This paper presents a novel method for monitoring continuous welded rails (CWR) to estimate longitudinal stress and determine the rail neutral temperature (RNT). The technique combines vibration measurements, finite element analysis (FEA), and machine learning (ML). FEA establishes the relationship between boundary conditions and stress, serving as the foundation for training an ML algorithm using field data from accelerometers on the track. In the field tests, the method accurately predicted RNT and stress levels, with the ML model demonstrating the ability to learn effectively from experimental data. This approach holds promise for improving rail safety, maintenance, and performance optimization.
|