ANN & regresssion analysis based predictions of BOD5 & COD for refinery wastewater

Abstract

Industrial wastewater quality is indicated by several physico-chemical and biological parameters. If a Suitable correlation is established between them, some difficult and not instantaneously available parameters can be easily predicted. Such correlations are traditionally achieved by regression analysis. However, non-linear fluctuations are not easily represented by these correlations. Models based on artificial neural networks (ANNs) are fast emerging as an alternative tool to predict and forecast water quality parameters based on a well-defined set of training data that are easily available. The present study reports the correlations for BOD<inf>5</inf> and COD with TOC for a refinery wastewater. Additionally, 12 ANN based models were developed to forecast the BOD<inf>5</inf> and COD by considering other water quality indices as the input data. The results from this study indicate that ANNs are simple and reliable, under adequately trained conditions.