REVERSE OSMOSIS PILOT PLANT EFFICIENCY ASSESSMENT USING ARTIFICIAL NEURAL NETWORK (ANN)

Reverse osmosis (RO) is widely used for desalination and water treatment, but its efficiency depends on several operational factors. This study evaluates the performance of a commercial RO unit in treating concentrate water and applies an artificial neural network (ANN) model to predict permeate total dissolved solids (TDS) and RO removal efficiency. Data was collected over 60 days monitoring period, water quality parameters including pH, total dissolved solids (TDS), electrical conductivity (EC), and temperature were measured for feed, permeate, and concentrate water of the RO unit. The results indicated that the RO unit has shown good performance in removing salt from concentrate water and achieved an average removal efficiency of 95.7% with permeate water meeting Iraqi and WHO drinking water standards. A slight decrease in efficiency was observed with rising feedwater temperature, with the highest performance near (96.9%) recorded near 21.8 °C. The ANN model showed moderate predictive accuracy for permeate TDS (R² = 0.870) with TDS variability explaining the difference in model performance. In addition, the ANN model presented a strong predictive performance for removal efficiency (R² = 0.902). Compared with previous studies, this work demonstrates the feasibility of using ANN to model and predict RO performance with small experimental datasets. The results provide practical insights for employing RO units for the reuse of RO concentrate through a secondary RO unit, which could minimize water wastage and mitigate environmental impacts