Modeling and prediction of COD and turbidity removals from dairy wastewaters by Fenton process using RSM and ANN

The main objective of this study was to use response surface methodology (RSM) and artificial neural network (ANN) to model and predict chemical oxygen demand (COD) and turbidity elimination from a synthetic dairy wastewater treated by the Fenton process. The experimental design was realized using RSM and, in particular, a face-centered composite (CCF) design. The responses were fitted by a second-order model in the form of quadratic polynomial equation, and the experimental data were analyzed by ANOVA (analysis of variance). The obtained results showed acceptable coefficients of determination (R²) for COD (0.836) and turbidity (0.870), indicating that the predicted values fit well with the real data when using quadratic models. Moreover, it was noticed from the iso-surface plots that the processing parameters have a great influence on COD. Among these factors, the most important one was the pH. However, they had a little effect on turbidity. The COD removal percentage was increased with pH decreasing and increasing reaction time, and iron sulfate concentration. In parallel, the same data generated from RSM were utilized to create the ANN model. In order to evaluate the accuracy of the predictions of the models, the R², the mean square error (MSE), and the mean absolute error (MAE) were used, and the actual and predicted responses were compared. The results confirmed that the created ANN model has sufficient reliability in predicting the outputs for a different set of input values, with R² of 0.980 for COD removal and 0.952 for turbidity removal.