分子动力学和人工神经网络的算法耦合研究  

Studies of arithmetic coupling between molecular dynamics and neural network

作　　者：雷博[1] 饶金理[1] 覃武[1] 何克波[1] 唐玲[1] 霍明[1] 江建军[1] 

机构地区：[1]华中科技大学电子科学与技术系,湖北武汉430074

出　　处：《功能材料》2009年第2期284-286,290,共4页Journal of Functional Materials

摘　　要：对比了分子动力学和人工神经网络两种不同模拟算法的主要特点,提出了将这两种算法相互耦合,即将分子动力学的模拟结果作为人工神经网络的训练样本,训练后的人工神经网络用来预测。利用分子动力学建立了金刚石表面化学气相沉淀的模型,运用两种算法的耦合计算了碳原子在金刚石表面吸收概率,解吸收概率和散射概率。计算表明,这两种算法的耦合可节省计算资源,同时保证了一定的精确度。From the view of computer simulation, this article compared the major characters of two different simulation arithmetics, molecular dynamics and neural network. The method of arithmetic coupling between the two simulation arithmetic is figured out, that is to make the simulation result as the the training sample of neural network and to use the trained neural network to predict new input data. Using molecular dynamics,we set up the the model of chemical vapour deposition on the diamond surface. Using the coupling of the two arithmetics, we have canculated the the chemisorption probability of the carbon atom on the diamond surface. Simulation resuits show that the coupling of the two arithmetic can save the calculation resourse and make sure of the precision at the same time.

关 键 词：分子动力学模拟 人工神经网络 算法耦合 薄膜生长 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构] TP183[自动化与计算机技术—计算机科学与技术]