Modeling of Cu(II) adsorption on the activated Phragmites australis waste by fuzzy-based and neural network-based inference systems

In this study, soft computing models were used to predict Cu(II) adsorption on activated Phragmites australis waste (PAC) and commercial activated carbon (CAC). The effects of pH, adsorbent dose, contact time, initial concentration, and temperature were evaluated in batch mode. Cu(II) adsorption of both adsorbents was better described by the pseudo-second-order kinetic and Langmuir isotherm models. The maximum adsorption capacity was found as 48.31 mg/g and 45.46 mg/g for PAC and CAC, respectively. From thermodynamics, Cu(II) adsorption onto PAC and CAC had an exothermic, randomness, feasible, and spontaneous nature, as physical adsorption. Desirability levels were above 90% in the optimization of the adsorbent parameters that constitute the Mamdani Fuzzy Inference System (MFIS) and Feed-Forward Neural Network (FFNN) inputs. FFNN and MFIS showed superior prediction performance with an error percentage of less than 1% in 2 of 6 experimental designs and were successful with a percentage error of approximately 2–3% in 2 of them. In others, the error percentage of 6–8% was at a level that indicates acceptable and competitive prediction performance. As a result of the hypothesis tests, it was proven that there was no statistically significant difference between PAC and CAC.