Molten salt phase diagrams calculation using artificial neural network or pattern recognition-bond parameters: Part 2. Prediction of phase diagrams of Ternary molten salt systems

摘要

Using pattern recognition method or the three layered-artificial neural network (ANN) method with back propagation algorithm, and with chemical bond parameters as features or inputs, the prop- erties or characteristics of the phase diagrams of some ternary molten salt systems have been investigated. The trained ANN was used to predict the formability of ternary intermediate compound, the type of phase dia- gram. Moreover, the results of computerized prediction for two unknown systems are confirmed by experimen- tal work.