Empirical prediction of ground-borne vibration from railway systems; validating the HS2 model in Sweden, west coast

Abstract

Trains are becoming a more and more viable option as a fast, environmentally friendly travel choice. Along with this cities grows larger and civilizations becomes more urbanised. This results in an increase of human exposure to vibrations from trains, creating a need of vibration level predictions. Predictions enables at an early stage to be able to tell the impact of a new railway system. Whether residence buildings can be built at a certain proximity to the rail or hospitals, research facilities with vibration sensitive instruments, what trains with what speeds can pass and so on. One model to do this is the British empirical model named High Speed 2 (HS2). In Sweden, especially the west coast, the lithology differs from that of the UK as a result of the melting glaciers. Soft ground material, such as the Swedish glacial clay have tendencies to generate low frequency vibrational disturbances. This thesis is an evaluation of the HS2 model under Swedish conditions. It is limited to predicting the vibration levels in the ground at a certain distance from a surface running train. In 2021 a thesis work of Gustav Vågfelt was done were the HS2 model was implemented in Matlab and evaluated against a measurement data set at a Swedish site with help of Efterklang - part of AFRY [1] [15]. This thesis will continue on that work and in cooperation with Efterklang use the Matlab implementation to evaluate the model by conducting measurements at two different sites, one being post glacial fine sand and the other glacial clay. As there are still uncertainties in the model parameters a final conclusion of its precision can not yet be set. With the data available however the HS2 performs decent on the fine sand with quite consistent results and with the evaluation tools a table of values presenting the expected precision of the model is presented. The HS2 model did however not perform as well on the glacial clay. Because of this it was extended with an additional lithology to fit the Swedish glacial clay. This lithology performed excellent for the glacial clay site measurements, however that is expected as it originated from them. It was also tested on external measurement data from Greby, Skövde which also consisted of glacial clay. Here the extension of the model was less accurate. This indicates that measurement data from more glacial clay ground sites are needed to get a more accurate model extension.