Calibração de um modelo de concentração de sólidos na descarga de espessadores de minério de ferro via Algoritmo Genético

Abstract

The importance to use mathematical models properly adjusted to represent precisely the dynamics of processes, thus allowing more accurate analysis and more assertive decisions, is unquestionable in several industrial sectors, such as the mineral processing sector, for instance. In this sense, this work proposes an application of Genetic Algorithm to configure a nonlinear mathematical model of a conventional thickener, with variable transverse section and submitted to flocculant addition. In order to simulate and apply the Genetic Algorithm, data banks of a real concentrate thickener from Vale S.A. in Carajás, Pará, Brazil, are used. The comparative analyses focus on the underflow concentration values, one output of the model and one variable frequently measured in the thickening process. Finally, metrics associated to estimation error are used to evaluate the precision of the estimates in relation to the real values of concentration, evidencing a significant improvement of the results by using the proposed technique.