Parametric optimization of single-effect distillation via artificial neural network for desalination process

With the growing global demand for clean water, efficient and sustainable desalination technologies are becoming increasingly critical. This study explores the optimization of a single-effect distiller (SED) for desalination using artificial neural network modeling, specifically a multi-layer perceptron (MLP), benchmarked against support vector regression (SVR) and linear and polynomial regression models. The models predict the distillate mass flow rate under varying operating conditions, including hot-water temperature, cold-water temperature, and vacuum pressure. Among the models, the MLP demonstrates superior performance, achieving an R 2 of 0.9671 and an RMSE of 0.3867 g/min, compared to those of SVR with R 2 of 0.9658 and RMSE of 0.3884 g/min. Contour plots generated by the MLP model provided actionable insights, identifying optimal operating conditions for maximizing the distillate mass flow rate. Specifically, the optimal performance is achieved with high hot-water temperatures (above 85 °C), low cold-water temperatures (approximately 22 °C), and vacuum pressures below 10 kPa. These conditions are practical, as room-temperature seawater can effectively cool the condenser without additional energy for cooling, while solar energy can heat the hot water. A conceptual design for a solar-powered SED that integrates these findings, offering a sustainable and energy-efficient desalination solution is proposed. The machine learning-driven optimization framework presented here provides a valuable pathway for addressing global water scarcity challenges. • Machine learning models optimize single-effect desalination performance effectively. • MLP achieves superior predictive accuracy with R 2 = 0.9671, RMSE = 0.3867 g/min. • Contour plots reveal critical insights into the influence of operational parameters. • Cold-water temperature is critical for maximizing desalination efficiency.