Random Dynamic Analysis of Wind-Vehicle-Bridge System Based on ARMAX Surrogate Model and High-Order Differencing

Abstract

To investigate the stochastic characteristics of vehicle-bridge (VB) system under crosswind, an efficient method which combines AutoRegressive Moving Average with eXogenous inputs (ARMAX) model, high-order differencing (HOD) and important sample was proposed in this paper. First, the wind turbulence spectra relative to a moving vehicle and equivalent static gust load method were adopted to simplify the turbulent wind field of VB system, and a wind-vehicle-bridge (WVB) model was established and verified. Then, an analysis framework for WVB system based on ARMAX model was proposed, and HOD method and important sample were used to improve the prediction performance of the surrogate model. Prediction accuracy and calculation efficiency of proposed AMRAX model were verified and compared by Monte Carlo simulation (MCS). Finally, the impacts of vehicle speed and wind velocity on the stochastic characteristics of train response were discussed. Results indicate that the HOD method has significantly improved the prediction performance of ARMAX model for lateral response of trains, and the train responses predicted by ARMAX model based on HOD and important sample show perfect agreement with target results. Compared with MCS, the calculation efficiencies of proposed ARMAX model are improved by about two orders of magnitude. The extreme values of the train response with different vehicle speed and wind velocity gradually obey right skewness distribution, especially the lateral acceleration.