Deep Fuzzy-NN modeling for the prediction of Zn(II) adsorption in columns using alkaline modified biochar: Integrated experimental and computational insights.

The precise prediction of adsorption process is significant in the optimization of pollutant removal systems. In this research, deep fuzzy neural network (DFNN) model was developed for the prediction of Zn(II) removal efficiency using alkaline activated neem bark biochar. The study has been tested using conventional Artificial Neural Network (ANN) model which was trained using levenberg-marquandt algorithm exhibiting limited predictive performance. Modeling performance of conventional ANN trained using Levenberg-Marquandt algorithm showed limited predictive ability with R2 of 0.4670. In the view of addressing the limitations, DFNN model with integrated sugeno based inference system of 81 fuzzy rules was optimized using Gaussian membership function. Among the five tested membership functions, gauss2mf achieved the best performance with R2 of 0.9999 and mean squared error of 0.00335. DFNN model exhibited high predictive accuracy with R2 of 0.9999 with low errors. The minimal deviation among experimental and DFNN predicted values was confirmed by the residual analysis. Training convergence has been achieved within 100 epochs with reducing errors from 0.0125 to 0.0016. Hybrid optimization algorithm with DFNN model outperformed ANN model with regard to predictive accuracy and computational efficiency. The integration of fuzzy logic with neural network modeling resulted in better predictive accuracy which has been useful in predicting the adsorption dynamic patterns.