Prediction on soil-ground vibration induced by high-speed moving train based on artificial neural network model

Abstract

Vibration prediction plays a vital role in the environmental assessment of new railway lines and the vibration analysis of new buildings near railway lines. We use theoretical analysis, numerical simulation, and field experiment, which can help predict vibration with high accuracy. These methods have their own advantages, but are time-consuming or costlier. This article outlines a fast, high-accuracy prediction method combining the advantages of numerical simulation, experiment, and artificial neural network. The method takes artificial neural network as one of the supervised machine learnings, and it is founded upon a fully validated three-dimensional finite element model. First, establish a three-dimensional finite element model using ANSYS software and generate series of samples by changing field conditions. Then, create an artificial neural network and train it by the samples obtained from FEM and experiment. Finally, use the artificial neural network that has been trained to predict the level of vibration on certain conditions. The result states that the predicted method can control the maximum error below 6.41% and the average error below 2.29% when it is used to predict acceleration vibration level. An advantage of this method is that it spends zero time on predicting vibration level. A key benefit from this decreased prediction time is that it potentially reduces the detailed vibration analyses required for a new train line so that it can reduce project costs. In addition, it could also consider the effect of soil parameters.