Advancing wastewater treatment: removal of methylene blue using sustainable low-cost activated carbon—comparative and efficient prediction by ANN and ANFIS

Artificial neural network (ANN) and adaptive neuro-fuzzy inference system (ANFIS) models were developed to predict the removal efficiency of the cationic dye methylene blue from wastewater using activated carbon derived from Calotropis gigantea leaves (CGAC). The adsorbent was characterized by SEM-EDAX, FT-IR, XRD, BET, and XPS analyses. Batch adsorption studies showed that the equilibrium data were best described by the Freundlich isotherm, followed by Langmuir and Temkin models over the investigated temperature range, while the adsorption kinetics followed a pseudo-second-order model. A dataset comprising 128 experimental runs (100 for training and 28 for testing) was employed for model development using initial dye concentration, contact time, temperature, adsorbent dosage, and pH as input variables. Principal component analysis (PCA) was applied, but did not significantly alter model performance compared to the raw data. Among 14 ANN training functions and three transfer functions, the trainbr–tansig combination provided the highest predictive accuracy. In the ANFIS framework, the Gaussmf membership function with four memberships yielded optimal results. Although ANFIS achieved excellent training accuracy, ANN demonstrated more stable and reliable generalization across both training and testing datasets. Sensitivity analysis identified contact time as the most influential parameter governing dye removal. ANN and ANFIS are confirmed as effective modelling tools for predicting and optimizing dye adsorption by CGAC, with ANN showing superior robustness.

Hali tarjima qilinmagan