Degree Name

Doctor of Philosophy


Department of Civil, Mining and Environmental Engineering


Railway ballast forms a major component of a conventional rail track and is used to distribute the load to the subgrade, providing a smooth running surface for the train. Currently, high speed (frequency) and heavy haul trains are increasing the loads experienced by the ballast layer. This research aims to investigate the densification and degradation behaviour of ballast under high frequency cyclic loading. Laboratory experiments were conducted using large-scale cyclic triaxial equipment to assist in the understanding of ballast deformation and degradation under high frequency loading. Furthermore, the Discrete Element Method (DEM) was employed to study the ballast behaviour from a particulate approach. Finally, a cyclic densification model was proposed as an extension of a previous monotonic loading model developed by Salim and Indraratna (2004). The model was calibrated and validated using laboratory and published data.

That there is an increase in load on the ballast layer due to the increase in train speed has been widely accepted. However, past cyclic laboratory experiments conducted at various frequencies have shown that there is little effect of the change in frequency (speed) on the densification of granular material. Unless the magnitude of the load is also changed with change in frequency during triaxial tests, the effect of frequency cannot be observed clearly. This shortcoming was overcome in this study by increasing the amplitude of cyclic loading with corresponding increase in frequency. The ballast contact pressures (loading amplitude) were calculated based on the wellknown empirical relationship proposed by Esveld (2001).

In this research, a series of drained cyclic triaxial tests were conducted at various frequencies and confining pressures using the large-scale cyclic triaxial apparatus. The experimental results revealed that both the densification and breakage of ballast increase with the increase in frequency and number of cycles. However, there is a range of frequency (20 Hz ≤ f ≤ 30 Hz) where cyclic densification occurs without any significant particle breakage. Moreover, the resilient modulus of the ballast was found to increase with the increasing number of cycles and confining pressures, while it was found to decrease with the increasing frequency. Furthermore, it was observed that the ballast layer requires a certain confinement to operate a train at high speed efficiently. An empirical approach was devised to calculate the required confining pressure and resilient modulus of the ballast layer for a desired level of axial strain (vertical deformation) for a given train speed.

A DEM-based commercially available program, Particle Flow Code in 2-Dimensions (PFC2D), was used to investigate the ballast behaviour at the particulate (micromechanics) level. 2-D projections of real ballast particles were modelled to create various shapes and sizes of ballast particles, and to represent adequate angularity. These created particles were treated as breakable and unbreakable in order to study the effect of particle breakage on cyclic densification and to explore the breakage mechanism. As PFC2D does not have the capability to apply real-time cyclic loading, a new subroutine was developed to apply the cyclic loading so that the effect of loading frequency could also be investigated.

The DEM simulation results revealed that particle breakage is a governing aspect of the actual behaviour of granular material. Unbreakable particles underestimated the densification of the sample, while breakable particles simulated the densification of the sample very close to that observed during laboratory tests. Initial rapid deformation observed in case of breakable particles is higher than that observed in the case of unbreakable particles. Higher initial deformation during breakable particles can be attributed to particle breakage. Moreover, the DEM results confirmed that the particles break under tension and the breakage is mainly oriented and concentrated in the direction of particle movement. In addition, a direct relationship between maximum particle displacement and bond breakage was also observed.

As cyclic loading is a very complex phenomenon, a semi-empirical approach was adopted. A non-associated flow rule was used and modified based on the experimental observations that captured particle breakage. The model is based on a critical state framework and has been calibrated and validated using cyclic triaxial experiments conducted in this study, as well as data reported in published literature. The model represents ballast behaviour under a wide range of stress conditions.