A knowledge-based expert system for the prediction of ternary azeotrope formation in organic mixtures

New functions of AZEOPERT [Kim and Simmrock, 1997] were investigated to predict the occurrence of ternary azeotropes and their azeotropic compositions in an organic mixture. This study describes its new problem-solving strategy. The knowledge base of AZEOPERT for ternary azeotropes is hierarchically structured with the several levels of domain-specific knowledge on ternary azeotropy. First, an azeotropic data bank including ternary azeotropic experimental data was implemented in AZEOPERT as the lowest level. It may be used to determine whether or not ternary azeotropic experimental data for the consulted organic mixture are already available. Moreover, compiled heuristic knowledge as the second level and class-oriented model-based knowledge as the highest level were implemented in the knowledge base. The problem-solving strategy through the integration of model-based reasoning into compiled reasoning gives a very efficient, general way for the prediction of ternary azeotrope formation in a wide varitey of organic mixtures, and especially, in unknown mixture systems.