Application of non-linear regression and soft computing techniques for modeling process of pollutant adsorption from industrial wastewaters

The process of pollutant adsorption from industrial wastewaters is a multivariate problem. This process is affected by many factors including the contact time (T), pH, adsorbent weight (m), and solution concentration (ppm). The main target of this work is to model and evaluate the process of pollutant adsorption from industrial wastewaters using the non-linear multivariate regression and intelligent computation techniques. In order to achieve this goal, 54 industrial wastewater samples gathered by Institute of Color Science & Technology of Iran (ICSTI) were studied. Based on the laboratory conditions, the data was divided into 4 groups (A-1, A-2, A-3, and A-4). For each group, a non-linear regression model was made. The statistical results obtained showed that two developed equations from the A-3 and A-4 groups were the best models with R2 being 0.84 and 0.74. In these models, the contact time and solution concentration were the main effective factors influencing the adsorption process. The extracted models were validated using the t-test and F-value test. The hybrid PSO-based ANN model (particle swarm optimization and artificial neural network algorithms) was constructed for modelling the pollutant adsorption process under different laboratory conditions. Based on this hybrid modeling, the performance indices were estimated. The hybrid model results showed that the best value belonged to the data group A-4 with R2 of 0.91. Both the non-linear regression and hybrid PSO-ANN models were found to be helpful tools for modeling the process of pollutant adsorption from industrial wastewaters.