Surrogate-based probabilistic modeling of pavement deterioration under traffic-induced stress

This study proposes a surrogate-based probabilistic modeling framework to assess pavement deterioration under traffic-induced stress using survival analysis. Surrogates such as the Pavement Condition Index (PCI) were used to represent cumulative distress progression, enabling deterioration prediction in the absence of direct loading measurements. A semi-parametric Cox Proportional Hazards (CPH) model was employed to estimate failure risk over time without requiring a predefined baseline hazard, offering both flexibility and interpretability. Principal Component Analysis (PCA) was applied to address multicollinearity among candidate variables (PCI, RUT, and All Cracks), with the first principal component explaining 76.91% of total variance and PCI demonstrating the strongest loading. Based on this, PCI was retained as the primary covariate. Due to missing event-time data, only 84 out of 1668 samples (5.0%) contained valid survival records. Despite the limited analytical size, the model serves as a proof of concept under real-world constraints. Kaplan–Meier survival curves stratified by PCI categories demonstrated significant differences in deterioration trends, with median survival dropping from 8.3 years (Good) to 4.5 years (Poor). A log-rank test confirmed group differences ( X 2 = 14.02, p < 0.001 ). The final model indicated a statistically significant hazard ratio for PCI (HR = 1.164, p = 0.004 ), where each unit reduction in PCI increased failure risk by 16.4%. This surrogate-based approach offers a practical pathway for proactive asset management, particularly under constrained data environments.