Optimizing an‎d prediction of fracture an‎d cracking resistance behavior of asphalt mixtures using machine learning algorithms

9/20/2026 12:00:00 AM

The durability of asphalt structures has always been regarded as one of the fundamental issues in the field of engineering, since deterioration an‎d the propagation of cracks are among the most significant factors contributing to the reduction of their service life. Therefore, the development of innovative methods for identifying an‎d predicting the behavior of asphalt mixtures is an unavoidable necessity to improve the quality of roadway infrastructure. In this study, with the aim of providing a comprehensive approach to analyzing these behaviors, a unique experimental dataset consisting of 430 data samples along with 5 features was utilized. These data were processed through 12 machine learning algorithms an‎d artificial neural networks, along with their hyperparameter tuning, in order to detect the relationships between material properties an‎d the indices associated with the fracture an‎d resistance of asphalt mixtures. Furthermore, to determine the optimal conditions, multi-objective optimization approaches such as NSGA-II an‎d manual search were applied for detailed analysis of different scenarios, combined with multi-criteria decision-making TOPSIS method , where the relative weight of each parameter in crack stability was precisely eva‎luated. The findings of the research indicate that the artificial neural network model is the most effective for predicting the output variables. The coefficient of determination, mean absolute error an‎d root mean square error based on cross-validation, respectively, 0.3641, 0.1933, an‎d 0.3877 for fracture energy based on crack mouth opening displacement (CMOD), 0.8744, 0.3417, an‎d 0.5661 for fracture energy based on vertical displacement, 0.9202, 0.0354, an‎d 0.0552 for stress intensity factor mode I an‎d 0.9412, 0.0228, an‎d 0.0422 for stress intensity factor mode II. These results clearly demonstrate the outstan‎ding performance of deep learning methods in capturing the complex fracture behavior of asphalt mixtures.

يادگيري ماشين , مخلوط آسفالت , بهينه سازي , يادگيري عميق , تشخيص

Machine Learning , Asphalt Mixture , Optimization , Deep Learning , Recognition

Ali Akhiani

Dr. Mohammad Reza Mohammad Aliha