A Machine Learning-Assisted Framework for Performance Prediction and Optimization of Graphene Oxide Nanofiltration...

Area of Science:

• Materials Science
• Chemical Engineering
• Nanotechnology

Background:

• Graphene oxide (GO) membranes show potential for nanofiltration due to their unique 2D structure and surface chemistry.
• Quantifying the relationship between GO membrane structure and separation performance is crucial but challenging.

Purpose of the Study:

• To develop a data-driven approach for predicting GO nanofiltration membrane performance.
• To identify key structural and chemical parameters influencing water flux and ion rejection.
• To optimize GO membrane design for enhanced separation efficiency.

Main Methods:

• Constructed a comprehensive dataset of GO membrane structural and chemical parameters.
• Employed machine learning models (XGBoost) combined with Bayesian optimization.
• Utilized SMILES representation and Morgan fingerprint analysis for chemical feature identification.

Main Results:

• Achieved accurate predictions of water flux and ion rejection rates (XGBoost R²=0.92).
• Identified zeta potential and interlayer spacing as critical factors for separation performance.
• Discovered that amino-functionalized cross-linkers significantly impact membrane performance.
• Identified 14 optimal parameter combinations for balanced performance.

Conclusions:

• A data-driven methodology enables quantitative prediction and optimization of GO nanofiltration membranes.
• Insights gained will guide the rational design of advanced nanofiltration systems.
• This approach offers a pathway to developing highly efficient and selective membranes.