Prediction of organic matter removal from pulp and paper mill wastewater using an artificial neural network

Authors:
Open Access
Publisher website
Google Scholar
Abstract
The Multilayer Perceptron Model was developed for predicting organic matter removal from pulp and paper mill wastewater. The original database covered a period of 1,427 consecutive days and contained the most frequently measured parameters. Three models were constructed by applying the technique of Principal Component Analysis, which selected principal components, discarded original variables and excluded possible outliers. The data were randomized and divided into training, validation and testing sets. The training algorithm was the Levenberg–Marquardt type, which is an adaptation of the back-propagation algorithm. The learning rate was 0.05, and the evaluation criteria used were the mean square error and the linear correlation coefficient. A marked difference was observed in the predictive performance when the organic matter load was used as an input. The model M4, which was built by discarding the two variables pH and EC, proved to be the most suitable and the simplest model obtained. However, choosing the best model cannot be done arbitrarily. It will be necessary to use various statistical parameters and perform comparisons of models with different sizes and structures in order to select the best model.

This publication has 18 references indexed in Scilit: