On using artificial neural network models for a thermodynamically-balanced humidification-dehumidification system: Design and rating analysis

Abstract

This research focuses on rating and sizing a thermodynamically-balanced humidification-dehumidification (HDH) system with zero, single, and double extractions of humid air from the humidifier to the dehumidifier. In order to implement thermodynamic balancing appropriately, optimal thermal design is essential to be estimated via a reliable and straightforward model. Thus, this study suggests an artificial neural network to predict the system performance and size parameters. The artificial neural network is trained, tested, and validated using a rich data set created to cover possible operating ranges. The considered performance parameters are gained output ratio, water-to-air mass flow rate ratio, recovery ratio, seawater and air temperatures, energy effectiveness, and critical enthalpy pinch, and the design parameters involve humidifier volume and dehumidifier area. Four inputs are used: seawater inlet temperature, heating temperature, enthalpy pinch, and extractions count. The developed model works best for inlet temperature, heating temperature, enthalpy pinch, and extractions count within 20 – 40 °C, 60 – 80 °C, 1 – critical enthalpy pinch in kJ kgd-1, and 0 – 2 extractions, respectively. The results indicate that the developed artificial neural network has high accuracy in predicting the performance and design parameters, demonstrating minimum accuracy of 98.3%, 96.4%, and 98.9% for zero, single, and double extractions, respectively. Furthermore, the generated neural network is validated against numerical and experimental data from the literature, showing a maximum discrepancy percentage of less than 4%. Thus, the neural network provided by this study is helpful for HDH desalination engineers, designers, researchers, and scientists.