The determination of viscosity transporting property of essential oils, vegetable oils and biodiesel is mandatory for the chemical, pharmaceutical, petroleum and food industries. Over the past decades many mathematical models have been developed to predict the viscosity of mixtures, among them we can highlight the UNIFAC-VISCO and GC-UNIMOD models. These two models based on the theory of group contribution rely on the determination of group interaction parameters to have a viable application. Studies conducted in SLE (Separations Engineering Laboratory - FZEA / USP) has shown that the interaction parameters already determined, for the two models, predict unsatisfactorily viscosity non-ideal mixtures. Two of the factors for the low predictive ability of existing interaction parameters are: low numbers of data used in the optimization and the limited number of components. In order to address this deficiency proposes to build a comprehensive database, large and robust, for new optimization of interaction parameters for UNIFAC-VISCO and GC-UNIMOD models using genetic algorithm. Additionaly, it is proposed to use an artificial neural network for determining the viscosity of non-ideal mixtures based on the concept of molecular composition. The database will be developed in SQL, using as database manager MySQL. MATLAB is used as the programming language due to its ease of handling arrays and provide graphical user interface. The effectiveness of process optimization will be assessed by the average deviation between the calculated data and experimental data. The results of this project should contribute to the development of processes and projecting equipment in industries that use as raw materials essential oils, vegetable oils and biodiesel.
News published in Agência FAPESP Newsletter about the scholarship: