A comparative study of different numerical models for predicting train noise in high-rise cities

Large-scale outdoor field measurements were carried out on a residential building to assess the noise levels caused by pass-by trains that run on a nearby viaduct. The experimental results were compared with different schemes for predicting noise from trains. The octave band sound power levels of the train passing by, which are required as input parameters for the Nordic prediction method for train noise (NMT), CSTB 92 and ISO 9613-2 provided in the Mithra software, were determined by an inversion method. The method of calculation of railway noise (CRN) from the UK gives the best agreement with the measured results. The NMT prediction scheme also provides a good prediction of the general trend of the experimental data, but it always overestimates the measured noise levels. As far as the quantitative agreement with experimental data is concerned, the CSTB 92 and ISO 9613-2 prediction schemes are comparatively less satisfactory.