Improving the Efficiency of a Multistage Evaporator Station for Sugar Production Using Neural Networks

The paper shows the process of developing a model of multistage evaporator station of a sugar factory based on a neural network. The neural network predicts the main performance indicators of the multistage evaporator station, such as the concentration of syrup at the output of the evaporator and the amount of evaporated water at the station. Unlike existing solutions, forecasting is performed in real time based on a set of indicators of the automation system (temperatures, flow rates, levels, pressure, density in the apparatus) and industrial laboratory (concentration of solids at the inlet to the evaporation station). The proposed neural network model has an MLP structure with 27 inputs, 2 outputs and one hidden layer with 20 neurons and predicts these indicators with an error of less than 5%. This allows to increase the informative support of the operator-technologist on the quality indicators of the enterprise without additional load on the industrial laboratory and without additional cost automatic quality devices, as well as to perform timely adjustment of the technological mode of the department.