Considerations in Modelling Freight Rail Noise

Abstract

An environmental noise model, developed in specialist software such as LimA, SoundPLAN or CadnaA is typically at the core of noise impact assessments for rail projects. These software packages include rail noise modules that account for such variables as train speed and length, and propagation algorithms that allow large-scale models to reliably predict noise levels at thousands of receivers neighbouring the track. The ease with which results are generated by these software packages can easily provide a false sense of accuracy however, as the old maxim of noise modelling still applies—garbage in equals garbage out. This is particularly the case for freight rail operations which involve complex noise sources. These include multiple moving noise sources, such as locomotive exhaust noise and rolling noise, that exhibit distinct relationships with train speed. There are localised sources such as crossovers and curve noise which have a particularly annoying character. There are sources that exhibit different characteristics under different operational scenarios. For example, locomotive noise can be dominated by the engine exhaust under traction on an uphill grade and ruled by dynamic brake noise on a downhill grade. This paper presents a discussion of issues to be considered to successfully model freight rail noise. It describes the different noise sources which apply, when they apply and how to incorporate them in a computer noise model. An outline is provided on the most appropriate propagation algorithms to apply and how sources such as curve noise and brake noise, can be accounted for. Attention is also given to validation of the noise model at each stage of its development, including how to make best use of noise measurements to verify model predictions.