Modeling Strategies Based on Multiple Neural Network Systems Applied for a Monopolar and Bipolar Electrocoagulation

Abstract

The objective of this research was to evaluate the efficiency of an electrocoagulation system using iron and aluminum electrodes, arranged in both monopolar and bipolar arrangements, for the removal of acid red 18 dye. Experimental and modeling approaches were employed to investigate the system’s performance. The effects of operating parameters: including initial pH (3–9), current density (0.4–5.6 mA cm−2), charge passed (2.16–21.6 C cm−2), and initial dye concentration (50–300 mg l−1) were studied. The results demonstrated that an increase in electric current intensity and passed charge enhanced the removal of COD and dye. However, to minimize energy consumption, these parameters were optimized for different dye concentrations. The monopolar arrangement exhibited favorable performance for the both electrodes, primarily due to reduced ohmic drop effect, although the iron electrode generated sludge with better settling characteristics. The monopolar iron electrode consumed the least energy (38.3 kWh kg−1 COD). Experimental evaluation was conducted to assess the influence of key electrolysis process parameters on dye and COD removal. Additionally, neural network models, employing radial basis function and multilayer perceptron approaches, were utilized to predict system outputs based on initial characteristics (COD and dye) and operation conditions. The neural network models provided accurate predictions, offering practical insights for experimental applications.