Machine learning-driven constructing relationship between processing and mechanical properties of TPV: A view of one stacking model

Thermoplastic vulcanizate (TPV) exhibit excellent processability and mechanical properties, making them widely applicable in industrial fields. Their mechanical properties, including tensile strength, fracture toughness, and hardness, play a crucial role in determining their suitability for various applications. However, due to the numerous factors influencing the mechanical properties of TPV, the traditional trial-and-error experimental approach not only would consume significant resources but also fail to provide a comprehensive and quantitative understanding of the relationship between the processing parameters and its mechanical performance. To address this challenge, this study has developed one novel stacking model to achieve one high-precision prediction for tensile strength, elongation at break, and Shore hardness for TPV, thereby accelerating the experimental and development process of TPV. Additionally, SHapley Additive exPlanations (SHAP) feature analysis has been employed to interpret the black-box characteristics of the stacking model, revealing how different features influence the overall mechanical performance of TPV. Finally, bivariate dependence plots have been generated to establish the specific mechanisms governing the relationship between TPV processing parameters and mechanical properties. This study would provide an effective approach for guiding TPV experimentation while offering a more efficient method for optimizing TPV performance.