Application of artificial neural networks in the development of elastomeric composite formulations with recovered carbon black from end–of–life tires

In 2023, global tire production reached 2.3 billion units. After use, end-of-life tires pose a major environmental challenge due to their high resistance to degradation and the pollutant potential of vulcanization reagents. Tire pyrolysis has emerged as a sustainable alternative, generating gases, high-calorific-value oils, and recovered carbon black (rCB), which can be applied as a filler in elastomeric compounds. Although rCB exhibits properties similar to virgin carbon black (CB), variations in the composition of original tires compromise standardization and the mechanical performance of resulting composites. Numerous studies aim to enhance rCB performance using empirical trial-and-error approaches that require significant time, material, skilled labor, and energy. To optimize this process, this study proposes the use of artificial neural networks (ANNs) to predict the rheometric and mechanical properties of rCB-filled elastomeric composites prior to their fabrication, thereby reducing cost and minimizing unnecessary waste generation. Polybutadiene (BR) composites were prepared with rCB contents ranging from 0 to 50 phr. Experimental data were used to train nine ANNs to predict optimal cure time ( t 90 ), hardness, density, abrasion resistance, crosslink density, tensile strength at break, elongation at break, and tear force and displacement at tear. The networks were trained using the Levenberg-Marquardt algorithm with Bayesian regularization. Predictions showed low error margins compared to experimental validation, confirming the accuracy of the models. The use of ANNs proved to be a reliable and efficient tool for the sustainable development of rCB-filled elastomeric composites, improving the vulcanization process and promoting resource optimization in the formulation of rubber materials.