Towards post-curing parameters optimization of phthalonitrile composites through the synergy of experiment and machine learning

Phthalonitrile composites are highly anticipated in fields such as aerospace, marine, and electronics due to their exceptional heat resistance, excellent high-temperature mechanical properties, and outstanding processability. The conversion of nitrile functional groups to macrocyclic structures during post-curing is essential for performance enhancement. However, the influence of post-curing conditions on the mechanical properties of phthalonitrile composites is complicated, and the underlying mechanism remains unclear. In this study, an experimental and machine learning synergic strategy was employed to optimize the post-curing parameters of the phthalonitrile composites, including temperature, time, and pressure. Through mechanical experiments conducted at both room temperature and 400 °C, scanning electron microscope, and dynamic mechanical analysis, the underlying mechanism of the influence of post-curing parameters on the composite mechanical properties was revealed. The enhancement of the polymerization degree brought about by post-curing is the decisive factor for high-temperature performance, while the room-temperature properties are a tradeoff between the polymerization degree improvement and the defect proliferation. The genetic algorithm-optimized backpropagation (GA-BP) neural network was trained for the efficient and accurate prediction of the optimal post-curing parameters. The tendency of mechanical properties with the post-curing parameters predicted by machine learning is consistent with the mechanism revealed by experiments.