IMPLEMENTING MACHINE LEARNING DISTRESS PREDICTION MODELS TO ASSESS THE EFFECT OF BINDER MODIFICATION ON ASPHALT PAVEMENT PERFORMANCE

Asphalt pavement endures a decline owing to the repeated stress from vehicles, shifts in environmental conditions, and binder degradation, which causes distress. This study develops a comprehensive framework that employs machine learning to forecast pavement impairment and sustainable binder modification techniques to enhance both infrastructure durability and environmentally friendly practices. Long Term Pavement Performance (LTPP) data, encompassing FWD deflection indices, traffic load, environmental, and material factors, were employed to train Artificial Neural Network (ANN) models and various classification frameworks, which precisely predicted rut depth (R² > 0.88), fatigue crack area (R² > 0.9), and classified distress types with nearly 90% accuracy. Laboratory examination of PG 64-22, PG 70-22, and PG 76-22 binders augmented with reclaimed facemasks (FM) and sugarcane bagasse fiber (SBF) demonstrated substantial advancements in rutting and fatigue resistance. Integrating binder rheology with predictive modeling established a robust connection between laboratory advancements and their practical applications, providing a unified, data-driven approach for sustainable pavement design and management.