Designing Multicomponent\nThermosetting Resins through\nMachine Learning and High-Throughput Screening

Designing cyanate copolymers with improved comprehensive\nproperties,\nsuch as simultaneously with high thermal resistance, moisture resistance,\nand improved modulus, by copolymerizing cyanate ester monomers is\nchallenging because of the intrinsic restrictions among these properties.\nThe generally employed trial-and-error method is labor-intensive and\nthus insufficient to meet the demands of the efficient development\nof new materials. Herein, we proposed a property-oriented high-throughput\nscreening strategy, integrating state-of-the-art machine learning\ntechniques with automated feature engineering, to expedite the discovery\nof cyanate copolymers with enhanced comprehensive properties. Several\nscreened samples with varied ranks of comprehensive properties were\npicked up for proof-of-concept experiments to validate the predicted\nhygroscopicity, glass transition temperature, and tensile modulus.\nFurthermore, the importance of the feature to properties was evaluated\nto reveal the factors essential for improving moisture resistance,\nthermal stability, and mechanical properties. The developed strategy,\nefficiently bypassing cumbersome dimensionality reduction and minimizing\nmanual interference via automated feature engineering, offers versatile\napplicability in designing a broad range of multicomponent copolymerized\nand blend resins.