Abstract:Thermoplastic vulcanizates (TPVs) have significant advantages in replacing traditional rubber due to their low energy consumption and recyclability, but their application in dynamic sealing is still limited by their resilience performance. This study proposed a method for investigating the resilience performance of TPV. A 2D random multi-particle model representing the microstructure of TPV was established through Abaqus and Python. Simulation model parameters and results were extracted through programming, and an automated simulation process on the Isight platform generated a material simulation database, effectively addressing the lack of machine learning datasets in the TPV field. The resilience performance of TPVs was predicted using K-fold cross-validation and random forest algorithms, and the model accuracy was verified. Feature importance analysis and bivariate partial correlation plots reveal that the matrix plastic strain and rubber-plastic ratio have the greatest impact on the resilience performance of TPV, while the influence of the structural parameters of rubber particles is limited.