Mecsol 2022 Webinar Series #3

This free webinar is part of Mecsol 2022's Webinar Series. Visit our website to see the list of all webinars.

Railway induced vibration in the built environment: assessment, prediction, and mitigation

Considerable progress was made during the last decades in developing numerical prediction models for railway induced vibration in the built environment. Accurate numerical prediction of railway induced vibration in buildings remains very challenging, however, due to the complexity of the coupled dynamic soil-structure interaction problems involved, the wide frequency range of interest, the large number of determining parameters, and the uncertainty involved. 

These challenges are illustrated by an extensive measurement campaign that was conducted in a three-storey reinforced concrete building, located at 40 m from a ballasted track on an embankment. Synchronous vibration measurements were performed in 84 directions using accelerometers mounted on the track, in the free field and in the building. Transfer functions were determined using impact hammer excitation at 17 sleeper positions, while the response due to freight and commuter trains was measured during one week. Bayesian updating was used to infer soil profiles from in situ geophysical tests (SASW, seismic refraction, SCPT). 2.5D and periodic coupled finite element – boundary element (FE-BE) models of the track and soil were calibrated by means of measured track receptance and transfer functions. Model updating was also applied to a coupled FE-BE model of the reinforced concrete building, accounting for dynamic soil-structure interaction, and using the building’s modal characteristics that were identified using ambient and forced excitation.

Despite all efforts taken to identify the properties of the track, soil, and building, the predicted and measured response in the free field and in the building differ by 10 dB in the dominant range of frequencies. The incident wave field due to impacts on the sleepers and train passages is very sensitive to uncertain dynamic soil properties, which in turn also explains the observed discrepancy between the predicted and measured response of the building.

Industry experiences a strong need (1) to conduct extended parametric studies, (2) to optimize vibration mitigation measures in a robust way, and (3) to quantify and reduce model and parameter uncertainties. These applications all require detailed vibration prediction models that are fast to run; we will therefore investigate the potential of model order reduction techniques to considerably speed-up state-of-the-art prediction models. There also remains a need for scoping models that are fast to run and can be used in an early design phase on a larger urban scale; we therefore also plan to further develop hybrid prediction models that combine numerical prediction with experimental results.

Chaired by Prof. Josué Labaki - University of Campinas (Unicamp)