(33f) Optimal Machine Learning Approach for Anion Conductivity Prediction in Anion Exchange Membranes (AEMs)

Electrically driven separations have transformative applications across water desalination, fuel cells, and pharmaceutical purification processes. Among these systems, anion exchange membranes (AEMs) play a critical role due to their selective transport properties. AEM conductivity, a pivotal feature influencing membrane efficiency and performance, remains challenging to optimize due to limited experimental datasets, which restrict the application of data-driven approaches such as machine learning (ML).

In this study, we address the challenge of scarce data by employing advanced data augmentation techniques to synthetically expand our dataset. Leveraging a combination of ML algorithms and synthetic data generation, we significantly increased the volume and diversity of training data while maintaining relevant molecular and material characteristics. This approach enabled us to train more robust models, resulting in substantial improvements in predictive accuracy and reduced error rates on validation with unseen datasets.

Our results underscore the potential of ML-driven data augmentation in enhancing the predictive power of models in the realm of membrane technology, particularly for applications where experimental data collection is resource-intensive or limited. This work not only demonstrates the feasibility of augmenting AEM datasets but also sets the stage for broader applications of ML in membrane science, with implications for accelerating the development of high-performance membranes across various industries.