Abstract:The temperatures of the liquidus lines or surfaces in some binary or ternary systems have been estimated using artificial neural network(ANN) with the chemical bond parameters of ions and the melting points, and using the compositions of the constituents as inputs and the properties of phase diagrams as outputs. The computerized values are in good agreement with the experimental ones. Moreover, an expert system has been built up, which consists of two parts, one is the database produced by the experimental data of the various known phase diagrams and chemical parameters for the retrieval of the features of the existing phase diagrams; the other is the knowledge base produced by the trained ANN for the estimation or prediction of some features of the unknown phase diagrams.

Key words: phase diagram calculation　artificial neural network　bond parameter　molten salt system　 expert system