DEM modelling of railway ballast using the Conical Damage Model: a comprehensive parametrisation strategy

Abstract

Despite ongoing research, the parametrisation of a DEM model is a challenging task, as it depends strongly on the particle shape representation used, particle-particle contact law and the simulated applications: for railway ballast e.g. lab tests or track conditions. The authors previously modelled railway ballast with a DEM model using a simple particle shape. The DEM model was parametrised, by trial-and-error, to compression and direct shear test results. A good agreement between DEM model and experimental results was achieved only when the Conical Damage Model (CDM) was used as the contact law. Compared to the well-known linear-spring Cundall-Strack law or the Hertz-Mindlin law, this contact law takes into account additional physical effects (e.g. edge breakage) occurring in the experiment. Little is known on the influence of the CDM model parameters on the simulation results or on possible parameter ambiguities. This lack of knowledge hinders a reliable and efficient parametrisation of DEM models using different particle shapes. Both points are addressed in this work in detail by investigating a DEM model for railway ballast using one simple particle shape. Suggestions for a parametrisation strategy of reduced computational effort are formulated and tested using a second particle shape. In future works, the newly presented parametrisation strategy can help to calibrate different DEM models and to study the influence of particle shape.

Efficient DEM simulations of railway ballast using simple particle shapes

“…Sample generation and pre-compaction is done using the method developed in [28] and will be summarised here briefly. The samples were generated using the CDM law (2)…”

“…The parametrisation of the DEM models to the compression and direct shear tests is conducted using the method described in [28]. In the parametrisation step the parameters of the CDM law are searched, E, E∕ max , , , which bring the simulation results close to the experimental ones.…”

“…In a similar manner a cost function for the direct shear test was defined, shear (p i ) , which takes into account DEM simulations of three levels of applied normal load for three different initial configurations. A combined cost function (p i ) is defined as the mean value of the cost functions for compression and direct shear test, see [28] for details.…”

“…In [28], a parametrisation strategy for the CDM law was developed. Wherever possible, measured values were used to define ranges of parameters.…”

“…Single stone measurements on the coefficient of friction of both types of ballast showed a high amount of scatter, see [21]. As the CDM law involves four parameters, the problem of possible parameter ambiguity was investigated in detail in [28]. High parameter ambiguity means that different (non neighbouring) parameter sets give the same (or similar) simulation results.…”