Abstract Manager

Vehicle forces from braking and steering in horizontal curves results in a differential rate of pavement friction degradation compared to tangent road segments. This presents a challenge for safety-conscious road agencies seeking to reduce friction-related crashes and develop a pavement friction management program or maintenance decision framework. This report utilizes five years and thousands of centerline miles of continuous pavement friction measurement (CPFM) data collected across a US State road network conflated with available network and surface characteristics to estimate the remaining useful life (RUL) of pavement for a selection of curves and associated tangents. Preliminary results evaluate the feasibility of using AI-driven survival analysis on high-resolution pavement friction data to better understand friction degradation behavior under the high friction demand conditions present in horizontal curves.

Survival analysis is employed to model the time until friction falls below specified thresholds, representing the remaining friction service life. Machine learning–based survival models capture nonlinear interactions while maintaining interpretability for engineering decision-making. Roadway geometry, network, and surface characteristic variables are treated as predictors to estimate time-to-intervention distributions. Hazard ratios quantify the relative risk of friction degradation across different curve severity classes and pavement types, while variable sensitivity analysis identifies which factors most significantly influence degradation rates in curves versus tangent sections.

The insights from this analysis support the development or enhancement of a decision support framework to enable road agencies to transition from reactive friction management to predictive, risk-based maintenance scheduling. By identifying early warning indicators of accelerated friction degradation in high-risk curves, agencies can apply targeted surface treatment options, optimize maintenance timing, and proactively manage resource allocation. The methodology demonstrates how longitudinal continuous pavement friction monitoring combined with artificial intelligence can extract actionable insights in support of more cost-effective pavement management strategies for improved safety outcomes.