Artificial neural network prediction of chemical oxygen demand in dairy industry effluent treated by electrocoagulation

Abstract

We used electrocoagulation to reduce the chemical oxygen demand of dairy industry effluent. The effects of operating parameters were evaluated, including the electric current density, initial effluent pH, electrolysis time and distance between electrodes. The characteristics of the effluent, namely, the solids content and its fractions, turbidity and chemical oxygen demand, were also considered. An artificial neural network was constructed to model chemical oxygen demand after electrocoagulation; it was trained and validated, yielding a correlation coefficient of 0.96 between predicted and experimental values. Input variables were ranked by their relative importance for the prediction of chemical oxygen demand after treatment by electrocoagulation. Among effluent the Total Dissolved Solids concentration had the greatest relative importance, followed by the chemical oxygen demand. It can be concluded that an artificial neural network can predict chemical oxygen demand after treatment by electrocoagulation. In practice, operating parameters may be adjusted to obtain a greater reduction of chemical oxygen demand and to allow automation of the handling process.