The International Journal of Cuban Studies
(Online) ISSN 1756-347X
Neural network modelling of reference purity of Cuban molasses
Dr C.O. Gozá León, Dr C.H. Kazemian, Dr C. J. Hormaza, Lic. Yatali Montero-Sanchez Rojas and Dr C. R. Santana report on ground-breaking multi-disciplinary research related to improving the profitability of Cuba’s sugar industry.
Summary
The purity of molasses is a very important process variable in sugar manufacturing; it depends on many factors, both external (related to agriculture) and internal (related to the industry). In order to assess the purity of molasses, a benchmark is necessary, not only for sugar factories but also for ethanol distilleries. This is generally available in the form of a 'target purity equation'. A number of different target purity equations have been proposed over the years. In spite of the fact that many target purity formulae have been developed, they still have many limitations and do not cover all the uncertainties associated with the purity of molasses. Artificial neural networks allow us to take into account a greater number of variables and are also very interactive. Consequently a neural networks technique is very suitable to model the highly non-linear reference purity of molasses. In this paper the results obtained in modelling of reference purity of Cuban molasses are presented using a novel neural networks approach. Computer simulation results demonstrate that the neural networks technique can overcome the limitations of the traditional approach.
Download the full article in pdf
