Statistical modelling of endocrine disrupting compounds adsorption onto activated carbon prepared from wood using CCD-RSM and DE hybrid evolutionary optimization framework: Comparison of linear vs non-linear isotherm and kinetic parameters

Abstract In this research, the efficiency of two adsorbents, including powdered and granular activated carbon (obtained from wood) was investigated on BPA removal in a batch-mode reactor. ANOVA analysis based on the central composite design-response surface methodology (CCD-RSM) showed a good fit between quadratic model predictions with experimental values, thus resulting in R2 of 0.9992 and 0.9997 for PAC and GAC respectively. The proposed 3 layered backpropagation artificial neural network (ANN) model predictions results with R2 = 0.9839 and 0.9992 for PAC and GAC respectively. The CCD-RSM optimised results indicated a maximum removal efficiency of 99% BPA in the case of the PAC under the optimal conditions, whereas, it is 89% for GAC. Genetic algorithm (GA) is also implemented to find the optimal values that can result high removal efficiency. The set (pH, contact time, adsorbent dosage and initial BPA concentration) of GA based optimised values for both PAC and GAC are [7.18, 90 min, 18 mg/L, 1.6 mg/L] and [7.76, 90 min, 18 mg/L, 1.67 mg/L] respectively which results in 99% and 89.95% removal efficiency. In this study, the Qmax for powdered and granular activated carbon was found to be 93.89 and 74.62 mg/g, respectively. The adsorption process is following the Langmuir isotherm and Pseudo 2nd order kinetic models. The thermodynamic study also signifies a favourable and spontaneous removal process. Overall the results confirm that the low-cost powder activated carbon favours high removal efficiency of BPA from aqueous environment.