PREDIÇÃO DE PARÂMETROS DE QUALIDADE DO BIODIESEL UTILIZANDO REDES NEURAIS ARTIFICIAIS

Abstract

The prediction is an action for decisions making. The smaller the error of the prediction, the lower the loss by decision taken. One technique that has been used very successfully to predict is the Artificial Neural Network (ANN). It is a computational tool, inspired by the human brain and has a great ability to model complex real-world problems through a phase called training enables it to learn the characteristics of the proposed problem. The ANNs have the potential to discover complex relations between input data and response so they become very useful in the prediction area. For optimal performance in various areas of knowledge, it was decided to assess its efficiency in a sector, the Biodiesel. It is a renewable fuel generated from oils of animal or vegetable to be used in diesel engines. It has several benefits over mineral diesel and at the time, their production and use gets great stimuli federal government, through the National Program for Production and Use of Biodiesel. Thus, interest in research grew and Biodiesel in the state of Maranhão is performed by the Laboratory of Chemical Research & Analytics (LPQA / UFMA). It was then that their research can be accelerated if the analyzes of quality biodiesel that are influenced by the chemical composition are replaced by a single analysis of chemical composition and this composition is used to predict the results of other analyzes of quality. For this, was developed the Sys-PANN, software able to find the best ANN for each type of analysis and predict your results.