An AIoT system for real-time monitoring and forecasting of railway temperature

Abstract

Excessive deformation of railway tracks caused by thermal loadings critically affects the efficiency and safety of railway transportation. Accurately quantifying the thermal variations in railway tracks is essential for mitigating heat-related risks. Nevertheless, the complex thermal regime influenced by multiple meteorological factors has posed challenges in understanding the nature of heat-related incidents in railway infrastructure. To investigate the thermal behaviors of railway tracks, this study implemented an IoT monitoring system to measure the temperature along a railway stretch from Changdong to Ssangmun station in Seoul, Korea. Furthermore, a railway temperature forecast model was developed based on Bayesian long short-term memory (BLSTM) trained by the monitoring data. Analyzing the 2-year monitoring results revealed the thermal patterns of the railway, characterized by long seasonal periods and trend stationary. The increasing trend of railway temperature during frequent high-temperature occurrences raised urgent concerns for the railway administration to adapt existing infrastructure to the impacts of climate change. The BLSTM model demonstrated comparable performance with the SARIMA model, a well-established statistical model, and physical models in forecasting the railway temperature, exhibiting a relatively low root mean squared error of 2.21 °C and a bias of − 0.04 °C. Moreover, a notable advantage of the presented BLSTM model is its capacity to provide probabilistic upper and lower bounds of railway temperature, making it suitable for supporting railway safety management. Importantly, using monitoring data as the exclusive input enabled the integration of the BLSTM model into the monitoring system, facilitating the development of a hybrid temperature control system for real-time railway safety management.