Métodos de construção de analisadores virtuais para estimação de teor de enxofre de hidrocarbonetos

Abstract

The acquisition of a process analyzer to obtain, in real time, a products sulfur content may be expensive and its installation can take a while, due to the need of stop the processing unit. Are presented in this work, solutions for building soft sensors with the objective to infer the sulfur content in the product. To achieve that goal, multiple linear regression and artificial neural network techniques are used and their performance are compared in order to obtain the best configuration for constructing a sulfur content soft sensor.