In present study, a hierarchical knowledge-based expert system, named HYDROPERT, to predict binary azeotrope formation in hydrocarbon mixtures was created and investigated. Specific knowledge included in azeotropy on hydrocarbon mixtures and the implementation of the expert system are described along some of the several components of expert system applications : knowledge representation strategy and levels of knowledge abstraction, inference machine, user interface, explanation facilities. The knowledge base is hierarchically structured with the multiple levels of domain-specific knowledge such as the azeotropic data bank as the lowest level, component-specific compiled heuristic rules as the second level, group-oriented compiled heuristic rules as the third level, and generic class-oriented model-based heuristic rules as the highest level. The predictive capabilities and generality of the expert system can be highly enhanced through the integration of different kinds of domain knowledge into the hierarchical structure. The expert system predicting the binary azeotrope formation in hydrocarbon mixtures may be a useful tool for many chemical engineering activities, especially such as process synthesis and design.
Buchanan BG, Sutherland GL, Feigenbaum EA, "Heuristic DENDRAL; A Program for Generating Explanatory Hypo-theses in Organic Chemistry," Machine Intelligence, Vol. 4, Edinburgh University Press, 1969
Edan T, Abraham T, Michael BK, Chem. Eng. Sci., 36, 759, 1980
Rebon R, "Knowledge Engineering Techniques and Tools in the PROSPECTOR Environment, Technical Note 243 SRI, Projects 5821, 6415 and 8172," Artificial Intelligence Center, SRI International, June, 1981