Regression model, artificial intelligence, and cost estimation for phosphate adsorption using encapsulated nanoscale zero-valent iron

• Published in: Separation science and technology Vol. 54; no. 1; pp. 13 - 26
• Main Authors: Mahmoud, Ahmed S., Mostafa, Mohamed K., Nasr, Mahmoud
• Format: Journal Article
• Language: English
• Published: Abingdon Taylor & Francis 02.01.2019; Taylor & Francis Ltd
• Subjects: Adsorption; Artificial intelligence; Artificial neural network; Artificial neural networks; Cost analysis; Dosage; Encapsulation; Iron; Mathematical models; Neural networks; nZVI/Ca-beads; pH effects; phosphate adsorption; Phosphates; Regression analysis; Regression models; Removal; response surface methodology; Sensitivity analysis
• ISSN: 0149-6395; 1520-5754
• DOI: 10.1080/01496395.2018.1504799

This study investigated the adsorption of PO 4 3− onto encapsulated nanoscale zero-valent iron (nZVI). At initial PO 4 3- : 10 mg · L −1 , the optimum condition was initial pH: 6.5, nZVI dosage: 20 g · L −1 , stirring-rate: 100 rpm, and time: 30 min, achieving PO 4 3− removal of 42%. The effect of pH and time on the PO 4 3− removal efficiency was quadratic-linear concave up, whereas the curve of nZVI dosage was quadratic-convex. Artificial neural network with a structure of 5−7−1 adequately predicted PO 4 3− removal (R 2 : 97.6%), and the sensitivity analysis demonstrated that pH was the most influential input. The cost of the adsorption unit was 3.15 $USD · m −3 .