Predicting Indirect Tensile Strength of Rejuvenated Asphalt Mixes Using Machine Learning with High Reclaimed Asphalt Pavement Content

Predicting the performance of hot mix asphalt (HMA) is crucial for ensuring pavement durability, especially as the use of rejuvenated reclaimed asphalt pavement (RAP) increases in sustainable construction. Indirect tensile strength (ITS) is a critical parameter that indicates a pavement’s resistance to cracking and distress under traffic loads. This study developed statistical and machine learning models—linear regression, support vector machine (SVM), and artificial neural network (ANN)—to predict ITS and ITS loss in RAP-incorporated HMA rejuvenated with waste cooking oil (WCO) and waste engine oil (WEO). The models used key input variables, including rejuvenator type and the composition of asphalt, rejuvenator, and RAP. Results showed that WCO increased initial ITS, while WEO enhanced durability by reducing ITS loss. Additionally, lower RAP and asphalt content contributed to improved pavement durability. Among the predictive models, ANN demonstrated the highest accuracy, exhibiting lower error metrics and less variation in scatterplots compared to regression and SVM models. The only exception was ITS loss percentage prediction, where the mean absolute error was nearly identical across all models. These predictive models provide valuable insights for designing and testing modified asphalt mixtures, particularly those containing RAP. By optimizing mix design and enabling proactive maintenance strategies, they contribute to the development of more durable and sustainable pavement infrastructure with the provision of accurate and workable models for prediction of ITS and loss prediction which can be used for design.