Determination of Critical Water Film Thickness for Ultra‐Wide Expressway Pavements to Improve Road Safety

To address the difficulty in accurately evaluating rain‐induced surface water accumulation and hydroplaning risk in superelevation transition sections of ultra‐wide expressways, this study proposes a water film thickness (WFT) prediction method that considers critical WFT and roadway geometric conditions. First, a three‐dimensional finite element model coupling the tire, water film, and pavement was developed. Different tire angular velocities were applied to simulate vehicle operating conditions, and the effects of tread depth, vehicle speed, and WFT on hydrodynamic pressure were analyzed. Based on this analysis, the critical WFT corresponding to different vehicle speeds was determined. Subsequently, a numerical simulation model of pavement water film in superelevation transition sections was established. The effects of rainfall intensity, longitudinal slope, cross slope, pavement width, superelevation transition rate, and mean texture depth on WFT were systematically investigated. Finally, using the numerical simulation results as the dataset, a WFT prediction model based on a genetic algorithm‐optimized back‐propagation (GA‐BP) neural network was developed and validated through indoor artificial rainfall experiments. The results show that the critical WFT corresponding to design speeds of 80, 100, and 120 km/h were 9.74, 6.97, and 4.03 mm, respectively. The influence of the investigated factors on WFT followed the order of longitudinal slope, rainfall intensity, pavement width, superelevation transition rate, cross slope, and mean texture depth. The coefficient of determination of the multiple linear regression model was only 0.38, while that of the back propagation neural network model was 0.82. In contrast, the GA‐BP neural network model achieved a coefficient of determination of 0.98, with a maximum prediction error of less than 2 mm. In addition, the coefficient of determination between the indoor experimental measurements and model predictions reached 0.90. These findings indicate that the proposed GA‐BP WFT prediction model has high accuracy and stability. This approach enables the identification of potential ponding and hydroplaning risks in superelevation transition sections of ultra‐wide expressways and provides a reference for wet‐weather driving safety assessment and roadway drainage design.