Abstract
Abstract To address the issue of the complexity and high computational cost of ground vibration prediction models induced by high-speed trains crossing bridges, this study proposes a prediction method based on machine learning. Initially, this research derives a dataset of ground vibration data based on the analytical solution of a moving load acting on an elastically supported beam and the Green’s function of a simple harmonic point load acting on an elastic half-space, and uses these datasets to pre-train a neural network model. Subsequently, a series of experiments on ground vibrations induced by high-speed trains crossing bridges were conducted along the high-speed railway line from Hangzhou to Changsha, and the obtained vibration data were used to further train and optimize the neural network model to enhance the prediction accuracy. Finally, the optimized model was applied to predict the measured ground vibration data and compared with actual measurement results to verify the effectiveness and generalization capability of the neural network model in practical applications. Machine learning technology provides an efficient method for predicting environmental vibrations in the field of railway transportation, playing a key role in promoting the green and sustainable development of the railway industry.