Delft University of Technology

Excavation of hard deposits and rocks

On the cutting of saturated rock
Helmons, Rudy

DOI
10.4233/uuid:7a46bca3-4105-4cdc-952d-a6d9fcfced76
Publication date
2017
Document Version
Final published version
Citation (APA)
Helmons, R. (2017). Excavation of hard deposits and rocks: On the cutting of saturated rock. [Dissertation
(TU Delft), Delft University of Technology]. https://doi.org/10.4233/uuid:7a46bca3-4105-4cdc-952d-
a6d9fcfced76

Important note
To cite this publication, please use the final published version (if applicable).
Please check the document version above.

Copyright
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent
of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Takedown policy
Please contact us and provide details if you believe this document breaches copyrights.
We will remove access to the work immediately and investigate your claim.

This work is downloaded from Delft University of Technology.

For technical reasons the number of authors shown on this cover page is limited to a maximum of 10.
Excavation of Hard Deposits and Rocks
On the Cutting of Saturated Rock
Excavation of Hard Deposits and Rocks
On the Cutting of Saturated Rock

Proefschrift

ter verkrijging van de graad van doctor

aan de Technische Universiteit Delft,
op gezag van de Rector Magnificus prof. ir. K. C. A. M. Luyben,
voorzitter van het College voor Promoties,
in het openbaar te verdedigen op woensdag 10 mei 2017 om 10.00 uur

door

Rudolfus Lambertus Jacobus Helmons

werktuigbouwkundig ingenieur
geboren te Halsteren, Nederland.
Dit proefschrift is goedgekeurd door de promotor:
Prof. dr. ir. C. van Rhee
en copromotor:
Dr. ir. S. A. Miedema
Samenstelling promotiecommissie:
Rector Magnificus, voorzitter
Prof. dr. ir. C. van Rhee, Technische Universiteit Delft, promotor
Dr. ir. S. A. Miedema, Technische Universiteit Delft, copromotor

Onafhankelijke leden:
Prof. dr. ir. M. L. Kaminski Technische Universiteit Delft
Prof. dr. ir. L.J. Sluys Technische Universiteit Delft
Prof. dr. rer. nat. S. Luding Universiteit Twente
Prof. dr. C. Drebenstedt Technische Universität Bergakademie Freiberg
Prof. dr. E. M. Detournay, University of Minnesota
The research presented in this thesis has been financed by Agentschap NL under
grant agreement IMA1100007, together with Royal Dutch Shell, Royal Boskalis
Westminster, Van Oord Dredging and Marine Contractors and Royal IHC.

Keywords: Rock mechanics, Discrete Element Method, Smoothed Particle,

Rock cutting
Printed by: Ridderprint
Front & Back: Resultant outflow of a lava flow under water, located near the Axial
Volcano (300 miles offshore from Washington state (USA). Credit:
NSF-OOI/UW/CSSF

Copyright c 2017 by R.L.J. Helmons

ISBN 978-94-6186-790-2
An electronic version of this dissertation is available at
http://repository.tudelft.nl/.
http://www.researchgate.net/profile/Rudy_Helmons.
 To Corine
and the little boy
 Summary
As a result of the worldwide population and welfare growth, the demand for energy
(oil, gas and renewable sources) and raw materials increases. In the last decades,
oil and gas are produced from more and more offshore sites and deeper waters.
Besides energy, the demand for diverse metals and rare earth elements increases as
well. These raw materials are often at the basis of new sustainable technologies
e.g. permanent magnets for wind energy and battery packs for electric cars. The
availability of these raw materials is essential for a stable development of the world
economy. Unfortunately, for some of the crucial raw materials, the availability is
sometimes very local and in various cases there is a monopoly forming. To reduce
this economic risk, investments are needed to search and extract minerals from
new locations. Large, metal-rich fields are found at the bottom of the sea, such
as phosphate nodules, manganese nodules, cobalt-rich crusts and vulcanic sulphide
deposits (often referred to as Seafloor Massive Sulphide, SMS). These deposits are
mainly located in the deep sea, at depths ranging from several hundreds of meters
to several kilometers.
One of the technical challenges to enable production from these locations is the
cutting or excavation process. Experiments have shown that the energy needed to
excavate the material increases with water depth. Besides that, it is demonstrated
that rock that fails brittle in atmospheric conditions can fail more or less in a plastic
fashion when present in a high pressure environment, as would be the case at large
water depths. The goal of this research is to identify the physics of the cutting
process and to develop this into a model in which the effect of hydrostatic and pore
pressures is included.
The cutting of rock is initiated by pressing a tool into the rock. As a result, at
the tip of the tool a high compressive pressure occurs, which leads to the formation
of a crushed zone. Depending on the shape of the tool and the cutting depth,
shear failures might emanate from the crushed zone, which will eventually expand
as tensile fractures that can reach to the free rock surface. Through this process
intact rock will be disintegrated to a granular medium. Additionally, the presence
of water in the pores of and surrounding the rock influences the cutting process
through drainage effects. The most relevant effects are weakening when compaction
and hardening when dilation occurs in shearing and tension. Deformation of the
rock causes the pore volume to change, resulting in a under or over pressure. As a
result, the pore fluid needs to flow. The magnitude of the potential under pressure
is limited through cavitation of the pore fluid, limiting further reduction of the
pore pressure. The drainage effects cause the rock cutting process in a submerged
environment to show a stronger dependency of both the hydrostatic pressure as well
as the deformation rate.
The numerical simulations are performed with a 2D DEM (Discrete Element

vii
viii Summary

Method). In DEM, the mechanical behavior of a rock is mimicked by gluing loose

particles together with brittle bonds. Such a method shows strong resemblance with
sedimentary rock. In order to include the effect of an ambient pressure as a result
of the water depth and to include the presence of a fluid in the pores of the rock,
a pore pressure diffusion equation is added to the model. The discontinuous results
obtained with DEM are interpolated to a continuum field through the use of a SP-
method (Smoothed Particles). Additionally, SP is used to solve the pore pressure
diffusion equation. For that reason, the methodology used in this dissertation is
referred to as DEM-SP.
Thus far no direct coupling has been found between the input microscopic pa-
rameters, that define the properties of and interactions between the particles in
DEM, and the resulting bulk properties of the particle assembly. For that reason, a
sensitivity analysis is performed in which the effect of the micro-properties on the
macroscopic behavior is investigated. Additionally it is proven that the addition
of the pore pressure diffusion process to the DEM-SP model corresponds with the
effective stress theory. It is also proven that when air is used as a medium in the
pores, no significant changes compared to simulations without pore pressure cou-
pling occur. Comparison of the numerical model with a set of tri-axial experiments
on shale, in which the deformation rate is varied, shows that the model is well
capable to describe both compaction weakening and dilatant hardening.
In order to further validate DEM-SP, several experiments from literature are
simulated. A comparison of 2D cutting experiments on tiles shows a good match
for the chip size, chip shape and the required cutting force. DEM-SP is used to
simulated drilling experiments on marble, in which the hydrostatic pressure is var-
ied. These results show that the simulated behavior of the cutting process matches
qualitatively with the experiments, i.e. the trend of increasing cutting force with
increasing hydrostatic pressure. Furthermore a series of cutting experiments for the
purpose of deep sea mining has been simulated. These results match both qual-
itatively and quantitatively. Additionally, both the experiments and simulations
show the existence of a hyperbaric effect. This means that at a hydrostatic pressure
which is significantly larger than the tensile strength of the rock the cutting pro-
cess shear and cataclastic failure are more dominant, while at hydrostatic pressures
significantly smaller than the tensile strength the cutting process is dominated by
tensile failure and chipforming.
Finally, DEM-SP is used to simulate the full cutting motion of a pick point on a
rotating cutterhead, in order to investigate the applicability of the method to shallow
water depths (<30 m) and to investigate the use of the method for the dredging
practice. Even at shallow water depths the effect of an increased hydrostatic pressure
shows significant differences. Furthermore, the simulations show a transition from
cataclastic towards ductile cutting process based on the cutting depth. Additionally
a transition between stick-slip friction of the cut material along the tool is observed,
which is an indication for different wear processes.
It is proven that DEM-SP is capable of solving drainage related effects in defor-
mation of saturated rock. A range of rock cutting experiments are simulated and
the results match well both qualitatively and quantitatively with respect to cutting
Summary ix

force and hydrostatic pressure. Further improvement of the model can be achieved
by extending the model towards 3D.
 Samenvatting
Door de wereldwijde bevolkings- en welvaartsgroei neemt de vraag naar energie (olie,
gas en herwinbare bronnen) en grondstoffen toe. In de laatste decennia worden olie
en gas steeds meer uit de kust en in steeds diepere wateren gewonnen. Naast energie
is er ook een toename in de vraag naar diverse metalen en zeldzame aardelementen
toegenomen. Deze grondstoffen liggen veelal aan de basis van de nieuwe duurzame
technologiën zoals bijvoorbeeld permanent magneten voor windmolens en accu’s
voor elektrische auto’s. De beschikbaarheid van deze grondstoffen is essentieel voor
een stabiele ontwikkeling van de wereldeconomie. Helaas is de beschikbaarheid van
enkele cruciale grondstoffen soms zeer lokaal en in een aantal gevallen is er sprake
van monopolie vorming. Om dit economisch risico te beperken moet er ge investeerd
worden in het vinden en ontginnen van nieuwe locaties. Op de zeebodem zijn grote
metaalrijke velden gevonden van o.a. fosfaatknollen, mangaanknollen, kobaltrijke
korsten en vulkanische sulfide afzettingen. Deze locaties bevinden zich voornamelijk
in de diepzee, op dieptes van enkele honderden meters tot meerdere kilometers.
Een van de technische uitdagingen om winning uit deze gebieden mogelijk te
maken is het snij- of ontgravingsproces. Experimenten hebben uitgewezen dat de
benodigde energie om het materiaal te ontginnen toe neemt met de waterdiepte.
Daarnaast is gebleken dat gesteente dat bros kapot gaat bij een atmosferische druk
min of meer plastisch kan bezwijken bij een hoge omgevingsdruk, zoals het geval is
op grote waterdiepte. Het doel van dit onderzoek is om de fysica van het snijproces
in kaart te brengen en dit uit te werken tot een model waarin het effect van de
hydrostatische druk wordt meegenomen.
Het snijden van gesteente wordt geïnitieerd doordat het gereedschap in de steen
gedrukt wordt, waarbij aan de punt van het gereedschap een hoge drukspanning
ontstaat, wat leidt tot het ontstaan van een vergruizingszone. Afhankelijk van de
vorm van het gereedschap en de snedediepte kunnen vanuit de vergruizingszone af-
schuivingsbreuken ontstaan, welke naar het oppervlak uitgroeien als trekscheuren.
Door middel van dit proces wordt een intacte steen opgebroken tot een granulair
medium. Daarnaast be invloedt de aanwezigheid van water in de pori en van en
rond het gesteente het verspaningsproces door drainage effecten. De voornaamste
effecten hiervan zijn verzwakking door compactie en versteviging door dilatantie
bij afschuiving en trek. Vervorming van het gesteente zorgt ervoor dat het volume
van de pori en verandert, waardoor een onder- of bovendruk ontstaat. Het gevolg
is dat de porievloeistof zal moeten toe- of afstromen. Het grootte van de even-
tuele onderdruk wordt beperkt door het caviteren van de porievloeistof, waardoor
verdere afname van de poriedruk wordt tegengegaan. De drainage effecten zorgen
ervoor dat het steensnijproces onder water een sterkere afhankelijkheid van zowel
de hydrostatische druk als de vervormingssnelheid heeft.
De numerieke simulaties zijn uitgevoerd met een 2D DEM (discrete elementen

xi
xii Samenvatting

methode) model. Hierin wordt het gedrag van gesteente nagebootst door losse
korrels met behulp van brosse bindingen aan elkaar te plakken, wat een sterke
analogie heeft met sedimentair gesteente. Om het effect van een omgevingsdruk ten
gevolge van de waterdiepte en om de aanwezigheid van een vloeistof in de pori en
van het gesteente te simuleren is een poriedruk-diffusie vergelijking aan het model
toegevoegd. De discontinue resultaten van DEM is met behulp van een SP-methode
(smoothed particles, letterlijk vertaald gladgestreken deeltjes) geïnterpoleerd naar
een continuumsveld. Tevens is SP gebruikt om de poriedruk-diffusievergelijking op
te lossen. Vandaar dan ook dat naar deze methode wordt gerefereerd als DEM-SP.
Tot op heden is er nog geen directe koppeling gevonden tussen de op te geven
micro-parameters voor de eigenschappen van en de interacties tussen de korrels in
DEM en de daaruit resulterende eigenschappen van de korrels als bulk. Om die reden
is een gevoeligheidsanalyse gemaakt van de invloed van de micro-eigenschappen op
het macro-gedrag. Daarnaast is aangetoond dat het toevoegen van een poriedruk-
diffusie proces aan het DEM-SP model overeenkomt met de theorie van effectieve
spanning. Tevens is bewezen dat wanneer lucht als medium in de pori en gebruikt
wordt dit geen significante verschillen oplevert ten opzichte van simulaties zonder
poriedruk-koppeling. Vergelijking van het numerieke model met een serie tri-axiaal
proeven op schaliegesteente, waarbij de vervormingssnelheid sterk is gevarieerd laat
zien dat het model goed in staat is om compactie verzwakking dilatantie versteviging
te beschrijven.
Om DEM-SP te valideren zijn diverse experimenten uit literatuur nagerekend.
Vergelijking van 2D snijproeven op tegels laten zien dat zowel de brokvorm, brok-
grootte als de benodigde snijkracht overeen komen. Daarnaast is DEM-SP gebruikt
om experimenten van het boorproces in marmer waarbij hydrostatische druk is ge-
varieerd te simuleren. Deze resultaten laten zien dat kwalitatief het gesimuleerde
gedrag overeenkomt met de experimenten, de trend van de benodigde snijkracht
t.o.v. de hydrostatische druk komt overeen. Tevens is er een serie snijproeven ten
behoeve van diepzeemijnbouw, waarbij zowel de hydrostatische druk als de snij-
snelheid zijn gevarieerd, gesimuleerd. Zowel kwalitatief als kwantitatief komen de
resultaten overeen. Verder illustreren zowel de experimenten als de simulaties het
bestaan van een hyperbaar effect. Hiermee wordt bedoeld dat bij een hydrostati-
sche druk die significant groter is dan de treksterkte van het gesteente het snijpro-
ces wordt gedomineerd door afschuiving en vergruizing, terwijl voor hydrostatische
drukken significant lager dan de treksterkte het snijproces gedomineerd wordt door
falen op trek en brokvorming.
Tot slot is DEM-SP gebruikt om de volledige snijbeweging van een tand op een
roterende cutterkop te simuleren en om de toepassing van de methode voor ondiepe
wateren (<30 m) te onderzoeken. Ook op deze kleine waterdiepten is het effect
van een verhoogde hydrostatische druk significant aanwezig. Verder tonen de si-
mulaties een transitie van een cataclastisch naar een bros snijproces op basis van
de snedediepte. Tevens is er een overgang tussen kleef-slip wrijving van het gesne-
den materiaal langs de beitel waargenomen, wat een indicatie geeft voor mogelijke
slijtage.
Het is aangetoond dat DEM-SP in staat is om drainage gerelateerde effecten in
Samenvatting xiii

verzadigd gesteente op te lossen. Een scala aan rots snijproeven zijn gesimuleerd
en de resultaten komen zowel kwalitatief als kwantitatief met de snijkrachten en
hydrostatische druk overeen. Verdere verbetering van het model is mogelijk door
het model uit te breiden naar 3D.
 Contents
Summary vii
Samenvatting xi
List of Symbols xix
1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Research objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3 Outline of this thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2 Physics of Rock in Relation to Rock Cutting Process 9
2.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 Rock Failure Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.1 Failure modes . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.2 Brittle Failure Modes . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.3 Brittle-Ductile Transition . . . . . . . . . . . . . . . . . . . . 13
2.2.4 Ductile failure mode (cataclasis) . . . . . . . . . . . . . . . . . 14
2.2.5 Post-failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.6 Grain size effects . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2.7 Strain rate effects . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.8 Specimen size effects . . . . . . . . . . . . . . . . . . . . . . . 17
2.3 Fluid Saturated Rock. . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.1 Physico-chemical effects . . . . . . . . . . . . . . . . . . . . . 18
2.3.2 Hydro-mechanical effects . . . . . . . . . . . . . . . . . . . . . 19
2.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3 Rock Cutting Process 27
3.1 Phenomenological rock cutting model . . . . . . . . . . . . . . . . . . 28
3.1.1 General concepts . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.1.2 Dredging and seabed mining . . . . . . . . . . . . . . . . . . . 29
3.1.3 Drilling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.1.4 Effect of geometry . . . . . . . . . . . . . . . . . . . . . . . . 30
3.1.5 Effect of rock properties . . . . . . . . . . . . . . . . . . . . . 39
3.1.6 Effect of cutting speed . . . . . . . . . . . . . . . . . . . . . . 41
3.1.7 Effect of hydrostatic pressure . . . . . . . . . . . . . . . . . . 42
3.1.8 Identification of other parameters . . . . . . . . . . . . . . . . 46
3.2 Analytic rock cutting models . . . . . . . . . . . . . . . . . . . . . . 47
3.3 Discussion rock cutting models . . . . . . . . . . . . . . . . . . . . . 55
3.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

xv
xvi Contents

4 Modeling Approach 57
4.1 Numerical modeling of rock cutting . . . . . . . . . . . . . . . . . . . 58
4.1.1 Continuum based modeling . . . . . . . . . . . . . . . . . . . 58
4.1.2 Discontinuum based modeling . . . . . . . . . . . . . . . . . . 59
4.2 Hydromechanical coupling . . . . . . . . . . . . . . . . . . . . . . . . 59
4.3 Modeling Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.4 Solid Modeling - Discrete Element Method . . . . . . . . . . . . . . . 64
4.4.1 Constitutive model . . . . . . . . . . . . . . . . . . . . . . . . 65
4.5 Fluid Modeling - Smoothed Particle . . . . . . . . . . . . . . . . . . . 68
4.5.1 Boundary conditions and detection . . . . . . . . . . . . . . . 72
4.6 Initial geometry generation. . . . . . . . . . . . . . . . . . . . . . . . 73
4.7 Smoothed particle averaging of discrete elements . . . . . . . . . . . . 74
4.8 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5 Validation - Material tests 79
5.1 Parameter sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.1.1 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.1.2 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.2 Fluid stress effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.2.1 Effective stress . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.2.2 Strain rate effects . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.2.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
6 Validation - Tool-Rock Interaction 103
6.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.2 2D tile cutting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.2.1 Experiments. . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.2.2 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
6.3 Drilling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.3.1 Experiments. . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.3.2 Simulated results . . . . . . . . . . . . . . . . . . . . . . . . . 110
6.3.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6.4 Dredging and (Deep) Seabed Mining . . . . . . . . . . . . . . . . . . 116
6.4.1 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
6.4.2 Comparison of numerical and experimental results . . . . . . . 126
6.4.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
6.5 Testcase: rotational cutting . . . . . . . . . . . . . . . . . . . . . . . 129
6.5.1 Simulation setup . . . . . . . . . . . . . . . . . . . . . . . . . 129
6.5.2 Simulated results . . . . . . . . . . . . . . . . . . . . . . . . . 129
6.5.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
6.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
7 Conclusions and Recommendations 139
7.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
7.2 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
Contents xvii

A Practical Rock Properties 145

Bibliography 147
List of Publications 163
Curriculum Vitæ 165
Acknowledgements 167
 List of Symbols

Roman symbol Description Unit

A Fictional parameter [-]
a Surface area [m2 ]
Cf Rock fabric compressibility [Pa-1 ]
Cl Liquid compressibility [Pa-1 ]
Cp Pore fluid compressibility [Pa-1 ]
Cs Compressibility of solid [Pa-1 ]
C0 Constant depending on particle packing [-]
c Cohesion [Pa]
c1,2,... Constant [-]
D (hydraulic) Diffusion coefficient [m2 /s]
D Damage parameter [-]
d Specimen diameter [m]
E Young’s modulus [Pa]
Esp Specific energy [Pa]
E0 Effective Young’s modulus of pack of grains [GPa]
Etot Total Energy [J]
F Force [N]
Fc Cutting force [N]
Fh Horizontal component of cutting force [N]
Fn Normal force [N]
Fs Shear force [N]
Fv Vertical component of cutting force [N]
fs Sampling frequency [Hz]
g Crack extension force [N]
G Mechanical energy release rate [N]
h Smoothing length [m]
htool Height of tool [m]
I Moment of inertia [kgm2 ]
i Particle index [-]
j Neighoring particle index [-]
Kf Bulk modulus of fluid [Pa]
Khyd Hydraulic conductivity [m/s]
Km Bulk modulus of rock matrix [Pa]
Ks Bulk modulus of solids [Pa]
KIc Fracture toughness [Pa]
K1 Grain force on shear plane [N]

xix
xx List of Symbols

Roman symbol Description Unit

K2 Grain force on tool surface [N]
kn Normal stiffness [N/m2 ]
ks Shear stiffness [N/m2 ]
l Length [m]
m Particle mass [kg]
m Strength ratio [-]
N Deborah number [-]
Ncr Critical Deborah number [-]
n Porosity [-]
n Normal vector [-]
nc Number of contacting neighbors [-]
p Pressure [Pa]
ph Hydrostatic pressure [Pa]
pi Initial pore pressure [Pa]
Pc Cutting power [W]
pundr Undrained pore pressure [Pa]
Qc Cutting production [m3 ]
Qr Relieved cutting production [m3 ]
Qu Unrelieved cutting production [m3 ]
q Fluid flux [m/s]
R Dimensionless distance (SP) [-]
R0 Surface tension force [N]
r̄ Mean particle radius [m]
rmax Maximum particle radius [m]
rmin Minimum particle radius [m]
Si Initial saturation degree [-]
s Specific storage capacity [-]
st Tool spacing [m]
T Torque [Nm]
Tn Normal bond strength [N/m]
Ts Shear bond strength [N/m]
t Time [s]
~t Tangential vector [-]
tc Cutting depth [m]
tdef Characteristic time of deformation [s]
thd Characteristic time of hydraulic diffusion [s]
tsim Simulated timespan [s]
Ua Surface energy [J]
Ue External energy [J]
Ui Internal energy [J]
u Position vector [m]
u̇ Velocity vector [m/s]
ü Acceleration vector [m/s2 ]
u Unknown parameter [-]
List of Symbols xxi

Roman symbol Description Unit

v Gas fraction adsorbed by solids [-]
vc Cutting velocity [m/s]
W Kernel function [-]
W1 Force resulting from pore under pressure on [N]
shear plane
W2 Force resulting from pore under pressure on [N]
tool surface
w Tool width [m]
z Water depth [m]

Greek symbol Description Unit

α Rake angle with respect to vertical [◦ ]
αd Numerical damping coefficient [-]
αes Effective stress coefficient [-]
β Shear angle [◦ ]
δ External friction angle [◦ ]
 Strain [-]
˙ Strain rate [1/s]
 Strain [-]
˙ Strain rate [s-1 ]
˙cr Critical strain rate [s-1 ]
V Volumetric strain [-]
f Strain to failure [-]
ζ Fluid content [-]
η Dynamic viscosity [Pas]
κ Intrinsic permeability [m2 ]
κ0 Intrinsic permeability at zero effective stress [m2 ]
λ Stress distribution factor [-]
µ Coulomb friction (grain-grain) [-]
µtg Coulomb friction (tool-grain) [-]
ν Poisson’s ratio [-]
ξP e Pore Peclet number [-]
ρ Density [kg/m3 ]
σ Stress [Pa]
σBT S Tensile strength, determined as Brazilian disc [Pa]
σc Compressive strength [Pa]
σcd Compressive strength of specimen with diam- [Pa]
eter d
σc50 Compressive strength of specimen with diam- [Pa]
eter of 50 mm
σt Tensile strength [Pa]
σU CS Unconfined compressive strength [Pa]
xxii List of Symbols

Greek symbol Description Unit

σU T S Unconfined tensile strength [Pa]
σ1,2,3 Principal stresses [Pa]
σ0 Effective stress [Pa]
τs Shear strength [Pa]
τ0 Yield stress in shear [Pa]
Φs Particle sphericity [-]
φ Internal friction angle [◦ ]
ω Angular velocity [rad/s]
θ Half top angle of tool [◦ ]

Abbreviation Description
AE Acoustic Emissions
BTS Brazilian Tensile Strength
CRM Critical Raw Material
CSD Cutter Suction Dredge
DEM Discrete Element Method
DEM-SP Discrete Element Method-Smoothed Particle
DSM Deep Sea Mining
EXHADERO Excavation of Hard Deposits and Rocks
IDM Inhomogeneous Deformations on Micro-scale
LCM Liner Cutting Machine
PDC Polycrystalline Diamond Compact
RQD Rock Quality Designation
SP Smoothed Particle
SPH Smoothed Particle Hydrodynamics
TBM Tunnel Boring Machine
UCS Unconfined Compressive Strength
 1
Introduction
"There are no such things as applied sciences, only applications of science."
Louis Pasteur

This chapter gives an introduction to deep sea mining and its technological chal-
lenges, one of them being the excavation process of rock like materials from the sea
bed. The work in this thesis also considers rock cutting applications for drilling and
dredging (shallow water). Development of a modeling approach to simulate the rock
cutting process of a single tool is the main topic of this dissertation.

1
 2 1. Introduction

1 1.1. Background
Currently, the demands for raw materials increase. It is expected that these demands
continue to increase in the near future. This expectation is based on several trends
on a global scale.
The world population keeps growing. At the end of the year 2015, the world
population already consists of more than seven billion people. Prospects of the
United Nations on the world population show that it is expected that the world
population will increase to nine billion people within the next 30 years (Population
Division of the Department of Economic and Social Affairs of the United Nations
Secretariat, 2015) . Together with the growth of the population, the demand for
food will increase as well. Artificial fertilizers become almost a necessity to obtain
a sufficient food supply. These fertilizers are based on phosphorous material, which
is extracted from terrestrial mines. However, with the growing world population,
the current production rate will not suffice.
Another trend is that developing countries get more and more developed in
economical terms. Due to their development their demand for energy (mostly fossil
fuels) and raw materials increases. While developed countries are trying to become
less dependent on fossil fuels for their energy supplies, it is expected that the demand
for fossil fuels will not decline in the near future.
The development and production of new technologies such as wind turbines,
computers, televisions, mobile devices, solar cells and electric cars requires large
amounts of energy and raw materials, including rare earth elements. Currently
these are extracted from terrestrial mines. Production from these mines may not
suffice in the future.
The European Commission has created a list of critical raw materials (CRMs).
CRMs are defined as those materials that combine a high economic importance
to the EU and a high risk associated in their supply assurance (EU Commission,
2015). The supply risk is based on indicators like accountability, political stability,
government effectiveness and regulatory quality. In figure 1.1 an overview of the
main CRMs suppliers. Special notice is made on the resources located in China,
which might be of a significant influence on the development of the world economy
and geo-political relations. Especially because for several types of raw materials
China has (almost) a monopoly position, especially for antimony, magnesium, rare
earth elements and tungsten. Besides that, some of the locations are situated in
political unstable regions (e.g. cobalt in central Africa).
In order to reduce the dependence on the supply of raw materials by these
countries, other deposits and means of producing the raw materials are needed.
Recycling can help to supply in the demand of the raw materials needed. However,
it is expected that recycling will is not sufficient, and other mining locations are
needed to meet the demand. Besides mining sites on land, asteroid mining and
deep sea mining are considered an option.
In the deep sea, various types of metal rich deposits have been found, e.g. phos-
phorous nodules, manganese nodules, volcanic sulphide deposits, metal-rich crusts.
Most of these deposits are found on the seafloor of the Pacific and Indian ocean.
These deposits are especially interesting of because they contain many of the mate-
1.1. Background 3

Figure 1.1: Overview of highest production per country on critical raw materials, taken from EU
Commission (2015)

rials that are needed for development of the newer technologies, like for use in bat-
tery packs, permanent magnets, etc. Table 1.1 gives an overview of the prospected
resources of the Prime Crust Zone, an area of metal-rich crust in the Pacific of ap-
proximately the size of the United States. A remark has to be made, the total bulk
of materials prospected in the Prime Crust Zone is the total amount of material
expected to be found in that area, not necessarily being economically viable.
In this context, the perspective of deep sea mining is being considered. Besides
the environmental impact that deep sea mining can have, several technological chal-
lenges still have to be solved or are investigated. First of all, the material needs to
be excavated, which is investigated in this thesis. At large water depths, the hydro-
static pressure can be of substantial influence on the excavation process, (Zijsling,
1987). This is especially the case for rock-like materials. It is expected that rock
that fails in a brittle fashion at ambient pressure, might fail in a more ductile way in
a high pressure environment. Besides the change in failure mode, it is expected that
the cutting forces increase with increasing hydrostatic pressure as well. To what
extent this change in behavior occurs and what the implications will be is not yet
fully understood.
Furthermore, the excavated material needs to be transported to the surface,
where it can be further processed. Various ideas exist on how to transport the
materials, one of these ideas is investigated at our group of dredging engineering, i.e.
vertical hydraulic transport through a riser with a series of booster stations based on
centrifugal pumps (van Wijk, 2016). In vertical hydraulic transport, approximately
80% of the pumped volume will be water. In order to lower the environmental
impact of an offshore mining operation, it is required that the water that acts as
carrier fluid in the vertical transport system has to be deposited near the seabed
 4 1. Introduction

Table 1.1: Estimated metal contents in cobalt crust compared with global land-based reserves and
1 resources, after World Ocean Review (2014). Metal contents in millions of tonnes

Elements Cobalt crusts in Global reserves global reserves

the Prime Crust on land (econom- and resources
Zone (PCZ) ically minable de- (including sub-
posits today) economic de-
posits)
Manganese (Mn) 1714 630 5200
Titanium (Ti) 88 414 899
Rare earth oxides 16 110 150
Nickel (Ni) 32 80 150
Vanadium (V) 4.8 14 38
Cobalt (Co) 50 7.5 13
Tungsten (W) 0.67 3.1 6.3
Niobium (Nb) 0.4 3 3
Arsenic (As) 2.9 1 1.6
Bismuth (Bi) 0.32 0.3 0.7
Yttrium (Y) 1.7 0.5 0.5
Platinum group 0.004 0.07 0.08
Tellurium (Te) 0.45 0.02 0.05
Thallium (Tl) 1.2 0.0004 0.0007

in the vicinity of the mining operation. Besides filtration, no other treatments are
carried out on the seawater before it is returned to the seafloor. Fine sediments
remain in the water and during deposition of the return flow, sediment plumes can
occur that can travel for over tens of kilometers. Such a plume may be harmful
for the ecosystems in the region of a mining site. Near and far-field models will
be developed to allow for assessment of the environmental impact of such a mining
operation (Ortega, 2014; de Wit, 2015; van Grunsven et al., 2016).
The cutting of saturated rock is of interest in several fields of industry. Some
examples are addressed:

Example 1: Deep Sea Mining

Several of the mineral rich deposits in the deep sea, such as Seafloor Massive Sulfide
(SMS) deposits and ferro-manganese (Fe-Mn) crusts require rock cutting equipment
to retrieve the materials contained in these structures. SMS deposits are typically
located in water depths greater than 1 km and are in close proximity to tectonic
plate boundaries and submarine volcanic activities, e.g. the Solwara 1 SMS deposit
is located at a water depth of 1600 meter. Fe-Mn crusts form at water depths
of about 400-7000 m, with the thickest and most metal-rich crusts occurring at
depths of about 800-2500 m (World Ocean Review, 2014). There is limited data
about the mechanical properties of the crusts and the deposits. The typical range
of mechanical properties that have been found are presented in tables 1.2 and 1.3.
1.1. Background 5

Table 1.2: Range of mechanical properties in SMS deposits, after Yamazaki and Park (2003).
1
Parameter Min Max
Wet bulk density [kg/m3 ] 2.4 · 103 4.0 · 103
Solid density [kg/m3 ] 3.6 · 103 5.5 · 103
Porosity [-] 0.15 0.53
Unconfined compressive strength [MPa] 3.1 38
Tensile strength [MPa] 0.14 5.2
Typical water depths [m] > 1000

Table 1.3: Range of mechanical properties in Fe-Mn crusts, after Chung (1996)

Parameter Min Max

3 3
Wet bulk density [kg/m ] 1.65 · 10 2.17 · 103
Porosity [-] 0.43 0.74
Unconfined compressive strength [MPa] 0.5 16.8
Tensile strength [MPa] 0.1 2.3
Shear strength [MPa] 1.7 2.5
Typical water depths [m] 400 7000

In 2005 Nautilus Minerals Inc. started exploring the SMS deposits in the Ex-
clusive Economic Zone of Papua New Guinea. In 2010 several drilling trials in the
Solwara 1 project in Papua New guinea showed the presence of high graded copper
deposits. For the excavation of the SMS deposit, Nautilus designed to use three Re-
motely Operated Vehicles. An auxiliary cutter will be used deal with rough terrain
and creates benches for the other machines to work on. The bulk cutter is designed
to have a higher cutting capacity, but is limited to working on benches. Both ma-
chines will leave the cut material on the seafloor, which will be later collected by
the collecting machine that is connected to the vertical transport system. Due to
the relatively high hydrostatic pressure (with respect to the tensile strength of the
rock), tensile failure of the rock is less likely to occur and it is expected that shear
failure of the chips will be the dominating failure mechanism. Mining at the Solwara
1 project site is expected to start within a couple of years from 2016. Various com-
panies have comparable design concepts for deep sea mining operations, e.g. Royal
IHC, Bauer, Soil Machine Dynamics. An impression of the concept design of Royal
IHC is presented in figure 1.2.

Example 2: Dredging
Whether it is for the construction of new ports or the deepening of canals, in dredg-
ing often densely compacted sand, clay or rock have to be excavated. For these kind
of excavation projects, often a Cutter Suction Dredge (CSD) is used. The CSD is a
stationary dredger equipped with a cutter device, most often a crown cutter, which
excavates the soil before it is sucked up by the flow of the dredge pumps. While
operating, the dredger moves around a spud pole by pulling and slacking of the two
 6 1. Introduction

Figure 1.2: Preliminary designs of deep sea mining vehicles, courtesy of Royal IHC

side wires. The cutter head is installed at the end of the ladder, which is used to
position the cutter head.
CSD’s are typically used in water depths up to 30 m and have installed cutter
powers of 50 kW up to 8.5 MW. Most of the dredging works deal with rocks with
compressive strength up to 20 MPa where rock cutting is economically the most
preferable method of excavation. Tougher rocks, with a UCS of 60 - 80 MPa, can
be excavated with a CSD if blasting is not an option because of its environmental
impact. An example of a CSD is presented in figure 1.3a and an impression of
a crown cutter head is shown in figure 1.3b. When excavating rock with a CSD
cutterhead, the rock cutting process is dominated by brittle failure and chips are
generated by brittle shear and tensile failures. One of the market trends is that the
CSD’s will be used more and more on tougher types of rock.

(b) Cutterhead for cutter suction dredges,

(a) Cutter Suction Dredger Athena
IHC lightduty rock

Figure 1.3: Cutter suction dredge equipment.

Example 3: Drilling
In petroleum well drilling, the polycrystalline diamond compact (PDC) bits have
become one of the most common drilling tools. PDC-bits are composed by multiple
cutters that are positioned on fixed blades, see figure 1.4 for an example. The bits
are positioned at a negative rake angle, meaning that the bit ’drags’ through the
1.2. Research objective 7

rock. Due to the design of the bit, the small cutting depth (up to a few millimeters)
and the high hydrostatic pressure that is applied, PDC drilling works by shearing 1
the formation.
The typical rock properties with respect to drilling almost cover the whole range
of rocks that can be found, with UCS values ranging up to 200 MPa. The drilling
process often takes place in high pressure conditions (> 10 MPa). Special interest
is shown towards the drilling process of low permeable rocks (e.g. shale) at great
depths (Zijsling, 1987). Besides that, the drilling industry is interested in the appli-
cation of fluid flow with other fluids than water (e.g. non-Newtonian, drilling fluid)
through the pores of the rock considered.

Figure 1.4: Drill head with PDC bit inserts

1.2. Research objective

In engineering practice very often the information obtained from the field about what
kind of rock has to be cut is very limited. It is quite common that both the number
of samples as well as the number of measured rock parameters is very limited. It
happens quite often that only the UCS value and the type of rock are known, all
other parameters might have to be estimated based on empiricism (Zijsling, 2013).
In the industries, many questions about the cutting process of saturated rock-like
materials still remain. Especially on how to optimize and develop (new) excavation
and drilling equipment and how to predict the cutting forces. A proper understand-
ing of the behavior of saturated rock, the failure mechanisms and their interactions
is essential to answer these questions.
The main objectives of this PhD-study are:

• To describe the physical phenomena that occur during the cutting of saturated
rock, with an emphasis on the fluid pressure effects.

• To develop a physical and implement a mathematical model to predict the

rock cutting process, in which the hydrostatic and pore pressure effects are
incorporated.
 8 1. Introduction

The wide range of applications with respect to rock properties, ambient condi-
1 tions and tool design, more or less implies the need for a generic modeling approach.
For that reason, the emphasis of the model is on an approach that will be generally
applicable towards the cutting of saturated rock, independent of the type of rock
and tool that are considered.
To meet the objectives, several steps are defined:
• Step 1: Identification of the significant physical phenomena and parameters
for the cutting of saturated rock

• Step 2: Setup of physical and mathematical model to simulate the rock cut-
ting process. Furthermore, the numerical implementation of the models needs
to be verified
• Step 3: The developed model has to be validated. Validation cases and test
cases are simulated for various tool-rock interactions, e.g. tile cutting, drilling,
seabed mining and dredging.

1.3. Outline of this thesis

In chapter 2 the different processes influencing the behavior of saturated rock are
introduced. How these processes work together and affect the rock cutting process
is elaborated upon in chapter 3. The state of the art of rock cutting models is
discussed in this chapter as well.
Chapter 4 elaborates on the modeling approach, setup and implementation of
the physical model in a mathematical model. The methodology is tested for several
numerical cases to prove that the mathematics of the model are calculated correctly.
In chapter 5 validation cases for the methodology with respect to material tests of
the saturated rock are presented. Validation and test cases of the methodology with
respect to tool-rock interactions are investigated in chapter 6.
Each chapter contains conclusions and recommendations concerning that specific
topic. The overall conclusions and recommendations are presented in chapter 7.
These are subdivided in the topics saturated rock, rock cutting models and the
numerical modeling approach developed within this project.
 Physics of Rock in Relation
2
to Rock Cutting Process
"The first step to be taken, is to study carefully the fundamental phenomena above
described, and to examine all the various circumstances under which they present
themselves."
Jean-Baptiste Biot

The different processes influencing the mechanics, deformation and failure of rocks
are introduced and some characteristic properties of deformation and failure of sat-
urated rock are discussed. In this chapter the emphasis is on the phenomena and
effects that are related to (saturated) rock mechanics. First the most common charac-
teristics of properties that are of interest to the cutting of dry rock. This is followed
by an overview on how the presence of a pore fluid can influence the mechanical
response of the rock. Tool-rock interactions will be treated in chapter 3.

9
 10 2. Physics of Rock in Relation to Rock Cutting Process

2.1. Introduction
In all rock cutting processes, the material response of rock is essential. The presence
of a pore fluid can significantly affect the mechanical response of the rock. How the
mechanical behavior can be affected by a fluid is analyzed in detail in this chapter.
2 To start with, some basic definitions are given and a brief overview of the type of
rocks that are of special interest within the industrial practice is presented. This is
followed by an overview of the mechanical response of dry rock, with an emphasis on
the most relevant aspects of the mechanical behavior of the rock with respect to the
cutting process. Thereafter the influence of a pore fluid on the mechanical behavior
of a rock is analyzed. Although the main focus of this research is on fluid saturated
rock, most of the research carried out in the field of rock cutting concerns dry rock.
A pore fluid can affect the mechanical behavior of a rock through physico-chemical
effects (independent of time scale) and through hydro-mechanical effects (dependent
on time scale of experiments). First an overview of the physico-chemical effects is
presented to show what changes can occur when comparing dry and saturated rock
samples. This will be followed by an analysis on which hydro-mechanical coupling
effects can have a significant influence on the rock deformation process.
The topic of this dissertation covers several fields of research and engineering,
each with their own sign conventions. To avoid confusion, the following sign con-
ventions are used throughout this dissertation.
• Compressive direction is positive, tensile direction is negative.
• Vectors are denoted in bold face, e.g. F = mü.
There are three general types of rocks, igneous, metamorphic and sedimentary.
Although all types of rocks can be found in abundance, sedimentary rock is the
most common kind within the drilling, dredging and seabed mining industries that
are considered in this dissertation. In several cases in this dissertation the data
presented in the references does not provide sufficient information, the estimations
for these missing material properties are based on empirical data and rules of thumb,
these are presented in appendix A.
As stated in the research objective, see section 1.2, there is a need to investigate
the rock cutting process in general. Due to the fact that the industries have to
deal with a wide range of rocks and detailed rock properties often lack in practice,
emphasis is put on modeling the rock cutting process for a generic type of rock. For
that reason, the work presented in this chapter is restricted to the most relevant
physical phenomena and trends observed in literature.

2.2. Rock Failure Mechanics

2.2.1. Failure modes
At the micro-scale, several failure mechanisms can be distinguished in practical
rock mechanics tests (e.g. uni-axial compression, tri-axial compression, Brazilian
splitting tests):
• Tensile failure (brittle).
2.2. Rock Failure Mechanics 11

• Shear failure (brittle).

• Compressive failure (ductile).
Which failure mechanism will occur depends on the rock properties and the stress
conditions applied on the specimen. The strength ratio m is often used to distinguish
the dominant failure mechanisms in a rock. The strength ratio is defined as 2
σU CS
m= (2.1)
σU T S
with uni-axial compressive and uni-axial tensile strength, respectively σU CS and
σBT S . Low values of the strength ratio are typical for ’ductile’ rock types (e.g.
shale (m = 6 ± 2), chalk (m = 7 ± 2), gypsum (m = 8 ± 2)), high values are typical
for brittle rock types (e.g. granite (m = 32 ± 3), quartzite (m = 20 ± 3)). The values
of m are presented here are from Hoek and Brown (1997), for an overview of more
typical values for m, see appendix A.
An overview of the macro-scale failure modes that occur with changes in confin-
ing stress is presented in figure 2.1. Tensile failure is associated with the separation
of grains. It can also occur as axial splitting, which can occur in uni-axial com-
pressive tests and tri-axial compressive tests with low confining stresses for rock
specimens with a high strength ratio. In this failure mode the specimen splits along
the axis of the highest principle stress. Specimens with low strength ratios will not
fail in axial splitting or even in tensile failure at zero confining stress. At higher
confining stresses, the sample will fail along a shear plane. Further increase of the
confining stress can result in shear bands. At sufficiently high confinement, the
sample will fail in a compressive ductile fashion. In the extreme case this results in
barreling of the specimen. This failure mode is especially of interest to high porous
rocks, where the porous structure of the rock collapses as a result of high mean
stress, resulting in densification of the rock.

2.2.2. Brittle Failure Modes

There are three modes of brittle deformation that can occur in a three-dimensional
body, (Lawn, 1993). These modes are shown in figure 2.2. In tensile fracturing
(mode I), the fracture surfaces open with respect to each other after fracturing has
taken place. The energy needed for creation of the surface is provided by tensile
strain energy. In the shearing mode (mode II), the fracture acts along the fracture
surface and parallel to the propagation direction of the fracture. The energy required
for the fracture to propagate is provided by shear strain relative to the fracture plane
ate the tip. The tearing mode (mode III) also acts along the fracture surface, but
perpendicular to the propagation direction of the fracture. This is due to a torsional
component applied to the fracture front. In fracture mechanics, often combinations
of these three modes occur. Rock cutting processes, dependent on the conditions,
are dominated by failure modes I and II, as far as brittle failures are concerned.

Linear Fracture Mechanics

Linear elastic fracture mechanics (LEFM) help to interpret the opening and propa-
gation of fractures in a linear elastic continuum.
 12 2. Physics of Rock in Relation to Rock Cutting Process

failure modes

tensile failure shear failure ductile failure

direct shear plane shear band σ3
splitting
tension
2
σ1=σ2

brittle ductile
compression q
σ3

remoulded
shear failure
tension cut-off shear failure cap compressive cap

pbd p

extension q σ1=σ2

Figure 2.1: Overview of the range of deformation modes for rock with increasing confining stress,
after Winterwerp and van Kesteren (2004).

Figure 2.2: Definition of the deformation modes.

The first law of thermodynamics can be used to determine whether a crack

will propagate. In order for a crack to propagate, the total amount of mechanical
energy that is supplied to a material volume per unit of time must be transferred
into internal energy, surface energy, dissipated energy and kinetic energy. Internal
2.2. Rock Failure Mechanics 13

energy is the stored energy. The surface energy changes when a new free surface
is generated (i.e. when a crack propagates). The kinetic energy is the result of
material velocity. The dissipation may occur in various ways, but is mostly due to
friction and plastic deformation, which results in temperature changes. According to
Griffith (1921), the propagation of a crack is determined by the transfer of internal
and external energy into surface energy, as in 2
dUe dUi dUa
− = (2.2)
da da da
with external energy Ue , internal energy Ui , surface energy Ua and change in surface
da. The propagation of a crack is then determined by
dUe dUi dUa
− < → no crack growth
da da da
dUe dUi dUa
− > → unstable crack growth
da da da
dUe dUi dUa
− = → critical crack length
da da da
(2.3)

An unstable crack growth corresponds to a brittle macroscopic failure. Like in

many physical processes, the total free energy of a system will be maximized. Based
on this, the direction of propagation of a crack is favored in the orientation that
maximizes the decrease in total system free energy. In an isotropic system this
corresponds with seeking a maximum of the mechanical energy release rate G. In
other words, the crack-extension force g = G − R0 is maximized, (Lawn, 1993,
p.45) with surface tension force R0 . However, Lawn also mentions that the entire
propagation history of a crack is predestined by the existing stress state before
fracture has even begun.

2.2.3. Brittle-Ductile Transition

There is a large variety of the potential deformation mechanisms in the ductile
field, and thus for the brittle-ductile transition as well. The nature of brittle-ductile
transitions at relatively low temperatures (as is the case for rock cutting), two
extreme cases of ductile deformation mechanisms are considered, purely cataclastic
and purely crystal plastic. Although the purely cataclastic ductile behavior is most
likely to occur in rock cutting processes, purely crystal plastic ductile behavior has
been observed in drilling experiments as well (Zijsling, 2013).
According to Paterson and Wong (2005), the brittle-ductile transition in case of
cataclastic failure can be explained by a lower stress dependency of the fracture stress
with respect to the frictional sliding stress. The cross-over of these two stress trends
is where the brittle ductile transition occurs, as depicted in figure 2.4a. However,
this transition is based on a macroscopic observation, on a microscopic scale, the
following occurs. With increasing confining stress not only the growth of micro-
cracks becomes more difficult, the growth of the cracks at high confining stresses
tends to be more stabilized. This eventually leads to fragmentation or disintegration
 14 2. Physics of Rock in Relation to Rock Cutting Process

of the specimen by a rapid increase in the number of stable micro-cracks, resulting

in a cataclastic failure mode. Francois and Wilshaw (1964) explained, based on
dislocation models for the nucleation of cracks, that because the additional stress
from a confining stress is small compared to the local stresses near the dislocations,
the stress for microfracture initiation will be nearly independent of the confining
2 stress while crack propagation depends on the confining stress.
The cataclasis can mainly result in two different types (Paterson and Wong,
2005). In compact rock and strongly cohesive rock, stabilization of the micro-cracks
leads to the occurrence of more micro-cracks as the loading increases, eventually
leading to the coalescence of the cracks, resulting in the breaking down of the rock
into a granular mass. In weakly cohesive porous rock, cataclasis is more a result of
grain crushing, (Zhang et al., 1990; Wong et al., 2004).
Deformation by crystal plasticity occurs when an increase of the confining stress
on a specimen raises the brittle fracture strength to a level that exceeds the yield
stress for crystal plastic flow, see figure 2.4b. Depending on the type of rock, it will
be possible that before the crystal plastic flow occurs, first a cataclastic flow regime
exists (Wong et al., 2004).

2.2.4. Ductile failure mode (cataclasis)

The macroscopic compressive failure of rocks can be caused by three different types
of failure at the micro-scale (van Kesteren, 1995), especially for sedimentary rock:

• Failure of the bonds between particles

• Failure of the particles (often referred to as particle crushing)

• Failure of the skeleton (often referred to as pore collapse)

All three failure mechanisms have a similar effect, i.e. they increase the mobility
of particles within the rock fabric. Based on the Inhomogeneous Deformations
on Micro-scale (IDM) theory (van Kesteren, 1995) it follows that this mobility on
micro-scale is obtained through heterogeneous shear deformation between particles
(i.e. sliding and/or rolling of particles) in order to allow for homogeneous isotropic
plastic deformation on the macro-scale. A visual representation is presented in
figure 2.3. For that reason, quasi-ductile behavior requires mobility of (components
within) the rock skeleton to allow the plastic shear strain within the skeleton. In
general, this mobility is created by failure of the bonds between grains or failure of
the grains themselves.

2.2.5. Post-failure
The brittle-ductile transition can also be observed with respect to the post-failure
behavior of a rock specimen. The maximum stress before failure increases with
increasing confining stress and the amount of strain softening decreases. At low
confining stresses, when approaching the peak strength, strain softening will occur
weakening the specimen with increasing strain. Further strain will tend to be inho-
mogeneous and it concentrates in the weaker elements of the rock that have already
2.2. Rock Failure Mechanics 15

(a) Continuum based (b) Particle based

Figure 2.3: Homogeneous deformation paradox, based on Winterwerp and van Kesteren (2004).
σ1-σ3 at failure
σ1-σ3 at failure

s s stress
stre yield
re l ss
tu na str
e
frac tio re
r fric tu
f o
ss lidin
g frac
r e
st s
ductile brittle ductile
brittle
Confining stress Confining stress
(b) Transition to plastic flow, through crys-
(a) Transition to cataclastic flow.
tal plasticity.

Figure 2.4: Simple models for the brittle-ductile transition, after Paterson and Wong (2005).

been subjected to the largest amount of strain. Following peak stress, these zones of
concentrated strain or shear planes develop as large cracks throughout the specimen.
In the case of strain-hardening deformation, specimens of rock become stronger
as they deform. As a consequence, the strain tends towards homogeneity throughout
the confined specimen, since those elements of the rock which have strained most
will be stronger than those that have strained less, (Farmer, 1983, p.85). See figure
2.5 for an illustration of the effects with respect to an increase in confining pressure.
Wawersik and Fairhurst (1970) classified rocks according to their post peak be-
havior as class 1 and 2.
 16 2. Physics of Rock in Relation to Rock Cutting Process

ductile

σ1-σ3
brittle-ductile

2 Increasing σ3
brittle
ε

Figure 2.5: Illustration of brittle to ductile transition with increasing confining pressure σ3 .

1. Rock can absorb more energy and continue axial deformation after the peak
load

2. Energy needs to be extracted from the rock sample and axial displacement
needs to be reduced for a class 2 rock if a quasi-static rock behavior is to
obtain.

Class 1 type of rocks are most common to be found. Gowd and Rummel (1980)
found that for porous rocks, the brittle-ductile transition is characterized by an
abrupt change from dilative behavior at low pressures to compaction during inelastic
axial strain at high pressures. This abrupt change they found in experiments on
sandstone from SW-Germany. In low porosity rocks (e.g. Carrara marble with
1% porosity) dilation persists well into the ductile regime (Edmond and Paterson,
1972). The compaction that occurs during ductile deformation in porous rocks at
high confining pressures is due to pore collapse and the rearrangement of the grains
to allow for a denser packing.
Nearly all rocks, as well as concrete, become dilatant prior to fracture in com-
pression, even under high confining pressure(Brace and Martin III, 1968). Dilation
represents an increase in porosity, and changes in porosity can lead to changes in
pore pressure. Although pore pressure within their samples could have changed.
In experiments on the failure strength of mica, Obreimoff (1930) noted that
the confining pressure is of influence on the crack propagation velocity. In the
experiment a glass wedge is inserted into a crack. It was observed that the crack
did not grow immediately to its equilibrium length: in air (atmospheric pressure)
equilibrium was reached within seconds, whereas in a vacuum the crack continued
to creep for several days. Thus the confining stress can also be of influence on the
time scale of the mechanical behavior of a rock.

2.2.6. Grain size effects

The size and shape effects of the component minerals in the rock are of significant
effect, i.e. in general a smaller grain size leads to a tougher rock. Additional
features of the micro-structure that influence the mechanical response of the rocks
is the degree of interlocking of the grains. Fracture is more likely to occur along the
grain boundaries instead of through the grains and therefore fracture propagation
2.3. Fluid Saturated Rock 17

is more difficult in an irregular structure. Onodera and Kamura (1980) found that
a linear relationship exists between compressive strength and grain size for granite,
i.e. strength increases with decreasing grain size.
In many sedimentary rocks the bond between the grains is provided by cement
rather than the interlocking of grains. The amount and type of cement is important
as it not only influences the strength and elasticity of the rock, but density, porosity 2
and permeability as well.

2.2.7. Strain rate effects

Li et al. (2013) performed direct tension tests with dry materials and confining
pressures. In these tests they varied both the strain rate and the confining pressures
on samples of gypsum. They varied the strain rate in the range of 10−5 to 3.0 · 10−2 .
In case of no confining pressure, the tensile strength increases from 1.72 MPa to
3.57 MPa from low to high strain rates. They compare this result with BTS tests
with varying strain rates as well, which gives comparable results. Furthermore,
they measured that with increasing side confinements, the strengthening effect due
to increasing strain rates decreases. They noted the same phenomenon for granite.

2.2.8. Specimen size effects

Experimental results show that the rock strength decreases significantly with in-
creasing sample size. Hoek and Brown (1980) suggested that the UCS of a specimen
with a diameter of d mm is relates to the UCS value of a specimen with a 50 mm
diameter, based on published data, through
 0.18
50
σcd = σc50 (2.4)
d
with σcd as the compressive strength of a sample with diameter d and compressive
strength of a specimen with diameter of 50 mm σcd . Medhurst and Brown (1996)
found that with decreasing specimen size not only the UCS value increases, but
also the internal friction angle increases. Hoek and Brown (1997) suggest that the
reduction in strength with increasing specimen sizes is due to the larger opportunity
for failures to occur. When the specimen size is sufficiently large, the strength of
the sample reaches a constant value.

2.3. Fluid Saturated Rock

Unless stated otherwise it is assumed that the fluid in the rock is a liquid. The
presence of a pore fluid in rock can have effects on the mechanical response of the
rock. This can either be the result of one or a combination of two possible effects
(Duda and Renner, 2013):
1. Physico-chemical interactions between solid constituents of rocks and pore
fluids are controlled by mineralogical composition, pore fluid chemistry and
micro-structural features.
2. Hydro-mechanical effects, also often referred to as drainage effects. Volumetric
deformation of the local pore volumes result in a change of the local pore
 18 2. Physics of Rock in Relation to Rock Cutting Process

pressure. When the deformation process is more rapid than that the pore
fluid can flow, the local pore pressure gradients will affect the mechanical
response of the rock as well.

This thesis focuses on the hydro-mechanical effects of a pore fluid on the rock cutting
2 process. Physico-chemical effects might be of interest when comparing dry and
saturated conditions of the same type of rock. Such a comparison is beyond the
scope of this research, as rock will be cut when it is submerged by seawater or
drilling fluid, that . For sake of completeness, a brief overview of the physico-
chemical effects is presented here.

2.3.1. Physico-chemical effects

Most of the research towards physico-chemical effects of a pore fluid in rocks uses
water as the pore fluid. The effect of water on the strength of rock is highly variable
across different types of rocks, as is shown in table 2.1. This is understandable, as
the physico-chemical interactions between water and the constituent of rock grains
are largely influenced by mineralogical composition, pore volume and shape, grain
size and other micro-structural properties, which vary greatly between different
rock types. However, the effect of water is more pronounced in clay-rich rocks and
siliceous rocks (Atkinson, 1984; Reviron et al., 2009) than in quartz-rich rocks. Clay
minerals in water-saturated rock weaken its strength by two mechanisms; chemical
reactions with water (Cook, 1999) and reducing the frictional coefficient of rock
(Byerlee, 1978; Morrow et al., 2000).

Table 2.1: Ratio of unconfined compressive strength at saturated condition to that of dry condi-
tions, after Zhang et al. (2005).

σc(sat)
σc(dry) Rock type Originally published in
0.50 Shale and Quartzitic sandstone (Colback and Wild, 1965)
0.76 Penrith sandstone (Dyke and Dobereiner, 1991)
0.75 Bunter sandstone (Dyke and Dobereiner, 1991)
0.66 Waterstone (Dyke and Dobereiner, 1991)
0.97 Oolitic limestone (Lashkaripour and Ghafoori, 2002)
0.62 Sandstone and sandy limestone (Lashkaripour and Ghafoori, 2002)
0.81 Oolitic limestone and limy sandstone (Lashkaripour and Ghafoori, 2002)
0.52 Shale (Lashkaripour and Ghafoori, 2002)
0.76 British sandstone (Vasarhelyi, 2003)
0.66 Miocene limestone (Vasarhelyi, 2005)

Vutukuri (1974) used various pore fluids (e.g. water, glycerine, alcohols) to test
their effects on the tensile strength of Indiana limestone. The results suggested
that with increasing dielectric constant and surface tension of the liquid, the tensile
strength of the limestone increases as well. Swolfs (1972) found that with aluminum
and ferric iron salt solutions in water react with the surface structure of quartz and
silicates. Resulting in a reduction in surface energy, surface cohesion and breaking
2.3. Fluid Saturated Rock 19

strength of the rock. However, the coefficient of internal friction remains the same.

2.3.2. Hydro-mechanical effects

The flow and pressure of a pore fluid can significantly affect the mechanical response
of a rock. Depending on the field of application it is referred to as hydro-mechanical
coupling (e.g. rock mechanics), or drainage mechanisms (e.g. dredging industry).
2
Hydro-mechanical coupling is mainly caused by the combination of (local) volumet-
ric deformation of the pores and fluid flow. The analysis starts with a discussion of
the concept of permeability and to what extent the permeability can change due to
deformation of the rock, and thus affecting the flow of the fluid. This is followed by
a discussion on how the hydro-mechanical coupling affects the mechanical response
of a rock.

Effective stress
Drainage mechanisms control the effective stress in a rock. Pore pressures work as a
counteracting effect on the normal stress in a rock, which is expressed by Terzaghi’s
law of effective stress Terzaghi (1943). The effective stress law is defined as

σ0 = σ − p (2.5)

with effective (particle-particle) stress σ 0 , total stress σ and pore fluid pressure p.
Several corrections on the effective stress for rock were proposed by Skempton
(1960), which were experimentally verified by Nur and Byerlee (1971). According
to them the effective stress in rocks is given by
 
0 0 Cs
σ = σ + αes p = σ + 1 − p (2.6)
Cf

with αes the effective stress factor, Cs compressibility of the solids (i.e. of a single
grain) and Cf compressibility of the fabric (i.e. solid skeleton, bonded grains). It
must be noted that the influence of porosity is considered indirectly in (2.6) in
the parameter for the compressibility of the skeleton. When considering a material
with high porosity, the skeleton is far more compressible compared to the individual
grains Cs  Cf , i.e. αes ≈ 1. In the limit of a material with very low porosity, the
compressibility of the skeleton is almost comparable to the compressibility of the
individual grains (αes ≈ 0), where the effect of the pore pressure is negligible. It is
mentioned by van Kesteren (1995) that (2.6) has limitations on its applicability. He
states that the correction is only applicable when only deformation of the skeleton
is considered. Van Kesteren therefore suggests the use of pore pressure dissipation.
In the perfectly undrained case, the limit pressure of the pore-water that can be
generated with respect to the total/effective stress, which is then given by:
σ Cp − Cs
=1+n (2.7)
pundr Cf − αes Cs
van Kesteren (1995) mentions that the compressibility of the pore-water is influ-
enced by the presence of a gaseous phase in the pore-water. When the dissolution
 20 2. Physics of Rock in Relation to Rock Cutting Process

of the gas is negligible (e.g. the loading is sufficiently fast) the compressibility of
the pore-water is given by Cools (1984):
  
1−v 1 pi
Cp = + Cl − 1 − v − (1 − Si − v) (2.8)
u p p
2 with compressibility of the pore water Cp , initial saturation degree Si , initial
pore-fluid pressure pi and the gas fraction adsorbed by solids v. Due to the very
high pressures in deep sea applications and at the tool tip, it is assumed that the
compressibility of the pore-water is equal to pure water.
Brady and Brown (2005) show that the influence of pore water pressure on
the behavior of porous rock in tri-axial compression tests is strongly dependent on
the ’effective’ confining pressure (similar to Terzaghi’s principle): A series of tri-
axial compression tests was carried out on a limestone with a constant value of
σ3 = 69M P a, but with various levels of pore pressure in the range u = 0 − 69M P a
applied. There is a transition from ductile to brittle behavior as u is increased from
0 to 69M P a. In this case the mechanical response is controlled by the effective
confining pressure, σ30 = σ3 − u, calculated using Terzaghi’s classical effective stress
law. For less permeable rocks than this limestone, it may appear that the classical
effective stress law does not hold, as is stated by Brace and Martin III (1968)
concerning the dilatancy hardening effect.

Permeability of rock
The permeability of a porous material is a measure for its capability to transmit a
fluid. The term permeability might lead to confusion, in practice two permeability
related parameters are used, i.e. the permeability coefficient and the intrinsic per-
meability. The permeability coefficient (or hydraulic conductivity) Khyd is defined
as a discharge velocity through a unit area under a unit hydraulic gradient and is
dependent on the properties of the porous medium as well as the fluid, e.g. density
and viscosity. The intrinsic permeability κ is a property of only the porous medium
itself and thus independent of the fluid properties. The two are related (in the case
of hydraulic conductivity with water properties) through
ρg
Khyd = κ (2.9)
η
with fluid density ρ, gravitational acceleration g and dynamic viscosity η. The
intrinsic permeability is used in this dissertation, because it is independent of the
fluid properties. According to Darcy’s law, the quantity of flow through a porous
medium is determined by
κ
q = − ∇p (2.10)
η
where q is fluid flux.
Various empirical relations relate the intrinsic permeability with porosity, or
grain size (Schön, 1996). Both correlations can be used to estimate the permeability
of a rock.
The intrinsic permeability of a rock can change due to the loading that is applied
to the rock. There are various ways to describe such a change in permeability. In a
2.3. Fluid Saturated Rock 21

direct way, the effective stress state can be linked to the permeability of intact rock.
Various researchers found an empirical relationship based on a negative exponent,
a negative power law or Hertz theory to describe the relationship between effective
stress and the permeability at zero effective stress (Tiller, 1953; Louis et al., 1977;
Gangi, 1978):
2
−c2
κ = c1 σef f (σ 0 > σthreshold ) Tiller (1953) (2.11)
0
κ = κ0 e−σ Louis et al. (1977) (2.12)
 0 2/3 !4
σ + σi
κ = κ0 1 − C0 Gangi (1978) (2.13)
E0

with constants A and m, effective stress σef f , threshold stress above which the equa-
tion is valid σthreshold , permeability at zero effective stress κ0 , constant depending
on the packing C0 , equivalent pressure due to permanent deformation and cementa-
tion of the grains and the effective elastic modulus of the grains E0 . In a less direct
way, the permeability can change due to deformation of the matrix of the porous
material. The deformation can be described linked to a change in porosity. As an
example, the Carman-Kozeny relation describes the relation between porosity and
permeability for a packed bed of solids.

Φ2s d2p n3
κ= (2.14)
180 (1 − n2 )
with particle sphericity Φs , particle diameter dp and porosity n. Due to the asymp-
totes that occur near the limits of zero porosity or full porosity (0-1), close to those
boundaries the validity of the Carman-Kozeny relation is doubtful. For a graphical
representation and the asymptotic limits, see figure 2.6.

Carman Kozeny
4
10
[m2 ]

100
Φ2s d2p
180

10−4
0 0.2 0.4 0.6 0.8 1
φ [−]

Figure 2.6: Carman Kozeny relation for permeability.

Although the permeability depends on the deformation or effective stress state

of the rock matrix, the resulting change in permeability while the rock is still intact
 22 2. Physics of Rock in Relation to Rock Cutting Process

results in a permeability of the same order of magnitude, which is already close

to the accuracy of the measured permeability that is used in the industrial appli-
cations for rock cutting. Besides, the change in permeability due to deformations
strongly depends on the micro-structure of the rock matrix. In the case of damaged
rock, it can no longer be assumed that the permeability is isotropic and homoge-
2 neously distributed throughout the specimen and the use of more advanced porosity
and permeability models might be required (e.g. dual porosity/permeability, large
discontinuities).

Drainage response
When considering the compressive strength of a rock, the effective stress law is
only valid below a critical strain rate, below which effective drainage of the sample
is achieved Brace and Martin III (1968). A sample is considered to be effectively
drained when the local pressure differences due deformation rate are significantly
smaller than the compressive and/or tensile strength of the rock, i.e. ∆p  σc , σt .
However, at higher deformation rates, the local pressure differences can contribute
to the strength of the rock, an example of such a phenomenon is dilation hardening
with respect to compressive strength Brace and Martin III (1968). Another effect
that might occur is compaction weakening. Which effect is more likely to occur
depends on the micro-structure of the rock. It is reasonable to assume that similar
trends are valid for tensile strength (if it is tested at a sufficiently high hydrostatic
pressure, to allow for a significant pressure difference to build up without having
the fluid to cavitate). As a result, the observed material behavior can change with
deformation rate and total hydrostatic pressure to which it is subjected.
Bulk compaction is related to the collapse of the rock matrix, which causes
a reduction in pore volume. This reduction leads to an increase in pore pressure,
which can lead to a decrease in effective stress, resulting in a reduction in compressive
strength Bernabe (1987). During dilation, the porosity increases due to the creation
of new and the extension of existing micro-cracks. As a result, the pore pressure of
an effectively undrained sample drops locally, resulting in an increase of the effective
stress and thus an increase in compressive strength, which is often referred to as
dilatancy hardening Brace and Martin III (1968). The effect of dilatancy hardening
is associated with a critical strain rate ˙cr and strongly depends on the hydraulic
properties of a rock. At strain rates above ˙cr , the apparent strength of a rock
increases more compared to that of a drained or dry sample.
Two mechanisms can limit the effect of dilatancy hardening, i.e. cavitation and
grain failure (Zijsling, 1987). When during dilation the pore pressure decreases
to the vapor pressure of the pore fluid, the fluid will vaporize and the fluid bulk
modulus and viscosity both decrease by several orders of magnitude. As a result
the pore pressure in the cavitated region does not decrease anymore. In the case of
grain failure, the contact stresses at the grain boundaries can exceed the strength
of the grain. As a result, grains will fail, giving rise to a smoother surface and
reducing the amount of dilation needed before grains can slide along each other
and thus effectively a smaller drop in pressure in the dilating region. A schematic
representation of the dilatancy hardening effect and its limits is shown in figure 2.7.
2.3. Fluid Saturated Rock 23

Grain failure limit

Strength
Cavitating limit Increasing
hydrostatic
Dilatancy pressure 2
ϵcr hardening

Strain rate
Figure 2.7: Phenomenon of dilatancy hardening and its limitations, based on Brace and Martin
III (1968)

Analysis of Rudnicki and Chen (1988) on data of Martin III (1980) shows that
slip-induced dilation coupled with the flow of pore fluid can stabilize rapid slip. Only
a small amount of dilation, corresponding to uplift of the order of a few percent of
the slip is needed. However, the stabilizing effect decreases with decreasing pore
fluid bulk moduli. As a result, when instabilities in the pore pressure occur due to
cavitation or the dissolution of gases from the pore fluid, ultimately lowering the
resistance towards dilation.
Even in the absence of dilation, strengthening occurs when local fluid motion
is restricted, because of the coupling between elastic deformation and pore fluid
movement around the crack tip. Rice and Cleary (1976) analyze propagation of
shear cracks in porous elastic media, and they find a dependence of the energy
release rate on the velocity of propagation of the crack. This effect arises from the
elastic distortion of the material around the crack and means that higher driving
forces are required at higher crack velocities. In other words, the undrained response
at the tip of the crack has a higher stabilizing effect with respect to drained response.
Rutter (1972) performed experiments with varying strain rates and effective
stresses on Solenhofen limestone saturated with water. Several trends are distin-
guished by Rutter, i.e. the ductility decreases (more significant strain softening)
and the yield strength increases with increasing strain rates; both the yield strength
and ductility increase with increasing confining stress.
Brace and Martin III (1968) show that the law of effective stress does not always
apply for low porosity (0.001-0.03) rock, for deformations above a critical strain rate,
Brace and Martin measured an increase in strength beyond the failure criterion.
They showed that the validity of the law of effective stress depends on the strain
rate, the intrinsic permeability and the pore-fluid viscosity (note that it also depends
on the fluid compressibility, but the compressibility of the fluids used by Brace and
Martin were too similar to distinguish this effect). Something similar has been
observed by Rutter (1972); due to dilation hardening, the observed effective stress
in the specimen is higher at the moment of failure compared to that of the initial
state of the experiment.
Swan et al. (1989) observed a much larger effect of the increase of the compressive
 24 2. Physics of Rock in Relation to Rock Cutting Process

strength with increasing strain rates for saturated shale from Kimmeridge Bay,
Dorset, U.K. They noticed an increase in both strength and strain to failure of the
order of four and seven times respectively over three orders of increasing strain rates.
The hydro mechanical behavior of saturated rock is directly related to the
amount of drainage allowed by the applied deformation, which is often referred
2 to as the Deborah number (here N ). According to Duda and Renner (2013), for
compression tests this can be characterized by the ratio
tc
N= (2.15)
tdef

of the characteristic time of hydraulic diffusion tc to the characteristic time of de-

formation tdef . The characteristic times are defined as

l2
tc = (2.16)
D
with characteristic length scale l and hydraulic diffusion coefficient D, and
∆ 0.1f
tdef = = (2.17)
˙ ˙
with strain , strain to failure f and strain rate .
˙ In a similar fashion to Duda and
Renner (2013), it is assumed that the characteristic strain should be one order of
magnitude smaller than the strain to failure f . The Deborah number shows strong
resemblance with the pore Peclet number that is used in dredging applications,
which will be treated in section 3.1.6. The transition between drained and undrained
behavior occurs at Ncr ≈ 1. Duda and Renner derived the critical strain rate above
which the effective internal drainage is lost, with respect to compression tests. The
critical strain rate is given by
Ncr f D
˙cr ≈ 0.1 (2.18)
l2
with critical strain rate ˙cr , strain to failure f , hydraulic diffusion coefficient D,
and specimen length l. The hydraulic diffusion coefficient is determined based on
the specific storage capacity
κ κ
D= = (2.19)
ηs η (Cf − αCs + n (Cp − Cs ))

with intrinsic permeability κ, fluid viscosity η, specific storage capacity s, coefficient

for solid compression α, porosity n, compressibility of the rock matrix, solid grains
and fluid, respectively Cm , Cs and Cf (cf. Duda and Renner (2013); van Kesteren
(1995)).
Another explanation for the change in mechanical response beyond the critical
strain rate is given by Swan et al. (1989), which is that since only localized fluid
movement is allowed at high strain rates, the nucleation and growth of a crack in one
part of a specimen does not affect the pore pressure distribution in the bulk. Thus
2.4. Conclusions 25

when one crack is formed and subsequently strengthened, it is easier for another
crack to grow elsewhere in the specimen than it is to continue to grow the original
crack. This process of stabilization is repeated for each crack that forms, so that
the deformation becomes de-localized, linkage of cracks is suppressed and the strain
to failure increases.
Some other qualitative observations on the influence of the strain rate on the 2
rock mechanics were observed by Masuda et al. (1987) in experiments on granite,
i.e. the AE rate is accelerated at a stress level closer to the failure stress as the
strain rate decreases. In other words, slow strain rate results in a more perfect
macro-crack. Furthermore, Masuda et al. (1987) observed that the strength of the
rock linearly increases as the logarithm of the strain rate increases and that the
strain rate dependence is enhanced at high confining stresses.

2.4. Conclusions
To summarize, the most relevant physical phenomena that can have a significant
influence on the rock cutting process are identified. Only physical effects that di-
rectly affect the rock cutting process should be considered, assuming that the in-situ
conditions are reasonably recreated in lab tests. In dry rocks, a distinct effect is
noticeable with the presence of a confining stress. That is, with an increase in confin-
ing stress the compressive strength of the rock increases. Furthermore, a transition
from brittle to ductile failure might occur.
In the case of saturated rock, one should also consider the effects of dilation
hardening, compaction weakening, cavitation, bulk stiffening, together with a more
profound strain rate dependency of the mechanical properties. Due to differences in
mechanical response through physico-chemical effects, one has to be cautious when
comparing mechanical properties of dry and saturated bulk material with respect
to the rock cutting process.
 3
Rock Cutting Process
". . . once a theory appears on the question sheet of a college examination, it turns
into something to be feared and believed, and many of the engineers who were
benefited by a college education applied the theories without even suspecting the
narrow limits of their validity"
Karl von Terzaghi

In this chapter the current state of the art of the knowledge concerning the rock
cutting process is discussed. Starting with the phenomena that occur when cut-
ting saturated rock, with an emphasis on dredging, seabed mining and drilling tools.
Various rock cutting models concerning the use of different kinds of equipment and
different types of rocks are presented. These models can be useful to aid the design of
excavation equipment. However, due to the simplifications and assumptions that are
made to develop these models, their use for research into the cutting process itself
have their limitations.

27
 28 3. Rock Cutting Process

This chapter starts with an overview of the cutting mechanisms and the phe-
nomenological models for (saturated) rock cutting in general is presented. This is
followed by an overview of the most relevant cutting parameters and how these af-
fect the cutting process itself. Many researchers have developed analytical and/or
semi-empirical models for rock cutting. Within this dissertation, the focus is on
the models for drilling and dredging/seabed mining. Although these models can be
powerful tools for the engineering and design of equipment, they are less suited to
investigate the rock cutting process itself. The assumptions and limitations of these
models are discussed as well.
3
3.1. Phenomenological rock cutting model
3.1.1. General concepts
The drilling and dredging/deep sea mining processes show many similarities. In all
cases saturated rock is cut at a higher hydrostatic pressures, although the increase
in hydrostatic pressure with respect to the strength of the rock is rather limited
for dredging applications. The most significant differences between drilling and
dredging/mining are the rake angle and the typical thickness of the layer being cut.
In drilling the rake angle is often negative while in dredging applications a positive
rake angle is more common. The definitions used in literature vary. For that reason,
in this work, a positive rake angle is defined as the cases where the cutting edge
precedes, and in the case of a negative rake angle the cutting edge follows in the
direction of motion, see figure 3.1 for clarification. The typical cutting thicknesses
are within the range of sub-millimeter to several millimeters for drilling, while in
dredging often layers of several centimeters thick are excavated in a single motion.

Figure 3.1: Definition of positive and negative rake angles.

Similar to the failure mechanisms that result in macro-scopic failure of a rock

specimen, the same failure mechanism can dominate the rock cutting process given
the tool, rock properties, environmental conditions, etc. These can be divided into
either continuous (ductile) plastic, discontinuous shear (brittle) dominated or dis-
continuous tensile dominated (brittle) cutting, as depicted in figure 3.2. These char-
acteristic failure mechanisms are often used as a basis for the various rock cutting
models, which are treated in section 3.2.
3.1. Phenomenological rock cutting model 29

The rock cutting process is a complex process to model, as it concerns the

transition of intact rock, through fragmentation towards granular media.

Figure 3.2: Dominant failure types in rock cutting.

A measure that is often used to compare rock cutting equipment is the specific
energy Esp , which is a property of the cutting process itself. Specific energy is
defined as the amount of energy that is needed to excavate a certain volume of rock,
which can be calculated as
RT
Pc dt Fc v c Fc
Esp = R 0T which is often simplified to = (3.1)
Qc dt tc wv c tcw
0

with total time T , power of equipment that is used to excavate the rock Pc , the
excavation production Qc , the cutting force Fc , the cutting velocity vc , cutting
depth tc and width of the tool w. Note that the dimension of Esp is often noted
as Pa, which is identical to J/m3 . In general, the Esp increases when the cutting
process enforces the rock to fragment into smaller pieces.

3.1.2. Dredging and seabed mining

van Kesteren (1995) developed a phenomenological model which explains the chip
formation process in terms of the critical state theory of rock failure, see figure
3.3. First the tool moves towards the rock, resulting in penetration of the rock. At
this stage, the rock directly underneath the tool undergoes crushing in compression,
creating a crushed zone. In the crushed zone (I), the stress state is beyond the
volumetric cap. At the boundary of the crushed zone (II), the stress state is on the
volumetric cap, near the brittle-ductile transition stress (BD). When the tool moves
further into the intact rock, the boundary of the crushed zone will move further into
the rock. The growth of the crushed zone is proportional to the penetration depth
of the tool. Just outside the crushed zone, shear failure will occur (III). Further
away in the intact rock, the stress state will be below the brittle-ductile transition
and the isotropic stress decreases. The decrease in stress results in a shear crack
that bifurcates into a tensile crack (IV). After bifurcation unstable crack growth
will result in chip formation. As is shown in figure 3.3 intact rock is subjected to
failure mechanisms along the complete failure envelope. To what extent each of
 30 3. Rock Cutting Process

the mechanisms contributes to the cutting process depends on the rock properties,
environmental conditions (e.g. ambient pressure, surrounding fluid) and the tool
(geometry, machining conditions).

chip
tool IV tensileL
crack
III shearLplane

I II
3 crushedLzone
BRITTLELFRACTURING DUCTILELCATACLASTICL
FAILURE

p
LM
deviatoricLcap qL=
II
gth BD
qL[MPa] st ren
ar L
he
ak
Ls
III gth
pe en volumetricLcap
str
u alL crushedLrock
re sid
I
IV
tension intactLrockVLstableLdeformation compression
pL[MPa]

Figure 3.3: Failure during rock cutting involves the entire failure envelope, given in a hydrostatic-
deviatoric stress diagram. Figure is based on Verhoef (1997). (p = 1/3(σ1 + 2σ3 ) and q = σ1 − σ3 )

3.1.3. Drilling
In drilling engineering, the cutting process is somewhat different. The main differ-
ences are the rake angle, which is in general negative for drilling purposes, and the
characteristic cutting depth. In drilling, cutting depths are in the range of (sub-
)millimeter scale compared to range of several centimeters for dredging applications.
Both due to the smaller cutting depth and the negative rake angle, the failure mech-
anisms mainly consist of the crushing of the rock along the whole cutting depth and
that the removal of material is in the ductile regime. Brittle failures still can occur,
but due to the negative rake angle mostly brittle shear occurs. Tensile failures are
less likely to occur because the negative rake angle has the tendency to constrain the
propagation of tensile cracks by forcing them to close. A resultant of the crushed
zone is the heavily damaged and plastic deformed layer that is left behind after the
tool has passed. In industry this is often referred to as filter cake. An overview is
given in figure 3.4.

3.1.4. Effect of geometry

Many aspects of the cutter geometry can have an influence on the rock cutting
process. Here only the most important geometrical parameters are discussed, being
the rake angle, cutting depth, wear flat and tool spacing.
3.1. Phenomenological rock cutting model 31

PDC cutter

Wear flat
Filter cake Shear zone
(grains and
pulverized rock) 3
Crushed zone
Plastically deformed
rock

Intact rock

Figure 3.4: Conceptual model of a boundary layer underneath a blunt drill bit, after Dagrain and
Richard (2006).

Cutting depth
Two modes of failure can be identified depending on the cutting depth (Richard
et al., 2012). At small cutting depths, (typically larger than the grain size and
less than 1 mm in sandstones), rock fails in a ductile mode, which looks like a
continuous plastic flow of crushed rock. The cutting mode is also often referred to as
scratching or grinding. As the cutting depth increases beyond a threshold the cutting
mode transitions towards a brittle chipping mode. The differences in behavior are
noticeable on force measurements as well. In measurements in the ductile regime,
the force signal looks more like white noise, while for measurements in the brittle
regime, the force signal behaves more like a saw -tooth pattern with peaks and valleys
corresponding to the chip formation (Verhoef, 1997; Huang, 1999). Chaput (1991)
made similar observations on the failure mechanism in rock cutting. Chaput also
noted that the critical cutting depth at which the brittle-ductile transition occurs
decreases with increasing uni-axial compressive strength of the rock. Based on
measurements of the cutting force with respect to cutting depth on Berea sandstone,
Richard et al. (1998) suggest that the critical depth at which the brittle-ductile
regime occurs scales according to

 2
KIc
tc,crit ∝ (3.2)
σc

with critical cutting depth tc,crit and fracture toughness KIc . The ductile regime is
associated with cohesion or compressive strength. Therefore the mean cutting force
 32 3. Rock Cutting Process

in the ductile regime scales according to (Richard et al., 1998)

Fc
∝ σc t c (3.3)
w
with cutting force Fc and width of the cutting tool w. In the brittle mode the
formation of cracks dominates the process. Therefore scaling of the brittle regime
is according to
Fc √
∝ KIc tc (3.4)
w
3 For a graphical representation of the scaling properties, see figure 3.5 Similar obser-
vations to the brittle-ductile transition with respect to cutting depth are observed
by Liefferink (2013); Motzheim (2016) in cutting experiments on frozen clay, al-
though in a different tool configuration, i.e. positive rake angle of 22◦ , cutting
depths ranging from 5 to 25 mm.
In discrete element simulations of He and Xu (2015), in which the micro-structure
of the simulated rock is varied, similar trends are observed. Rock cutting simula-
tions of intact homogeneous rock are compared with a heterogeneous rock simulation
based on a clustered particle assembly. Various simulations are performed with vary-
ing cutting depths. The obtained results show that the clustered particle assembly
has a less steep increase in cutting force compared to the homogeneous rock. Ad-
ditionally it is observed that in cutting simulations on the heterogeneous sample
failure occurs along the cluster boundaries when the cutting depth is of similar size
or larger than the characteristic length scale of the clusters. Both simulated rocks
have a similar compressive strength (155 MPa), although the tensile strength of
the heterogeneous sample is significantly lower (36 vs. 16.5 MPa), so the less steep
increase in cutting force for the heterogeneous sample might be as well a result of
the significant difference in tensile strength, but this was not tested by He and Xu
(2015). In rock cutting, the importance of the tensile strength of the rock is too
often neglected, while tensile failure can have a significant influence on the cutting
process.
2
���
tc,crit =
��
F

Brittle
Ductile
d
Figure 3.5: Scaling of cutting force with respect to cutting depth, based on Richard et al. (1998).
.

In scraper tests studied by Deketh (1995), in a setup similar to a turning machine.

In these experiments, Deketh observed a transition in cutting forces and cutting
regimes. However, instead of a cutting force that increases monotonously with
3.1. Phenomenological rock cutting model 33

(g/m)
0.1 I Mode5II Mode5III

rate5of5wear
dredge5steel
wedge5steel
0
350 normal5force
(N)
3
forces

cutting5force
(wedge5steel)
0
0 0.2 0.4 0.6 0.8 1.0 1.2
feed5(mm/rev)

Figure 3.6: Scaling of cutting force with respect to cutting depth and rate of wear with respect to
cutting depth, based on Deketh (1995).
.

respect to the cutting depth, as would have been expected based on the experimental
results of Richard et al. (1998), a significant peak is observed at the transition from
ductile cutting (scraping) to chipping, see figure 3.6. The difference in trends for the
cutting force with respect to the cutting depth between the experiments of Deketh
(1995) and Richard et al. (1998) is most likely due to the setup that has been used.
There are two explanations to why both set of experiments show different behavior
at the brittle ductile transition.
Both do scratching tests on cylindrical specimens, but the direction of the
scratch/cutting differs. Richard et al. (1998) performed the experiments in axial
direction (milling) of the specimen while Deketh (1995) did experiments in which
the main cutting direction is tangential along the cylindrical specimen (turning).
Besides, the difference in behavior between the two sets of experiments might as
well be caused by the difference in rake angle, being a positive rake angle in the
case of Deketh (1995) and a negative rake angle in the case of Richard et al. (1998).
Furthermore, the experiments of Deketh are performed on many different types
of mortar in which specific parameters are compared to see the effect of it in the
scraper test. Not all mortars have shown this type of behavior, based on the re-
sults of Deketh there is no direct relation between material parameter(s) and this
phenomenon.

Rake angle
Obviously, there is a large distinction to be made between positive and negative
rake angles. In general, negative rake angles are more often used in drilling and at
smaller cutting depths (in the range up to a few millimeters) and in tougher rocks,
while positive rake angles are typical for dredging and seabed mining applications,
 34 3. Rock Cutting Process

with typically large cutting depths (in the range of several centimeters) and is often
applied in weak rock (with UCS up to 20 MPa). Unfortunately, not many experi-
mental results for cutting experiments on saturated rock with varying rake angles
are published. This can either be due to the lack of the existence of such exper-
imental data, or that the existence of such data is kept as confidential within the
industry, this is especially the case for tools with a positive rake angle.
One of the few datasets with varying rake angles is that of Bilgin (1977), in which
small negative rake angles are used for cutting experiments on limestone and granite,
respectively having compressive strengths of 127 MPa and 179 MPa. The tensile
3 strength of these rocks are not presented by Bilgin, however, based on the rules of
thumb as given in appendix A, the tensile strengths of these rocks are estimated as
respectively as 12-16 and 6-10 MPa. See figure 3.7 for the peak cutting forces with
respect to the rake angle.

Peak horizontal cutting force

·103
20
Limestone
Granite

15
Fh [N ]

10

0
−22 −20 −18 −16 −14 −12 −10 −8 −6 −4 −2 0
◦
α[ ]

Figure 3.7: Effect of rake angle on peak cutting force, data from Bilgin (1977).

Detournay and Defourny (1992) propose, based on the rigid plasticity, three
types of flow regimes based on the (negative) rake angle, namely a forward flow
regime, a backward flow regime and a flow regime with a build-up edge, see fig-
ure 3.8. The intermediate regime with a build-up edge involves a static wedge of
dead material in contact with the tool, in which case both backward and forward
flow take place simultaneously. This has also been observed by Zijsling (1984) and
Adachi et al. (1996). Similar observations with respect to the occurrence of a build-
3.1. Phenomenological rock cutting model 35

up edge are made for metal (Eggleston et al., 1959), dry sand (Hettiaratchi and
Reece, 1967; Hatamura and Chijiiwa, 1975) and saturated sand (He and Vlasblom,
1998; Miedema and Frijters, 2003; Miedema, 2014). In the experiments of He and
Vlasblom (1998) vertical bars of colored sand were implemented in the sand cutting
setup. The observed deformation of these bars during the cutting experiment indi-
cate that the occurring wedge of built-up material is not static, grains in the wedge
are transported relative to the tool. However, the transport velocity of grains in the
wedge is significantly lower than that of the layer cut. See figure 3.9 for an impres-
sion of the transport of grains through the build-up edge. Numerical analysis of the
built-up edge has been performed by Huang (1999); Huang et al. (2013) through 3
the use of the discrete element method.

α2 α3
α1

α1>α2>α3
Figure 3.8: Flow regimes: forward flow, backward flow and flow with a build-up edge, all after
Detournay and Defourny (1992). Note that because of the definition of the rake angle used in this
thesis, α1 > α2 > α3 is valid.

Layer cut

Tool

Wedge

Figure 3.9: Flow lines in the dynamic wedge when cutting sand, after Miedema (2014).

Wear flat
Various researchers have investigated the wear of rock cutting tools (e.g. Deketh
(1995); Verhoef (1997); Dagrain and Richard (2006)). In general, the emphasis in
these studies is on the abrasive wear of the tool (catastrophic failure or breakage of
the tool is disregarded). Most studies distinguish abrasive and non-abrasive wear,
 36 3. Rock Cutting Process

based on the combination of tool material and rock. Most of the rock cutting tools
have the shape of a wedge, e.g. PDC-bit, dredging pick-point, dredging chisel. In
linear cutting processes, wear results in the removal of a prism of tool material from
the cutting edge of the bit.
Although the wear rate is an interesting topic for the research and development
of cutting tools, it is the change in shape of the tool that determines the rise in
forces. If a tool is able to maintain its shape while wearing down, it would result in
a less significant change in cutting force. Some rock cutting tools are designed in
such a way they are self-sharpening while during wear. Such a design property can
3 be used in closed form cutting processes like in drilling (Feenstra, 1988) or tunnel
boring (Buchi, 1984). Although there are some patents for dredging equipment
claiming that they describe self sharpening tools, e.g. (van Opstal, 2015), but the
cutter head is filled with bits instead of chisels or pick points.
Experience in practice reveals that the use of sharp tools is merely limited to a
short time before the tools become blunt by wear. In linear cutting processes, a wear
flat is generated. In drilling, the cutting depth per cutter typically ranges between
0.05 to 2 mm, which shows that PDC bits rarely can be considered as perfectly
sharp tools (Dagrain and Richard, 2006). Field and laboratory tests indicate that
the drilling response of a blunt bit is characterized by two distinct regimes, see figure
3.10. At low applied W , weight on bit (WOB), the drilling forces with respect to the
depth of cut (or rate of penetration (ROP)) is dominated by the frictional processes
taking place across the wear flats, as the effective contact area increases with W .
Once the maximum forces admissible on the wear flats are mobilized, the evolution
of the drilling forces is solely controlled by the cutting process taking place ahead
of the cutting faces. Obviously, the first regime is far less efficient.
Perfectly Blunt cutters
sharp cutter
l2 > l 1
l=0 l1
V Increasing wear
s
es
roc
gp
ttin

V1
cu
nt
icie
Eff

tional
Inefficient frinc
V2 < V 1 process
W

Figure 3.10: Rate of penetration versus weight on bit curves, based on Dagrain and Richard (2006).

Tool spacing
Thus far, the cutting process is only considered with respect to a single cutting tool.
Depending on the cutting tool, geometry and rock properties the cross-sectional area
of the groove that is cut by a single cutter is often larger than the projection of the
tool in that groove. Experiments of Alvarez Grima et al. (2015) clearly show that
3.1. Phenomenological rock cutting model 37

the shape of the groove is strongly influence by the operating conditions. In their
experiments, hydrostatic pressure and cutting velocity were varied. Figure 3.11
gives an impression to what extent the shape of the groove can be influenced by the
operating conditions.

width [mm] width [mm]

-100 -80 -60 -40 -20 0 20 40 60 80 100
-100 -80 -60 -40 -20 0 20 40 60 80 100
0
0

3
depth [mm]

depth [mm]
10 10

20 20

30 30
40 40
(a) (b)

width [mm] width [mm]

-100 -80 -60 -40 -20 0 20 40 60 80 100 -100 -80 -60 -40 -20 0 20 40 60 80 100
0 0
depth [mm]

depth [mm]

10 10
20 20
30 30
40 40

(c) (d)

Figure 3.11: Composition of laser scan cut geometry, a) ph = 0.1 MPa and vc = 0.2 m/s b) ph = 18
MPa and vc = 0.2 m/s, c) ph = 0.1 MPa and vc = 2.0 m/s and d) ph = 18 MPa and vc = 2.0
m/s. Each black line represents a single laser scan, the red lines represent the mean groove shape.
Data from Alvarez Grima et al. (2015).

Based on the knowledge of the shape of the groove that will be cut, the distance
between two consecutive tools can be adjusted to obtain a minimum in energy re-
quired to cut the bulk of the rock. Research into the effect of cutter spacing is
investigated by Roxborough and Phillips (1975) for applications in TBM’s. Rox-
borough observed the influence of the spacing and penetration of a disc cutter with
respect to the magnitude of the cutting forces and specific energy from a liner cutting
machine (LCM) test, which is a test on full scale. See figure 3.12 for an overview
of the observations made by Roxborough and Phillips (1975). Similar tests were
performed by Roxborough and Sen (1986) for the use of picks.
At a tool-spacing of 0, the second tool will move exactly in the shadow of the
tool producing the relieving groove, hence the forces and yield are 0. As spacing
increases so do the cutting forces and the yield. The maximum in yield is obtained
when the grooves have sufficiently distant that they create their own grooves, but
still have interaction with the groove that is closest by. Lateral cracks from the two
neighboring grooves combine to remove the ridge that would have otherwise existed.
As the spacing further increases, the grooves lose the interaction with neighboring
grooves, resulting in purely unrelieved cutting. See figure 3.13 for an overview,
where Θ is the edge angle of the cutter disc. Here yield ratio is defined as the ratio
of relieved (Qr ) to unrelieved (Qu ) rock cutting.
Various researchers have investigated what the optimum distance between the
cutting tools should be used in the design of the equipment. Evans (1972) estimates
that the optimum spacing for chisel shaped tools, by using the width of the chisel
 38 3. Rock Cutting Process

Disc forces Yield

30 2
FT hrust
FN ormal
1.5
20

[kN]
F [kN]

Qu
Qr
10
3 0.5

0 0
0 20 40 60 80 0 10 20 30
st [mm] st /p [mm]
(a) Disc forces (b) Yield ratios for relieved cutting
Specific energy

Θ = 100◦
100 Θ = 90◦
Θ = 70◦
Esp [MJ/m3 ]

Θ = 60◦

50

0
0 10 20 30
st /p [mm]
(c) Specific energy

Figure 3.12: Effect of spacing on various parameters, from Roxborough and Phillips (1975).

and the cutting depth, is determined as

s !
w 20t2
st = 0.001 1+ 1 + 2c (3.5)
2 w

with optimum spacing st , width of the tool w and depth of cut tc . For pick-points,
Evans (1984) found the following approximate relation
√
st = 2tc 3 (3.6)
For roller discs, Roxborough and Phillips (1975) determine the optimum spacing
3.1. Phenomenological rock cutting model 39

s s s s s s

d d d
Picks)too)close Optimum)spacing Loss)of)interaction

d=10)mm d=10)mm
specific)energy)(MJ/m3)

specific)energy)(MJ/m3)
8 8
d=20)mm d=20)mm
6
d=30)mm
6
d=30)mm 3
4 4

2 2

0 20 40 60 80 100 120 0 1 2 3 4 5 6 7
s/d
pick)spacing)(mm)

Figure 3.13: Effect of pick spacing on specific energy (Roxborough and Sen, 1986).

based on the rock properties to be

st σU CS
= (3.7)
tc τs

with the rock’s unconfined compressive strength σU CS and the rock’s simple shear
strength τs . Snowdon et al. (1982) shows that there is a constant stct ratio for min-
imum tool specific energy. However, they explain that there is no simple relation
between the optimized stct ratio and any combination of rock strength parameters.
Rostami and Ozdemir (1993) postulate that there are three different chipping pat-
terns in terms of specific spacing, being too close, optimum spacing and loss of
interaction due to a too large space between the tools, see figure 3.13.

3.1.5. Effect of rock properties

Verhoef (1997) makes a distinction between brittle and ductile cutting to explain
differences in wear rate with respect to the type of rock being cut, Verhoef makes
this distinction based on the rock properties. Verhoef’s definition of ductile cutting
might lead to misinterpretation, it is more accurate to refer to this as brittle shear
dominated or cataclastis dominated cutting. In such a case, the stress state in the
crushed zone reaches well above the brittle-ductile transition, resulting in a large
crushed zone near the tool tip. In the case of brittle cutting the stress level in the
crushed zone is slightly above the brittle-ductile transition, resulting in a relatively
smaller crushed zone.
Furthermore, the distinction between brittle tensile and brittle shear dominated
cutting processes can be made depending on the strength ratio as well. Unfortu-
nately, there is no clear consensus on what values of m the failure regimes occur.
Natau et al. (1991) states the criteria for the application of roadheaders (tunnel
boring) and Verhoef (1997) states the criteria for use in dredging applications, while
 40 3. Rock Cutting Process

both state that their criteria are based on Gehring (1987) (which is an internal
report of Voest Alpine Zeltweg). The criteria are presented in table 3.1

Table 3.1: Characteristics of strength ratio for rock cutting

Failure mode Natau et al. (1991) Verhoef (1997)

Roadheader Dredging
Brittle shear m<7 m<9
3 Intermediate
Brittle tensile
7<m<9
9<m
9 < m < 15
15 < m

Cools (1993) performed large-scale cutting experiments to measure the temper-

atures developing at the wear flat of a dredger cutting tooth during the cutting
of rock, see figure 3.14 for the geometry of the chisel that was used. In these ex-
periments, two 2.5 meter long blocks were cut, one block of limestone (UCS 25.2
MPa, BTS 1.5 MPa) and one block of sandstone (UCS 19.5 MPa, BTS 1.7 MPa).
High normal forces were measured indicating that during cutting the normal stress
beneath the machined wear flat amounted approx. 300 MPa in the limestone test,
and approx. 130 MPa in the sandstone test. Tri-axial test results indicated that the
brittle-ductile transition occurred at confining-maximum stress set (σ3 −σ1 ) of 23-96
MPa for the limestone and 41-119 MPa for the sandstone. Although the sandstone
is more abrasive, the experiments of Cools show that the wear rate of the limestone
cutting is significantly higher than that of the sandstone. He explains this by the
increase in temperature of the tool during cutting, in the sandstone temperatures
up to 550◦ C are measured, which is too low to cause thermal softening of the wear
flat. Temperatures measured in the limestone reached up to 800◦ C, high enough to
cause softening of the steel of the tool in a fraction of a mm thin zone at the wear
flat, leading to higher wear rate. With respect to the rock, Cools and Verhoef both
explain that this is caused by a large ductile crushed zone that occurs around the
chisel, based on the large normal stresses that have been measured.
Majidi et al. (2011) performed drilling experiments with varying hydrostatic
pressures and varying RPM on Carthage marble and Indiana limestone. They ob-
served that in case of atmospheric pressure, above a critical RPM, the Carthage
marble did not fail in a chipping mode, but the marble got crushed. In this crushed
regime they also observed that the required cutting force is actually lower than in
the chipping regime. When they increased the hydrostatic pressure, they noticed
that the reduction in required cutting force with respect to lower RPM was lost.
Their explanation for the disappearance of this reduction is that the friction of the
crushed material along the bit face is higher due to the low pore pressures at the
bit face. However, they do not give an explanation for the transition from chip-
ping to crushed cutting. A possible explanation for the transition from chipping
to crushed cutting could be that with increasing cutting velocity the rock behaves
in an undrained fashion, leading to compaction weakening of the rock. Elevated
hydrostatic pressure might prevent the compaction weakened material to dilate, re-
sulting in no significant difference in failure mechanism with respect to low and high
3.1. Phenomenological rock cutting model 41

3
41 54
58
4

Figure 3.14: Chisel geometry as used in the experiments of Cools (1993).

velocity cutting.

3.1.6. Effect of cutting speed

Most of the experimental results for coal and rock cutting (in dry conditions) in-
dicate that the cutting force is not affected by the cutting velocity in its practical
range of applications (Potts and Shuttleworth, 1958; O’Dogherty and Burney, 1963;
Roxborough, 1973). For example, tensile fractures in coal propagate at a speed
of more than 500 m/s, which is more than 100 times faster than in practical rock
cutting applications. However, it is noted that the cutting velocity can have a signif-
icant influence on the tool life, once the cutting velocity is above a critical velocity
(Kenney and Johnson, 1976; Hurt and MacAndrew, 1985; von den Driesch, 1994).
Velocity effects might be of relevance when cutting saturated rock. van Kesteren
(1995) describes two limit conditions, drained and undrained. In the drained condi-
tions, flow of the pore fluid due to the pressure gradients in the pore fluid is possible
without affecting the mechanical behavior of the rock skeleton. In the undrained
case, deformation of the rock skeleton is that rapid that the pore fluid cannot flow
sufficiently to relieve the pore fluid pressure gradients. As a result, the pore fluid
contributes to the mechanical response of the rock skeleton. In a similar fashion to
the rate dependent effects as described in section 2.3.2, van Kesteren (1995) distin-
guishes these two limit conditions based on the pore Peclet number, which shows
a strong resemblance to the Debora number in equation (2.15). The pore Peclet
number is given by

vc t c vc tc η (Cf − αCs + n (Cp − Cs ))

ζP e = = (3.8)
D κ
According to van Kesteren (1995), the drained response is valid when ζP e < 1 and
undrained behavior will occur when ζP e > 10. Detournay and Atkinson (2000)
independently defined a similar dimensionless number to distinguish drained and
 42 3. Rock Cutting Process

undrained behavior. However, the criteria they use, based on the form as in equation
(3.8), are drained behavior when ζP e < 0.004 and undrained behavior when ζP e >
40. No direct comparison is made between the two criteria, the difference between
the two criteria is likely due to the differences in cutting processes, i.e. a negative
rake angle at small cutting depths in the case of Detournay and Atkinson (2000)
and a positive rake angle at large cutting depths in the case of van Kesteren (1995).
In the experimental program that Van Kesteren uses as a basis for his theory,
it is explained that the drained/undrained conditions are more directly concerned
with the size of the crushed zone (Combinatie Speurwerk Baggertechniek, 1984). In
3 the drained case, fluid can flow sufficiently away from the crushed zone and as a
result the crushed zone shows significant compaction, resulting in a large crushed
zone. On the other hand, in the undrained case the crushed zone is significantly
smaller. This is because the fluid cannot flow sufficiently from the crushed zone and
as a result the fluid in the crushed zone will be compressed, effectively increasing
the bulk modulus of the rock in the crushed zone, resulting in a lower critical strain
that is needed until chipping of the rock will occur.
Uittenbogaard (1980) and Luger (1981) performed indention tests with large
cones on saturated rock to investigate the influence of a pore fluid (water) on rock
deformation and its relation to velocity and length scales. The experiments are
performed on St. Lieu limestone and WL II, an artificial type of rock developed
by Delft Hydraulics for dredging research (Uittenbogaard, 1980; Luger, 1981; van
Kesteren, 1995). The intrinsic permeability of the two rocks is respectively 3 · 10−14
and 10−17 m2 . The cone has a top angle of 60◦ and is used to penetrate to depths
up to 32 mm. The range of velocities is chosen in such a way that both drained and
undrained cases are tested. At low velocity (0.01 m/s), the pore fluid is allowed to
flow out of the porous system. In a zone around the cone the rock is compacted
to a level that balances the penetrated volume of the cone. At a larger penetration
rate, the penetrated volume cannot be balanced anymore by pure compaction of
the rock matrix. As a result, the elevated fluid pressure in the rock will lead to
shearing towards the free surface, eventually resulting in chips. At a velocity of 1
m/s the cone penetration occurs in an undrained regime. Due to the high porefluid
pressures, the crushed rock is liquefied and squeezed out of the specimen along the
cone face. See figure 3.15 for an overview of these failure phenomena.
Crack initiation and propagation are influenced by a pore fluid as well. The
growth of the macro-crack and the micro-cracks in the shear zone is determined by
the flow of the pore fluid towards the dilating zone. The resulting pressure gradient
prohibits dilation and reduces the stress concentrations at the crack tips. Depending
on the applied strain rate and (initial) pore pressure, the pore pressure gradient can
result in a higher apparent shear or tensile strength.

3.1.7. Effect of hydrostatic pressure

Alvarez Grima et al. (2015) performed rock cutting experiments on saturated Savon-
nieres limestone at elevated hydrostatic pressures and large cutting depths (ca 20
mm). Their work focuses on the effect of a high hydrostatic pressure (to which they
refer as hyperbaric pressure) and cutting velocity with respect to the cutting process
3.1. Phenomenological rock cutting model 43

Figure 3.15: Failure mechanisms in cone indention tests in saturated st. Leu limestone (κ =
3.6 · 10−14 m2 ), with varying penetration velocity, after van Kesteren (1995).

with a positive rake angle. In these experiments, the hydrostatic pressure ranges
from 0-18 MPa, compared to the rock strengths σc = 7.92-10.64 MPa and σt =
0.86-1.13 MPa. The tests were executed in a hyperbaric tank and were performed
on a 1:1 scale. In total, test results of fifteen experiments are presented, in which
the hydrostatic pressure varies between 0, 1.5, 3, 6 and 18 MPa and the cutting
velocity varies between 0.01, 0.2, 0.6, 1.2 and 2.0 m/s. The cutting experiments are
performed with a positive rake angle of 22 ◦ with respect to the vertical axis, a wear
angle of -10◦ , a chisel width of 21 mm and a cutting depth of 20 mm.
The average cutting force in horizontal direction with respect to hydrostatic
pressure and cutting velocity is respectively shown in figures 3.16 and 3.17. The
experimental results clearly show that with increasing hydrostatic pressure (water
depth) the cutting forces increase as well. Besides that, the cutting force increases
with cutting velocity as well. It seems that there are two regimes that can be
distinguished based on these experiments, after ph = 6 MPa, the cutting force
seems to increase more rapidly for experiments with high cutting velocity. Alvarez
Grima et al. (2015) explain that in rapid deformations the hydrostatic pressure acts
as an additional resistance to deformation of the rock. The extent to which this can
contribute to resist deformation of the rock is limited by cavitation (similar to the
dilatancy hardening concept as explained in 2.3.2). It is expected that the size of
the crushed zone increases, the shear failure at the edge of the crushed zone occurs
later and that bifurcation of the shear crack towards a tensile crack is postponed.
As a result, the cutting forces will increase and the average chip size will reduce.
See figure 3.18 for a visualization of this explanation.
Particle size distributions of the debris and measurements of the groove that is
created by a single cut support this theory. With an increase in hydrostatic pressure,
the side breakout angle of the chips becomes more steep and the averaged cross-
 44 3. Rock Cutting Process

Average cutting force

·103
30
vc = 0.01m/s
vc = 0.2m/s
25
vc = 0.6m/s
vc = 1.2m/s
20 vc = 2.0m/s

3
Fh [N ]

15

10

0
105 106 107
ph [P a]

Figure 3.16: Average cutting force with respect to hydrostatic pressure, data from Alvarez Grima
et al. (2015).

sectional area of the groove significantly reduces, as is already shown in figure 3.11
in section 3.1.4.
The drilling industry was, until the early 1950’s, unaware of the effect of down-
hole pressure on the rock and thus the drilling process. Kuhne (1952) suggested
that downhole pressure strengthens the rock and that a Mohr-Coulomb criterion
may be used for taking the effect of strengthening into account. Through experi-
ments, Cunningham and Eenink (1959) point out that it is not the effective stress
that determines the strengthening of the rock matrix, but that the differential pres-
sure (the difference between borehole pressure and the pore pressure). Furthermore,
the experiments of Cunningham and Eenink (1959) also show that differential pres-
sure had a more profound effect on the rate of penetration than would be expected
by the increase in strength as a result of a Mohr-Coulomb material and that there
had to be other mechanisms at work. Garnier and van Lingen (1958) suggested this
to be the phenomenon of "chip hold down". Chip hold down refers to the force that
the drilling fluid may exert on a cutting, or a bed of crushed material, due to the
differential pressure.
Additionally, the industry also recognized that the permeability of the intact
rock can have a significant influence on the differential pressure (Feenstra and van
Leeuwen, 1964). When a drill bit shears rock, the rock will dilate, causing the pore
3.1. Phenomenological rock cutting model 45

Average cutting force

·103
30
ph = 0.1 MPa
ph = 1.5 MPa
25
ph = 3 MPa
ph = 6 MPa
20 ph = 18 MPa

3
Fh [N ]

15

10

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
vc [m/s]

Figure 3.17: Average cutting force with respect to cutting velocity, data from Alvarez Grima et al.
(2015).

Figure 3.18: Trend of change in failure type with increasing water depth, after Alvarez Grima et al.
(2015). An increase in water depth leads to larger cutting forces, larger crushed zone and smaller
cuttings, resulting in a larger specific energy.

volume to increase. At low permeabilities this will cause a reduction of the pore
pressure, that increases the differential pressure and ultimately strengthening the
rock. This effect is documented in several laboratory experiments concerning the
 46 3. Rock Cutting Process

cutting of shales, (Zijsling, 1987; Cook et al., 1991; Andersen and Azar, 1993; Gray-
Stephens et al., 1994), as well as it is studied analytically, (Kolle, 1996; Detournay
and Tan, 2002).
Full scale experiments of Black et al. (2008) and Judzis et al. (2009) have shown
that the strengthening of the rock due to shear dilation only partly explains the
reduction in efficiency and rate of penetration (ROP) when cutting rock under
pressure. In the experiments of Judzis et al. (2009) a significant reduction in the
penetration rate is noted. They explain that this is due to the unproductive work
that is done on the crushed material that is held in place on the hole bottom by the
3 local differential pressures. This effect is also known as a filter-cake, which acts as
a layer of lower permeability, restricting the inflow of drilling fluid into the virgin
rock. As a result, the pressure difference between the bottomhole pressure and the
pore pressure remains larger, and thus requires more energy to actually cut a second
time through a completely disintegrated rock.
In drilling experiments on Mancos shale, Zijsling (1987) observes that the cutting
force is proportional to the total bottomhole pressure. Furthermore, the condition-
ing pressures do not have a significant effect on the cutting forces, implying that
the bottomhole pressure rather than the overbalance (and thus the pore pressure
level in the sample) governs the cutting process of a PDC in Mancos shale, see
figure 3.19. In similar tests on Pierre shale, Zijsling observes a significant effect of
the conditioning process of the shale samples on the cutting forces, see figure 3.20.
The effect of the total bottomhole pressure on the cutting process is apparent at
pressure levels in excess of 20 MPa, at which the cutting forces no longer increase
with increasing total bottomhole pressure. Zijsling gives two arguments that can
explain the lower effect of bottomhole pressure on the Pierre shale samples. First
of all, it is expected that it has a less pronounced effect of dilation during shearing
due to its higher porosity. Secondly, the dilation is governed by relatively weak clay
particles sliding across each other. These particles will fail if the contact stress,
which depends on the bottomhole pressure, exceeds the strength of the particles.
As a result, the skeleton fails and the cavitation limit is not reached, thus a lower
pressure difference needs to be overcome to cut through the rock.
To analyze the rock-drilling process from an energy point of view, Simon(1963)
and Teale (1965) use the concept of mechanical specific energy Esp . In their experi-
ments they find that the Esp for drilling rocks under atmospheric conditions strongly
correlates with the UCS of the rock. Pessier and Fear (1992) questioned whether
the same comparison could be done under pressurized conditions using the confined
compressive strength instead of the UCS. However, in all their experiments, even
under ideal hydraulic conditions, the Esp they have measured was much higher than
the confined compressive strength of that rock under the same confining pressures.

3.1.8. Identification of other parameters

Mishnaevsky Jr (1995) states that probably up to 90 % of the energy for cutting
is spent in the crushing of the rock near tool tip. Analysis of Mishnaevsky also
shows that the confining stresses near the tool tip can be very high, high enough to
exceed the brittle-ductile transition in rock. This explains why rock can be cut in a
3.2. Analytic rock cutting models 47

Average cutting force

3
unconditioned tc = 0.3 mm
pc = 35M P a
2.5
pc = 30M P a
pc = 21M P a
2

3
Fh [kN ]

1.5

tc = 0.15 mm
0.5

0
0 10 20 30 40 50 60 70 80 90 100
Bottomhole pressure [M P a]

Figure 3.19: Cutting forces in Mancos shale with respect to bottomhole pressure and borehole
depth, data from Zijsling (1987).

ductile mode. Based on rock cutting experiments, Chiaia (2001) concludes that the
cutting performance of a tool can be significantly improved by reducing the crushing
component and enhancing the chipping ability of the tools.
On the basis of test results, the ratio of peak to mean cutting force is reported as
5-7 in coal cutting (Pomeroy, 1963). In rock cutting, more dispersed values are re-
ported, 1.4-7.7 for the wedge shaped bit (Muirhead and Glossop, 1968; Roxborough,
1973) and 1.6-3.0 for pick points (Kenney and Johnson, 1976; Hurt, 1980; Goktan
and Gunes, 2005)

3.2. Analytic rock cutting models

Various rock cutting models exist in literature. An overview of the existing rock cut-
ting models will be presented. These models can be distinguished into two groups,
i.e. analytic and semi-empirical. Due to the different assumptions and parameters,
a direct comparison between these models is difficult. Several dominating failure
processes are identified (Helmons and Miedema, 2013), as is already shown in figure
3.2. In general a distinction can be made between continuous and discontinuous
failure mechanisms. In a continuous failure, the cut material moves as a plastic flow
along the tool (or it can ball up over a portion of the tool length). This typically
occurs for cutting in cohesionless materials (e.g. sand, Miedema (1987)) and it can
 48 3. Rock Cutting Process

Average cutting force

0.6
pc = 16M P a
pc = 8M P a
0.5
unconditioned

0.4

3
Fh [kN ]

0.3

0.2

0.1

0
0 5 10 15 20 25 30 35 40 45 50
Bottomhole pressure [M P a]

Figure 3.20: Cutting forces in Pierre shale with respect to bottomhole pressure and borehole depth
at cutting depth of 0.3 mm, data from Zijsling (1987).

occur in drilling processes, as is the case in the experiments of Zijsling (1987) on

Mancos and Pierre shale. The other group of cutting processes is based on discon-
tinuous cutting processes. Within this group a distinction can be made between
tensile and shear dominated failure processes. These failure mechanisms are used
for the basis of the analytical models.
In literature, various analytic rock cutting models exist. Due to the complexity
of the rock cutting process, these analytical models are often limited to a certain
dominant failure mechanism. Note that the way the models are described here might
differ from the exact form that is described in the original work of the authors. Here
all models are adjusted to be applicable in the same frame of reference, that is with
respect to rake angle, compressive forces and stresses are positive, tensile forces are
negative.

Evans
The cutting model of Evans (1965) is specifically for tensile failure dominated cutting
processes. His model is based on the following assumptions:

• The process is two dimensional.

• No velocity effects are considered.

3.2. Analytic rock cutting models 49

• Movement is in horizontal direction.

• The crack to separate the chip occurs instantly.
• The penetration of the wedge is small compared to cutting depth.
A schematic representation of Evans (1965) model is given in figure 3.21. It is
assumed that along the curve C-D, tensile strength of the rock occurs simultaneously.
The exact shape of C-D is determined by a minimizing the required energy to
generate the chip with respect to the shear angle β. In the original model assumes
symmetry, resulting in zero total vertical cutting force. The total horizontal cutting 3
force and the vertical cutting force per side of the tool are given by respectively

2 sin (θ + δ)
Fh,Evans = σt tc w (3.9)
1 − sin (θ + δ)

cos (θ + δ)
Fv,Evans = σt tc w (3.10)
1 − sin (θ + δ)
with tensile strength σt , cutting depth tc , cutting width w, half top angle of the
tool θ and external friction angle (tool-rock) δ.

β β

r R

δ
D
A
β tc
Fc θ
C
θ

δ R

Figure 3.21: Rock cutting model of Evans (1965) for tensile chipping.

Evans (1984) presents an improvement of the cutting model for the use of conical
pick points. In the case of conical picks, the assumption of a 2D model is not valid
anymore. Total cutting force

16πσt2 t2c
Fc,Evans−pickpoint = (3.11)
σc cos2 (θ)
 50 3. Rock Cutting Process

Goktan (1997) made some suggestions to improve the theory of Evans (1984), re-
sulting in a cutting force calculated by

4πσt t2c sin2 (θ + δ)

Fc,Evans−Goktan = (3.12)
cos (θ + δ)
resulting in elimination of one of the shortcomings identified by Evans himself, that
the cutting force does not reduce to zero when θ = 0◦ for a frictionless tool (δ = 0◦ )
and Goktan eliminated the compressive strength in the model for tensile failures, see
3 equation (3.11). A semi-empirical model is derived by Goktan and Gunes (2005),
by fitting equation (3.12) to experimental data, resulting in

12πσt t2c sin2 45 − α2 + δ

Fc,Goktan = (3.13)
cos 45 − α2 + δ

Furthermore, Goktan found that the mean cutting force is about one third of the
peak cutting force as defined in equation (3.13). A similar observation is made by
Ranman (1985).
Vlasblom (in (Miedema, 2014)) presents an extension of the Evans model for the
use of worn chisels, by assuming that the tool can enter the rock under an angle ε,
here chosen to have the same value as θ, and that it has a wear flat, as presented in
figure 3.22. Here the direction of the cutting velocity is still horizontal. The cutting
force, decomposed in horizontal and vertical components, are respectively given by
 
2 sin (θ + δ) cos (θ) + cos (δ)
Fh,Evans−wearf lat = σt tc w (3.14)
1 − sin (2θ + δ)
 
2 sin (θ + δ) sin (θ) + sin (δ)
Fv,Evans−wearf lat = σt tc w (3.15)
1 − sin (2θ + δ)

Merchant
Most of the (semi-)analytical rock cutting models are adaptations of the Merchant
(1945) cutting model for metals. The model is based on plastic deformation of the
metal and a continuous chip formation. In this theory, and in the extensions to this
theory, the following assumptions are made
• Tool width is much larger than the depth of the cut, i.e. plane strain.
• Tool moves with constant velocity over a constant depth.
An overview of the geometry and parameters of the Merchant cutting model and its
corresponding extensions is presented in figure 3.23.
Furthermore, the following assumptions are specifically for Merchant’s model
• Constant stress state.
• Coulomb friction acts along tool-chip surface.
• Plastic yield of the metal is based on Mohr-Coulomb criterion.
3.2. Analytic rock cutting models 51

ε
r
β
β

T
Fc R
δ D
A
θ t 3
θ
C
β tc

R
δ

Figure 3.22: Extension of Evans’ model with wear flat and tool entrance under an angle, according
to Miedema (2014).

vc

N2
δ S2
htool K2
α S1
C
ϕ N1
K1
β tc

Figure 3.23: Model definition of Merchant (1945), combined with the extensions based on
Miedema’s model.

• Failure stress acts along the whole shear plane.

Based on these assumptions and through equilibrium of forces, the cutting force is
derived as
cos (α − δ)
Fh,M erchant = τ0 tc w (3.16)
sin (β) cos (α − β − δ)
cos (α − δ) 1
Fv,M erchant = τ0 tc w (3.17)
sin (β) cos (α − β − δ) tan (δ − α)
 52 3. Rock Cutting Process

with yield stress in shear τ0 , rake angle α and shear angle β. Note that the shear
angle β is a parameter that is not directly known. In order to obtain β, the minimum
energy principle is used. Minimization of the cutting force with respect to β gives
π α−δ
β= + (3.18)
4 2
By determining the minimum of the cutting force with respect to β, one finds
meaning that the minimum in cutting force is determined with respect to β.
3 Nishimatsu
The model of Nishimatsu (1972) is specifically for brittle shear failure dominated
rock cutting. This model is based on the Merchant model, with the following as-
sumptions:
• Brittle shear failure occurs according to a linear Mohr envelope.
• Failure through indention of the tool is neglected, i.e. no crushed zone.
 n
tc
• Stress distribution along the shear plane distributed as σ = σ0 sin(β) −λ .
Where n represents a stress distribution factor, λ is the distance from tool tip
along the shear plane to the surface.
A graphical representation of the geometry used in Nishimatsu’s model is shown in
figure 3.23 as well. The cutting force according to the Nishimatsu model, decom-
posed in horizontal and vertical component, respectively given by
1 2ctc w cos (φ) cos (α − δ)
Fh,N ishimatsu = (3.19)
1 + n 1 + sin (α − δ − φ)
1 2ctc w cos (φ) sin (α − δ)
Fv,N ishimatsu = (3.20)
1 + n 1 + sin (α − δ − φ)
where n stands for the stress distribution along the shear plane. Nishimatsu states
that n depends on the rake angle, based on the empirical relation

n = 11.3 − 0.18α (3.21)

This empirical relation is based on experiments of Nishimatsu (1972) in the range

of α = 10 − 40◦ . Note that beyond this range, the validity of this equation can be
questioned, especially for rake angles larger than 68◦ where the sign of n changes.
In that case, the cutting force changes sign as well and unrealistic results will occur.

Detournay and Atkinson

In the case of drilling, the problem definition differs from the models described
before. Detournay and Atkinson (2000) derive a model based on the facts that the
rock now is saturated with a pore fluid at p0 , the surface of the rock is exposed to a
constant hydrostatic pressure ph , caused by mud pressure, and the rock is considered
to be shear dilatant. Detournay and Atkinson introduce additional assumptions to
extend Merchant’s model:
3.2. Analytic rock cutting models 53

• The effective normal stress and (σn − p) and the shear stress σs across the
failure plane satisfy the Mohr-Coulomb criterion
|σs | = c + (σn − p) tan φ (3.22)

• The effective normal stress and shear stress across across the tool-rock interface
are related by the frictional law
|σs | = (σn − p) tan δ (3.23)
where δ is the contact friction angle. 3
• The interstitial fluid pressure along the cutter/rock interface is equal to ph
• The pore pressure variation in the intact rock is governed by pore pressure
diffusion equation.
Based on these assumptions, the cutting forces are given by
2 cos φ cos (δ − α)
Fh,Detournay = tc w λ (c + (pm − pb ) tan φ) (3.24)
1 − cos (φ + δ − α)
2 cos φ sin (δ − α)
Fv,Detournay = tc w λ (c + (pm − pb ) tan φ) (3.25)
1 − cos (φ + δ − α)
with cavitating limit factor λ. In the case of a fully cavitating cutting process λ = 1,
in the limit that the cutting process is infinitely slow, the fluid pressure component
does not contribute at all to the cutting force, and thus in such a case λ = 0. The
corresponding angle of the shear plane with respect to the direction of motion of
the cutter is given by (Detournay and Atkinson, 2000)
π α+φ+δ
β= − (3.26)
4 2
The expression for β indicates that the Merchant failure mechanism can only be
constructed when
π
α < − (φ + δ) (3.27)
2
When the rake angle approaches the upper bound given by (3.27), another flow
mechanism may occur. This flow mechanism involves the presence of a wedge of
material sticking to the cutting face (Detournay and Defourny, 1992). They suggest
the existence of three flow regimes (all considering negative rake angles): forward
flow at small rake angles (sub-vertical cutter), flow with a build-up edge at inter-
mediate rake angles and backward flow at large rake angles (sub-horizontal). In
the experiments of Zijsling (1984); Adachi et al. (1996) a built-up edge of crushed
material in front of the PDC-cutter is observed, supporting the claims of Detournay
and Defourny (1992).
The pressure drop that can be attained in the dilating shear zone depends on
the pore Peclet number (as is discussed in 3.1.6). At low pore Peclet numbers the
pressure difference will be insignificantly small, resulting in a simplified form of
equations (3.24) and (3.25), which is similar to Nishimatsu’s model.
 54 3. Rock Cutting Process

Miedema
The cutting model of Miedema (2014) finds its origin as an adaptation of Mer-
chant’s model to be applicable to the cutting of (saturated) sand (Miedema, 1987).
Later on, Miedema extended this model to be applicable to the cutting of clay and
rock by incorporating brittle tensile and brittle shear failure criteria in the cutting
model. Furthermore, Miedema’s model has been extended to incorporate effects of
hydrostatic pressure and to cases where the crushed zone reaches the surface of the
rock (drilling). Miedema (2014) gives an extensive derivation for all the different
cases and adjustments, here only the most relevant aspects of this model are pre-
3 sented. Besides the basic assumptions of Merchant’s model, the following additional
assumptions are made:

• Failure occurs instantly.

• Hydrostatic pressure (water depth) and pore pressure of virgin rock are in
equilibrium.

• The rock is shear-dilatant.

• Two zones with underpressure are assumed 1) along the shear zone between
the cut layer and the virgin rock and 2) between the cut layer and the tool.

The model follows a similar structure for each failure mechanism, the forces as
depicted in figure 3.23 acting on the cut material and through that on the seabed
and the tool are calculated and again equilibrium of forces is assumed. The resultant
force between the cut material and the tool K2 is given by

W2 cos (α − β − φ) + W1 sin (φ) + C cos (φ)

K2 = (3.28)
cos (α − β − φ − δ)

with pore pressure force acting along the shear plane W1 , pore pressure force acting
along the tool W2 , cohesive force C. The maximum pore pressure forces for the
cavitating (maximum) and non-cavitating are respectively given by

ρf g (z + 10) tc w p1m tc w
W1 = or W1 = (3.29)
sin (β) sin (β)

ρf g (z + 10) htool w p2m htool w

W2 = or W2 = (3.30)
cos (α) cos (α)
and the cohesive force is given by
ctc w
C= (3.31)
sin (β)

resulting in the total cutting force given by

This results in the horizontal and vertical cutting forces, respectively given by

Fh,M iedema = −W2 cos (α) + K2 cos (α − δ) (3.32)

3.3. Discussion rock cutting models 55

Fv,M iedema = −W2 sin (α) + K2 sin (α − δ) (3.33)

Note that in shallow water conditions (low hydrostatic pressure with respect to the
strength of the rock), the pore pressure terms W1 and W2 are negligible. However,
the cutting model underestimates the cutting forces measured in experiments of
Alvarez Grima et al. (2015) significantly. Therefore an empirical correction factor
is used, which is of the form
  
c1 c2
ccorr = 1 − 1+ (3.34)
z + 10 z + 10 + c3 3
with values c1 = 3.33, c2 = 200 and c3 = 400. Care must be taken when using this
correction, as it is only based on one set of experiments. Additionally, the correction
factor they use is to correct for the upper limit (with respect to cutting velocity)
of the mean rock cutting forces, while the model derivation is based on the peak
cutting forces. Furthermore, it can be argued whether the pore pressure force W2 is
actually valid in the case of brittle cutting processes. The debris of the experiments
of Alvarez Grima et al. (2015) still show rather large chips (up to several cm’s in
size) even for cutting experiments at ph = 18 MPa. How the debris moves along
the tool during cutting is unclear, since it was not possible to visualize due to the
pressure tank.

3.3. Discussion rock cutting models

The analytical rock cutting models can be used as a means to design and engineer
the equipment used to cut through rock. However, one must be careful when using
these models because of the simplifications that are made and the limited range
of applicability of these models. Basically all of the described models have similar
problems, that to some extent can be corrected for.
First of all, all the models that are applied for brittle rock failure (both shear
and tensile) can only be used to determine the peak value of the cutting forces,
not the mean cutting force. Furthermore, all models essentially assume 2D cutting
processes, while in practice rock cutting is almost always a 3D process. In 2D, the
cut layer can only move directly along the tool (or curling/balling up might occur,
(Zijsling, 1987; Miedema and Zijsling, 2012)), neglecting the possibilities of sideways
movement of the cut material. Furthermore, it can be argued whether the assump-
tion that the debris acts as a plastic layer moving along the tool surface can be true.
The assumption could be valid, because of the crushed material that is generated
in the crushed zone is forced to move along the tool surface. However, it might
as well be the case that the crushed material is not forced along the tool surface
but that it moves along with the generated chips. Unfortunately, such observations
are not described in the work of Alvarez Grima et al. (2015). Furthermore, it is
observed that the size of the crushed zone significantly differs in 2D tile cutting ex-
periments (Uittenbogaard, 1980; Luger, 1981) compared to 3D cutting experiments
(Combinatie Speurwerk Baggertechniek, 1984). Additionally, the break-out shape
of the groove cut by one chisel and the beneficiary effects of tool spacing are not
considered.
 56 3. Rock Cutting Process

All the presented rock cutting models are linear with respect to the cutting depth,
while in experiments of Richard et al. (1998), Liefferink (2013) and Motzheim (2016)
a transition is observed, at small cutting depths the cutting force scales as√Fc ∝ tc ,
while for larger cutting depths the cutting force scales according to Fc ∝ tc .
A large amount of the equipment used in dredging and seabed mining is based
on circular motion (e.g. cutterhead, roadheader, drumcutter), while the cutting
models that are used for engineering of the equipment are based on linear cutting
models. The movement from a small cutting depth towards a large cutting depth
or vice versa might result in different behavior than that would be expected of the
3 linear cutting models because the assumption that the velocity of the tool is parallel
to the surface of the rock is not met, it actually changes during the movement of the
tool. Besides the lack of capturing the brittle-ductile transition with cutting depth,
differences might occur as well in e.g. friction, specific energy, wear rate.
Basically all analytic cutting models are based on peak force estimations, while
in practice averaged cutting forces are of interest as well. Average to peak cutting
forces are thus far estimated based on empiricism. Besides, all models are based on
constant parameters over time and assume rigid body movements of the tool. As
a result, effects like oscillations on the cutting motion (e.g. percussion) or inertia
effects of the complete cutting tool are neglected.

3.4. Conclusions
It is shown that the cutting of rock is a complex process. All failure types along the
failure envelope can occur in a single cut. Many parameters can have a significant
influence on the cutting process itself, e.g. rake angle, cutting depth, rock properties,
ambient conditions, pore fluid, tool geometry, velocity.
The (empirical-)analytical models can be a powerful tool to aid the design of the
rock cutting equipment. However, most of the models require the failure process to
be identified a priori, while it is desired to have this as a result rather than an input
parameter. If one wants to be able to investigate the effect of hydrostatic pressure
and pore fluid pressure on the rock cutting process, beyond the level of rock cutting
forces, a more detailed approach compared to the presented rock cutting models
is necessary. Furthermore, variations in time like inertia, oscillations and a more
complex shape of the cutting tool are not captured with these cutting models.
Numerical techniques are powerful tools to aid in the research of the rock cutting
process. What simulation techniques already exist and have been used to simulate
rock cutting processes, as well as the modeling approach that is used within this
dissertation, will be discussed in the next chapter.
 4
Modeling Approach
"If the facts don’t fit the theory, change the facts."
Albert Einstein

As presented in previous chapters, the combination of the physical phenomena, to-

gether with the fact that the whole range of failure mechanisms occurs in the cutting
process, make it that the cutting of fluid saturated rock is a complex process to model
and simulate. At the start of this research, there was no methodology available that
was sufficiently capable to model the cutting of saturated rock. First an overview is
presented on the state-of-the-art of rock cutting simulations and what attempts have
been made to incorporate fluid pressure effects in rock deformations. Furthermore
the ideas and the approach on how the rock cutting process is modeled within this
research is presented.

57
 58 4. Modeling Approach

Simulating the rock cutting process of a single tool requires a model that is
capable of capturing the transition from intact rock through (micro-)cracking to
fragmented rock, ultimately leading to a granular flow. Additionally, the failure
mechanisms that need to be considered should be able to cover the whole failure
envelope, ranging from the brittle tensile and shear fractures towards cataclastic
failure.

4.1. Numerical modeling of rock cutting

In general, a distinction can be made between continuum based and discontinuum
based modeling techniques.

4
4.1.1. Continuum based modeling
Several attempts have been made to model the rock cutting process with the use
of a continuum based method, e.g. the Finite Element Method (FEM), e.g. (van
Kesteren, 1995; Liu et al., 2002) . However, in the case that such a methodology is
used, several problems arise. First of all, for continuum based problems one needs to
use a set of equations that is capable of describing the desired physical model. The
more complex the physical problem is, the more difficult it will be to find, derive
and solve the required set of equations. In the case of rock cutting, it is especially
difficult because it has to deal with the whole range from intact rock to granular
flow, also combined with fluid pressure.
Secondly, mesh-based continuum methods, like FEM, have difficulties with the
large deformations that are applied in the cutting process. In the case of brittle
failures it requires that the elements through which the cracks occur can be split.
If too much distortion is applied to the underlying mesh, remeshing might become
necessary. For that reason, various researches avoid this problem by investigating
the chip forming process until the chip has formed, stopping the simulations when
remeshing becomes a necessity (van Kesteren, 1995; Liu et al., 2002; Yu, 2005). An
option to allow the simulations to continue beyond this point would be to use an
element deletion routine (Cho et al., 2010; Pittino et al., 2015), i.e. when tensile
failure occurs in an element, it will be immediately eliminated. When modeling the
cutting process involved in a single cutter such a simplification is prone to modeling
errors. The generated debris is completely neglected while its presence can result in
significant stresses. Another disadvantage is that mass is not conserved.
A different solution could be to use remeshing techniques based on the combi-
nation of discrete and finite elements, (Mahabadi et al., 2004; Oñate and Rojek,
2004; Zárate and Oñate, 2015). In such an approach, finite elements are used to
model the behavior of intact rock. The nodes of the elements are treated as discrete
elements, which become active when a crack occurs through the element boundary.
This way it is possible to allow large deformations of the mesh, by making adapting
the discretisation from a continuum base to a discontinuum base.
4.2. Hydromechanical coupling 59

4.1.2. Discontinuum based modeling

The discrete element method (DEM) is often used to simulate the rock cutting
process of a single tooth bit, e.g. (Huang, 1999; Lei and Kaitkay, 2003; Rojek et al.,
2011). The unique feature of the DEM is that it can be used to model solids as a
collection of particles consisting of arbitrary shapes (most often simple shapes like
discs, spheres and polygons are used because of their mathematical simplicity). The
particles can move independently from one another and they interact through the
contacts or interfaces between them. In contrast to the continuum based methods,
the DEM is not limited to specific failure modes, it can deal with cataclastic damage
and brittle failure mechanisms simultaneously. These phenomena occur as a generic
consequence of the effects that occur on the micro-scale, being tensile and or shear
failure, resulting in particle separation and sliding.
The use of the DEM in geomechanics originates from the study of granular mate-
rials (Cundall and Strack, 1979). Especially in the last decade, due to the increased 4
computational power that is available, the interest in the DEM has increased. Nowa-
days it finds its use in a wide range of applications, e.g. granular flow (Guo and
Curtis, 2015), solid-liquid mixtures (Robinson et al., 2014), pedestrian flows (Lohner
et al., 2016) and rock failure (Potyondy, 2015). Various researchers have used the
DEM to model tool-rock interactions, like indention and cutting. Interests in using
DEM for rock cutting simulations started with the PhD thesis of Huang (1999), fo-
cusing on 2D simulations of rock indention and rock cutting tests. Especially within
the last few years, the DEM has gained popularity to do rock cutting simulations,
using it for a wide range of applications. A brief overview of the work that has been
done is presented in table 4.1.
To summarize, the distinction between the continuum and discontinuum-based
modeling approaches results in a shift in where the complexity of the modeling
approaches will reside. For continuum based models the equations that need to be
solved increase in complexity and have a rather simple geometry, while the opposite
is true for the DEM. The physical models that are solved are simple on a microscopic
level, but the complexity of the model comes from the complexity of the geometry.
Furthermore, the cutting process concerns the transition from intact rock towards a
discontinuous assembly of grains and chips. That is why a discontinuous approach
is more favorable to adequately model the rock cutting process.

4.2. Hydromechanical coupling

The DEM has been useful to model tool-rock interactions, however, without consid-
ering fluid-pressure effects. Several approaches have been developed to incorporate,
to some extent, fluid pressure effects in rock. The first group are methods that
are based on a discontinuum approach by modeling the flow of the pore fluid along
the element contacts/bonds (Shimizu et al., 2011; Zhang and Wong, 2013; Wang
et al., 2014). The aperture between two neighboring particles is used to determine
the width of the pore channel, allowing to adapt to a change in permeability due
to the occurrence of opening (micro-)cracks, see figure 4.1 for an overview. These
methods are often used to model the hydraulic fracturing process, with typically
mode-I (tensile) failures as the dominant failure phenomena. In case of large defor-
 60 4. Modeling Approach

Table 4.1: Research on rock cutting simulations using DEM.

Publication Application Special interests

Huang (1999) PDC bit, wedge 2D, indention and cut-
ting tests
Lei and Kaitkay (2003) PDC bit, drilling 2D, confining stress on
top layer
Ledgerwood III (2007) PDC bit, drilling 2D, confining stress on
top layer
Alvarez Grima et al. pick point, DSM 2D, confining stress on
(2011) top layer
Moon and Oh (2011) disc cutter, TBM 2D, tool spacing
4 Rojek et al. (2011) pick point, roadheader 2D/3D, comparison dif-
ferences between 2D and
3D
Su and Akcin (2011) PDC bit 3D
Huang et al. (2013) PDC bit 2D, improved on work of
Huang (1999)
Huang and Detournay wedge indention 2D, improved on work of
(2013) Huang (1999)
Mendoza Rizo (2013) PDC bit, drilling 2D, confining stress on
top layer, crushable par-
ticles
Carrapatoso et al. (2014) PDC bit, drilling 3D, effect of chamfer,
sharp vs blunt tool
Rojek (2014) pick point, dredging 2D, heat distribution and
wear rate analysis
van Wyk et al. (2014) PDC bit, drilling 3D, tribology effects
Tropin et al. (2014) PDC bit, drilling 3D, confining stress on
top layer
He and Xu (2015) PDC bit, drilling 2D, comparison single
particles and clustered
particles, brittle-ductile
transition based on cut-
ting depth
Liu et al. (2015) disc cutter, TBM 2D, effect of tool spacing
and confining stress
Li et al. (2016) disc cutter, TBM 2D, effect of cutting
velocity and confining
stress

mations, occurrence of new contacts or significant distortion of the contact network

(i.e. "mesh") of the fluid, it will be necessary to remesh the discretisation of the fluid
phase. This will be problematic for the cases concerning the transition from intact
4.2. Hydromechanical coupling 61

rock to moving fragments of rock, as remeshing will be too frequently necessary.

True pore volume

ri F

Domain
fc
Flow channel
w
Lp

Particle
Flow
fc
channel
rj 4
Domain volume
(a) Channel-domain model (b) Fluid flow in a channel

Figure 4.1: Pore channel model, taken from Shimizu et al. (2011).

The second group of hydro-mechanical coupling methods are those that apply an
adaptive confining pressure boundary condition (Lei and Kaitkay, 2003; Ledgerwood
III, 2007; Mendoza Rizo, 2013; Tropin et al., 2014). It is considered adaptive in the
sense that it allows the boundaries to deform, in contrast to the servo-controlled
walls that are sometimes used in bi/tri-axial tests (Potyondy and Cundall, 2004).
Various algorithms exist to determine which particles are part of the free surface and
which are not. Lei and Kaitkay (2003); Ledgerwood III (2007); Alvarez Grima et al.
(2011) all identify the boundary particles based on the chain of contacting particles
along the boundary, see figure 4.2. Starting with particles A and B which are
located based on a geometric criteria. Then the algorithm works counterclockwise
from particle B to locate all the other boundary particles, moving on to the particle
that makes the smallest angle with the vector of BA is identified as a boundary
particle, etc.

θ1

θ2 B
C A F

Figure 4.2: Boundary detection method as suggested by Lei and Kaitkay (2003).
 62 4. Modeling Approach

In the approach of Mendoza Rizo (2013) the boundary is identified by an al-

gorithm that loops over through the particles within a window and identifies the
nearest particle with respect to a frame of reference, see figure 4.3. A somewhat
similar approach is used by Tropin et al. (2014) in what they call the shining light
method. Each particle that can be directly ’seen’ from a plane is identified as a
boundary particle. Because such a method is only applicable to a static boundary,
they also make use of an α-shape algorithm based on Delauny triangulation, which
is based on the work of Edelsbrunner and Mücke (1994).

DATUM
δi+1

δi
α,δ=f(R)
α
4
2R

Window
i
Window
i+1

Figure 4.3: Boundary detection method as suggested by Mendoza Rizo (2013).

In all of the mentioned boundary detection methods, a force is applied to each

boundary particle based on the confining stress and its weighted contribution to the
boundary. Although these approaches mimic the effect of a confining stress, they
do not consider the pores of the rock to be filled with a fluid.

DEM-SP coupling
In this research, the DEM is extended with fluid pressure effects. These fluid effects
are based on a Smoothed Particle (SP) approach, which is a meshless Lagrangian
particle based method used to solve continuum based problems. SP finds its origin
in astrophysical problems Lucy (1977), nowadays it is used in many fields, e.g.
hydrodynamics (SPH) Monaghan (2012), applied mechanics (SPAM) Hoover et al.
(1994). Various researchers have used the combination of DEM and SP techniques
in the field of rock mechanics. These can be distinguished into two groups.
The first group of methodologies are those in which DEM and SP are used
to model the two phases fully separated, particles represent either solid or fluid
(Lagrangian-Lagrangian model description). Such methods are used for problems
like hydraulic fracturing Komoróczi et al. (2013), rock blasting Fakhimi and Lanari
(2014), boudinage Komoróczi et al. (2013). Besides, such a two phase description
for fluid rock interactions, it is often used in modeling of multi-phase flows and/or
4.3. Modeling Approach 63

fluid structure interactions as well, for example Potapov et al. (2001); Cleary et al.
(2006); Sun et al. (2013).
The other group of methods are based on methods that treat the DEM and SP
particles in a co-located fashion, meaning that the properties of both fluid and solid
phases are based on the same particles and these properties are transported with the
same particles as well. Thus far, the use of these kind of approaches is rather limited
due to a strong but necessary assumption, transport of the two phases should be
almost similar. However, it can be used to relate the discontinuous properties of
DEM to a continuum field, e.g. heat, stress. A strongly simplified approach similar
to SP that describes the stress state of DEM particles is used in SDEM (stress-
based DEM) (Egholm, 2007; Egholm et al., 2007). In SDEM, the calculated stress
is used as a failure criterion for the DEM. SDEM is too simplified in the sense of
number of neighboring particles, only the direct neighbors are considered, and the
interpolation kernels that are used are insufficient to properly model a continuum. 4
As a result, their application is limited to an interpolation scheme for the DEM.

4.3. Modeling Approach

As presented in section 4.1.2, the DEM is a powerful method that is well capable
and accepted to model the cutting of dry rock and therefore the DEM can be used
as a basis for modeling the cutting of saturated rock. The large deformations of
the rock require to model the fluid effects on a continuum level to avoid expensive
remeshing of the fluid phase. The Smoothed Particle method is a particle based
method that is used to solve problems based on a continuum field. That way, the
SP can be used to interpret the discontinuous properties of the DEM to a continuum
level.
The research objective of this dissertation, modeling the cutting of rock for
shallow and deep water applications, is a research area with a rather small academic
and industrial community. This, together with the complexity of the cutting process,
makes it that a pragmatic approach is desirable. In the industrial practice very
little is known about the rock that has to be cut, which is often limited to the
UCS, the type of rock (e.g. limestone, sandstone) and water depth. Several other
mechanical properties of the rock are often estimated based on empiricism, some of
these relations are presented in A. In order to stay close to the industrial application,
the model is set up according to the following design choices:
1. Emphasis is on the influence of the fluid pressure on the cutting process for a
generic type of rock.
2. Limit the number of required input parameters.
3. Aim for those effects that can be validated based on existing experimental
data.
4. The characteristic domain size should resemble the cutting process of a single
tool. As a result, the computational domain is required to be in the order of
0.1-2 m. Acceptable computation time is approximately one day per cutting
simulation.
 64 4. Modeling Approach

The methodology is developed within the author’s own implementation, written

in C++ and CUDA (C++ environment of NVIDIA to support GPU programming).
The implementation of the DEM and SP model are described in next section. In this
study, the algorithms and simulations are set up in 2D to limit the computational
expense. Besides, first the suitablity of the technique needs to be proven before
further investments will be made into extension to 3D. Luckily, extension to 3D is
rather straightforward.

4.4. Solid Modeling - Discrete Element Method

The translational and rotational motion of a particle is governed by the standard
equations for rigid body mechanics

4 mi üi = Fi (4.1)
Ii ω̇i = Ti (4.2)

where F and T are the sums of all forces and moment, m the particle mass, I the
particle moment of inertia, u the particle centroid displacement in a fixed coordinate
frame, ω the angular velocity, which are all applied to particle i . The vectors Fi
and Ti are calculated through
c
ni
+ F∇p
X
Fi = Fext
i + Fcont
ij + Fdamp
i i (4.3)
j=1
c
ni
X
Ti = Text
i + scij × Fcont
ij + Tdamp
i (4.4)
j=1

where Fext and Text are external loads, Fcont ij the contact interactions with neigh-
boring particles j = 1, ..., nci , with nci the number of contacting neighboring elements,
numerical damping loads Fdamp and Tdamp , scij is the vector that connects the cen-
ter of mass of particle i with the contact point with particle j and F∇p i the coupling
force resulting from the fluid pressure gradient over the particle. The numerical
damping is defined by

u̇i
Fdamp
i = −α|Fext
i + Fcont
i | (4.5)
|u̇i |
ω̇i
Tdamp
i = −α|Text
i + Tcont
i | (4.6)
|ω̇i |

where α is the numerical damping coefficient. The numerical damping is similar to

what is first used by Potyondy and Cundall (2004). Their definition of numerical
damping is included in the commercial DEM-code PFC (Particle Flow Code) de-
veloped by Itasca and it is included in DEMPack (Rojek et al., 2011), developed by
CIMNE.
4.4. Solid Modeling - Discrete Element Method 65

The equations of motion (4.1) and (4.2) are integrated in time by using the
explicit velocity Verlet scheme.

1
uin+1 = uin + u̇in ∆t + üin ∆t2 (4.7)
2
n+1/2 1
u̇i = u̇in + üin ∆t (4.8)
2
1 X n+1
üin+1 = Fi (4.9)
mi
n+1/2 1
u̇in+1 = u̇i + üin+1 ∆t (4.10)
2
The advantage of the velocity Verlet scheme is that it only requires to sum all particle
interactions once per time step, it is fully reversible and has no artificial damping in 4
itself. Because the scheme is an explicit integration scheme, the timestep is limited
by
mi
r
∆tcrit = min P (4.11)
nc k n

4.4.1. Constitutive model

Models of contact interactions Fcont
ij are based on the decomposition of Fcont
ij into
normal and tangential components, respectively

Fcont = Fncont n + Fscont t (4.12)

where n is the normal vector with respect to the particle surface at the contact
point, t is the unit tangent vector aligned with the tangent force, Fncont and Fscont
are the scalar quantities that are obtained by projection of Fcont on the base vectors.
See figure 4.4 for an overview.

kn Tn
A B

ks

μ Ts

Figure 4.4: Contact interaction model

The interaction force between a pair of particles can either consist of a bond
type or a collision type of interaction. At the beginning of each simulation, the
particles that are in contact are bonded together. These bonds are modeled with
an elastic perfectly brittle model. The interaction between two particles is modeled
as a collision when either their bond is broken or never a bond has existed between
 66 4. Modeling Approach

these contacting particles. The interaction models are in the form of

Fncont = kn un (4.13)
∆Fscont = ks ∆us (4.14)

where kn is stiffness in normal direction, ks stiffness in shear (tangential) direc-

tion, un normal relative displacement which is positive in compressive direction and
negative in tensile direction and us tangential relative displacement. Here compres-
sive forces are considered to be negative (overlapping particles), and tensile forces
positive (distance between particles). Note that the shear force is calculated in an
incremental form, because it is less prone to drift in this form. Bonds fail instanta-
neously when either the shear strength (tangential direction) or the tensile strength
(normal direction) is exceeded. Bonds remain intact when
4
Fn ≥ Tn (4.15)
|Fs | ≤ Ts (4.16)

with tensile strength of the bond Tn and shear strength of the bond Ts . Although a
compressive interaction force between the particles does not result in bond breakage,
on a macroscopic scale compression is properly represented by allowing the shear
and tensile breakage of the bonds on the micro-scale.
In case of a collisional interaction between particles, the tangential force is eval-
uated assuming the Coulomb model of friction, that can either inhibit stick and slip
conditions: 
µCoulomb Fn
|Fs | = min (4.17)
ks |us |
with friction coefficient µCoulomb . A graphical representation of the bond model and
its failure criteria is shown in figure 4.5.

Fn Fs
Fs Ts
Tn
Ts
μ|Fn|

kn ks us ks us
un
μ|Fn|
Ts
un<0 Ts
compression tension un>0
(b) Bond in shear direc-
(a) Bond in normal direc- tion, with contact. (c) Bond in shear direction,
tion. without contact.

Figure 4.5: Bond models

In this research, it is opted to use some of the simples DEM interaction models for
both bonds and collisions. The main reasons to do this is that often the introduction
of more advanced models requires more input parameters, while the experiments on
which the validation of this research is based have rather limited detail on the
mechanical properties of the tested rock. Addition of more input-parameters does
4.4. Solid Modeling - Discrete Element Method 67

not help to improve the understanding of the rock cutting process, although it
might improve the results. In a later stage of the research, one might think about
on using more advanced bond and collision models to improve the results. Some
of the most interesting improvements are the parallel bond model (Potyondy and
Cundall, 2004), or one can add viscous or history terms to the bond model as is
done in powder sintering models, e.g. (Luding, 2011), (Rojek et al., 2017).

Particle Geometry
Dilative, compactive and frictional (post-failure) effects are likely to be misinter-
preted when using disc-shaped particles. To improve on these effects, besides tun-
ing the particle size distribution of the generated mesh, complex particle shapes are
often used. Although beyond the scope of this research, due to the introduction of
new variables, several options to improve on these effects are suggested in literature: 4
• Rolling resistance: Besides the stick/slip behavior that is modeled for the
tangential interaction force, an additional resistance that restricts the rolling
of the particle can be included. This rolling resistance is a commonly accepted
method to adjust for the non-sphericity of the particles (Luding, 2008).

• Particle clusters: Particles are combined to create arbitrarily shaped particle

clusters. The micro-models of the inter- and intra-cluster interactions differ.
Often the intra-cluster interactions have a higher stiffness and strength com-
pared to the inter-cluster interactions. For example, see the work of Thomas
and Bray (1999) and Potyondy and Cundall (2004).

• Particle clumps: Similar to particle clusters, except for now the particles are
allowed to overlap. The overlapping particles act as a rigid body clump, see
(Cho et al., 2007).

• Grain interlocking: When the initial geometry is set up, particles are allowed
to have a small overlap. When installing the bonds, these bonds are installed
with an equivalent distance in such a way that the initial state of the rock can
be stress free. However, this method has its limitations, the initial overlap
cannot be larger than the break-up length for tensile failure, i.e. when a
particle fails in tensile direction, the particles must not have contact. See
(Scholtès and Donzé, 2013).

However, the disadvantage of these improvements is that each of them introduces one
or multiple parameters extra. In chapter 5 an analysis on the parameters and the
parameter sensitivity will be presented. The improvements on particle geometry
as mentioned before are not implemented in the model of this dissertation. The
main reason to not do this is that the suggested improvements only give rise to
slightly different types of rocks, but are expected to have little influence on the
hydro-mechanical coupling of the pore fluid. However, these improvements might
be included in a later phase of the research if it is needed for a specific type of rock.
 68 4. Modeling Approach

4.5. Fluid Modeling - Smoothed Particle

Due to the low permeability of rock-like materials, the contribution of hydro me-
chanical effects of a fluid to the mechanical properties of a rock is dominated by the
generated pressure gradient. The effect of fluid velocity on the mechanical behavior
of rock is in that range negligible. For that reason, the fluid phase can be simplified
in such a way that velocity of the fluid is not fully solved. This can be achieved
through a pore pressure diffusion equation Coussy (2004). This equation is based
on the combination of mass conservation, Darcy flow and a constitutive equation
for the compressibility of the pore fluid, given respectively by

Dζ
+∇·q =γ (4.18)
Dt
κ
4 q = − (∇p − f )
µ
(4.19)

p = M (ζ − αV ) (4.20)

with pore fluid content ζ, fluid flux q, fluid source/sink γ, intrinsic permeability
κ, pore pressure p, body force f , Biot modulus M , effective stress coefficient α and
volumetric strain V . By combining the equations (4.18), (4.19) and (4.20), the pore
pressure diffusion equation is obtained.
   
Dp κ DV κ
− M∇ · ∇p = −αM + M γ − ∇f (4.21)
Dt µ Dt µ

The right hand terms respectively describe the volumetric strain (dilation/compaction
of pore volume), fluid source/sink and a gradient in an external force. The last two
terms are of negligible effect on the rock cutting process considered in this thesis.
Here purely mechanical effects and deformations are considered, effects of jetting
or pumping are beyond the scope of this research. Furthermore, it is assumed that
thermal effects are negligible with respect to the fluid pressure effects and therefore
the models is assumed to be iso-thermal as well.
Due to the discontinuum properties of DEM, it is not possible to directly solve a
continuum equation, like equation (4.21). The discontinuous data of DEM needs to
be interpolated towards continuum properties, which can be achieved by applying
an interpolation technique based on the smoothed particle (SP) approach, which
weighs each close neighbouring particle contribution to an interpolation point based
on a kernel function. In this research, the interpolation points of SP are collocated
with the particle centers of the DEM particles. These interpolation points will also
move along with the movement of the DEM particles. The advantage of using SP in
combination with DEM is that the structure of both algorithms is identical, except
the particle interactions differ.
A proper kernel function is consistent, has small support and is sufficiently
smooth. Small support means a small domain of influence, which is desirable to
limit the computational expense of the simulation. The integral of the kernel func-
tion within its domain of influence should be equal to unity, and the integral of the
kernel gradient should be equal to zero. A smoothing function with smoother value
4.5. Fluid Modeling - Smoothed Particle 69

of the function and derivatives usually yields better results. Using a sufficiently
continuous kernel function makes the algorithm less sensitive to particle disorder
and the errors in approximating the integral interpolants are small provided that
the particle disorder is not too extreme. Here the Wendland C2 kernel function
(Wendland, 1995) is used, because of its consistency, small support and smoothness
properties. The kernel is given by
4
7
(1 − R) (4R + 1) if R ≤ 1
W = π (4.22)
0 if R > 1

where
|ri − rj |
R= (4.23)
h
with smoothing length h and dimensionless relative distance R. 4
In the DEM-SP model, the discretized particles are taken from a particle size
distribution and randomly stacked together. The unstructured positions and ran-
dom size and mass of the particles can easily cause stability issues. Close to the
boundaries of the domain, the kernel support may not be entirely contained within
the problem domain. To preserve linear consistency near the boundaries, the Cor-
rective Smoothed Particle Method (CSPM) is used (Chen et al., 1999).

In CSPM, the kernel interpolation of a field quantity A is calculated by

P
j Aj mj W (ri − rj , h)
A (ui ) = P (4.24)
j mj W (ri − rj , h)

with particle mass m, kernel function W , smoothing length h, index i for the particle
under consideration and index j for the neighboring particles (including particle i ).
In the CSPM, the equations for the first derivatives have to be calculated in a coupled
fashion (Chen et al., 1999). This can be done by solving the set of equations

Xαβi Aαi = Yβi (4.25)

where
N
X
rjα − riα ∇β mj W (ri − rj , h)


Xβαi = (4.26)
j=1

N
X
Aα α


Yβi = j − Ai ∇β mj W (ri − rj , h) (4.27)
j=1

gives the first derivative approximations. In above equations, α and β are the indices
of the spatial derivatives.
Unfortunately, higher order derivatives in SP, necessary for the diffusive term in
(4.21), cannot be calculated directly by taking the second derivative of the kernel
function. Fatehi et al. (2009) give several options to calculate the second derivative,
i.e. double summation, second order kernel derivation, difference scheme. The dou-
ble summation scheme requires to first calculate the first order derivative of each
 70 4. Modeling Approach

particle, resulting in a large extra computational effort (looping twice over each par-
ticle). Although the second-order kernel derivation and the difference schemes can
compute the second derivative in the same loop as the first derivative, implementa-
tion of Neumann boundaries are not straightforward. Cleary and Monaghan (1999)
derived a model for heat conduction based on the difference scheme. In this scheme
the diffusive term is calculated directly, it can deal with discontinuities in material
parameters and it automatically adopts Neumann boundary conditions when not
stated otherwise. Implementation of Dirichlet conditions is straightforward in this
method. That is why in this research the methodology of Cleary and Monaghan
(1999) is implemented, which is done in the form of

  X
κ mj (κi + κj ) nij · ∇W (ri − rj , h)
∇· ∇p = (pi − pj ) (4.28)
4 µ j
ρi (µi + µj ) |ri − rj |

where nij is the normal unit vector of the neighboring particle centers. Note that
here κ and µ are considered as particle properties, meaning that these parameters
can differ throughout the rock sample. In the case of DEM-SP, these parameters
are assumed to be constant.
Two way coupling is applied at every timestep. DEM is advanced half a timestep
as in equation 4.9, based on the intermediate velocities of the DEM-particles, the
volumetric strain rate is calculated for the fluid. This is then used to advance the
pore pressure diffusion equation one timestep. The resultant local pressure gradient
of the fluid is calculated based on the new pore pressure distribution. The pressure
gradient on a particle is then added as an interaction force to the sum of forces
acting on the particles in equation 4.8 through

Fi = −∇pπri2 (4.29)

with particle radius ri . See figure 4.6 for an overview of the algorithm. Care must
be taken when using this pore pressure coupling. When the coupling force is much
larger than the strength of the particle bonds, i.e.

∇pπri2  Tn (4.30)

the rock will disintegrate and the resulting behavior will be unrealistic. The ef-
fect becomes apparent when increasing either the particle size, whose effect scales
quadratically, or by increasing the hydrostatic pressure, which leads to a higher
maximum achievable pore pressure gradient, assuming that the cavitating limit can
be met. If it is not met, larger hydrostatic pressure does not necessarily lead to
different ∇p. Note that in the case of rock cutting for deep sea applications, the
maximum achievable pore pressure gradient can be in the order of GPa/m.
Following the same analogy as in Cleary and Monaghan (1999), the critical time
step for a pure diffusion problem of the fluid phase is given by

βµf h2
∆tf = (4.31)
Mκ
4.5. Fluid Modeling - Smoothed Particle 71

Figure 4.6: Flowchart of the calculations done per timestep in DEM-SP.

where β is a constant. According to Cleary and Monaghan (1999), β ≤ 0.15 is

required for a stable time integration for a structured set of particles. Although the
set of particles is unstructured and a non-uniform particle distribution is used, in
our case the time integration of DEM-SP is mostly limited by the stability of the
solid phase. In the case of high permeability (κ ≥ 10−11 m2 when assuming water as
pore fluid and gravelstone like rock), the critical timestep of the fluid phase might
become limiting. However, in such a case the peaks in the fluid pressure will diffuse
sufficiently fast that its influence on the failure mechanics of the rock is negligible.
Due to the full coupling between solid and fluid phase, both phases have to march
in time with the same step size, therefore a forward Euler scheme is sufficient for
the fluid phase.
At the beginning of each simulation, the intrinsic permeability of the simulated
rock is set as a global homogeneous and isotropic material property. Changes in
permeability due to deformation of the rock are not considered within this research.
The main reason to do so is that with the large range in strain rates that are
considered within this thesis (up to 6 orders of magnitude) in this research, the
local variations due to dilation and compaction are considered to be insignificant.
When cracks occur and they are opened sufficiently, they are considered as a new
(internal) boundary and hydrostatic pressure is then instantly applied, assuming
that in such a case the permeability of the crack is several orders of magnitude
larger.
It is possible that the fluid pressure drops below the vapor pressure during sim-
ulations, resulting in cavitation. When cavitation occurs, the compressibility of
the fluid increases with several orders of magnitude. In the case of failure of sat-
urated rock, the pore pressures can easily obtain a pressure gradient in the order
 72 4. Modeling Approach

of O (GPa/m). Therefore it is chosen to model cavitation through a simplified ap-

proach, by the assumption that the phase transition of liquid to vapor or vice versa,
are of negligible effect, except for the local pressure values (in case of cavitation of
water the compressibility of the fluid increases with approximately the same factor
as the viscosity decreases, which merely affects the source term in equation (4.21)
and the combination of these two parameters is of negligible effect on the diffusiv-
ity of the pore pressure). When the pressure drops below the vapor pressure, the
pressure value is fixed at the minimum pressure. From thereon it is only possible to
increase the pressure by having fluid flow towards the cavitating zone, so if
 
∂pi ∂pi
pi < pmin then pi = pmin and = max ,0 (4.32)
∂t ∂t

4 and both the fluid bulk modulus and the dynamic fluid viscosity are adjusted in
those areas. In the case
of water as a pore fluid this results in a bulk modulus that
is approximately

O 103 smaller and a dynamic viscosity that is also approximately
3
O 10 smaller. As a result, the fluid pressure in the cavitating zone is close to
uniform, resulting in an almost negligible pressure gradient force acting on the solid
and dilative effects of the solid are no longer counteracted upon by the pore pressure
gradient. In the end, cavitation has no effect on the numerical stability of the time
integration scheme (∆tcrit = constant). Simulations of rock saturated with air show
no significant differences compared to simulations without pore fluid interaction.
This is due to the high compressibility of air compared to liquids, like water, oil,
resulting in a factor of O(103 − 104 ) in difference.

4.5.1. Boundary conditions and detection

Often in smoothed particle simulations, the boundary conditions are applied by
using either repulsive forces (Monaghan, 1994) or ghost particles (Colagrossi and
Landrini, 2003). However, in this case the SP are coupled in a co-located fashion
with the DEM, complicating the use of such methods. In these methods, the motion
of the particles are described by SP as well, while in DEM-SP the motion of the par-
ticles is dictated by the DEM. Therefore it is desired to apply boundary conditions
on the particles that are part of the boundary itself.
By using the discretisation of Cleary and Monaghan (1999), as in equation (4.28),
∂p
undrained boundary conditions (no fluid flow, i.e. ∂n |Γ = 0) are automatically
applied. In order to apply a drained boundary condition (Dirichlet, i.e. hydrostatic
pressure), the boundary particles on which it needs to be applied have to be detected,
because interior particles can become boundary particles and vice versa. Therefore
a method that can deal with a disordered particle structure and a non-uniform
particle size/mass distribution is used. Besides that, the method needs to be able
to detect the opening of a crack (in case of mode I failure) in the interior domain.
The divergence of position, as suggested by Muhammad et al. (2013), meets these
requirements. Divergence of position is given by

∂rx ∂ry
∇·r= + =2 (4.33)
∂x ∂y
4.6. Initial geometry generation 73

The above equation is applied to the standard Smoothed Particle approach of Lucy
(1977), in order to prevent the CSPM in equation (4.25) from adjusting the di-
vergence of position to the particle deficiency that occurs at a boundary. In the
interior domain of the rock, equation (4.33) is valid. However, at the boundaries
this value differs due to particle deficiency, see figure 4.7. The divergence of position
is more strict in comparison to other SP-based criteria, such as the density, sum of
the kernel, etc. An advantage of the divergence of position is that it is insusceptible
to what the shape or orientation of the boundary of the rock specimen is. For some
of the algorithms used to determine the boundary particles in DEM this is still an
issue, since these approaches determine the boundary particles with respect to a
fixed or a priori known observer plane (Mendoza Rizo, 2013; Tropin et al., 2014).
In this work, particles with ∇ · r ≤ 1.5 are considered to be part of the bound-
ary. Dirichlet conditions can be applied by fixing the pressure value of the boundary
particles to the desired hydrostatic pressure. In order to avoid instabilities origi-
4
nating from particle deficiency, particles that have less than five particles within
their interpolation domain are considered to be fully surrounded by the hydrostatic
pressure. As a result, the coupling force of these particles is zero.

∇ · ~r < 2 ∇ · ~r = 2

Figure 4.7: Schematic overview of particle deficiency. The incomplete kernel support of a surface
particle gives a value for ∇ · r ≤ 2 (2D).

4.6. Initial geometry generation

The discrete element method is best suited for materials which are inherently dis-
ordered on the microscopic scale, meaning that the material is composed of grains
with various sizes and/or shapes with interactions between the grains. In the ideal
case, the model gives a high quality representation of the micro-structure of the spe-
cific material. However, due to computational expenses the particles are idealized
as discs with varying radii. This way, the problem of discretisation of a geometry
reduces to the generation of a random homogeneous, isotropic packing of particles
 74 4. Modeling Approach

with a prescribed size distribution. In general, two classes of generation methods

can be distinguished, i.e. the dynamic and the constructive methods.
Dynamic methods start with an assembly of randomly distributed particles which
are moved together by either sedimentation or applying compression through bound-
ary elements until the desired porosity is achieved. Particles in the assembly that
have insufficient number of contacts can be increased in size until the desired num-
ber of contacts are present. Overviews of such methods are presented in Potyondy
and Cundall (2004) and in Feng et al. (2003). In these methods, the motion of each
particle has to be simulated with the DEM code, which can lead to undesirable
computational expense, especially for the larger particle assemblies. Furthermore
the initial stress in the assembly needs to be sufficiently low to not let the initial
stress affect the simulated results.
Constructive methods use a different basis. In most of such approaches particles
4 are taken randomly from a particle size distribution and than placed alongside each
other. This way, particles can be placed exactly touching its neighboring particles,
resulting a zero initial stress condition. Various strategies can be applied when
constructing a particle assembly.

1. Closed front methods start from the center of the geometry and add particles
moving outward, e.g. (Feng et al., 2003; Zsaki, 2009).

2. Inwards packing methods start from the boundary of the geometry that needs
to be generated and start moving inward, e.g. (Bagi, 2005; Benabbou et al.,
2009).

In this thesis, the methodology of Bagi (2005) is used because it allows to ac-
curately determine the boundary shape of the specimen, gives full control over the
particle size distribution, has low initial stress has a sufficiently dense contact net-
work and is fast. The porosity and the coordination number are parameters that are
used as a measure of the density of the contact network. The coordination number
is defined as the average number of bonds per particle. For proper rock modeling in
2D, the coordination number should be larger than 4 and the porosity lower than
0.19 (Bagi, 2005). Furthermore, the constructive geometry generation methods are
fast (computation within few seconds for an assembly of > 100k particles on a single
CPU).

4.7. Smoothed particle averaging of discrete el-

ements
Cleary and Monaghan (1999) have shown that their discretisation of the diffusive
term is suitable for structured (regular Carthesian grid) and unstructured parti-
cle positions. In their paper they consider the particles to have the same mass
and/or volume. A structured particle distribution is not realistic in the case of SPH
(and DEM as well), so in order to obtain a realistic disorder of the particle struc-
ture, Cleary and Monaghan (1999) performed simulations of fluid flow with SPH.
However, in DEM-SP this is not completely valid. The particles have a particle size
4.7. Smoothed particle averaging of discrete elements 75

Figure 4.8: Particle position distribution for particles with unequal mass.

distribution and the positions of the particles are dictated by the underlying physics 4
of the DEM.

In order to verify the applicability of equation (4.28) to solve diffusive problems

related to the DEM, results of DEM-SP are compared with the exact solution of
the diffusion equation
∂u ∂2u
=D 2 (4.34)
∂t ∂x
u(0, t) = u(1, t) = 0 (4.35)
u(x, 0) = sin (πx) (4.36)
in which D is the diffusion coefficient. The exact solution of the boundary value
problem stated above is given by
2
u(x, t) = sin (πx) e−π Dt
(4.37)

The settings that have been used to determine the DEM-SP solution of the diffusion
equation (4.34) are a height and width of 0.5 and 1.0, respectively, with a uniform
particle size distribution with radii in the range of 0.005-0.01, resulting in a total
of 2142 particles. The distribution of the particle sizes and positions used for this
calculation is shown in figure 4.8. The diffusion coefficient D has been set to 1.0 and
for the time integration 1000 timesteps of 5 · 10−6 are used. The numerical diffusion
coefficient of equation (4.28) differs with varying smoothing length, as is shown in
figure 4.9. The solution of (4.28) shows a spread closely comparable to the exact
solution. This spread is caused by both the disordered particle positions and the
non-uniform particle masses, see figure 4.10. It is clear that with a smoothing length
that is too small, the diffusion coefficient is significantly underestimated, which is
due to the lack of interacting particles. For larger smoothing lengths (h > 6r̄), the
diffusion coefficient is slightly overestimated, but is still within acceptable accuracy
(error < 5%) for application in rock mechanics.
The choice of the smoothing length requires a compromise between a minimum
amount of neighboring particles to assure stability, while on the other hand it is
desired to use as few neighboring particles as possible to limit the computational
 76 4. Modeling Approach

expense and to avoid the loss of detail due to smoothing. Therefore it is chosen to
use a smoothing length in the range of 5r̄ − 6r̄.

Relative diffusion of SP on DEM particle structure

1
D∗ [s−1 ]

0.8

4
0.6

0.4
4 6 8 10 12 14
h
r̄ [−]

Figure 4.9: Diffusion coefficient with respect to the relative smoothing length. D∗ is the relative
diffusion of the SP solution w.r.t. the exact diffusion coefficient. The relative diffusion coefficient
is considered for the particle that is nearest to the line x = 0.5.

4.8. Conclusions
Regarding the rock cutting process, discontinuum methods have a preference over
continuum based methods. The two main arguments are: 1)continuum based meth-
ods will require remeshing when significant distortion is applied to the rock, 2) a
complex set of equations that will be able to cover the transition from intact rock
through fragmentation towards granular flow.
The approach used within this thesis is a combination of the discrete element
method to model the rock and a smoothed particle based approach to model the
hydro-mechanical coupling, currently implemented in two dimensions. Due to the
low permeability, fluid flow is modeled as a pore pressure diffusion process. This
way, fluid velocity does not need to be solved, but it is modeled through the diffusive
term. The combination of physics allows to use the DEM and SP in a co-located
fashion. It is the first attempt to incorporate full hydro-mechanical coupling in
DEM on this scale.
Solid boundaries of the rock are identified by calculation of the divergence of
position with SP. The use of SP to model diffusive processes with particle size and
position distributions based on DEM problems show good resemblance with the
exact solution. A smoothing length of 5r̄ − 6r̄ gives the best compromise between
4.8. Conclusions 77

Global solution
1
Simulations
0.8 Exact
0.6
u [−]

0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
x [−]
Detail of solution 4
0.84 Simulations
Exact
0.82
u [−]

0.8
0.78
0.76
0.4 0.42 0.44 0.46 0.48 0.5 0.52 0.54 0.56 0.58 0.6
x [−]

Figure 4.10: Solution of (4.34), results obtained by solving the diffusive term with equation (4.28)
and the exact solution according to equation (4.37).

computational effort and accuracy to solve the pore pressure diffusion equation.
 5
Validation - Material tests
"There are known knowns. These are things we know that we know. There are
known unknowns. That is to say, there are things that we know we don’t know. But
there are also unknown unknowns. There are things we don’t know we don’t know."
Donald Rumsfeld

The combination of the DEM and the hydro-mechanical coupling effect, which is
modeled with a smoothed particle approach, gives rise to a large number of inde-
pendent parameters. First a parameter sensitivity analysis is presented. Here the
most relevant micro-parameters of the DEM are compared between dry and satu-
rated conditions. Additionally, it is shown that the methodology corresponds with
Terzaghi’s effective stress theory. Furthermore, simulations of shale samples over a
range of several orders of magnitude in strain rate are compared with experiments
in literature.

Part of this chapter has been published in Helmons et al. (2016b).

79
 80 5. Validation - Material tests

5.1. Parameter sensitivity

Huang (1999) developed a methodology to calibrate the macro-parameters of a rock
specimen based on the combination of dimensional analysis and numerical simula-
tions of the standard laboratory tests, e.g. unconfined compression test, Brazilian
test. Other researchers have adopted this methodology as well, (Yang et al., 2006;
Fakhimi and Villegas, 2007; Rojek et al., 2011).
Dimensional analysis is performed, based on the Buckingham π theorem. This
theorem states that any physically meaningful relationship of N variables can be
expressed equivalently by a function of N − c dimensionless parameters, where c is
the number of primary dimensions, in this case being mass, length and time. The
macroscopic parameters that define the behavior of dry rock are: Young’s modulus
E, Poisson’s ratio ν, uni-axial compressive strength σU CS and tensile strength σBT S .
The hydro-mechanical coupling effects are characterized by the fluid’s bulk modulus
M , dynamic viscosity η, hydrostatic pressure ph , pore pressure pp , the rock’s in-
trinsic permeability κ and effective stress coefficient α. The micro-parameters that
define the macro-scopic properties of the simulated rock are kn , ks , Tn , Ts , µ, αd .
5 Furthermore, the properties of the particle assembly are of influence as well: average
particle diameter d¯p , porosity of the particle assembly n (not necessarily the same
porosity as that of the real specimen), material density ρ, loading velocity V and
characteristic specimen dimension L (possible scale effects, (Yenigül and Alvarez
Grima, 2010)). This leads to a total of 16 parameters, which can be reduced to 13
independent parameters. Note that κ and η only occur once in the whole set of
equations in the same term that they cannot be treated as separate parameters.
The set of parameters is not unique and can be modified by considering other
parameters as well. One can argue whether the type of particle size distribution
(e.g. Gaussian, uniform) should be considered or not. Here it is assumed that the
geometrical porosity of the particle assembly is directly related to the particle size
distribution. The set of independent parameters considered in this study is:

kn Tn kn d¯p V
Dry rock , , , µ, αd p (5.1)
ks Ts Tn kn /ρ
dp
Geometry ,n
L
M ph κ M
Hydromechanical coupling , αd , ,
kn pp η V dp

Even more parameters could be necessary to include when considering more

complex bond and/or collision models. Due to the large number of independent
variables, some assumptions and simplifications are made. As stated in chapter
4, the numerical damping is chosen to be exactly the same as used by Potyondy
and Cundall (2004) and is thus fixed at the value of 0.7. Geometry scaling effects
of particle size with respect to specimen size are already thoroughly investigated
by Yenigül and Alvarez Grima (2010). Concerning the hydro-mechanical coupling,
various parameters are limited to the application of water or air. Other fluids are
beyond the scope of this dissertation. In order to investigate hydro-mechanical
5.1. Parameter sensitivity 81

coupling effects on a generic type of rock, these effects will be investigated in a

parameter study on material tests (UCS and BTS) for dry and saturated rock, in
the undrained regime.
Due to the large number of simulations needed in the parameter sensitivity
analysis, post processing to determine the macro parameters is automated. The
criteria that are used are:
• σc : maximum force acting on a plate in a UCS-test, divided by W , the width
of the sample, to obtain stress.
2Fmax
• σt : maximum force acting on a plate in a BTS test, calculated as σt = πW .

• E 0 : determined at 0.2 millistrain before maximum force in a UCS-test, calcu-

lated as E 0 = σyy1 .

• ν 0 , determined at 0.2 millistrain before maximum force in a UCS-test, cal-

culated as ν 0 = − xx
yy
, where xx is determined by measuring the averaged
x-coordinate of the outermost particles at these free boundaries.
5
Currently, the DEM-SP is in 2D, inherently assuming plane strain conditions (that is
why 0 is used). Therefore the Poisson’s ratio and the Young’s modulus are corrected
for 3D analysis through
ν0
ν= (5.2)
1 − ν0
E0
E= (5.3)
1 − ν2
For each test both the damaged sample and the stress-strain curves are checked
to verify that the obtained result indeed corresponds with an artificial rock and
the criteria mentioned before. Combinations of micro parameters that would result
in behavior that does not correspond with the criteria mentioned before are not
considered within the scope of this dissertation. Some examples of what might
happen are:
• Maximum stress is not the same as stress at failure, i.e. the frictional strength
of the sample is larger than its structural strength.
• Fluid pressure gradient forces are significantly larger than the bond strength,
resulting in disintegration of the sample.

5.1.1. Results
The parameter study concerns the comparison of the dry and saturated macroscopic
properties of generic rock. In this study, the micro-parameters kn , ks , Tn , Ts are
varied. The friction coefficient µ is not varied, since it is only of limited effect when
considering brittle failures up to the occurrence of the macroscopic failure (Huang,
1999). The properties of all rock samples are in this case kept identical, using the
same particle assembly as well. For an overview of the constant and the varying
parameters, see respectively tables 5.1 and 5.2 for an overview of respectively the
 82 5. Validation - Material tests

Table 5.1: Constant properties for sensitivity analysis.

Parameter Quantity Unit

α 0.5 [-]
µ 1.0 [-]
κ 1 · 10−19 [m2 ]
η 0.001 [Pas]
Kf 2.0 [GPa]
ph 0.1 [MPa]
pi 0.1 [MPa]
H 0.07 [m]
W 0.035 [m]
Rmin 0.2 [mm]
Ravg 0.3 [mm]
Rmax 0.4 [mm]
h 0.0015 [m]
5
Table 5.2: Varying properties for sensitivity analysis.

Parameter Quantities
kn 1.25 2.5 5.0 7.5 10 [GPa]
ks 0.5kn 0.75kn 1.0kn 1.5kn 2.0kn [GPa]
Tn 1 · 10−6 kn 2 · 10−6 kn 4 · 10−6 kn 6 · 10−6 kn 8 · 10−6 kn [N/m]
Ts 1 · 10−6 ks 2 · 10−6 ks 4 · 10−6 ks 6 · 10−6 ks 8 · 10−6 ks [N/m]

constant and the varying parameters. For each combination of settings, four tests are
performed and analyzed, being both dry (without any fluid coupling) and saturated
for both UCS and BTS tests. Due to the random generation of the set of parameters,
not all combinations lead to a stable solution to simulate rock. Especially the cases
with high bond stiffness and low bond strength might result in this unrealistic
behavior. These results are omitted from this analysis.
The results of the parameter sensitivity study on the macroscopic rock properties
are shown in figures 5.1 to 5.8. All figures are based on UCS tests, except figures 5.4
and 5.7 are based on BTS tests and figures 5.5 and 5.8 are based on the combinations
of UCS and BTS tests.

5.1.2. Discussion
Figure 5.1 presents the variation of the macroscopic Young’s modulus as a function
of the microscopic parameters for both dry and saturated conditions. Based on the
results, several observations can be made. First of all, there is a significant difference
in the measured Young’s modulus between saturated and dry rock. In the saturated
cases, the Young’s modulus is significantly larger. This is due to the additional stiff-
ness that is contributed by the fluid against any volumetric deformation. Secondly,
5.1. Parameter sensitivity 83

with an increase in the ratio of shear to normal stiffness ks /kn the bulk modulus of
the rock sample slightly increases as well. And a third significant effect is that with
an increase in the ratio of bond stiffness over bond strength kn Ravg /Tn , the Young’s
modulus increases as well. The ratio can be used as a measure of the relative bond
length with respect to the particle size. Low ratios allow for a relatively large bond
length. As a result, the particle assembly allows for larger mobility of the particles
before the bonds will fail, effectively lowering the bulk stiffness of the assembly.

Variation of E
1.5
E/kn [−]

0.5

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4
ks /kn [−] 5
(a) Dry conditions.
Variation of E
1.5
E/kn [−]

0.5

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4
ks /kn [−]

(b) Saturated conditions.

Figure 5.1: Variation of Young’s modulus as a function of Ks /Kn . Micro-parameters correspond

with input presented in tables 5.1 and 5.2.

The dependence of the Poisson’s ratio ν on the bond stiffness ratio ks /kn is
shown in figure 5.2. When the stiffness ratio increases, the Poisson’s ratio decreases
significantly. In saturated conditions, a similar trend is shown. However, it shows
a higher Poisson’s ratio compared to dry conditions, for all combinations of the
micro-parameters used in this parameter sensitivity analysis.
Both trends for the dry Young’s modulus and Poisson’s ratio show similar trends
compared to Huang (1999); Fakhimi and Villegas (2007); Rojek et al. (2011).
 84 5. Validation - Material tests

Variation of ν
0.6

0.4
ν [−]

0.2

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4
5 ks /kn [−]

(a) Dry conditions.

Variation of ν
0.6

0.4
ν [−]

0.2

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4
ks /kn [−]

(b) Saturated conditions.

Figure 5.2: Variation of Poisson’s ratio as a function of Ks /Kn Micro-parameters correspond with
input presented in tables 5.1 and 5.2.
5.1. Parameter sensitivity 85

Figures 5.3, 5.4 and 5.5 the compressive, tensile and the ratio of compressive
to tensile failure with respect to the input of the micro parameters. These figures
show that there is a weak effect of the average bond length kn Ravg /Tn on the scaled
strengths σc Ravg /Tn and σt Ravg /Tn , when assuming a constant bond strength ratio
Ts /Tn . This is the case for both dry and saturated conditions. Interesting to notice
is that the compressive strength is lower for saturated compared to dry conditions
for values of kn Ravg /Tn > 75. The same holds for the tensile strength, but here the
reduction is less significant. With respect to the strength ratio of the rock sample,
a slight increase in possible ratios is noted for larger values of kn Ravg /Tn , see figure
5.5. When comparing dry and saturated conditions, it can be observed that the
strength ratio is lower and has a wider spread in saturated conditions than is the
case in dry conditions.

5
 86 5. Validation - Material tests

Variation of σc
σc Ravg /Tn [−]

0.5

0
0 50 100 150 200 250 300 350
Kn Ravg /Tn [−]
5
(a) Dry conditions.
Variation of σc
σc Ravg /Tn [−]

0.5

0
0 50 100 150 200 250 300 350
Kn Ravg /Tn [−]
log Ts /Tn

−2.5 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2 2.5

(b) Saturated conditions.

Figure 5.3: Variation of compressive strength as a function of Kn Ravg /Tn Micro-parameters cor-
respond with input presented in tables 5.1 and 5.2.
5.1. Parameter sensitivity 87

Variation of σt
0.2
σt Ravg /Tn [−]

0.15
0.1
0.05
0
0 50 100 150 200 250 300 350
Kn Ravg /Tn [−]
5
(a) Dry conditions.
Variation of σt
0.2
σt Ravg /Tn [−]

0.15
0.1
0.05
0
0 50 100 150 200 250 300 350
Kn Ravg /Tn [−]
log Ts /Tn [−]

−2.5 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2 2.5

(b) Saturated conditions.

Figure 5.4: Variation of tensile strength (BTS) as a function of Kn Ravg /Tn Micro-parameters
correspond with input presented in tables 5.1 and 5.2.
 88 5. Validation - Material tests

Variation of σc /σt
8
6
σc /σt [−]

4
2
0
0 50 100 150 200 250 300 350
kn Ravg /Tn [−]
5
(a) Dry conditions.
Variation of σc /σt
8
6
σc /σt [−]

4
2
0
0 50 100 150 200 250 300 350
kn Ravg /Tn [−]
log Ts /Tn [−]

−2.5 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2 2.5

(b) Saturated conditions.

Figure 5.5: Variation of strength ratio σc /σt as a function of Kn Ravg /Tn Micro-parameters cor-
respond with input presented in tables 5.1 and 5.2.
5.1. Parameter sensitivity 89

Figure 5.6 presents the compressive strength with respect to the ratio of the bond
shear and tensile strengths. Based on this figure, two regimes can be identified, i.e.
for Ts /Tn < 1, the compressive strength is dominated by microscopic tensile failures,
while for values > 1 microscopic shear failures are dominant. This holds for both
dry and saturated conditions. Similar trends can be observed for the scaled tensile
strength, see figure 5.7. Figure 5.8 shows the strength ratio with respect to the
bond strength ratio on a semilog scale. The transition between tensile and shear
dominated failure mechanisms is located near Ts /Tn = 1. In the figure it is also
shown that the higher values for σc /σt are obtained for cases with smaller bond
lengths (larger values for kn Ravg /Tn ).

Variation of σc
σc Ravg /Tn [−]

0.5
5
0
0 2 4 6 8 10 12 14 16 18 20
Ts /Tn [−]

(a) Dry conditions.

Variation of σc
σc Ravg /Tn [−]

0.5

0
0 2 4 6 8 10 12 14 16 18 20
Ts /Tn [−]
log Kn Ravg /Tn [−]

3.8 4 4.2 4.4 4.6 4.8 5 5.2 5.4 5.6

(b) Saturated conditions.

Figure 5.6: Variation of compressive strength as a function of Ts /Tn . Micro-parameters correspond

with input presented in tables 5.1 and 5.2.
 90 5. Validation - Material tests

Variation of σt
0.3
σt Ravg /Tn [−]

0.2

0.1

0
0 2 4 6 8 10 12 14 16 18 20
Ts /Tn [−]
5
(a) Dry conditions.
Variation of σt
0.3
σt Ravg /Tn [−]

0.2

0.1

0
0 2 4 6 8 10 12 14 16 18 20
Ts /Tn [−]
log Kn Ravg /Tn [−]

3.8 4 4.2 4.4 4.6 4.8 5 5.2 5.4 5.6

(b) Saturated conditions.

Figure 5.7: Variation of tensile strength (BTS) as a function of Ts /Tn . Micro-parameters corre-
spond with input presented in tables 5.1 and 5.2.
5.1. Parameter sensitivity 91

Variation of σc /σt
10
8
σc /σt [−]

6
4
2
0
0 2 4 6 8 10 12 14 16 18 20
Ts /Tn [−]
5
(a) Dry conditions.
Variation of σc /σt
10
8
σc /σt [−]

6
4
2
0
0 2 4 6 8 10 12 14 16 18 20
Ts /Tn [−]
log Kn Ravg /Tn [−]

3.8 4 4.2 4.4 4.6 4.8 5 5.2 5.4 5.6

(b) Saturated conditions.

Figure 5.8: Variation of strength ratio σc /σt as a function of Ts /Tn . Micro-parameters correspond
with input presented in tables 5.1 and 5.2.
 92 5. Validation - Material tests

In order to verify whether DEM-SP does not lead to significant differences, the
same set of simulations have been performed with air as pore fluid. The fluid
properties that are used are a bulk modulus of 2 MPa and a fluid viscosity of 1·10−6
Pas. A comparison of the resulting Young’s modulus and compressive strength of
fluid-less DEM and DEM-SP with air is shown in figure 5.9. Both properties show
almost a perfect match, only slightly some differences in Young’s modulus can be
observed. The differences that are noted are well within what is reasonable for the
accuracy of the methodology.

Young’s modulus Compressive strength

4 40

σU CS with air [MPa]

E with air [GPa]

3 30

2 20
5
1 10

0 0
0 1 2 3 4 0 10 20 30 40
E without fluid coupling [GPa] σU CS without fluid coupling [MPa]

(a) Comparison of Young’s modulus. (b) Comparison of compressive strength.

Figure 5.9: Comparison of macro-properties based on DEM without any fluid coupling and DEM-
SP with air
5.2. Fluid stress effects 93

5.2. Fluid stress effects

In this section, various tests are presented that specifically look more into the effects
of the hydro-mechanical coupling itself. First a set of tests is presented in which the
effective stress theory of Terzaghi is verified. This is followed by a set of simulations
concerning a range of strain rate effects in shale, which is based on a comparison
with the experiments of Swan et al. (1989).
The micro and macro parameters that are used for these tests are shown in table
5.3. Note that in the 2D case to obtain practical units for the micro and macro
parameters that a unit thickness of 1 is assumed for all particles. A comparison
between the simulated rock properties and the properties of the shale are shown
in table 5.4. Note that a range of permeabilities is used in the numerical rock to
correct for the Deborah number N , which will be discussed in section 5.2.2.

Table 5.3: Micro and macro parameters of the artificial rock

Micro parameter Quantity Macro parameter Quantity

kn 9
5 · 10 [Nm ] -2
UCS (dry) 21.4 [MPa] 5
ks 2.5 · 109 [Nm-2 ] E (dry) 3.5 [GPa]
Tn 2 · 104 [Nm-1 ] ν (dry) 0.27 [-]
Ts 104 [Nm-1 ] Sample height 0.07 [m]
µ 1 [-] Sample width 0.035 [m]
α 0.5 − 0.75 − 1.0 [-] Nr of particles 7201
αdamping 0.7 [-] ˙ 1.43 [s-1 ]
rmin 0.2 [mm] H 0.07 [m]
rmax 0.4 [mm] W 0.035 [m]
# of particles 7196 Nc 3.98 [-]
n 0.18 [-]
Pmin 0 [Pa]

Table 5.4: Comparison of simulated rock and Kimmeridge Bay shale Swan et al. (1989)

Parameter Kimmeridge Bay shale Swan et al. (1989) DEM-SP

U CS [MPa] 23.5 21.4
E [GPa] 3.4 3.5
ν [-] 0.14 0.27
κ [m2 ] 8 · 10−21 0 − 10−15

5.2.1. Effective stress

The basic principle of the effective stress theory applied to rock mechanics has been
verified with DEM-SP. About 100 simulations of jacketed bi-axial tests have been
performed. In these simulations, the initial pore pressure, the confining pressure
and thus the effective confining pressure are varied (with pressure values of 0.1,
 94 5. Validation - Material tests

Strength with constant effective stress

pef f = 0.1 MPa

40
σ1 − σ3 [M P a]
pef f = 0.5 MPa
pef f = 1 MPa
pef f = 2 MPa
20 pef f = 4 MPa
pef f = 8 MPa
pef f = 16 MPa
0
0 10 20 30
ppore [MPa]

Figure 5.10: Simulated strength values with respect to initial pore pressure.

5 0.5, 1, 2, 4, 8, 16 and 32 MPa). A comparison of the experiments with constant

effective confining stress is shown in figure 5.10. In this figure it is clearly shown
that the strength of the rock does not differ for constant effective stress. Figure 5.11
shows the damage and pressure distribution in the simulated rock that is tested in
saturated atmospheric conditions, i.e. ppore = pconf = 0.1 MPa. Damage is defined
#bonds
as 1 − #initialbonds . The damage and pressure distribution for the sample tested at
ppore = 0.1 and pconf = 16 MPa are shown in figure 5.12.
5.2. Fluid stress effects 95

5
(a) Damage distribution. (b) Pore pressure distribution.

Figure 5.11: Simulation results of artificial rock tested with ppore = pconf = 0.1 MPa. Results
shown at the moment of failure of the rock.

(a) Damage distribution. (b) Pore pressure distribution.

Figure 5.12: Simulation results of artificial rock tested with ppore = 0.1 MPa and pconf = 16 MPa.
Results shown at the moment of failure of the rock.
 96 5. Validation - Material tests

5.2.2. Strain rate effects

In this section, biaxial compression simulations with varying strain rate are com-
pared with experiments on Kimmeridge Bay shale, as those are described by Swan
et al. (1989). Only few references are found that describe bi- or tri-axial compression
tests with varying strain rates on rock samples, simulated in DEM. Zhang and Wong
(2013) performed an analysis of strain rate effects on gypsum in uniaxial compression
tests in DEM with a bonded particle model. In Zhang and Wong (2014) a further
extended this analysis based on data from literature to the effect of loading rate on
uniaxial compression tests on rock-like materials in DEM. Most other publications
that deal with DEM and a rock-like material are based on split Hopkinson bar tests
(both tensile and compressive), e.g. Li et al. (2013), or are based on impact loading
of debris falling down or impacted at high velocity, e.g. Carmona et al. (2008);
Zhao et al. (2017). Biaxial tests with varying strain rates in DEM are also applied
in DEM for modeling of granular media, e.g. (Singh and Luding, 2010; Stahl and
Konietzky, 2010). Although these are not directly applicable for rock-like materials,
those simulations can help to obtain insight in the effect of friction between particles
and particle shapes on the residual strength of a fragmented rock specimen.
5 Swan et al. (1989) performed two series of experiments on Kimmeridge Bay shale
in which they investigated the effect of strain rate on the deformation, strength
and failure properties of the rock. In the first, no drainage of pore fluid could
occur across the sample boundary. The deviatoric stress is applied at the required
constant rate of axial deformation. In the second series, isotropic consolidation was
achieved under fully drained conditions. Upon completion of the isotropic loading
they closed the drainage lines and the deviatoric stress applied under nominally
undrained conditions, as in the first series. Swan et al. (1989) performed experiments
at various confining pressures, of which most of the experiments were done at a
confining pressure of 50 MPa.

Strain rate dependency of simulated dry rock

30
σ1 [M P a]

20

10

0
10−1 100 101
−1
˙ [s ]

Figure 5.13: Strength simulated dry rock with respect to applied strain rate

It is not possible to directly apply the strain rates in the simulations, due to the
5.2. Fluid stress effects 97

large range in orders of magnitude. Not only the required computational time for
the slowest strain rates would take too long, but also the strain rate dependency
of the simulated dry rock is of influence, as is shown in figure 5.13. Zhang and
Wong (2013) obtained a similar result, based on DEM simulations with varying
strain rates in compression tests on gypsum blocks with a pre-existing flaw. The
problem is that in the case of the higher strain rates, work is applied in a too high
rate to ensure sufficient numerical damping to result in a realistic solution, which
is also the reason why it is often stated that the velocities in the simulation need
to be significantly smaller than the speed of sound of the artificial rock (Huang,
1999). The failure mechanism that is observed in simulated UCS-tests shows strong
resemblance when in practice, equipment with a low stiffness is used (Hemami and
Fakhimi, 2014). After failure, the strain energy that is stored in the machine needs
to be released. If this release of strain energy allows for a large displacement, the
failed specimen is immediately again loaded and the frictional strength of the failed
specimen might get higher than the strength of the residual intact rocks, leading
towards a higher apparent strength and a transition from brittle to ductile behavior.
In order to simulate the strain rate effect, scaling of the intrinsic permeability is
applied through the dimensionless number N , as in equation (2.15). 5
The simulation and experimental results of the unconsolidated compression tests
and consolidated compression tests are shown in figures 5.14 and 5.15 respectively.
The cumulative accoustic emissions (total nr. of broken bonds) measured in the
simulations are shown in figure 5.16. Damage and pressure distribution of a sample
tested at lower and higher Deborah number are respectively shown in figures 5.17
and 5.18.

5.2.3. Discussion
Based on figure 5.14, the critical strain rate for the shale occurs at a deformation
rate with N ≈ 2.3 · 105 . Below this value there is no significant increase in strength
with increasing strain rate. For experiments with N > 104 , a significant increase is
noted. The strengthening rate of the DEM-SP seems less significant compared to
the experiments. It needs mentioning that the simulation at λ = 1.7 · 109 is basically
the upper limit in strain rate dependency of the DEM-SP. Here the pore pressure
diffusion is almost identical to the case where the rock is considered to be perfectly
impermeable (i.e. κ = 0). Besides that, in this simulation it is also noted that the
pore fluid in the macro crack is fully cavitating, resulting in a limited strengthening
effect.
 98 5. Validation - Material tests

Unconsolidated undrained tri/bi-axial test at 50 MPa

5 80 Experiments
σ1 − σ3 [M P a]

Simulations
60

40

20

0
103 104 105 106 107 108 109
N [−]

Figure 5.14: Unconsolidated tri-axial tests at 50 MPa confining pressure (experiments of (Swan
et al., 1989)), compared with bi-axial simulations (2D). N is a representation of the Deborah
number, as given in equation (2.15).
5.2. Fluid stress effects 99

In the consolidated experiments and simulations both strengthening and weak-

ening effects are observed, see figure 5.15. For experiments performed at N < 104 ,
the material tends to weaken. Due to the compression that is applied to the samples,
the volume decreases, resulting in an increase in pore pressure. This increase results
in a lower effective stress and eventually in a lower strength of the sample. For ex-
periments and simulations with λ > 104 , dilative strengthening causes the strength
of the samples to increase again. Both the weakening and the strengthening effect
seem to be less pronounced in the simulations with respect to the experiments.

Consolidated undrained tri/bi-axial test at 50 MPa

80
σ1 − σ3 [M P a]

60

40

20 Experiments 5
Simulations
0
103 104 105 106 107 108 109
N [−]

Figure 5.15: Consolidated tri-axial tests at 50 MPa confining pressure (experiments of (Swan et al.,
1989)), compared with bi-axial simulations (2D). N is a representation of the Deborah number, as
given in equation (2.15).

Undrained tri-axial test at 50 MPa

1,500
[−] bonds

1,000
#broken

500
Consolidated
Unconsolidated
0
103 104 105 106 107 108 109
N [−]

Figure 5.16: Number of broken bonds in simulations. N is a representation of the Deborah number,
as given in equation (2.15).

The trend in the accoustic emissions with respect to the applied strain rate
 100 5. Validation - Material tests

corresponds with the observations of Rutter (1972), that with increasing strain rate,
rock tends to show less strain softening (i.e. more ductile) and that the amount of
damage (number of accoustic emissions) increases.

5
(a) Damage distribution. (b) Pore pressure distribution.

Figure 5.17: Simulation results of artificial rock tested with ppore = pconf = 50 MPa and N = 170.
Results shown at the instance of macro-failure of the rock.

Several drainage related mechanisms have been identified in simulations. The

phenomena of compaction weakening, stiffening and the occurrence of cavitation
correspond with the trends described in chapter2. Except for the occurrence of
dilatancy hardening. In DEM-SP, the rock starts to dilate when the sample has
already failed, while in practice, often the sample dilates before it fails (Bernabe,
1987). This is most likely caused by the strongly simplified geometry of the rock,i.e.
the grains are idealized as circular particles, the simulations are based on 2D and not
3D. Additionally, near internal discontinuities (micro-cracks), the SP will smoothen
the pore pressure, ignoring the local discontinuity.

5.3. Conclusions
A parameter sensitivity analysis is performed and analyzed with the use of the
Buckingham-π theorem in which the microscopic input parameters for bond stiffness
and bond strengths are varied for both UCS and BTS tests. The results show
that for saturated conditions, given that a constant Deborah number is used, a
decrease in compressive strength and strength ratio. Furthermore an increase in
Young’s modulus, Poisson ratio is noted. No significant differences with respect to
the tensile strength of the sample are noted. Mechanical properties of DEM-SP filled
with air simulations show little difference compared to DEM simulations without
fluid coupling effects.
The DEM is found to be highly sensitive above a critical strain rate. Low strain
5.3. Conclusions 101

(a) Damage distribution. (b) Pore pressure distribution.

Figure 5.18: Simulation results of artificial rock tested with ppore = pconf = 50 MPa and N =
1.7 · 107 . Results shown at the instance of macro-failure of the rock.
5
rates are undesirable in DEM due to the large increase in computational effort.
Therefore strain rate effects on saturated rock should be modeled by scaling the
Deborah number through permeability instead of strain rate.
It is shown that the effective stress theory is still valid when using the DEM-
SP approach. Furthermore, tri-axial tests with varying strain rates on Kimmeridge
Bay shale have been simulated and compared with experiments. Scaling of the
strain rate in the simulations is achieved by scaling through the Deborah number
by adjusting the permeability, instead of the applied strain rate. The results show
that for the case where the initial pore pressure is equal to the hydrostatic pressure,
an increase in strain rate first results in compactive weakening, and for a further
increase of the strain rate dilatancy hardening becomes dominant. In the case that a
higher hydrostatic pressure is used compared to the initial pore pressure, failure the
resultant compressive strength of the samples increases significantly due to dilatant
hardening.
Additionally, based on the results presented on the range in strain rate/permeability
in the simulation, it is expected that the use of an adaptive permeability will be of
lesser relevance to the model. Especially because in practice the uncertainties con-
cerning the rock properties are already quite large, since rock is a natural material.
For practical applications the accuracy of the model is thus limited by the tests that
are done to measure the rock properties.
 Validation - Tool-Rock
6
Interaction
"The first principle is that you must not fool yourself and you are the easiest
person to fool."
Richard Feynman

Several validation cases with respect to rock cutting processes are presented in this
chapter. A set of simulations are compared to 2D rock (tile) cutting experiments that
are performed on dry artificial rock. Additionally, the method is used to compare the
tool-rock interactions that are more comparable to the practices of the industry. One
set of simulations based on a confining stress is compared to drilling experiments.
A comparison is also made between the DEM-SP methodology and the existing tech-
niques found in literature. Another set of simulations comprises the effect of water
depth on the rock cutting process for deep sea mining applications. These results
are also validated with experimental results found in literature. As a test case, the
cutting motion of a single tooth on a cutter head is simulated.

Section 6.3 is based on improved simulations of those published in Helmons et al. (2015).
Section 6.4 has been published in Helmons et al. (2016a).
Section 6.5 has been published in Helmons et al. (2016c).

103
 104 6. Validation - Tool-Rock Interaction

6.1. Introduction
Through the 1970’s till the beginning of 1990’s, an extensive research program on
rock cutting processes was commissioned by the Dredging Research Association
(in Dutch: Commissie Speurwerk Baggertechniek, nowadays Stichting Speurwerk
Baggertechniek, currently the joint research organisation of Boskalis and Van Oord)
to Delft Hydraulics and Delft Geotechniques (now Deltares). Within this research
program, emphasis was on the cutting processes that are of relevance to the, in
dredging commonly used, crown cutter. Some of the experiments are made available
to use for validation in this thesis. These are the 2D tile cutting experiments (Meijer,
1972, 1973b,a).
As already described in section 4.2, several researchers modeled the confining
stress in the drilling process as a boundary condition in the simulations, inherently
assuming dry rock material. The DEM-SP method is used to recreate the cases
described by Kaitkay and Lei (2005) in order to show that application of a confining
stress in the DEM gives proper results.
Furthermore, experimental data concerning the rock cutting process at hyper-
baric (ph > σc , σt ) conditions Alvarez Grima et al. (2015) for deep sea mining
purposes, carried out by Royal IHC, is publicly available and can be used for valida-
tion. On a similar type of rock, the effect of both hydrostatic pressure and cutting
velocity have been tested.
6 As a test case for dredging applications, the cutting process of a single tooth
mounted on a cutter head is simulated. Other than the results of the high pressure
cases in drilling and deep sea mining, in dredging the range in hydrostatic pressure
is limited to 30m (0.3 MPa). Additionally, all cutting models presented in chapter
3 and all cutting simulations performed in this thesis thus far are based on linear
cutting. Most significant differences in rotational compared to linear cutting are the
non constant cutting depth and that the tool no longer moves in a direction parallel
to the rock bed.
The various aforementioned cases of tool-rock interactions are discussed in this
chapter. Each of these cases starts with a brief overview of the experiments per-
formed, although these experiments were not performed with this thesis in mind.
Therefore, for sake of completeness, a brief description of these experiments is pre-
sented. This is then followed by the input parameters and simulation geometry
setup for these experiments. The simulated results, after post-processing, will be
compared with the experimental results. At the end of this chapter, an overall
discussion of the presented tool-rock interactions is given.

6.2. 2D tile cutting

6.2.1. Experiments
The experiments, described by Meijer (1972, 1973b), consist of a chisel that is
pushed into a tile of artificial rock. The thickness of the tile is smaller than the
width of the tool to ensure a 2D cutting process. The force needed to cut the tile
is generated a hydraulic cylinder. The cylinder pushes a cart on which the tool is
mounted to enforce a linear movement of the tool into the tile. A force sensor is
6.2. 2D tile cutting 105

positioned between the cylinder and the cart. The height and angle of the chisel can
be altered. The chisel is pushed at a velocity of 1 mm/s into the rock. See figure
6.1 for a schematic overview of the test setup. The movement of the tool stopped
when a large chip occurred, limiting each experiment to one chip.
Hydr. cylinder
Force sensor

Sled
Filler plates
Signal to decoder α
Chisel

H
Tile

32 cm
Clamping

40 cm

Figure 6.1: Experimental setup used for the tile cutting experiments, as in (Meijer, 1972, 1973b).

The experiments are performed on tiles artificial rock with dimensions 40 x 32

x 2 cm. The artificial rocks are based on the mixture as given in table 6.1. The
material properties of the artificial rock are presented in table 6.3, together with a
comparison of the material properties of the simulated rock. The Young’s modulus 6
of the tiles in the experiments was not measured. Its value for the simulations is
estimated as approximately 450·σU CS , which is comparable for an average type of
limestone, see appendix A.

Table 6.1: Mixture components for artificial rock used in 2D tile cutting experiments

Component %mass
Sand 69.2
Cement 13.3
Water 13.3
Clay 3.5
Calcium-chloride 0.7

Results and comparisons of the cutting forces and penetration depth are pre-
sented in respectively tables 6.4 and 6.5. A comparison between the typical chip
shapes that are obtained in the simulations and experiments is shown in figure 6.2.

6.2.2. Discussion
As depicted in table 6.4, the required cutting forces correspond well with those
observed in experiments. All of the simulated results lie within one standard de-
viation (std) from the mean cutting force measured in the experiments. Although,
at smaller cutting depths, the simulated results seem to underestimate the required
cutting forces. With respect to the penetration depth, these values can significantly
 106 6. Validation - Tool-Rock Interaction

Table 6.2: Input parameters for tile cutting experiments in DEM-SP

Symbol Quantity
2
kn [N/m ] 5.0 · 109
ks [N/m2 ] 5.0 · 109
Tn [N/m] 1.25 · 104
Ts [N/m] 1.25 · 104
µ [-] 1.0
rmin − rmax [mm] 0.5 − 1.0
r̄ [mm] 0.75
∆t [s] 3 · 10−7
#particles [-] 56179
vc [m/s] 0.03125

Table 6.3: Mechanical properties of rock used in 2D tile cutting experiments. Material properties
are found in (Meijer, 1972, 1973b)

Property Experiment DEM-SP

(mean - std)
σU CS [MPa] 11.38 - 0.88 10.36
6 σBT S [MPa] 1.17- 0.08 1.17
E [GPa] Unknown 4.51

Figure 6.2: Comparison of chip shape of simulations and experiments, with α = 45◦ and tc = 6.5
cm. Simulated results depict damage, blue being intact and red completely disintegrated.

differ from those observed in the experiments. First of all, calculations of pene-
tration depths that small in DEM, i.e. tp ≈ r̄ are prone to discretisation errors.
The simulated rock is not a perfect block, but an assembly of finite sized particles
that are connected together. Hence, the penetration depth is strongly influenced
by at which exact location the tool enters the rock, i.e. whether the tooltip moves
directly towards a particle center or whether it moves exactly between two particles
can make a big difference, see figure 6.3 for clarification.
6.2. 2D tile cutting 107

Table 6.4: Tile cutting forces obtained in experiments and simulations. Experimental data is found
in (Meijer, 1972, 1973b)

Rake angle Cutting depth Experiments DEM-SP

α [◦ ] tc [m] Fc [kN] (mean - std) Fc [kN]
45 3.5 1.589 - 0.216 0.915-1.122
30 6.5 1.97 - 0.216 2.2
45 6.5 1.76 - 0.39 1.68-1.8
60 6.5 1.69 - 0.177 1.243-1.767
30 6.5 1.97 - 0.216 2.2

Table 6.5: Tile cutting penetration depth obtained in experiments and simulations. Experimental
data is found in (Meijer, 1972, 1973b)

Rake angle Cutting depth Experiments DEM-SP

α [◦ ] tc [m] tp [m] (mean - std) tp [mm]
45 3.5 1.22 - 0.36 0.3
30 6.5 1.97 - 0.216 2.2
45 6.5 1.29 - 0.38 2.6
60 6.5 2.21 - 0.55 5.8
6
The shape and size of the chips generated both in the experiments and simula-
tions correspond well. The simulated chip shows a less smooth crack trajectory than
those in the experiments. This is likely caused by the limited brittleness of the sim-
ulated rock. In the current model, only shear and normal bonds are incorporated,
which restrict deformation in normal and in tangential directions along the particle
bonds, but it does not restrict the bending moment that particles can make. This
might be improved upon by using the parallel bond model of Potyondy and Cundall
(2004).

(a) Attacking particle through center

(b) Attacking between particles

Figure 6.3: Discretisation effects in tool - rock interactions

 108 6. Validation - Tool-Rock Interaction

6.3. Drilling
In this section, the DEM-SP approach is validated against the experimental data on
drilling experiments of Kaitkay and Lei (2005). These experiments are especially
interesting because Lei and Kaitkay (2003) also tested their DEM-based modeling
approach against these results, allowing for a comparison between their method and
the DEM-SP approach.

6.3.1. Experiments
The single point cutting experiments are performed on Carthage marble. The rock
sample is placed in a pressure vessel in which the desired hydrostatic pressure is
applied. Vertical feed and the rotary motion of the cutter are controlled. Table 6.6
lists the machining conditions tested by Kaitkay and Lei (2005). The properties of
the Carthage marble samples they have used are presented in 6.7. The DEM-SP
requires that for a reasonable solution of the cutting process, a cutting depth of at
least several particles needs to be achieved. The particle size is chosen in such a
way that on average the cutting depth is five particles deep. In order to limit the
computational expense of the simulations, a smaller block with a height of 0.016 m
and a width of 0.05 m is used, similar to the size used by Lei and Kaitkay (2003).

Table 6.6: Machining conditions for drilling experiments, (Kaitkay and Lei, 2005).

6 Parameter Setting
Feed 0.8 [mm/r]
Cutting speed 1 [m/s]
Rake angle -15, -25 ◦
External pressure 0.1, 3.44, 34.4 [MPa]

Table 6.7: Properties of Carthage marble and simulated rock.

Property Carthage marble Lei and Kaitkay (2003) DEM-SP

E [GPa] 44.8 41.8
ν [-] 0.24 0.21
σc [MPa] 103 105
σc at σ2,3 = 34.5 [MPa] 186 154
σBT S [MPa] unknown 14

The input parameters used in DEM-SP are shown in table 6.8. The compressive
strength at high confining pressures does not match, the current implementation
of the DEM models does not allow for such an increase in strength with confining
pressure. On the other hand, Cunningham and Eenink (1959) already showed by
experiments that the differential pressure (bottomhole - pore pressure) has a more
profound effect than the increase in compressive strength of the rock due to the
confinement of the differential pressure. Furthermore, the tensile strength of the
6.3. Drilling 109

simulated material is given. Unfortunately, the tensile strength of the rock used
in the experiments is not mentioned (Kaitkay and Lei, 2005). Fixed displacement
boundary conditions are applied on the particles at the bottom and right end of the
rock specimen.

Table 6.8: Input parameters used generate the simulated rock as in 6.7.

Symbol Quantity
2
kn [N/m ] 7 · 1010
ks [N/m2 ] 2 · 1010
Tn [N/m] 2 · 104
Ts [N/m] 1.3 · 104
µ [-] 0.75
rmin − rmax [mm] 0.0533 − 0.01067
r̄ [mm] 0.08
∆t [s] 1 · 10−8
ph [MPa] 0.1 − 34.5
h [m] 4 · 10−4

Care must be taken when analyzing the data of Kaitkay and Lei (2005), as the
data presented at atmospheric conditions is actually based on a different type of 6
experiment. The experiments at elevated pressure are performed with the drill axis
parallel to gravity, while the experiments at atmospheric conditions are performed
with the drill axis perpendicular to gravity and without a confining fluid, see figure
6.4 for clarification. The drill bit is positioned such that the generated debris falls
to the side of the bit (in DEM-SP that would mean out of plane movement of the
debris).

Atmospheric, horizontal feed Pressurized, vertical feed

chuck clamps
workpiece rotation

rotation
toolholder cutting tool

workpiece

feed pressure vessel

cutting tool

Figure 6.4: Experimental setups as used by Kaitkay and Lei (2005).

 110 6. Validation - Tool-Rock Interaction

6.3.2. Simulated results

Figure 6.8 shows a set of image sequences representing the damage that occurs dur-
ing the cutting process with a rake angle α = −15◦ at atmospheric conditions and
at confining stress pc = 34.5 MPa. Because of its negligible effect on the cutting
process, gravity is neglected in this simulation. That is why the large horizontal
ribbon of debris occurs at atmospheric conditions. The force signals of both simu-
lations are shown in figures 6.5 and 6.6 for respectively the horizontal and vertical
direction.

Horizontal cutting force

6,000
pc = 0.1 MPa
pc = 34.5 MPa
4,000
Fc,x [N]

2,000

0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045
t [s]
6
Figure 6.5: Horizontal cutting force over time for both atmospheric and high pressure confinement
(34.5 MPa) with rake angle of -15◦ .

Vertical cutting force

pc = 0.1 MPa
2,000 pc = 34.5 MPa
Fc,y [N]

1,000

0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045
t [s]

Figure 6.6: Vertical cutting force over time for both atmospheric and high pressure confinement
(34.5 MPa) with rake angle of -15◦ . The tendency is that the tool will be ’pushed’ out of the rock.

The damage inflicted over time in both cutting processes is shown in figure 6.7.
More simulations are performed to investigate the behavior of the cutting process
6.3. Drilling 111

Total damage infliced

6,000
ph = 0.1 MPa
#brokenbonds [-]

ph = 34.5 MPa
4,000

2,000

0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045
t [s]

Figure 6.7: Total number of accumulated broken bonds over time for both atmospheric and high
pressure confinement (34.5 MPa) with rake angle of -15◦ .

with respect to the confining stress, for confining stresses ranging from 0.1 − 34.5
MPa and for both rake angles α = −15, −25◦ . The time averaged cutting forces,
both horizontal and vertical are respectively shown in figures 6.9 and 6.10. The
cutting forces as predicted by Lei and Kaitkay (2003) and the experimental results 6
of Kaitkay and Lei (2005) are also shown in these graphs. The total damage applied
over time for each of the simulations is shown in figure 6.11.
 112 6. Validation - Tool-Rock Interaction

t = 0.0 s t = 0.0 s

t = 0.00625 s t = 0.00625 s

t = 0.0125 s t = 0.0125 s

t = 0.01875 s t = 0.01875 s
6

t = 0.025 s t = 0.025 s

t = 0.03125 s t = 0.03125 s

t = 0.0375 s t = 0.0375 s

t = 0.04375 s t = 0.04375 s
(i) No confining stress. (r) At confining stress ph = 34.5 MPa.

Figure 6.8: Damage of cutting simulations with α = −15◦ . For damage, red corresponds with fully
disintegrated particles and blue corresponds with perfectly intact material. Note that only the top
layer where observable changes occur is shown in the figures
6.3. Drilling 113

6.3.3. Discussion
A significant difference in behavior is noted between low and high pressure condi-
tions. At low confining stresses, the debris generated behaves much like a granular
flow and separate small clusters and single particles move freely away from the cut-
ting process. At high confining stresses a continuous chip is generated and at some
extent this chip can curl up (balling as described by Zijsling (1987) and theoretically
calculated by Miedema and Zijsling (2012)). It seems that the curling up of the chip
in front of the tool gives no significant effect on the cutting force. Furthermore it
is shown that the friction area of the chip along the face of the bit changes over
time. Based on the number of experiments that are performed thus far it is not yet
possible to give a more in depth analysis on the curling behavior of the chip.
The type and amount of damage that is applied to the non-cut rock differs as
well when comparing low and high confining stresses. At low stresses, the total
damage to the rock is lower and some cracks that move deeper into the virgin rock
are shown. At high confining stresses the amount of damage that is applied to the
virgin rock is significantly larger and it shows little to no damage deeper into the
virgin rock. The highly damaged layer acts like a filter cake, as described in chapter
3.

Averaged horizontal cutting force

DEM-SP α = −15◦ 6
4,000 DEM-SPα = −25◦
LK sim α = −15◦
Fc,x [N]

LK sim α = −25◦
2,000 LK exp α = −15◦
LK exp α = −25◦

0
0 5 10 15 20 25 30 35
ph [MPa]

Figure 6.9: Time averaged horizontal cutting force with varying confining pressures.

In the graphs showing the cutting forces, although highly rapid fluctuations are
noticed, the mean cutting force over time does not differ significantly. The spikes in
these graphs are a result of the stress build until bonds break. The highest spikes
are a result of the discretisation, depending on where the tool touches on a specific
particle and whether it has some mobility whether such a spike can occur, as is
discussed in previous section 6.2.
The time averaged components of the cutting forces of the simulations in figure
6.9 and 6.10 show a similar trend to the drilling experiments of Kaitkay and Lei
(2005). Although the vertical cutting force significantly underestimates the cutting
force as measured in the experiments (approx 50-60 %). The cutting forces as
 114 6. Validation - Tool-Rock Interaction

Averaged vertical cutting force

DEM-SP α = −15◦
4,000 DEM-SPα = −25◦
LK sim α = −15◦
Fc,y [N]

LK sim α = −25◦
2,000 LK exp α = −15◦
LK exp α = −25◦

0
0 5 10 15 20 25 30 35
ph [MPa]

Figure 6.10: Time averaged vertical cutting force with varying confining pressures.

simulated with DEM-SP show closer resemblance with the experiments in horizontal
direction compared to the simulated method of Lei and Kaitkay (2003). In vertical
direction, no significant differences are noted with respect to the simulations of Lei
and Kaitkay (2003).
6
Total damage inflicted
8,000
α = −15◦
#brokenbonds [-]

6,000 α = −25◦

4,000

2,000

0
0 5 10 15 20 25 30 35
ph [MPa]

Figure 6.11: Total number of accumulated broken bonds over time, presented for varying confining
pressures.

The improvement shown by DEM-SP can be due to the higher level of detail in
the cutting zone, a layer of on average five particles thick is being cut in DEM-SP
while Lei and Kaitkay (2003) cut a layer of one particle thickness. Additionally it
might be that their chain method causes some unrealistic results, as is shown by
Mendoza Rizo (2013). It could as well be that both arguments are valid.
Only the cutting force at atmospheric conditions is significantly overestimated.
As mentioned before, this experiment is performed with a different setup, which
6.3. Drilling 115

might lead to the significantly different cutting force. Simulations with lower tensile
strength (σBT S = 10 MPa) at confining stress pc = 0 and 34.5 MPa did not result
in a significant decrease (< 10%) in cutting force that might explain the difference.
In the cutting theory of Miedema (2014) and in the experiments of Zijsling (1987)
this such a transition as depicted by Kaitkay and Lei (2005) might not be present
when all experiments are performed with the same setup.
Total damage over time that is applied in the cutting process shows a linear
trend with increasing confining stress for both rake angles. The trend corresponds
with expectations, that with increasing confining stress the cutting process goes
through the brittle-ductile transition.

6
 116 6. Validation - Tool-Rock Interaction

6.4. Dredging and (Deep) Seabed Mining

The DEM-SP with fluid pressure effects has been applied to the simulation of linear
rock cutting tests using a 2D geometry. Both the pore Peclet number and the water
depth (assuming both internal pore pressure and ambient pressure are initially in
equilibrium) are varied in the set of simulations. The rock specimen is discretized
using 96 991 cylindrical elements with a uniform particle size distribution with radii
0.2-0.4 mm. The cutting tool is discretized in 1050 straight line segments with a
maximum length of 0.4 mm per line segment. As a boundary condition, the position
of the particles at the bottom and the right end of the slab is fixed. For the fluid
counterpart, all boundary particles are subsequent to Dirichlet boundary (ambient
hydrostatic pressure). Except for the particles that are in contact with the cutting
tool, there a Neumann boundary is applied (no fluid flow through the cutting tool).
The geometry of the tooth and the rock properties that are used are based on
the experiments of Alvarez Grima et al. (2015). Linear rock cutting simulations are
carried out on a block with dimensions 0.1 and 0.35 m. The tooth is placed at a
cutting angle of 68 ◦ and has a clearance angle of 10 ◦ , is 95 mm in height and the
tip of the tool has a width of 21 mm. In the 2D simulations, the cutting force is
calculated per unit thickness. Therefore the cutting force is scaled with the width
of the cutting tool. A fixed timestep of 11.25 ns is used and the data is stored every
1250 timesteps, resulting in a sampling frequency of 7.1 kHz.
6 Table 6.9: Micro (input) and macro (output) parameters used in the rock cutting simulations.

Input parameters Output parameters

Micro parameters (DEM)
kn 10 GPa σU CS 9.89 MPa
ks 2.5 GPa σBT S 1.47 MPa
Tn 10 kNm-1 E 8.03 GPa
Ts 20 kNm-1 ν 0.28
µ 1
R 0.0002-0.0004 m
αd 0.7
ρ 2800 kgm-3

Fluid parameters (SP)

O 10−16 − 10−14 m2


κ
n 0.18
h 0.0015 m
M 2 GPa
µf luid 0.001 Pas
αef f 0.5

The micro and macro parameters used are shown in table 6.9. The properties of
the rock are chosen such that they are well within the range of the measured rock
properties in the experiments of Alvarez Grima et al. (2015). These macroscopic
6.4. Dredging and (Deep) Seabed Mining 117

parameters are obtained by calibration of the micro-parameters on both UCS and

BTS tests. Several simulations have been performed in which the permeability and
the ambient pressure are varied. Here the permeability is used to mimic a change in
cutting velocity in the simulation. Decreasing the cutting velocity in the simulation
would result in a too large computational expense. Note that for the weak type
of rock that is modeled in this study, the contact bond model is sufficient. When
simulating tougher rock types, it is advised to extend the contact bond model to
the parallel bond model (Potyondy and Cundall, 2004).

6.4.1. Results

(a) t = 0 s (b) t = 0.07 s 6

(c) t = 0.14 s (d) t = 0.21 s

(e) t = 0.28 s

Figure 6.12: Damage for rock cutting simulation at atmospheric pressure, ph = 0.1 MPa, and
κ = 10−14 m2 .
 118 6. Validation - Tool-Rock Interaction

(a) t = 0 s (b) t = 0.07 s

(c) t = 0.14 s (d) t = 0.21 s

6

(e) t = 0.28 s

Figure 6.13: Damage for rock cutting simulation at ph = 10.1 MPa and κ = 10−14 m2 .

Image sequences of rock cutting simulations at both atmospheric and 10.1 MPa
ambient pressure (comparable to 1 km water depth) are presented in figures 6.12
and 6.13. In both image sequences, the depicted damage is defined as

#intact bonds
D =1− (6.1)
#initial bonds

so that with D = 0 all bonds of the particle are still intact and with D = 1 all bonds
of that particle are broken.
The total cutting force with respect to time is shown in figure 6.14. The total force
is calculated by summation of all force vectors acting on the surface of the tool.
The averaged total cutting force components for various ambient pressures and
6.4. Dredging and (Deep) Seabed Mining 119

permeabilities are shown in figures 6.15 and 6.16. Start-up effects are omitted by
determining the average cutting forces in the range of 0.07-0.28 seconds. Although it
is shown in figures 6.12b and 6.13b that the front of the tooth is not yet completely
covered with cut material, it has an insignificant effect on the total cutting force.

Cutting force over time

120
ph = 0.1 MPa κ = 10−14 m2
100 ph = 0.1 MPa κ = 10−16 m2
ph = 18.1 MPa κ = 10−14 m2
80 ph = 18.1 MPa κ = 10−16 m2
Fh [kN ]

60

40

20

0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 0.26 0.28 0.3 6
t [s]

Figure 6.14: Horizontal cutting force over time for atmospheric and hyperbaric conditions.
 120 6. Validation - Tool-Rock Interaction

Horizontal cutting force at varying ambient pressure

κ = 10−16 m2
40 κ = 10−15 m2
Fh [kN ]

κ = 10−14 m2

20

105 106 107

ph [P a]

Figure 6.15: Simulated average horizontal cutting force with respect to varying ambient pressures.
Cutting velocity is set at 1 m/s.

Vertical cutting force at varying ambient pressure

20
κ = 10−16 m2
15 κ = 10−15 m2
Fv [kN ]

κ = 10−14 m2
10

105 106 107

ph [P a]

Figure 6.16: Simulated averaged vertical cutting force with respect to varying ambient pressures.
Cutting velocity is set at 1 m/s.
6.4. Dredging and (Deep) Seabed Mining 121

Figure 6.17 shows a comparison for the total amount of damage incurred over
time for the most diverse cases, i.e. atmospheric vs hyperbaric and low permeability
vs high permeability. A comprehensive comparison of the total cumulative damage
incurred at the end of all simulations is shown in figure 6.18.

Total number of broken bonds over time

·103
40 ph = 0.1 MPa κ = 10−14 m2
ph = 0.1 MPa κ = 10−16 m2
ph = 18.1 MPa κ = 10−14 m2
AE [−]

20 ph = 18.1 MPa κ = 10−16 m2

0
0 0.1 0.2
t [s]

Figure 6.17: The number of broken bonds over time for atmospheric and hyperbaric conditions.

Cumulative accoustic emissions at t = 0.28s and varying ambient pressure 6

·103
40 κ = 10−16 m2
κ = 10−15 m2
35
κ = 10−14 m2
AE [−]

30

25

105 106 107

ph [P a]

Figure 6.18: Number of broken bonds at the end of the simulations. Hydrostatic pressure and
permeability of the rock are varied.

Two different regimes can be distinguished, being tensile dominated and shear
dominated cutting. The transition is observed in both the cutting force components
and the amount of damage inflicted to the rock, see figures 6.15, 6.16 and 6.18. For
higher permeabilities, this transition occurs at approximately ph = 3 MPa, while
for the less permeable rocks this transitions seems to occur at ph = 1.5 MPa. The
first regime occurs at low ambient pressures, the rock cutting process is dominated
by the occurrence of large chips that are created by long macroscopic tensile cracks.
The second regime occurs at higher ambient pressures (ph > 2 − 3σBT S ) is dom-
inated by shear failures that result in smaller chips and further fragmentation of
 122 6. Validation - Tool-Rock Interaction

the cut material. A similar effect can be noticed in figures 6.14 and 6.17, where
the frequency of stress build-up and release through brittle failure increases with
increasing ambient pressure. The transition from tensile dominated to a shear dom-
inated cutting seems to correlate with the BTS value of the simulated rock, i.e. the
transition is at a hydrostatic pressure with a magnitude of 100-200% of the BTS
value depending on the rocks permeability.
The simulations show qualitatively that the size of the crushed zone increases
with increasing water depth. This can be observed when comparing figures 6.12d and
6.13d. Pore collapse and particle crushing are failure modes that can occur in the
crushed zone van Kesteren (1995), while these failure phenomena are not considered
in the current DEM modeling approach. The bonds between the particles can fail
in either tensile or shear mode, while the particles themselves cannot fail. In this
research, particle crushing is considered to be of lower significance on the cutting
process compared to the fluid pressure effect, see section 2.3.2.

6
1
2
Figure 6.19: Location of the pressure measurements with respect to the shape of the tool. The
area over which the pore pressures are measures is a square with 2x2 mm.

More interestingly, with the use of simulation techniques it is possible to inves-

tigate effects that are difficult to measure, like the pore pressure at the tip of the
tool. The averaged pore pressure over time at the tip of the tool is measured at
two different positions with respect to the tool. The position of these measuring
areas are shown in figure 6.19. Measurements of these pressure sensors are shown
in figures 6.20 and 6.21. Note that there are no particles present in the first few
time steps of the simulations to determine the excess pore pressure in the sensor
regions. For the set of simulations, the time averaged excess pore pressures with re-
spect to the hydrostatic pressure are shown in figures 6.22 and 6.23. An impression
of an instantaneous spatial pore pressure distribution during a cutting simulation
performed at ph = 18.1 MPa and κ = 10−14 m2 is shown in figure 6.24.
6.4. Dredging and (Deep) Seabed Mining 123

Excess pore pressure at sensor 1 over time

40
ph = 0.1 MPa κ = 10−14 m2
p1 − ph [M P a]

ph = 0.1 MPa κ = 10−16 m2

20 ph = 18.1 MPa κ = 10−14 m2
ph = 18.1 MPa κ = 10−16 m2

0 5 · 10−2 0.1 0.15 0.2 0.25 0.3

t [s]

Figure 6.20: Excess pore pressure over time, measured at the front of the tip of the tool (position
1, as depicted in figure 6.19). Note that there are no particles present in the first few time steps
to measure a pore pressure.

Excess pore pressure at sensor 2 over time

10
ph = 0.1 MPa κ = 10−14 m2
p2 − ph [M P a]

ph = 0.1 MPa κ = 10−16 m2

0 ph = 18.1 MPa κ = 10−14 m2
ph = 18.1 MPa κ = 10−16 m2
−10

−20
0 5 · 10−2 0.1 0.15 0.2 0.25 0.3
t [s]

Figure 6.21: Excess pore pressure over time, measured at the front of the tip of the tool (position
2, as depicted in figure 6.19). Note that there are no particles present in the first few time steps
to measure a pore pressure.
 124 6. Validation - Tool-Rock Interaction

The simulations show that in the crushed zone (at pressure sensor 1, figure
6.22), the averaged pore pressure increase with respect to the hydrostatic pressure is
constant. To what extent this pore pressure increase occurs, seems to depend on the
permeability of the rock, i.e. a lower permeability leads to a higher pressure increase
in the crushed zone. Below the tip of the tool a different trend is observable (see
figure 6.23). At low hydrostatic pressures an increase in pore pressure is measured
while at large hydrostatic pressures, the average pore pressure below the tool seems
to decrease. It is not yet fully understood why this happens.

Average pore pressure at pos 1

·107
2 κ = 10−16 m2
κ = 10−15 m2
p1 − ph [P a]

1.5 κ = 10−14 m2

0.5
105 106 107
ph [P a]
6
Figure 6.22: Averaged pore pressure with respect to hydrostatic pressure, measured at the front of
the tip of the tool (position 1, as depicted in figure 6.19).

Average pore pressure at pos 2

·106
κ = 10−16 m2
κ = 10−15 m2
p2 − ph [P a]

0
κ = 10−14 m2

−5

105 106 107

ph [P a]

Figure 6.23: Averaged pore pressure with respect to hydrostatic pressure, measured at bottom of
the tip of the tool (position 2, as depicted in figure 6.19).

A possible explanation for the decrease in pore pressure below the tool might
be that in front of the tool due to dilation in the shear zone, fluid has to flow from
the surrounding regions. As depicted in figure 6.24, a lower pore pressure region
is present directly beneath the tip of the tool. Due to differences in the maximum
6.4. Dredging and (Deep) Seabed Mining 125

achievable pressure gradient (limited by ph ), the amount of fluid flow from this region
can differ. Interesting to mention as well, even at ph = 18 MPa, the simulations
show that the rapid deformations applied to the rock still might result in cavitation
of the pore fluid, as is depicted in figure 6.24.

Shear zone (dilative)

Crushed zone

Figure 6.24: Pore pressure distribution in simulation with ph = 18.1 MPa, κ = 10−14 m2 at t=0.14
s. Note that cavitation occurs (dark blue)
 126 6. Validation - Tool-Rock Interaction

6.4.2. Comparison of numerical and experimental results

Although Alvarez Grima et al. (2015) have used multiple blocks with a slight de-
viation in the material properties, for sake of simplicity only one set of macro-
parameters are used in the simulations, see table 6.10 for an overview of the com-
pared data. The simulated macro-properties are well within the range of the mea-
sured properties in the experiments, except that the tensile strength is slightly
overestimated in the simulations. In the current implementation, the maximum
achievable brittleness ratio (given by σσUBTCSS ) is limited to approximately 7. Due to
the wide range of hydrostatic pressures (up to 2σU CS ) it is assumed that a match-
ing compressive strength is of higher importance compared to a matching tensile
strength of the rock. For that reason we favored the UCS over the BTS value to
match the experiments. Figure 6.25 shows a comparison for the averaged cutting
forces with respect to water depth for both the experiments and simulations. Also
the range of the minimum and maximum forces measured in the experiments is
shown to give an impression of the fluctuations with respect to the average cutting
force.

Table 6.10: Comparison of material properties and experimental setup of Alvarez Grima et al.
(2015) and the simulations.

Test σU CS E ν σBT S κ n vc ph Favg

6 no. [MPa] [GPa] [-] [MPa] [m2 ] [%] [ms-1 ] [MPa] [kN]
1 7.92 5.95 0.31 0.88 3.1E-13 37.86 0.188 0 7.22
2 7.92 5.95 0.31 0.88 3.1E-13 37.86 0.178 18 9.25
3 7.92 5.95 0.31 0.88 3.1E-13 37.86 0.200 1.5 10.42
4 8.75 7.53 0.25 1.09 8.5E-14 34.64 1.826 0 8.09
5 8.75 7.53 0.25 1.09 8.5E-14 34.64 1.717 1.5 11.17
6 8.75 7.53 0.25 1.09 8.5E-14 34.64 1.740 3 12.23
7 8.75 7.53 0.25 1.09 8.5E-14 34.64 1.702 6 13.19
8 9.29 5.89 0.27 1.15 1.4E-14 33.17 1.577 18 20.70
9 10.62 8.32 0.23 1.05 2.8E-14 31.66 0.618 18 22.72
10 10.64 9.01 0.27 1.13 2.2E-15 33.92 0.010 18 4.94
11 8.86 8.20 0.31 0.86 1.5E-14 35.12 0.017 0 4.72
12 8.86 8.20 0.31 0.86 1.5E-14 35.12 0.202 3 11.36
13 8.86 8.20 0.31 0.86 1.5E-14 35.12 0.207 6 11.29
14 10.54 9.98 0.33 x 3.4E-16 35.89 1.238 18 12.74
15 10.54 9.98 0.33 x x x 1.188 6 10.90
DEM-SP
9.89 8.03 0.28 1.47 varies 17.86 1.000 varies varies

The results obtained, as shown in figure 6.25, indicate that the most dominant
effects regarding pore fluid effects, hydrostatic pressure and deformation rate, are
well predicted. Unfortunately, no other clear trends can be distinguished based on
the data presented in Alvarez Grima et al. (2015). In the case of lower hydrostatic
6.4. Dredging and (Deep) Seabed Mining 127

Comparison of Fh simulations vs experiments

·104
5 exp v = 0.2m/s κ = 3 · 10−13 − 10−16 m2
exp v = 2m/s κ = 3 · 10−13 − 10−16 m2
sim v = 1m/s κ = 10−16 m2
4
sim v = 1m/s κ = 10−15 m2
sim v = 1m/s κ = 10−14 m2
3 Miedema
Fh [N ]

0
105 106 107
ph [P a]
6
Figure 6.25: Comparison between experiments and simulations. Plot shows averaged cutting
force with respect to hydrostatic pressure. The errorbars of the experiments correspond with the
minimum and maximum measured cutting forces.

pressure (up to ph = 3.1 MPa), the cutting force is slightly underestimated compared
to the region at higher hydrostatic pressures. Most likely these phenomena are
resultant to simulating a 3D process with a 2D method.

6.4.3. Discussion
At lower ph , the side break out angle results in a larger area being cut than the width
of the tool, while for higher ph the side break out angle is almost 90◦ resulting in
an almost perfect box cut, as observed by Alvarez Grima et al. (2015). Another
effect caused by the 2D-3D conversion is where the cut material will move to. In
2D, the cutting process mimics an infinitely wide tool (tool width is used to scale
the forces to practical applications), giving the debris that does not immediately
move away from the tool has to move along and over the tool. A reason for the
debris to ’stick’ together is the fluid pressure coupling. Particles that get to distant
from another will mimic dilation, resulting in a large local under pressure, which will
be counteracted upon by the surrounding fluid pressure. In the current model, the
permeability is assumed to be uniform and constant over time. Overestimation of
dilation effects might be reduced when the permeability can change with changing
porosity.
 128 6. Validation - Tool-Rock Interaction

In 3D, the debris will be able to move to the side of the tool, reducing the
amount of friction that the debris will exert on the tool. Another side effect of
working with 2D simulations compared to 3D simulations is that most likely the
normal component of the cutting force is underestimated (Rojek et al., 2011).

6
6.5. Testcase: rotational cutting 129

6.5. Testcase: rotational cutting

In this section, simulations based on the geometry and motion of a real cutterhead
will be discussed.

6.5.1. Simulation setup

The input and output parameters of the simulated rock are shown in table 6.11.
In this study, the parameters were chosen to represent a tight sandstone. On this
rock, several cutting simulations are performed with a circular cutting motion, in
which the cutter head moves with both an angular and a swing velocity, respectively
with 1.62 rads-1 and 0.2 ms-1 . The combination of the angular and swing velocities
result in a cutting velocity of 1 ms-1 . The tool is modeled as a rigid body and its
movement is prescribed. The water depth has been varied between 0 and 30 m and
therefore the initial pore and hydrostatic pressures are respectively set to 0.1 and
0.4 MPa. In these simulations, a pick point positioned at a 26◦ angle (with respect
to the radial direction) is used. In the circular cutting motion, the layer thickness
starts at 0 m and will gradually increase to to 0.20 m. The geometry of the rock
cutting experiment is shown in figure 6.26. No displacement boundaries are placed
at the bottom and east side of the rock specimen by fully restricting the movement
of the particles at these boundaries. Although the cutting velocity is lower than in
common practice and the initial rock geometry is intact, these conditions are chosen
for academic purposes. 6
Table 6.11: Input, output and geometrical parameters of simulated cutting experiment

Input parameters Simulation

Kn 20 GPa σU CS 20 MPa
Ks 4 GPa σBT S 3 MPa
Tn 100 kNm-1 E 5 GPa
Ts 100 kNm-1 vc 0.2 ms-1
µgrain 0.8 ω 1.62 rads-1
R 1 - 1.5 mm Tool angle 64 ◦
α 0.5 Water depth 0-30 m
κ 1 · 10−17 m-2 Tool width 0.02 cm
n 0.18
ρgrain 2800 kgm-3
h 0.005 m
M 2 GPa
µf luid 0.001 Pas
µtool−rock 0.5

6.5.2. Simulated results

To give an impression of the cutting process, an image sequence of the simulation
with water depth of 30 m is given in figure 6.27. This image sequence depicts
 130 6. Validation - Tool-Rock Interaction

250 vc
320

700

760
150

1400

Figure 6.26: Model geometry for simulation of rock cutting with one pick of a dredge cutter head

the damage that is applied to the rock over time, with damage defined as D =
#intact bonds
1− # initial bonds
, where D = 0 corresponds with perfectly intact rock and D = 1
with fully disintegrated rock, all the bonds at that particle are broken. The pore
pressure distribution during cutting is shown in figure 6.28. Note that in this figure
a cavitating zone occurs deeper inside the rock, which is caused by the opening
6 of tensile crack. At the same instance in time, an impression of the solid stresses,
depicted as the Von Mises stress, is shown in figure 6.29.
In 2D, the simulations automatically adopt a unit thickness of the rock specimen.
The thickness of the specimen has been scaled to the width of the tool (in this case
0.02 m), all forces and energies are scaled accordingly. In the simulations, a sampling
frequency of 1 kHz is used. To allow for a better analysis, a moving average filter
of 25 samples wide is used.
The total cutting force is calculated by integrating all resultant tool-grain contact
forces over the whole area of the cutter tool, in both x- and y-components. The
magnitude of the total cutting force is shown in figure 6.30. For each tool-grain
interaction, the work delivered by the tool can be estimated by
Z
E = mi |~vlt |dt = mi |~vlt |∆t (6.2)

with particle mass mi , local tool velocity vlt and timestep ∆t. Note that the work
delivered by the tool in the interaction is corrected for the reversible energies of the
grain (i.e. kinetic, normal strain and shear strain potential energy). The cumulative
total work delivered by the tool is calculated by integrating all tool-grain interactions
over the tool surface and in time. Figure 6.31 shows the cumulative total energy
and its decomposition in local normal and shear directions.
The tip of the tool is especially of interest, since most of the work is done by
this area. The tip has been analyzed for two small sections, the wear flat and the
front of the tip. Both sections are 4 mm in length. These areas are shown in figure
6.32. Figure 6.33 shows the total forces that are applied by the tooltip regions.
The direction of the force applied by the tooltip regions with respect to the tooltip
6.5. Testcase: rotational cutting 131

t=0s t = 0.15 s

t = 0.3 s t = 0.45 s
6

t = 0.6 s t = 0.75 s

Figure 6.27: Damage occuring in simulation at 0 meter water depth (but still fully submerged,
ph = 0.1 MPa). Red corresponds with fully damaged particles (of which all bonds are broken) and
blue with intact particles.

surface area is shown in figure 6.34 (0◦ corresponds with pure shear and thus 90◦
is perpendicular to the surface). Beneath and in front of the cutting tool, pore
pressures tend to increase due to the compression of the rock (crushed zone). The
averaged pore pressures directly below the tool are shown in figures 6.35 and 6.36,
note that it only shows the pore pressure under the tip of the tool and not in front,
most of the time there are too few particles in front of the tool to measure this.
The total acoustic emissions (AE) in the simulation are shown in figure 6.37. The
 132 6. Validation - Tool-Rock Interaction

Figure 6.28: Pore pressures in simulation at Figure 6.29: Von Mises stress contributions
ph = 0.1 MPa at 0.45 s of the solids at ph = 0.1 MPa at 0.45 s

Work required in cutting process

Total cutting force ·106
·103 100
100 pth = 0.1 MPa
80 pth = 0.4 MPa
80 psh = 0.1 MPa
60 psh = 0.4 MPa
pnh = 0.1 MPa
E [J]

60
|F̄c | [N ]

6 40
pnh = 0.4 MPa
40
20
20 ph = 0.1 MPa
ph = 0.4 MPa 0
0 0 0.2 0.4 0.6 0.8 1
0 0.2 0.4 0.6 0.8 1
Time [s]
Time [s]

Figure 6.31: Total work required in cutting pro-

Figure 6.30: Magnitude of the cutting force in- cess. Superscripts t ,s ,n respectively give total
tegrated over the complete tool work spent and it decomposition in shear and
normal directions.

cumulative AE can be used as a means of estimating and comparing the amount of

damage that has been induced to the simulated rock.

6.5.3. Discussion
Based on the total force applied by the cutting tool as in figure 6.30, three regimes
can be distinguished:

1. Pseudo-ductile cutting (0 < t < 0.15): Here defined as the cutting process is
dominated by complete fragmentation of the bonded particles into a granular
6.5. Testcase: rotational cutting 133

2
1
Figure 6.32: Detail of the tool tip. Area 1 corresponds with the wear flat and area 2 with the front
of the tool tip.

Force magnitude at the tool tip Angle of force vectors at tooltip

·103
Bottom 80
15 Front
Stick
60 Slip
φ [◦ ]
F̄c [N ]

10
6
40
5
Bottom
20 Front
0 0 0.2 0.4 0.6 0.8 1
0 0.2 0.4 0.6 0.8 1
Time [s] Time [s]

Figure 6.34: Angle of the cutting force with re-

Figure 6.33: Magnitude of the cutting forces in-
spect to the faces of the tip of the tool, as shown
tegrated over tool tip area, as shown in figure
in figure 6.32. The dashed line corresponds with
6.32
the stick-slip criterion

material, figure 6.27b shows no chips.

2. Brittle cutting (0.15 < t < 0.6): Within this range, chips can break close to
the surface of the rock. It is noticed that the size of the chips increases with
increasing cutting depth, as can be seen in figures 6.27c and 6.27d.

3. Edge chipping: The overburden of the intact rock is not sufficiently strong to
further constrain the cutting process and a large chunk of material will be cut
loose. This boundary effect causes a significant reduction in the cutting force.
 134 6. Validation - Tool-Rock Interaction

Pore pressure below the tool tip (1) Pore pressure in front of the tool tip (2)
4
ph = 0.1 MPa ph = 0.1 MPa
3 ph = 0.4 MPa ph = 0.4 MPa
3
ppore [M P a]

ppore [M P a]
2
2

1 1

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Time [s] Time [s]

Figure 6.35: Averaged pore pressure distribution Figure 6.36: Averaged pore pressure distribution
below the tip of the tool, measurement area is in front of the tip of the tool, measurement area
moving along with the tool is moving along with the tool
6
The cutting force does not drop to 0 due to the presence of a thin layer of rock
that still has to be cut.

It is worth mentioning that the forces (and thus work delivered) in the various
regimes have a different scaling effect when considering a 3D cutting process. In the
pseudo-ductile cutting regime, the volume of rock that is affected grows spherically
with respect to the layer thickness. Because of the complete destruction of the
structure of the rock in this region, the energy and thus the force required to cut
the material will scale with t3c . On the other hand, in the chipping regime with
increasing volume, the surface of the cracks will scale with surface area and thus the
required energy to cut the rock will scale with crack area (although fragmentation
might occur), so the required energy will scale approximately with t2c . In practice it
will be slightly higher, because for larger tc fragmentation of the chip might occur.
When comparing the total energy required to cut through the rock, an increase
of 30 meter of water depth compared to atmospheric pressure results in a 5 % in-
crease of required energy. To allow for an objective comparison based on the same
amount of rock removal and thus the contribution of the edge chipping effect is not
considered (this would only be valid when many simulations are considered so that
an averaged excavated production can be used). Based on the tensile strength of
the simulated rock, a maximum energy increase of 10% could be expected (30 meter
water depth difference is equal to 10% of the tensile strength) in the regime where
tensile failure dominates the cutting process. However, that is based on the limiting
situation that the process is sufficiently fast that the cracks instantaneously open,
6.5. Testcase: rotational cutting 135

Cumulative
3
accoustic emissions
·10

ph = 0.1 MPa
15
ph = 0.4 MPa

AE [−] 10

0
0 0.2 0.4 0.6 0.8 1

T ime [s]

Figure 6.37: Accoustic emissions during cutting simulation at various water depths.

causing the lack of fluid in the cracks, resulting in a local underpressure which acts
as an extra toughness for tensile failure. A similar analysis for the shear strength 6
cannot be done without more detailed knowledge of the micro-structure. The effect
of pore volume change in shear failure depends strongly on the micro-structure of
the rock, in shear failure, rock can behave dilatant, compactant or no net change at
all. The cutting regimes and the difference in required energy are also noted when
comparing the amount of damage that is induced by the cutting process, as is shown
in figure 6.37.

Based on the complete cutting tool, approximately half of the energy that is
applied by the cutting tool is spent on friction in the scratching and brittle cutting
regimes. When the edge chipping effect starts to dominate the cutting process, the
cumulative energy spent on tool-grain friction will reduce to 30-35 % of the total
energy spent in the cutting process. The highest intensity of the energy spent in
the cutting process occurs at the tip of the tool (approximately 25 %). In figure
6.33 a saw-tooth cycle is clearly shown (it is also observed in the pore pressure near
the tool tip, see figure 6.28). This corresponds with the stress build up, creating a
crushed zone. The drops in the force correspond with the growing of macro-cracks.
The occurrence of these cracks and the resulting chips is less clear in the image
sequence in figure 6.27, most of the chips that occur are broken down into smaller
pieces because these chips seem to get stuck between the tool and the still remaining
rock body. The graph also shows a range where no angle was determined, in that
range that area of the tool is not touching any rock, and thus no force is exerted.
The sudden drop in the angle for the bottom of the tool tip is caused by the start
of the edge chipping regime and in that range the front of the tool dominates the
cutting process.
 136 6. Validation - Tool-Rock Interaction

The friction factor of the tool-grain interactions gives a critical friction angle
of 26.6◦ . This implies that for any angle of the cutting force in figure 6.34 larger
than 90 − 26.6 = 63.4◦ , the shear force at the tool surface is not large enough to
generate slip with the given normal force. This occurs at the tip of the tool in the
period of 0.4-0.6 sec. During this period, the interacting grains actually stick to the
tool, causing the grains to roll along the surface of the tool. Note that the effect of
rolling might be overestimated, because in the simulations perfect discs are used as
particles and these have a lower rolling resistance compared to real grains. It could
well be possible that due to the rolling effect, wear of the tool mainly occurs in the
scratching regime.
As is shown in figure 6.29, the solid stresses near the tip of the tool can easily
exceed 200 MPa. With these high stresses, the individual grains might fail them-
selves (which is not considered as a failure mechanism in the simulations). One
needs to be careful with interpreting this phenomena, because this might as well be
a scaling size effect of the simulation. In these simulations, particle diameters are
in the range of 2-3 mm, while in real sedimentary rock (e.g. sandstone, limestone)
the typical grain sizes are smaller than that. Complementary to that, smaller grains
can withstand a higher stress before they will fail. For example, the critical nominal
stress that sand particles can withstand scales σc,nom ∝ dp−0.25 (Brzesowsky et al.
(2011)).
The pore pressure build-up beneath the pick point in the range of 0-30 meter
6 water depth does not differ significantly. A linear increasing trend can be noticed
in the scratching regime, in the brittle chipping regime a saw tooth like behavior
is observed. The peaks in pressure build up do not differ significantly with varying
water depth. However, the total time averaged pore pressure below and in front of
the tip of the tool are respectively 9.5% and 4.5% more at 30 meter water depth
compared to the slightly submerged case (atmospheric pressure).
One needs to be careful when interpreting the results of these simulations. The
results presented in this dissertation are based on a fully two-dimensional simulation,
while in practice the rock cutting process is a fully three-dimensional problem. In
3D, chips are expected to show a different breakout pattern and shape compared to
2D. Besides that, the effect of friction of the rock along the cutting tool will be more
realistic. That is, in 2D, the cut material cannot move in any other direction than
along the movement of the tool (the 2D process can be compared with an infinite
wide cutting tool), while in 3D the cut material may fall next to the equipment,
which will significantly reduce the total force applied by the tool. In the near future,
the simulation software will be extended into 3D to improve on those issues. Besides
the estimation of friction along the tool, the force scaling effects in the brittle and
pseudo-ductile regimes will be improved.

6.6. Conclusions
The tile cutting simulations show that if sufficient detail is applied, the DEM is
capable of properly simulating the chip forming process. Furthermore, simulations
of rock cutting processes like drilling and dredging show good resemblance with
experiments. The effect of a hydrostatic pressure is properly included.
6.6. Conclusions 137

Simulations of rock cutting for both drilling and deep sea mining applications
show that the cutting force increases with increasing hydrostatic pressure. This
observation is similar to what is observed in experiments. The amount of damage
applied to the rock increases as well and the size of the crushed zone increases. In
drilling the chip balling and the presence of a filter cake are noted at high pressure.
Furthermore, the trend of an increase of the cutting velocity on the cutting process
for deep sea mining applications corresponds with the theory as well.
The results obtained with DEM-SP show better resemblance with the deep sea
mining cutting experiments than the model described in Alvarez Grima et al. (2015).
The results obtained in DEM-SP are based on the average cutting forces, as are the
experimental results that they presented. However, the model that they present is
based on peak forces and is compared with time averaged forces.
Research on equipment with cutting tools mounted on an axis not perpendicular
to the rock surface, can be improved by simulating the cutting process based on a
rotating cutter, instead of linear cutting. Obviously the assumption of a constant
cutting depth is not valid in such experiments, but more assumptions do not hold
as well, i.e. movement of the cutting tool is no longer parallel to the rock bed, a
wear flat might induce a secondary cutting process and a stick-slip transition of the
cut material along the tool is observed.

Recommendations
Although the 2D simulations show that the current methodology works well, signif- 6
icant improvements are expected when the method is extended towards 3D. Espe-
cially improvements on the effective width of the tool (breakout angle), increased
resistance of the overburden to withhold the crushed zone and the flow of the cut
material no longer has to move over the tool are expected.
All simulations show results based on a single cut in intact rock. After a tool
has passed by, it still leaves some damaged rock behind. It will be interesting to
see to what extend the specific energy will decrease due to the pre-damage that is
applied by the previous tool.
The data that is (publicly) available to validate the rock cutting simulations is
rather limited. The experiments that are available are often limited in the level of
detail to properly model the macro-scopic mechanical properties of the rock, e.g.
Kaitkay and Lei (2005).
Furthermore, the rock cutting simulations and the validation through experi-
ments are all based on sharp tools. For industrial practice it would be more of
interest to see how the rock cutting process is affected when wear of the tool is
taken into consideration as well.
 Conclusions and
7
Recommendations
"If we knew what it was we were doing, it would not be called research, would it?"
Albert Einstein

In this chapter the overall conclusions and recommendations of this project are pre-
sented.

139
 140 7. Conclusions and Recommendations

To recap, in the introduction, the aim of this study was defined as

• To describe the physical phenomena that occur during the cutting of saturated
rock, with an emphasis on the fluid pressure effects.
• To develop and implement a physical and mathematical model to predict the
rock cutting process, in which the hydrostatic and pore pressure effects are
incorporated.

7.1. Conclusions
Based on a literature survey, it is found that this is the first time the DEM and SP are
combined to model saturated rock. Mechanics of the solids (rock) is modeled with
DEM, with elastic perfect brittle contact bond models for intact bonds and Coulomb
friction model as collision model. Fluid is modeled as a pore pressure diffusion
process and is solved with a smoothed particle method. The smoothed particle
technique is also used to interpolate the discrete nature of DEM towards a continuum
field. Coupling between the rock and fluid phase is achieved by calculating the
volumetric strain of the rock, which acts as a source term in the pore pressure
diffusion process. The resultant fluid pressure gradient is again applied as a particle
force in DEM.

Mechanics of Saturated Rock

The presence of a pore fluid can have significant influence on the rock mechanics,
through either

7 • Physico-chemical effects
• Drainage related effects
Within the scope of this thesis, only drainage related effects are considered. The
most profound drainage related mechanisms are:
• Dilation hardening
• Compaction weakening
Both mechanisms can occur and the behavior depends on the local conditions. Fur-
thermore, the effect of dilation hardening can be limited by either cavitation of the
pore fluid or failure of the grains.

DEM-SP
The extension of DEM with the smoothed particle approach to model pore pressure
effects works properly. Thus far, it is the first approach that allows for simulation
of saturated rock. Simulated results match well with strain rate effects on saturated
rock (shale).

The smoothed particle approach can be used to apply adaptive boundary con-
ditions on a particle assembly, independently of the shape and orientation of the
7.2. Recommendations 141

boundary.

The effect of a high hydrostatic pressure is well captured in simulations of drilling

experiments.

Both effects of cutting velocity and hydrostatic pressure are simulated properly.
In the case of hyperbaric rock cutting experiments with positive rake angle, the
DEM-SP shows better correspondence than the presented analytical models.

Rock Cutting Process

Rock cutting simulations for tile cutting, drilling, deep sea mining and dredging
have been performed successfully. They show a good comparison with experiment
data found in literature.

Simulations with DEM-SP of rock cutting with a positive rake angle show that
with increasing hydrostatic pressure a transition occurs from tensile dominated to-
wards shear dominated cutting. The transition seems to correlate with the tensile
strength of the rock sample.

The size of the crushed zone, the total applied damage and the total required
energy to cut the rock all increase with water depth.

The use of linear rock cutting models is a too strong simplification compared
to the cutting motion of a tooth on a cutter head. Especially when tool wear is of
interest. 7
Both the set of experiments cited in this dissertation as well as the simulated
results of DEM-SP show that the use of the compressive strength of the rock only
is insufficient to properly model the rock cutting process.

7.2. Recommendations
Mechanics of Saturated Rock
The rock cutting process covers the transition of an intact rock, through fragmen-
tation towards a granular medium. In rock mechanics, and especially in the case of
saturated rock, most of the research is concerned up to failure of the rock. More re-
search is needed to cover the transition from intact rock towards a granular medium.

DEM-SP
Although the methodology works well for 2D cases, the rock cutting process in itself
is a full 3D process, which cannot be fully captured in 2D (plain stress). Extension
towards 3D should improve the simulated results.
 142 7. Conclusions and Recommendations

The Smoothed Particle method can be used to interpolate DEM properties to-
wards a continuum and solve a range of a continuum equations as well. This property
should be used to model other physical processes that might be of interest on a con-
tinuum level, such as heat generation and dissipation (conduction), tool wear rate,
etc.

The results of the DEM-SP depend on the particle assembly that is used. Thus
far it is assumed that time averaging is sufficient to compare with experiments.
Running similar simulations but with a new particle assembly might give different
results. Similar to experiments, it will increase the confidence in the simulated re-
sults when multiple simulations are performed at the same settings with different
particle assemblies.

In this thesis, the contact bond model and circular particles are used. Due to
the simplicity of these models, the range of practical rocks that can be simulated is
limited, especially with respect to high strength rock (> 100 MPa) and high ratios
of compressive over tensile strength (> 8). Various methods could be used to extend
the applicability of the DEM-SP for a wider range of mechanical rock properties.
Several suggestions to improve the methodology are presented in chapter 4.

Rock Cutting Process

In the current model, the rock cutting process is considered as a 2D process. As a
result, most of the debris generated moves over the tool surface, effectively mimick-
7 ing an infinitely wide tool. In 3D, the debris can move sideways as well and thus
the material cut does not have to move along the tool surface. Furthermore, it is
expected that the accuracy of the vertical cutting force will increase as well. Other
effects that cannot be properly simulated in 2D can be investigated as well, e.g. tool
spacing, worn tools (wear flat).

Thus far, particle crushing is neglected in the simulations. Incorporating particle

crushing can aid to improve the simulations where heterogeneities become signifi-
cant, e.g. deep drilling, arctic clay (ice lenses).

The simulations that are described in this thesis all consider intact rock as a
starting material, while in practice this will not be the case. Consecutive tool pas-
sages can give insight on interaction between multiple tools.

Thus far, the publicly available rock cutting tests as well as most of the rock
cutting theories assume that the cutting motion is parallel to the rock surface and
that the cutting velocity is constant. If the inertia and stiffness of the tool is low
compared to the stiffness and strength of the rock, the load on the tool prescribes
whether the tool can cut through the rock or not. Furthermore, experiments have
been performed in which an oscillation is added to the cutting motion.
7.2. Recommendations 143

Further validation of the simulations is desired. Thus far the experimental data
that is (publicly) available is rather limited. Especially concerning experiments on
saturated rock. Special interest in new experiments would be on cutting velocity,
non-constant cutting depth and hydrostatic pressure.

In the current implementation the rock cutting process is modeled as rigid body
movement of the tool. Modeling the movement of the tool with oscillations in the
velocity, or based on a prescribed load will allow to do research into e.g. cutting in
tougher rock, percussive drilling.

Practical Implications
In this dissertation it is shown that the cutting of saturated rock can be significantly
influenced by the cutting velocity and the hydrostatic pressure. Based on results
obtained from both literature and this research, several key aspects for rock cutting
in the industrial practice are noted. These aspects mostly concern the design and
operation of the excavation equipment.

The equipment should be designed in such a way that it can easily deliver the
required normal force with respect to the designed cutting depth. If the equipment
cannot deliver the desired normal force, the tool will not penetrate deep enough.
The loss of penetration depth can also result in a loss of interaction between con-
secutive teeth, leading to a dramatic loss in excavation rate. Design parameters
like cutting velocity and tool spacing can help to extend the applicability of the
equipment.

Both compressive and tensile strength are required to estimate dominant failure 7
mode of the cutting process.

It is shown that for lower cutting velocities, the cutting force is lower as well.
This is caused by a reduction of the pressure drop when dilation occurs in the cutting
process. It might be possible to take advantage of this phenomenon by lowering the
viscosity of the fluid, e.g. using a drilling fluid with lower viscosity can be beneficial
for the drilling process.
 A
Practical Rock Properties
For basically all industrial applications of cutting of saturated rock, the information
that the engineers have at hand to design and estimate the production, cutting
forces, etc. is limited. Most often only the type of rock (sandstone, limestone,
etc.) and the UCS value of the rock are given. Furthermore, on a larger scale
some information about the RQD of the rock may be available. Here the rules of
thumb that are used in this thesis to estimate the rock properties when these are
not provided in the cited publication. These are the classification of rocks based
on their compressive strength, as in table A.1. Typical values for Young’s modulus,
Poisson’s ratio and the strength ratio m are respectively given in tables A.2, A.3
and A.4.

Table A.1: Classification of rocks on the basis of UCS, after Attewell and Farmer (1976).

Strength Strength range Typical rock types

classification [MPa]
Very weak 10-20 Weathered and weakly compacted sedimentary rocks
Weak 20-40 Weakly cemented sedimentary rocks
Medium 40-80 Competent sedimentary rocks,
coarse igneous rocks
Strong 80-160 Fine-grained sandstones,
competent igneous rocks
Very strong 160-320 Quartzites,
fine-grained igneous rocks

145
 146 A. Practical Rock Properties

A
Table A.2: Typical values of Young’s modulus of intact rocks, based on Officials and Transportation
(1989), in (Zhang et al., 2005)

Rock type No of samples Max Minimum Mean Std

Granite 26 100 6.41 52.7 24.5
Marble 14 73.8 4.00 42.6 17.2
Sandstone 27 39.2 0.62 14.7 8.21
Shale 30 38.6 0.007 9.79 10.0
Limestone 30 89.6 4.48 39.3 25.7

Table A.3: Typical values of Poisson’s ratio of intact rocks, based on Officials and Transportation
(1989), in (Zhang et al., 2005)

Rock type No of samples Max Minimum Mean Std

Granite 22 0.39 0.09 0.20 0.08
Marble 5 0.40 0.17 0.28 0.08
Sandstone 12 0.46 0.08 0.20 0.11
Shale 3 0.18 0.03 0.09 0.06
Limestone 19 0.33 0.12 0.23 0.06

Table A.4: Typical values of strength ratio m for different rocks, based on Hoek and Brown (1997),
in (Zhang et al., 2005)

Rock type Texture Mean [MPa] Std [MPa]

Granite Coarse 32 3
Marble Coarse 9 3
Sandstone Medium 17 4
Shale Very fine 6 2
Limestone Coarse 12 3
Limestone Medium 10 2
Limestone Fine 9 2
 Bibliography
Adachi, J., Detournay, E., and Drescher, A. (1996). Determination of rock strength
parameters from cutting tests. In Proceedings of the 2nd North American Rock
Mechanics Symposium (NARMS 1996), Rotterdam, the Netherlands.
Alvarez Grima, M., Heeren, J., Verichev, S. N., and Wijk, J. M. V. (2011). Into the
deep: a risk based approach for research to deepsea mining. In CEDA Dredging
Days 2011, Rotterdam,the Netherlands.
Alvarez Grima, M., Miedema, S., van de Ketterij, R., Yenigül, N., and van Rhee, C.
(2015). Effect of high hyperbaric pressure on rock cutting process. Engineering
Geology, 196:24–36.
Andersen, E. and Azar, J. (1993). PDC-Bit performance under simulated borehole
conditions. In SPE Drilling and Completion, number September, pages 184–188.
Atkinson, B. K. (1984). Subcritical crack growth in geological materials. Journal of
Geophysical Research, 89(B6):4077.
Attewell, P. and Farmer, I. (1976). Principles of engineering geology. Chapman and
Hall, London, UK.
Bagi, K. (2005). An algorithm to generate random dense arrangements for discrete
element simulations of granular assemblies. Granular Matter, 7(1):31–43.
Benabbou, a., Borouchaki, H., Laug, P., and Lu, J. (2009). Geometrical modeling of
granular structures in two and three dimensions. Application to nanostructures.
International Journal for Numerical Methods in Engineering, 80(4):425–454.
Bernabe, Y. (1987). The effective pressure law for permeability during pore pressure
and confining pressure cycling of several crystalline rocks. Journal of Geophysical
Research, 92(B1):649–657.
Bilgin, N. (1977). Investigations into the mechanical cutting characteristics of some
medium and high strength rocks. Phd thesis, University of Newcastle upon Tyne.
Black, A. D., Bland, R. G., Hughes, B., Curry, D. A., Iii, L. W. L., Christensen, H.,
Robertson, H. A., Judzis, A., Prasad, U., and Grant, T. (2008). Optimization of
deep drilling performance with improvements in drill bit and drilling fluid design.
In 2008 IADC/SPE Drilling Conference, page 112731.
Brace, W. and Martin III, R. (1968). A test of the law of effective stress for crys-
talline rocks of low porosity. International Journal of Rock Mechanics and Mining
Sciences, 5:415–426.

147
148 Bibliography

Brady, B. and Brown, E. (2005). Rock mechanics for underground mining. Springer
Science Business Media, Inc, Dordrecht.

Brzesowsky, R. H., Spiers, C. J., Peach, C. J., and Hangx, S. J. T. (2011). Failure
behavior of single sand grains: Theory versus experiment. Journal of Geophysical
Research, 116(B6):B06205.

Buchi, E. (1984). Einfluss geologischer parameter auf die Vortriebsleistung einer

Tunnelbohrmachine. (Influence of geolocial parameters on the performance of a
tunnel boring machine. Doctoral, University of Bern, Switzerland.

Byerlee, J. (1978). Friction of rocks. Pure and Applied Geophysics, 116(4).

Carmona, H., Wittel, F., Kun, F., and Herrmann, H. (2008). Fragmentation Pro-
cesses in Impact of Spheres. Physical Review E, 77:051302.

Carrapatoso, C., Inoue, N., and Curry, D. (2014). New Developments for Single-
Cutter Modeling of Evaporites using Discrete Element Method. In Rock mechanics
for Natural Resources and Infrastructure.

Chaput, E. (1991). Observations and analysis of hard rocks cutting failure mecha-
nisms using PDC cutters. Technical report, Imperial College of Science, Technol-
ogy and Medicine, London, UK.

Chen, J., Beraun, J., and Carney, T. (1999). A corrective smoothed particle method
for boundary value problems in heat conduction. International Journal for Nu-
merical Methods in Engineering, 252(June 1998):231–252.

Chiaia, B. (2001). Fracture mechanisms induced in a brittle material by a hard

cutting indenter. International Journal of Solids and Structures, 38(44-45):7747–
7768.

Cho, J. W., Jeon, S., Yu, S. H., and Chang, S. H. (2010). Optimum spacing of
TBM disc cutters: A numerical simulation using the three-dimensional dynamic
fracturing method. Tunnelling and Underground Space Technology, 25(3):230–
244.

Cho, N., Martin, C., and Sego, D. (2007). A clumped particle model for rock.
International Journal of Rock Mechanics and Mining Sciences, 44(7):997–1010.

Chung, J. S. (1996). Deep-ocean mining: Technologies for manganese nodules and

crusts. International Journal of Offshore and Polar Engineering, 6(4).

Cleary, P. and Monaghan, J. (1999). Conduction modelling using smoothed particle

hydrodynamics. Journal of Computational Physics, 148(1):227–264.

Cleary, P., Sinnott, M., and Morrison, R. (2006). Prediction of slurry transport in
SAG mills using SPH fluid flow in a dynamic DEM based porous media. Minerals
Engineering, 19(15):1517–1527.
Bibliography 149

Colagrossi, A. and Landrini, M. (2003). Numerical simulation of interfacial flows by

smoothed particle hydrodynamics. Journal of Computational Physics, 191(2):448–
475.

Colback, P. and Wild, B. (1965). The influence of moisture content on the com-
pressive strength of rock. In Proceedings of the 3rd Canadian Rock Mechanics
Symposium, pages 65–83.

Combinatie Speurwerk Baggertechniek (1984). Het snijden van gesteente.

Cook, J. (1999). The effects of pore pressure on the mechanical and physical prop-
erties of shales. Oil & Gas Science and Technology, 54(6):695–701.

Cook, J. M., Sheppard, M. C., Houwen, O. H., and Forex, S. (1991). Effects of
strain rate and confining pressure on the deformation and failure of shale. In SPE
Drilling Engineering, number June, pages 100–104.

Cools, P. (1984). Determination of the changes in saturation of a porous rock sample

when varying the porewater pressure. Technical report, Delft.

Cools, P. (1993). Temperature measurements upon the chisel surface during rock
cutting. International Journal of Rock Mechanics and Mining Sciences and,
30(1):25–35.

Coussy, O. (2004). Poromechanics. John Wiley & Sons Ltd, West Sussex.

Cundall, P. and Strack, O. (1979). A discrete numerical method for granular assem-
blies. International Journal for Rock Mechanics, Mining Sciences and Geome-
chanics Abstracts, 29:47–65.

Cunningham, R. A. and Eenink, J. G. (1959). Laboratory study of effect of over-

burden , formation and mud column pressures on drilling rate of permeable for-
mations. In 33rd Annual Fall Meeting of Society of Petroleum Engineers, volume
217, pages SPE 1094–G, Houston, Texas, USA.

Dagrain, F. and Richard, T. (2006). On the influence of PDC wear and rock type
on friction coefficient and cutting efficiency. Eurock 2006: Multiphysics Coupling
and Long Term Behaviour in Rock Mechanics, (MAY 2006):577–584.

de Wit, L. (2015). 3D CFD modelling of overflow dredging plumes. Phd thesis, Delft
University of Technology.

Deketh, H. J. R. (1995). Wear of rock cutting tools. Laboratory experiments on the

abrasivity of rock. Phd thesis, Delft University of Technology.

Detournay, E. and Atkinson, C. (2000). Influence of pore pressure on the drilling

response in low-permeability shear-dilatant rocks. International Journal of Rock
Mechanics and Mining Sciences, 37(7):1091–1101.
150 Bibliography

Detournay, E. and Defourny, P. (1992). A Phenomenological Model for the Drilling

Action of Drag Bits. International Journal of Rock Mechanics and Mining Sci-
ences, 29(I):13–23.

Detournay, E. and Tan, C. (2002). Dependence of drilling specific energy on bottom-

hole pressure in shales. In SPE/ISRM Rock Mechanics Conference, number 2,
pages 1–10, Irving, Texas, USA.

Duda, M. and Renner, J. (2013). The weakening effect of water on the brittle failure
strength of sandstone. Geophysical Journal International, 192(3):1091–1108.

Dyke, C. G. and Dobereiner, L. (1991). Evaluating the strength and deformabil-

ity of sandstones. Quarterly Journal of Engineering Geology and Hydrogeology,
24(1):123–134.

Edelsbrunner, H. and Mücke, E. (1994). Three-dimensional alpha shapes. ACM

Trans Graph, 13(1):43–72.

Edmond, J. and Paterson, M. (1972). Volume changes during the deformation of

rocks at high pressures. International Journal of Rock Mechanics and Mining
Sciences & Geomechanics Abstracts, 9(2):161–182.

Eggleston, D., Herzog, R., and Thomsen, E. (1959). Observations on the angle re-
lationships in metal cutting. ASME Journal of Engineering for Industry, 81:263–
279.

Egholm, D. L. (2007). A new strategy for discrete element numerical models: 1.

Theory. Journal of Geophysical Research, 112(B5):1–16.

Egholm, D. L., Sandiford, M., Clausen, O. R., and Nielsen, S. B. (2007). A new
strategy for discrete element numerical models: 2. Sandbox applications. Journal
of Geophysical Research, 112:B05204: 1–12.

EU Commission (2015). Report on critical raw materials for the EU. Technical
Report Retrieved April, 30.

Evans, I. (1965). The force required to cut coal with blunt wedges. Interna-
tional Journal of Rock Mechanics and Mining Sciences & Geomechanics Abstracts,
2(1):1–12.

Evans, I. (1972). Line spacing of picks for effective cutting. International Journal
of Rock Mechanics and Mining Sciences and, 9(3):355–361.

Evans, I. (1984). A theory of the cutting force for point-attack picks. International
Journal of Mining Engineering, 2(1):63–71.

Fakhimi, A. and Lanari, M. (2014). DEM–SPH simulation of rock blasting. Com-

puters and Geotechnics, 55:158–164.
Bibliography 151

Fakhimi, a. and Villegas, T. (2007). Application of Dimensional Analysis in Cali-

bration of a Discrete Element Model for Rock Deformation and Fracture. Rock
Mechanics and Rock Engineering, 40(2):193–211.

Farmer, I. (1983). Engineering Behaviour of Rocks. Chapman and Hall, 2nd edition.

Fatehi, R., Fayazbakhsh, M. A., and Manzari, M. T. (2009). On discretization

of second-order derivatives in smoothed particle hydrodynamics. International
Journal of Aerospace and Mechanical Engineering, 3(1):41–44.

Feenstra, R. (1988). Status of polycrystalline-diamond- compact bits: part 1-

development. Journal of Petroleum Technology, 40(06):675–684.

Feenstra, R. and van Leeuwen, J. (1964). Full-scale experiments on jets in imper-

meable rock drilling. Journal of Petroleum Technology, 16(3):329–336.

Feng, Y. T., Han, K., and Owen, D. R. J. (2003). Filling domains with disks:
an advancing front approach. International Journal for Numerical Methods in
Engineering, 56(5):699–713.

Francois, D. and Wilshaw, T. (1964). The effect of hydrostatic pressure on the

cleavage fracture of polycrystalline materials. Journal of Applied Physics, 39:4170–
4177.

Gangi, A. F. (1978). Variation of whole and fractured porous rock permeability with
confining pressure. International Journal of Rock Mechanics and Mining Sciences
and, 15(5):249–257.

Garnier, A. and van Lingen, N. (1958). Phenomena affecting drilling rates at depth.
In 33rd Annual Fall Meeting of Society of Petroleum Engineers, pages 232–239.

Gehring, K. (1987). Rock testing procedures at VA’s geotechnical laboratory in

Zeltweg. Technical report, Voest Alpine Zeltweg, Austria.

Goktan, R. (1997). A suggested improvement on Evans’ cutting theory for conical

bits. In Proc. of the 4th Int. Symp. on mine mechanization and automation, pages
A4–57/A4–61, Brisbane, Queensland, Australia.

Goktan, R. M. and Gunes, N. (2005). A semi-empirical approach to cutting force

prediction for point-attack picks. Journal of the South African Institute of Mining
and Metallurgy, 105(April):257–264.

Gowd, T. and Rummel, F. (1980). Effect of Confining Pressure on the Fracture

Behaviour of a Porous. International Journal of Rock Mechanics and Mining
Sciences, 17:225–229.

Gray-Stephens, D., Cook, J., and Sheppard, M. (1994). Influence of pore pres-
sure on drilling response in hard shales. In SPE Drilling & Completion, number
December, page SPE 23414.
152 Bibliography

Griffith, 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.

Guo, Y. and Curtis, J. S. (2015). Discrete element method simulations for complex
granular flows. Annu. Rev. Fluid Mech, 47:21–46.

Hatamura, Y. and Chijiiwa, K. (1975). Analysis of the mechanism of soil cutting

(1st report cutting patterns of soils). Bulletin of the JSME, 18(120):619–626.

He, J. and Vlasblom, W. (1998). Modelling of saturated sand cutting with large
rake angles. In WODCON XV, Las Vegas, USA. WODA.

He, X. and Xu, C. (2015). Discrete element modelling of rock cutting: from ductile
to brittle transition. International Journal for Numerical and Analytical Methods
in Geomechanics, pages n/a–n/a.

Helmons, R. and Miedema, S. (2013). Cutting through hard rock-like materials, a

review of the process. In WODCON XX: The art of dredging, Brussels, Belgium.

Helmons, R., Miedema, S., Alvarez Grima, M., and van Rhee, C. (2016a). Modeling
fluid pressure effects in cutting of saturated rock. Engineering Geology.

Helmons, R., Miedema, S., and Rhee, C. V. (2016b). Simulating hydro mechanical
effects in rock deformation by combination of the discrete element method and the
smoothed particle method. International Journal of Rock Mechanics and Mining
Sciences, 86:224–234.

Helmons, R., Miedema, S., and van Rhee, C. (2015). Inclusion of pore pressure ef-
fects in discrete element modeling of rock cutting. In IV International Conference
on Particle-based Methods - Fundamentals and Applications, PARTICLES 2015,
pages 1–12.

Helmons, R., Miedema, S., and van Rhee, C. (2016c). Discrete element modeling of
circular rock cutting with evaluation of pore pressure effects. In WODCON XXI:
Innovations in dredging, Miami, Florida, USA.

Hemami, B. and Fakhimi, A. (2014). Numerical simulation of rock-loading machine

interaction. In 48th US Rock Mechanics/Geomechanics Symposium.

Hettiaratchi, D. and Reece, A. (1967). Symmetrical three-dimensional soil failure.

Journal of Terramechanics, 4(3):45–67.

Hoek, E. and Brown, E. (1980). Underground excavation in rock. Institution of

Mining and Metallurgy, London, UK.

Hoek, E. and Brown, E. T. (1997). Practical estimates of rock mass strength.

International Journal of Rock Mechanics and Mining Sciences, 34(8):1165–1186.
Bibliography 153

Hoover, W. G., Pierce, T. G., Hoover, C. G., Shugart, J. O., Stein, C. M., and
Edwards, a. L. (1994). Molecular-dynamics, smoothed-particle applied mechanics,
and irreversibility. 28(19):155–174.

Huang, H. (1999). Discrete element modeling of tool-rock interaction. Phd thesis,

University of Minnesota.

Huang, H. and Detournay, E. (2013). Discrete element modeling of tool-rock inter-

action II : rock indentation. International Journal for Numerical and Analytical
Methods in Geomechanics, 37(13):1930–1947.

Huang, H., Lecampion, B., and Detournay, E. (2013). Discrete element modeling
of tool-rock interaction I : rock cutting. International Journal for Numerical and
Analytical Methods in Geomechanics, 37(13):1913–1929.

Hurt, K. (1980). Rock cutting experiments with point attack tools. Colliery
Guardian Coal Int, 228:47–50.

Hurt, K. and MacAndrew, K. (1985). Cutting efficiency and life of rock cutting
picks. Min. Sci. Tech., 2:139–151.

Judzis, A., Bland, R. G., Curry, D. A., Black, A. D., Robertson, H. A., Mein-
ers, M. J., and Grant, T. C. (2009). Optimization of deep-drilling performance
— benchmark testing drives ROP improvements for bits and drilling fluids. In
SPE/IADC Drilling Conference 2007, volume 2008, page SPE 105885, Amster-
dam, the Netherlands.

Kaitkay, P. and Lei, S. (2005). Experimental study of rock cutting under external
hydrostatic pressure. Journal of Materials Processing Technology, 159(2):206–213.

Kenney, P. and Johnson, S. (1976). The effect of wear on then performance of

mineral cutting tools. Colliery Guardian, 224:246–252.

Kolle, J. (1996). The effects of pressure and rotary speed on the drag bit drilling
strength of deep formations. In SPE Annual Technical Conference and Exhibition,
page SPE 36434, Denver, Colorado, USA.

Komoróczi, A., Abe, S., and Urai, J. L. (2013). Meshless numerical modeling of
brittle–viscous deformation: first results on boudinage and hydrofracturing using
a coupling of discrete element method (DEM) and smoothed particle hydrody-
namics (SPH). Computational Geosciences, 17(2):373–390.

Kuhne, I. (1952). Die Wirkungsweise von Rotarymeiseln und anderen drehenden

Gesteinsbohren. Sonderdruck aus der Zeitschrift, Bohrtechnik-Brunnenbau, pages
1–5.

Lashkaripour, G. R. and Ghafoori, M. (2002). The engineering geology of the

Tabarak Abad Dam. Engineering Geology, 66(3-4):233–239.

Lawn, B. (1993). Fracture of Brittle Solids. Cambridge university press.

154 Bibliography

Ledgerwood III, L. W. (2007). PFC modeling of rock cutting under high pressure
conditions. In 1st Canada -U.S. Rock Mechanics Symposium, volume 2, pages
ARMA–07–063, Vancouver, Canada.

Lei, S. and Kaitkay, P. (2003). Distinct element modeling of rock cutting under
hydrostatic pressure. Key Engineering Materials, 250:110–117.

Li, H., Li, J., Liu, B., Li, J., Li, S., and Xia, X. (2013). Direct tension test for rock
material under different strain rates at quasi-static loads. Rock Mechanics and
Rock Engineering, 46(5):1247–1254.

Li, X., Li, H., Liu, Y., Zhou, Q., and Xia, X. (2016). Numerical simulation of rock
fragmentation mechanisms subject to wedge penetration for TBMs. Tunnelling
and Underground Space Technology, 53:96–108.

Liefferink, D. (2013). Failure mechanism of cutting submerged frozen clay in an

arctic rrenching process. Master thesis, Delft University of Technology.

Liu, H. Y., Kou, S. Q., and Lindqvist, P.-A. (2002). Numerical simulation of the
fracture process in cutting heterogeneous brittle material. International Journal
for Numerical and Analytical Methods in Geomechanics, 26(13):1253–1278.

Liu, J., Cao, P., and Han, D. (2015). Sequential indentation tests to investigate the
influence of confining stress on rock breakage by tunnel boring machine cutter in
a biaxial state. Rock Mechanics and Rock Engineering, 49(4):1479–1495.

Lohner, R., Baqui, M., Haug, E., and Muhamad, B. (2016). Real-time micro mod-
elling of a million pedestrians. Engineering Computations, 33(1):217–237.

Louis, C., Dessenne, J.-L., and Feuga, B. (1977). Interaction between water flow
phenomena and the mechanical behavior of soil or rock masses. In Gudehus, G.,
editor, Finite Elements in Geomechanics, pages 479–511. Wiley, New York.

Lucy, L. (1977). A numerical approach to the testing of the fission hypothesis.

Astronomical Journal, 82:1013–1024.

Luding, S. (2008). Cohesive Frictional Powders: Contact Models for Tension. Gran-
ular Matter, 10:235–246.

Luding, S. (2011). From discrete particles to solids - about sintering and self-healing
(review). Computer Methods in Material Science, 11:53–63.

Luger, H. (1981). Invloed van wateroverspanning en vervormingssnelheid bij indring-

proeven op St. Leu (in Dutch). Technical report, Delft Soil Mechanics Laboratory.

Mahabadi, O. K., Lisjak, A., Grasselli, G., Lukas, T., and Munjiza, A. (2004).
Numerical modelling of a triaxial test of homogeneous rocks using the combined
finite-discrete element method.
Bibliography 155

Majidi, R., Miska, S., and Tammineni, S. (2011). PDC single cutter: the effects of
depth of cut and rpm under simulated borehole conditions. Wiertnictwo Nafta
Gaz, 28(1-2):283–295.

Martin III, R. (1980). Pore pressure stabilization of failure in westerly granite.

Geophysical Research Letters, 7(5):404–406.

Masuda, K., Mizutani, H., and Yamada, I. (1987). Experimental study of strain-
rate dependence and pressure dependence of failure properties of granite. J. Phys.
Earth, 35:37–66.

Medhurst, T. and Brown, E. (1996). Large scale laboratory testing of coal. In

Proc. 7th Australian-New Zeland Conf. on Geomech., pages 203–208, Canberra,
Australia.

Meijer, K. (1972). Eerste serie rotssnijproeven op tegels (in Dutch). Technical

report, Delft Hydraulics, Delft, Netherlands.

Meijer, K. (1973a). Brokvorm bij tweedimensionaal snijden (in Dutch). Technical

report, Delft Hydraulics, Delft, Netherlands.

Meijer, K. (1973b). Het snijden van rots: tweede serie rotssnijproeven op tegels (in
Dutch). Technical report, Delft Hydraulics, Delft, Netherlands.

Mendoza Rizo, J. (2013). Considerations for discrete element modeling of rock

cutting. Phd thesis, University of Pittsburgh.

Merchant, M. (1945). Mechanics of the metal cutting process. II. plasticity condi-
tions in orthogonal cutting. Journal of Applied Physics, 16(6):318.

Miedema, S. (1987). Calcuation of the cutting forces when cutting water saturated
sand. Phd thesis, Delft University of Technology.

Miedema, S. (2014). The Delft sand, clay and rock cutting model. IOS Press, Delft,
Netherlands.

Miedema, S. and Frijters, D. (2003). The mechanism of kinematic wedges at large

cutting angles - velocity and friction measurements. In 23rd WEDA Technical
Conference & 35th TAMU Dredging Seminar.

Miedema, S. and Zijsling, D. (2012). Hyperbaric rock cutting. In OMAE 2012:

31st International Conference on Ocean, Offshore and Arctic Engineering, Rio de
Janeiro, Brazil.

Mishnaevsky Jr, L. (1995). Physical mechanisms of hard rock fragmentation under

mechanical loading : a review. International Journal of Rock Mechanics and
Mining Sciences, 32(8):763–766.

Monaghan, J. (1994). Simulating free surface flows with SPH. Journal of Compu-
tational Physics, 110:399–406.
156 Bibliography

Monaghan, J. (2012). Smoothed particle hydrodynamics and its diverse applications.

Annual Review of Fluid Mechanics, 44(1):323–346.

Moon, T. and Oh, J. (2011). A study of optimal rock-cutting conditions for hard rock
TBM using the discrete element method. Rock Mechanics and Rock Engineering,
pages 837–849.

Morrow, C., Moore, D., and Lockner, D. (2000). The effect of mineral bond strength
and adsorbed water on fault gouge frictional strength. Geophysical Research Let-
ters, 27(6):815–818.

Motzheim, R. (2016). Vibration-impact cutting of submerged forzen clay in arctic

trenching process. Master thesis, Delft University of Technology.

Muhammad, N., Rogers, B., and Li, L. (2013). Understanding the behaviour of
pulsed laser dry and wet micromachining processes by multi-phase smoothed par-
ticle hydrodynamics (SPH) modelling. Journal of Physics D: Applied Physics,
46(9):095101.

Muirhead, I. and Glossop, L. (1968). Hard rock tunnelling machines. Trans. Inst.
Min. Metall. Sect A, 77:1–21.

Natau, O., Mutschler, T., and Lempp, C. (1991). Estimation of the cutting rate
and the bit wear of partial-face tunnelling machines. In 7th ISRM Congress, page
289, Aachen, Germany.

Nishimatsu, Y. (1972). The mechanics of rock cutting. International Journal of

Rock Mechanics and Mining Sciences, 9:261–270.

Nur, A. and Byerlee, J. (1971). An exact effective stress law for elastic deformation
of rocks with fluids. Journal of Geophysical Research, 76(26):6414–6419.

Oñate, E. and Rojek, J. (2004). Combination of discrete element and finite element
methods for dynamic analysis of geomechanics problems. Computer Methods in
Applied Mechanics and Engineering, 193(27-29):3087–3128.

Obreimoff, J. W. (1930). The splitting strength of mica. Proceedings of the Royal

Society of London. Series A, Containing Papers of a Mathematical and Physical
Character, 127(805):290–297.

O’Dogherty, M. and Burney, A. (1963). A laboratory study of the effect of cutting

speed on the performance of two coal cutter picks. Colliery Eng. (London), 40:51–
54.

Officials, A. A. o. S. H. and Transportation (1989). Standard specifications for

highway bridges. Washington DC, USA, 14th editi edition.

Onodera, T. and Kamura, A. (1980). Relation between texture and mechanical

properties of crystalline rocks. Bulletin International Association Engineering
Geology, 22:173–177.
Bibliography 157

Ortega, A. (2014). Towards zero impact of deep sea offshore projects. Technical
report.

Paterson, M. and Wong, T.-f. (2005). Experimental Rock Deformation - The Brittle
Field. Springer Berlin Heidelberg, 2nd edition.

Pittino, G., Galler, R., Mikl-resch, M., Tichy, R., Ecker, W., Gimpel, M., and Kargl,
H. (2015). The rock cutting process of road headers – Overview of laboratory tests
and numerical simulation. In ISRM Regional Symposium EUROCK 2015 - Future
Development of Rock Mechanics, pages 243–248.

Pomeroy, C. (1963). The breakage of coal by wedge action. Colliery Guardian1,

207(5354):672–677.

Population Division of the Department of Economic and Social Affairs of the United
Nations Secretariat (2015). World population prospects: the 2015 revision. Tech-
nical report, New York.

Potapov, A. V., Hunt, M. L., and Campbell, C. S. (2001). Liquid-solid flows us-
ing smoothed particle hydrodynamics and the discrete element method. Powder
Technology, 116(2-3):204–213.

Potts, E. and Shuttleworth, P. (1958). A study on the ploughability of coal, with

special reference to the effects of blade shape, direction of planing to the cleat,
planing speed and the influence of water infusion. Trans. Inst. Min. Eng., 117:519–
553.

Potyondy, D. O. (2015). The bonded-particle model as a tool for rock mechan-

ics research and application: current trends and future directions. Geosystem
Engineering, 18(1):1–28.

Potyondy, D. O. and Cundall, P. a. (2004). A bonded-particle model for rock. Inter-

national Journal of Rock Mechanics and Mining Sciences, 41(8 SPEC.ISS.):1329–
1364.

Ranman, K. (1985). A model describing rock cutting with conical picks. Rock
Mechanics and Rock Engineering, 18(2):131–140.

Reviron, N., Reuschlé, T., and Bernard, J.-D. (2009). The brittle deformation
regime of water-saturated siliceous sandstones. Geophysical Journal International,
178(3):1766–1778.

Rice, J. R. and Cleary, M. P. (1976). Some basic stress diffusion solutions for
fluid saturated elastic porous media with compressible constituents. Reviews of
Geophysics and Space Physics, 14(2):227–241.

Richard, T., Dagrain, F., Poyol, E., and Detournay, E. (2012). Rock strength
determination from scratch tests. Engineering Geology, 147-148:91–100.
158 Bibliography

Richard, T., Detournay, E., Drescher, A., Nicodeme, P., and Fourmaintraux, D.
(1998). The scratch test as a means to measure strength of sedimentary rocks. In
SPE/ISRM Eurock ’98, Trondheim, Norway.

Robinson, M., Ramaioli, M., and Luding, S. (2014). Fluid-particle flow simulations
using two-way-coupled mesoscale SPH-DEM and validation. International journal
of multiphase flow, 59:121–134.

Rojek, J. (2014). Discrete element thermomechanical modelling of rock cutting with

valuation of tool wear. Computational Particle Mechanics, 1(1):71–84.

Rojek, J., Nosewicz, S., Mazdziarz, M., Kowalczyk, P., Wawrzyk, K., and Lumel-
skyj, D. (2017). Modeling of a Sintering Process at Various Scales. Procedia
Engineering, 177:263–270.

Rojek, J., Oñate, E., Labra, C., and Kargl, H. (2011). Discrete element simulation
of rock cutting. International Journal of Rock Mechanics and Mining Sciences,
48(6):996–1010.

Rostami, J. and Ozdemir, L. (1993). A new model for performance prediction of hard
rock TBMs. In Proceedings of the Rapid Excavation and Tunneling Conference.
Society for mining, metallogy and exploration inc.

Roxborough, F. (1973). Cutting rock with picks. Trans. Inst. Min. Eng., 132:445–
454.

Roxborough, F. and Phillips, H. (1975). Rock excavation by disc cutter. Interna-

tional Journal of Rock Mechanics and Mining Sciences, 12:361–366.

Roxborough, F. and Sen, G. (1986). Breaking coal and rock. In Martin, C., editor,
Australian coal mining practice, pages 130–147. Aimm monog edition.

Rudnicki, J. W. and Chen, C.-H. (1988). Stabilization of rapid frictional slip

on a weakening fault by dilatant hardening. Journal of Geophysical Research,
93(B5):4745.

Rutter, E. (1972). The effects of strain-rate changes on the strength and ductility
of Solenhofen limestone at low temperatures and confining pressures. Interna-
tional Journal of Rock Mechanics and Mining Sciences & Geomechanics Abstracts,
9(2):183–189.

Scholtès, L. and Donzé, F.-V. (2013). A DEM model for soft and hard rocks: Role
of grain interlocking on strength. Journal of the Mechanics and Physics of Solids,
61(2):352–369.

Schön, J. (1996). Physical properties of rocks - fundamentals and principles of

petrophysics. Pergamon, Oxford, UK.
Bibliography 159

Shimizu, H., Murata, S., and Ishida, T. (2011). The distinct element analysis for
hydraulic fracturing in hard rock considering fluid viscosity and particle size distri-
bution. International Journal of Rock Mechanics and Mining Sciences, 48(5):712–
727.

Singh, A. and Luding, S. (2010). Flow behavior at different shear rates for dry pow-
ders. In Proceedings of World Congress on Particle Technology WCPT6. Nürnberg
Messe GmbH.

Skempton, A. (1960). Effective stress in soils, concrete and rocks. In Pore pressure
and suction in soils, pages 4–16.

Snowdon, R. a., Ryley, M. D., and Temporal, J. (1982). A study of disc cutting
in selected British rocks. International Journal of Rock Mechanics and Mining
Sciences and, 19(3):107–121.

Stahl, M. and Konietzky, H. (2010). Discrete element simulation of ballast and

gravel under special consideration of grain-shape, grain-size and relative density.
Granular Matter, 13(4):417–428.

Su, O. and Akcin, N. A. (2011). Numerical simulation of rock cutting using the
discrete element method. International Journal of Rock Mechanics and Mining
Sciences, 48(3):434–442.

Sun, X., Sakai, M., and Yamada, Y. (2013). Three-dimensional simulation of a

solid-liquid flow by the DEM-SPH method. Journal of Computational Physics,
248:147–176.

Swan, G., Cook, J., Bruce, S., and Meehan, R. (1989). Strain rate effects in Kim-
meridge bay shale. International Journal of Rock Mechanics and Mining Sciences
& Geomechanics Abstracts, 26(2):135–149.

Swolfs, H. (1972). Chemical effects of pore fluids on rock properties. In Underground

Waste Management and Evironmental Implications, chapter 18, pages 224–234.

Terzaghi, K. (1943). Theoretical Soil Mechanics. John Wiley & Sons Ltd.

Thomas, P. and Bray, J. (1999). Capturing nonspherical shape of granular media

with disk clusters. Journal of Geotechnical and Geoenvironmental Engineering,
1235(3):169–178.

Tiller, F. (1953). Chem. Engrg. Prog, 49:467.

Tropin, N., Manakov, V., and Bocharov, O. (2014). Implementation of boundary

conditions in discrete element modeling of rock cutting under pressure. Applied
Mechanics and Materials, 598:114–118.

Uittenbogaard, R. (1980). Indringproeven op WL-II met grote indringdiepte:

Meetverslag en analyse (in Dutch). Technical report, Delft Hydraulics Laboratory.
160 Bibliography

van Grunsven, F., Keetels, G., and van Rhee, C. (2016). Modeling offshore mining
turbidity sources. In WODCON XXI: Innovations in dredging, Miami, Florida,
USA.

van Kesteren, W. (1995). Numerical simulations of crack bifurcation in the chip

forming cutting process in rock. Fracture of Brittle Disordered Materials: con-
crete, Rock and Ceramics, pages 505–524.

van Opstal, T. (2015). cutter head for removing material from a water bed patent
IHC Holland.pdf.

van Wijk, J. (2016). Vertical hydraulic transport for deep sea mining. Phd thesis,
Delft University of Technology.

van Wyk, G., Els, D., Akdogan, G., Bradshaw, S., and Sacks, N. (2014). Dis-
crete element simulation of tribological interactions in rock cutting. International
Journal of Rock Mechanics and Mining Sciences, 65:8–19.

Vasarhelyi, B. (2003). Some observations regarding the strength and deformability

of sandstones in dry and saturated conditions. Bulletin of Engineering Geology
and the Environment, 62(3):245–249.

Vasarhelyi, B. (2005). Statistical analysis of the influence of water content on

the strength of the miocene limestone. Rock Mechanics and Rock Engineering,
38(1):69–76.

Verhoef, P. (1997). Wear of rock cutting tools implications for the site investigation
of rock dredging projects. Phd thesis, Delft University of Technology, Delft.

von den Driesch, S. (1994). Schneidtechnische und verfahrentechnische Weiteren-

twicklungen fur den maschinellen Vortrieb (Operational and cutting technical
developments for tunneling machines). Gluckauf-Forschunshefte, 55:156–160.

Vutukuri, V. (1974). The effect of liquids on the tensile strength of limestone.

International Journal of Rock Mechanics and Mining Sciences, 11:27–29.

Wang, Y., Lau, Y., and Gao, Y. (2014). Examining the mechanisms of sand creep
using DEM simulations. Granular Matter, 16(5):733–750.

Wawersik, W. and Fairhurst, C. (1970). A study of brittle rock fracture in laboratory

compression experiments. International Journal of Rock Mechanics and Mining
Sciences, 7:561–575.

Wendland, H. (1995). Piecewise polynomial positive definite and compactly sup-

ported radial functions of minimal degree by wendland. Advances in Computa-
tional Mathematics, 4:389–396.

Winterwerp, J. and van Kesteren, W. (2004). Introduction to the physics of cohesive

sediment in the marine environment. Elsevier, Amsterdam, the Netherlands.
Bibliography 161

Wong, T.-f., David, C., and Menendez, B. (2004). Mechanical compaction. In

Gueguen, Y. and Bouteca, M., editors, Mechanics of fluid saturated rocks, pages
55–114. Elsevier Acadamic Press, Amsterdam, the Netherlands.
World Ocean Review (2014). Marine resources - opportunities and risks. Technical
report.
Yamazaki, T. and Park, S. (2003). Relationships between geotechnical engineering
properties and assay of seafloor massive sulfides. In 13th International Offshore
and Polar Engineering Conference, volume 5, pages 310–316, Honolulu, Hawaii,
USA.
Yang, B., Jiao, Y., and Lei, S. (2006). A study on the effects of microparame-
ters on macroproperties for specimens created by bonded particles. Engineering
Computations, 23(6):607–631.
Yenigül, N. B. and Alvarez Grima, M. (2010). Discrete element modeling of low
strength rock. Numerical Methods in Geotechnical Engineering, pages 207–212.
Yu, B. (2005). Numerical simulation of continuous miner rock cutting process nu-
merical aimulation of continuous miner rock cutting process. Phd thesis, West
Virginia University.
Zárate, F. and Oñate, E. (2015). A simple FEM–DEM technique for fracture predic-
tion in materials and structures. Computational Particle Mechanics, 2(3):301–314.
Zhang, J., Ong, S., All-Bazali, T., Chenevert, M., Sharma, M., and Yi, X. (2005).
Effects of Strain Rate on Failure Characteristics of Shales. In The 40th U.S.
Symposium on Rock Mechanics (USRMS): Rock Mechanics for Energy, Mineral
and Infrastructure Development in the Northern regions, number 1958, pages
ARMA/USRMS 05–673.
Zhang, J., Wong, T.-F., and Davis, D. M. (1990). Micromechanics of pressure-
induced grain crushing in porous rocks. Journal of Geophysical Research,
95(B1):341.
Zhang, X. and Wong, L. (2013). Loading Rate Effects on Cracking Behavior of
Flaw-contained Specimens under Uniaxial Compression. International Journal of
Fracture, 180:93–110.
Zhang, X. and Wong, L. (2014). Choosing a Proper Loading Rate for Bonded-
Particle Model of Intact Rock. International Journal of Fracture, 189:163–179.
Zhao, T., Crosta, G., Utili, S., and De Blasio, F. (2017). Investigation of Rock
Fragmentation during Rockfalls and Rock Avalanches via 3D Discrete Element
Analyses. Journal of Geophysical Research: Earth Surface.
Zijsling, D. (1984). Analysis of temperature distribution and performance of poly-
crystalline diamond compact bits under field drilling conditions. In SPE Annual
Technical Conference and Exhibition, Dallax, Texas, USA. Society of Petroleum
Engineers.
162 Bibliography

Zijsling, D. (1987). Single cutter testing - A key for PDC bit development. In
Offshore Europe 87, page SPE16529/1, Aberdeen, UK.

Zijsling, D. (2013). Personal communication.

Zsaki, a. M. (2009). An efficient method for packing polygonal domains with disks
for 2D discrete element simulation. Computers and Geotechnics, 36(4):568–576.
 List of Publications
Journal Publications
2. R.L.J. Helmons, S.A. Miedema, M. Alvarez Grima, C. van Rhee, Modeling
fluid pressure effects in cutting of saturated rock, Engineering Geology 250 (2016),
50-60.
1. R.L.J. Helmons, S.A. Miedema, C. van Rhee, Simulating hydro mechanical
effects in rock deformation by combination of the discrete element method and the
smoothed particle method, International Journal of Rock Mechanics and Mining Sci-
ences 86 (2016), 224-234.

Conferences
6. R.L.J. Helmons, S.A. Miedema, C. van Rhee, Discrete element modeling of
circular rock cutting with evaluation of pore pressure effects, WODCON XXI World
Dredging Conference, Innovations in Dredging, 13-17 June 2016, Miami, USA.
5. R.L.J. Helmons, S.A. Miedema, C. van Rhee, Modeling the effect of wa-
ter depth on rock cutting processes with the use of discrete element method, CEDA
Dredging Days 2015, 5-6 November 2015, Rotterdam, Netherlands.
4. R.L.J. Helmons, S.A. Miedema, C. van Rhee, Inclusion of pore pressure ef-
fects in discrete element modeling of rock cutting, IV International Conference on
Particle-based Methods - Fundamentals and Applications, PARTICLES 2015, 28-30
september 2015, Barcelona, Spain
3. R.L.J. Helmons, S.A. Miedema, C. van Rhee, A new approach to model hyper-
baric rock cutting processes, ASME 2014 International Conference on Ocean, Offshore
and Arctic Engineering, OMAE 2014, 8-13 June 2014, San Francisco, USA
2. R.L.J. Helmons, S.A. Miedema, Rock cutting for deep sea mining: an extension
into poromechanics, Poromechanics V: Fifth Biot Conference on Poromechancis, 10-
12 July 2013, Vienna, Austria
1. R.L.J. Helmons, S.A. Miedema, Cutting through hard rock-like materials, a
review of the process, WODCON XX: The art of dredging, 3-7 June 2013, Brussels,
Belgium

Magazines
1. R.L.J. Helmons, S.A. Miedema, C. van Rhee, Modeling the effect of water
depth on rock cutting processes with the use of discrete element method, Terra and
Aqua 142, March 2016.

Awards
1. IADC Young Authors Award 2015

163
 Curriculum Vitæ
Rudolfus Lambertus Jacobus Helmons

27-04-1987 Born in Halsteren, the Netherlands.

Education
1998-2005 Secondary School
Regionale Scholengemeenschap RSG ’t Rijks, Bergen op Zoom

2005-2010 BSc, Mechanical Engineering

2009-2011 MSc, Mechanical Engineering

Track: Thermo-Fluids Engineering
with great appreciation
Eindhoven University of Technology

2012-2017 PhD, Offshore and Dredging Engineering

Delft University of Technology
Thesis: Excavation of Hard Deposits and Rocks - on the cutting of satu-
rated rock
Promotor: Prof. dr. ir. C. van Rhee
Co-promotor: Dr. ir. S. A. Miedema

Work Experience
2011-2017 Research Engineer, IHC MTI (Royal IHC)
2017 Researcher, Delft University of Technology

165
 Acknowledgements
Ben je net lekker op dreef met je afstuderen, beginnen je begeleiders ineens (onafhankelijk
van elkaar) of promoveren niet iets is voor na het afstuderen. Dit terwijl bij mij altijd het
idee leefde om na de studie direct de industrie in te gaan. Het feit dat je dit proefschrift
nu al in de hand hebt laat zien dat het toch anders is gelopen dan vooraf gedacht.
De grens tussen (academisch) onderzoek en de industrie heeft me altijd blijven boeien.
In de tijd dat ik afstudeerde ging het erg goed met de baggersector, waardoor vrijwel direct
bij mijn aanstelling als R&D Engineer bij toen nog MTI Holland (nu IHC MTI) zich de
kans voor deed om voor IHC een promotieonderzoek uit te voeren. Ik ben dan ook Henk
van Muijen en Robert van de Ketterij dankbaar dat zij dit vanuit IHC zo snel mogelijk
maakten.
Een van de projecten waar IHC aan deel nam was het onderzoek naar het snijden
van gesteente op grote waterdieptes. Een project dat vooral mijn aandacht trok door
zijn toepassing voor o.a. diepzeemijnbouw en zijn combinatie in zowel de bagger, olie en
gasindustrie.
Veel dank gaat uit naar mijn begeleiders Cees van Rhee en Sape Miedema. Cees, als
prof heb je het altijd druk, maar de deur stond altijd open. Wanneer ik ergens op vast liep
kon ik altijd bij even aankloppen om mij op weg te helpen, soms waren enkele suggesties
voldoende, soms met uitgebreidere discussies. Sape, ook bij jou kon ik altijd langslopen
om even wat te vragen, vaak waren enkele suggesties voldoende, maar liep het toch uit op
een uitgebreidere discussie, al dan niet vakinhoudelijk. Ik ben ook erg blij dat ik samen
met jou les heb mogen geven in Vietnam. Naast de unieke ervaring heb ik je op een heel
andere manier leren kennen. Het was een erg leuke tijd.
Joep, ik leerde je kennen tijdens mijn afstuderen, toen als begeleider. Niet wetende dat
we later als collega’s en de laatste jaren ook als kamergenoot verder zouden gaan. Door
de jaren heen heb ik veel gehad aan de vele discussies, zowel voor het onderzoek, andere
werkzaamheden als alle heerlijk onzinnige onderwerpen. Xiuhan, both our projects were
on the same topic. It really helped me a lot to be able to discuss up to every little detail in
our projects. Zeker ook de andere collega’s in de vakgroep hebben erg veel geholpen bij het
gemotiveerd blijven voor het onderzoek. Met name de goede sfeer, wat zeker ook is terug
te zien in de vele (on)zinnige discussies bij de koffie-automaat. Ik heb ook veel genoten van
de gezamenlijke congresbezoeken en de aansluitende reizen. Kortom, ook Arno, Bas, Dave,
Ershad, Frans, Geert, Jort, Rik en Thijs en de inmiddels voormalige collega-promovendi
Lynyrd en Ralph, bedankt allemaal.
Verder wil ik de collega’s vanuit de industrie bedanken. Met name Ronald, Michel,
Djurre, Rihard, Mario, Robert, Rick, Roeland en alle anderen die namens de deelnemende
bedrijven hebben bijgedragen aan het tot stand komen van dit onderzoek en het sturen
naar de gekozen aanpak.
Naast de vakinhoudelijke steun heb ik ook veel medeleven ervaren van familie en vrien-
den. Mannen van ’t 51e, ik heb alle jaren enorm genoten van de chill-weekenden. Het is
een heerlijke traditie geworden, ik hoop dat we die nog vele jaren door kunnen zetten.
Ik wil vooral ook Corine bedanken. Ik weet dat ik het je de afgelopen jaren niet altijd
gemakkelijk heb gemaakt. Maar je hebt me in al die jaren altijd gesteund. Tot slot, dit

167
168 Acknowledgements

alles zou niet mogelijk geweest zijn door de steun van mijn ouders om te gaan studeren en
tegelijkertijd iets te doen wat ik leuk vind. Helaas hebben zij dit niet meer mee kunnen
maken, maar ik ben er enorm trots op, hoewel het niet altijd makkelijk is geweest, dat zij
mij tijdens mijn eerste 18/19 levensjaren hebben geholpen te worden tot wie ik nu ben.