Insights into predictive modeling, isotherms and kinetic studies in the removal of methylene blue from water using pineapple Peel activated carbon

Abstract Mathematical and empirical models have been used to study the adsorption process of methylene blue (MB) using pineapple peel activated carbon, through optimization and prediction. The models applied in this study include Adaptive neuro-fuzzy inference system (ANFIS), Artificial Neural Networks (ANNs), and Response surface methodology (RSM). The adsorption of MB was optimized and predicted by varying the dose, pH, and time using ANFIS, ANN, and RSM. The optimal conditions were obtained at a pH (5.05), time (5.69 min), and dose (95.13 mg) which gave a removal of 87.67%. The models demonstrated high predictive accuracy, as confirmed by the MSE, RMSE, SAE, SSE, Adj R 2 , and R 2 . The ANFIS model was the best followed by RSM and ANN had the least predictive data. In addition, non-linear regression analysis on MB adsorption was carried out using both isotherm and kinetic models. Adsorption isotherm models, including Langmuir, Freundlich, Sips, Toth, Dubinin-Radushkevich, Jossens, Khan, and Temkin, revealed high correlation coefficients (R 2 > 0.94). However, the mode is best explained by the Freundlich model as such multilayer heterogeneous adsorption occurs on the surface. The Langmuir mode with a q max of 15.72 mg/g also shows the strong adsorption capacity of the adsorbent. Kinetic studies involving Pseudo-First Order (PFO), Pseudo-Second Order (PSO), Elovich, Intraparticle diffusion, and Fractional Power models showed high correlation coefficients (R 2 > 0.98). The PSO model with a relatively high R2 () best describes the adsorption process as such the mechanism was by chemisorption. It is worth noting that activated carbon was characterized using Fourier Transform Infrared Spectroscopy (FTIR), Scanning Electron Microscopy (SEM), and powder X-ray Diffraction (XRD). Lastly, this research successfully demonstrates a sustainable approach for the effective use of ANFIS, ANNs, and RSM in optimizing and predicting adsorption processes.