DATABASES AND PREDICTION OF NEW INORGANIC COMPOUNDS. N.N. Kiselyova, A.A.Baikov Institute of Metallurgy of Russian Academy of Sciences

Application of artificial intelligence (AI) methods to processing large information volumes of databases (DB) on substance and material properties allows to find the regularities in data and to use of found regularities for the prediction of the possibility of forming inorganic compounds and for the estimation of their properties.

DB on inorganic ternary compounds' properties, DB on phase diagrams of the systems with intermediate semiconducting phases, and DB on crystals with acousto-, electro- and nonlinear optical properties are discussed.

The advantages of AI methods (a possibility of processing large information volumes, the search for complicated and multidimensional regularities and so on) are illustrated by the example of predicting the new ternary compounds and estimating some their properties. The method of computer learning (the method of concept formation in pyramidal networks) was used for processing information of DB inorganic ternary compounds' properties. The examples of known compounds, taken from DB, were used for computer learning. Each ternary compound was described in computer memory as set of component (chemical elements and/or binary compounds) property values with indication of information about belonging to one of the considered class of compounds. The search for regularities is based on formation and analysis of the semantic pyramidal network which receptors are the component property values. The prediction requires only the knowledge of the values of the component properties. The comparison of these predictions with the new experimental data shows that the average reliability of predicted ternary phases exceeds 80 percent.

The information-predicting system is developing for the data retrieval on the known compounds and for automation of the prediction of inorganic compounds, not yet synthesized, and the forecasting of its properties. This system will employ the DB on inorganic compounds properties, AI system, DB on elements properties, knowledge base, conversational processor and monitor.