Jujube stones based highly efficient activated carbon for methylene blue adsorption: Kinetics and isotherms modeling, thermodynamics and mechanism study, optimization via response surface methodology and machine learning approaches

Water contaminated by methylene blue (MB) dye was treated with activated carbon based on locally collected jujube stones. This activated carbon was characterized by physico-chemical methods after preparation and chemical activation. Response surface methodology (RSM) was used to maximize the MB uptake a dependent variable in Box-Behnken Design (BBD) with the initial concentration of methylene blue (400-700 mg/L), adsorbent dosage (0.6-1.6 g/L), contact time (30-540 min), pH (7-11) and temperature (20-50 degrees C) as inde-pendent variables. Then, the database created by BBD was further modeled using Gaussian Process Regression coupled with Bagging (Bootstrap Aggregation_Bag) and Dragonfly optimization algorithm (GPR_DA_Bootstrap). The results of the optimization analysis using the GPR_DA_Bootstrap model are shown to be superior to those of the BBD model. The experimental validation of the optimal conditions of the GPR_DA_Bootstrap model (X1 = 700 mg/L, X2 = 0.6 g/L, X3 = 540 min, X4 = 11, and X5 =50 degrees C) led to an MB uptake (501.01 mg/g) significantly higher than that of BBD (456.00 g/g). In addition, the very low error between the experimental and the predicted values given by the GPR_DA_Bootstrap model (8.64 mg/g) compared to that of the BBD model (22.19 mg/g), should be highlighted. This clearly shows the efficiency and the performance of the GPR_DA_Bootstrap model on the one hand; as well as the effectiveness of activated carbon prepared from jujube stones (PJAC) as a low-cost adsorbent on the other hand.

Automated summary

In this study, water contaminated by methylene blue (MB) dye was treated with activated carbon based on locally collected jujube stones. Response surface methodology (RSM) and GPR_DA_Bootstrap model were used to optimize the treatment process and obtain the optimal conditions. The results showed that the GPR_DA_Bootstrap model outperformed the BBD model in terms of treatment efficiency.