Background and Aims:The aim of the current study was to investigate Hong Kong

A detailed experimental programme investigated how the coefficient of friction at particle contacts for a typical UK railway ballast varied during cyclic loading. Despite a decrease of roughness, the friction coefficient increased steadily, stabilising in the region of tens of cycles. In contrast to previous work on the contact behaviour of sands, water inundation caused a significant but reversible decrease of the coefficient, while the generation of a significant amount of abraded fines did not affect it, nor did the load level.

A newly developed apparatus was used to investigate the contact mechanics between particles of a common UK railway ballast. The data were then compared with the models currently and commonly used in geotechnical DEM analyses but the discrepancy between the predictions and measurements is large even at small loads. The first contact behaviour was much softer than the Hertz-Mindlin model for smooth spheres, both in normal and shear. Even if the normal loading could be fitted better with a model that accounted for roughness, the elastic nature of these models could not capture the plasticity that is evident on unloading. Large displacement shear cycles caused not only an increase of inter-particle friction as discussed previously, but also a significant degree of particle interlock arising from the wear of the contacts. While there were no rate effects for sliding shearing, contact ageing displacements could be observed at pre-failure loads, although these stabilised after a few days. Pre-failure cyclic loading resulted in stiffnesses that increased in lateral loading but decreased in normal loading, but in both cases stabilised after a few tens of cycles, while the stiffnesses at the reversals of the large displacement cycles did not evolve with continued cycling. The presence of water did not affect the lateral stiffness and the influence of the normal load was consistent with Hertz-Mindlin, even if the absolute values were much softer. It can be concluded that the current commonly used models for DEM analyses are not applicable for ballast/crushed rock and alternate models should also focus on plasticity at the asperity scale.