Machine learning-driven multi-objective optimization of electrospun nanofibrous membranes design for membrane...

Area of Science:

• Materials Science and Engineering
• Chemical Engineering
• Separation Technologies

Background:

• Electrospinning fabricates tunable membranes for membrane distillation (MD).
• Optimizing electrospun membrane properties requires complex control of solution and processing parameters.
• Conventional trial-and-error methods are inefficient for navigating the vast parameter space.

Purpose of the Study:

• To develop a machine learning (ML) workflow for predicting and optimizing electrospun membrane properties.
• To enable data-guided selection of polymer systems and process conditions for MD applications.
• To establish a general framework for the forward prediction and inverse design of electrospun membranes.

Main Methods:

• Dataset construction, ML model training, and tuning.
• Feature-contribution analysis to understand parameter influences.
• Integration of ML model with Bayesian optimization for inverse design.

Main Results:

• ML workflow accurately predicts key membrane attributes (thickness, contact angle, fiber diameter, morphology).
• Feature analysis quantifies the impact of polymer descriptors and operating parameters.
• Membranes prepared using ML-guided conditions exhibited predicted properties.

Conclusions:

• A holistic ML framework facilitates rapid screening and optimization of electrospun membranes for MD.
• The developed workflow enables efficient inverse design, accelerating material discovery.
• This approach provides a general and extensible solution for designing advanced separation membranes.