Monday, April 30, 2007

Evaluation of fuzzy logic and artificial neural networks for the development of software sensors for bioethanol production monitoring

Elmer Ccopa Rivera, Nadson Lima, Rafael Ramos de Andrade, Aline Carvalho da Costa, and Rubens Maciel Filho. School of Chemical Engineering, State University of Campinas, Laboratory of Optimization, Design and Advanced Control, Cidade Universitária, Barão Geraldo, CEP 13081-970, Campinas-SP, Brazil

One of the main difficulties in the bioethanol fermentation process nowadays is the lack of robustness of the fermentation in the presence of fluctuations in the quality of raw material, which leads to changes in the kinetic behavior and affects yield, productivity and conversion. This can be avoided by frequent adjustments in the operational conditions and control settings of the process, which requires an efficient monitoring with reliable sensors. In this work we developed software sensors based on a Fuzzy Relationship of Takagi-Sugeno Type and a Multilayer Perceptron Neural Network to infer the process key variables (concentrations of biomass, bioethanol and substrate) from secondary measurements, such as pH, turbidity and CO2 flow rate. The objective is to have a robust sensor, which describes the process even with operational conditions changes.