Genetic algorithm-assisted artificial neural network modelling for remediation and recovery of Pb (II) and Cr(VI) by manganese and cobalt spinel ferrite super nanoadsorbent

Abstract

MnFe₂O₄ and CoFe₂O₄ nanoparticles were hydrothermally synthesized to examine their capability in adsorption of Pb (II) and Cr (VI). The adsorbents exhibited a high rate of adsorption, reaching 90% of their adsorption capacity in less than 30 min. Furthermore, the adsorption capability of the Magnetic Nanoparticles (MNPs) was noticeably greater at initial pollutant concentrations smaller than 40 mg/L. Maximum adsorption capacity on MnFe₂O₄ and CoFe₂O₄ nanoparticles were 40 and 25.38 mg/g for Cr (VI) and 523.32 and 476.19 mg/g for Pb (II), respectively. A data-driven model of Artificial Neural Network was used for prediction of adsorption capacity at both equilibrium and non-equilibrium condition. The model parameters including the numbers of neuron (n = 7) and data portioning for training (49.5%), validation (40.5%), and testing (10%) were obtained using Genetic Algorithm. The results indicated that the model could predict the data with high accuracy (R² = 0.998). The input parameters were initial concentration, time, pH, temperature, adsorbent dosage, and other parameters that is dependent to the physico-chemical properties of ions and adsorbents' surface (ε, α₁, α₂). The mechanism involved in Cr(VI) and Pb(II) adsorption are electrostatic physisorption and a combination of ion exchange chemisorption and electrostatic physisorption, respectively. Desorption capability and adsorbent reuse capability were also examined.

Keywords: Adsorption; Artificial neural network; Heavy metals; Spinel ferrite; Superparamagnetic.