Data-Driven Exploration of Polymer Processing Effects on the Mechanical Properties in Carbon Black-Reinforced Rubber Composites

Abstract

The performance and corresponding applications of polymer nanocomposites are highly dominated by the choice of base material, type of fillers, and the processing ways. Carbon black-filled rubber composites (CRC) exemplify this, playing a crucial role in various industries. However, due to the complex interplay between these factors and the resulting properties, a simple yet accurate model to predict the mechanical properties of CRC, considering different rubbers, fillers, and processing techniques, is highly desired. This study aims to predict the dispersion of fillers in CRC and forecast the resultant mechanical properties of CRC by leveraging machine learning. We selected various rubbers and carbon black fillers, conducted mixing and vulcanizing, and subsequently measured filler dispersion and tensile performance. Based on 215 experimental data points, we evaluated the performance of different machine learning models. Our findings indicate that the manually designed deep neural network (DNN) models achieved superior results, exhibiting the highest coefficient of determination (R²) values (>0.95). Shapley additive explanations (SHAP) analysis of the DNN models revealed the intricate relationship between the properties of CRC and process parameters. Moreover, based on the robust predictive capabilities of the DNN models, we can recommend or optimize CRC fabrication process. This work provides valuable insights for employing machine learning in predicting polymer composite material properties and optimizing the fabrication of high-performance CRC.