Me-Me'-X系三元化合物形成规律的人工神经网络

来源期刊：中国有色金属学报1998年第4期

论文作者：李重河 荔建锋 王世华 陈念贻

文章页码：678 - 680

关键词：三元化合物； 相图； 计算机预报； 人工神经网络

Key words：ternary compound; phase diagram; computer prediction; artificial neural network

摘    要：用r_Me, r_Me', r_X, x_Me和x_Me'等原子参数作为人工神经网络的输入， 203个可信的Me-Me'-X体系的三元化合物的形成情况作为输出，研究了Me-Me'-X系三元化合物的形成规律，这里r是离子半径，x是电负性， Me代表一价金属， Me'代表二价金属，X代表卤族元素。利用所得规律预报了M-Eu-I（M=Li，Na， K， Rb和Cs）系三元碘化物的形成情况， 应用差热分析和粉末X射线法测定了它们的相图，预报结果和实验测定结果的对比是令人满意的。

Abstract: Using the atomic parameters of r_Me, r_Me', r_X, x_Me and x_Me as the inputs, the regularity of formation of ternary complex halides of MeMe'-X halides systems has been investigated by artificial neural networks（ANNs）. Where, r is the radius of the ion, x is the electronegativity of element, Me is the mono-valent metal, Me' is di-valent metal, X represents the F, Cl, Br or I. The regularity was found by training the ANNs with 203 known samples (such as Ag-Ca-Cl system and K-Mg-Cl system etc.). The formation of ternary complex iodides in M-Eu-I systems (where M represents Li, Na, K, Rb or Cs) was predicted by this trained ANNs. The predicted results are completely in agreement with the experimental facts.