In any DEM simulation, the chosen particle shape will greatly influence the simulated material behaviour. For a specific material, e.g. railway ballast, it remains an open question how to model the particle shape, such that DEM simulations are computationally efficient and simulation results are in good accordance with measurements. While DEM shape modelling for railway ballast is well addressed in the literature, approaches mainly aim at approximating the stones’ actual shape, resulting in rather complex and thus inefficient particle shapes. In contrast, very simple DEM shapes will be constructed, clumps of three spheres, which aim to approximate shape descriptors of the considered ballast material. In DEM simulations of the packing behaviour, a set of clump shapes is identified, which can pack at porosities observed at track sites, as well as in lab tests. The relation between particle shape (descriptors) and obtained packing (characteristic) is investigated in a correlation analysis. The simulated packing’s porosity is strongly correlated to four shape descriptors, which are also strongly correlated among each other. Thus, to derive simple shape models of a given particle shape, matching one of these shape descriptors, might be a good first step to bring simulated porosities closer to measured ones. The conducted correlation analysis also shows that packing’s coordination number and isotropic fabric are correlated to more shape descriptors, making it more difficult to estimate the effect of particle shape on these quantities. Graphic
Ballasted tracks are the commonly used railway track systems with constant demands for reducing maintenance cost and improved performance. Elastic layers are increasingly used for improving ballasted tracks. In order to better understand the effects of elastic layers, physical understanding at the ballast particle level is crucial. Here, discrete element method (DEM) is used to investigate the effects of elastic layers – under sleeper pad ( $$\text {USP}$$ ) at the sleeper/ballast interface and under ballast mat ( $$\text {UBM}$$ ) at the ballast/bottom interface – on micro-mechanical behavior of railway ballast. In the DEM model, the Conical Damage Model (CDM) is used for contact modelling. This model was calibrated in Suhr et al. (Granul Matter 20(4):70, 2018) for the simulation of two different types of ballast. The CDM model accounts for particle edge breakage, which is an important phenomenon especially at the early stage of a tamping cycle, and thus essential, when investigating the impact of elastic layers in the ballast bed. DEM results confirm that during cyclic loading, $$\text {USP}$$ reduces the edge breakage at the sleeper/ballast interface. On the other hand, $$\text {UBM}$$ shows higher particle movement throughout the ballast bed. Both the edge breakage and particle movement in the ballast bed are found to influence the sleeper settlement. Micro-mechanical investigations show that the force chain in deeper regions of the ballast bed is less affected by $$\text {USP}$$ for the two types of ballast. Conversely, dense lateral forces near to the box bottom were seen with $$\text {UBM}$$ . The findings are in good (qualitative) agreement with the experimental observations. Thus, DEM simulations can aid to better understand the micro-macro phenomena for railway ballast. This can help to improve the track components and track design based on simulation models taking into account the physical behavior of ballast. Graphical
The prediction quality of discrete element method (DEM) models for railway ballast can be expected to depend on three points: the geometry representation of the single particles, the used contact models and the parametrisation using principal experiments. This works aims at a balanced approach, where none of the points is addressed with excessive depth. In a first step, a simple geometry representation is chosen and the simplified Hertz–Mindlin contact model is used. When experimental data of cyclic compression tests and monotonic direct shear tests are considered, the model can be parametrised to fit either one of the two tests, but not both with the same set of parameters. Similar problems can be found in literature for monotonic and cyclic triaxial tests of railway ballast. In this work, the comparison between experiment and simulation is conducted using the entire data of the test, e.g. shear force over shear path curve from the direct shear test. In addition to a visual comparison of the results also quantitative errors based on the sum of squares are defined. To improve the fit of the DEM model to both types of experiments, an extension on the Hertz–Mindlin contact law is used, which introduces additional physical effects (e.g. breakage of edges or yielding). This model introduces two extra material parameters and is successfully parametrised. Using only one set of parameters, the results of the DEM simulation are in good accordance with both experimental cyclic compression test and monotonic directs shear test.
Railway ballast is an angular and coarse material, which demands careful DEM modelling and validation. Particle shape is often modelled in high accuracy, thus leading to computational expensive DEM models. Whether this effort will increase the DEM model’s overall prediction quality will also vitally depend on the used contact law and the validation process. In general, a DEM model validated using different types of principal experiments can be considered more trustworthy in simulating other load cases. Here, two types of railway ballast are compared and DEM model validation is conducted. Calcite and Kieselkalk are investigated under compression and direct shear test. All experimental data will be made openly accessible to promote further research on this topic. In the experiments, the behaviour of Calcite and Kieselkalk is surprisingly similar in the direct shear test, while clear differences can be seen in the stiffnesses in the compression test. In DEM modelling, simple particle shapes are combined with the Conical Damage Model contact law. For each type of ballast, one set of parameters is found, such that simulation and experimental results are in good accordance. A comparison with the simplified Hertz-Mindlin contact law shows several drawbacks of this model. First, the model cannot be calibrated to meet both compression and shear test results. Second, the similar behaviour in shear testing but differences in compression cannot be reproduced using the Hertz-Mindlin model. For these reasons, the CDM model is considered the better choice for the simulation of railway ballast, if simple particle shapes are used.
Open research data practices are a relatively new, thus still evolving part of scientific work, and their usage varies strongly within different scientific domains. In the literature, the investigation of open research data practices covers the whole range of big empirical studies covering multiple scientific domains to smaller, in depth studies analysing a single field of research. Despite the richness of literature on this topic, there is still a lack of knowledge on the (open) research data awareness and practices in materials science and engineering. While most current studies focus only on some aspects of open research data practices, we aim for a comprehensive understanding of all practices with respect to the considered scientific domain. Hence this study aims at 1) drawing the whole picture of search, reuse and sharing of research data 2) while focusing on materials science and engineering. The chosen approach allows to explore the connections between different aspects of open research data practices, e.g. between data sharing and data search. In depth interviews with 13 researchers in this field were conducted, transcribed verbatim, coded and analysed using content analysis. The main findings characterised research data in materials science and engineering as extremely diverse, often generated for a very specific research focus and needing a precise description of the data and the complete generation process for possible reuse. Results on research data search and reuse showed that the interviewees intended to reuse data but were mostly unfamiliar with (yet interested in) modern methods as dataset search engines, data journals or searching public repositories. Current research data sharing is not open, but bilaterally and usually encouraged by supervisors or employers. Project funding does affect data sharing in two ways: some researchers argue to share their data openly due to their funding agency’s policy, while others face legal restrictions for sharing as their projects are partly funded by industry. The time needed for a precise description of the data and their generation process is named as biggest obstacle for data sharing. From these findings, a precise set of actions is derived suitable to support Open Data, involving training for researchers and introducing rewards for data sharing on the level of universities and funding bodies.