Abstract
This study focused on simulating the adsorption-based separation of Methylene Blue (MB) dye utilising Oryza sativa straw biomass (OSSB). Three distinct modelling approaches were employed: artificial neural networks (ANN), adaptive neuro-fuzzy inference systems (ANFIS), and response surface methodology (RSM). To evaluate the adsorbent's potential, assessments were conducted using Fourier-transform infrared spectroscopy (FTIR) and scanning electron microscopy (SEM). The evaluation of RSM, ANN, and ANFIS included the quantification of R(2), mean squared error (MSE), root mean square error (RMSE), and mean absolute error (MAE) metrics. The regression coefficients from the process modelling demonstrated that RSM (R(2) = 0.9216), ANN (R(2) = 0.8864), and ANFIS (R(2) = 0.9589) all accurately predicted MB adsorptive removal. However, comparative statistical analysis revealed that the ANFIS model exhibited superior accuracy in data-based predictions compared to ANN and RSM models. The ideal pH for MB adsorption utilizing OSSB was established as 7. Additionally, favourable outcomes were obtained with 60-minute contact durations, 20 mg adsorbent quantities, and temperatures of 30 °C. The pseudo 2nd -order kinetic model for MB adsorption by OSSB was confirmed. The equilibrium data exhibited a superior fit with the Langmuir isotherm model in comparison to the Freundlich model. The thermodynamic adsorption parameters, including (∆G = -9.1489 kJ/mol), enthalpy change (∆H = -1457.2 kJ/mol), and entropy change (∆S = -19.03 J mol(-1) K(-1)) indicated that the adsorption of MB onto the OSSB surface is exothermic and spontaneous under the experimental conditions. This research effectively showcased the potential of RSM, ANN, and ANFIS in simulating dye removal using OSSB. The generated parameter data proved valuable for the design and control of the adsorption process.