A data-driven framework for assessing horizontal curve safety using friction and reliability analysis: a case study on an Indian highway

Horizontal curves are frequently associated with elevated crash risk because of the combined effects of vehicle speed, roadway geometry, and tire–pavement friction. This study examines these interactions using field data collected from five horizontal curves on NH-340C in India. An empirical model for estimating the maximum available side friction was developed using measured vehicle speeds, superelevation, and pavement surface texture represented by Mean Texture Depth (MTD). Curve safety was evaluated using three complementary indicators: the Safety Index (SI), Change in Safety Index (ΔSI), and Dynamic Curve Safety Index (DCSI), which together describe local stability conditions and safety variations along the curve. The calibrated friction model shows a decrease in available friction with increasing speed and a moderate increase with pavement texture, producing values in the range of 0.16–0.28 that are consistent with international design guidelines. Model reliability was evaluated using uncertainty and sensitivity analysis techniques, including the delta method, bootstrap resampling, the First-Order Reliability Method, and global sensitivity analysis using Sobol and Morris approaches. The results indicate that pavement texture and operating speed are the dominant factors influencing friction variability, while superelevation has a comparatively smaller effect. In addition, a regression-based model was developed to estimate mid-curve operating speeds, which can support the determination of advisory speeds for curve safety management.