Modeling the removal of Sunfix Red S3B from aqueous solution by electrocoagulation process using artificial neural network

Abstract

This study presents an application of artificial neural networks (ANNs) to predict the dye removal efficiency (color and chemical oxygen demand value) of Electrocoagulation process from Sunfix Red S3B aqueous solution. The Bayesian regulation algorithm was applied to train the networks with experimental data including five factors: pH, current density, sulphate concentration, initial dye concentration (IDC), and electrolysis time. The predicting performance of the ANN models was validated through the low root mean square error value (9.844 %), mean absolute percentage error (13.776 %) and the high determination coefficient value (0.836). Garson, Connection weight method and neural interpretation diagram were also used to study the influence of input variables on dye removal efficiency. For decolorization, the most effective inputs are determined as current density, electrolysis time and initial pH, while COD removal is found to be strongly affected by initial dye concentration and sulphate concentration. Through these steps, we demonstrated ANN’s robustness in modeling and analysis of electrocoagulation process.

Keywords

• color removal
• Garson’s algorithm
• sensitivity analysis
• textile wastewater