Otimização de misturas no setor sucroalcooleiro: padronização de álcool e redução de cor de caldo

Abstract

Brazil is the major producer of sugar cane in the world, being sugar and ethanol its main products. Different types of ethanol and sugar are produced throughout the industries in Brazil. Problems regarding quality in the final products may bring economical prejudice due to fines and reprocessing. It has been seen in previous work that it is possible to produce blends of ethanol with different quality standard. Furthermore, it is known that in Brazilian industry there is an issue about producing sugar with lower colors. The mixing of sugar cane juice with different properties can be used in sugar cane plants to use darker juices so that the rise in color due to a certain sugar cane variety may be compensated by the addition of a juice with lighter color so that the resulting mixture is ideal for the process. The ethanol and juice mixtures are studied in this work through mixture optimization techniques. In this work, Statistica v. 12.0 and GAMS v. 22.2 softwares were used respectively for mixture design and mixture optimization. For the mixture modeling the centroid-simplex was used in both cases. The data were adjusted by linear and quadratic models according to variance analysis and the two F tests. Computational codes in GAMS have been developed in which the user can enter industrial data (juice properties, ethanol properties, prices, etc.) and the algorithm returns the optimal proportions of the components for the maximum profit and attendance to all pertinent constraints. It has been observed in this work the importance of using non-linear models to adjust some parameters, such as pH and acidity for the ethanol mixing. The quality and economic impacts were observed according to the used model. It the juice problems, it was observed a linear behavior for pH and ICUMSA color whereas there was a quadratic behavior for the brix. Using study cases, the application of the developed models was evaluated to support the production planning in the sugar cane industry. The developed strategy showed a viable alternative to optimize mixtures using a minimum and relevant amount of experimental data in the model development.