INFORMATION-PREDICTING SYSTEMS FOR THE DESIGN OF NEW MATERIALS

In this paper we discuss the main principles for the cybernetic prediction of inorganic substances which would have predefined properties. These predictive techniques are based on machine learning strategies. The efficiency of the proposed approach is illustrated by comparing the results of predicting the properties of new substances with experimental data. Examples showing the application of these techniques to the prediction of those materials used in the electronics industry are described. Also discussed are the components and the organization of the computer-aided information-predicting system developed by the Institute of Metallurgy of the Russian Academy of Sciences. A detailed description of the databank for the properties of the ternary inorganic phases is provided, as well as an explanation of the main difference between the information-predicting system and the expert systems.