(中南工业大学矿物工程系，长沙 410083；
中国科学院上海冶金研究所， 上海 200050)
摘 要: 以化学键参数作为人工神经网络的输入，实测相图数据为输出，将采用误差反向传播算法训练好的神经网络用于对未知相图作计算机预报，将数据库与知识库相结合，设计和开发出了一个检索和预报二元及部分三元熔盐系相图特征的专家系统。该数据库包括各类已知熔盐相图特征的实验数据及熔盐系各种元素的化学键参数，而知识库为训练好的人工神经网络，通过人机对话形式提供相图特征的各种信息。给出了3个实际应用例子， 对预报结果进行的实验验证表明该专家系统对未知相图特征的预报是可靠的。
关键字: 专家系统 人工神经网络 熔盐相图 化学键参数
（Department of Mineral Engineering, Central South University of Technology, Changsha 410083, P. R. China
Shanghai Institute of Metallurgy, Chinese Academy of Sciences, Shanghai 200050, P. R. China）
Abstract:An expert system for retrieval and prediction of the properties in some binary and ternary phase diagrams of molten salt systems has been built. The models obtained by chemical bond parameters-artificial neural network method has been used for computerized prediction. The data base consisting of the known properties of phase diagrams and chemical bond parameters and the knowledge base produced by the trained artificial neural network were included in this expert system. By man-machine interfacing, the formability, chemical stoichiometry, melting type and melting point of the intermediate compound of phase diagram can be retrieved or predicted.
Key words: expert system artificial neural network phase diagram of molten salt chemical bond parameter
