DESENVOLVIMENTO E APLICAÇÃO DE REDES NEURAIS ARTIFICIAIS PARA A OTIMIZAÇÃO DA FILTRAÇÃO DIRETA NO TRATAMENTO DE ÁGUA

Abstract

The goal of the work was the mathematical modelling of particle removal of a direct up flow filtration through the development of an artificial neural network. The mathematical modeling by artificial neural network type MLP – Multi Layers Perceptron was, programmed in the language of MatLab. The RNA was trained with experimental data taken from a filtration of up flow filtration in a pilot plant of double filtratio. During the training was analyzed the behavior of RNA according to the ratio of the amount of neurons and the activation function performing changes in search of their optimization, the results obtained were satisfactory with coefficient of determination (R²) between 93 and 99%. Therefore, the mathematical model is suitable to describe the particle removal in direct up flow filtration.

Index Terms - RNA, direct filtration, doble filtration, MLP,particle removal.