Data‐augmented machine learning for inverse design of homopolymers with targeted glass transition temperature
Authors:
- 7 June 2025
- journal article
- Published by Wiley in Polymer International
- Vol. 75 (2) , 67-74
Abstract
Establishing a reliable workflow of inverse design by data‐driven machine learning (ML) models offers significant potential to accelerate molecular design of polymeric materials. Nevertheless, there exist scarcity issues of training datasets in current data‐driven models for polymers. In this contribution, we integrate the ML method with a data augmentation strategy to build upon a workflow of inverse design of polymeric materials with targeted glass transition temperature T g . Results show that the data‐augmented ML model significantly enhances the prediction accuracy of T g in spite of a small training dataset. Furthermore, the data augmentation strategy has the capability of generating the monomers of homopolymers with higher novelty and uniqueness, whose T g values are validated by the simulations of all‐atomic molecular dynamics. The ML‐assisted inverse design workflow offers significant advantages in establishing structure–property relationships and also provides an accelerated pathway for the targeted design of polymer systems. © 2025 Society of Chemical Industry.Keywords
Funding Information
- National Natural Science Foundation of China
This publication has 38 references indexed in Scilit:
- Discovery of Copolymer Resins with Optimal Viscosity–Toughness–Heat Resistance Trade-Offs via the Material Genome ApproachChemistry of Materials, 2024
- Machine Learning in Soft Matter: From Simulations to ExperimentsAdvanced Functional Materials, 2024
- Molecular dynamics simulations of several linear homopolymers: Assessment and comparison of shock propertiesPolymer, 2023
- Recent Progress and Future Prospects on All-Organic Polymer Dielectrics for Energy Storage CapacitorsChemical Reviews, 2021
- Machine learning discovery of high-temperature polymersPatterns, 2021
- Polymer informatics: Current status and critical next stepsMaterials Science and Engineering: R: Reports, 2021
- Machine learning in polymer informaticsInfoMat, 2021
- Wrapper-based feature selection using regression trees to predict intrinsic viscosity of polymerEngineering with Computers, 2021
- Polymers for Extreme Conditions Designed Using Syntax-Directed Variational AutoencodersChemistry of Materials, 2020
- Evaluation of parallel particle swarm optimization algorithms within the CUDA™ architectureInformation Sciences, 2011