A hybrid artificial neural network with particle swarm optimization for estimation of heavy metals of rainwater in the industrial region-a case study

Abstract

The objective of this study was to explore the application of hybrid artificial neural network methods to predict heavy metals in rainwater based on major elements. Measurements of the heavy metals Pb, Cu, Zn, As, Ni, Hg, and Fe in soluble rain fractions were performed in rainwater collected at the Arak plain during the rainy seasons of 2012. In the soluble fractions, the concentrations of the heavy metals decreased in the order Fe, Pb, Zn, Ni, Cu, As and Hg. Enrichment factor related to the relative abundance of elements in crustal material were calculated using Fe as reference. The high enrichment factor (EF_crustal) suggested that, in general, heavy metals had an anthropogenic origin. Industrial activity and traffic are the source of heavy metals in the rainwater samples in the Arak city. Prediction of the heavy metals in the rainwater is important in developing any appropriate remediation strategy. This paper attempts to predict heavy metals of rainwater in Arak city using a new approach based on hybrid artificial neural network (ANN) with particle swarm optimization (PSO) algorithm by taking major elements (Cl, Mg, Na, SO₄) in rainwater. For this purpose, contamination sources in rainwater were recorded 50 data samples and several models were trained and tested using collected data. It determined the optimum model in each model based on four inputs and five outputs. The results obtained indicate that ANN-PSO model has strong potential to estimation of the heavy metals in the rainwater with high degree of accuracy and robustness.