Shedding light on the transfer of tetracycline in forward osmosis through experimental investigation and machine learning modeling.

ABSTRACT

Tetracycline in wastewater can pose adverse impacts on the environment and human health. Forward osmosis (FO) is a promising method to reject antibiotics due to its low energy demand and high rejection rate. Tetracycline rejection during FO is a complicated process. Mechanistic models have been developed to describe antibiotic rejection by the FO membrane under ideal conditions but cannot be applied to real wastewater. Herein, the effects of draw concentration, pH, and solute type on the fate of tetracycline during FO were investigated by combining experimentation, factor analysis, and artificial neural network (ANN) modeling. High draw concentrations led to high convection that favored tetracycline diffusion. Low draw pH helped reject antibiotics potentially due to the decreased tortuosity and pore size of the FO membrane. When different draw solutes were tested, both convection and electrostatic interaction exerted effects on tetracycline retention on the FO membrane surface, and steric hindrance could further affect the amount of tetracycline in the draw solution. Exploratory factor analysis (EFA) showed that tetracycline rejection was a combined result of convection, steric hindrance, and electrostatic interactions. Path analysis revealed the significant roles of initial conductivity and draw pH in tetracycline rejection. Eight representative input variables were selected from 13 observed explanatory variables using redundancy analysis (RDA), based on which an ANN was trained and successfully predicted tetracycline diffusion and transfer through the FO membrane. These results have provided practical and predictive insights in the development of FO processes for efficient treatment of pharmaceutical wastewater.