Despite advances in machine learning for accurately predicting material properties, forecasting the performance of thermosetting polymers remains a challenge due to the sparsity of historical experimental data and their complicated crosslinked structures. We proposed a machine-learning-assisted materials genome approach (MGA) for rapidly designing novel epoxy thermosets with excellent mechanical properties (high tensile moduli, high tensile strength, and high toughness) through high-throughput screening in a vast chemical space. Machine-learning models were established by combining attention- and gate-augmented graph convolutional networks, multilayer perceptrons, classical gel theory, and transfer learning from small molecules to polymers. Proof-of-concept experiments were carried out, and the structures designed by the MGA were verified. Gene substructures affecting the modulus, strength, and toughness were also extracted, revealing the mechanisms of polymers with high mechanical properties. The developed strategy can be employed to design other thermosetting polymers efficiently.
Keywords: machine learning; materials genome approach; modulus; tensile strength; thermosetting polymers; toughness.