Statistical Model For Prediction Of Railway Temperature With An Application To Controlling Thermal Buckling

Abstract

Thermal buckling is one of the most critical concerns regarding railroad management, as it poses several dangers to railway transportation and the economy, especially in high-temperature conditions. Owing to the increasing demands for railway transportation caused by explosive urbanization, thermal buckling is also a significant concern in Vietnam. In conjunction with the increasingly profound effect of climate change, thermal buckling poses a severe threat to railroad transportation safety in Vietnam. Several conventional methods (e.g., improve rail quality and change rail shape) were used to prevent thermal buckling. However, they are costly, hard to implement for a whole railway system and they are harmful for the environment. Therefore, this thesis aims to utilize the Statistical Analysis model, especially the Seasonal autoregressive integrated moving average (SARIMA) model to develop a rail-temperature-prediction-model (RTPM) and predict the variation of railroad temperature with time. This model was trained and validated on a monitored database of an ongoing Metropolitan Railroad project. The performance of the SARIMA model is evaluated through five standard evaluation metrics used for appraising the accuracy and performance of a time series forecasting model. The obtained results proved that the model achieved efficient performance and accuracy values. Given the flexibility, simplicity, and high computational performance, the proposed forecast model is expected to be a robust tool for accurately and fast predicting railway temperature variations compared with other RTPMs.