Computer‐Aided Design of Crosslinked Polymer Membrane Using Machine Learning and Molecular Dynamics

Abstract

The formation of crosslinking network between polymer chains has significant influence on polymer properties. In particular, the crosslinked structure of ionic networks like proton exchange membrane affects the conductivity performance. To further develop in this area, a framework for polymer membrane design based on the developed quantitative prediction model of the properties of crosslinked polymer is proposed. First, polymers with different crosslinking degrees are constructed by a crosslinking algorithm. Next, molecular dynamics is used to calculate the properties of crosslinked polymers. Then, the quantitative relationship between crosslinked polymer structures and macroscopical properties is developed. Subsequently, computer-aided polymer design method is integrated with the developed quantitative predict model. The crosslinked polymer design problem is expressed as an optimization problem to obtain the optimal crosslinking degree. Bayesian optimization strategy is used to solve the established optimization model. Finally, two case studies of perfluoro sulfonic acid and perfluoro imide acid design are given to illustrate the application of the proposed polymer design framework.

Design of Polymeric Membranes for Air Separation by Combining Machine Learning Tools with Computer Aided Molecular Design

“…By placing limitations on safety and health risks during CAMD, they were able to generate molecules that were less toxic while still exhibiting outstanding product performance. CAMD approaches have also been used in the design of polymeric membranes based on group contribution methods [24] and molecular dynamics [25]. In these approaches, a set of desirable properties of the polymer has been targeted and the polymeric structures that meet those properties have been generated.…”

Prediction of Electrical Conductivity of Ionic Liquids: From COSMO-RS Derived QSPR Evaluation to Boosting Machine Learning

“…One approach to create more reliable and robust ML models is to integrate physically derived model and ML through ensemble methods, which could combine the flexibility and sophisticated training strategies inherent in ML with the adherence to certain physical laws and their associated predictive capability. , In this regard, successful examples have already been reported on the prediction of infinite dilution activity coefficients ( γ ∞ ) and self-diffusion coefficients of fluid mixtures by applying a typical ensemble method known as boosting. , Such boosting ML models were trained on the residuals of a specific physically derived model, and then their predictions were added to the physically derived predictions to obtain the final prediction, thereby correcting the error and providing higher model accuracy. For example, Medina et al trained graph neural network models on the residuals of several physical models (e.g., Hansen Solubility Parameters, COSMO-RS and UNIFAC) for predicting γ ij ∞ over 1000 binary systems, demonstrating that such hybrid models can overall improve the performance of the individual model instances.…”