Biosorptive removal of Zn(II) ions by Pongamia oil cake (Pongamia pinnata) in batch and fixed-bed column studies using response surface methodology and artificial neural network

• Published in: Journal of environmental management Vol. 227; pp. 216 - 228
• Main Authors: Shanmugaprakash, Muthusamy, Venkatachalam, Sivakumar, Rajendran, Karthik, Pugazhendhi, Arivalagan
• Format: Journal Article
• Language: English
• Published: England Elsevier Ltd 01.12.2018
• Subjects: Artificial neural networks; Biosorption; Isotherm; Kinetics; Pongamia oil cake; Response surface methodology; Zn(II) ions; Biosorption; Artificial neural networks; Pongamia oil cake; Zn(II) ions; Isotherm; Response surface methodology; Kinetics
• ISSN: 0301-4797; 1095-8630
• DOI: 10.1016/j.jenvman.2018.08.088

Design of experiment and artificial neural networks (ANN) have been effectively employed to predict the rate of uptake of Zn(II) ions onto defatted pongamia oil cake. Four independent variables such as, pH (2.0–7.0), initial concentration of Zn(II) ions (50–500 mg/L), temperature (30ºC-50 °C), and dosage of biosorbent (1.0–5.0 g/L) were used for the batch mode while the three independent variables viz. flowrate, initial concentration of Zn(II) ions and bed height were employed for the continuous mode. Second-order polynomial equations were then derived to predict the Zn(II) ion uptake rate. The optimum conditions for batch studies was found to be pH: 4.45, metal ion concentration: 462.48 mg/L, dosage: 2.88 g/L, temperature: 303 K and on the other hand the column studies flow rate: 5.59 mL/min, metal ion concentration: 499.3 mg/L and bed height: 14.82 cm. Under these optimal condition, the adsorption capacity was 80.66 mg/g and 66.29 mg/g for batch and column studies, respectively. The same data was fed to train a feed-forward multilayered perceptron, using MATLAB to develop the ANN based model. The predictive capabilities of the two methodologies were compared, by means of the absolute average deviation (AAD) (4.57%), model predictive error (MPE) (4.15%), root mean square error (RMSE) (3.19), standard error of prediction (SEP) (4.23) and correlation coefficient (R) (0.99) for ANN and for RSM AAD (16.27%), MPE (21,25%), RMSE (13.15%), SEP and R (0.96) by validation data. The findings suggested that compared to the prediction ability of RSM model, the properly trained ANN model has better prediction ability. In batch studies, equilibrium data was used to determine the isotherm constants and first and second order rate constants. In column, bed depth service time (BDST) and Thomas model was used to fit the obtained column data. [Display omitted] •Predictive modeling of biosorption of Zn(II) ions on DPOC was studied.•ANN had better prediction capability than RSM for Zn(II) uptake rate on DPOC.•pH and temperature are significant factors for removal of Zn(II) in batch mode.•Metal ions concentration and bed height are significant factors in continuous mode.