Optimizing forced circulation crystallizer performance in zero discharge desalination using artificial neural networks

Abstract

Brine discharge from desalination processes poses ecological risks due to thermal and salinity stress. Zero discharge desalination (ZDD) technologies like forced circulation crystallization (FCC) recover salts from concentrated brine. This study investigates FCC performance and develops an artificial neural network (ANN) model to optimize operational parameters. Key variables considered include crystallizer capacity, residence time, heat exchanger temperature, cooling water flow rate, crystal size distribution, water recovery, and energy consumption. A feed-forward backpropagation ANN with 14 hidden neurons, trained using the Levenberg–Marquardt algorithm, was implemented. Model predictions were validated against experimental data. The results of the laboratory studies show that the optimal residence time and the best flow rate of incoming cold water are 4 h and 10 kg/min, respectively. Also, 70% can be introduced as the optimal occupied volume. This study shows that the total cost of the conventional crystallizer is about 30% higher than the total cost of the optimized crystallizer.