Empirical prediction of ground-borne vibration from railway systems

Abstract

In today’s society with the trends of urbanization and rapid growth comes an increasing demand for fast and eco-friendly land travel, with low noise and vibration emission. To predict ground-borne noise and vibration from railway systems, several empirical and numerical models have been developed. Generally, soft ground materials can generate low frequency disturbances in the vicinity of the railway. Stiffer materials can transmit a higher frequency ground-borne noise. Investigations in Sweden have shown that areas with non-stiff soil materials, for example very loose clay, can produce high vibration levels at low frequencies. This is true for some of the soil materials deposited during the melting of the last land covering ice cap, which can be found at various sites throughout Sweden (glacial soils). This thesis presents an investigation of ground-borne vibration prediction using an empirical model named High-Speed 2 (HS2), developed in two major high-speed railway projects in the United Kingdom. The relevant underlying theory and the prediction model was studied during the initial literature review. The model was programmed in Matlab, and has been utilized to compute predictions of vibration levels arising at a receiver position, as a result of a train passage. A measurement data set of train passages at a Swedish site where the ground material is constituted by glacial clay, has been acquired and processed. This was done in order to compare measurements and predictions for an evaluation of the model’s accuracy under Swedish conditions. The findings indicate that the HS2 model can compute relatively accurate vibration level spectra using the default reference data within the model, for the studied case. The average single value difference (vertical particle velocity at the soil surface) between vibration level of predictions and measurements is approximately 2.0 dB, with slight over-estimation of levels for most of the studied frequency range, 6.3 to 250 Hz. This is considered as good accuracy for general noise and vibration assessment. However, it should be noted that these results are based on a comparison of the model with measurement data from one single Swedish site. Further, the model seem to produce lower levels than measurement data at very low frequencies. It is possible that new reference source spectra and propagation terms for soft ground materials would increase the accuracy of predictions for low frequency vibration under Swedish conditions. This would require further measurements for sufficient statistical confidence.