Simulation the adsorption capacity of polyvinyl alcohol/carboxymethyl cellulose based hydrogels towards methylene blue in aqueous solutions using cascade correlation neural network (CCNN) technique

Abstract

Almost all industries produce a large quantity of dye-contaminated wastewater. Wastewaters containing a high dosage of methylene blue (MB) are a menace for human beings, the environment, and the ecosystem. Thus, the MB molecules are needed to be removed before wastewater discharge to the environment. Adsorption is the most well-known process for dye removal from water and wastewater. This study focuses on the intelligent simulation of the MB removal by the bio-based hydrogel. Indeed, the cascade correlation neural network (CCNN) employs to simulate the adsorption mechanism of MB molecules by the bio-based (polyvinyl alcohol/carboxymethyl cellulose) hydrogel reinforced by graphene oxide nanoparticles and bentonite. The Levenberg-Marquardt algorithm trains the CCNN to estimate the hydrogel capacity for MB uptake as a function of adsorbent type, temperature, initial dye concentration, pH, and contact time. Trial-and-error analyses justified that the CCNN with one hidden layer containing six neurons is the most reliable model for the given problem. This model predicts the collected experimental data from the literature with excellent agreement (i.e., RMSE = 2.00, AARD = 2.4%, and R2 = 0.9980). Experimental measurements and modeling findings approved that MB uptake at 30–40 °C intensifies by increasing temperature, contact time, pH, and initial dye concentration. Increasing the mobility of the large dye ions at 50 °C reduces the adsorption capacity of all bio-adsorbents. Furthermore, the reinforced bio-based hydrogel with bentonite/GO nanoparticles is the best adsorbent for the methylene blue uptake.