A pilot platform was set up to operate three 8-inch nanofiltration (NF) and reverse osmosis (RO) modules in parallel. Five membranes were evaluated using a set of 25 emerging organic contaminants spiked at 100 times the detection levels to allow the determination of removal efficiencies. The list of compounds was defined based on occurrence in drinking water, toxicological relevance and chemical properties to ensure chemical diversity in the data set. Three samples were collected at three monitoring points (feed, permeate, concentrate) to control the consistency of the analyses. The NF/RO modules were operated at high recovery rates to evaluate performance in conservative conditions. Results were compared to predicted rejection levels as determined by the qualitative approach proposed by Verliefde et al. that classifies organic contaminants into 8 categories based on hydrophobicity, molecular weight and charge [Verliefde et al. (2007)]. Differences in rejection were observed for a number of compounds, particularly for membranes that remove a lesser extent of mineral content. These "looser" membranes are of particular interest for drinking water production due to the lesser need for post-treatment to ensure corrosion control in the distribution system and compliance with regulations such as the Lead and Copper Rule. Overall, the prediction model appears to be conservative in anticipating the performance of the membranes that prove to be more effective than expected in many cases.
展开▼
