基于混合遗传算法的人工神经网络模型及其对有机化合物熔点的预测  

Artificial neural network based on hybrid genetic algorithm and prediction of melting points of organic compounds

作　　者：彭黔荣[1] 杨敏[2] 石炎福[1] 余华瑞[1] 刘钟祥 

机构地区：[1]四川大学化工学院 [2]四川大学化学学院,四川成都610064 [3]贵阳卷烟厂技术中心,贵州贵阳550003

出　　处：《化工学报》2005年第10期1922-1927,共6页CIESC Journal

摘　　要：为了避免BP神经网络在训练过程中收敛于局部极小的缺陷,采用自适应交叉变异、最优保存的混合遗传算法对BP网络的权值和阈值进行优化,从而提出一种新的基于混合遗传算法的神经网络模型.该算法首先对一给定的网络结构,采用混合自适应交叉变异和最优保存策略,取各自的长处,用尽可能少的搜索代数找到问题的最优解,从而既防止算法陷入局部最优,又保证算法有较好的平均适应值和最佳的适应值个体.采用上述优化策略的人工神经网络可明显改善收敛的稳定性和收敛速度,并确保网络收敛于全局极小点.人工神经网络运用于物性数据的预测是一个具有潜力和有待开发的领域.运用该模型,根据有机化合物的分子量、临界密度、正常沸点和偶极矩,对其熔点进行预测.预测结果表明:提出的混合遗传算法神经网络优于其他算法神经网络,而且预测结果优于文献上已有的Joback方程和许氏方程的计算值.In order to avoid local minimum points during training BP network, adaptive crossover mutation algorithm and elitist model algorithm were used, which could optimize the weight and bias of BP network so as to build a hybrid genetic algorithm neural network. The algorithm consisted of the respective advantages of adaptive crossover mutation algorithm and elitist model algorithm. The best solution of problems was obtained by using as few generations as possible. In this way, local minimum point was prevented and the better average fitness and individual of best fitness were obtained in the algorithm. The artificial neural network with the above optimization strategy could greatly improve stability and speed of convergence and guarantee convergence to a global minimum point. Application of artificial neural network to predict physical properties of materials is a promising field to be developed. In this paper the proposed artificial neural network based on hybrid genetic algorithm was used to predict melting points of organic compounds. The molecular mass, critical density, normal boiling point and dipole moment of organic compounds were used in prediction. The predicted values based on artificial neural network with hybrid genetic algorithm were better than the values calculated with the Joback equation and Xu equation.

关 键 词：BP神经网络 遗传算法 自适应交叉变异 最优保存 物性预测 熔点 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP274[自动化与计算机技术—控制科学与工程]