Train-bridge interaction under correlated wind and rain using machine learning

The stochastic response of the high-speed train running along the bridge is susceptible to significant effects of wind and rain simultaneously. In this study, an efficient probability method for simulating the train-bridge interaction (TBI) was developed by combining wind-rain and train-bridge interaction (WRTBI) models with machine learning methods. The high-speed train was defined as multibody system (MBS) and the bridge was described over finite element method (FEM). The wind load was modeled based on the standard turbulent model and the rain load was defined according to Euler multi-phase model. The WRTBI was solved using a co-simulation method between finite element analysis and multibody system (FEA-MBS). A machine learning consisting of support vector machine (SVM) and Latin hypercube sampling (LHS) was introduced to substitute further FEA-MBS simulations, to overcome time-consuming issue. The results show that when the train runs along the bridge at the speed range of 260 km/h to 300 km/h, the maximum error is 12% in the case where the wind and rain loads are considered with the case where only the wind load is considered. The maximum displacement of the bridge in z direction increase to around 33.33%. from 25 to 100 years return period.