自适应递阶遗传神经网络测井岩性识别方法研究

摘要

Aimed at the disadvantages of slowly convergence rate and the difficulty of choosing appropriate network topology and learning parameters in lithology identification model,a neural network model based on adaptive hierarchical genetic algorithm is proposed.The network topology and learning parameters are coded by hierarchical genes in this mode.The Sigmoid function with nonlinear characteristic is introduced into the design of genetic operators,making the crossover and mutation in genetic evolutionary changed adaptively,which reducing the probability of premature convergence in genetic algorithm.Compared with the three-layers neural network model and the optimized neural network model by genetic algorithm,The experimental results show that the average recognition accuracy of this method is 95.67％,which is superior to the other two models in accuracy,iteration convergence,time-consuming and accuracy.%针对传统神经网络岩性识别模型存在收敛速度慢、难以选择合适的网络拓扑结构和学习参数等问题,提出一种自适应递阶遗传优化神经网络模型.该模型通过采用递阶遗传染色体编码方式优化神经网络拓扑结构和权阈值参数,并将具有非线性特性的Sigmoid函数引入到遗传操作算子设计中,使得交叉变异算子在遗传进化阶段能自适应变化,从而减少遗传算法陷入“早熟”的机率.与三层传统神经网络和基于遗传算法优化的神经网络模型相比较,实验结果显示,该方法平均识别准确率达到95.67％,在准确率、迭代收敛次数、运算耗时和精度指标上均优于其它两种模型.