Functionalized chitosan-magnetic flocculants for heavy metal and dye removal modeled by an artificial neural network.

[Display omitted] • A high-efficiency, multifunctional, salt/pH-tolerant magnetic flocculant was successfully prepared. • The growth characteristics of flocs are systematically elucidated. • There is a positive correlation between the fractal dimension of flocs and flocculation efficiency. • Ca(II) and Fe(III) will promote the removal of DB56 and Cu(II), respectively. • The neural network models can accurately predict the removal rate of DB56 and Cu(II). In this study, an amphoteric magnetic chitosan (CS)-based flocculant MFe 3 O 4 @CS-g-PIA was prepared from CS, Fe 3 O 4 , and itaconic acid (IA), and its apparent morphology and characteristic structure were systematically studied. The flocculation performance and mechanism of the fabricated material were also investigated in different pollution systems, and the effects of total monomer concentration, m(CS):m(IA), IA pre-neutralization degree, reaction temperature, reaction time, and initiator concentration on the synthesis of MFe 3 O 4 @CS-g-PIA were studied. Characterization results showed that MFe 3 O 4 @CS-g-PIA forms a three-dimensional network with excellent magnetic induction. The optimal removal rates of Cu(II) and Disperse Blue 56 (DB56; 90.2% and 97.0%, respectively) were obtained under the conditions of 150 mg·L−1 MFe 3 O 4 @CS-g-PIA, pH 6.0, and 300 rpm stirring speed. MFe 3 O 4 @CS-g-PIA maintained removal rates of over 80.0% for Cu(II) and DB56 after five consecutive cycles of regeneration/flocculation and demonstrated excellent acid resistance stability. Changes in the particle size distribution, fractal dimensions, and zeta potentials of the flocs indicated that the relevant flocculation mechanism involves the synergistic functions of chelation, charge neutralization, and adsorption bridging. An artificial neural network model was finally established on the basis of the experimental flocculation data to predict the removal rates of Cu(II) (R = 0.97) and DB56 (R = 0.98) accurately. [ABSTRACT FROM AUTHOR]