Long‐term validation of virtual sensing of a railway bridge with ballasted superstructure

Railway bridges have a long lifespan, which is challenged by the constant development of vehicles leading to increased loads that they were not originally designed for. To ensure the longest possible use of existing structures, a sensor‐based structural health monitoring system can make a significant contribution. However, due to economic reasons and the inaccessibility of many points of interest, sensors cannot be installed everywhere. Therefore, in most cases, only a few sensors are available at a few points of interest, and methods that aim to reconstruct structural responses at unmeasured points from these measurements are referred to as virtual sensing. In this paper, we have analyzed 19,075 passages recorded on a steel trough bridge with a ballast superstructure and a span of 16.4 m, together with weather data. Our findings show that the influence of train type and speed has a significantly higher impact on the results than environmental factors. The investigation revealed that the model‐based analysis produced similar results to the data‐driven analysis concerning acceleration signals. However, when analyzing strain signals, the two approaches yielded distinctly different results.