@Article{sdhm.2026.082304,
AUTHOR = {Hassan Ibrahim Tasiu, Hassan Shuaibu Abdulrahman, Ali Almusawi},
TITLE = {Experimental and Machine Learning-Based Evaluation of the Rheological Properties of Bitumen Modified with Graphene Nanoplatelets},
JOURNAL = {Structural Durability \& Health Monitoring},
VOLUME = {},
YEAR = {},
NUMBER = {},
PAGES = {{pages}},
URL = {http://www.techscience.com/sdhm/online/detail/27229},
ISSN = {1930-2991},
ABSTRACT = {Conventional bitumen used in road construction often exhibits insufficient resistance to rutting, fatigue cracking, and thermal distress, particularly in hot climates. This study investigates graphene nanoplatelets (GNPs) as a bitumen modifier to enhance rheological performance and develops data-driven predictive models for key rheological parameters. Penetration-grade bitumen was modified with GNP contents ranging from 1% to 5% by weight of binder. Conventional tests, including penetration, softening point, ductility, and flash point, alongside rheological characterization using a Dynamic Shear Rheometer (DSR), were performed to evaluate the complex shear modulus (G*) and phase angle (<roman>δ</roman>). Experimental results demonstrated that GNP incorporation significantly enhanced binder stiffness and thermal stability, as reflected by reduced penetration values and elevated softening and flash points. The complex modulus (G*) increased by up to approximately 35%. In comparison, the phase angle (<roman>δ</roman>) decreased by approximately 12% at higher GNP dosages, indicating an improved elastic response and greater resistance to permanent deformation. A progressive reduction in ductility with increasing GNP content was also observed, revealing a trade-off between high-temperature performance and low-temperature flexibility. Performance Grade (PG) analysis confirmed that GNP-modified binders attained higher temperature grades than the unmodified base binder, with optimal overall performance at GNP contents of 2% to 3%. To complement the experimental findings, predictive models were developed using Multiple Linear Regression (MLR) and Artificial Neural Networks (ANN). The MLR model yielded coefficients of determination (R<sup>2</sup>) of approximately 0.87 and 0.82 for G* and δ, respectively. In contrast, the ANN model demonstrated superior predictive capability, achieving R<sup>2</sup> values of approximately 0.95 and 0.93 for the same parameters. The ANN effectively captured the nonlinear relationships among conventional binder properties, GNP content, and rheological response, which linear approaches cannot adequately represent. The findings confirm that moderate GNP dosages substantially improve the high-temperature performance of bitumen and that machine learning models, particularly ANN, offer a reliable and efficient framework for predicting rheological properties. This integrated experimental-computational approach provides a robust basis for data-driven optimization of nanomodified binders, supporting the design of more durable pavement systems in hot-climate environments.},
DOI = {10.32604/sdhm.2026.082304}
}