Analysis and preparation of data during modeling of plasma-chemical processes of obtaining nanosystems

Abstract

One of the new directions of industrial chemical technology is plasma chemical. Despite sig-nificant experimental and theoretical research, physicochemical processes are complex and not ful-ly understood for scaling and control in mass production. In low-temperature plasma, chemical processes, regularities of reactions, and the basics of plasma chemical technology require computer modeling. The experimental data of natural experiments require additional analysis and prepara-tion for their further application and the construction of adequate simulation models of plasma-chemical processes for the development of nanosystems. In the course of the preliminary experi-ment, which includes two stages, the input and output variables of the application task were exam-ined. A software toolkit has been developed for the analysis of experimental data, which allows col-lecting statistical information and visualizing some processes. The software approaches to building analytical and simulation models of plasma-chemical processes of obtaining nanosystems were considered and implemented. Two possible approaches to the construction of homomorphic models are reviewed: classical and with the use of artificial intelligence. Without limiting judgment, linear and non-linear regres-sion models were reviewed. The first type of model is multiple regression and methods that use reg-ularization. As an alternative, the construction of nonlinear models based on the method of support vectors is proposed. Vector and multi-vector regressions are considered. Multidimensional regres-sion is aimed at studying and constructing a mapping of a multidimensional input feature space into a multidimensional output space. The obtained numerical results were analyzed and it was pro-posed to use nonlinear models of plasma-chemical processes for obtaining nanosystems for com-puter simulation, in particular: multi-vector regression.