Berest O. Methods of functional status classification in the automated process control system of crystal growth

Thesis is devoted to the increasing of the functional efficiency of decision support system (DSS) of automateted control system of the growing large alkali halide scintillation single crystals from the melt under conditions of a priori uncertainty. Providing DSS with adaptability is reached by machine learning using extreme intellectual information technology of data analysis, which is based on maximizing the ability of information system. A new information-extreme method of analysis and synthesis, which is based on cylindroid shapes of class recognition containers was developed. Also it is proposed to use three alternative information criterion of functionality optimization, which allows to generate unmistakable decision rules for conditions of complex configurations and distributions of realizations in feature space. Method of time interval formation of corrective action and observation interval optimization of unsteady process when regulators parameters remain unchanged was proposed. In addition, visualization method of multidimensional data presented in binary form is also improved. This allows to represent current functional status of the process as a point in the plane. It was also developed necessary DSS software and it was considered physical implementation as a separate module to the second level of the automated control system of the growing large alkali halide scintillation single crystals from the melt. Method of data visualization under the machine learning is also was improved.