Understanding and Designing a High-Performance Ultrafiltration Membrane Using Machine Learning

Abstract

Ultrafiltration (UF) as one of the mainstream membrane-based technologies has been widely used in water and wastewater treatment. Increasing demand for clean and safe water requires the rational design of UF membranes with antifouling potential, while maintaining high water permeability and removal efficiency. This work employed a machine learning (ML) method to establish and understand the correlation of five membrane performance indices as well as three major performance-determining membrane properties with membrane fabrication conditions. The loading of additives, specifically nanomaterials (A_wt %), at loading amounts of >1.0 wt % was found to be the most significant feature affecting all of the membrane performance indices. The polymer content (P_wt %), molecular weight of the pore maker (M_Da), and pore maker content (M_wt %) also made considerable contributions to predicting membrane performance. Notably, M_Da was more important than M_wt % for predicting membrane performance. The feature analysis of ML models in terms of membrane properties (i.e., mean pore size, overall porosity, and contact angle) provided an unequivocal explanation of the effects of fabrication conditions on membrane performance. Our approach can provide practical aid in guiding the design of fit-for-purpose separation membranes through data-driven virtual experiments.

A Meta-Analysis to Revisit the Property–Aggregation Relationships of Carbon Nanomaterials: Experimental Observations versus Predictions of the DLVO Theory

“…Data-driven research has received wide attention in the scientific community and has great potential in improving the design of materials. − This study serves as a proof-of-concept for applying a meta-analytical approach to synthesize pre-existing data to determine patterns and make predictions, thereby obtaining a better understanding of the property–aggregation relationships of nanomaterials, which will be required for the development of accurate predictive models to estimate nanomaterial aggregation. This meta-analysis focused on carbon nanomaterials, and future analysis can include more types of nanomaterials (e.g., metal nanoparticles and nanoplastics) to evaluate the generalizability of the findings in this analysis.…”

“…Other than the aggregation behavior of nanomaterials, their electronic structure and chemical activity are also property-dependent, and leveraging a meta-analytic approach to discover and define such relationships is well-suited . Also, with more data being generated, future studies could utilize more advanced meta-analytic techniques, such as machine learning , to better understand the interplay of multiple properties on the aggregation of nanomaterials and to provide potential practical aid in the design of nanomaterials.…”

Machine Learning in Membrane Design: From Property Prediction to AI-Guided Optimization

“…After a ML model is trained to predict membrane properties in the supervised manner, the SHAP package can be used to quantitatively analyze the correlation between the properties and membrane features. 36 In the work of Yang et al 33 (Figure 3), the SHAP value was used to interpret the multitask gas permeability prediction by a random forest (RF) and deep neural network (DNN) ensemble model. The models were trained to predict the permeability of 6 different gases (He, H 2 , O 2 , N 2 , CO 2 , and CH 4 ) through polymeric membranes.…”

“…The SHapley Additive exPlanations (SHAP) package is a unified framework built on the basis of Shapley value to interpret ML predictions. After a ML model is trained to predict membrane properties in the supervised manner, the SHAP package can be used to quantitatively analyze the correlation between the properties and membrane features . In the work of Yang et al (Figure ), the SHAP value was used to interpret the multitask gas permeability prediction by a random forest (RF) and deep neural network (DNN) ensemble model.…”

Non-Equilibrium Block Copolymer Self-Assembly Based Porous Membrane Formation Processes Employing Multicomponent Systems

“…Sensitivity analysis revealed that simulated permeability was more sensitive to changes in pore diameter than pore density (Figure 7b,c). Further studies, perhaps involving machine learning (ML) [46], could help elucidate how fabrication conditions and other structural parameters (e.g., pore connectivity, macrovoids in the substructure, and separation layer thickness) contribute to the experimentally observed property profiles, bringing us one step closer to realizing inverted designer cycles in the making of SNIPS-based UF membranes.…”