Field monitoring and prediction of the thermal response of an in-service curved continuous welded rail using distributed fiber optic strain measurements

In this paper, distributed fiber optic sensing (DFOS) was used to monitor indicators of track buckling, i.e., the axial strain, curvature, and lateral deflection along a 20-m section of curved track during the summer when ambient temperatures are highest. The objectives of the research were to (i) evaluate the performance of DFOS for this type of monitoring, (ii) develop an improved understanding of the lateral and longitudinal thermal response of a curved rail, and, (iii) create Gaussian process regression (GPR) models to assist in the prediction of the rail thermal response. The DFOS provided temperature measurements along the rail; however, local variations in temperature over the cross-section were not captured by the single temperature fiber installed. Multiple temperature fibers are required to correct each strain measurement for temperature to accurately capture the track behavior. The DFOS data were used to evaluate two of the critical parameters that indicate the potential for buckling due to temperature variation: the axial strain and lateral bending curvature. Gaussian process regression was used to predict the rail response to temperature and enable buckling assessment with uncertainty bounds.