Machine learning for interfacial evaporation rate prediction and implementation cost optimization

Interfacial evaporation is an emerging technology for sustainable water purification and resource recovery, offering an energy-efficient approach for wastewater minimization. Optimizing interfacial evaporator designs for both high evaporation rates and cost-effectiveness is crucial for practical, large-scale deployment. This work investigates a wind-driven ”sailboat” evaporator configuration with two water passes, analyzing the influence of key design parameters: sail height, sail width, pass spacing, and wind velocity. Models were developed using data from 35 experimental conditions, supplemented by data augmentation techniques. To capture the complex non-linear relationship between the evaporation rate and design parameters, four machine learning algorithms were evaluated: multiple linear regression (MLR), support vector regression (SVR), random forest regression (RFR), and artificial neural network (ANN). The ANN demonstrated superior performance, achieving an R 2 of 0.95 and a mean absolute percentage error (MAPE) of 0.10. Feature importance analysis via the ANN identified the width of the evaporation surface as the most influential parameter for the evaporation rate, a finding validated by experimental results. Leveraging the differentiability of the ANN, the model was integrated with cost functions to perform a rigorous cost-effectiveness optimization. A parametric sensitivity analysis revealed that increasing the sail height is preferable to increasing the width when cost minimization is the dominant objective. This work provides an approach for optimizing evaporator designs for different objectives, which can leverage experimental efforts in deploying interfacial evaporators in real-world applications.