Remarkable dynamic adsorption of Hg2+ ions on the magnetic MOF nanocomposite (CuNi-BTC@Fe3O4): Experimental and modeling using GA+CFD+ANFIS method

The magnetic bimetallic metal-organic framework (MOF) nanocomposite was fabricated using the facile preparation method for the removal of mercury ions in the fixed bed column. The several analyses including XRD, TEM, BET and VSM were also carried out to characterize the physico-chemical properties of the synthesized sample ([email protected]3O4). The Langmuir model with a maximum adsorption capacity of 144.32 mg/g was fitted well with equilibrium data. The performance of the adsorption column was evaluated under the various operating conditions including length of bed (10, 15, 20 cm), flow rate of feed (10, 15, 20 ml/min) and initial concentration of mercury ions in the inlet (25, 50, 75 mg/l). The breakthrough time was measured to be about 581.1 min at the proper operating conditions including 15 cm of bed height, 50 mg/l of feed concentration and 10 ml/min of feed flow rate. The commercial models including Thomas and Yoon-Nelson were investigated for the determination of the adsorption behavior of Hg (II) in the adsorption column. The results indicated the good agreement between the experimental data and Yoon-Nelson model (R2 > 0.99). The combination model consisting of the unsteady computational fluid dynamics (CFD), an adaptive neuro-fuzzy inference (ANFIS) and genetic algorithm (GA) was developed to simulate the uptake of Hg (II) ions in dynamic bed. The employed ANFIS was also coded by using five layers to predict the source term of mass transfer equation as the most significant parameter of the modeling which must be calculated with high precision. The generated Gaussian membership functions i n the ANFIS model was optimized by applying GA. Finally, the CFD (the first order implicit) was linked with ANFIS to describe the dynasmic adsorption of mercury ions. The validation illustrated a good matching between the experimental data and the output of the combination model (R2 > 0.999).