Determination the parameters of model the technological process of diffusion in sugar production on the basis of neuro network identification

Abstract

The article considers the sugar industry's importance in the country's food supply. The complexity of technological processes characterizes sugar factories. A system analysis of the technical operation of diffusion and an analysis of indicators of the diffusion process in the diffusion equipment were carried out. An analysis of the mathematical support of automated control systems of complex technological processes is given. The assessment of material, quality, and energy indicators was carried out, as well as the formation of the necessary information variables, which makes it possible to determine the structure of the technological process control system. To effectively manage the complex dynamic technological process of diffusion, it is proposed to use neurocontrol. The use of linear regression models was considered to determine the type of mathematical models in the system of automated control of technological processes in the diffusion department of the sugar factory. When building regression models, a significant number of input and output components of the technology diffusion process were considered. Two variants of regression models of the diffusion process were considered. The obtained regression models were analyzed, and the significance of the indicators of the diffusion process was determined. For an adequate description of the complex diffusion process in an automated control system, it is proposed to use models based on neural network identification. Significant indicators of regression models were taken as the main indicators of the process. The construction of mathematical models that make it possible to respond to changes in the technological process adequately boils down to the issue of building neural network models based on a multilayer perceptron and a radial base network. The most effective model structures were determined from various types of neural network models. The control activation time when using the proposed model was 3 minutes. The error of identification of the received models was 5-7%. Keywords: diffusion equipment, mathematical model, neural network, regression equation, control, identification.