【摘要】 针对地质条件复杂、参数变化大的油藏地层对比问题 ,提出了一类模式识别方法 ;该方法利用粗集理论中的决策规则最小化算法 ,去掉冗余的属性特征、样本及相互矛盾的学习样本 ,将数字化测井曲线和地层参数预处理 ,转化为二值点阵图像模式 ,压缩点阵数据编码 ,提取和记忆曲线所表征的地层模式特征 ,并利用多层神经网络训练条件属性与模式类别之间的映射关系 ;所得神经网络的记忆能力和推广能力强 ,用于解决地层对比问题具有良好的适应性更多还原

【Abstract】 In this paper we propose a pattern recognition method for the oil field stratum contrast problems of complicated geological condition and varying parameters. First, we use rule set’s least decision\|making algorithm in rough set theory, get rid of redundancy and incompatible characteristics in pattern set. Second, we change digital well\|logging curves and stratum parameters into two\|value lattice image pattern by pretreatment, distill and memory stratum pattern characters that the curves expressed by compress lattice data code. Lastly, we use multi\|layer neural networks to train mapping relation between condition property and pattern sort. The neural network obtained has strong memory ability and extension ability and better adaptability to solve stratum contrast problem.更多还原