Investigation of the Relationship between the Mechanical Performance and Contact Structure Characteristics for Asphalt Concrete Using a Digital Image Processing Method and a Machine-Learning Algorithm

The internal structure of asphalt concrete (AC) plays a crucial role in determining its mechanical performance. However, current studies lack accurate assessments of AC’s mechanical performance based on its internal contact structure, and the relationship between the internal contact structure and mechanical performance remains unexplored. This study aims to analyze the distribution of the internal contact structure in asphalt concrete and predict its mechanical performance through contact morphology. A semicircular bending test was performed on 112 asphalt concrete specimens (S1–S112) using a universal testing machine (UTM) to measure the peak force (Pmax). Contact parameters were extracted using digital image processing (DIP) technology, establishing evaluation indices for the contact state. The improved light gradient boosting machine (LightGBM) algorithm was employed to develop a mechanical performance prediction model. The results indicate that variations in the internal contact structure lead to change in AC’s mechanical performance. The distribution of indicators varies, the absolute value of the difference between the average radial orientation and 90° shows significant variation, while the average contact line length exhibits lower variability. The prediction model achieved R2 values of 0.914 and 0.703 for the training and test sets, respectively, effectively predicting AC’s mechanical performance. Among the indicators, the average contact line length index, ACL, has the most significant influence on mechanical performance. These findings provide valuable insights into the relationship between AC’s internal structure and its mechanical performance, offering a technical reference for the on-demand and targeted design.