![]()
1. Sample preparation and characterisation![]()
Investigation into 3D printing of granular media
O. Adamidis, S. Alber & I. Anastasopoulos
ETH Zürich, Switzerland
ABSTRACT: Advances in additive manufacturing have recently motivated interest into 3D printing of geomaterials for an array of applications within geotechnical research. However, certain obstacles exist, which are investigated in this paper. Firstly, the geometry of the proposed particles has to be chosen, so that it is representative of an actual geomaterial. Here, Hostun sand was chosen as a reference material. It was found that CT scanning, microscopy, and morphology analysis were useful tools in assessing whether the proposed medium was representative of the reference material. Secondly, a specific 3D printing technology has to be chosen. Amongst the commercially available technologies investigated, PolyJet was found to be the most appropriate for the creation of small particles. Available materials, printing resolution, and support material removal were considered. Support material removal was found to be a limiting factor as very small particles can delaminate during this process. As a result, a particle diameter of 2 mm is proposed as the lower limit that can be reliably reproduced with the current level of PolyJet 3D printing.
1 INTRODUCTION
Advances in the field of 3D printing have sparked an interest for this methodology within geotechnical research. Interest stems from the unique opportunity provided by 3D printing for the creation of granular media with full and independent control of particle size, morphology, and material properties.
Significant promise exists in furthering our understanding of the influence of micro-scale particle characteristics on the macro-scale behaviour of granular media. For instance, Miskin and Jaeger (2013) focused on the role of particle shape on packing stiffness. They altered the shape of particles in the context of artificial evolution and used 3D-printing to recreate the computed shapes and verify the results of their simulations.
Another area of application is the potential synergy between 3D printing and Discrete Element Method (DEM) modelling. 3D printing could be used to recreate DEM generated particles and particle arrangements, offering a unique opportunity for meaningful validation of the numerical simulations. Kondo et al. (2017), performed such a validation by comparing the permeability of a granular medium, as calculated through numerical simulation using DEM and as measured for a 3D printed specimen which recreated the same particle arrangement.
Lastly, the use of 3D printing to produce standardised geomaterials for quality control and calibration of geotechnical testing devices has been proposed by Hanaor et al. (2015). By fully controlling the material properties and the size and morphology of particles, uniform and consistent specimens could be reliably reproduced.
Though the potential for the application of 3D-printing in geotechnical research is significant, several important issues remain. Firstly, the geometry of the particles that are to be 3D-printed has to be generated. Secondly, the appropriate 3D printing technology has to be chosen. Finally, quality assessment of the printed particles is necessary, in order to verify that the required geometries have been successfully reproduced. This last step is of increased importance since small particles deviate significantly from typical 3D printed objects, both in size and morphology.
2 DEFINING PARTICLE GEOMETRY
2.1 Choosing a reference material
Two avenues are available in defining the geometry of particles. The first is to choose simplified geometries in order to validate DEM simulations, as done by Kondo et al. (2017), who used spheres. The second is to produce particles that represent real geomaterials. In this case, geometries can be either directly obtained from a real material or stochastically generated through an algorithm, e.g. as proposed by Hanaor et al. (2015).
Here, the focus was on the capacity of 3D printing to recreate particles which include the small features that exist in real geomaterials. To that end, Hostun sand grains were chosen as reference particles to be replicated. Hostun sand is a poorly graded sand with d50 = 0.34 mm that has been widely used in geotechnical research (Mokni and Desrues 1999, for instance). It is an angular sand whose grains contain small features and are therefore appropriate for the pursued investigation.
Figure 1. The 65 CT scanned Hostun sand grains.
2.2 Micro Computed Tomography (CT) scanning
In order to capture the 3D geometry of Hostun sand grains, micro computed tomography (CT) was used. CT scanning has been used rather extensively in the geotechnical field, primarily to study the morphology of granular media and rocks (Desrues et al. 2006, for instance). Here, micro CT scanning was performed on 65 grains of Hostun sand. The number of grains chosen from different diameter size intervals corresponded to the relevant mass fraction calculated from the Particle Size Distribution (PSD) curve of the sand.
A μCT scanner (μCT50) of SCANCO Medical AG (Brüttiselllen, Switzerland) was used. Scanning was performed at the Institute for Biomechanics of ETH Zürich. In order to define the resolution for the scans, the approach suggested by Fonseca et al. (2012) was used. For Reigate Sand, they proposed a resolution of 5 μm, which corresponds to 0.018 × d50. Extending this approach to Hostun sand, for which 0.018 × d50 = 6 μm, it was decided that a resolution of 5 μm would be sufficient. An overview of the relevant additional parameters to be set for CT scanning is given by Stauber and Müller (2008).
For the image processing of CT scans, the aspects that need to be considered include image filtering, segmentation, and component labelling (Stauber and Müller 2008). Image filtering reduces the noise within the image and hence enhances the visualisation of important features. During segmentation, images are discretized into two phases for further processing. Here, a threshold value was defined and voxels above it were set to white (object) while voxels below it were set to black (background). The last image processing step is component labelling, during which all small unconnected and undesired objects are removed. These small objects, which emerge because of image noise or improper sample preparation, are labelled by an algorithm and all particles below a certain size are eliminated. The sand grains as obtained after these processes are shown in Figure 1.
3 COMPARISON WITH REAL MEDIUM
In order to assess whether the 65 grains scanned were enough to ...
