A predictive model for the relationship between processing conditions and properties of thermoplastic vulcanizates (TPVs) via machine learning

Abstract

Thermoplastic vulcanizates (TPVs) could be extensively used in industrial fields due to their excellent processing performance and mechanical properties. Polymer processing, including the experimental formula and process technology, is the most direct and crucial factor that determines the performance of TPV. However, it is difficult to fully understand the relationship between polymer processing and property of TPV via the experimental means. Herein, a BP neural network model optimized by genetic algorithm has been designed via machine learning method. The R value of the prediction model could reach to 0.95, which could accurately predict the mechanical properties of different types of TPV. The impact of the important processing factors on the mechanical properties of TPV have been analyzed quantitatively via the bivariate partial dependence plots of TPV. This work could effectively guide the processing of TPV and optimize the properties of TPV.