Computational evaluation using machine learning for analysis of membrane desalination process powered by solar energy

Area of Science:

• Chemical Engineering
• Environmental Science
• Computational Modeling

Background:

• Membrane contactors are crucial for water purification and desalination.
• Accurate estimation of permeate volume is vital for optimizing membrane performance.

Purpose of the Study:

• To develop and compare predictive models for permeate volume in membrane desalination.
• To evaluate the efficacy of ensemble methods and advanced hyperparameter tuning.

Main Methods:

• Numerical simulation of mass transfer in membrane contactors.
• Development and comparison of Multilayer Perceptron (MLP), Extreme Gradient Boosting (XGBoost), and Stacking ensemble models.
• Application of Jellyfish Optimizer (JO) for hyperparameter tuning and data preprocessing techniques.

Main Results:

• The Stacking ensemble model demonstrated superior predictive accuracy.
• Achieved high R² values (0.99060 training, 0.97627 cross-validation, 0.99043 testing) and low test error (RMSE = 6.17921).
• Jellyfish Optimizer enhanced model robustness and hyperparameter tuning.

Conclusions:

• Stacking ensemble methods significantly improve prediction accuracy for permeate volume.
• The JO-tuned Stacking model offers a robust and efficient solution for membrane desalination performance prediction.