RSM and ANN-GA modeling of colloidal particles removal from paint wastewater via coagulation method using modified Aguleri montmorillonite clay

Abstract

In this study, response surface methodology (RSM) and artificial neural network (ANN) based on statistically designed experiments (CCD) were used as tools for simulation and optimization of total dissolved and suspended particles (TDSP)(colloidal particles) removal from paint effluent using coagulation method. A feed forward neural network model with Levenberg - Marquard (LM) back propagating training algorithm was adapted to predict the response (TDSP). The studied input variables were dosage, time and temperature. The raw montmorillonite clay (RMC) and modified montmorillonite (MMC) clay were characterized for elemental/structural elucidation using XRF, FTIR and SEM techniques and the result indicates that RMC is predominantly sodium montmorillonite. The performance of the ANN and RSM model showed adequate prediction of the response with R2 of 0.9504 and 0.9403, respectively. The RSM model predicted an optimal TDSP removal efficiency of 91.9% at 3 ​g/L, 38 ​min and 37 ​°C and validated experimentally as 89.8%. The artificial neural network-genetic algorithm (ANN-GA) predicted optimal TDSP removal of 90% at 3 ​g/L, 25 ​min and 30 ​°C and validated as 90.6%. The results obtained indicate that the ANN-GA was a better and more effective optimization tool than RSM in consideration of its higher R2.