Assessing the impact of reverse osmosis plant operations on water quality index improvement through machine learning approaches and health risk assessment

Reverse osmosis (RO) is used to improve drinking water quality, but knowledge of outlet quality and system performance is required. This study evaluated RO performance in treating groundwater in southern Iran, focusing on the water quality index (WQI) and health risks. The parameters in the inlet flow that were above the standard level in some samples include total hardness (TH), sulfate (SO4), chloride (Cl), total dissolved solids (TDS), electrical conductivity (EC), and Turbidity. But in the outlet Flow, the parameters exceeding the standard level were TH, SO4, Cl, and TDS. The highest mean removal efficiencies in the RO system were for free chlorine residual (FCR) (98.43 %) and SO4 (82.89 %). The WQI of the inlet, classified as good in 97.67 % of the total samples, improved to excellent in 95.35 % of the samples at the outlet after treatment by the RO system. Machine learning results revealed that the random forest (RF) model was the most accurate in predicting the WQI, with TDS and EC as the key influencing factors. The non-carcinogenic risk from fluoride (F) and nitrate (NO3) in children group exceeded the permissible limit in approximately 4.6 % and 6.9 % of inlet water samples, respectively. The 95th percentile hazard index (HI) for children was 2.32 for inlet water and 1.10 for outlet water, while for adults, it was 1.08 and 0.52, respectively. The F level and ingestion rate (IR) were the most effective parameters on HI. These findings highlight the need for RO-purified water and emphasize regular monitoring of treatment plants.