Rapid evaluation of drivers’ ride comfort on long-span suspension bridges under VIV using Gaussian process regression

Vortex-induced vibration (VIV) significantly affects ride comfort and may necessitate traffic restrictions, disrupting economic and social activities. The combined impact of VIV and traffic on ride comfort is not well understood, and existing studies are often too time-consuming for timely bridge management decisions. This study aims to explore ride comfort on long-span suspension bridges (LSSBs) during VIV and develop an online prediction model for real-time evaluations, aiding bridge management decisions. A vortex-traffic-bridge (VTB) simulation platform is established to extract vehicle dynamic responses and calculate motion sickness incidence (MSI) for evaluating ride comfort during VIV. MSI is treated probabilistically due to traffic flow's stochastic nature. The optimal probabilistic distribution model (PDM) for MSI data is identified using Jensen-Shannon divergence. A Gaussian process regression (GPR) surrogate model is constructed with VIV mode, VIV amplitude, and traffic density as inputs, and PDM parameters for MSI as outputs. A case study of a prototype LSSB using the GPR surrogate model thoroughly investigates the influence of VIV mode, VIV amplitude, and traffic density on MSI. This model can timely predict drivers' MSI under VIV, aiding effective bridge management decisions.