Extreme response analysis of maglev train-guideway system under high-dimensional random track irregularity excitation by surrogate model

Abstract

The extreme responses of maglev train-guideway systems (MTGS) subjected to random irregularities are important in evaluating operation safety and ride comfort. However, further improvements are still required in high-dimensional track irregularity simulation and small probability response prediction. This study presents an efficient analysis framework to predict the extreme response of MTGS at small probability levels. First, a spectral energy-driven dimensional stratification method, combined with an optimized sample selection strategy, is proposed to address the high-dimensional challenges in track irregularity simulation effectively. Then, a tail prediction method, which uses the Support Vector Regression (SVR) surrogate model to predict the probability of exceedance (POE) of extreme response, is introduced to enhance the analysis efficiency of small probability events. Finally, the proposed framework is validated by a two-degree-of-freedom (DOF) maglev train and a complex MTGS. The result shows that the spectral energy-driven stratification method reduces the dimensionality of representative point selection by approximately an order of magnitude. The surrogate model method exhibits significant advantages in predicting small-probability extreme responses. Compared to the Monte Carlo method (MCS), the proposed analysis method significantly enhances computational efficiency in obtaining extreme responses at a 10^-5 probability level.