For many years, the ability to predict the behaviour of agglomerates during various processing and transportation techniques has been difficult to achieve. Fundamental knowledge of the parameters and properties of the granular material that controls breakage behaviour has at best been assumed and application to granule processing have been arbitrary in nature.
Computer simulations using Discrete Element Method (DEM) have been developed over the last 20 years to help with the understanding of fundamental behaviour of material properties and its affect on granular breakage. The fault with this technique though is the lack of experimental verification of the simulation code and the underlying theories that it has been based around. This thesis attempts to rectify this shortcoming.
Two model granules, using glass ballotini as the model particle and two separate binders - sodium chloride to analyse brittle
crystalline bonds and poly vinyl alcohol to analyse ductile viscoelastic bonds - were characterised using a number of quantitative techniques. Nanoindentation was utilised to measure mechanical properties i.e. hardness and Young's modulus of the particles and binders. Tensile testing was used to measure the pull-off force (and hence the surface energy) of the binder between two particles. Granule structure was also measured with x-ray microtomography, which provided quantitative data on the spatial locations of the particles within the agglomerate. Qualitatively, SEM analysis of the granular bonds was undertaken to obtain a visual perspective of the bonding behaviour of the two binders.
For past simulations, the granules were created from a random seeding of simulated particles within a volume space, followed by the addition of adhesion during centripetal motion towards the centre of the volume space. For the comparison between real and simulated
agglomerates, a new approach was implemented to input the previously mentioned parameters into the simulation code. Analysis of the x-ray microtomography cross-sectional slices provided three dimensional spatial data of the particles within the agglomerate. Individual particle diameters were also recorded. The simulation code was altered to allow particle information (X-, Y- and Z-coordinates, particle diameters and material type) to be read in from an external file. Adhesion was added to the applied granule structure with values obtained from characterisation tests. Visual analysis between the two granules showed that the granules were identical in shape, with a few minor discrepancies observed.
The observation of slight discrepancies between the experimental and simulated agglomerates lead to an analysis, using DEM simulations, into the importance of agglomerate structure during compression testing. While the effects of impact angle on the breakage
behaviour of agglomerates has been analysed by other authors, the effect of structure and orientation on breakage behaviour during compression has not previously been investigated. Two granules, one spherical in shape, the other highly irregular, were compressed between two plates with the properties of steel. The irregular agglomerate, as a comparison, was compressed in both the Y- and Z-directions. The results of the simulated compressions showed that the structure of randomly packed, polydisperse agglomerates is significant in determining the breakage mechanism, granule strength and the time to fracture of the agglomerate.
Comparison between experimental and simulated compression of two types of granules were completed with quantitative and qualitative analysis. Quantitatively, top wall force plots (which indicate granule strength) show that for the simulated compression of the sodium chloride bound granule, the granule strength is within the same order of
magnitude and the time to fracture was within one order of magnitude from the actual compression. On the other hand, the simulated granule that contained the properties of the poly vinyl alcohol bound granule could not be compared to the experiments due to early termination of the simulation, caused mainly by numerical instability of the simulation.
While the quantitative output appears to match, the visual analysis of the granule does not. Again, the simulation of the poly vinyl alcohol bound granule terminated before an accurate analysis could be completed. However, the sodium chloride bound granule showed discrepancies between the simulated and actual behaviour. The simulated granule fractures much earlier than that of the actual granule, with the simulated granule breaking approximately in one tenth of the time it took to fracture the actual granule. The simulated granule also completely shatters, which is not observed in the experimental agglomerate,
which breaks into three large fragments.
Results indicate that the simulation provides a close approximation of actual behaviour during compression testing. However, there are a number of discrepancies between the two results to indicate that there are significant issues with the use of the simulation code for granular breakage modelling. These issues include numerical instability caused by "bugs" within the simulation code, the inability to model viscoelastic bond behaviour with JKR theory and the inclusion of non-spherical particles within the actual agglomerate.
Despite the unfavourable comparison between the simulated and experimental results, the work here shows that, for the first time, a true experimental granule can be implemented into a computer simulation to predict its behaviour under a wide variety of conditions. This opens many new avenues of research by allowing true granule structures to be tested
without trying to relate ideal mono-sized, structured agglomerates to the behaviour of real granules.