Models for colour formation in evaporation stations Part 1: Conventional models

Abstract

This study is dedicated to making a comparison among models that describe juice colour formation during thickening in an evaporator station. The self-developed neural model for juice colour formation, which is outlined in this study, is compared with three conventional models, already developed by Vukov, de Visser and Smejkal. The initial data for developing respective verification of both conventional and neural models were acquired during field studies, carried out from 2005 to 2009 at four sugar factories operated by Südzucker Polska. It was claimed that the colour of sugar juice predicted by a conventional mathematical model, when the juice is thickened in an evaporator station, adheres to the actual process only for beginning compartments of the evaporation station: 1, 2, 3 – i.e. when the juice is relatively weakly coloured and its dry substance is relatively low. The application of unmodified conventional models to the simulation of juice colour formation, when the evaporator station is to be newly designed or upgraded, demonstrates poor convergence with data from measurements. Part 2 of the paper (“A static neural and a recurrent neural model”) shall provide results of anticipated thin juice colour change during thickening in sugar factory evaporator stations, where the results were obtained from the model of artificial neural networks.