Aplicação das tecnicas de redes neurais e de analise de componentes principais na modelagem de uma lagoa aerada da RIPASA S/A

AUTOR(ES)
DATA DE PUBLICAÇÃO

2000

RESUMO

In recent years, computer-based methods have been applied to many areas of environmental chemistry. In the process industry the use of modern control strategies is required due to increasing demands on the quality of its effluent treatment systems. In this work a wastewater treatment system of a pulp and paper industry has been studied using Artificial Neural Networks (ANN) and the Principal Components Analysis to predict output environment parameters (BOD). Control process data sets generated from input and output of the current treatment system (an aerated lake) are used in this research. Variation within sampling of some auxiliary and process parameters including chemical oxygen demand (COD), biochemical oxygen demand (BOD), flow, pulp and paper production, pH and suspended solids are evaluated over a two-year period. Predictive models are presented calculated from ANN and Principal Component Regressions (PCR) for the estimation of biochemical oxygen demand, one of the main process control variables. The results show that neither principal component regression nor artificial neural network treatment is satisfactory when used separately in modeling and simulation. Neural network presents superior results for the training set but poorer ones than those from PCR for the test set. One explanation is that there are too few data resulting in an overfit of the training set. Best prediction performance is achieved when the data are preprocessed using PCA, before they are fed to a backpropagated neural network composed of three neurons in a hidden layer and the Delta-Bar-Delta (DBD) learning algorithm. The PCA technique orthogonalizes the input original variables and helps the ANN nonlinear mapping

ASSUNTO(S)

redes neurais (computação) meio ambiente papel - industria analise de componentes principais

ACESSO AO ARTIGO

Documentos Relacionados