ANN Model to Predict the Influence of Chemical and Biological Parameters on Iron and Manganese Accumulation

Water companies have adopted sophisticated risk-based management systems for managing water quality in water distribution networks (WDNs). Despite their efforts to comply with the standards for drinking water, they continue to receive customer complaints related to the water quality; discoloration is one such customer complaint. These complaints greatly undermine customers’ confidence in water companies. Discoloration is the result of release of accumulated material on pipe walls under stressed conditions. Therefore, understanding the causes of accumulation and the processes that influence accumulation of material is of paramount importance to water companies. In this paper, initially we identified various chemical and biological processes that highly influence the process of accumulation. Thereafter, using six years of water quality data, collected randomly, an artificial neural network (ANN) model was developed to predict Iron (Fe) and Manganese (Mn) accumulation potential. From the prediction profiler graph of the model, it was observed that increasing aluminium in the range 0 to 120 μg/l resulted in an increase in Fe and Mn accumulation potential due to increased sorption capabilities. It was also observed that free chlorine residual FCR has a dual effect on Fe and Mn accumulation potential. A cross-validation coefficient of determination, R2 of 0.70 for the ANN model indicates that the model is likely to predict accumulation potential well on new datasets.