Micro-scale behaviour of recycled construction and ... Behaviour of Recycled Construction and Demolition Materials: Discrete Element Method Simulations and Physical Testing A thesis

Department of Civil and Construction Engineering

Faculty of Science, Engineering, and Technology

Micro-scale Behaviour of Recycled Construction and

Demolition Materials: Discrete Element Method

Simulations and Physical Testing

A thesis submitted for the degree of Doctor of Philosophy

by

Tabassom Afshar

October 2017

Swinburne University of Technology, Melbourne, Australia

iii

To my love

iv

Abstract Granular recycled Construction and Demolition (C&D) materials used in pavements,

roads, and embankments experience static and dynamic loading during their service life.

As a result, particle breakage can cause serious issues such as settlement and reduction in

hydraulic conductivity. Particle breakage depends on different factors such as particle

mineralogy, loading condition, particle size, and particle shape. A new insight into the

importance and effect of particle shape on the degree of crushing is presented. The

particle behaviour of C&D particles, with different mineralogy and microstructure, at

different scales from a single particle to an assembly of particles, is presented and

discussed. The fracture characteristics were found to be highly dependent on the particle

shape factor, and a modified particle tensile strength as a function of particle aspect ratio

is introduced. It has also been found that particle shape plays a more prominent role in

the particle breakage phenomena than the mineralogy and microstructure of C&D

particles. Discrete Element Modelling/Method (DEM) was also used to evaluate the

evolution of cracks through the particles. DEM assisted in measuring breakage energy

more accurately by partitioning and tracking the energy dissipation especially through the

creation of new surfaces during fragmentation. More accurate three-dimensional particle

shapes were generated by a large number of bonded spherical sub-particles and used to

model single particle crushing and particle assembly crushing. The results demonstrated

that brittle C&D granular materials with a higher degree of sphericity and lower flakiness

index would show higher resistance to breakage. Moreover, in order to obtain a better

understanding of C&D particle behaviours in relation to grain-scale damage and post-

breakage changes in grain properties, Synchrotron Radiation-based X-ray Micro-

Computed Tomography (SR-µCT) was used to capture high-resolution 4D images (i.e.

3D monitoring over time) from grain assemblies subjected to compression at different

loading intervals. To conduct CT scanning, a novel loading apparatus, capable of

conducting compression tests at the high stress on an assembly of grains, was designed

and developed. Changes to grain properties, stemming from breakage during compression

of specimens with various particle size distributions, were studied with the aid of three-

dimensional SR-µCT. The fractal distribution of C&D assemblies showed that breakage

becomes dominant in smaller grains rather than larger ones, where an increase in the

amount of newly generated fine fragments leads to a high coordination number

surrounding the larger grains. More importantly, the results of morphological changes in

v

the particulate assemblies revealed that there is a reversal trend in the grain morphology

evolution with increasing stress. The grains tended to create more spherical fragments

with higher aspect ratio whereas, by increasing the stress, this trend completely reversed.

In addition, it was found that the general trend of changes in particle shape obeys

universality. In other words, the same generic evolution by increasing stress was observed

irrespective of material types or sizes.

vi

Acknowledgments I would first like to express my sincere gratitude to Dr. Mahdi Disfani for his untiring

support, guidance, and mentorship during this research. In fact, the work presented in this

thesis would have really been difficult to pursue without his consistent encouragement by

asking questions and providing ideas. Thank you for giving me this opportunity to pursue

the research presented here. I would also like to thank Prof. Arul Arulrajah and Dr.

Guillermo Narsilio for your invaluable support during this project. Whenever a problem

was raised, I could always count on your assistance and scientific guidance.

The CT scanning discussed in the present work was undertaken in the Imaging and

Medical Beam Line at Australian Synchrotron (National Centre for Synchrotron

Science), Victoria, Australia. I wish to give my best thanks to beam scientists at

Australian Synchrotron, technicians in Swinburne workshop, Swinburne advanced

geotechnical engineering laboratory manager, and Swinburne microfabrication and

micro-analytical facility manager for their priceless assistance in conducting

experimental work presented in this dissertation. I wish to acknowledge Alex Fraser

Group for providing samples of recycled concrete, brick, and crushed rock for this

research. The support of Itasca Australia during numerical simulations is also gratefully

acknowledged.

Finally yet importantly, I would like to thank my family, friends, and colleagues at

Swinburne and the University of Melbourne, in particular my parents, Farzaneh and Ali,

and also my sister, Tarannom, for their support and inspiration throughout my studies.

Words could not ever sufficiently show my gratitude for all that you do for me. Special

thanks to my love, Moji, you are a wonderful husband, and I feel so lucky that I have you

on my side. Your love, kindness, patience, and understanding during these years know

no bounds. I love you with all my heart.

vii

Declaration I, Tabassom Afshar, declare that this thesis entitled:

“Micro-scale Behaviour of Recycled Construction and Demolition Materials: Discrete

Element Method Simulations and Physical Testing”

is my own work and has not been submitted previously, in whole or in part, in respect

of any other academic awards.

Tabassom Afshar

Department of Civil and Construction Engineering

Faculty of Science, Engineering, and Technology

Swinburne University of Technology, Melbourne, Australia

viii

List of publications 1. T. Afshar, M. M. Disfani, A. Arulrajah, G. A. Narsilio, & S. Emam, 2017.

“Impact of Particle Shape on Breakage of Recycled Construction and Demolition Aggregates”, Powder Technology, 308:1-12, doi:10.1016/j.powtec.2016.11.043. (IF: 2.825)

2. T. Afshar, M. M. Disfani, G. A. Narsilio, & A. Arulrajah, 2017. “Post-breakage changes in grain properties using synchrotron tomography”, Powder Technology, revised version submitted on 04/10/2017.

3. T. Afshar, M. M. Disfani, G. A. Narsilio, & A. Arulrajah, 2017. “Changes to Grain Properties due to Breakage in a Sand Assembly using Synchrotron Tomography”, European Physical Journal-Web of Conferences, Vol. 140, Article No. 07004, doi:10.1051/epjconf/201714007004.

4. T. Afshar, M. M. Disfani, A. Arulrajah, & G. A. Narsilio, 2017. “Microstructural analysis of particle crushing in Construction and Demolition materials using synchrotron tomography”, Proceedings of the 19th International Conference on Soil Mechanics and Geotechnical Engineering, Seoul, 2017, pp. 1003-1006.

5. T. Afshar, M. M. Disfani, A. Arulrajah, & G. A. Narsilio, 2015. “Discrete Element Modelling of Recycled Waste Rock: Particle Shape Simulations and Effects”, Proceedings of 12th Australia New Zealand Conference on Geomechanics (ANZ 2015), Wellington, New Zealand, 2015, Article No. 186.

6. T. Afshar, M. M. Disfani, A. Arulrajah, & G. A. Narsilio, 2014. “Discrete Element Modelling of Recycled Waste Rock under Monotonic Loading”, Proceedings of the 67th Canadian Geotechnical Conference (GeoRegina), Saskatchewan, Canada, 2014, Article No. 236.

List of grant and award

• M.M. Disfani, T. Afshar, G.A. Narsilio, & A. Arulrajah; “Micro-scale behaviour of recycled construction and demolition materials: focus on particle shape and breakage”; beamtime CT imaging, grant No. AS161/IM/10502, Australian Synchrotron (National Centre for Synchrotron Science), Feb. 2016

• Itasca Educational Partnership Award, Itasca Consulting Group, Feb. 2015

ix

List of abbreviations 2D: two-dimensional

3D: three-dimensional

AR: Aspect Ratio

C&D: Construction and Demolition

CB: Crushed Brick

CCD: Charge-Coupled Device

DEM: Discrete Element Modelling/Method

EDS: Energy-Dispersive X-ray Spectroscopy

FI: Flakiness Index

IDT: Indirect Diametral Tensile strength

IMBL: Imaging and Medical Beam Line

LVDT: Linear Variable Differential Transformer

MASSIVE: Multi-modal Australian ScienceS Imaging and Visualization

Environment

MDD: Maximum Dry Density

OMC: Optimum Moisture Content

PAC: Particle Assembly Crushing

PCM: Portland Cement Mortar

PFC: Particle Flow Code

PIV: Particle Image Velocimetry

PSD: Particle Size Distribution

x

RCA: Recycled Concrete Aggregate

RDF: Relative Distribution Factor

RP: Representative Particle

SEM: Scanning Electron Microscopy

SPC: Single Particle Crushing

SR-µCT: Synchrotron Radiation-based X-ray Micro-Computed Tomography

UCS: Unconfined Compressive Strength

WR: Waste Rock

1

Contents Abstract .................................................................................................................................... iv

Acknowledgments ................................................................................................................... vi

Declaration .............................................................................................................................. vii

List of publications ................................................................................................................ viii

List of grant and award .......................................................................................................... viii

List of abbreviations ................................................................................................................ ix

Contents ................................................................................................................................... 1

List of figures ........................................................................................................................... 6

List of tables .......................................................................................................................... 11

1. INTRODUCTION ...................................................................................................... 12

1.1. Problem statement and research significance ......................................................... 12

1.2. Objective and scope ................................................................................................ 13

1.3. Research method ..................................................................................................... 14

1.4. Thesis outline .......................................................................................................... 16

2. LITERATURE REVIEW ........................................................................................... 18

2.1. Construction and Demolition materials .................................................................. 18

2.2. Necessity of micro-scale studies on C&D materials ............................................... 19

2.3. Micro-scale study of geomaterials .......................................................................... 21

2.3.1. Discrete Element Modelling ........................................................................... 22

2.3.1.1. Laboratory test simulations ......................................................................... 24

2.3.1.2. Different materials ...................................................................................... 24

2.3.2. Experimental methods ..................................................................................... 26

2.3.2.1. X-ray Tomography ...................................................................................... 27

2.3.2.2. Particle Image Velocimetry (PIV) .............................................................. 28

2.4. Particle breakage in granular materials ................................................................... 28

2.4.1. Factors governing particle breakage ............................................................... 29

2

2.4.2. Particle breakage in DEM ............................................................................... 30

2.4.3. Breakage energy .............................................................................................. 32

2.5. Particle shape .......................................................................................................... 33

2.5.1. Shape measurement methods .......................................................................... 33

2.5.2. Shape factors/descriptors ................................................................................ 34

2.5.3. Particle shape in DEM .................................................................................... 35

3. RECYCLED CONSTRUCTION AND DEMOLITION MATERIALS .................... 37

3.1. Geotechnical characteristics .................................................................................... 38

3.1.1. Sample preparation ......................................................................................... 38

3.1.2. Particle size distribution .................................................................................. 39

3.1.3. Specific gravity ............................................................................................... 40

3.1.4. Flakiness Index ............................................................................................... 41

3.1.5. Optimum Moisture Content ............................................................................ 42

3.1.6. Unconfined Compressive Strength ................................................................. 42

3.2. Mineralogy and microstructure ............................................................................... 44

3.3. Particle shape .......................................................................................................... 48

3.3.1. Measurement methods .................................................................................... 48

3.3.1.1. Shape measurement of coarse grains .......................................................... 48

3.3.1.2. Shape measurement of fine grains .............................................................. 49

3.3.2. Analyses .......................................................................................................... 53

3.4. Summary ................................................................................................................. 56

4. EXPERIMENTAL METHODOLOGY AND ANALYSIS TECHNIQUES .............. 58

4.1. Single Particle Crushing .......................................................................................... 58

4.2. Particle Assembly Crushing .................................................................................... 59

4.3. Synchrotron tomography ......................................................................................... 60

4.3.1. Experiment Design .......................................................................................... 60

4.3.2. Synchrotron Radiation-based X-ray Micro-Computed Tomography ............. 62

4.3.2.1. Synchrotron source ..................................................................................... 62

3

4.3.2.2. Synchrotron light ......................................................................................... 63

4.3.2.3. Ruby detector .............................................................................................. 65

4.3.3. Experimental compression set-up ................................................................... 66

4.3.3.1. Sample chamber .......................................................................................... 67

4.3.3.2. Loading and data acquisition system .......................................................... 67

4.3.4. Image processing............................................................................................. 69

4.3.4.1. Density contrast ........................................................................................... 69

4.3.4.2. CT artefacts ................................................................................................. 70

4.3.4.2.1. Ring noise ................................................................................................... 70

4.3.4.2.2. Motion artefact ............................................................................................ 70

4.3.4.3. Segmentation ............................................................................................... 71

4.3.4.4. 3D reconstruction ........................................................................................ 74

4.4. Summary ................................................................................................................. 76

5. PARTICLE BREAKAGE ACROSS THE DIFFERENT SCALES ........................... 77

5.1. Single Particle Crushing .......................................................................................... 77

5.1.1. Qualitative analysis of fragmentation ............................................................. 77

5.1.2. Quantitative analysis of particle breakage ...................................................... 78

5.1.3. Modified particle tensile strength ................................................................... 81

5.2. Particle Assembly Crushing .................................................................................... 83

5.2.1. Post-breakage visual inspection of particles ................................................... 83

5.2.2. Particle shape and cushioning effect ............................................................... 87

5.3. Summary ................................................................................................................. 90

6. DISCRETE ELEMENT MODELLING OF PARTICLE BREAKAGE .................... 91

6.1. Principles of Discrete Element Modelling .............................................................. 91

6.1.1. Updating particle locations ............................................................................. 92

6.1.2. Contact models ................................................................................................ 93

6.1.2.1. Simple linear model .................................................................................... 93

6.1.2.2. Linear contact bond model .......................................................................... 95

4

6.1.2.3. Linear parallel bond model ......................................................................... 96

6.2. 2D modelling of particle breakage and effect of particle shape .............................. 97

6.2.1. Model calibration ............................................................................................ 97

6.2.2. Particle shape modelling ............................................................................... 100

6.2.3. Particle breakage modelling .......................................................................... 101

6.2.4. Simulation of biaxial tests ............................................................................. 102

6.2.5. Effect of particle shape on the macro-scale behaviour of the WR assembly 103

6.3. 3D modelling of particle breakage and effect of particle shape ............................ 103

6.3.1. Precise particle shape modelling ................................................................... 105

6.3.2. Calibration of micro-parameters ................................................................... 106

6.3.3. Internal stress distribution ............................................................................. 108

6.3.4. Breakage energy ............................................................................................ 110

6.4. Summary ............................................................................................................... 113

7. BREAKAGE AND PARTICLE CHARACTERISTICS EVOLUTION THROUGH

SYNCHROTRON TOMOGRAPHY ....................................................................................... 114

7.1. Crack propagation in different C&D granular materials ....................................... 115

7.1.1. Effect of shape .............................................................................................. 115

7.1.2. Effect of internal microstructure on crack patterns ....................................... 118

7.2. Evolution of grain property due to breakage ......................................................... 121

7.2.1. Soil grading and fractal distribution .............................................................. 122

7.2.2. Changes in external morphology .................................................................. 128

7.2.3. Universality of grain property evolution due to breakage............................. 129

7.3. Summary ............................................................................................................... 138

8. CONCLUSIONS AND RECOMMENDATIONS ................................................... 140

8.1. Major conclusions ................................................................................................. 140

8.1.1. Experimental observations and analyses from SPC and PAC tests .............. 140

8.1.2. Discrete Element Modelling ......................................................................... 141

8.1.3. Post-breakage analyses using synchrotron tomography................................ 141

5

8.2. Recommendations for future research .................................................................. 143

References ............................................................................................................................ 145

6

List of figures Figure 1.1. Research methods at a glance ................................................................................... 15

Figure 2.1. Distribution of Melbournian basalt (after Osborne et al. (2010)) ............................. 19

Figure 2.2. Common methods to study geomaterials at micro-scale .......................................... 23

Figure 2.3. Relative number of publications related to DEM simulation of different materials in the last decade ............................................................................................................................. 25

Figure 2.4. Classification of various experimental methods used in micro-scale studies of geomaterials (after Evans (2005)) ............................................................................................... 27

Figure 2.5. Breakage mechanisms (after Pitchumani et al. (2004)) ............................................ 29

Figure 2.6. Particle breakage with a breakage criterion (each particle with a coordination number smaller than 3 is allowed to break if σ > σmax) (Lobo-Guerrero, 2006) ......................... 32

Figure 2.7. 10-balls clump with eight small balls (asperities) bonded as a ballast particle (Lu and McDowell, 2008) ........................................................................................................................ 35

Figure 2.8. Simulation of semi-real-shaped particles with overlapped balls in each clump (Mollanouri Shamsi and Mirghasemi, 2012) .............................................................................. 36

Figure 3.1. C&D stockpiles at Alex Fraser Group Ltd. .............................................................. 37

Figure 3.2. C&D granular materials: a) WR, b) RCA, c) CB ..................................................... 38

Figure 3.3. Riffle splitter as a sample divider ............................................................................. 38

Figure 3.4. Cone and quartering method ..................................................................................... 39

Figure 3.5. Particle size distribution of C&D materials .............................................................. 40

Figure 3.6. Flakiness index sieves and gauge ............................................................................. 42

Figure 3.7. a) Modified compaction machine, b) Compaction curve for unbound WR ............. 43

Figure 3.8. USC results; a) WR sample after failure, b) Stress-strain curve .............................. 43

Figure 3.9. Sample preparation for SEM testing: a) Cold moulding, b) Grinded samples, c) Diamond polishing with Struers Tegramin-25, d) Gold coating by K975X Turbo-Pumped Thermal Evaporator, e) Gold coated specimens ......................................................................... 45

Figure 3.10. Zeiss Supra 40VP Scanning Electron Microscope ................................................. 46

Figure 3.11. SEM images of a) WR, b) RCA, and c) CB ........................................................... 46

Figure 3.12. EDS elemental analysis: (a) WR, (b) RCA, (c) CB ................................................ 47

Figure 3.13. Some example images of WR coarse grains: a) original images, b) binary images 49

Figure 3.14. Some example images of RCA coarse grains: a) original images, b) binary images .................................................................................................................................................... 50

Figure 3.15. Some example images of CB coarse grains: a) original images, b) binary images 51

Figure 3.16. CILAS 1190 Particle Size Analyser: a) schematic view of measurement with particle size analyser, b) Particle Size Analyser setup ................................................................ 51

Figure 3.17. Some example images of C&D fine grains: a) WR, b) RCA, c) CB ...................... 52

Figure 3.18. RCA fine grains: a) microscopic image, b) noise-free and segmented image ........ 53

Figure 3.19. Degree of sphericity distribution of RCA grains: a) coarse grains (>1.18mm), b) fine grains (<1.18mm); N is the number of studied grains ......................................................... 54

Figure 3.20. Degree of sphericity distribution of WR grains: a) coarse grains (>1.18mm), b) fine grains (<1.18mm); N is the number of studied grains ................................................................ 55

Figure 3.21. Degree of sphericity distribution of CB grains: a) coarse grains (>1.18mm), b) fine grains (<1.18mm); N is the number of studied grains ................................................................ 55

7

Figure 3.22. Degree of sphericity of WR grains in different shape categories; particle size fraction: 13.2 to 19 mm, (N: Number of grains. Circles, ‘o’, and asterisks, ‘’, are related to outliers) ....................................................................................................................................... 56

Figure 3.23. WR flaky particle.................................................................................................... 56

Figure 4.1. Single Particle Crushing: a) Schematic view, b) WR grain under loading by Geocomp LoadTrac II ................................................................................................................. 58

Figure 4.2. Particle Assembly Crushing setup ............................................................................ 59

Figure 4.3. Particle Assembly Crushing: (a) Grains before loading, (b) Test setup, (c) Grain crushing after loading ................................................................................................................. 60

Figure 4.4. Experiment design for synchrotron tomography experimets on samples with different particle sizes and under various loading levels ............................................................ 61

Figure 4.5. Maquette of synchrotron light machine in Australian Synchrotron ......................... 62

Figure 4.6. Beamline ................................................................................................................... 63

Figure 4.7. High intensity/brightness of synchrotron light compared to other types of light ..... 64

Figure 4.8. Ruby detector ............................................................................................................ 65

Figure 4.9. Schematic diagram of the experimental layout and the loading setup ..................... 66

Figure 4.10. Schematic view of the loading apparatus and sample chamber .............................. 68

Figure 4.11. Density contrast in a CT image .............................................................................. 69

Figure 4.12. Basaltic WR particle ............................................................................................... 69

Figure 4.13. CT image showing severe ring artefact .................................................................. 70

Figure 4.14. Main steps of image processing in order to obtain quantitative results .................. 72

Figure 4.15. Histogram-based binarization of the sand sample .................................................. 72

Figure 4.16. Image processing illustration: a) Original binary image, b) Image after median filtering, c) Image after Gaussian blur filtering .......................................................................... 73

Figure 4.17. Segmentation: a) Original image, b) Image after filtering, c) Watershed segmentation ............................................................................................................................... 73

Figure 4.18. Watershed segmentation basics: a) Greyscale image as a topographical surface in terms of intensity, b) Watershed transformation ......................................................................... 73

Figure 4.19. Comparison of classical watershed with Watershed Irregular Features ................. 74

Figure 4.20. a) Original and ‘segmented and labelled’ 2D slices, b) 3D reconstructed, segmented, and labelled WR sample at the initial condition (i.e. 0 MPa) .................................. 76

Figure 5.1. Single WR particle crushing: a) Initial state, b) Post-breakage, c) Schematic description of the one-dimensional compression and induced tension ....................................... 77

Figure 5.2. Single RCA particle crushing: a) Initial state, b) Post-breakage .............................. 78

Figure 5.3. Single CB particle crushing: a) Initial state, b) Post-breakage ................................. 78

Figure 5.4. Single Particle Crushing results: Load-displacement comparison of a) WR, b) RCA, and c) CB particles in different shape categories (bulky, solid line; elongated, dashed line; flaky, dotted line) .................................................................................................................................. 79

Figure 5.5. Yielding point range of different types of C&D particles with various shapes (N: Number of particles) ................................................................................................................... 80

Figure 5.6. SEM image of vesicular basaltic WR ....................................................................... 81

Figure 5.7. Particle Assembly Crushing: a) Initial state of bulky WR particles, b) Different layers of particles in the mould, c) Post-breakage state of particles including Representative Particle (RP) after 5 MPa vertical compression .......................................................................... 84

8

Figure 5.8. Particle Assembly Crushing: a) Initial state of bulky WR particles, b) Different layers of particles in the mould, c) Post-breakage state of particles including Representative Particle (RP) after 7.5 MPa vertical compression ....................................................................... 84

Figure 5.9. Particle Assembly Crushing: a) Initial state of bulky WR particles, b) Different layers of particles in the mould, c) Post-breakage state of particles including Representative Particle (RP) after 10 MPa vertical compression ........................................................................ 85

Figure 5.10. Particle Assembly Crushing: a) Initial state of elongated WR particles, b) Different layers of particles in the mould, c) Post-breakage state of particles including Representative Particle (RP) after 5 MPa vertical compression .......................................................................... 85

Figure 5.11. Particle Assembly Crushing: a) Initial state of elongated WR particles, b) Post-breakage state of particles including Representative Particle (RP) after 7.5 MPa vertical compression ................................................................................................................................ 85

Figure 5.12. Particle Assembly Crushing: a) Initial state of elongated WR particles, b) Different layers of particles in the mould, c) Post-breakage state of particles including Representative Particle (RP) after 10 MPa vertical compression ........................................................................ 86

Figure 5.13. Particle Assembly Crushing: a) Initial state of flaky WR particles, b) Post-breakage state of particles including Representative Particle (RP) after 2.5 MPa vertical compression ... 86

Figure 5.14. Particle Assembly Crushing: a) Initial state of flaky WR particles, b) Different layers of particles in the mould, c) Post-breakage state of particles including Representative Particle (RP) after 5 MPa vertical compression .......................................................................... 86

Figure 5.15. Particle Assembly Crushing results: a, b, and c) Load-displacement relationship of WR assemblies in different shape categories at different load levels ......................................... 89

Figure 5.16. Comparison of different WR particle shapes in PAC tests in terms of crushing strength and stiffness (K); (solid black line is drawn based on 100 periods moving average) ... 90

Figure 6.1. Simplified linear contact model ................................................................................ 94

Figure 6.2. Linear contact bond model: (a) normal force and (b) shear force versus relative displacement; Fi

c is bond strength, (after Cho et al. (2007)) ....................................................... 95

Figure 6.3. Illustration of parallel bond model provided in PFC; a parallel bond acts like a beam resisting moments as well (after Cundall (2004)) ....................................................................... 96

Figure 6.4. 2D synthetic specimen particle size distribution compared to the real material ....... 98

Figure 6.5. Different clump shapes used for 2D DEM modelling of Crushed Waste Rock, WR Specimen, and its three basic particle shapes............................................................................ 100

Figure 6.6. Cluster modelling; black parallel bonds are three times stronger than red parallel bonds (Ghazvinian, 2010) ......................................................................................................... 101

Figure 6.7. The proposed cluster model; black bonds are parallel bonds and yellow bonds are contact bonds ............................................................................................................................ 102

Figure 6.8. Numerical steps of the biaxial test simulation: a) Isotropic consolidation, b) Shearing phase ......................................................................................................................................... 102

Figure 6.9. Deviator stress versus vertical strain for crushed basaltic Waste Rock from the results of the triaxial test and simulation of biaxial tests with PFC2D ..................................... 104

Figure 6.10. Bond breakage simulation leading to grain crushing (black: parallel bonds, yellow: contact bonds) ........................................................................................................................... 104

Figure 6.11. Examples of different shapes of WR particles: (a) actual particles (top view), (b) 3D scans (side view) ................................................................................................................. 105

9

Figure 6.12. 1D compression on single bulky WR particles: (a) Laboratory and DEM results, (b) Initial state of the particle, (c) Force chain at the failure moment, (d) Total fragmentation at yielding point ............................................................................................................................ 106

Figure 6.13. (a) Laboratory and 3D DEM results of 1D compression on assemblies of bulky WR particles, (b) Contact force network distribution and bond breakage from DEM simulation of PAC test on bulky WR particles ............................................................................................... 109

Figure 6.14. Parallel bond state after loading (i.e. Max load 10kN or 5 MPa). Colours represent tensile (blue) and shear (green) bond breakages and bonded (red) ........................................... 110

Figure 6.15. Input energy versus energy dissipation through breakage based on DEM simulations of different assemblies of WR, (The amount of energy was calculated up to the onset of ‘representative particle’ breakage) .............................................................................. 112

Figure 6.16. Breakage energy per volume change of the WR sample versus applied force in relation to particle shape ........................................................................................................... 112

Figure 7.1. Crack propagation in different grains in a CB assembly: a) Density profile, b) The initial phase, c) After 5 kN compression ................................................................................... 116

Figure 7.2. Bending failure of elongated RCA grains: a) Initial state, b) After 10.2 MPa compression .............................................................................................................................. 117

Figure 7.3. Basaltic Crushed WR under a) 0, b)10, and c)20 MPa compression (asterisks, ‘*’, are highlighting bulky grains not experiencing severe breakage) ............................................. 117

Figure 7.4. SEM image of an agglomerated PCM grain ........................................................... 119

Figure 7.5. Fracture propagation in WR grains: a) Initial ortho-slice, b) After 10 MPa vertical

compression ( : tensile event; : shear event; : crack branching) ........................... 120

Figure 7.6. Porphyritic and vesicular texture of a WR grain .................................................... 120

Figure 7.7. Microstructural effect on grain tensile splitting: a) A close-up of the vesicular WR grains in an assembly, b) After 5 kN (10.2 MPa) compression ................................................ 121

Figure 7.8. Fractures following cleavage in WR grains............................................................ 121

Figure 7.9. Sand particles: a) CT ortho-slice, b) natural sand used in this research ................. 122

Figure 7.10. Changes in WR particle size distribution under different loading levels: a) Sample’s initial particle size: 0.425-4.75 mm, b) 0.425-1.18 mm, c) 1.18-2.36 mm, and d) 2.36-4.75 mm .................................................................................................................................... 124

Figure 7.11. Changes in RCA particle size distribution under different loading levels: a) Sample’s initial particle size: 0.425-4.75 mm, b) 0.425-1.18, c) 1.18-2.36, and d) 2.36-4.75 mm .................................................................................................................................................. 125

Figure 7.12. Changes in CB particle size distribution under different loading levels: a) Sample’s initial particle size: 0.425-4.75 mm, b) 0.425-1.18 mm, c) 1.18-2.36 mm, and d) 2.36-4.75 mm .................................................................................................................................................. 125

Figure 7.13. Sand specimen under different loading levels (0, 5, 10, and 20 MPa): a) Changes in grain size distribution, b) The fractal distribution ..................................................................... 126

Figure 7.14. Fractal distribution of WR samples: a) Sample’s initial particle size: 0.425-4.75 mm, b) 0.425-1.18 mm, c) 1.18-2.36 mm, and d) 2.36-4.75 mm ............................................. 126

Figure 7.15. Fractal distribution of RCA samples: a) Sample’s initial particle size: 0.425-4.75 mm, b) 0.425-1.18 mm, c) 1.18-2.36 mm, and d) 2.36-4.75 mm ............................................. 127

Figure 7.16. Fractal distribution of CB samples: a) Sample’s initial particle size: 0.425-4.75 mm, b) 0.425-1.18 mm, c) 1.18-2.36 mm, and d) 2.36-4.75 mm ............................................. 127

10

Figure 7.17. Morphology evolution of WR grains under different loading levels: Aspect Ratio and True sphericity distribution: a) Sample’s initial particle size: 0.425-4.75 mm, b) 0.425-1.18 mm, c) 1.18-2.36 mm, and d) 2.36-4.75 mm ............................................................................ 130

Figure 7.18. Morphology evolution of RCA grains under different loading levels: Aspect Ratio and True sphericity distribution: a) Sample’s initial particle size: 0.425-4.75 mm, b) 0.425-1.18 mm, c) 1.18-2.36 mm, and d) 2.36-4.75 mm ............................................................................ 131

Figure 7.19. Morphology evolution of CB grains under different loading levels: Aspect Ratio and True sphericity distribution: a) Sample’s initial particle size: 0.425-4.75 mm, b) 0.425-1.18 mm, c) 1.18-2.36 mm, and d) 2.36-4.75 mm ............................................................................ 132

Figure 7.20. Morphology evolution of sand grains under different loading levels: a) Aspect Ratio, b) True Sphericity ........................................................................................................... 133

Figure 7.21. Changes in mean values of AR and true sphericity due to breakage: a) WR, b) RCA, and c) CB ........................................................................................................................ 134

Figure 7.22. Relative Distribution Factor of AR and true sphericity distributions: a) WR, b) RCA, and c) CB ........................................................................................................................ 135

Figure 7.23. Skewness of AR distributions: a) WR, b) RCA, and c) CB ................................. 136

Figure 7.24. EDS elemental analysis of sand (insert: SEM image of a sand particle) .............. 137

Figure 7.25. Changes in statistical properties of shape factor distributions of sand fragments due to breakage: a) Mean values of AR and true sphericity, and b) Relative Distribution Factor .. 137

Figure 7.26. Schematic explanation of particle shape evolution due to breakage .................... 138

11

List of tables Table 2.1. A brief list of different topics on granular materials at micro-scale level.................. 23

Table 2.2. Recent DEM simulation of different laboratory tests ................................................ 25

Table 2.3. Recent work on DEM simulation of different materials ............................................ 25

Table 2.4. Factors affecting particle breakage ............................................................................ 31

Table 2.5. Definition of shape factors ......................................................................................... 34

Table 2.6. Representative ballast particles using in the DEM simulation (after Indraratna et al. (2010)) ........................................................................................................................................ 36

Table 3.1. Classification characteristics of C&D materials ....................................................... 40

Table 3.2. Specific gravity of C&D materials and natural sand ................................................. 41

Table 3.3. Flakiness index of C&D materials ............................................................................. 41

Table 3.4. Weight percentage of different elements in different kinds of C&D materials ......... 47

Table 6.1. Calibrated micro-parameters of crushed waste rock used in 2D simulations ............ 99

Table 6.2. Strength and modulus of elasticity of WR resulting from Unconfined Compressive Strength test; DEM simulations and laboratory tests ................................................................ 100

Table 6.3. The clumps’ shape factor used in 2D DEM simulations ......................................... 101

Table 6.4. Micro-parameters used in 3D DEM modelling of WR particles ............................. 107

12

1. INTRODUCTION

1.1. Problem statement and research significance

Recycled Construction and Demolition (C&D) materials are particulate waste

materials usually produced during construction and demolition of buildings and structures

or commercial and industrial activities. C&D materials have been recognised as having

suitable geotechnical properties to be reused as pavement subbase/base materials

commonly used in Victoria, Australia (Arulrajah et al., 2013a). Recycling and reusing of

waste materials leads to decreasing the demand for limited natural resources, and

simultaneously lowers the disposal landfill cost. Among different types of C&D

materials, crushed basaltic Waste Rock (WR), Recycled Concrete Aggregate (RCA), and

Crushed Brick (CB) are of interest in this study. Crushed WR used in this study originates

from surface excavation of Quaternary aged basaltic rock, which normally occurs near

the surface to the west and north of Melbourne, Australia (McAndrew and Marsden,

1973). RCA and CB are by-products of construction and demolition activities of buildings

and structures.

The response of particulate layers under traffic loading is normally characterised by

the resilient modulus test. However, the true nature of the deformation mechanism of

aggregates in pavement layers is still not fully understood (Leiva-Villacorta et al., 2017).

It has been accepted that the deformation of granular materials under loading is the

consequence of three major mechanisms: consolidation, particle rearrangement, and

particle breakage (Luong, 1982). The consolidation mechanism is the alteration in

compressibility of grain assemblies, while the particle rearrangement mechanism includes

sliding and rolling of particles. The breakage mechanism is the crushing that happens

when the applied load exceeds the strength of the particles. Crushing is a progressive

process that can initiate at relatively low stresses, change the soil fabric and packing

gradually, and cause serious issues, such as settlement and reduction in hydraulic

conductivity of the soil (Lekarp et al., 2000a, Lekarp et al., 2000b). In addition, the

engineering characteristics of a granular assemblage, including friction angle, shear

strength, and constitutive behaviour, have been proven to be dependent on properties

altered by breakage (Miao, 2015). In fact, particle breakage is a detrimental phenomenon,

13

particularly in granular pavement layers, and deserves to be fully understood in terms of

its causes, consequences, and the ways to avoid it.

Particle crushing is governed by particle size and shape, the applied stress level, and

mineralogy and microstructure of individual particles (Afshar et al., 2017). Considering

the inherent links between particle size and form/shape, it is reasonable to maintain that

wherever size matters, shape needs to be considered (Jia and Garboczi, 2016).

Nevertheless, a comprehensive review of the most recent literature reveals that studies of

the effect of shape on particle breakage are lagging behind. The most likely reason for

this shortcoming is the fact that particle shape is more difficult to measure compared to

other characteristics, such as particle size or size distribution (Cho et al., 2006).

Due to experimental limitations on measuring force chains, i.e. contact force

distribution, and monitoring crack propagation at particle scale, Discrete Element

Modelling (DEM) has been recently used in a number of studies to explore the

micromechanical breakage behaviour; however, most of them only provide information

on the evolution of the particle size distribution during breakage (Miao and Airey, 2013).

Some studies have utilised agglomerate-based models to study breakage (e.g. Bono et al.

(2014); Cil and Alshibli (2014)), but they failed to quantify the resulting particle shapes

formed.

In the present research, a new insight into the importance and effect of particle shape

on the degree of crushing of C&D materials is presented. Breakage of C&D particles,

individually and in granular assemblies, with different mineralogy and microstructure,

was investigated. A range of advanced laboratory and numerical techniques were applied

and developed to study, not only the breakage phenomenon at particle-scale, but also to

monitor and analyse post-breakage changes in the soil properties.

1.2. Objective and scope

This research contributes to the geotechnical industry and sustainability in

geotechnical engineering, particularly in pavement, road, and embankment applications.

The present research also facilitates a better understanding of particle mechanics and

strength properties of three major C&D materials, namely Waste Rock, Recycled

Concrete Aggregate, and Crushed Brick used in road and pavement constructions. The

14

focus of this study is specifically on the intimate relation between degree of crushing and

particle shape. The main objectives of this research are summarised as follows:

• Quantify the effect of particle-/micro-scale properties, such as

morphology and microstructure, on degree of crushing of different C&D granular

materials (i.e. basaltic crushed Waste Rock, Recycled Concrete Aggregate, and

Crushed Brick);

• Highlight the significant effect of particle shape on crushing strength of

grains by introducing a modified particle tensile strength as a function of particle

shape factor;

• Using three-dimensional Discrete Element Modelling, identify crack

propagation mechanisms by analysing the stress distribution and energy

dissipation in individual grains during single particle and particle assembly

compression; and

• Using three-dimensional Synchrotron Radiation-based Micro-Computed

Tomography (SR-µCT), identify the changes to grain properties of granular

assemblies, with different mineralogy, microstructure, size, and gradation, due to

breakage.

1.3. Research method

As the first step, the particle shape distribution of C&D materials was measured for

coarse and fine particles using image analysis techniques and laser-based microscopy.

The microstructure and mineralogy of each type of C&D materials were also determined

using Scanning Electron Microscopy (SEM) and Energy-Dispersive X-ray Spectroscopy

(EDS). Later, to investigate the effect of micro-scale properties, including textural,

microstructural, mineralogical, and morphological properties on the degree of crushing,

Single Particle Crushing (SPC) was conducted. Subsquently, Particle Assembly Crushing

(PAC) was designed and carried out on particle assemblies to take into account the effect

of coordination number. However, in-situ experimental characterisation of the evolution

of grain fracture within a granular system is difficult utilising conventional methods.

Consequently, Discrete Element Modelling was used as an indispensable tool to monitor

force chains developed in the sample, and also to measure energy dissipation due to

breakage.

15

Moreover, due to the experimental difficulty in examining post-breakage changes in

particle-scale properties of the soil, synchrotron tomography was used to conduct 4D

imaging (i.e. 3D monitoring over time). Synchrotron radiation-based tomography was

used in this research owing to its high flux density in contrast to medical or industrial X-

ray CT devices, enabling CT scanning with extremely high spatial resolution. To conduct

scanning during loading, a new loading apparatus capable of conducting compression

tests at the high stress on assemblies of grains was also designed and developed. After

obtaining CT images, post-processing of the images was conducted using a range of

image processing techniques, including local adaptive kriging, in order to visualise and

statistically analyse the changes in particle properties due to breakage in 3D. The research

methods are summarised in Figure 1.1.

Figure 1.1. Research methods at a glance

16

1.4. Thesis outline

This thesis contains eight chapters. A brief description of each chapter is outlined as

follows:

Chapter 1: Introduction

Chapter 1 is composed of an introduction to the present research and highlights the

problem statement and research significance, objective and scope, research method, and

the outline of the study.

Chapter 2: Literature review

In this chapter, after a brief introduction of C&D materials, an extensive review of the

recent research work on micro-scale studies of geomaterials is presented. Common

methods used to study granular materials at particle-scale, including experimental and

numerical methods, are also reviewed. Along the same line, studies on particle breakage

and its governing factors are reviewed. Particle shape measurement techniques and

descriptors used in past research work are also summarised in this chapter.

Chapter 3: Recycled Construction and Demolition materials

In this chapter, a variety of laboratory tests, carried out on C&D materials in order to

determine their geotechnical, mineralogical, microstructural, and morphological

characteristics, are described, and the results presented.

Chapter 4: Experimental methodology and analysis techniques

Different experimental designs conducted and methods used in this research are

explained in this chapter in terms of the design process, procedure, equipment, and

standard followed. After a brief explanation of single particle and particle assembly

crushing, the basics of synchrotron-based radiation tomography are presented. In the

current research, a loading apparatus was developed to perform constraint compression

tests during CT scanning. A number of technical challenges, and the ways of overcoming

these challenges, are discussed in relation to the chamber and loading system design. In

the final sections of this chapter, the image processing techniques applied for segmenting

and 3D reconstructing of CT images are described.

17

Chapter 5: Particle breakage across the different scales

The effect of particle shape on particle breakage at different scales for three different

types of recycled C&D materials was investigated using Single Particle Crushing and

Particle Assembly Crushing tests. The results assisted in proposing a new relationship

between particle tensile strength and particle shape. The results of this chapter have been

published in publication No.1 (see list of publications).

Chapter 6: Discrete Element Modelling of particle breakage

Chapter 6 is dedicated to DEM simulations of particle fracture and shape to study

fragmentation mechanisms and the effect of particle shape. SPC and PAC tests were

simulated using 3D DEM to track the in-situ evolution of force chains and stress

distribution, along with energy dissipation in the system. In this chapter, the principle of

DEM, calibration procedure, and validation results are also presented. The results of this

chapter have been published in publication No. 1, 5, and 6 (see list of publications).

Chapter 7: Breakage and particle characteristics evolution through synchrotron

tomography

The fast scanning and high-resolution 4D imaging were utilised to capture images from

the interior body of the granular assemblies during loading. The changes to grain

properties due to breakage, in particular the evolution of fractal distribution and particle

shape distribution, were investigated. The comprehensive statistical interpretation of

results is discussed in this chapter. The results of this chapter have been accepted for

publication (see publications No. 3 and 4 in the list of publications), and also submitted

as part of publication No. 2.

Chapter 8: Conclusion and recommendation

The findings are summarised in this chapter, along with the recommendations for

potential future research work.

18

2. LITERATURE REVIEW

2.1. Construction and Demolition materials

Construction and Demolition (C&D) materials are solid waste materials normally

collected near curbsides or generated by construction and demolition of buildings and

structures (SustainabilityVIC, 2010). Reusing and recycling of waste materials decrease

the demand for scarce virgin natural resources and simultaneously reduce disposal cost

into the landfills (Disfani et al., 2012). It has been proven that reuse and recycling of C&D

materials in pavement and road constructions are a sustainable option, lowering carbon

footprints in comparison with using traditional quarried materials (Arulrajah et al.,

2013b). In Australia, 3.3 million tons of crushed waste rock, nearly 8.7 million tons of

demolition concrete, and 1.3 million tons of demolition brick are stockpiled per annum

(Arulrajah et al., 2012b, Arulrajah et al., 2014). Recycling and subsequent reuse of C&D

materials provide enormous benefits in terms of waste disposal cost and environmental

impacts (Tam and Tam, 2007).

In this research, three main categories of C&D materials are studied, Waste Rock

(WR), Recycled Concrete Aggregate (RCA), and Crushed Brick (CB). Waste Rock used

in this study originates from surface excavation of basaltic rock, which normally occurs

near the surface to the west and north of Melbourne, Australia (Figure 2.1). Crushed

waste rock is normally produced during excavation for residential development or near

surface and subsurface infrastructure. In terms of mineralogical constituents,

Melbournian basalt primarily consists of pyroxene, olivine, and plagioclase, with rare

apatite, alkali feldspar, and glass. The rock is normally altered to some degree. This

alteration commonly appears in form of filled pores with clay minerals increasing the

density of the rock (Peck et al., 1992). Recycled Concrete Aggregate and Crushed Brick

are by-products of construction and demolition activities of buildings and structures

(Rahman et al., 2014). Concrete chunks from demolition of concrete structures are usually

crushed into aggregates of different sizes, and CB often includes impurities such as dry

mortar paste pieces. WR, RCA, and CB have been recognized as exhibiting geotechnical

properties equivalent or similar to typical quarry pavement subbase materials commonly

used in Victoria, Australia (Arulrajah et al., 2011, Arulrajah et al., 2012a).

19

Figure 2.1. Distribution of Melbournian basalt (after Osborne et al. (2010))

2.2. Necessity of micro-scale studies on C&D materials

To date, there has been some research on the use of C&D materials in pavement

application with the focus mainly on experimental laboratory scale research (Arulrajah et

al., 2014). Despite these efforts, there is still a level of uncertainty and knowledge gap in

the behaviour of C&D material which is one of the main obstacles in further usage of

Port Phillip Bay

Melbournian basalt

20

these waste materials in road infrastructures. Pavements are complicated structures and

normally consist of materials that differ in nature and properties. The concept of design

and methods of analysis of road and pavement structures have been significantly

developed in recent decades, such as material characterization, i.e. the shift from physical

to mechanical tests, and analysis tools, i.e. the development of various models based on

different constitutive models such as elasto-plastic and visco-plastic (Peng, 2014, Lekarp

et al., 2000a). Despite all these advances, the behaviour of granular assemblies in road

performance is still not fully understood. Some pavements and roads designed to a

specified service life demonstrate severe rutting only a few years of being subjected to an

appropriate design (Collop et al., 2006). Earlier, road studies were mostly focused on the

top surface (i.e. asphalt layer) since it exhibited failures such as rutting and cracking and

was easily accessible for maintenance or remediation studies. As a result, a variety of

asphalt mixes are currently available to solve several road problems. Nevertheless, the

persistence and severity of some distress types have led to the new assumption that

considers base, subbase, and subgrade layers as the main reasons for these issues (Lekarp

et al., 2000b). Consequently, some research during the last decade has focused on

improvement of characterization of unbound aggregate materials. The best example of

this progress is shifting from using the California Bearing Ratio test to the resilient

modulus test (Zeghal, 2004).

Geomaterials also show multi-scale behaviours that are associated with the

interactions of individual particles; from the pattern of force chains to the thickness of

shear bands and from laboratory samples to the full geotechnical engineering scope

(Shamy and Zeghal, 2007). It is evident that conventional laboratory tests are unable to

provide all the answers; hence, other methods have to been applied to complement them

(Zeghal and Edil, 2002). Owing to the discrete, heterogeneous, and anisotropic nature of

granular materials, Discrete Element Modelling (DEM) as a numerical approach emerges

for complementing conventional laboratory testing.

Terzaghi (1920) highlighted the importance of micro-scale studies: ‘[Coulomb]

purposely ignored the fact that sand consists of individual grains. Coulomb’s idea proved

very useful as a working hypothesis, but it developed into an obstacle against further

progress as soon as its hypothetical character came to be forgotten by Coulomb’s

successors. [. . .] The way out of the difficulty lies in dropping the old fundamental

21

principles and starting again from the elementary fact that sand consists of individual

grains’. Although discrete element method provide a powerful tool to study geomaterials

at particle-scale, but as Sibille et al. (2007) stated: ‘This has led to the paradox of

micromechanics of granular materials as a science based almost entirely on “virtual

evidence”’.

Recently, non-destructive testing methods have become popular in fields such as

material sciences and geomechanics to observe the interior microstructure of a sample

without penetrating its surface by physical means (Viggiani, 2013). Experimental, along

with numerical, access to particle-scale information facilitates our understanding of

complicated mechanisms as well as helps to explore new mechanisms happening across

the different scales, from micro to macro (Viggiani et al., 2015).

2.3. Micro-scale study of geomaterials

Most of geomaterials, from clays to sands and crushed rocks, have a microstructure

that is sometimes visible to the naked eye, such as sand grains, or is not visible, such as

clay particles. Nevertheless, conventional methods for estimating the mechanical

behaviour of geomaterials generally ignore their microstructure and assume them as a

continuum medium leading to offer a relatively simple framework. It cannot, however,

explain complicated phenomena where microstructure affects the macroscopic behaviour

of a material. As an exemplification, the initiation of a failure zone occurring under

certain conditions cannot be adequately described (Evans, 2005).

Global granular assembly response during loading is a necessary feature of interest for

studying particulate media. It is also essential to relate observed global stress-strain

response of granular materials with local force and displacement at micro-/particle-scale.

For continuum and isotropic materials, the stress-strain relation can be fully explained

macroscopically based on continuum mechanics (Narsilio et al., 2010). Nonetheless, for

a particulate assemblage, it is well-known that the stress-strain relation is complicated,

and is dependent on both the original state of the granular assembly (e.g. local and overall

porosity and particle coordination numbers) and the loading condition. A granular

assembly is an anisotropic, heterogeneous, and non-linear medium, particularly in terms

of the contact forces between particles (Majmudar and Behringer, 2005). The strain

experienced by a granular material stems from two fundamental mechanisms; firstly, the

22

relative motion such as rolling and sliding between grains, and secondly, distortion and

breakage of individual grains (Behringer et al., 1999, Penumadu et al., 2009). A number

of techniques have been developed to interpret the microscopic behaviour of granular

materials in terms of the interaction between particles. Figure 2.2 summarizes different

common methods which have been used to study micro-scale behaviour of geomaterials.

Moreover, various applications of these methods in geoscience are shown in Table 2.1.

2.3.1. Discrete Element Modelling

Discrete Element Modelling/Method (DEM), first introduced by Cundall and Strack

(1979), is a numerical modelling approach that can simulate granular materials taking

into account particle interactions. A “virtual” DEM-simulated test can be calibrated or

validated by comparing the macro-scale response observed in a real physical test with

model’s response. The detailed particle scale information provided in the DEM

simulation can then be utilised to enhance our understanding of the material behaviour

(Cheung and O’Sullivan, 2008).

Itasca PFC2D or PFC3D (Particle Flow Code) are commercial numerical codes based

on DEM used for analysing and testing of granular materials where the interaction of

several discrete objects causes large-strain and/or fracturing (Itasca Consulting Group,

2008). In modelling by PFC, it is necessary to calibrate the micro-parameters to match

the macro-response (Cho et al., 2007). Since this research uses PFC, some fundamental

assumptions and principles of this coding program are presented as follows (Itasca

Consulting Group, 2008, O'Sullivan, 2011):

• Particles are basically presented by rigid circular disks (2D) or rigid

spheres (3D).

• Particles can overlap at their contacts although these overlaps are

extremely small compared to the size of the particles.

• Contacts are characterized by a force-displacement law, and the contact

force is associated with the magnitude of the overlap.

PFC is based on an explicit numerical approach. The calculations made by the program

are based on the numerical integration of the Newton’s second law applied to every

particle and the force-displacement laws applied to every contact. A contact force has two

23

components: the normal and shear component. Thus, two different types of stiffness, the

normal stiffness and the shear stiffness, need to be specified. Broadly speaking, the

constitutive model for each contact includes three different models: the contact-stiffness

model, the slip model, and the bonding model (Kim et al., 2012).

Figure 2.2. Common methods to study geomaterials at micro-scale

Table 2.1. A brief list of different topics on granular materials at micro-scale level

Application Method Example references Level of compaction Numerical/

Experimental Otani et al. (2013)

Shear banding Numerical/ Experimental

Viggiani et al. (2010); Evans (2005)

The time behaviour of materials such as clay

Experimental Yigit and Cinicioglu (2013)

The effect of treatment Experimental Minder and Puzrin (2013) Bonding between grains Numerical Jiang et al. (2013) Shear wave propagation Numerical Ning and Evans (2013);

O’Donovan et al. (2012) Flow of granular

material Numerical Tomac and Gutierrez (2013)

Grain breakage Analytical/ Experimental/

Numerical

Cil and Alshibli (2014); Caicedo et al. (2013); Elghezal et al. (2013);

Hossain et al. (2007); Lobo-Guerrero (2006)

Imaging of stress in samples

Numerical/ Experimental

Lesniewska and Wood (2009); Mitra and Westman (2009)

Ice formation in soils Experimental Viggiani et al. (2015) Soil permeability Experimental/

Numerical Zheng and Tannant (2016); Kress

et al. (2012)

24

• ‘Contact-stiffness model’ is the relationship between force and magnitude

of the overlap at a contact. Linear and simplified Hertz-Mindlin models can be

used in PFC.

• ‘Slip model’ allows particles to slip by defining a maximum shear force at

the contact considering the contact friction coefficient μ.

• ‘Bonding model’ is like a glue between two particles. Two well-known

bonding models are the contact bond model and the parallel bond model.

Maximum allowable normal and shear forces should be specified in advance to

describe the strength of the bond (Lobo-Guerrero, 2006).

2.3.1.1. Laboratory test simulations

Laboratory tests (e.g. triaxial or direct shear tests) on granular specimens have

provided important information about materials’ responses. Discrete Element Modelling

is a powerful numerical tool in the study of granular mechanics, providing details of the

evolution of particle displacements, rotations, and interactions that cannot be readily

measured in the laboratory. The full potential of DEM can only be realised when the

results and findings of DEM simulations are related to the existing (macroscopic)

experimental data. In addition to physical laboratory tests, the particle scale information

from DEM simulations can be used to analyse the distribution of stresses within the

specimen or provide additional insight for experimentalists (Cheung and O’Sullivan,

2008).

Several researchers have tried to simulate different laboratory tests such as triaxial

loading, Unconfined Compressive Strength (UCS), cyclic loading, direct shear, bender

element, plane-strain compression, and Indirect Diametral Tensile strength (IDT). Some

examples of recent DEM simulation of laboratory tests are presented in Table 2.2.

2.3.1.2. Different materials

As shown in Table 2.3 and Figure 2.3, sand and asphalt have recently been of interest

of numerous researchers studying their behaviour using DEM. The most likely reason for

the recent focus on clean uniform sand particles may be the fact that there is less

complexity involved in the simulation of monodisperse sand particles.

25

Table 2.2. Recent DEM simulation of different laboratory tests

Tests Researchers and References Cyclic Loading Phusing and Suzuki (2015); Xin (2013); O’Donovan et al.

(2012); Sazzad and Suzuki (2010); Indraratna et al. (2010); Hossain et al. (2007); Zeghal (2004)

Direct Shear Khalili and Mahboubi (2013), Keppler et al. (2016); Indraratna et al. (2012); Salot et al. (2009); Shafipour and Soroush (2008)

Triaxial Loading Bono et al. (2014), Elghezal et al. (2013); Belheine et al. (2009); Lu and McDowell (2008); Cheung and O'Sullivan (2008)

Plane-Strain Compression Yan et al. (2009); Evans (2005) IDT Khanal et al. (2005); Thornton et al. (2004) UCS McDowell (2002)

Table 2.3. Recent work on DEM simulation of different materials

Materials Example researchers and references Asphalt Liu et al. (2012); Yu and Shen (2012); Kim et al. (2009);

Collop et al. (2006) Ballast Indraratna et al. (2012); Lu and McDowell (2008) Perlite Elghezal et al. (2013) Sand Zhao et al. (2017); Cil and Alshibli (2014); Obermayr et al.

(2013); Yan et al. (2009); Harireche and McDowell (2003) Sugar Particles Lobo-Guerrero (2006)

Steel Balls O'Sullivan and Cui (2009)

Figure 2.3. Relative number of publications related to DEM simulation of different materials in the last decade

26

2.3.2. Experimental methods

Non-destructive testing methods have recently become popular in fields such as

material sciences and geomechanics. Among different methods, wave-based techniques

such as X-ray tomography and contact measurement techniques such as particle image

velocimetry are emerging as powerful tools to study a wide range of materials in terms of

deformation and density. Contact measurement techniques are more conventional

approaches based on the use of transducers located at the specimen boundaries and are

more suitable for studying homogeneous specimens. In contrast, the development of

comprehensive models to explain non-homogeneous processes has historically been

hampered by a lack of access to the core of the material. X-ray Computed Tomography

(CT) is used to observe the interior microstructure of a sample without penetrating its

surface by physical means. 3D CT images offer rich information about the whole

specimen in contrast with point-wise data (Evans, 2005, Viggiani et al., 2004, Viggiani

and Hall, 2004). The recent advances in X-ray Micro-Computed Tomography (µCT),

with synchrotron sources and sensitive detectors, have provided a powerful tool to obtain

much finer spatial resolution of geomaterials, such as particle-scale characterization of

sand undertaken by Zhao et al. (2015) and Cil and Alshibli (2012).

Evans and Frost (2010) classified experimental methods, used for micro-scale studies

of geomaterials, into three main groups: external analysis, wave-based analysis, and

internal analysis (Figure 2.4). They stated that each of these techniques has its own merits

and drawbacks; however, external methods are arguably the most predominant. In

external methods, in other words contact measurement techniques, images are normally

captured at regular intervals during testing from the surface of the specimen, and particle

displacements next to the confining membrane are analysed.

Tomography has been defined as an ‘imaging technique which generates a cross-

sectional picture (a tomogram) of an object by utilizing the object’s response to the non-

destructive, probing energy of an external source’ (Gondrom et al., 1999). This technique

was first introduced by Radon in 1917, who claimed that the interior of a body can be

scanned by analysing energy which has attenuated from one boundary to another

(Herman, 1979). Ultrasonic waves are commonly utilised in laboratory-scale

tomographic studies whereas seismic waves are employed in field studies. Both methods

are commonly used in the analysis of brittle rock samples where micro-cracks are closed

27

during loading. This causes the elastic waves to travel at greater speed through rock. The

main difference between seismic and ultrasonic tomography is in the frequency ranges

utilised. Seismic tomography utilizes low frequency waves whereas in ultrasonic

tomography, the waves have smaller wavelengths. Consequently, seismic tomography in

contrast with ultrasonic tomography is more suitable to measure large-scale anomalies,

like fractures or high stressed zones, since the low frequencies correspond to long

wavelengths which cause greater penetration depth (Mitra and Westman, 2009).

Figure 2.4. Classification of various experimental methods used in micro-scale studies of geomaterials (after Evans (2005))

Among the different experimental methods available, X-ray tomography and Particle

Image Velocimetry (PIV), which can be used to study granular materials, are briefly

explained.

2.3.2.1. X-ray Tomography

CT systems are diverse, from laboratory scanners to synchrotron micro-tomographs.

They are mainly different in X-ray source and energy and detector geometrical

specifications. Full reviews in terms of principles of computed tomography are presented

in the work of Stock (1999), Ketcham and Carlson (2001), and Wildenschild et al. (2002).

X-ray CT is commonly not free of artefacts, which stem from obscuration of a major

feature or misreading of attenuation values. The magnitude of linear attenuation depends

on the chemical composition and density of the material, in addition to the X-ray energy.

These issues can be alleviated through the application of specific filters, precise detector

28

calibration, and sample centring. Moreover, errors can be compounded during CT

reconstruction. This is a common problem with laboratory scanners that use wide cone

beam geometry (Ikeda et al. (2000); (Rebuffel and Dinten, 2007)).

X-ray tomography was first used in the 1960s to measure 2D deformation in sand

samples (Viggiani et al., 2010). Later, X-ray tomography was used by Desrues et al.

(1996) and Alshibli et al. (2000). X-ray micro-Computed Tomography with synchrotron

sources or laboratory scanners has offered fine spatial resolution opening up new

possibilities for understanding and exploring the mechanics of granular materials (in 3D)

at grain-scale. As an example, Takahashi et al. (2004) presented CT images of sand grains

inside a shear band. Viggiani et al. (2010) also provided valuable 3D information on

localization patterns in sand, and showed the potential of X-ray tomography as a

measurement tool, especially for measuring the changes in the global and local void ratio

due to a shear band.

2.3.2.2. Particle Image Velocimetry (PIV)

Particle Image Velocimetry (PIV) is a velocity-measuring technique used for the

analysis of displacements in soil samples during different testing (White et al., 2003). PIV

is based on tracking the spatial changes in brightness within an image, which is divided

into a mesh of different PIV patches, by comparing sequential images. The procedure can

then be automated to extract the displacement data from sequential digital images,

captured during tests in plane-strain conditions (Meinhart et al., 1999).

The fundamental limitations of this method are the need for a transparent boundary;

for example, a glass or Perspex, and that the particulate material should have an adequate

texture that the image analysis software can consistently detect changes from one

photograph to another (Lesniewska and Wood, 2009).

2.4. Particle breakage in granular materials

Granular materials used in pavement structures, embankments, foundations, and even

rail track structures experience static and dynamic loading conditions. Consequently,

particle breakage in the shape of abrasion or asperity breakage and total fragmentation

may occur (Elghezal et al., 2013). Pitchumani et al. (2004) suggested two main

mechanisms for particle breakage: body mechanism and surface mechanism (Figure 2.5).

29

Brittle fracture is owing to the existence of infinitesimal flaws and is controlled by the

critical tensile stresses even if the applied load is compressive (Lawn, 1993). Jaeger

(1967) studied breakage of rock particles between two flat platens and showed that tensile

strength () of rock particles is a function of both the vertical force at failure (F) and the

diameter of the particle (d) when it is compressed diametrically:

2

Fd

Eq. (2.1)

Figure 2.5. Breakage mechanisms (after Pitchumani et al. (2004))

Single particle compression tests, in which a particle is compressed between two rigid

plates, are usually used to measure the strength of particles. Single particle compression

tests have been conducted on sand grains by Nakata et al. (1999), McDowell (2002),

Cavarretta et al. (2010), and Cil and Alshibli (2012). This test can also be used to calibrate

discrete element model of crushable particles. However, there are limited studies on

assemblies of grains where coordination number (i.e. number of neighbouring grains) also

plays a significant role in crushing strength of the grains.

Particle breakage causes various issues such as settlement or reduction in the hydraulic

conductivity of the granular material. Furthermore, the elastic properties and the shear

strength could also be adversely influenced (Lobo-Guerrero and Vallejo, 2005, Bono et

al., 2014). Coop et al. (2004) showed that granular sand samples experience a decrease in

the internal friction angle due to particle breakage before reaching a constant value of

residual strength.

2.4.1. Factors governing particle breakage

Among a number of factors affecting the degree of crushing, the inherent strength of

the particles and effective stress state have been reported as the most important

(Yamamuro and Lade, 1996). This is exemplified in the work undertaken by Indraratna

et al. (2014) who introduced an elastoplastic constitutive model to capture ballast

degradation under monotonic loading. Table 2.4 summarises factors affecting pattern and

the probability of particle breakage. Particle shape is one of the governing factors in the

30

particle breakage phenomena (Afshar et al., 2017). Jia and Garboczi (2016) stated that

particle shape is equally as crucial as particle size distribution in the characterization of

particulate media; however, its importance has been largely overlooked due to difficulties

in obtaining particle shape information. Santamarina and Cho (2004) summarised particle

shape irregularity into three main scales: sphericity, roundness, and smoothness, and

explained how angularity causes difficulty in particle rotation while roughness hinders

slippage.

Le Pen et al. (2013) and Sun et al. (2014) also investigated ballast particle shapes in

relation to particle sizes. To date, several studies have confirmed the notable effect of

particle shape on packing characteristics of granular materials (Cho et al., 2006, Gan et

al., 2004, Williams and Jia, 2003). However, so far studies on the effect of particle shape

on particle breakage are still limited.

2.4.2. Particle breakage in DEM

Particle breakage and fracture is a detrimental phenomenon that can only be fully

understood at the particle scale. Due to experimental limitations on measuring force

chains and monitoring crack propagation at this scale, DEM has been widely used in the

past few years. However, original DEM used circular/spherical balls to simulate particles

(Cundall and Strack, 1979). Later, rolling resistance was added at contact points between

balls to indirectly model angular particles (Iwashita and Oda, 1998). Nevertheless, the

rolling resistance behaves isotropically around the spheres; and cannot accurately

represent the behaviour of elongated particles (Ferellec and McDowell, 2010). Clustering

of balls, first introduced by Thomas and Bray (1999), brought about some improvement

in representing irregular crushable particles. In recent years, crushable and irregular

shaped particles have been simulated by replacing crushed particles with smaller

fragments, and the selection of the failure criterion (Figure 2.6) or the bond strength

between sub-particles has been based on inherent tensile characteristics of the material

(Lobo-Guerrero and Vallejo (2005); Harireche and McDowell (2003); Cheng et al.

(2003); Tsoungui et al. (1999); Åström and Herrmann (1998)).

31

Table 2.4. Factors affecting particle breakage

Factors Example studies Notes Material type/

Mineralogy Lobo-Guerrero and

Vallejo (2005), Bono et al. (2014)

Study of sugar and sand particles, respectively

Loading condition

Thakur (2011), Indraratna et al. (2014)

Ballast breakage under cyclic and monotonic loading was studied.

Particle size Rozenblat et al. (2011); Marsal (1975), Hardin

(1985)

• Breakage probability of different particles with different sizes was investigated.

• Breakage index was introduced with the focus on particle size distribution before and after loading.

Shape Antony et al. (2006); Golchert et al. (2004);

Tavares and King (1998)

• Influence of non-spherical particles during shearing was investigated using DEM.

• Study was conducted on glass ballotini micro-particles with NaCl binder forming a greater agglomerate; totally different from soil particles. The focus was on issues related to chemical and process engineering.

• Particle fracture under impact loading was investigated to understand comminution process in mineral processing field.

Coordination number

Cil and Alshibli (2012); Lobo-Guerrero and

Vallejo (2005); Mishra and Thornton

(2001);

• Studies on three well-rounded sand particles in a compression column, highly affected by boundary conditions

• Breakage criterion only applied to particles having a coordination number smaller than three

• Impact breakage of individual agglomerates

32

Figure 2.6. Particle breakage with a breakage criterion (each particle with a coordination number smaller than 3 is allowed to break if σ > σmax) (Lobo-Guerrero, 2006)

2.4.3. Breakage energy

The vital role of particle breakage in the macro-scale mechanical behaviour of

particulate media has been proven by various researchers such as Nakata et al. (1999), Cil

and Alshibli (2014). However, the quantification of the impact of particle crushing

particularly on stress-strain behaviour of a granular assembly still remains a challenge

(Wang and Yan, 2012, Coop et al., 2004). From the particle-scale point of view, the

difficulty is primarily from two aspects, the role of grain crushing and fragment

rearrangement and accordingly their roles in energy dissipation, which eventually create

a linkage between micro- and macro-scale response of a granular soil (Bolton et al., 2008).

In the well-known Griffith theory, the decrease in strain energy, while a crack is

propagating, is assumed to be equal to the rise in surface energy owing to the increase in

surface area (Griffith, 1921). Thus far, it is accepted that the term ‘breakage energy’ refers

to the energy dissipation due to the creation of new surfaces during fragmentation. Einav

(2007a) proposed a novel constitutive model based on a thermodynamics approach to

estimate energy dissipation due to breakage by considering changes in particle size

distribution.

DEM provides a unique tool to track and ‘quantitatively’ measure energy dissipation

due to breakage which is impossible to measure during conventional experimental tests.

For example, Antonyuk et al. (2006) and Wang et al. (2012) used DEM to monitor energy

distribution and dissipation mechanisms during impact breakage occurring in mills

(during mineral extraction). Khanal et al. (2005) also compared the input energy with new

surface generation and number of broken bonds during a single particle crushing test

using DEM. In this study, DEM was employed to partition energy components precisely,

particularly during breakage, in granular assemblies.

33

2.5. Particle shape

Broadly speaking, the behaviour of a particulate medium is governed by various

factors: stress-dependency and material-dependency. Particle size, degree of cementation,

grading, inherent anisotropy, and particle shape can be categorised as material-dependant

factors (Abbireddy and Clayton, 2015). Even though particle shape has been recognised

as a significant factor by many authors (e.g. Katagiri et al. (2010); Gan et al. (2004);

Nouguier-Lehon et al. (2003); Hansson and Svensson (2001)), due to the lack of practical

methods to measure particle shape, so far its effect, particularly on particle breakage, has

not been properly investigated.

Das (2007) stated that particle shapes are equally as important as the particle size

distribution; “… however, not much attention is paid to particle shape because it is more

difficult to measure”. In addition, it has been accepted that for non-spherical grains, no

definite particle size exists since it is highly affected by the particle form/shape (Jia and

Garboczi, 2016).

2.5.1. Shape measurement methods

Shape measurement can be performed on actual physical objects, photographs, and

micrographs. Further analyses can then be conducted by computers using different image

processing tools. 2D measurements are typically obtained from 2D projections of

particles. Digital images from 2D projections are commonly binarized to measure each

particle’s area or perimeter based on the number of solid pixels (Mora and Kwan, 2000).

2D images can also contain textural information when Scanning Electron Microscopy

(SEM) or Transmission Electron Microscopy is used. The laser diffraction method, first

proposed by Ma et al. (2000) in order to gain particle shape information, is also another

approach to obtain size information and shape factors based on 2D projection areas. In

diffraction methods, the distances between light blobs are normally used to measure

particle dimensions (Blott and Pye, 2006).

3D imaging with the aid of X-ray techniques is essential for characterising hard-to-

describe, i.e. irregular, particle shapes. In fact, a full 3D shape visualisation can be

conducted using CT images; however, extremely sophisticated separation/segmentation

methods are required in order to obtain realistic and accurate results (Cepuritis et al.,

2017b). Cepuritis et al. (2017a) stated that it is very difficult to analyse particles with

34

sizes of less than 10 µm via CT scanning, particularly when they are highly packed. CT

scanning can also be combined with X-ray fluorescence spectroscopy to investigate

materials containing different metallic components (Jia and Garboczi, 2016).

2.5.2. Shape factors/descriptors

Particle shape can be described using quantitative or qualitative factors/descriptors.

The formers are typically dimensionless factors linking particle lengths to particle volume

or surface area (Muszynski and Vitton, 2012). For many years, particle shape has been

characterised by comparing a limited number of particles with a reference chart, such as

visual estimation chart of Krumbein et al. (1949). In recent years, particle shape

information has become more accessible because of the advent of digital cameras and

image analysis techniques. Common shape factors are summarised in Table 2.5.

Table 2.5. Definition of shape factors

Name Formula Reference True sphericity Se/S0 Wadell (1935) Degree of sphericity 2(√𝐴/𝜋)/𝑑𝑐−𝑚𝑖𝑛 Wadell (1933) Circularity 4𝜋𝐴/𝑃2 IOS (2006) Sphericity Ri-max/Rc-min Krumbein et al. (1949), Cho et al.

(2006) Ellipseness Pe/P Le Pen et al. (2013) Aspect ratio D/d IOS (2006) Solidity A/Ac IOS (2006)

Note: Se: Surface area of the sphere with same volume as the particle, S0: Real surface area of the particle, A: Area, P: Perimeter, Ri-max: Radius of the maximum inscribed circle to the section A, Rc-min: Radius of the minimum circumscribed circle to the section A, D and d: the smallest diameter and the intermediate diameter orthogonal to each other, Pe: Perimeter of an ellipse having the same area as the projection of a particle, Ac: Convex area

Although 3D quantification of the bodies’ geometry is more precise and accurate, it

requires application of more advanced techniques. Cavarretta et al. (2009) after

comparing various 2D shape factors with true (3D) sphericity of grains showed that the

two-dimensional shape factor, ‘degree of sphericity’, proposed by Wadell (1933) is

sufficient for 3D characterization of particle sphericity. Nevertheless, the degree of

sphericity is not appropriate to describe platy or flaky shapes. Gantenbein et al. (2011)

suggested Aspect Ratio (AR) for quantifying shape of flaky particles, which is the ratio

of the laminar thickness to intermediate axis length.

35

2.5.3. Particle shape in DEM

A prime feature of particulate DEM is that the particles themselves are idealized. All

numerical models simplify the physical reality. In a particulate DEM simulation, the

particles’ geometries are typically disks (in 2D DEM simulations) or spheres (in 3D DEM

simulations). These particle shape idealisations are popular as it is relatively easy to

recognize whether the particles are in contact or almost touching; besides, the geometry

of the contact point, including the inter-particle contact overlap or separation, can easily

be calculated with a high level of accuracy. At every time increment in a DEM simulation,

every contact is considered individually, and the geometry of that contact point is

calculated. There will be many more contacts than particles in a DEM simulation, and

contact resolution is usually the most computationally expensive part of a DEM algorithm

(O'Sullivan, 2011).

The clump logic introduces another way to generate a group of attached particles that

act as a rigid body to achieve a more realistic particle shape. Clumped particles may have

overlaps to any extent. Contact forces are not created between these particles, so such a

deformable body is not capable of breaking apart regardless of the forces acting upon it.

Therefore, clumped particles are assumed to be a single slaved particle moving as a rigid

body. In this sense, a clump is totally different from a group of particles that are bonded

to one another, i.e. clustered particles (Cho et al., 2007). Lu and McDowell (2008)

compared simulation of ballast particles with different shapes under monotonic loading.

They concluded that the asperity breakage model is an efficient way to study the micro-

scale behaviour of railway ballast (Figure 2.7). Liu et al. (2015) also used four general

geometrically different clumps to model ballast particles, i.e. trapezoidal, triangular,

rectangular, and hexagonal.

Figure 2.7. 10-balls clump with eight small balls (asperities) bonded as a ballast particle

(Lu and McDowell, 2008)

The level of sophistication in the use of overlapping clumps has increased significantly

in recent years. Various researchers have proposed algorithms to create clumped particles

36

from digital images of real particles. For example, the algorithm used by Das et al. (2008)

and Mollanouri Shamsi and Mirghasemi (2012) is shown in Figure 2.8. Their results also

confirmed the fact that an accurate simulation of mechanical behaviour of a granular

assembly relies on the simulation of realistic grain shapes. However, their ‘clump

generation algorithm’ is not capable of simulation of particle breakage or fracture

initiation in a grain.

Indraratna et al. (2010) proposed a slightly different approach for the simulation of

ballast particles. Firstly, the ballast particles were divided into different sieve sizes, then

some representative ballast particles, i.e. three from each size fraction, of different shapes

were selected. The ballast particles’ geometries are finally filled with tangential circles

(Table 2.6).

Figure 2.8. Simulation of semi-real-shaped particles with overlapped balls in each clump

(Mollanouri Shamsi and Mirghasemi, 2012)

Table 2.6. Representative ballast particles using in the DEM simulation (after Indraratna et al. (2010))

Ballast particles

PFC particles

37

3. RECYCLED CONSTRUCTION AND DEMOLITION MATERIALS

The process of recycling Construction and Demolition (C&D) materials involves

initial crushing using jaw and cone crushers. Impurities such as steel, glass, and plaster

are then removed from the crushed materials. Afterwards, the materials are sieved in order

to sort the boulders into different sizes. Recycled materials used in this study were

collected from stockpile storages of various sites in Victoria, Australia, that are owned

and operated by Alex Fraser Group Ltd. (Figure 3.1). The ASTM (2014) standard for

sampling aggregates was adopted to collect representative samples from stockpiles, and

then the collected materials in plastic bags, each of which had a capacity of 20 kg, and

were transported to the Advanced Geotechnical Engineering Laboratory at Swinburne

University of Technology.

Among different kinds of C&D materials, Waste basaltic Rock (WR), Recycled

Concrete Aggregates (RCA), and Crushed Brick (CB) are studied in this project (Figure

3.2). In this chapter, geotechnical, along with microstructural, mineralogical, and

morphological, properties of the C&D materials are presented.

Figure 3.1. C&D stockpiles at Alex Fraser Group Ltd.

38

Figure 3.2. C&D granular materials: a) WR, b) RCA, c) CB

3.1. Geotechnical characteristics

3.1.1. Sample preparation

Riffle splitting and the cone and quartering method were utilized to prepare samples

for further testing. To control for bias in sample preparation, a riffle splitter, which has

an even number of riffles, was used to sub-sample the material (Figure 3.3). Before

splitting, the riffle was levelled and then the material was poured gradually into the

device. In addition, cone and quartering technique was used to reduce the sample size

without any systematic bias (Gerlach and Nocerino, 2003). The split material was piled

in the form of a cone, and after flattening the surface of the cone, it was quartered as

shown in Figure 3.4. Alternate quarters were mixed to make representative sub-samples.

The aforementioned procedure was practiced for the preparation of all WR, RCA, and

CB specimens used in this study.

Figure 3.3. Riffle splitter as a sample divider

39

Figure 3.4. Cone and quartering method

3.1.2. Particle size distribution

The Particle Size Distribution (PSD) of the C&D materials (> 0.075 mm) is determined

by sieve analysis. The sieve analysis tests were performed based on ASTM (2009b). Prior

to the analysis, the test materials were dried in the oven, 105ºC to 110ºC, for 24 hours,

then a minimum of 1.3 kg from each of C&D materials was selected to conduct the sieve

analysis. The analysis was carried out using a sieve shaker apparatus with the standard

shaking period of 10 to 20 minutes.

As demonstrated in Figure 3.5 and Table 3.1, C&D materials are coarse-grained

materials with a maximum particle size of 19 mm. The fine content (< 0.075 mm) is below

5% for all types of material; thus, the hydrometer test was not conducted on the materials.

Based on Table 3.1 and Unified Soil Classification System, the WR, RCA, and CB can

be classified as Well-Graded Gravels (GW).

40

Figure 3.5. Particle size distribution of C&D materials

Table 3.1. Classification characteristics of C&D materials

Characteristic/ Type WR RCA CB

Fine content (< 0.075 mm), (%) 3.19 2.56 1.34

Sand content (0.075-2.36 mm), (%) 23.53 37.46 34.97

Gravel content (> 2.36 mm), (%) 73.28 59.98 63.69

d50 (mm) 6.71 4.75 4.85

Coefficient of uniformity, Cu 18 28 23.33

Coefficient of curvature, Cc 2 0.8 1.9

3.1.3. Specific gravity

Specific gravity (Gs) is the specific gravity of the solids, which is the ratio of the

density of the solid particles relative to the density of water. The specific gravity of the

coarse and fine fractions of the C&D materials was calculated based on ASTM (2004)

and ASTM (2000), respectively. A minimum mass of 3 kg retained on the 4.75 mm sieve

was used to measure the specific gravity of coarse fraction of the C&D materials. Besides,

50 grams passing the 4.75 mm sieve was selected to calculate the particle density of the

fine fraction using a 250 mL pycnometer, along with a suction pump and vibration table,

as specified by ASTM (2000).

41

The average values of specific gravity of C&D materials are presented and compared

with natural sand particles in Table 3.2. The values presented are the average of at least

three test results and are also comparable to the ones reported by Arulrajah et al. (2013a).

Table 3.2. Specific gravity of C&D materials and natural sand

Material/Specific gravity (Gs) Coarse fraction Fine fraction Average

WR 2.85 2.75 2.80

RCA 2.68 2.64 2.66

CB 2.54 2.66 2.60

Sand (Das and Sobhan, 2013) - - 2.64-2.66

3.1.4. Flakiness Index

The existence of flaky/flat particles significantly affects the geotechnical

characteristics of a soil matrix particularly the packing density and internal friction angle

(Hansson and Svensson (2001), Disfani (2011), and Sivakugan et al. (2011)). An

aggregate is classified as flaky when its thickness, smallest dimension, is less than 60%

of its average dimension, where the average dimension is the average sieve size of the

size fraction into which the aggregate falls (BS, 2000). Based on BS 812-105.1(2000),

the Flakiness Index is a dimensionless value representing the percentage of flaky particles

in the material; the FI sieves and gauge are shown in Figure 3.6. As shown in Table 3.3,

the highest Flakiness Index of 26.32 % was recorded for the WR grains, while the

percentage of flaky particles in RCA and CB is much lower. The maximum FI of 35% is

acceptable for the use of crushed recycled materials in pavement construction

(SustainabilityVIC, 2010). Table 3.3 indicates that all kinds of C&D materials of interest

to this study meet the FI requirement for base and subbase applications.

Table 3.3. Flakiness index of C&D materials

Material Flakiness Index (FI), (%)

WR 26.32

RCA 15.20

CB 12.89

42

Figure 3.6. Flakiness index sieves and gauge

3.1.5. Optimum Moisture Content

Broadly speaking, compaction improves soil strength by making the soil fabric denser.

Approximately 95% of the Proctor maximum dry unit weight is a requirement for several

geotechnical infrastructures such as road pavements (Lekarp et al., 2000a). The

Maximum Dry Density (MDD) and Optimum Moisture Content (OMC) were obtained

for unbound WR using the modified dynamic compaction test based on ASTM

D1557(2012) (Figure 3.7a). Later, the measured MDD and OMC (Figure 3.7b) were

used as a basis for determining the degree of compaction of samples, particularly for UCS

testing.

3.1.6. Unconfined Compressive Strength

The static compaction method was initially used to prepare representative samples for

conducting Unconfined Compressive Strength (UCS) tests. A split compaction mould

with dimensions of 100200 mm (diameterheight) was used to compact the soil sample

in 8 layers. A constant pressure of 12.5 MPa was used to compact each layer at OMC,

which was previously measured from the modified compaction curve (Figure 3.7b). Gabr

and Cameron (2011) and Mohammadinia (2016) utilised a similar approach to prepare a

43

more ‘uniform’ sample for UCS tests. To minimise cracking along the layer interfaces

and increase interlocking between the layers, the surface of each layer was scarified

before adding the next layer.

Figure 3.7. a) Modified compaction machine, b) Compaction curve for unbound WR

UCS tests are commonly used to predict the strength and ultimately performance of

materials in different engineering applications. ASTM (2009a) was followed to conduct

UCS tests on WR samples. The axial load with a rate of 1 mm/min was continuously

applied to each sample, with a height-to-diameter ratio of 2.00. Attention was also paid

to ensure a smooth load application, without any sudden impacts to the top of sample.

The failure mechanism of a WR sample in the form of axial splitting is shown in Figure

3.8a. Figure 3.8b also shows that 780 MPa is WR’s strength under an unconfined

compression. The UCS results were later used to calibrate the 2D DEM model as

discussed in Chapter 6.

Figure 3.8. USC results; a) WR sample after failure, b) Stress-strain curve

44

3.2. Mineralogy and microstructure

To observe and identify different minerals and also the microstructure of the C&D

grains, Scanning Electron Microscopy (SEM) and Energy-Dispersive X-ray

Spectroscopy (EDS) were utilized. Prior to SEM and EDS testing, a specified sample

preparation procedure was practised. Firstly, a coarse grain, with the mean size of 13.2-

19 mm, from each of WR, RCA, and CB was selected and placed in a mould containing

liquid resin. The soaked grains were cured for 12 to 24 hours until resin set (Figure 3.9a).

Secondly, in order to flatten and smooth the sample surface, the samples were grinded

and polished using different sized SiC abrasive papers from coarse (i.e. 300 grit) to fine

(i.e. 1200 grit) (Figure 3.9b). Following the sawing and grinding operations, further

polishing with diamond paste was conducted in order to remove surface scratches.

Diamond paste is a water-based diamond suspension which enables further efficient

surface polishing and removes contaminants (Stutzman and Clifton, 1999). Diamond

polishing was carried out with Struers Tegramin-25 on polishing cloths with the grit size

ranging from 1 to 3 μm (Figure 3.9c). Once the samples were polished, the ultrasonic

cleaning process using ethanol was performed to clean contaminants which might adhere

to the sample surface. Finally, the samples were coated with a thin layer of gold alloy

(approximately 5 nm) to increase the electrical conductivity of the surface in order to

obtain high resolution images (Solanki and Zaman, 2012). Figure 3.9d shows the K975X

Turbo-Pumped Thermal Evaporator used to spatter gold layer on the specimens and the

coated samples.

After sample preparation, the Zeiss Supra 40VP Scanning Electron Microscope was

used to capture microscopic images from each material (Figure 3.10). A Scanning

Electron Microscope scatters a beam of electrons to scan samples. The electron beam

interacts with atoms in the specimen and produces a variety of signals containing

microstructural information of sample surfaces. Figure 3.11 demonstrates SEM images

of WR, RCA, and CB grains captured at a voltage level of 20 kV and a working distance

of 8 mm, which is the distance from the sample to the beam tip. As shown in Figure 3.11,

among different types of C&D materials, basaltic WR has a vesicular texture with a

relatively large interior void in its microstructure, while CB and RCA have more uniform

microstructures.

45

Figure 3.9. Sample preparation for SEM testing: a) Cold moulding, b) Grinded samples, c) Diamond polishing with Struers Tegramin-25, d) Gold coating by K975X Turbo-Pumped

Thermal Evaporator, e) Gold coated specimens

Energy-Dispersive X-ray Spectroscopy tests were also carried out for elemental

analysis of samples. When the incident beam causes electron migrations at the atomic

level, the energy released in the form of X-rays, is measured by the energy-dispersive

spectrometer. The energy emitted from the specimen reveals the atomic structure of the

specimen elements (Harding, 2002). Figure 3.12 and Table 3.4 show the percentage and

intensity of different elements in C&D grains. Referring to Table 3.4 and Figure 3.12,

clay Crushed Brick is mainly made of silica, alumina, iron dioxide, and lime, and the

chemical composition of Recycled Concrete Aggregate is mostly silica and calcium

alumino- ferrite. As expected, the basaltic Waste Rock consists of albite, olivine, and

pyroxene, and also a low titanium content was observed.

46

Figure 3.10. Zeiss Supra 40VP Scanning Electron Microscope

Figure 3.11. SEM images of a) WR, b) RCA, and c) CB

47

Table 3.4. Weight percentage of different elements in different kinds of C&D materials

Elements/Weight % WR RCA CB

Si 23.1 38.3 30.7

Al 8 8.3 9.1

Fe 6.6 5.4 4.6

Ca 5.2 4.1 4.9

Mg 3.8 0.0 0.0

Na 2.9 2.7 0.0

K 1.1 2.0 2.8

Ti 1.1 0.0 0.0

Figure 3.12. EDS elemental analysis: (a) WR, (b) RCA, (c) CB

48

3.3. Particle shape

As suggested by Cavarretta et al. (2009), the two-dimensional shape factor, ‘degree of

sphericity’, is sufficient for three-dimensional characterization of particle sphericity.

Accordingly, to determine the particle shape distribution of C&D materials, degree of

sphericity was measured for each material. However, the degree of sphericity is not

appropriate to describe the shape of flaky particles. Based on Table 3.3, the highest

Flakiness Index of 26.32 % was recorded for WR grains, while the percentage of flaky

particles in RCA and CB is negligible. Hence, prior to sphericity measurements, flaky

particles were separated from WR grains based on BS (2000) 812-105.1, and the Aspect

Ratio (AR) was measured to quantify the shape of the flaky WR particles.

3.3.1. Measurement methods

3.3.1.1. Shape measurement of coarse grains

In this study, to categorize particle shapes, several images from 1.3 kg of coarse grains

(i.e. >1.18 mm) from each of WR, RCA, and CB were taken. The value of 1.3 kg was

selected based on ASTM (2009b) D6913, the minimum mass requirement of specimens

for sieve analysis. The grains were spread out on white sheets at a reasonable distance

from one another. The most advanced DSLR Nikon camera with a 24.2 megapixel CMOS

sensor and EXPEED 4 image-processing engine was used to capture images from grains

on every sheet (Figure 3.13a, 3.14a, and 3.15a). The images were then imported to an

Image Processing and Analysis software program written in Java (ImageJ) to calculate

the degree of sphericity of particles. Firstly, the images were converted to binary images

using Otsu thresholding. Otsu is a thresholding method used to segment the image into

different homogenous components based on the grey-level histogram of the image

(Jianzhuang et al., 1991) (Figure 3.13b, 3.14b, and 3.15b). Secondly, noise and artefacts

in every image were corrected during image analyses by removing outliers. The number

of removed outliers were 2 to 10 in each dataset containing approximately 300 to 30000

data points. IBM SPSS Statistics 23 (Green and Salkind, 2010) was used for statistical

analyses.

49

Figure 3.13. Some example images of WR coarse grains: a) original images, b) binary images

3.3.1.2. Shape measurement of fine grains

In order to investigate the particle shape distribution of fine C&D grains (i.e. <1.18

mm), the CILAS 1190 Particle Size Analyser was utilized. The particle size analyser is a

laser-based microscope and is suitable to measure particle shapes and sizes ranging from

0.04 μm to 2,000 μm. A flow of moving particles is examined by CILAS 1190, i.e.

dynamic image analysis. Dynamic image analysis enables the investigation of a large

sample size with more randomly oriented particles and causes reduction in overlapping

particles (Altuhafi and Coop, 2011). CILAS dispersing unit brings about a well-dispersed

flow of grains passing through the scanning beam emission. An exposure time of less

than 1 ns is used to ensure that motion blur is negligible. CILAS 1190 also uses laser

diffraction and Charge-Coupled Device (CCD) cameras allowing measurement of

particles between 0.04 and 2,000 μm in different single shots (Cilas, 2008) (Figure 3.16).

The laser diffraction method can be employed to identify particle size and shape from the

light intensity distribution. The CILAS laser diffraction spectrometry is based on the

principle of Fraunhofer diffraction (Weiner, 1984). It should be noted that the

50

measurements of particle properties are more accurate when particles are large enough in

comparison with the wavelength of light. Mie-scattering patterns of single particles can

also be used to extend the method to lower particle size ranges (de Boer et al., 1987). The

Fraunhofer diffraction and Mie scattering theory were utilized to estimate the particle

shape distribution of fine C&D grains. The particle properties were measured using a real-

time Fast Fourier Transform of the images gained with the CCD camera equipped with

an image processing unit (Cilas, 2008) (Figure 3.16).

Figure 3.14. Some example images of RCA coarse grains: a) original images, b) binary images

51

Figure 3.15. Some example images of CB coarse grains: a) original images, b) binary images

Figure 3.16. CILAS 1190 Particle Size Analyser: a) schematic view of measurement with particle size analyser, b) Particle Size Analyser setup

52

A total of 50 grams from each of WR, RCA, and CB were selected, and 3 grams were

poured into the device in each attempt to capture the shape of the fine particle shape

(Figure 3.17). The images were then imported into ImageJ (Abràmoff et al., 2004) to

calculate the degree of sphericity of the particles. Noise in every image was again

corrected during the image analyses. Particles touching each other in an image were also

segmented in order to identify each individual particle in an image. Besides, particles

touching, the image border was removed from further statistical analyses since their full

shapes were not clear in the images. An example is shown in Figure 3.18. More details

on image processing and segmentation methods are provided in Chapter 4.

Figure 3.17. Some example images of C&D fine grains: a) WR, b) RCA, c) CB

53

Figure 3.18. RCA fine grains: a) microscopic image, b) noise-free and segmented image

3.3.2. Analyses

Figure 3.19 illustrates the ‘degree of sphericity’ distribution of 2,879 of RCA coarse

grains and 1,746 of RCA fine grains. The sphericity distribution of coarse RCA grains in

particular, here particle diameter > 1.18 mm, shows mainly two different modes of 0.70

and 0.84 which are related to elongated and bulky particles, respectively (Figure 3.19a).

The bimodal distribution of sphericity of coarse WR and CB grains is also evident in

Figure 3.20a and 3.21a. Figure 3.22 also shows the mean and median of the degree of

sphericity for elongated and bulky WR grains from the largest size fraction of 13.2-19

mm.

As shown in Figure 3.19b, 3.20b, and 3.21b, it is noticeable that the sphericity

distributions of fine C&D grains are slightly skewed to the left. Although a relatively

large number of particles with a sphericity of approximately 0.7 (i.e. elongated grains)

were observed (Figure 3.19b, 3.20b, and 3.21b), the percentage of bulky particles with

a larger value of sphericity is higher in the fine fraction of C&D materials compared to

the coarse fraction. The average degree of sphericity is 0.804, 0.81, and 0.82 for coarse

grains of RCA, WR, and CB, respectively (Figure 3.19a, 3.20a, and 3.21a). On the other

hand, the measured mean value of sphericity is 0.009-0.029 higher for the fine fraction of

each material (Figure 3.19b, 3.20b, and 3.21b). Comparison between the mean value of

sphericity of the coarse and fine fraction of each material confirms that the sphericity

increased slightly as the particle size decreased. Sun et al. (2014) also reported that the

roundness of ballast particles diminished slightly as particle size increased.

54

Moreover, the average Aspect Ratio of 0.13, which is the ratio of laminar thickness to

the intermediate axis length, was measured for WR flaky grains (Figure 3.23). In

summary, based on the average degree of sphericity (i.e. ϕ̅), AR, and FI presented in

Table 3.3, each material was divided into different particle shape categories as follows:

• WR: bulky (ϕ̅ = 0.84 or AR 1), elongated (ϕ ̅= 0.70 or AR 0.33), and

flaky (AR 0.13)

• RCA: bulky (ϕ ̅= 0.84 or AR 1) and elongated (ϕ̅ = 0.70 or AR 0.33)

• CB: bulky (𝜙 ̅= 0.84 or 𝐴𝑅 1) and elongated (�̅� = 0.70 or 𝐴𝑅 0.33)

Figure 3.19. Degree of sphericity distribution of RCA grains: a) coarse grains (>1.18mm), b) fine grains (<1.18mm); N is the number of studied grains

55

Figure 3.20. Degree of sphericity distribution of WR grains: a) coarse grains (>1.18mm), b) fine grains (<1.18mm); N is the number of studied grains

Figure 3.21. Degree of sphericity distribution of CB grains: a) coarse grains (>1.18mm), b) fine grains (<1.18mm); N is the number of studied grains

(a) (b)

Mean=0.819Std. Dev.=0.109

N=1,971

Mean=0.810Std. Dev.=0.090

N=3,075

(a) (b)

Mean=0.820Std. Dev.=0.088

N=3,367

Mean=0.830Std. Dev.=0.084

N=1,458

56

Figure 3.22. Degree of sphericity of WR grains in different shape categories; particle size fraction: 13.2 to 19 mm, (N: Number of grains. Circles, ‘o’, and asterisks, ‘’, are related to

outliers)

Figure 3.23. WR flaky particle

3.4. Summary

Different tests were conducted to determine geotechnical, mineralogical,

microstructural, and morphological characteristics of C&D materials. The results are

summarised as follows:

• C&D materials are coarse-grained granular materials with maximum

particle size of 19 mm.

• Results of SEM and EDS testing show that C&D materials have a variety

of mineralogical and microstructural characteristics, i.e. from basaltic vesicular

waste rock to clayey crushed brick.

57

• Shape measurement of a large number of fine and coarse particles of

different C&D materials was carried out showing a slight increase in degree of

sphericity as particle size decreased.

• Considering the particle shape distribution and Flakiness Index, particle

shape of each type of C&D materials can be classified into two or three main

categories of bulky (ϕ̅ = 0.84 or AR 1), elongated (ϕ ̅= 0.70 or AR 0.33), and

flaky (AR 0.13).

58

4. EXPERIMENTAL METHODOLOGY AND ANALYSIS TECHNIQUES

In order to investigate the effect of particle shape on particle breakage across the

different scales, Single Particle Crushing (SPC) and Particle Assembly Crushing (PAC)

tests were conducted. SPC and PAC tests were carried out on a variety of C&D grains in

the particle size fraction of 13.2-19 mm. Whereas the mean size (i.e. d50 or d30) of C&D

materials, which determines the governing particle-level forces and following macro-

scale behaviour, is between 0.425-4.75 mm. Hence, Synchrotron tomography was

performed on different C&D assemblies to further analyse particle breakage at a smaller

scale, which was impossible to achieve by using conventional laboratory tests.

4.1. Single Particle Crushing

Displacement-controlled single grain compression tests, in which individual particles

are compressed between two rigid plates, are often utilized to measure the strength of

granular materials. The individual C&D grains were subjected to vertical compression

between two flat platens until the induced horizontal tensile stress inside the grain caused

breakage (Figure 4.1a). The Geocomp LoadTrac II with a load/frame capacity of 22 kN

was used to perform the SPC tests (Figure 4.1b). The loading rate was set to 1 mm/min

between the two rigid loading plates.

Figure 4.1. Single Particle Crushing: a) Schematic view, b) WR grain under loading by Geocomp LoadTrac II

59

Single Particle Crushing (SPC) experiments were conducted on 21 single grains of

WR, 14 single grains of RCA, and 14 grains of CB (7 grains from each shape category).

To eliminate the effect of size on particle crushing, grains in the size range of 13.2 to 19

mm were selected. The materials’ largest size fraction was selected since larger particles

have a higher probability of breakage in a particulate assembly. The particles were placed

resting on their longest dimension (lowest potential energy).

4.2. Particle Assembly Crushing

In addition, in order to investigate the effect of particle shape at a larger scale, eight

one-dimensional compression experiments were conducted on assemblies of WR

particles for each shape category (i.e. bulky, elongated, or flaky). The WR particles, with

a size between 13.2 and 19 mm, were placed inside a cylindrical mould and were

vertically compressed at a constant displacement rate of 1 mm/min using Geocomp

LoadTrac II (Figure 4.2). In order to designate one particle as the ‘representative particle’

in the middle of the assembly, a cylindrical mould with a dimension of 50 mm diameter

and height was selected. Particles in the Particle Assembly Crushing (PAC) tests were

coloured prior to testing with spray paint based on their positions in the mould (i.e. close

to the top cap, middle, and close to the base). One particle was deliberately coloured green

in every test specimen (here called the ‘representative particle’) and placed in the middle

of the assembly so that the other particles completely surrounded it to minimize smooth

boundary effects on the representative particle, i.e. simulate in situ conditions (Figure

4.3).

Figure 4.2. Particle Assembly Crushing setup

60

Figure 4.3. Particle Assembly Crushing: (a) Grains before loading, (b) Test setup, (c) Grain crushing after loading

4.3. Synchrotron tomography

4.3.1. Experiment Design

Due to experimental difficulty in examining crack propagation at the particle-scale,

synchrotron tomography was used to conduct 4D imaging (i.e. 3D monitoring over time)

on 12 samples of the C&D materials. Assemblies of C&D grains in the size fractions of

0.425-1.18, 1.18-2.36, and 2.36-4.75 mm were studied under different monotonic loading

conditions, ranging from 5 to 15 kN (i.e. from 10 to 30 MPa). Moreover, to further

investigate the interaction of coarse and fine particles under loading and its consequent

impact on the overall crushing level, three specimens from a spectrum of each of the C&D

materials (i.e. 0.425 mm< particle size< 4.75 mm) were scanned. In addition, one natural

sand sample was scanned under different loading sequences for comparison purposes

(Figure 4.4). Referring to Figure 4.4, a maximum vertical load of 10 kN was applied to

samples in the size fraction of 1.18-2.36 and 2.36-4.75 mm, while assemblies with particle

sizes of 0.425-1.18 and 0.425-4.75 mm were compressed vertically up to 15 kN. Higher

load sequences were selected for assemblies containing smaller particles since smaller

grains normally exhibit higher tensile strength due to the statistical size effect. In total 36

CT scans were performed in this study, which provide a valuable data source for further

statistical and analytical analyses, which are presented in Chapter 7.

61

Figure 4.4. Experiment design for synchrotron tomography experimets on samples with different particle sizes and under various loading levels

62

4.3.2. Synchrotron Radiation-based X-ray Micro-Computed Tomography

In a synchrotron machine, electrons moving at velocities close to the speed of light,

are forced to change their direction under a strong magnetic field causing the electrons to

send out electromagnetic radiation (synchrotron light) (Cowie et al., 2010).

4.3.2.1. Synchrotron source

A synchrotron source consists of a number of main components (Figure 4.5). Firstly,

a barium cathode is heated to nearly 1000C in an electron gun in order to generate

electrons. Secondly, in an accelerator, the electrons are accelerated to 99.99% of the speed

of light. Thirdly, the electrons are transferred to a first ring, i.e. booster ring, where their

energy is increased/boosted using a radio frequency current of 3 GHz. Fourthly, electrons

circulate for about 30-40 hours in a second ring, i.e. storage ring. In this phase, the

synchrotron light is created by bending the path of the electrons using a powerful

magnetic field. Finally, the light is channelled from the ring to the pipelines, i.e.

beamlines, including different types of mirrors, filters, and optical components (Figure

4.6), so that it can ultimately be utilised for different research purposes (Boldeman and

Einfeld, 2004).

Figure 4.5. Maquette of synchrotron light machine in Australian Synchrotron

63

Figure 4.6. Beamline

4.3.2.2. Synchrotron light

Among the unique features of synchrotron light, tuneable, extremely intense, highly

collimated, polarised, pulsed, and non-destructive can be named. The synchrotron light

has a wide spectrum while a certain range of wavelengths can be set to be used for a

specific purpose. The synchrotron light is also highly intense, a million times brighter

than the sun (Figure 4.7). This brightness can reveal highly detailed information that is

impossible to capture by any other method. The synchrotron beam is also very focused

which allows investigating extremely small areas of a specimen. The emitted light is also

pulsed meaning small changes during a very short period of time, such as nanoseconds,

can be observed (Snigirev et al., 1995, Bilderback et al., 2005).

64

Figure 4.7. High intensity/brightness of synchrotron light compared to other types of light

Another benefit of synchrotron light is that it provides CT at much faster frame rates

compared to conventional facilities (Stevenson et al., 2010). All the high-resolution 4D

imaging tests (i.e. 3D monitoring over time) of this research were conducted in the

Australian Synchrotron, the Imaging and Medical Beam Line (IMBL), which provides

high-resolution, phase-contrast X-ray imaging. The cylindrical sample was placed

between the source and the detector to be scanned. The IMBL source is a multi-pole 4 T

wiggler with an optimal energy range of 20 - 120 keV (eV is Electron Volt) (Hausermann

et al., 2010). The optimal energy of 60 keV was used for imaging the granular C&D

materials with an exposure time of 400 msec. The exposure time was chosen based on the

X-ray attenuation of the sample and the power used. The experiment was conducted in

IMBL Hutch 3B whose application is for very high resolution large object 3D CT

scanning. The resolution (i.e. voxel size) depends on the distance between the sample and

the source, and its greatest lower bound was limited to 10 µm. There are always trade-

offs between sample size and resolution, likewise between scan time and process time.

Thus, the scans performed in this study were at the voxel size of 11.6 µm because of the

specimen size and desire to conduct fast scans, giving a scan time of only 10 minutes for

each scan.

65

4.3.2.3. Ruby detector

The vertical beam size Hutch 3B is approximately 25 mm, and scans were conducted

using the Ruby detector. Ruby is a custom designed IMBL detector. Its concept is based

on a photo-sensitive device coupled by a bright lens to a suitable X-ray sensitive

scintillator. The sensor is placed on a vertical motor-driven slide set within a light tight

enclosure. A mirror is utilized to observe a phosphor plate set perpendicular to the

direction of the beam. This permits protection of the sensor from direct and scattered

beam radiation using appropriate high-z materials. For the experiments presented in this

thesis, the sensor was equipped with a Nikon Micro-Nikkor 105 mm/f 2.8 macro lens

allowing the slide to be used as a zoom control (Figure 4.8). The scintillator was a 200

µm thick terbium doped gadolinium oxy-sulphide (Gadox, P43) screen with a thin

aluminium powder layer as an optical block (Hall et al., 2013, Hall, 2015).

Figure 4.8. Ruby detector

66

4.3.3. Experimental compression set-up

A loading apparatus was developed in order to apply one-dimensional compression to

the confined granular samples. The apparatus consists of two major parts: the sample

chamber and the loading and data acquisition system (Figure 4.9).

Wiggler Slits Filters Sample Detector

Load cell

Sample chamber(25 mm diameter)

Data logger

Distance from source to sample > 20 m

Sample to detector ~

0.2 m

X-ray

Figure 4.9. Schematic diagram of the experimental layout and the loading setup

67

4.3.3.1. Sample chamber

The design of the sample chamber involves consideration of a number of items. Firstly,

no barriers should be located between the sample chamber and the radiation path to avoid

blurring of the CT images. Secondly, the chamber must be made of radiolucent materials.

Thirdly, the chamber should resist high compressive stresses during testing. Finally,

efficient and quick removal and replacement of samples between tests are required.

Consequently, the sample chamber was made of a high-strength radiolucent tube of

aluminium with a thickness of 5 mm, 25 mm internal diameter and 25 mm height. The

height was selected based on the limited maximum vertical beam size. The thickness was

dictated by the linear attenuation coefficient of Al 6061 at a radiation energy of 60 keV

in order to be transparent to the beam, and also the required strength to resist up to 61.12

MPa compressive stress on the test samples. The chamber was bolted to the bearing frame

which is capable of resisting a load of 30 kN. Figure 4.10 shows a schematic view of the

loading apparatus and the test chamber. After placing the specimen in the chamber, the

chamber cap was bolted and then fixed onto the rotary base, since full rotation was

required for the sample to be scanned at different angular positions. To be able to place

the sample in the direct path of Synchrotron radiation, the sample chamber was located

in the bottom part of the apparatus.

4.3.3.2. Loading and data acquisition system

In addition to the limitation that radiation path to the sample should be clear, the

maximum mass capacity (70 kg) of the rotary unit was a second factor to be taken into

account. Accordingly, the use of a hydraulic jack or conventional stepping motor was

rejected in the early phase of the design due to their heavy weight and associated cables

causing disruption to the 360 rotation of the base. Hence, a reaction frame was used and

the force was manually exerted to the samples by screwing a threaded rod connected to

the load cell (Model SW, Tovey) (Figure 4.10), which allows sample scanning through

a full rotation without the need to displace or unload the system. A high strength thrust

ball bearing was used between the load cell and the threaded rod to avoid transferring a

moment to the loading piston. At the end of each test and upon completion of imaging,

after removal of the cap, the piston was pushed out to extrude the sample from the

chamber, allowing for post-test sample recovery. During each scan, a Linear Variable

68

Differential Transformer (LVDT) was used to measure deformation while the load cell

continuously recorded the force.

Due to manual application of load using a reaction frame and screw, an average of

0.50 kN load relaxation was observed after each scan, which normally took 15 min to

complete. Although 0.50 kN is negligible compared to the experimental loading

sequences (i.e. 5 kN, 10 kN, and 15 kN), utilization of a mini-actuator/stepping motor to

run load-controlled tests is recommended for future scanning in the situation that the

weight capacity of the rotary base and power supply to the rotary specimen can be

modified.

Figure 4.10. Schematic view of the loading apparatus and sample chamber

69

4.3.4. Image processing

The Multi-modal Australian ScienceS Imaging and Visualization Environment

(MASSIVE), fast data processing computers, were used to reconstruct high resolution

images from 1810 radiograms for each sample. The X-TRACT software was also used to

remove ring noises (which are discussed below) from 2159 reconstructed horizontal slices

of every sample.

4.3.4.1. Density contrast

Contrast within a CT image depends on differences in the density of particles. The

denser the particle, the more X-rays are attenuated. In fact, the greater the difference in

the density of the two phases causes the greater contrast between those phases in a CT

image (Figure 4.11). Low density phases, such as air, appear as black, while denser

materials are represented by brighter shades. Figure 4.12 shows a basaltic WR particle;

the denser pyroxene minerals can easily be distinguished from the lighter minerals such

as olivine and plagioclase.

Figure 4.11. Density contrast in a CT image

Figure 4.12. Basaltic WR particle

70

4.3.4.2. CT artefacts

Image artefacts are a discrepancy between the reconstructed values in an image and

the true attenuation coefficients of the sample. There are various categories of CT

artefacts; however, ring noises and motion artefacts are more likely to occur during each

CT scan.

4.3.4.2.1. Ring noise

Ring noises are commonly encountered in CT images and normally caused by a

miscalibrated detector (Figure 4.13). Calibration or occasionally replacement of the

detector is sufficient to reduce this kind of noise (Boas and Fleischmann, 2012).

4.3.4.2.2. Motion artefact

Motion in the sample during scanning causes blurring, as well as double imaging. A

fixed position sample is a simple means of preventing motion artefacts, which was the

procedure adopted in this research by fixing the test set-up to the rotary base.

Consequently, during a full rotation of the base, the sample did not wobble. Very rapid

scanning is another alternative technique to reduce motion artefacts (Hsieh, 2009).

Although calibrated detector and fixed samples can decrease the level of noise in CT

images, it is practically impossible to achieve perfect, noise-free images. In this research,

several filtering techniques were used to improve the image quality and image

segmentation.

Figure 4.13. CT image showing severe ring artefact

71

4.3.4.3. Segmentation

The main steps involved in image processing in this study are shown in Figure 4.14.

In order to perform statistical analysis on the CT images, each image must be segmented

into particles and the background (i.e. pores in this study). To do so, with the aid of

thresholding segmentation, particles were separated from the background and a binary

image was reproduced. A grey scale value is defined to separate regions based on analysis

of the image histogram (Iassonov et al., 2009). A histogram shape and the basis for

discretising the image are shown in Figure 4.15. However, as shown in Figure 4.14, the

image may still be noisy enough to prevent proper segmentation. In the next step, Median

filtering and Gaussian blur filtering were applied to reduce noise and achieve smooth

images. Median filter runs through every predefined radius of pixels and replaces them

with the median of intensity values of their neighbourhood, while the Gaussian blur filter

smooths the sharp edges using the Gaussian function (Gonzalez, 2009). Figure 4.16

illustrates the outcomes of the Gaussian blur and median filters with a radius of 7 pixels

on a piece of basaltic Waste Rock. In addition, since the basaltic crushed rock studied in

this research has a vesicular texture (Figure 4.17a), the internal grain pores were filled to

improve further segmentation (Figure 4.17b).

Despite denoising, individual fragments cannot be identified as long as they are in

contact with each other (Figure 4.17b); therefore, watershed segmentation technique was

further applied to resolve this issue as shown in Figure 4.17c. Watershed is a

segmentation algorithm that considers the image as a topographic surface (Figure 4.18a).

A watershed transformation segments the regions into catchment basins, while water is

assumed to collect into basins and separate the two basins from each other, as shown in

Figure 4.18b (Beucher and Meyer, 1992, Mangan and Whitaker, 1999). However, the

classical watershed normally leads to over-segmentation in an input image. In fact, the

classical watershed algorithm is more suited to segment connected, near circular

structures. Therefore, the Watershed Irregular Features plugin of Fiji, an image

processing package based on ImageJ, was used to separate, not only circular shaped

grains, but also irregular ones more efficiently by specifying two extra parameters, the

erosion cycle number and separator size.

72

1810 Radiograms

2159 Slices

Reconstruction Binary Images

Thresholding Segmented

Images

Filtering and Watershed Statistical

Files

Measurements

Figure 4.14. Main steps of image processing in order to obtain quantitative results

Figure 4.15. Histogram-based binarization of the sand sample

73

Figure 4.16. Image processing illustration: a) Original binary image, b) Image after median filtering, c) Image after Gaussian blur filtering

Figure 4.17. Segmentation: a) Original image, b) Image after filtering, c) Watershed segmentation

Figure 4.18. Watershed segmentation basics: a) Greyscale image as a topographical surface in terms of intensity, b) Watershed transformation

74

As shown in Figure 4.19, defining the appropriate erosion cycle numbers resolves the

issue of unwanted segmentation. Care should be taken since very high numbers of erosion

cycles will lead to the results being closer to the classical watershed (Schindelin et al., 2012).

The second parameter range (i.e. separator size) describes the length of the line separating

connected grains.

Figure 4.19. Comparison of classical watershed with Watershed Irregular Features

4.3.4.4. 3D reconstruction

As discussed earlier, locally adaptive thresholding was used to segment particles in the

images. This method uses a variable thresholding values based on local image characteristics

rather than using a single global value (Oh and Lindquist, 1999). To increase the accuracy of

the segmentation process, images went through different pre- and post-processing techniques

using erosion or dilation morphological operations (a summary of different techniques can

be found in Soille (2013)) to remove artefacts in the original image. Nevertheless, extensive

pre-processing or post-processing can cause reduction in segmentation quality by either

removing, not only noise, but also actual features, for example blurring of gray scale images,

75

or producing unwanted artefacts when using edge sharpening filters, and thus should be

monitored carefully and applied with care.

In this study, Avizo 9 (Westenberger, 2008), an image processing software package for

3D visualizing of CT images, is used to reconstruct each sample three-dimensionally. Local

adaptive kriging (developed by Oh and Lindquist (1999)) that utilizes both global and local

spatial information and Watershed Irregular Features were used to segment and then

accordingly label each particle or fragment (Figure 4.20). In the initial stage, two global

threshold values (V1 and V2) were chosen, either manually or with a global thresholding

approach. Voxels with grey scale values smaller than V1 and larger than V2 were defined as

the background and foreground, respectively. Later, all uncategorized voxels with grey

values within the range of V1 to V2 were determined as the background or foreground using

a kriging window centred on the unclassified voxel. According to Iassonov et al. (2009),

among available segmentation methods, best overall ‘3D’ segmentation quality can be

obtained with the kriging method; however, still this method requires adequate supervision

by a skilled operator. The segmented and labelled slices were stitched together in order to

visualize the whole 3D sample, where each particle or fragment was identified by a unique

identifier (Figure 4.20b).

Although different image enhancement techniques, such as pre-segmentation and post-

segmentation filtering, can dramatically alleviate noise problems in an image, every filtered

CT image still has some degree of artefacts (Kaestner et al., 2008). In this study, after image

processing, including filtering and segmentation, the resulting data files were processed

statistically using IBM SPSS Statistics 23 to remove outliers in terms of particle size and

shape factors (using ‘if’ commands).

76

Figure 4.20. a) Original and ‘segmented and labelled’ 2D slices, b) 3D reconstructed,

segmented, and labelled WR sample at the initial condition (i.e. 0 MPa)

4.4. Summary

To investigate the effect of particle shape on particle breakage Single Particle Crushing

(SPC) and in a larger scale Particle Assembly Crushing (PAC) tests were designed and

conducted. SPC and PAC tests were carried out on C&D coarse grains while the use of a

non-destructive technique was essential for further investigation of the C&D fine grains. A

novel constraint compression set-up was developed to conduct 4D imaging (i.e. 3D

monitoring over time) using Synchrotron tomography. Synchrotron light is a monochromatic

highly culminated X-ray source that produces a beam with a high flux and a specific energy

level. After CT scanning of the assemblies which were subjected to compression at different

loading intervals, a variety of image processing techniques were applied to reconstruct,

segment, and visualise the C&D samples in 3D. Finally, the data files, including size and

shape factors of each particle, or newly-generated fragment, were obtained for further

analyses.

(a) (b)

25 mm

77

5. PARTICLE BREAKAGE ACROSS THE DIFFERENT SCALES

Despite several efforts to characterize particle breakage (see Chapter 2), the dynamic

propagation and evolution of particle fractures, particularly in an assembly, had not been

studied extensively in the past. In this and subsequent chapters the in-situ evolution of particle

breakage and post-breakage particle characteristics are presented and discussed.

5.1. Single Particle Crushing

5.1.1. Qualitative analysis of fragmentation

Visual inspection of damage in individual particles provides valuable information about

the origin of fracture and the stress state (Quinn, 2007). Figure 5.1-5.3 show the initial and

post-breakage status of the C&D particles. Regardless of the particle shape or type, all

particles split mostly into two equal halves. Several minuscule fragments are also produced

during splitting due to asperity damage. It is evident that the major crack propagated in a

plane along the loading direction, confirming the tensile failure of a particle under diametrical

compression (Figure 5.1c). Visual inspection of single particle breakage under diametrical

compression match those observed in earlier studies on different materials, such as Cil and

Alshibli (2012) and Lobo-Guerrero and Vallejo (2005).

Figure 5.1. Single WR particle crushing: a) Initial state, b) Post-breakage, c) Schematic

description of the one-dimensional compression and induced tension

78

Figure 5.2. Single RCA particle crushing: a) Initial state, b) Post-breakage

Figure 5.3. Single CB particle crushing: a) Initial state, b) Post-breakage

5.1.2. Quantitative analysis of particle breakage

SPC tests on C&D particles from different shape categories are compared in Figure 5.4.

The brittle nature of the particles is evident in Figure 5.4. In addition, the highest yield point

79

(i.e. when a drastic load drop occurred) was observed in bulky particles of all of the different

kinds of C&D materials (Figure 5.4). Figure 5.5 compares the average yield point of the

different kinds of C&D particles in the different shape categories (i.e. bulky, elongated, and

flaky). The average yield point of bulky WR, RCA, and CB particles are 2.45, 3.32, and 2.90

times higher than the yielding point of the elongated ones, respectively, suggesting that the

bulky particles demonstrated the highest resistance against breakage among the other particle

shapes. This is in agreement with the observation of Tavares and King (1998) who conducted

single particle fracture tests under impact loading and reported that particle strength and

stiffness decreased as particles became more irregular.

Figure 5.4. Single Particle Crushing results: Load-displacement comparison of a) WR, b) RCA,

and c) CB particles in different shape categories (bulky, solid line; elongated, dashed line; flaky, dotted line)

0

1000

2000

3000

4000

5000

0 0.5 1 1.5

Load

: N

Vertical displacement: mm

Bulky Elongated Flaky

WR RCA

0

1000

2000

3000

4000

5000

0 0.5 1

Load

: N

Vertical displacement: mm

Bulky Elongated

0

1000

2000

3000

4000

5000

0 0.5 1 1.5

Load

: N

Vertical displacement: mm

Bulky Elongated

CB

(a) (b)

(c)

80

Figure 5.5. Yielding point range of different types of C&D particles with various shapes (N: Number of particles)

Interestingly, Figure 5.5 demonstrates that particles within the same shape category

fractured at approximately the same load range regardless of the material type (here WR, CB

or RCA). Aggregates with a wide range of mineralogy and microstructure were examined in

this study (see Chapter 3). As an example, basaltic WR particles have a unique microstructure

(i.e. vesicular texture) compared to the other types of C&D materials, which is shown in

Figure 5.6. Referring to results presented in Chapter 3 and Figure 5.6, it is evident that there

is a remarkable difference between the mineralogy and microstructure of WR, CB, and RCA

particles; however, particles falling in similar shape categories showed approximately similar

crushing strength. Hence, Figure 5.5 suggests that the influence of shape on C&D particle

breakage is more dominant than mineralogical or microstructural effects. Rozenblat et al.

(2011) also investigated the crushing strength of various individual particles. Interestingly,

81

even though their focus was entirely on the influence of particle size, their findings suggest

that marble, sugar, and salt particles with the same size range exhibited nearly equal crushing

strength in spite of the remarkable difference in their mineralogy.

Figure 5.6. SEM image of vesicular basaltic WR

5.1.3. Modified particle tensile strength

Despite recent interest in investigating breakage at the particle-scale, due to the versatile

characteristics of particulate media, there are still uncertainties, preventing the development

of a comprehensive description of the mechanical behaviour of particles during

fragmentation. Nonetheless, in general terms, particle breakage has been categorized in two

main forms: abrasion/asperity breakage and particle fracture/fragmentation (Aman et al.,

2010). Most detrimental impacts associated with breakage, such as settlements, are caused

by particle fragmentation. Particle fragmentation normally occurs when a grain is subjected

82

to a tensile stress higher than its tensile strength. To date, different theories, such as well-

known Rumpf’s model proposed for calculating tensile strength of granules. However,

Rumpf’s model is more appropriate for estimating the strength of bridges forming between

parent particles during a granulation process (Salman et al., 2006). Broadly speaking, the

nominal tensile strength of a non-granulated structure is usually defined as follows, when

geometrically similar structures are compared (Bažant, 1999):

NC FbD

Eq. (5.1)

Equation 5.1 is a more comprehensive form of Jaeger (1967)’s equation for measuring the

tensile strength of rock pieces under a quasi-static loading condition. The term D is the

structure thickness, but alternatively, can be defined as the separation between loading plates.

The term b is the dimension normal to D and can be chosen arbitrarily as it is not critical in

comparing geometrically similar structures. CN is an arbitrary coefficient which is typically

equated to unity. Based on the classical hypotheses, the nominal tensile strength is not

variable when geometrically similar structures made of similar substances are compared. In

other words, the ratio of the induced tensile force to the relevant area at yield is identical for

aggregates of the same material (Figure 5.1c). Nevertheless, in practice the tensile strength

of similar structures with different sizes are not identical due to the statistical size effect. The

size effect stems from variability in the tensile strength of the geomaterial due to random

internal flaws and weak zone distributions. This effect is normally expressed by the Weibull

modulus and measured from the particle survival probability versus tensile stress curve

(Jones and Ashby, 2005). However, the statistical size effect study is beyond the scope of

this thesis and is not related to the main discussion herein.

Particles in the same size fraction have a variety of shapes, from bulky to extremely flaky.

Therefore, usage of the mean particle diameter related to a sieve aperture leads to an

imprecise calculation of particle tensile strength. Since particles tend to rest on their longest

dimension (lowest potential energy), a particle thickness can be totally different from its

mean diameter measured from sieve analysis. Referring to Equation 5.1, b also must be

defined appropriately to obtain an accurate estimation of particle tensile strength due to the

83

discrepancy in particle shapes. Hence, to take the effect of particle shape into account,

Equation 5.2 is proposed to calculate the tensile strength of aggregates:

2

Fd AR

Eq. (5.2)

where AR is the Aspect Ratio and is equal to D/d (i.e. the ratio of laminar thickness to the

particle diameter normally measured in a sieve analysis). The Aspect Ratio is closely related

to the degree of sphericity, particularly for bulky and elongated particles. Furthermore, the

degree of sphericity is a 2D shape factor and simpler to measure than the aspect ratio for a

great number of particles (as measured for C&D materials in Chapter 3). Referring to Chapter

3, bulky C&D particles with a mean degree of sphericity from 0.80 to 0.84 had an AR close

to unity, while for elongated ones with a mean degree of sphericity from 0.70 to 0.74, the AR

was measured close to 0.33. Flaky particles usually have low aspect ratios, which were found

to be approximately 0.13 for WR particles in this study. To determine whether the percentage

of flaky particles in a sample is considerable, the Flakiness Index test, based on BS (2000)

812-105.1, can be conducted.

Moreover, one of the disadvantages of the sieve analysis is that it cannot measure the size

of individual particles and it is highly affected by the particle form (Fernlund, 1998). In

Equation 5.2, AR presents particle form with respect to both particle diameter and thickness,

giving a much more accurate particle tensile strength.

5.2. Particle Assembly Crushing

5.2.1. Post-breakage visual inspection of particles

PAC tests were performed to further study the effect of particle shape at a larger scale, in

an assembly, while the cushioning effect of other particles is considered. After compression,

the split mould was opened to examine the crushing level of the Representative Particle (RP),

here the green particle. Figure 5.7-5.14 demonstrate the initial and post-breakage state of

particles of different shapes in the assembly under different compression levels. The post-

breakage state of the particles shows that, regardless of the particle shape, the crushing level

is more dramatic at the upper part of the specimen, close to the piston, particularly at lower

84

loading levels, i.e. 5 and 10 kN or 2.5 and 5 MPa. Nevertheless, under higher loading levels,

fragmentation spread to other parts of the specimen, and severe fragmentation can be

observed even in the bottom half of the specimen (see orange particles in Figure 5.7-5.14).

This post-breakage observation is consistent with recent studies, such as Cil and Alshibli

(2014), indicating that less particle damage occurred in the bottom half of a mould containing

sand particles.

Figure 5.7. Particle Assembly Crushing: a) Initial state of bulky WR particles, b) Different layers of particles in the mould, c) Post-breakage state of particles including Representative

Particle (RP) after 5 MPa vertical compression

Figure 5.8. Particle Assembly Crushing: a) Initial state of bulky WR particles, b) Different layers of particles in the mould, c) Post-breakage state of particles including Representative Particle (RP)

after 7.5 MPa vertical compression

85

Figure 5.9. Particle Assembly Crushing: a) Initial state of bulky WR particles, b) Different

layers of particles in the mould, c) Post-breakage state of particles including Representative Particle (RP) after 10 MPa vertical compression

Figure 5.10. Particle Assembly Crushing: a) Initial state of elongated WR particles, b) Different layers of particles in the mould, c) Post-breakage state of particles including Representative

Particle (RP) after 5 MPa vertical compression

Figure 5.11. Particle Assembly Crushing: a) Initial state of elongated WR particles, b) Post-breakage state of particles including Representative Particle (RP) after 7.5 MPa vertical

compression

86

Figure 5.12. Particle Assembly Crushing: a) Initial state of elongated WR particles, b) Different layers of particles in the mould, c) Post-breakage state of particles including Representative

Particle (RP) after 10 MPa vertical compression

Figure 5.13. Particle Assembly Crushing: a) Initial state of flaky WR particles, b) Post-breakage state of particles including Representative Particle (RP) after 2.5 MPa vertical compression

Figure 5.14. Particle Assembly Crushing: a) Initial state of flaky WR particles, b) Different layers of particles in the mould, c) Post-breakage state of particles including Representative

Particle (RP) after 5 MPa vertical compression

87

5.2.2. Particle shape and cushioning effect

While single particle crushing provides valuable information about the crushing strength

of a grain, the cushioning effect of neighbouring particles plays a vital role in crack

propagation in an assembly (Tsoungui et al., 1999). The results from SPC experiments

demonstrate the significant effect of particle shape on breakage. However, normally, grains

are surrounded by other neighbouring particles in a particulate medium. The PAC

experiments were designed and conducted in this study to determine whether the effect of

particle shape remains significant at a larger scale when particle contact points (i.e. its

coordination number) increase in an assembly of particles.

Figure 5.15 shows PAC test results on bulky, elongated, and flaky particles of WR, which

also shows the ultimate state of RP after loading. No drastic drop can be seen in the load-

displacement curves (Figure 5.15), in contrast to the SPC tests (Figure 5.4). This is because

one-dimensional compression was applied to particles confined by the rigid lateral boundary,

and particle breakage progressed continuously until the predefined ultimate loading level was

reached. As compression continues, the number of minuscule particles produced during

fragmentation increased. Comparison between the results of PAC and SPC tests indicates

that single bulky WR particles fractured at a mean load level of 3.6 kN, but in the particle

assembly tests, RP only experienced asperity breakage at a load level of 10 kN. Single

elongated particles also split at an average load level of 1.5 kN, while in PAC, a major

fracture initiated through RP, surrounded by other particles, at a load level of 10 kN. These

results further confirm the cushioning effect of other particles on the Representative Particle

and suggest that when the coordination number surrounding a grain increases, the particle

resistance against breakage increases.

Remarkable differences between PAC test results of different particle shapes in terms of

particle strength and stiffness are shown in Figure 5.16. Based on different stages of particle

damage (Bolton et al., 2008), the representative bulky particle experienced asperity breakage

(Figure 5.16) whereas under a similar loading rate and level, the elongated particle split due

to internal tensile cracking. Interestingly, the flaky particle fractured completely under a

much lower load, i.e. 5 kN, as shown in Figure 5.16. The oscillatory form of the load-

88

displacement curves resulting from the PAC experiments stems from the resistance of newly-

generated fragments to the applied load after a sharp load drop (i.e. yielding). Besides, owing

to rearrangement of fragments during crushing and loading, some relatively low

gradient/plateau parts (i.e. increased displacement without increased force) can be seen in

load-displacement curves. Consequently, the stiffness of the assembly associated with pure

breakage was measured from certain parts of Figure 5.16, shown by the dashed black lines.

The findings indicate that the bulky assembly exhibited the highest stiffness while the

flaky assembly showed the lowest. The results further support the notion of the significant

influence of particle shape on particle breakage. In fact, particle shape played a crucial role

in particle breakage, not only when crushing of a single particle was examined, but also when

the cushioning effect of other particles (here in the same size, shape, and stiffness) in an

assembly was considered.

Moreover, referring to the fact that bulky particles showed higher resistance against

breakage, it can be concluded that less breakage would take place in an assembly containing

a high percentage of bulky particles compared to an assembly with a high Flakiness Index.

Thus, it is expected that the particle size distribution after loading would not be remarkably

different from the particle size distribution before loading in an assembly with a high

percentage of bulky particles. Further investigation into the effect of particle shape on

breakage, particularly in an assembly, is presented in Chapter 6 and 7.

89

Figure 5.15. Particle Assembly Crushing results: a, b, and c) Load-displacement relationship of WR assemblies in different shape categories at different load levels

kN

kN

kN

90

Figure 5.16. Comparison of different WR particle shapes in PAC tests in terms of crushing

strength and stiffness (K); (solid black line is drawn based on 100 periods moving average)

5.3. Summary

The investigation into a variety of Construction & Demolition materials demonstrated

that particle tensile strength is closely dependent on a particle’s shape factor. After studies

on several individual C&D particles, a modified particle tensile strength as a function of

particle Aspect Ratio is introduced, where the effect of particle shape/form is also taken

into account. In addition, it has been found that particle shape plays a more prominent

role in the particle breakage phenomena than mineralogy and the microstructure of C&D

particles. Further studies on particle fracture across the scale also showed that, although

boundary conditions and particle interactions have considerable effects on particle

breakage, the significant influence of particle shape on particle crushing is not diminished

even in an assembly of particles. Taken together, these findings suggest that brittle C&D

granular materials with a higher degree of sphericity (an Aspect Ratio closer to 1) and a

lower Flakiness Index would experience less particle breakage under loading.

kN

91

6. DISCRETE ELEMENT MODELLING OF PARTICLE BREAKAGE

Broadly speaking, conventional laboratory experimental tests often measure the

macro-scale behaviours of a specimen, such as the stress-strain behaviour normally

measured at the exterior boundary of a sample. The laboratory tests, designed and

conducted in this study (Chapter 5, SPC and PAC tests), provide valuable information

about the crushing strength of C&D particles. In addition, particle damage information

was obtained through post-breakage analyses of C&D particles by visually inspecting the

fractures in individual particles. However, particle breakage, controlled by

micromechanical properties of a material, needs to be analysed at a particle-scale.

Discrete Element Modelling (DEM) is an indispensable tool for studying granular

materials during loading. It provides a virtual laboratory to track and study not only

particle deformation and crack propagation but also the in-situ evolution of force chains

and stress distribution in a specimen, the latter is impossible to obtain by means of

laboratory tests (Katagiri et al., 2010).

The crushing strength of a particle under quasi-static loading can be estimated by using

Equations 5.1 or 5.2. Although the aforementioned equations are only an approximation

of the tensile strength of a brittle grain, where the plastic effects on breakage can be

neglected, it is one of the feasible ways to estimate the tensile strength of particles. Hence,

it is always helpful to include breakage energy calculations, along with particle tensile

strength, to gain a more accurate estimation of the material breakage, particularly when

an assembly of particles is studied. In this chapter, DEM was used to simulate the SPC

and PAC tests in order to gain a new insight into the internal stress and energy distribution

throughout the entire samples.

6.1. Principles of Discrete Element Modelling

Discrete Element Modelling is a computer simulation approach that can model

particulate media. The distinct advantage of DEM over other numerical modelling

approaches is that it considers the interaction of individual particles in a medium, in

contrast to a continuum model, such as the Finite Element Method, where the relative

displacements and rotations of particles are not taken into account.

92

The key assumptions made in a basic DEM simulation are listed as follows (Iwashita

and Oda (1999), Potyondy and Cundall (2004), and O'Sullivan (2011)):

• The particles are rigid and spherical (i.e. balls).

• The particles can translate and rotate independently.

• A DEM-based software can identify newly generated contacts between

particles.

• A contact occurs over an extremely small area between only two particles.

• Slight overlap is allowed at the contact points between particles, and it is

similar to the deformation occurring between actual particles.

• The compressive inter-particle forces are calculated from the overlap

magnitude.

• Tensile inter-particle forces are calculated by using the separation distance

between two adjacent particles. When the tensile force exceeds the tensile force

threshold, the contact is removed.

• The time step selected in a DEM model has to be adequately small that the

displacement of a particle in one time step is small enough to only affect its

immediate neighbouring particles.

In this study, Particle Flow Code (PFC 5) was utilised to conduct the simulations. PFC

is a DEM-based coding program which can be used to model cemented or unbounded

granular materials (Yoon, 2007).

6.1.1. Updating particle locations

The contact point and interactions are simulated by rigid springs in a particulate DEM

model. When particles move away from each other, the contact, in other words, the

related springs, are removed. Simultaneously new contacts are formed between particles

travelling toward each other and touching each other.

The principal translational and rotational equilibrium of a particle with mass m and the

moment of inertia I are (Zhu et al., 2007):

c ncma F F G Eq. (6.1)

I M Eq. (6.2)

93

where a and α are the acceleration and angular acceleration of the particle,

respectively, Fc are the contact forces, Fnc are the non-contact forces, G is the

gravitational force, and M is the moment. The sources of non-contact forces are often

capillary forces in an unsaturated soil which is outside the scope of the present research.

The acceleration of a particle can be calculated while the resultant forces acting on it are

known. In a DEM code, a time integration method, which is the same as the central-

difference approach, with a time step t is used:

/2 /2

1t t t ta V V

t

Eq. (6.3)

where Vt+t/2 and Vt-t/2 are the velocities of the particle at t+t/2 and t-t/2,

respectively (Rapaport, 2004). Accordingly, the velocity of the particle at time t+t/2 can

be calculated from Equation 6.3. Then, by knowing the velocity at time t+t/2, the

location of the particle x can be updated as (Munjiza, 2004):

/2t t t t tx x t V Eq. (6.4)

Likewise, the angular velocity is also used to update the position of the edges of an

irregular-shaped particle and to calculate the particle’s total rotation.

6.1.2. Contact models

A contact constitutive model typically consists of a combination of springs, sliders,

and dashpots. Generally, the main approach utilised in DEM is a penalty spring approach

which means the force is equated to the magnitude of overlap multiplied by the spring

stiffness, or as another option, the area/volume of the contact overlap is related to the

contact force (Peng, 2014). The following sub-sections describe the contact models used

in the present research.

6.1.2.1. Simple linear model

The simple linear model is the most fundamental contact model, providing elastic (but

no tension) and frictional behaviours along with viscous behaviour with a dashpot

component (Figure 6.1). This model cannot, however, resist relative rotations; thus, the

contact moment is equal to zero (Cundall, 2004). The force-displacement law for a linear

contact is as follows:

94

l dcF F F Eq. (6.5)

where Fl is the linear and Fd is the dashpot component of the contact force. Both forces

are expressed by normal and shear components, Fn and Fs:

ln n nF k and l

s s sF k , l ls nF F Eq. (6.6)

with 1 2

1 2n n

nn n

k kkk k

and 1 2

1 2s s

ss s

k kkk k

2d dn n i n nF m k and 2d d

s s i s sF m k (in the case of full shear) Eq. (6.7)

with 1 2

1 2i

m mmm m

where kn and ks refer to the linear normal and shear stiffness, respectively, µ is the

friction coefficient, and the terms n and s are the relative normal and shear displacement,

respectively. The parameters βn and βs are the normal and shear critical damping ratios,

and mi is the mass of body i. The parameters δnd and δs

d are the relative normal and shear

translational velocity, respectively. If the two balls, with masses m1 and m2, contacting

each other have different stiffness, then the contact normal and shear stiffness is a

combination of each ball’s stiffness, k1 and k2 (Cundall, 2004). The simple linear model

is suitable and efficient to be used where the deformation is elastic and relatively small.

Figure 6.1. Simplified linear contact model

95

6.1.2.2. Linear contact bond model

The linear contact bond model embeds both a linear repulsive part (identical to the

linear model) and a contact bond part, acting like a point of glue between two bonded

balls. The contact bond allows tensile and shear forces to develop at a contact point while

the tensile and shear force is limited by the predefined tensile and shear strength (Figure

6.2). If the normal or shear force, exceed the tensile or shear strength of the bond,

respectively, the bond breaks. In case of tensile bond breakage, the normal and shear

forces are equated to zero. However, in the case of bond failure due to shear, the contact

forces are not changed and if the normal force is compressive, the friction coefficient

multiplied by the normal force is used to update the slip state. If the bond breaks, the

contact is still active as long as the balls are touching, and the linear part is responsible

for the forces acting on the contact. However, if the bond breaks and the two engaged

balls move away enough from each other, the contact is eventually deleted by DEM

coding program, here PFC. Afterwards, if these balls again come close enough to each

other, a new contact is formed, and the contact type is assigned based on the

programmer’s specification (Potyondy et al., 1996).

The linear contact bond model does not resist moment either in bonded or unbonded

conditions. Later in 2004, the linear parallel bond model was introduced by Potyondy and

Cundall (2004) as a model allowing development of force and moment within the bond.

Figure 6.2. Linear contact bond model: (a) normal force and (b) shear force versus relative displacement; Fi

c is bond strength, (after Cho et al. (2007))

96

6.1.2.3. Linear parallel bond model

A parallel bond provides the physical behaviour of a cement-like material acting on

the finite area of a contact. It can also be assumed as a number of elastic springs

distributed over a cross-section centred at the contact point with constant normal and

shear bond stiffness (Figure 6.3).

Figure 6.3. Illustration of parallel bond model provided in PFC; a parallel bond acts like a beam resisting moments as well (after Cundall (2004))

The linear parallel bond model, similar to the linear contact bond model, embeds both

a linear repulsive part and a parallel bond part acting in parallel with the linear part

(Indraratna et al., 2011). A parallel bond also resists moment, and the increment of

twisting and bending moments (Mt and Mb ) are expressed as:

t s tM k J and b n bM k I Eq. (6.8)

where J and I are the polar moment of inertia and the moment of inertia of the parallel

bond cross-section. The parameters t and b are the twist and bend rotation

increments, respectively. The total force associated with the parallel bond is denoted by

Fi :and is given by ׳

n n nF k A and s s sF k A Eq. (6.9)

where kn׳ and ks

,refer to the linear normal and shear stiffness of the parallel bond ׳

respectively, and A is the cross-sectional area of the parallel bond (Cho et al., 2007). The

maximum tensile and shear stresses (max and max) applied to the bond are then calculated

as:

maxnF M R

A I

Eq. (6.10)

97

maxsF

A

Eq. (6.11)

where R׳ is the radius of the bond (Potyondy and Cundall, 2004).

A parallel bond breakage instantly causes stiffness reduction influencing not only the

adjacent balls’ stiffness but also the macro-scale stiffness of an assemblage. Hence, the

linear parallel model is known as a more realistic bond model for materials whereby the

bonds may break in tension or shearing resulting in a reduction in the macro-scale

stiffness of a specimen (Lee, 2007).

6.2. 2D modelling of particle breakage and effect of particle

shape

Two-dimensional DEM modelling, in this study by using PFC, is useful to undertake

initial model development, test ideas and examine phenomena before developing a

comprehensive 3D model. The decrease in computational time is notable in 2D

modelling, and one can obtain significant initial insights into physical processes that are

modelled in 2D. Besides, the interpretation of 3D results can be very time-consuming and

difficult (Cai, 2013). After gaining insights from 2D models, one can more readily

investigate the 3D counterparts. Therefore, in this study, first, 2D models have been

simulated, and then 3D simulations were conducted to thoroughly investigate the effect

of particle shape on the material behaviour, particularly on breakage and the crushing

strength.

6.2.1. Model calibration

PFC micro-parameters, unlike other geotechnical engineering codes or models, need

to be calibrated. The calibration of a PFC model requires adjusting the micro-parameters,

such as contact properties, the choice of the contact model, ball size, and so on (Hanley

et al., 2011).The macro-scale properties of the granular material are determined by

simulating laboratory tests. These macro-scale properties are equivalent to those that are

measured in the laboratory. In this study, the calibration was accomplished by simulating

uniaxial compression tests on rectangular samples (2D model) with the approximate

dimension of 100 × 200 mm, and then the results, in terms of deformability and strength

behaviours, were compared with those obtained from the laboratory test.

98

Due to the limited computing capacity and the large number of particles, the generation

of the synthetic specimen does not include the real particle size distribution. Thus, in

order to optimise the calculation time, a synthetic material with a simplified particle size

distribution (from 1 mm to 10 mm) was simulated (Figure 6.4). A similar approach was

followed by other researchers, such as Camusso and Barla (2009).

Figure 6.4. 2D synthetic specimen particle size distribution compared to the real material

The contact bond and parallel bond models were utilised in the 2D simulations. The

following parameters need to be calibrated for a contact bond model (Cundall, 2004):

• Ec: Particle-particle contact Young’s modulus

• kn/ks: Ratio of particle normal to shear stiffness

• μ: Particle friction coefficient (this applies when the contact bond has

broken)

• σc: Normal strength

• c: Shear strength

For a parallel bond model the following parameters need to be calibrated (Cundall,

2004):

• Ec: Particle-particle contact Young’s modulus

0

20

40

60

80

100

0.0 0.1 1.0 10.0 100.0

Tota

l pas

sin

g (%

)

Particle size (mm)

Crushed Waste Rock

Synthetic material

99

• kn/ks: Ratio of particle normal to shear stiffness

• μ: Particle friction coefficient

• λ׳: Radius multiplier used to set the parallel bond radii

• Ec Parallel bond Young’s modulus :׳

• knks/׳

Ratio of parallel bond normal to shear stiffness :׳

• c The parallel normal bond strength :׳

• c The parallel shear bond strength :׳

These input micro-parameters are normally unknown and can be determined by means

of a calibration process in which the behaviour of the simulated material is compared

with the relevant measured responses of the actual physical material in the laboratory

(Favier et al., 2010). The calibrated micro-parameters of the crushed basaltic Waste Rock

samples are listed in Table 6.1. More information about calibration of PFC micro-

parameters is provided in Section 6.3.2.

Table 6.1. Calibrated micro-parameters of crushed waste rock used in 2D simulations

Micro-parameters Value Ec 6 GPa kn/ks 2 µ 0.3 c 30 MPa c 60 MPa 1 Ec 6 GPa kn/ks 2 c 300 MPa c 300 MPa Rmin 1 mm Rmax 10 mm Sample porosity 0.21-0.22

Using the micro-parameters provided in Table 6.1, DEM simulation of the

Unconfined Compressive Strength (UCS) tests on WR resulted in strength values that

were in a good agreement with the actual laboratory results. Unconfined Compressive

Strength of WR from both numerical and experimental results (see Chapter 3 for the

experimental result) was between 700 to 800 kPa. Moreover, the modulus of elasticity of

the material calculated from UCS test results was also compared with the relevant DEM

100

results (Table 6.2). Since E is not constant, usually E50, the secant modulus at 50% of

peak strength, is utilized for numerical analyses (Lambe and Whitman, 2008). The

parameter E50 of the synthetic material is compared directly with the relevant measured

response of the physical material (Table 6.2).

Table 6.2. Strength and modulus of elasticity of WR resulting from Unconfined Compressive Strength test; DEM simulations and laboratory tests

Method UCS (kPa) E50 (kPa) DEM simulation (cluster model) 780 730

Laboratory tests (based on three attempts) 700 - 800 625-769

6.2.2. Particle shape modelling

Based on the true grain shapes of crushed Waste Rock, the particles were simplified

and categorised into three basic types: rectangular, circular, and triangular. Thakur (2011)

also used a slightly different method for simulating particle shapes of railway ballast

rather than using simple circular particles. Figure 6.5 shows the three basic shapes of

WR particles and the primary clump shapes of the proposed model. These primary clump

shapes were later changed to clusters as discussed in the following section. Table 6.3 also

demonstrates the degree of sphericity of the modelled WR particles during the 2D

simulations. The average degree of sphericity of the modelled clumps is 0.79, which is

quite close to the actual average degree of sphericity of WR grains, i.e. 0.81.

Figure 6.5. Different clump shapes used for 2D DEM modelling of Crushed Waste Rock, WR Specimen, and its three basic particle shapes

13 mm

101

Table 6.3. The clumps’ shape factor used in 2D DEM simulations

Shape

Degree of sphericity 0.80 0.56 1

6.2.3. Particle breakage modelling

An alternative to clumping is to replace the clumps with clusters that are not rigid and

can allow cracks to propagate through them. As an example, this method was used by

Ghazvinian (2010) who allocated parallel bonds between ball contacts in every clump

that was stronger than the existing parallel bonds between the individual clumps (Figure

6.6). In the current study, a subroutine was developed (using the Fish scripting language)

in PFC2D to replace clumps with clusters. Firstly, the Fish function searches through all

contacts and detects the contacts within and between the clumps. Secondly, the contact

bonds are installed between clumps and parallel bonds are applied between balls, which

form one individual clump. Then, the clumps are released and deleted to represent

breakable clusters (Figure 6.7). Since the adopted crushed Waste Rock was not stabilised

by any cement substances, the contact bond model was assigned between clumps.

Figure 6.6. Cluster modelling; black parallel bonds are three times stronger than red parallel bonds (Ghazvinian, 2010)

102

Figure 6.7. The proposed cluster model; black bonds are parallel bonds and yellow bonds are contact bonds

6.2.4. Simulation of biaxial tests

The simulation of biaxial test on WR sample was carried out in two steps (Figure 6.8).

The sample was isotropically consolidated and then sheared under a constant strain rate.

The specimen was confined and loaded by opposing walls during numerical biaxial tests.

While the top and base walls play the role of loading plates, the lateral walls apply a

constant confining stress by means of a numerical servomechanism controlling their

velocities. Since stress is a continuum quantity, it does not exist at each point in a 2D

discrete medium; therefore, an averaging procedure is required to calculate stress from

the contact forces. Hence, stress was computed by dividing the mean force acting on a

wall by the area of the corresponding sample cross section.

Figure 6.8. Numerical steps of the biaxial test simulation: a) Isotropic consolidation, b) Shearing phase

103

6.2.5. Effect of particle shape on the macro-scale behaviour of

the WR assembly

Figure 6.9 shows the stress-strain curve obtained using the DEM approach being in a

good agreement with the laboratory results. The experimental data, used to compare with

the DEM models in this section, has been reported previously by Arulrajah et al. (2012b).

Figure 6.9 also indicates the comparison of the clustered particle with the circular particle

assembly, in which the proposed cluster model is more accurate than the unbreakable and

simple shape (circular disks) model. Simulation of particles as disks and spheres clearly

leads to inaccurate results since the internal friction angle and shearing resistance of

spherical particles are less than the real values due to a lower resistance to rotation.

Moreover, the direction of the contact normal forces is always toward the centre in

spherical particles; as a result, no moment can be developed by the normal forces.

Therefore, the rotation is only affected by the contact tangential forces (Mollanouri

Shamsi and Mirghasemi, 2012). Apart from the unrealistic and over-idealised shape of

circular particles, the circular disk model did not represent particle breakage in the

specimen while degradation of particles, by corner breakage or splitting of particles into

two or more parts, was successfully simulated through the clustered particle model.

Figure 6.10 indicates the breakage and rearrangement of clusters in a part of the clustered

particle model. It is evident that the fragments of broken particles moved to the pores of

the assembly and consequently caused further deformation.

6.3. 3D modelling of particle breakage and effect of particle

shape

Initial 2D DEM simulations can help in testing ideas in a reasonable computational

time and also to estimate the approximate range of micro-parameters for a particular

material, accelerating the calibration process. However, it is clear that 3D DEM analyses

provide more realistic results in terms of physical responses of a material under loading.

For example, in a 2D DEM analysis, particles have fewer degrees of freedom, i.e. only

two translational and one rotational, compared to a 3D DEM analysis where a particle

has six degrees of freedom, i.e. three translational and three rotational (O'Sullivan and

Cui, 2009).

104

Figure 6.9. Deviator stress versus vertical strain for crushed basaltic Waste Rock from the results of the triaxial test and simulation of biaxial tests with PFC2D

Figure 6.10. Bond breakage simulation leading to grain crushing (black: parallel bonds, yellow: contact bonds)

In this section, DEM simulations of SPC and PAC tests on WR particles are presented

using Particle Flow Code (PFC3D 5.0). In addition, 3D DEM simulations were used to

measure breakage energy more accurately for WR samples. While the input energy can

be calculated from experimental results (the area under the load-displacement curve), it

contains not only breakage energy but also energy dissipation due to friction between the

particles and loading plates and during rearrangement of particle fragments. DEM was

105

used in this study to partition and track the energy dissipated through the creation of new

surfaces during fragmentation.

6.3.1. Precise particle shape modelling

Firstly, to simulate different particle shapes, WR particles (i.e.: bulky, AR=1;

elongated, AR=0.33; and flaky, AR=0.13) were scanned individually using a 3D laser

scanner (Figure 6.11). Subsequently, the particle shape geometries were imported into

PFC3D as a closed wall and filled with bonded smaller sub-particles (Figure 6.12b).

Hence, each WR particle was represented by an agglomerate built by bonding non-

uniformly-sized spherical sub-particles. Lim and McDowell (2007) stated that if the

computer is not able to handle large numbers of sub-particles efficiently and if the number

of sub-particles in the agglomerates falls below 500, the coordination number of the

agglomerates can be tracked and the bond strength scaled appropriately to compensate

for the influence of low coordination number. Accordingly, due to limited computational

capacity, the WR particle was simulated by an agglomerate of 239 bonded sub-particles.

Figure 6.11. Examples of different shapes of WR particles: (a) actual particles (top view), (b) 3D scans (side view)

AR1AR0.33

AR0.13

(a)

(b)

106

6.3.2. Calibration of micro-parameters

Cil and Alshibli (2012) showed that yielding of sand is directly controlled by the

crushing strength of individual grains. In the present study, the 3D model micro-

parameters were calibrated based on the single particle compression tests by using an

iterative calibration process.

A linear parallel bond model acting as a cementing material was used to bond sub-

particles at their contact points. Parallel bond normal stiffness is normally estimated using

the following expression:

1 2

cn

EkR R

Eq. (6.12)

where R1 and R2 are the radii of the bonded particles. It is necessary to calibrate and

adjust the micro-parameters, particularly contact model properties, in order to match the

macro-scale properties of the generated specimen with those measured in the laboratory

(Cho et al., 2007, Camusso and Barla, 2009).

Figure 6.12. 1D compression on single bulky WR particles: (a) Laboratory and DEM results, (b) Initial state of the particle, (c) Force chain at the failure moment, (d) Total

fragmentation at yielding point

107

In this study, PFC micro-parameters were determined by carrying out numerical SPC

tests on the generated agglomerate and comparing the results in terms of force-

displacement behaviour with those measured by laboratory tests (Figure 6.12a). All

bonds between sub-particles in every agglomerate have the same strength properties,

which were selected randomly from a normal distribution function based on the estimated

mean strength and standard deviation values listed in Table 6.4. To obtain this

distribution parameters, the standard deviation of material strength was initially assumed

to be equal to zero leading to mean bond strength values of approximately 500 and 300

MPa generating the highest and lowest tensile strengths measured in the SPC

experiments. Therefore, the bond strength is defined through a random number generator

using a mean bond strength of 400 MPa with a standard deviation of 100 MPa; the same

approach was also adopted by Cil and Alshibli (2012). Each agglomerate was compressed

between two flat platens at a constant loading rate up to fragmentation. The value of wall

stiffness was set at approximately one order higher than the particle stiffness to simulate

rigid steel walls (Coetzee, 2016).

Table 6.4. Micro-parameters used in 3D DEM modelling of WR particles

Parameter Value Wall stiffness: N/m 1×108 Mean diameter of agglomerates: mm 15 Spherical sub-particle properties Mass density: kg/m3 2800 Young’s modulus, Ec: GPa 8 Minimum radius, Rmin: mm 1 Rmax/Rmin 1.6 Friction coefficient, µ 0.5 Normal stiffness/shear stiffness 1.25 Parallel bond Young’s modulus, Ec

: GPa 8 Normal stiffness/shear stiffness 2 Mean normal strength: MPa 400 Mean shear strength: MPa 400 Normal and shear strength standard deviations:

MPa

100

As shown in Figure 6.12, the results of the DEM simulation of SPC testing on bulky

WR particles were, overall, in good agreement with the laboratory results. Moreover, in

order to simulate the PAC tests, a cluster template was created from the particle geometry,

and accordingly 13 breakable clusters were generated inside a cylindrical wall. A script

108

was developed to automate the aforementioned steps using Fish scripting language. The

previously calibrated micro-parameters (Table 6.4) were used to simulate the PAC tests.

To fill the numerical chamber with PAC simulations, clusters were generated at the upper

part of the chamber and then were forced into the chamber by applying only gravity.

Clusters were positioned without friction to optimise the ultimate porosity of the

specimen and the computational time.

6.3.3. Internal stress distribution

SPC and PAC experiments provided valuable information about particle crushing;

however, they do not provide sufficient information about the internal stress distribution

within the particles or the mechanism of cracks and whether they were shear or tensile

cracks. Therefore, DEM was used to estimate particle stress distribution leading to crack

propagation. Fragmentation was monitored by tracking bond breakage in the DEM

agglomerates. Figure 6.12c shows the development of force chains initiating the flaw

zone in an individual particle. Almost all cracks are concentrated in a plane along the

loading direction which is in accord with the tensile failure theory of particles under

compression. Figure 6.13a also provides the DEM result of PAC tests on bulky WR

particles which agrees well with the experimental results in terms of the force-

displacement curve. During compressive loading (Figure 6.13b), the evolution of cracks

and the rapid increase in bond breakages causes the formation of new force chains which

resist the axial load. After a 10 kN compressive loading, comparison of crack distribution

in bulky, elongated, and flaky particle assemblies also revealed that the flaky assembly

had the highest bond breakage percentage (Figure 6.14). Referring to Figure 6.14, the

crack mechanism is comparable to that found in single particle crushing, since tensile

cracks were the dominant type in comparison with shear cracks.

109

Figure 6.13. (a) Laboratory and 3D DEM results of 1D compression on assemblies of bulky

WR particles, (b) Contact force network distribution and bond breakage from DEM simulation of PAC test on bulky WR particles

kN

110

Figure 6.14. Parallel bond state after loading (i.e. Max load 10kN or 5 MPa). Colours represent tensile (blue) and shear (green) bond breakages and bonded (red)

6.3.4. Breakage energy

During a DEM simulation, the input energies (i.e. boundary forces or body forces)

dissipate in frictional sliding and the rupture of contact bonds. There is also an incessant

conversion of strain energy to kinetic energy and vice versa in the contact springs (Bardet,

1998). Referring to O'Sullivan and Bray (2004), the requirement for energy (E) balance

in a numerical system is:

Kinetic Internal InputE E E Eq. (6.13)

where EInternal consists of the strain energy stored in the contact springs and the

frictional energy dissipated during loading. Thus, to obtain an accurate estimation of

breakage energy, each energy component needs to be known. Research into breakage

energy in particulate media has a long history from the earliest work, such as research by

Rittinger (1867), to the most recent by Russell and Einav (2013). Thus far, it is accepted

that the term breakage energy refers to the energy dissipation due to the creation of new

surfaces during fragmentation. Breakage energy is normally measured based on the input

energy especially in single particle crushing experiments and is assumed to be equal to

the work calculated from a load versus displacement curve, which is exemplified in the

work undertaken by Zhao et al. (2015). Nonetheless, the major drawback of this

assumption is that it dramatically overestimates the breakage energy. The input energy

applied to a particulate medium is dissipated through, not only the fracture process, but

also other mechanisms, particularly frictional/slip dissipation between particles and

loading plates and also between newly generated fragments. The frictional dissipation

111

may be neglected in single particle crushing, but it certainly needs to be considered at a

larger scale, particulate assembly, where particles rearrange and rub against each other

during loading. DEM provides a powerful tool to track and partition energy dissipation

and its related mechanisms which are otherwise impossible to measure with experimental

tests. Antonyuk et al. (2006) and Wang et al. (2012) used DEM to monitor energy

distribution and dissipation mechanisms during impact breakage, while Khanal et al.

(2005) compared the input energy with new surface generation and the number of broken

bonds during a single particle crushing test. In this study, DEM was used as a virtual

laboratory to measure slip energy dissipated during PAC tests on WR samples.

Fundamentally, slip energy is updated after each time-step as follows:

12

with

s s so

s s os s

s

E F F

F Fk

Eq. (6.14)

where Eµ is the increment of slip energy, ½((Fs)o+Fs) is the average linear shear force

occurring during the time-step, sµ is the slip component of the adjusted relative shear

displacement increment. In this study, the linear contact bond model and the parallel bond

model were also utilized (more information is given by Cundall and Strack (1979) and

Potyondy and Cundall (2004)). Eventually, slip dissipation was subtracted from the total

work done by the load, to gain an accurate estimation of the breakage energy. Figure

6.15 shows the dramatic difference between total energy and pure breakage energy. The

total energy is 3.45 times higher than the breakage energy. Only 29% of the total energy

was dissipated through the creation of new surfaces during fragmentation, and the rest

was dissipated through other mechanisms, mostly frictional dissipation between particles

and newly generated fragments. The results are consistent with those obtained by Wu et

al. (2005), who reported that frictional energy dissipation forms a major portion of the

total energy dissipation during particle crushing under impact loading. Wang and Yan

(2012) also simulated a triaxial shear test on crushable soil particles using DEM and

pointed out that the major role of particle breakage is to enhance the inter-particle friction

dissipation rather than energy dissipation by itself.

Overall, these results suggest that the total energy dissipated during particle crushing

under quasi-static loading is much higher than the energy required to break inter-particle

bonds. Consequently, neglecting frictional dissipation, especially for an assembly of

112

particles, will lead to greatly overestimate breakage energy.

Figure 6.16 presents the breakage energy during loading in different WR samples in

relation to the samples’ Aspect Ratio, indicating a direct link between the breakage

energy and particle shape. The breakage energy per volume change of the sample clearly

rose during loading, and more energy was consumed to break the representative particle

in the bulky assembly compared to those with lower AR. It can thus be suggested that

under the identical compressive stress, less particle fragmentation would ocurr in an

assembly with a higher percentage of bulky particles in comparison with an assembly

consisting of a lower percentage of bulky particles.

Figure 6.15. Input energy versus energy dissipation through breakage based on DEM

simulations of different assemblies of WR, (The amount of energy was calculated up to the onset of ‘representative particle’ breakage)

Figure 6.16. Breakage energy per volume change of the WR sample versus applied force in relation to particle shape

y = 3.45x

y = x

0

30,000

60,000

90,000

0 10000 20000

Inp

ut

Ener

gy: N

.mm

Breakage Energy: N.mm

Bulky

Elongated

Flaky

113

6.4. Summary

Subroutines were developed to simulate different particle shapes and also particle

breakage of crushed waste rock during different testing, initially in 2D and ultimately in

3D. It has been confirmed that particle shape and particle breakage have significant

influences on the strength and deformation of a test specimen (using 2D DEM

simulations). The fracture mechanism of brittle WR particles with different shapes was

identified using 3D DEM simulations, indicating that tensile failures on the plane linking

the contact points of the particle with adjacent particles are the dominant type of particle

failure. Precise 3D particle shapes were generated by a large number of bonded spherical

sub-particles and used to model single particle crushing and particle assembly crushing.

The PAC simulations demonstrated that breakage energy is closely dependent on the

particles’ shape factor. Additionally, this study has shown that the energy calculated from

the force-displacement curve is far greater than the energy dissipated through particle

fragmentation. Breakage energy measured accurately by DEM simulations was less than

one-third of the total input energy. The results have indicated that a large portion of input

energy is dissipated through other mechanisms, particularly through frictional dissipation

between particles and fragments, and the input energy cannot be equated to the breakage

energy.

114

7. BREAKAGE AND PARTICLE

CHARACTERISTICS EVOLUTION THROUGH

SYNCHROTRON TOMOGRAPHY

Limited studies have been conducted to understand the relationship between particle

fracture and particle morphologies or internal microstructure, due to the experimental

difficulty in examining the evolution of particle microstructure during the fracture process

(Garcia et al., 2009, White et al., 2003). One of the aims of this study is to investigate and

track fracture patterns of granular C&D materials at the particle-scale. To accomplish this

goal, a series of experimental tests and 3D DEM simulations were conducted on WR,

RCA, and CB coarse grains (i.e. size fraction: 13.2-19 mm) across the scale from a single

particle crushing to assemblies of grains. The results demonstrated that particle shape is

one of the main factors governing particle breakage. However, the mean size (i.e. d50 and

d30) of C&D materials, which determines the governing particle-level forces and

associated macro-scale behaviour, is between 0.425 to 4.75 mm. Hence, Synchrotron

Radiation-based X-ray Micro-Computed Tomography was used as a powerful tool for

further analysing particle breakage at a smaller scale, which was impossible to achieve

by using conventional laboratory tests. Rapid advances in high-resolution 4D imaging

techniques have opened up unprecedented access to the grain scale, allowing one to ‘see’

inside the material, which can enhance the understanding of how microscale physics (the

cause) relate to various macroscale phenomena (the effect) (Viggiani et al., 2015).

As mentioned previously, synchrotron light is a monochromatic highly culminated X-

ray source that produces a beam with a high flux and a specific energy level (Boldeman

and Einfeld, 2004). The synchrotron beam, utilised in this study, in contrast with

laboratory scanners, provides much more rapid scanning than other available CT

scanners, and is highly efficient given the sequential loading-imaging cycles required to

characterise progressive deformation and breakage. An in-situ particle assembly

compression apparatus was also designed to carry out crushing tests on different kinds of

C&D particles under confinement during scanning. Crack propagation and particle

deformation, particularly particle morphology changes during loading, were precisely

observed and analysed.

115

7.1. Crack propagation in different C&D granular

materials

Grain fracture is common in the form of brittle fracture in geomaterials, causing rapid

propagation of several cracks through a stressed grain. The crack propagation patterns

and branching lead to identification of the point where and when a crack starts, and also

provide valuable information about the root of the crack, crack energy, and stress

condition (Quinn, 2007). CT images are the best available means by which to investigate

and interpret the fracture mechanism and its related theories owing to the remarkable

difference between X-ray attenuation in soil and air phases.

7.1.1. Effect of shape

A central vertical slice tomograph of a Crushed Brick (CB) specimen, in the particle

size fraction of 2.36-4.75 mm, is shown in Figure 7.1 at the initial phase and under 5 kN

(10.2 MPa) constraint compression. Since CB is mostly a by-product of demolition

activities of buildings, it usually consists of a relatively high percentage of other granular

materials, including crushed basaltic Waste Rock (WR), Recycled Concrete Aggregate

(RCA), and Portland Cement Mortar pieces (PCM). The density profile across the sample

helps to differentiate the various grain types, while the lowest density is related to PCM

particles (Figure 7.1a). It should be noted, as mentioned previously, that denser materials

appear in a brighter colour in an X-ray image in contrast to low density materials, such

as the air phases between particles, which are plain black. PCM grains can also be

recognised easily from their internal texture, an agglomeration of smaller particles

bonded together by mortar paste, as shown in Figure 7.1b. The unique microstructure

and low density of PCM result in severe fragmentation of them under 5 kN compression

(Figure 7.1c). Moreover, extensive bending failure occurred to the elongated CB grain,

and tip bending can clearly be observed in the WR-2 particle, where stress concentration

is much higher compared to other regions. Figure 7.2 also shows bending failure of

elongated particles in an RCA sample. However, referring to Figure 7.1b, WR-1 only

experienced asperity breakage since it lost loading contact points with its surrounding

grains. Among grains across the defined line in Figure 7.1b, the bulky RCA grain

remained almost intact under 5 kN compression. An overall view of Figure 7.1 shows

that, apart from low density and agglomerated PCM grains, bulky grains showed higher

116

resistance to breakage. The higher resistance of bulky grains to breakage can also be seen

in Figure 7.3. Central vertical slices of granular basaltic crushed WR, in the particle size

fraction of 2.36-4.75 mm, under 0, 10.2, and 20.4 MPa compression are shown in Figure

7.3. Mostly, bulky grains remained intact or experienced minor asperity breakage while

the rest of the grains crushed dramatically, particularly during the last phase of loading.

Figure 7.1. Crack propagation in different grains in a CB assembly: a) Density profile, b) The initial phase, c) After 5 kN compression

117

Figure 7.2. Bending failure of elongated RCA grains: a) Initial state, b) After 10.2 MPa compression

Figure 7.3. Basaltic Crushed WR under a) 0, b)10, and c)20 MPa compression (asterisks, ‘*’, are highlighting bulky grains not experiencing severe breakage)

0 MPa 10 MPa 20 MPa

**

*

*

*

*

(a) (b) (c)

25 mm

118

7.1.2. Effect of internal microstructure on crack patterns

As shown in Figure 7.1, PCM grains experienced severe fragmentation. Apart from

the low density of PCM grains, their microstructure, an agglomeration of small sub-

particles bonded together by mortar paste, influenced the grain’s strength against

breakage. A close-up of a PCM grain resulting from SEM imaging is shown in Figure

7.4. Different deformation response of sub-particles compared to their surrounding matrix

leads to stress concentrations along the sub-particle boundaries and ultimately facilitates

the crack propagation. This is also observed and reported by a number of researchers,

such as Katsaga (2010) testing concrete beams, and Tavares and das Neves (2008) testing

quarry rock samples.

The fractures typically have a complicated spatial distribution. A fast propagation of

numerous cracks through a stressed region is defined as a brittle fracture (Tattersall and

Tappin, 1966). Figure 7.5 presents crack paths in individual WR particles after 10.2 MPa

compression. Although it has been shown that the dominant type of cracks are tensile

cracks (Chapter 5 and 6), some particles experience shear cracking in the assembly, e.g.

No. 2 and 6 in Figure 7.5b. The main cause of the shear fracture could be the orientation

of the particle in relation to its contact points with surrounding particles. Although

particles tend to rest along their longest axis, there are a few cases, such as particle No. 2

and 6, whose longest axis are oblique to the loading direction. On the other hand, tensile

cracking is clear for particle No. 1 and 7, and the crack path is parallel to the loading

direction, being consistent with the theory of tensile fracture (Bažant and Oh, 1983).

Fracture complexities consist of crack deflection, crack branching, and crack arrest

(Katsaga, 2010). Crack branching, stemming from the vesicular texture of WR, is shown

in Figure 7.5, particle No. 3, 4 and 5. Basaltic crushed Waste Rock has a vesicular texture

as depicted in Figure 7.6 obtained using Scanning Electron Microscopy (SEM). When a

crack is arrested by a hole, higher stress is accumulated (Figure 7.7). The release of this

excess energy produces further crack branches (Hutchinson, 1968). In fact, the existence

of voids reduces the stress required for the development of the cracks, similar to the

observation made by Zhao et al. (2015) on single sand particles.

119

Figure 7.4. SEM image of an agglomerated PCM grain

Nevertheless, it is difficult to conclude that the cracks initiated from these internal

voids based solely on CT images, especially because a large number of internal voids

remain intact after grain splitting. It can thus be suggested that, although initial

microstructures such as voids can influence crack propagation paths, other factors such

as grain mineralogy and morphology first and foremost play critical roles in the initiation

of a fracture.

The cleavage pattern is also a structural weakness within a soil particle, along with

internal voids. As mentioned previously, Basaltic Waste Rock is mainly composed of

pyroxene, olivine, and plagioclase minerals (Peck et al., 1992). Figure 7.6 shows the

porphyritic texture of a basaltic WR particle; crystals of pyroxene (brighter colour) are

suspended in a groundmass/matrix of fine-grained dark and light (less dense) minerals

(Figure 7.8). Olivine has no cleavage and its fracture type is conchoidal; however,

pyroxene has a perfect cleavage in two directions that intersect at nearly right angles. This

crystallographic structure brings about fractures along the cleavage resulting in the almost

perpendicular fracture planes, as shown in Figure 7.8.

120

Figure 7.5. Fracture propagation in WR grains: a) Initial ortho-slice, b) After 10 MPa

vertical compression ( : tensile event; : shear event; : crack branching)

Figure 7.6. Porphyritic and vesicular texture of a WR grain

121

Figure 7.7. Microstructural effect on grain tensile splitting: a) A close-up of the vesicular WR grains in an assembly, b) After 5 kN (10.2 MPa) compression

Figure 7.8. Fractures following cleavage in WR grains

7.2. Evolution of grain property due to breakage

In order to gain a better understanding of soil behaviour in relation to grain-scale

damage, the changes to the grading and grain morphology of specimens were examined.

A uniformly graded and spectrum of C&D materials with grain diameter (d) varying

between 0.425 to 4.75 mm were scanned under 0, 5, 10 and 15 kN compressive loading

(i.e. 0, 10.2, 20.4 and 30.6 MPa). A sand sample, with particle size ranging from 1.18 to

2.36 mm, was also studied for comparison purposes (Figure 7.9). The grain size and

shape distribution after each loading sequence were obtained from precise 3D

reconstruction of 2159 CT horizontal slices using Avizo 9. Image processing techniques,

including image thresholding, filtering, and segmentation as explained in Chapter 4, were

used to separate and label each fragment in the assembly.

122

Figure 7.9. Sand particles: a) CT ortho-slice, b) natural sand used in this research

7.2.1. Soil grading and fractal distribution

While initial grain size distribution is one of the main factors governing the mechanical

behaviour of a granular assembly, the way it changes under loading can affect material

behaviour, especially the potential for further breakage. The grain size distribution of the

confined C&D assemblies under different compressive loading is illustrated in Figures

7.10, 7.11, and 7.12. Only fragments with a diameter larger than 0.075 mm were analysed

due to inaccuracy in segmenting and evaluating characteristics of extremely small

fragments. Figure 7.13a shows grain size distributions of a sand assembly under different

loading levels. After scanning under 20 MPa loading, the sand sample was recovered and

sieve analysed. As shown in Figure 7.13a, there is a slight discrepancy between the grain

size distribution measured by 3D reconstruction of the sample and the results from the

sieve analysis, particularly between the fine contents. This discrepancy is related to the

loss of fine grains during extruding the sample from the chamber at the end of the

scanning.

The samples’ grading clearly changed from a uniform to a well-graded distribution as

the stress increased and breakage progressed further (Figures 7.10, 7.11, 7.12, and

7.13a). Altuhafi and Coop (2011) also reported that toward the terminal state, the soil

gradation eventually tends to shift to well-graded, which explains the fact that very well-

123

graded and fractal soil experiences no dramatic breakage. This is evident in the spectrum

of C&D materials experiencing less breakage compared to uniformly-graded C&D

materials (Figures 7.10a, 7.11a, and 7.12a). Fractal theory, first used by Hartmann

(1969) to study crushing processes in meteorites, is widely used to quantify the amount

of breakage occurring under loading or after large displacement. It is widely accepted that

fragmentation is a fractal phenomenon, suggesting that it is also a scale-invariant

mechanism, and the fractal dimension (Df) describes the changes in grain size

distribution. Referring to Equation 7.1, the original fractal dimension can be calculated

based on number of grains, i.e. N(>r) which is the cumulative number of fragments with

a radius larger than r (Zhao et al., 2015). However, due to the difficulty in counting grains,

Einav (2007b) proposed a method based on the mass of grains smaller than a certain size.

3log (r) + log N(> r) (3-Df) log (r) Eq. (7.1)

With the aid of CT scanning, numbers of grains and newly generated fragments were

directly and precisely measured. The radius was calculated by measuring the volume of

each grain and defined as the radius of a sphere having the same volume as the grain. The

fractal condition is clear in Figures 7.14b-d, 7.15b-d, 7.16b-d and 7.13b, which shows

(3-Df) from Equation 7.1. The fractal dimension rises from 0.6 to 1.2 for WR samples, to

1.5 for RCA samples, to 1.3 for CB samples, and 0.3 to 1.4 for sand, with the increase in

compressive stress. It confirms that as further breakage occurs in the sample, the

gradation tends to be more fractal. This is the same trend observed by Turcotte (1986),

Einav (2007c), and Zhao et al. (2015). Nevertheless, it appears that some large grains did

not experience significant fragmentation, and that the fractal region has ended for grains

with diameter above 0.6, 1.4, and 1.6-2.4 mm within the initial size fraction of 0.425-

1.18, 1.18-2.36, and 2.36-4.75 mm, respectively. Interestingly, at all stress levels, the

upper bound of fractality is approximately the same and is close to d50 (i.e. median

diameter). In addition, this limit is also approximately the same for all kind of materials

from the same initial size fraction. This observation contrasts with the notion of size

effects in the strength of material, where breakage survival probability of a grain

decreases as its size increases (Jones and Ashby, 2005). McDowell et al. (1996) also

stated that while crushing strength of a grain depends on its size, high coordination

number (number of neighbouring grains) reduces the probability of grain fractures, and

attempts to model grain breakage based on these two effects (i.e. grain size and

124

coordination number) are very limited. Therefore, the termination of fractality in Figure

7.14b-d, 7.15b-d, 7.16b-d and 7.13b is related to the fact that when stress increases, more

fine fragments are generated by breakage, resulting in a denser packing and higher

coordination number around the larger grains. Consequently, breakage becomes more

dominant in smaller grains and less likely for larger ones. It should be noted that no clear

fractal region was observed in the spectrum of C&D materials under crushing (i.e. initial

size fraction of 0.425-4.75 mm, Figure 7.14a, 7.15a, 7.16a) due to the low level of

breakage occurring in the samples.

Figure 7.10. Changes in WR particle size distribution under different loading levels: a) Sample’s initial particle size: 0.425-4.75 mm, b) 0.425-1.18 mm, c) 1.18-2.36 mm, and d) 2.36-

4.75 mm

125

Figure 7.11. Changes in RCA particle size distribution under different loading levels: a)

Sample’s initial particle size: 0.425-4.75 mm, b) 0.425-1.18, c) 1.18-2.36, and d) 2.36-4.75 mm

Figure 7.12. Changes in CB particle size distribution under different loading levels: a)

Sample’s initial particle size: 0.425-4.75 mm, b) 0.425-1.18 mm, c) 1.18-2.36 mm, and d) 2.36-4.75 mm

126

Figure 7.13. Sand specimen under different loading levels (0, 5, 10, and 20 MPa): a) Changes in grain size distribution, b) The fractal distribution

Figure 7.14. Fractal distribution of WR samples: a) Sample’s initial particle size: 0.425-4.75 mm, b) 0.425-1.18 mm, c) 1.18-2.36 mm, and d) 2.36-4.75 mm

5 MPa

10 MPa

20 MPa

127

Figure 7.15. Fractal distribution of RCA samples: a) Sample’s initial particle size: 0.425-4.75 mm, b) 0.425-1.18 mm, c) 1.18-2.36 mm, and d) 2.36-4.75 mm

Figure 7.16. Fractal distribution of CB samples: a) Sample’s initial particle size: 0.425-4.75 mm, b) 0.425-1.18 mm, c) 1.18-2.36 mm, and d) 2.36-4.75 mm

128

7.2.2. Changes in external morphology

Statistical and quantitative analyses of 3D labelled images are presented in Figures

7.17, 7.18, and 7.19 in terms of grain shape/morphology evolution. Figures 7.17, 7.18,

and 7.19 show the changes in the Aspect Ratio distribution due to breakage, which is the

ratio of the short axis of the best-fitted ellipsoid to its intermediate axis. Figures 7.17,

7.18, and 7.19 also demonstrate the true sphericity distribution, defined as the ratio of the

surface area of a sphere having the same volume as the fragment to the actual surface area

of the fragment (Wadell, 1933). Figures 7.17, 7.18, and 7.19 indicate that after the first

loading sequence (i.e. 10 MPa or 20 MPa), both AR and true sphericity increased. The

same observation was also seen for the sand sample under the first loading sequence (i.e.

5 MPa) (Figure 7.20). This observation suggests that grains show a tendency to move

away from morphological extremes during crushing. Abbireddy and Clayton (2015) also

reported similar results on changes in grain forms during a triaxial shear test. Nonetheless,

under higher levels of stress, a reverse trend was observed in both true sphericity and

aspect ratio distributions, as both shape factors tend to decrease regardless of the material

types (Figures 7.17, 7.18, 7.19, and 7.20).

Figure 7.21 also illustrates the evolution of mean values of AR and true sphericity

versus Hardin’s relative breakage for different kinds of C&D materials, which is

calculated as the ratio of total breakage to breakage potential (more information is given

by Hardin (1985)). A decrease in the mean value of true sphericity and a notable drop in

average AR were measured at higher stress levels (i.e. 20 MPa or 30 MPa). It is interesting

that newly created fragments under higher stress levels are less spherical and have lower

aspect ratios than the initial grains. Takei et al. (2001) and Zhao et al. (2015) also noted

a slight reduction in sphericity and AR of different individual grains in a single particle

crushing test. Furthermore, to quantify the uniformity of shape distribution, the Relative

Distribution Factor (RDF) was calculated for each shape factor distribution (Equation

7.2):

RDF=AR90 /AR10

RDF=True Sphericity90 / True Sphericity10 Eq. (7.2)

where AR90 and AR10 or True Sphericity90 and True Sphericity10 are the values of the

shape factor at which 90% and 10% of fragments have a smaller value, respectively.

129

Figure 7.22 reveals that initially there has been a slight fall in RDF of true sphericity and

a larger drop in RDF of AR. This implies that the grains during the first phases of splitting

and breakage tend to create fragments with approximately uniform shapes. Then, the RDF

of both shape factors gradually has increased when the sample experienced more severe

fragmentation under the higher stress level. This suggests that, due to the complex

mechanisms of breakage, during severe fragmentation, newly generated fragments

possessed more diverse shapes. It should also be noted that true sphericity at every

loading stage is distributed normally with equal mean, median, and mode values. On the

contrary, considering the skewness of AR distribution, shown in Figure 7.23, the

distribution tends to skew to the right, toward lower values of AR, with increased

crushing in the sample. Skewness is a measure of the asymmetry of a distribution.

Skewness has been measured to determine the extent to which a distribution differs from

a normal distribution. Based on Bulmer (1979), if skewness is between −1 and −½ or

between +½ and +1, the distribution is moderately skewed; thus, the AR distribution is

moderately skewed for all kinds of C&D materials.

A sand sample with an initial particle size of 1.18-2.36 mm was also examined in terms

of morphology/form evolution. Interestingly, although the sand particles have different

mineralogical and microstructural characteristics from C&D materials (Figure 7.24), the

same trend was observed in morphological changes by increasing stress (Figure 7.25).

Overall, morphological changes during breakage showed a reversal trend as stress

increased, with a shift from more spherical and bulky to less spherical and flaky shapes.

7.2.3. Universality of grain property evolution due to breakage

The literature on particle ‘size’ and shape has highlighted that there is a dependency

between size and shape of particles. Domokos et al. (2015) showed that the shape of

fragments generated under (significant) dynamic loading on ‘large’ 15-150 mm particles

obeys universal scaling laws and that the larger crushed particles tend to be elongated.

However, Altuhafi and Coop (2011) and Sun et al. (2014) obtained contrasting results

(increase in sphericity and aspect ratios of sand and ballast particles as particle size

increased), for smaller diameter particle assemblies under different crushing loads than

that of Domokos et al. (2015).

130

Figure 7.17. Morphology evolution of WR grains under different loading levels: Aspect

Ratio and True sphericity distribution: a) Sample’s initial particle size: 0.425-4.75 mm, b) 0.425-1.18 mm, c) 1.18-2.36 mm, and d) 2.36-4.75 mm

131

Figure 7.18. Morphology evolution of RCA grains under different loading levels: Aspect Ratio and True sphericity distribution: a) Sample’s initial particle size: 0.425-4.75 mm, b)

0.425-1.18 mm, c) 1.18-2.36 mm, and d) 2.36-4.75 mm

132

Figure 7.19. Morphology evolution of CB grains under different loading levels: Aspect Ratio and True sphericity distribution: a) Sample’s initial particle size: 0.425-4.75 mm, b) 0.425-1.18

mm, c) 1.18-2.36 mm, and d) 2.36-4.75 mm

133

Figure 7.20. Morphology evolution of sand grains under different loading levels: a) Aspect Ratio, b) True Sphericity

The main focus of this chapter is on the way grain properties change with increasing

stress (rather than the existing correlation between particle shape and size). Different

C&D materials, along with a sand sample, were investigated in this study. The

composition of the sand sample was found to be mainly SiO2 in the form of quartz being

completely different from C&D materials (Figure 7.24). In the present study, the results

from studies on various grain assemblies of different size ranges, i.e. basaltic crushed

Waste Rock, Recycled Concrete Aggregate, Crushed Brick, and sand particles, show that

the general trend of changes in particle shape obeys an astonishing universality. The same

generic evolution with increasing stress was observed irrespective of material details or

initial size ranges, a reversal shift from more spherical and bulky to less spherical and

flaky shapes by increasing stress (Figures 7.17-7.20). The result can be explained by the

fact that less spherical particles normally fragmented earlier under loading. As discussed

earlier in Chapters 5 and 6, the less spherical particles (i.e. elongated or flaky particles)

showed lower crushing strength; therefore, their total fragmentation is likely at early

phases of loading (Figure 7.26). Tensile splitting in two nearly equal halves was shown

to be the dominant type of fracture in each sample (see Chapters 5 and 6). Therefore, an

increase in true sphericity or Aspect Ratio was observed under lower loading levels.

Asperity breakage or abrasion commonly occurring during early phases of loading also

causes generation of more rounded and spherical fragments.

134

Figure 7.21. Changes in mean values of AR and true sphericity due to breakage: a) WR, b) RCA, and c) CB

135

Figure 7.22. Relative Distribution Factor of AR and true sphericity distributions: a) WR, b) RCA, and c) CB

136

Figure 7.23. Skewness of AR distributions: a) WR, b) RCA, and c) CB

137

Figure 7.24. EDS elemental analysis of sand (insert: SEM image of a sand particle)

Figure 7.25. Changes in statistical properties of shape factor distributions of sand fragments due to breakage: a) Mean values of AR and true sphericity, and b) Relative Distribution Factor

However, by increasing the load, severe fragmentation and catastrophic splitting cause

more irregular fragments and accordingly cause a decrease in true sphericity or Aspect

Ratio. Zheng and Tannant (2016) reported a notable decrease in roundness and the degree

of sphericity of sand particles, with an initial size of 0.3-0.85 mm, under up to 40 MPa

vertical compression. They described the initial shape of their sand particles as rounded

and close to a spherical shape, with a roundness range of 0.8 to 0.9 and a degree of

sphericity between 0.75 to 0.95. In contrast, Abbireddy and Clayton (2015), after studies

on elongated and flaky glass nuggets with particle sizes ranging from 0.6 to 2 mm during

triaxial shear tests, pointed out an increase in aspect ratio of resulting broken particles. In

this study, samples containing particles with a variety of shapes were studied under

sequential loading, revealing the interesting reversal trend in morphology evolution as

138

stress increased. In fact, the findings presented in this study support both previous

aforementioned research studies and suggest that, where the percentage of elongated and

flaky particles is significant, an initial increase is expected in the sphericity and aspect

ratio of the newly generated fragments. However, this initial increase in shape factors is

not likely to occur for granular samples containing only rounded and spherical particles.

Moreover, at all stress levels, the upper bound of fractality was approximately the

same (Figures 7.14-16 and 7.13b). Along the same line, the termination of the fractal

region was dependent on the initial particle size but irrespective of material types or

loading levels (as the fractal region terminated at grains with diameters above 0.6, 1.4,

and 1.6-2.4 mm in samples with initial particle sizes of 0.425-1.18, 1.18-2.36, and 2.36-

4.75 mm, respectively, Figures 7.14-16 and 7.13b).

Figure 7.26. Schematic explanation of particle shape evolution due to breakage

7.3. Summary

Grain breakage brings about changes in characteristics of granular materials,

especially in grain size distribution and grain morphology. These shifts affect the

mechanical behaviour of granular media and, most prominently, the material’s crushing

strength against further breakage.

Various granular materials under different loading levels were studied using

synchrotron tomography. The effect of particle shape was identified as a notable factor

governing grain breakage. Crack formation and its related mechanisms were investigated

for different types of grains. Although the microstructural features such as a vesicular

texture tended to change the crack path, the findings attest to the dominant influence of

grain mineralogy and morphology on crack initiation.

139

Cutting-edge synchrotron tomography was used to study alterations in grain

characteristics of C&D and a sand specimens. The fractal distribution of the particle

assemblies due to crushing demonstrated that breakage becomes dominant in smaller

grains rather than larger ones where an increase in the amount of newly generated fine

fragments causes a high coordination number surrounding the larger grains.

The results of morphological changes also reveal that there is a reversal trend in the

grain shape evolution with increasing stress. The breakage process causes generation of

fragments with a greater isotropic shape, whereas by increasing the stress, this trend

reversed. Owing to severe breakage and splitting under higher stress levels, less spherical

fragments (i.e. anisotropic shape) with a lower aspect ratio compared to the original grains

were created. In addition, the results reported here are from studies on various granular

materials with different particle sizes, i.e. basaltic crushed Waste Rock, Recycled

Concrete Aggregate, Crushed Brick, and sand particles, showing that the general trend of

changes in particle shape obeys an astonishing universality. The same generic evolution

by increasing stress was observed irrespective of material details or sizes, a reversal shift

from more spherical and bulky to less spherical and flaky shapes by increasing stress.

140

8. CONCLUSIONS AND RECOMMENDATIONS

In this research, three different types of granular recycled Construction and Demolition

(C&D) materials (i.e. basaltic crushed Waste Rock, Recycled Concrete Aggregate, and

Crushed Brick) used in roads, pavements, and embankments were studied at particle-

scale. Particle fracture, causing serious issues such as settlements, is the focus of the

present research. In many industrial processes, particle breakage occurs, which may or

may not be desirable. As an example, in mining and ore processing, particle breakage is

desired during grinding and milling processes; however, proppant crushing causes a

dramatic reduction in the recovery rate of hydrocarbons in the oil and gas industry. Hence,

the implications of this research are not limited solely to the geotechnical engineering

industry.

Following geotechnical, mineralogical, microstructural, and morphological

characterisation of the C&D materials, different experimental and numerical tests,

including synchrotron tomography and Discrete Element Modelling (DEM) simulation,

were conducted to improve the understanding of particle fracture and subsequent changes

in grain properties.

8.1. Major conclusions

8.1.1. Experimental observations and analyses from SPC and

PAC tests

The investigation into a variety of C&D materials with various microstructure and

mineralogical aspects demonstrated that:

• Particle tensile strength is profoundly affected by particle shape, and it has

been found that particle shape plays a more prominent role in the particle

breakage phenomena than mineralogy and microstructure of C&D particles.

• Following studies on several individual C&D particles, a modified particle

tensile strength, as a function of particle Aspect Ratio is introduced, where the

impact of particle shape is also considered.

• Further studies on particle fracture using Particle Assembly Crushing (PAC) tests

showed that although boundary conditions and particle coordination number have

141

remarkable effects on particle breakage, the significant influence of particle

shape on particle crushing is not diminished even in an assembly of particles.

• The findings suggest that brittle C&D granular materials with a higher

degree of sphericity (an Aspect Ratio closer to 1) and a lower flakiness index

would experience less particle breakage under loading.

8.1.2. Discrete Element Modelling

Micromechanical analyses, such as contact force measurement, are difficult or

sometimes impossible using conventional laboratory tests. Hence, DEM was employed

to study particle breakage and the associated energy dissipation. Precise three-

dimensional particle shapes were generated using the Fish scripting language, within

PFC2D and PFC3D, and used to model single particle crushing and particle assembly

crushing. The outcomes are summarised as follows:

• 2D DEM simulations confirm the importance of accurate simulation of

particle shape and also the significant influence of particle shape and breakage on

the macro-mechanical behaviour of C&D materials.

• The fragmentation mechanism of brittle WR particles with different

shapes was investigated using 3D DEM simulations. The results imply that tensile

cracks on the plane linking the contact points of the particle with adjacent particles

are the dominant type of particle fracture.

• The PAC simulations proved that energy dissipation due to breakage

depends on the particles’ shape factor.

• Energy components, including strain energy, frictional energy, and

breakage energy, in the particulate system were monitored precisely using DEM.

Energy dissipation measurement showed that less than one-third of the total input

energy was dissipated due to particle breakage. The results indicated that a large

portion of input energy is dissipated through other mechanisms, particularly

through frictional dissipation between particles and newly generated fragments.

8.1.3. Post-breakage analyses using synchrotron tomography

Grain breakage results in changes to the properties of granular materials, particularly

grain size distribution and grain morphology. These alterations affect the macro-

142

mechanical behaviour of particulate systems and most prominently, the material’s

crushing strength against further breakage. Various granular materials, i.e. C&D and sand

specimens, were studied under different loading levels using cutting-edge synchrotron

tomography. Salient observations and findings are summarised as follows:

• The notable effect of particle shape on particle breakage was confirmed

using CT images.

• Crack propagation and associated mechanisms were studied for different

types of grains. The results suggest that, even though microstructural features,

such as the vesicular texture, can facilitate or arrest a propagating crack, other

factors, such as grain mineralogy and morphology, mainly govern crack initiation.

• The evolution of the particle size distribution has shown that some coarse

particles were left relatively unbroken and cushioned by smaller particles. This

contrasts with the concept that larger particles have a higher likelihood of

breakage since they have a higher probability of containing internal flaws. In fact,

the fractal distribution of the particle assemblies due to breakage demonstrated

that there is a competition between cushioning and size effects. Eventually,

breakage becomes dominant in smaller grains rather than larger ones, where an

increase in the amount of newly generated fine fragments causes high

coordination number surrounding the larger grains.

• The results of morphology evolution demonstrated that there is a reversal

trend in the grain shape evolution with increasing stress. The breakage

phenomenon generates fragments with a greater isotropic shape; however, by

increasing the stress, this trend reversed. Due to catastrophic and severe breakage

under higher stress levels, less spherical fragments (i.e. anisotropic shape), with a

lower aspect ratio in comparison to the initial particles, were generated.

• The more significant finding to emerge from this study is that particle

shape evolution due to breakage obeys universality. Studies on various granular

materials with different particle size and gradation, i.e. basaltic crushed WR,

RCA, CB, and sand particles, showed that the generic evolution in particle shape

by increasing stress is the same for each sample irrespective of material details or

sizes.

143

8.2. Recommendations for future research

The investigation carried out in this research has indicated new aspects of particle

mechanics in relation to particle breakage and shape; however, there remains areas where

further research can be undertaken, as follows:

• Investigation of three types of C&D materials during SPC testing showed

that shape plays a more prominent role in particle breakage phenomena than

mineralogy and microstructure of particles. Further investigation is required to

examine this notion for other types of geomaterials.

• All tests and simulations performed in the present research were under

quasi-static compression. A further study could compare the presented results and

observed trends with results obtained under different loading conditions, such as

dynamic, repeated, and cyclic loading.

• In this research, particle shape evolution was investigated for particles in

the size range of 0.425-4.75 mm. The investigation into changes in particle

properties due to breakage is recommended on particles with larger sizes, for

example particle diameters > 20 mm.

• It has been proven that the changes in particle properties, such as the

reduction in particle sizes, due to breakage have a significant effect on the

permeability of the soil (Zheng and Tannant, 2016). It would be useful to examine

the effect of particle morphology evolution during sequential loading on pore

conductivity and void ratio of the soil. The observed reversal trend is likely to

have a noteworthy and interesting effect on pore size or pore conductivity

distribution.

• The development of a model that can predict crushing levels using input

parameters derived from simple tests, such as initial particle size, density, and

particle shape, would be a fruitful area for further work. Although several

breakage indices, based on changes in particle size distribution are currently

available (e.g. Hardin (1985); Indraratna et al. (2014)), the development of a

crushing prediction model has historically been hampered by a lack of access to

particle-scale information of geomaterials. The findings of this study, particularly

the significant effect of particle shape, can facilitate this development.

144

• This research was conducted on unbound granular materials. Further

research to extend the results to stabilised granular materials is also

recommended.

145

References

ABBIREDDY, C. & CLAYTON, C. 2015. The impact of particle form on the packing and shear behaviour of some granular materials: an experimental study. Granular Matter, 17, 427-438.

ABRÀMOFF, M. D., MAGALHÃES, P. J. & RAM, S. J. 2004. Image processing with ImageJ. Biophotonics international, 11, 36-43.

AFSHAR, T., DISFANI, M. M., ARULRAJAH, A., NARSILIO, G. A. & EMAM, S. 2017. Impact of particle shape on breakage of recycled construction and demolition aggregates. Powder Technology, 308, 1-12.

ALSHIBLI, K. A., STURE, S., COSTES, N. C., FRANK, M. L., LANKTON, M. R., BATISTE, S. N. & SWANSON, R. A. 2000. Assessment of localized deformations in sand using X-ray computed tomography. ASTM geotechnical testing journal, 23, 274-299.

ALTUHAFI, F. & COOP, M. R. 2011. Changes to particle characteristics associated with the compression of sands. Géotechnique, 61, 459-471.

AMAN, S., TOMAS, J. & KALMAN, H. 2010. Breakage probability of irregularly shaped particles. Chemical Engineering Science, 65, 1503-1512.

ANTONY, S., MORENO-ATANASIO, R. & HASSANPOUR, A. 2006. Influence of contact stiffnesses on the micromechanical characteristics of dense particulate systems subjected to shearing. Applied physics letters, 89, 214103.

ANTONYUK, S., KHANAL, M., TOMAS, J., HEINRICH, S. & MÖRL, L. 2006. Impact breakage of spherical granules: experimental study and DEM simulation. Chemical Engineering and Processing: Process Intensification, 45, 838-856.

ARULRAJAH, A., ALI, M., PIRATHEEPAN, J. & BO, M. 2011. Geotechnical properties of waste excavation rock in pavement subbase applications. Journal of Materials in Civil Engineering, 24, 924-932.

ARULRAJAH, A., DISFANI, M. M., HORPIBULSUK, S., SUKSIRIPATTANAPONG, C. & PRONGMANEE, N. 2014. Physical properties and shear strength responses of recycled construction and demolition materials in unbound pavement base/subbase applications. Construction and Building Materials, 58, 245-257.

ARULRAJAH, A., PIRATHEEPAN, J., BO, M. & SIVAKUGAN, N. 2012a. Geotechnical characteristics of recycled crushed brick blends for pavement sub-base applications. Canadian Geotechnical Journal, 49, 796-811.

ARULRAJAH, A., PIRATHEEPAN, J., DISFANI, M. & BO, M. 2012b. Geotechnical and geoenvironmental properties of recycled construction and demolition materials in pavement subbase applications. Journal of Materials in Civil Engineering, 25, 1077-1088.

ARULRAJAH, A., PIRATHEEPAN, J., DISFANI, M. M. & BO, M. W. 2013a. Geotechnical and Geoenvironmental Properties of Recycled Construction and Demolition Materials in Pavement Subbase Applications. Journal of Materials in Civil Engineering, 25, 1077-1088.

ARULRAJAH, A., PIRATHEEPAN, J., DISFANI, M. M. & BO, M. W. 2013b. Resilient Moduli Response of Recycled Construction and Demolition Materials in Pavement Subbase Applications. Journal of Materials in Civil Engineering, 25, 1920-1928.

ASTM 2000. D854-00: Standard Test Method for Specific Gravity of Soil Solids by Water Pycnometer. American Society for Testing and Materials. Annual book of ASTM Standards, 4, 02.

ASTM 2004. C127: Standard Test Method for Density, Relative Density (Specific Gravity), and Absorption of Coarse Aggregate. Annual book of ASTM standards.

ASTM 2009a. D5102-09: Standard Test Method for Unconfined Compressive Strength of Compacted Soil-Lime Mixtures. ASTM International, West Conshohocken, PA, www.astm.org.

146

ASTM 2009b. D6913-04: Standard test methods for particle size distribution (gradation) of soils using sieve analysis. West Conshohocken, PA: ASTM International (Accessed: 2014. http://www.astm.org/Standards/D6913.htm).

ASTM 2012. D 1557: Standard Test Method for Laboratory Compaction Characteristics of Soil Using Modified Effort. ASTM, West Conshohocken, Pennsylvania, USA.

ASTM 2014. D75/D75M: Standard Practice for Sampling Aggregates. ASTM International, West Conshohocken, PA.

ÅSTRÖM, J. & HERRMANN, H. 1998. Fragmentation of grains in a two-dimensional packing. The European Physical Journal B-Condensed Matter and Complex Systems, 5, 551-554.

BARDET, J. 1998. Introduction to computational granular mechanics. Behaviour of granular materials. Springer.

BAŽANT, Z. P. 1999. Size effect on structural strength: a review. Archive of applied mechanics, 69, 703-725.

BAŽANT, Z. P. & OH, B. H. 1983. Crack band theory for fracture of concrete. Materials and structures, 16, 155-177.

BEHRINGER, R., HOWELL, D., KONDIC, L., TENNAKOON, S. & VEJE, C. 1999. Predictability and granular materials. Physica D: Nonlinear Phenomena, 133, 1-17.

BELHEINE, N., PLASSIARD, J. P., DONZÉ, F. V., DARVE, F. & SERIDI, A. 2009. Numerical simulation of drained triaxial test using 3D discrete element modeling. Computers and Geotechnics, 36, 320-331.

BEUCHER, S. & MEYER, F. 1992. The morphological approach to segmentation: the watershed transformation. OPTICAL ENGINEERING-NEW YORK-MARCEL DEKKER INCORPORATED-, 34, 433-433.

BILDERBACK, D. H., ELLEAUME, P. & WECKERT, E. 2005. Review of third and next generation synchrotron light sources. Journal of Physics B: Atomic, molecular and optical physics, 38, S773.

BLOTT, S. J. & PYE, K. 2006. Particle size distribution analysis of sand‐sized particles by laser diffraction: an experimental investigation of instrument sensitivity and the effects of particle shape. Sedimentology, 53, 671-685.

BOAS, F. E. & FLEISCHMANN, D. 2012. CT artifacts: causes and reduction techniques. Imaging in Medicine, 4, 229-240.

BOLDEMAN, J. & EINFELD, D. 2004. The physics design of the Australian synchrotron storage ring. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 521, 306-317.

BOLTON, M., NAKATA, Y. & CHENG, Y. 2008. Micro-and macro-mechanical behaviour of DEM crushable materials. Géotechnique, 58, 471-480.

BONO, J. P. D., MCDOWELL, G. R. & WANATOWSKI, D. 2014. DEM of triaxial tests on crushable cemented sand. Granular Matter, 16, 563-572.

BS 2000. 812-105.1: Method for Determination of Particle Shape; Flakiness Index. British Standard Institution, London, UK.

BULMER, M. G. 1979. Principles of statistics, Courier Corporation. CAI, W. 2013. Discrete Element Modelling of constant strain rate and creep tests on a graded

asphalt mixture. University of Nottingham. CAICEDO, B., OCAMPO, M. & VALLEJO, L. 2013. Modelling crushing of granular materials as a

poly-disperse mixture. Proc. of the 18th ICSMGE, Paris. CAMUSSO, M. & BARLA, M. 2009. Microparameters Calibration for Loose and Cemented Soil

When Using Particle Methods. International Journal of Geomechanics, 9, 217-229. CAVARRETTA, I., COOP, M. & O'SULLIVAN, C. 2010. The influence of particle characteristics on

the behaviour of coarse grained soils. Géotechnique, 60, 413-423. CAVARRETTA, I., O’SULLIVAN, C., COOP, M. R., NAKAGAWA, M. & LUDING, S. 2009. Applying 2D

shape analysis techniques to granular materials with 3D particle geometries. 833-836.

147

CEPURITIS, R., GARBOCZI, E. J., FERRARIS, C. F., JACOBSEN, S. & SØRENSEN, B. E. 2017a. Measurement of particle size distribution and specific surface area for crushed concrete aggregate fines. Advanced Powder Technology, 28, 706-720.

CEPURITIS, R., GARBOCZI, E. J., JACOBSEN, S. & SNYDER, K. A. 2017b. Comparison of 2-D and 3-D shape analysis of concrete aggregate fines from VSI crushing. Powder Technology, 309, 110-125.

CHENG, Y., NAKATA, Y. & BOLTON, M. 2003. Discrete element simulation of crushable soil. Geotechnique, 53, 633-641.

CHEUNG, G. & O'SULLIVAN, C. 2008. Effective simulation of flexible lateral boundaries in two- and three-dimensional DEM simulations. Particuology, 6, 483-500.

CHEUNG, G. & O’SULLIVAN, C. 2008. Effective simulation of flexible lateral boundaries in two- and three-dimensional DEM simulations. Particuology, (6):483-500.

CHO, G.-C., DODDS, J. & SANTAMARINA, J. C. 2006. Particle shape effects on packing density, stiffness, and strength: natural and crushed sands. Journal of Geotechnical and Geoenvironmental Engineering, 132, 591-602.

CHO, N., MARTIN, C. D. & SEGO, D. C. 2007. A clumped particle model for rock. International Journal of Rock Mechanics and Mining Sciences, 44, 997-1010.

CIL, M. B. & ALSHIBLI, K. A. 2012. 3D assessment of fracture of sand particles using discrete element method. Géotechnique Letters, 2, 161-166.

CIL, M. B. & ALSHIBLI, K. A. 2014. 3D evolution of sand fracture under 1D compression. Géotechnique, 64, 351.

CILAS 2008. CILAS Particle Size Analyzers - User manual. 8 Avenue Buffon, B. P. 6319, Z.I. La Source, 45063 ORLEANS CEDEX.

COETZEE, C. 2016. Calibration of the discrete element method and the effect of particle shape. Powder Technology, 297, 50-70.

COLLOP, A. C., MCDOWELL, G. R. & LEE, Y. W. 2006. Modelling dilation in an idealised asphalt mixture using discrete element modelling. Granular Matter, 8, 175-184.

COOP, M., SORENSEN, K., FREITAS, T. B. & GEORGOUTSOS, G. 2004. Particle breakage during shearing of a carbonate sand. Géotechnique, 54, 157-164.

COWIE, B., TADICH, A., THOMSEN, L., GARRETT, R., GENTLE, I., NUGENT, K. & WILKINS, S. The current performance of the wide range (90–2500 eV) soft x‐ray beamline at the Australian Synchrotron. AIP Conference Proceedings, 2010. AIP, 307-310.

CUNDALL, P. 2004. PFC user manual. Itasca Consulting Group Inc, Mineapolis. CUNDALL, P. A. & STRACK, O. D. L. 1979. A discrete numerical model for granular assemblies.

Geotechnique, 29, 47-65. DAS, B. 2007. Fundamentals of geotechnical engineering, Cengage Learning. DAS, B. & SOBHAN, K. 2013. Principles of geotechnical engineering, Cengage Learning. DAS, N., GIORDANO, P., BARROT, D., MANDAYAM, S., ASHMAWY, A. K. & SUKUMARAN, B. 2008.

Discrete Element Modeling and Shape Characterization of Realistic Granular Shapes. ISOPE Annual Conference.

DE BOER, G. B., DE WEERD, C., THOENES, D. & GOOSSENS, H. W. 1987. Laser diffraction spectrometry: Fraunhofer diffraction versus Mie scattering. Particle & Particle Systems Characterization, 4, 14-19.

DESRUES, J., CHAMBON, R., MOKNI, M. & MAZEROLLE, F. 1996. Void ratio evolution inside shear bands in triaxial sand specimens studied by computed tomography. Géotechnique, 46, 529-546.

DISFANI, M. M. 2011. Sustainable use of recycled glass: biosolids blends in road applications. PhD dissertation, Swinburne University of Technology.

DISFANI, M. M., ARULRAJAH, A., BO, M. W. & SIVAKUGAN, N. 2012. Environmental risks of using recycled crushed glass in road applications. Journal of Cleaner Production, 20, 170-179.

148

DOMOKOS, G., KUN, F., SIPOS, A. Á. & SZABÓ, T. 2015. Universality of fragment shapes. Scientific reports, 5, 9147.

EINAV, I. 2007a. Breakage mechanics—Part I: Theory. Journal of the Mechanics and Physics of Solids, 55, 1274-1297.

EINAV, I. 2007b. Breakage mechanics—Part II: Modelling granular materials. Journal of the Mechanics and Physics of Solids, 55, 1298-1320.

EINAV, I. Fracture propagation in brittle granular matter. Proceedings of the Royal Society of London A: Mathematical, Physical and Engineering Sciences, 2007c. The Royal Society, 3021-3035.

ELGHEZAL, L., JAMEI, M. & GEORGOPOULOS, I. O. 2013. DEM simulations of stiff and soft materials with crushable particles: an application of expanded perlite as a soft granular material. Granular Matter, 15, 685-704.

EVANS, T. M. 2005. Microscale physical and numerical investigations of shear banding in granular soils. PhD Dissertation, Georgia Institute of Technology.

EVANS, T. M. & FROST, J. D. 2010. Multiscale investigation of shear bands in sand: Physical and numerical experiments. International Journal for Numerical and Analytical Methods in Geomechanics, n/a-n/a.

FAVIER, J., CURRY, D. & LAROCHE, R. Calibration of DEM material models to approximate bulk particle characteristics. 6th World Congress on Particle Technology, Nuremberg, Germany, 2010.

FERELLEC, J.-F. & MCDOWELL, G. R. 2010. A method to model realistic particle shape and inertia in DEM. Granular Matter, 12, 459-467.

FERNLUND, J. M. 1998. The effect of particle form on sieve analysis: a test by image analysis. Engineering Geology, 50, 111-124.

GABR, A. & CAMERON, D. 2011. Properties of recycled concrete aggregate for unbound pavement construction. Journal of Materials in Civil Engineering, 24, 754-764.

GAN, M., GOPINATHAN, N., JIA, X. & WILLIAMS, R. A. 2004. Predicting packing characteristics of particles of arbitrary shapes. KONA Powder and Particle Journal, 22, 82-93.

GANTENBEIN, D., SCHOELKOPF, J., MATTHEWS, G. P. & GANE, P. A. 2011. Determining the size distribution-defined aspect ratio of platy particles. Applied Clay Science, 53, 544-552.

GARCIA, D., LIN, C. & MILLER, J. 2009. Quantitative analysis of grain boundary fracture in the breakage of single multiphase particles using X-ray microtomography procedures. Minerals Engineering, 22, 236-243.

GERLACH, R. W. & NOCERINO, J. M. 2003. Guidance for obtaining representative laboratory analytical subsamples from particulate laboratory samples, US Environmental Protection Agency, Office of Research and Development, National Exposure Research Laboratory, Environmental Sciences Division.

GHAZVINIAN, E. 2010. Modelling and Testing Strategies for Brittle Fracture Simulation in Crystalline Rock Samples. Master Thesis, Queen's University.

GOLCHERT, D., MORENO, R., GHADIRI, M. & LITSTER, J. 2004. Effect of granule morphology on breakage behaviour during compression. Powder Technology, 143-144, 84-96.

GONDROM, S., ZHOU, J., MAISL, M., REITER, H., KRÖNING, M. & ARNOLD, W. 1999. X-ray computed laminography: an approach of computed tomography for applications with limited access. Nuclear engineering and design, 190, 141-147.

GONZALEZ, R. C. 2009. Digital image processing, Pearson Education India. GREEN, S. B. & SALKIND, N. J. 2010. Using SPSS for Windows and Macintosh: Analyzing and

understanding data, Prentice Hall Press. GRIFFITH, A. A. 1921. The phenomena of rupture and flow in solids. Philosophical transactions

of the royal society of london. Series A, containing papers of a mathematical or physical character, 221, 163-198.

149

HALL, C. 2015. Large-Scale Imaging on the Australian Synchrotron Imaging and Medical Beamline (IMBL). Synchrotron Radiation News, 28, 24-29.

HALL, C., HAUSERMANN, D., MAKSIMENKO, A., ASTOLFO, A., SIU, K., PEARSON, J. & STEVENSON, A. 2013. Detectors for the Imaging and Medical Beam Line at the Australian Synchrotron. Journal of Instrumentation, 8, C06011.

HANLEY, K. J., O'SULLIVAN, C., OLIVEIRA, J. C., CRONIN, K. & BYRNE, E. P. 2011. Application of Taguchi methods to DEM calibration of bonded agglomerates. Powder Technology, 210, 230-240.

HANSSON, J. & SVENSSON, P. 2001. Stress distribution in unbound aggregate of road structure: influence of surface roughness and particle shape. Bulletin of Engineering Geology and the Environment, 60, 223-226.

HARDIN, B. O. 1985. Crushing of soil particles. Journal of Geotechnical Engineering, 111, 1177-1192.

HARDING, D. 2002. Mineral identification using a scanning electron microscope. Minerals and Metallurgical Processing.

HARIRECHE, O. & MCDOWELL, G. R. 2003. Discrete element modelling of cyclic loading of crushable aggreates. Granular Matter, 5, 147-151.

HARTMANN, W. K. 1969. Terrestrial, lunar, and interplanetary rock fragmentation. Icarus, 10, 201-213.

HAUSERMANN, D., HALL, C., MAKSIMENKO, A. & CAMPBELL, C. The imaging and medical beam line at the Australian Synchrotron. 6TH INTERNATIONAL CONFERENCE ON MEDICAL APPLICATIONS OF SYNCHROTRON RADIATION, 2010. AIP Publishing, 3-9.

HERMAN, G. T. 1979. Image reconstruction from projections. Image Reconstruction from Projections: Implementation and Applications, 1.

HOSSAIN, Z., INDRARATNA, B., DARVE, F. & THAKUR, P. 2007. DEM analysis of angular ballast breakage under cyclic loading. Geomechanics and Geoengineering: An International Journal, 2, 175-181.

HSIEH, J. Computed tomography: principles, design, artifacts, and recent advances. 2009. SPIE Bellingham, WA.

HUTCHINSON, J. 1968. Plastic stress and strain fields at a crack tip. Journal of the Mechanics and Physics of Solids, 16, 337-342.

IASSONOV, P., GEBRENEGUS, T. & TULLER, M. 2009. Segmentation of X‐ray computed tomography images of porous materials: A crucial step for characterization and quantitative analysis of pore structures. Water Resources Research, 45.

IKEDA, S., NAKANO, T. & NAKASHIMA, Y. 2000. Three-dimensional study on the interconnection and shape of crystals in a graphic granite by X-ray CT and image analysis.

INDRARATNA, B., NGO, N. T., RUJIKIATKAMJORN, C. & VINOD, J. 2012. Behavior of Fresh and Fouled Railway Ballast Subjected to Direct Shear Testing: Discrete Element Simulation. International Journal of Geomechanics, 14, 34-44.

INDRARATNA, B., SALIM, W. & RUJIKIATKAMJORN, C. 2011. Advanced rail geotechnology–ballasted track, CRC press.

INDRARATNA, B., SUN, Q. D. & NIMBALKAR, S. 2014. Observed and predicted behaviour of rail ballast under monotonic loading capturing particle breakage. Canadian Geotechnical Journal, 52, 73-86.

INDRARATNA, B., THAKUR, P. K. & VINOD, J. S. 2010. Experimental and numerical study of railway ballast behavior under cyclic loading. International Journal of Geomechanics, 10, 136-144.

IOS 2006. 35 (2006) Reference materials—general and statistical principles for certification. International Organization for Standardization (ISO), Geneva, 64.

ITASCA CONSULTING GROUP, I. 2008. PFC2D (Particle Flow Code in Two Dimensions) version 4.0 Manual.

150

IWASHITA, K. & ODA, M. 1998. Rolling resistance at contacts in simulation of shear band development by DEM. Journal of engineering mechanics, 124, 285-292.

IWASHITA, K. & ODA, M. 1999. Mechanics of granular materials: an introduction, CRC Press. JAEGER, J. Failure of rocks under tensile conditions. International Journal of Rock Mechanics

and Mining Sciences & Geomechanics Abstracts, 1967. Elsevier, 219-227. JIA, X. & GARBOCZI, E. J. 2016. Advances in shape measurement in the digital world.

Particuology, 26, 19-31. JIANG, M., LIU, F., ZHU, F. & XIAO, Y. 2013. A Simplified Contact Model for Sandy Grains

Cemented with Methane Hydrate. Proceedings of 18th International Conference on Soil Mechanics and Geotechnical Engineering.

JIANZHUANG, L., WENQING, L. & YUPENG, T. Automatic thresholding of gray-level pictures using two-dimension Otsu method. Circuits and Systems, 1991. Conference Proceedings, China., 1991 International Conference on, 1991. IEEE, 325-327.

JONES, D. R. & ASHBY, M. F. 2005. Engineering materials 2: an introduction to microstructures, processing and design, Butterworth-Heinemann.

KAESTNER, A., LEHMANN, E. & STAMPANONI, M. 2008. Imaging and image processing in porous media research. Advances in Water Resources, 31, 1174-1187.

KATAGIRI, J., MATSUSHIMA, T. & YAMADA, Y. 2010. Simple shear simulation of 3D irregularly-shaped particles by image-based DEM. Granular Matter, 12, 491-497.

KATSAGA, T. 2010. Geophysical imaging and numerical modelling of fractures in concrete. University of Toronto.

KEPPLER, I., SAFRANYIK, F. & OLDAL, I. 2016. Shear test as calibration experiment for DEM simulations: a sensitivity study. Engineering Computations, 33, 742-758.

KETCHAM, R. A. & CARLSON, W. D. 2001. Acquisition, optimization and interpretation of X-ray computed tomographic imagery: applications to the geosciences. Computers & Geosciences, 27, 381-400.

KHALILI, Y. & MAHBOUBI, A. 2013. Discrete simulation and micromechanical analysis of two-dimensional saturated granular media. Particuology.

KHANAL, M., SCHUBERT, W. & TOMAS, J. 2005. DEM simulation of diametrical compression test on particle compounds. Granular Matter, 7, 83-90.

KIM, B. S., PARK, S. W. & KATO, S. 2012. DEM simulation of collapse behaviours of unsaturated granular materials under general stress states. Computers and Geotechnics, 42, 52-61.

KIM, H., WAGONER, M. P. & BUTTLAR, W. G. 2009. Numerical fracture analysis on the specimen size dependency of asphalt concrete using a cohesive softening model. Construction and Building Materials, 23, 2112-2120.

KRESS, J., YUN, T. S., NARSILIO, G. A., EVANS, T. M. & LEE, D.-S. 2012. Evaluation of hydraulic conductivity in 3D random and heterogeneous particulate materials using network model. Computers and Geotechnics, 40, 45-52.

KRUMBEIN, W. C., SLOSS, L. L. & DAPPLES, E. C. 1949. Sedimentary tectonics and sedimentary environments. AAPG Bulletin, 33, 1859-1891.

LAMBE, T. W. & WHITMAN, R. V. 2008. Soil mechanics SI version, John Wiley & Sons. LAWN, B. 1993. Fracture of brittle solids, Cambridge university press. LE PEN, L., POWRIE, W., ZERVOS, A., AHMED, S. & AINGARAN, S. 2013. Dependence of shape on

particle size for a crushed rock railway ballast. Granular Matter, 15, 849-861. LEE, J. S. 2007. Time-dependent crack growth in brittle rocks and field applications to geologic

hazards, ProQuest. LEIVA-VILLACORTA, F., VARGAS-NORDCBECK, A. & AGUIAR-MOYA, J. P. 2017. Permanent

deformation and deflection relationship from pavement condition assessment. International Journal of Pavement Research and Technology.

LEKARP, F., ISACSSON, U. & DAWSON, A. 2000a. State of the art. I: Resilient response of unbound aggregates. Journal of transportation engineering, 126, 66-75.

151

LEKARP, F., ISACSSON, U. & DAWSON, A. 2000b. State of the art. II: Permanent strain response of unbound aggregates. Journal of transportation engineering, 126, 76-83.

LESNIEWSKA, D. & WOOD, D. M. 2009. Observations of stresses and strains in a granular material. Journal of engineering mechanics, 135, 1038-1054.

LIM, W. L. & MCDOWELL, G. R. 2007. The importance of coordination number in using agglomerates to simulate crushable particles in the discrete element method. Géotechnique, 57, 701-705.

LIU, Y., DENG, A. & JAKSA, M. Discrete element modelling of geocell-reinforced track ballast under static and cyclic loading. 12th Australia New Zealand Conference on Geomechanics (ANZ 2015), 2015 Wellington, New Zealand.

LIU, Y., YOU, Z. & ZHAO, Y. 2012. Three-dimensional discrete element modeling of asphalt concrete: Size effects of elements. Construction and Building Materials, 37, 775-782.

LOBO-GUERRERO, S. 2006. Evaluation of Crushing in Granular Materilas using the Discrete Element Method and Fractal Theory, Ph.D. Dissertation, University of Pittsburgh.

LOBO-GUERRERO, S. & VALLEJO, L. E. 2005. Discrete element method evaluation of granular crushing under direct shear test conditions. Journal of Geotechnical and Geoenvironmental Engineering, 131, 1295-1300.

LU, M. & MCDOWELL, G. R. 2008. Discrete element modelling of railway ballast under triaxial conditions. Geomechanics and Geoengineering: An International Journal, 3, 257-270.

LUONG, M. Mechanical aspects and thermal effects of cohesionless soils under cyclic and transient loading. Proc. IUTAM Conf. on Deformation and Failure of Granular materials, Delft, 1982. 239-246.

MA, Z., MERKUS, H. G., DE SMET, J. G., HEFFELS, C. & SCARLETT, B. 2000. New developments in particle characterization by laser diffraction: size and shape. Powder Technology, 111, 66-78.

MAJMUDAR, T. S. & BEHRINGER, R. P. 2005. Contact force measurements and stress-induced anisotropy in granular materials. Nature, 435, 1079-1082.

MANGAN, A. P. & WHITAKER, R. T. 1999. Partitioning 3D surface meshes using watershed segmentation. Visualization and Computer Graphics, IEEE Transactions on, 5, 308-321.

MARSAL, R. J. Mechanical properties of rockfill: In Embankment-dam Engineering. Textbook. Eds. RC Hirschfeld and SJ Poulos. JOHN WILEY AND SONS INC., PUB., NY, 1973, 92P. International Journal of Rock Mechanics and Mining Sciences & Geomechanics Abstracts, 1975. Pergamon, 67.

MCANDREW, J. & MARSDEN, M. 1973. Regional guide to Victorian geology. MCDOWELL, G. 2002. Discrete element modelling of yielding and normal compression of sand.

Géotechnique, 52, 299-304. MCDOWELL, G., BOLTON, M. & ROBERTSON, D. 1996. The fractal crushing of granular materials.

Journal of the Mechanics and Physics of Solids, 44, 2079-2101. MEINHART, C. D., WERELEY, S. T. & SANTIAGO, J. G. 1999. PIV measurements of a microchannel

flow. Experiments in fluids, 27, 414-419. MIAO, G. 2015. The Breakage Behaviour of Irregularly–Shaped Particles. University of Sydney. MIAO, G. & AIREY, D. 2013. Breakage and ultimate states for a carbonate sand. Géotechnique,

63, 1221. MINDER, P. & PUZRIN, A. 2013. Microstructural changes leading to chemically enhanced

drainage. Proc. of the 18th ICSMGE, Paris. MISHRA, B. & THORNTON, C. 2001. Impact breakage of particle agglomerates. International

Journal of Mineral Processing, 61, 225-239. MITRA, R. & WESTMAN, E. 2009. Investigation of the stress imaging in rock samples using

numerical modeling and laboratory tomography. International Journal of Geotechnical Engineering, 3, 517-525.

152

MOHAMMADINIA, A. 2016. Effect of geopolymer cement stabilisation on fatigue life of pavement sub-bases with recycled demolition aggregates. PhD dissertation PhD, Swinburne University of Technology.

MOLLANOURI SHAMSI, M. M. & MIRGHASEMI, A. A. 2012. Numerical simulation of 3D semi-real-shaped granular particle assembly. Powder Technology, 221, 431-446.

MORA, C. & KWAN, A. 2000. Sphericity, shape factor, and convexity measurement of coarse aggregate for concrete using digital image processing. Cement and concrete research, 30, 351-358.

MUNJIZA, A. 2004. The combined finite-discrete element methods, Wiley Online Library. MUSZYNSKI, M. R. & VITTON, S. J. 2012. Particle shape estimates of uniform sands: visual and

automated methods comparison. J. Mater. Civ. Eng., 24, 194-206. NAKATA, A., HYDE, M., HYODO, H. & MURATA 1999. A probabilistic approach to sand particle

crushing in the triaxial test. Geotechnique, 49, 567-583. NARSILIO, G. A., KRESS, J. & YUN, T. S. 2010. Characterisation of conduction phenomena in soils

at the particle-scale: Finite element analyses in conjunction with synthetic 3D imaging. Computers and Geotechnics, 37, 828-836.

NING, Z. & EVANS, T. 2013. Discrete Element Method Study of Shear Wave Propagation in Granular Soil. Proc. of the 18th ICSMGE, Paris.

NOUGUIER-LEHON, C., CAMBOU, B. & VINCENS, E. 2003. Influence of particle shape and angularity on the behaviour of granular materials: a numerical analysis. International Journal for Numerical and Analytical Methods in Geomechanics, 27, 1207-1226.

O'SULLIVAN, C. 2011. Particulate discrete element modelling, Taylor & Francis. O'SULLIVAN, C. & BRAY, J. D. 2004. Selecting a suitable time step for discrete element

simulations that use the central difference time integration scheme. Engineering Computations, 21, 278-303.

O'SULLIVAN, C. & CUI, L. 2009. Micromechanics of granular material response during load reversals: Combined DEM and experimental study. Powder Technology, 193, 289-302.

O’DONOVAN, J., O’SULLIVAN, C. & MARKETOS, G. 2012. Two-dimensional discrete element modelling of bender element tests on an idealised granular material. Granular Matter, 14, 733-747.

OBERMAYR, M., DRESSLER, K., VRETTOS, C. & EBERHARD, P. 2013. A bonded-particle model for cemented sand. Computers and Geotechnics, 49, 299-313.

OH, W. & LINDQUIST, B. 1999. Image thresholding by indicator kriging. IEEE Transactions on Pattern Analysis and Machine Intelligence, 21, 590-602.

OSBORNE, C. R., CAYLEY, R. A., SKLADZIEN, P. B. & MCLEAN, M. A. 2010. Eastern Victoria Structural Zones 1: 250 000 Scale 3D Geological Model Report, Including Tabberabbera, Omeo, Deddock, Kuark and Mallacoota Zones.

OTANI, J., MUKUNOKI, T., TAKANO, D. & CHEVALIER, B. 2013. Microscopic observation on compacted sandy soil using micro-focus Xray CT. Proc. of the 18th ICSMGE, Paris.

PECK, W., NEILSON, J., OLDS, R. & SEDDON, K. 1992. Engineering geology of Melbourne. Balkema, Rotterdam.

PENG, B. 2014. Discrete Element Method (DEM) Contact Models Applied to Pavement Simulation. Virginia Tech.

PENUMADU, D., DUTTA, A. K., LUO, X. & THOMAS, K. G. 2009. Nano and neutron science applications for geomechanics. KSCE Journal of Civil Engineering, 13, 233-242.

PHUSING, D. & SUZUKI, K. 2015. Cyclic behaviors of granular materials under generalized stress condition using DEM. Journal of Engineering Mechanics, 141, 04015034.

PITCHUMANI, R., ZHUPANSKA, O., MEESTERS, G. M. & SCARLETT, B. 2004. Measurement and characterization of particle strength using a new robotic compression tester. Powder Technology, 143, 56-64.

153

POTYONDY, D. O. & CUNDALL, P. A. 2004. A bonded-particle model for rock. International Journal of Rock Mechanics and Mining Sciences, 41, 1329-1364.

POTYONDY, D. O., CUNDALL, P. A. & LEE, C. A. Modelling rock using bonded assemblies of circular particles. 2nd North American Rock Mechanics symposium, 1996. American Rock Mechanics Association.

QUINN, G. D. 2007. NIST Recommended practice guide: fractography of ceramics and glasses. Washington, DC: National Institute of Standards and Technology, 960-16.

RAHMAN, M. A., ARULRAJAH, A., PIRATHEEPAN, J., BO, M. W. & IMTEAZ, M. A. 2014. Resilient Modulus and Permanent Deformation Responses of Geogrid-Reinforced Construction and Demolition Materials. Journal of Materials in Civil Engineering, 26, 512-519.

RAPAPORT, D. C. 2004. The art of molecular dynamics simulation, Cambridge university press. REBUFFEL, V. & DINTEN, J.-M. 2007. Dual-energy X-ray imaging: benefits and limits. Insight-non-

destructive testing and condition monitoring, 49, 589-594. RITTINGER, P. 1867. Lehrbuch der Aufbereitungskunde. ROZENBLAT, Y., PORTNIKOV, D., LEVY, A., KALMAN, H., AMAN, S. & TOMAS, J. 2011. Strength

distribution of particles under compression. Powder Technology, 208, 215-224. RUSSELL, A. R. & EINAV, I. 2013. Energy dissipation from particulate systems undergoing a single

particle crushing event. Granular Matter, 15, 299-314. SALMAN, A. D., HOUNSLOW, M. & SEVILLE, J. P. 2006. Granulation, Elsevier. SALOT, C., GOTTELAND, P. & VILLARD, P. 2009. Influence of relative density on granular materials

behavior: DEM simulations of triaxial tests. Granular Matter, 11, 221-236. SANTAMARINA, J. & CHO, G. Soil behaviour: The role of particle shape. Advances in geotechnical

engineering: The skempton conference, 2004. Thomas Telford, 604-617. SAZZAD, M. M. & SUZUKI, K. 2010. Micromechanical behavior of granular materials with

inherent anisotropy under cyclic loading using 2D DEM. Granular Matter, 12, 597-605. SCHINDELIN, J., ARGANDA-CARRERAS, I., FRISE, E., KAYNIG, V., LONGAIR, M., PIETZSCH, T.,

PREIBISCH, S., RUEDEN, C., SAALFELD, S. & SCHMID, B. 2012. Fiji: an open-source platform for biological-image analysis. Nature Methods, 9, 676-682.

SHAFIPOUR, R. & SOROUSH, A. 2008. Fluid coupled-DEM modelling of undrained behavior of granular media. Computers and Geotechnics, 35, 673-685.

SHAMY, U. E. & ZEGHAL, M. 2007. A micro-mechanical investigation of the dynamic response and liquefaction of saturated granular soils. Soil Dynamics and Earthquake Engineering, 27, 712-729.

SIBILLE, L., DONZÉ, F. V., NICOT, F., CHAREYRE, B. & DARVE, F. 2007. From bifurcation to failure in a granular material: a DEM analysis. Acta Geotechnica, 3, 15-24.

SIVAKUGAN, N., ARULRAJAH, A. & BO, M. W. 2011. Laboratory testing of soils, rocks, and aggregates, J. Ross Publishing.

SNIGIREV, A., SNIGIREVA, I., KOHN, V., KUZNETSOV, S. & SCHELOKOV, I. 1995. On the possibilities of x‐ray phase contrast microimaging by coherent high‐energy synchrotron radiation. Review of scientific instruments, 66, 5486-5492.

SOILLE, P. 2013. Morphological image analysis: principles and applications, Springer Science & Business Media.

SOLANKI, P. & ZAMAN, M. 2012. Microstructural and mineralogical characterization of clay stabilized using calcium-based stabilizers, INTECH Open Access Publisher.

STEVENSON, A. W., MAYO, S. C., HÄUSERMANN, D., MAKSIMENKO, A., GARRETT, R. F., HALL, C. J., WILKINS, S. W., LEWIS, R. A. & MYERS, D. E. 2010. First experiments on the Australian Synchrotron Imaging and Medical beamline, including investigations of the effective source size in respect of X-ray imaging. Journal of synchrotron radiation, 17, 75-80.

STOCK, S. 1999. X-ray microtomography of materials. International Materials Reviews, 44, 141-164.

154

STUTZMAN, P. E. & CLIFTON, J. R. Specimen preparation for scanning electron microscopy. Proceedings of the International Conference on Cement Microscopy, 1999. Citeseer, 10-22.

SUN, Y., INDRARATNA, B. & NIMBALKAR, S. 2014. Three-dimensional characterisation of particle size and shape for ballast. Géotechnique Letters, 4, 197-202.

SUSTAINABILITYVIC 2010. Victorian recycling industries annual report 2008-2009. ISSN 1836. Sustainability Victoria.

TAKAHASHI, M., TAKEMURA, T. & ODA, M. 2004. Microstructure in shear band observed by microfocus X-ray computed tomography. Geotechnique, 54, 539-542.

TAKEI, M., KUSAKABE, O. & HAYASHI, T. 2001. Time-dependent behavior of crushable materials in one-dimensional compression tests. Soils and Foundations, 41, 97-121.

TAM, V. W. & TAM, C. 2007. Crushed aggregate production from centralized combined and individual waste sources in Hong Kong. Construction and Building Materials, 21, 879-886.

TATTERSALL, H. G. & TAPPIN, G. 1966. The work of fracture and its measurement in metals, ceramics and other materials. Journal of Materials Science, 1, 296-301.

TAVARES, L. & DAS NEVES, P. B. 2008. Microstructure of quarry rocks and relationships to particle breakage and crushing. International Journal of Mineral Processing, 87, 28-41.

TAVARES, L. & KING, R. 1998. Single-particle fracture under impact loading. International Journal of Mineral Processing, 54, 1-28.

TERZAGHI, K. V. 1920. Old earth pressure theories and new test results. Engineering News-Record, 85, 632-637.

THAKUR, P. K. 2011. Cyclic densification of ballast and associated deformation and degradation. PhD Dissertation, University of Wollongong.

THOMAS, P. A. & BRAY, J. D. 1999. Capturing nonspherical shape of granular media with disk clusters. Journal of Geotechnical and Geoenvironmental Engineering, 125, 169-178.

THORNTON, C., CIOMOCOS, M. & ADAMS, M. 2004. Numerical simulations of diametrical compression tests on agglomerates. Powder Technology, 140, 258-267.

TOMAC, I. & GUTIERREZ, M. 2013. Particulate Modeling of Sand Slurry Flow Retardation. Proc. of the 18th ICSMGE, Paris.

TSOUNGUI, O., VALLET, D. & CHARMET, J.-C. 1999. Numerical model of crushing of grains inside two-dimensional granular materials. Powder technology, 105, 190-198.

TURCOTTE, D. 1986. Fractal and fragmentation. Journal of Geography Research, 91, 1921-1926. VIGGIANI, G. 2013. Geomechanics through the scales. Proceedings of the 18th International

Conference on Soil Mechanics and Geotechnical Engineering. Paris. VIGGIANI, G., BÉSUELLE, P., HALL, S. & DESRUES, J. 2010. Sand Deformation at the Grain Scale

Quantified Through X‐Ray Imaging. Advances in Computed Tomography for Geomaterials: GeoX 2010, 1-16.

VIGGIANI, G. & HALL, S. A. 2004. Full-field measurements in experimental geomechanics: historical perspective, current trends and recent results. Advanced experimental techniques in geomechanics, 3-67.

VIGGIANI, G., LENOIR, N., BÉSUELLE, P., DI MICHIEL, M., MARELLO, S., DESRUES, J. & KRETZSCHMER, M. 2004. X-ray microtomography for studying localized deformation in fine-grained geomaterials under triaxial compression. Comptes Rendus Mécanique, 332, 819-826.

VIGGIANI, G., TAKANO, D., ANDÒ, E. & SANTAMARINA, J. C. 2015. Laboratory X-ray tomography: A valuable experimental tool for revealing processes in soils. ASTM geotechnical testing journal.

WADELL, H. 1933. Sphericity and roundness of rock particles. The Journal of Geology, 310-331. WADELL, H. 1935. Volume, shape, and roundness of quartz particles. The Journal of Geology,

250-280.

155

WANG, J. & YAN, H. 2012. DEM analysis of energy dissipation in crushable soils. Soils and Foundations, 52, 644-657.

WANG, M., YANG, R. & YU, A. 2012. DEM investigation of energy distribution and particle breakage in tumbling ball mills. Powder technology, 223, 83-91.

WEINER, B. B. 1984. Particle and droplet sizing using Fraunhofer diffraction. Modern methods of particle size analysis, 135-172.

WESTENBERGER, P. AVIZO-3D visualization framework. Geoinformatics Conference, 2008. 1-11. WHITE, D., TAKE, W. & BOLTON, M. 2003. Soil deformation measurement using particle image

velocimetry (PIV) and photogrammetry. Geotechnique, 53, 619-631. WILDENSCHILD, D., VAZ, C., RIVERS, M., RIKARD, D. & CHRISTENSEN, B. 2002. Using X-ray

computed tomography in hydrology: systems, resolutions, and limitations. Journal of Hydrology, 267, 285-297.

WILLIAMS, R. A. & JIA, X. 2003. A new method for prediction of bulk particle packing behavior for arbitrary-shaped particles in containers of any shape. Particulate science and technology, 21, 195-205.

WU, C.-Y., LI, L.-Y. & THORNTON, C. 2005. Energy dissipation during normal impact of elastic and elastic–plastic spheres. International Journal of Impact Engineering, 32, 593-604.

XIN, L. 2013. Numerical Simulation of Granular Motion by Particle Flow Code under Vibration. YAMAMURO, J. A. & LADE, P. V. 1996. Drained sand behavior in axisymmetric tests at high

pressures. Journal of Geotechnical Engineering, 122, 109-119. YAN, G., YU, H.-S. & MCDOWELL, G. 2009. Simulation of granular material behaviour using DEM.

Procedia Earth and Planetary Science, 1, 598-605. YIGIT, I. & CINICIOGLU, S. 2013. A look into time dependent behaviour of clays at macro and

micro scale. Proc. of the 18th ICSMGE, Paris. YOON, J. 2007. Application of experimental design and optimization to PFC model calibration in

uniaxial compression simulation. International Journal of Rock Mechanics and Mining Sciences, 44, 871-889.

YU, H. & SHEN, S. 2012. Impact of aggregate packing on dynamic modulus of hot mix asphalt mixtures using three-dimensional discrete element method. Construction and Building Materials, 26, 302-309.

ZEGHAL, M. 2004. Discrete-Element Method Investigation of the Resilient Behavior of Granular Materials. Journal of Transportation Engineering, 130, 7.

ZEGHAL, M. & EDIL, T. B. 2002. Soil structure interaction analysis: modeling the interface. Canadian Geotechnical Journal, 39, 620-628.

ZHAO, B., CHOW, C. Y. D. & WANG, J. An Investigation of Single Particle Crushing Tests on Irregular-Shaped Quartz Sand with DEM. Proceedings of the 7th International Conference on Discrete Element Methods, 2017. Springer, 217-224.

ZHAO, B., WANG, J., COOP, M., VIGGIANI, G. & JIANG, M. 2015. An investigation of single sand particle fracture using X-ray micro-tomography. Géotechnique, 65, 625-641.

ZHENG, W. & TANNANT, D. 2016. Frac sand crushing characteristics and morphology changes under high compressive stress and implications for sand pack permeability. Canadian Geotechnical Journal, 53, 1412-1423.

ZHU, H., ZHOU, Z., YANG, R. & YU, A. 2007. Discrete particle simulation of particulate systems: theoretical developments. Chemical Engineering Science, 62, 3378-3396.