Computational intelligence for empirical modeling and optimization of methylene blue adsorption phenomena using available local zeolites and clay of Morocco

Abstract

The present study investigated the performance of natural Moroccan zeolites and clay in the removal of methylene blue (MB) from aqueous solutions. The deposits of the samples are extracted from the Teteoune (sample 1), Kenitra (sample 2), Khroubga (sample 3), and Benghrir (sample 4) regions of Morocco. The samples were characterized by the X-ray diffraction technique. Sorption experiments were carried out by batch experimental to examine the effects of contact time, solution pH, initial concentration of MB, and biosorbent dosage of samples. In the next step, an integrated methodology is adapted by applying the computational intelligence packages to this problem to develop empirical models and to conduct an optimization to maximize the removal percentage. The empirical or digital twin models are generated using machine learning techniques including regression tree (RT), support vector machine, ensemble, and gaussian process regression (GPR), a statistical technique including multivariate regression (MVR), and neural network techniques including artificial neural network (ANN) and group method of data handling (GMDH). The results have shown that the ANN technique having 10 hidden layers trained using the Bayesian regularization algorithm can accurately predict the removal percentage with an RMSE = 0.88, R2 = 0.9961, MSE = 0.67, and MAE = 0.43. The second-best technique is GPR followed by GMDH. The worst techniques are MVR and RT. The single-objective optimization was applied using simulated annealing (SA), particle swarm optimization (PSO), and genetic algorithm (GA) to maximize the removal percentage for each material to find the optimal mass, initial concentration, contact time, and pH. The results have revealed that the SA technique is not fully capable to find the true optima, whereas PSO and GA have shown conformity on the optimal removal percentage. The optimized indicator came out to be 85.32%, 90.23%, 93.25%, and 97.21% for material 1, 2, 3, and 4 which was 49%, 14.69%, 20.16%, and 45.11% for material 1, 2, 3, and 4, respectively. It is concluded that the sample from the region of Benghrir has the best removal performance of methylene blue from aqueous solutions marking up to a high digit of 97.21%.