A fast hybrid algorithm for the random vibration analysis of train-bridge systems under crosswinds

Abstract

In this study, a fast hybrid algorithm based on surrogate and theoretical models is proposed. It solves surrogate model failure with too many inputs to the system and improves the computational efficiency in random vibration analysis of train-bridge systems under crosswinds. First, the surrogate model to rapidly predict the wheel-rail force time history and the theoretical model (finite element model) of the bridge are established. Then, a large number of samples of fluctuating wind speeds and track irregularities are generated based on the Monte Carlo method. Next, the wheel-rail force and the dynamic response of the bridge are calculated using the surrogate and theoretical models, respectively, and the coupling of the train and bridge subsystems is realized through iteration. Finally, the random vibrations of the train-bridge system under crosswinds are analyzed based on the calculation results of all samples. The results of the hybrid algorithm agree well with those of the traditional calculation method. The maximum normalized mean square error (NMSE) is only 0.0105, and the computational efficiency of the hybrid algorithm is nearly 4 times higher than that of traditional calculation methods. When the train is in the midspan of the bridge, the bridge dynamic response has a large mean and standard deviation. Under crosswinds, wheels on the windward side determine the safe running of the entire train. Under the wind speed of 15 m/s and Chinese high-speed railway track irregularities, the train operation reliability on the bridge is 98 %.