Wind–vehicle–bridge interaction and driving safety reliability under skew winds

Wind-induced vehicle accidents on bridges depend not only on mean wind speed but also on turbulence and pavement roughness. Existing wind–vehicle–bridge interaction (W-VBI) studies typically rely on a limited number of stochastic wind and pavement realisations, restricting the quantification of record-to-record variability in driving safety. This study proposes a W-VBI framework that explicitly embeds this variability into the driving safety assessment. Applied to a twin box-girder bridge, the methodology combines stochastic simulations of turbulent wind and pavement profiles with aerodynamic coefficients obtained from wind-tunnel tests, which are used to define wind loads in the coupled W-VBI analysis. Over 85,000 nonlinear simulations are conducted for 11 wind angles, five gust levels, two high-sided vehicles, and speeds from 15–130 km/h. Vehicle overturning, side-slipping, and yawing probabilities are calculated, using bootstrapping techniques to evaluate the statistical stability of the simulations. Results show that wind and pavement randomness induce significant record-to-record variability in vehicle safety, while bootstrapping enables estimation of the minimum number of stochastic records required for statistically stable reliability assessments.