法具有收敛速度快、避免网络陷入局部最小和出现振荡现象、优化网络结构等优点。提出了一种基于统计的学习样本生成方法 ,使样本生成问题规范化。使用该方法生成的样本真实可靠 ,具有代表性 ,可大大提高样本质量。实际应用网络进行岩性识别和孔隙度预测 ,取得了令人满意的结果更多还原

【Abstract】 Application of BP Neural Network Technique in Log Interpretation of Uranium Deposits.WLT, 2001, 25(4): 308-310 The rock character identification and porosity prediction in log interpretation of uranium deposits are studied by artificial neural network. An improved BP algorithm and a learning sample generation method based on statistics are performed. The algorithm has many advantages, such as high convergence speed, avoid running into local minimum and occuring oscillation, in addition, it also optimizes the network structure and so on. The generation of learning sample used in this paper is normalized and this operation ensures the validity of the sample. A reliable result has been obtained in rock character identification and porosity prediction by the neural network.更多还原