2003:04

LICENTIATE THESIS

Numerical Modelling of the

Rock Fracture Process
Under Mechanical Loading

by

LIU HONGYUAN

Department of Civil and Mining Engineering

Division of Rock Engeneering

2003:04 • ISSN: 1402 - 1757 • ISRN: LTU - LIC - - 03/04 - - SE

 Licentiate Thesis

Numerical Modelling of the Rock Fracture

Process under Mechanical Loading

Hongyuan Liu

Department of Civil and Mining Engineering

Luleå University of Technology

December 2002

I
Preface

This thesis consists of two parts: on the one hand the research and development (R & D) for and the
calibration of the rock and tool interaction code (R-T2D), and on the other hand the preliminary industrial
application of R-T2D in rock fragmentation by mechanical loading. Some of the research in the first part was
conducted in China when I was studying at Northeastern University for a Master’s degree. The financial
support from LKAB’s Foundation for the Promotion of Research and Education at Luleå University of
Technology, Trelleborg AB’s Research and Education Foundation, the Foundation for Technology Transfer,
Arne S. Lundbergs Foundation, and the Knowledge Foundation is greatly appreciated.

I wish to express my sincere gratitude to my supervisor, Professor Per-Arne Lindqvist, for his effectual
supervision, steady support and valuable comments in the course of the thesis.

I want to give my heartfelt thanks to my examiner and scientific supervisor, Professor Shaoquan Kou, for
his heuristic supervision, inspiring encouragement and meticulous comments on all the work that my thesis
is based on.

I would like to thank my external supervisor, Professor Chun’an Tang, for his most valuable contributions
to the code development, positive encouragement and illuminating instructions during my studies both in
Sweden and in China.

I am grateful to Professor Karel Miskovsky, Mrs. Margarida Roquete and Mr. Peter Bergman for their
valuable comments, suggestions and discussions in the research work. The help from Dr. Zongxian Zhang
in the research work is gratefully appreciated. I particularly appreciate my colleagues at the Center for Rock
Instability and Seismicity Research (CRISR) at Northeastern University, China for their help when I was
conducting the original R & D for the R-T2D code.

My sincere gratitude is extended to Mrs. Eva Lambertsson for her kind help in many respects, to my other
colleagues at the Divisions of Rock Engineering and Rock Mechanics for their friendship and help, to Mr.
Paul McMillen for revising the English of the thesis, and to Mrs. Xiangqun Ai, Mrs. Mei Wang and other
friends for their help in various ways.

I would also like to thank my Swedish and Chinese friends for their hospitality during my studies in Sweden.
I will cherish forever the experience of studying in Sweden.

I wish to take this opportunity to express my gratitude to my parents. In order to create a good opportunity
for me to receive higher education, they have been shouldering a great burden since my childhood. Finally, I
would also like to thank my girlfriend, Cicy, for her encouragement and unconditional support during my
work on the thesis.

Hongyuan Liu

December, 2002

Luleå, Sweden

II
Abstract

The fracture of rock has been the subject of extensive research in mechanical fragmentation. With the
rapid development of computing power, interactive computer graphics and topological data structure,
numerical tools have become a good means of gaining some insight into the problem of rock fracture.
However, most of the commercial programs are not so robust that they can model the progressive fracture
process while taking the rock fabric characteristics (heterogeneity) into consideration. Consequently, a
novel numerical code called the rock and tool interaction code (R-T2D) has been developed to understand
the progressive fracture process of heterogeneous rock. The research and development (R & D) for and the
calibration of the R-T2D code (Paper A and B) mainly consist of (1) the characterization of rock
heterogeneity, (2) mesoscopic constitutive laws, non-linear behaviour and associated seismicity, (3) the
Mohr-Coulomb and the double elliptic strength criterion, and (4) the calibration of the R-T2D code by
simulating basic rock mechanics experiments.
In the second part of the thesis (Paper C and D), preliminary industrial applications of the R-T2D code in
mechanical fragmentation are conducted. In rock cutting, the peculiarities of cutting heterogeneous brittle
materials are investigated. Besides, the fracture processes induced by cutters with different back rake angles
are examined and compared with each other. In indentation, research on indentation-induced fractures is
performed using the R-T2D code. The influences of confining pressure on the formation of side cracks are
discussed. In simultaneous loading by multiple indenters, the propagation, interaction and coalescence of
side cracks induced by neighbouring indenters are simulated to check how the side cracks propagate,
interact and coalesce to form large rock chips. Finally, the indexing-effect in simultaneous loading is
discussed.
The major results for the two parts can be summarized as follows: (1) On the basis of the Weibull
distribution, a heterogeneous material model is proposed to characterize the heterogeneity of rock, and in
this model rock is specified by a few characteristic parameters: the homogeneous index and the elemental
seed parameters. For any specific rock, the homogeneous index can be determined on the basis of the defect
distribution of the microstructure and the elemental parameters can be obtained from laboratory tests. (2)
Rock heterogeneity has an important influence on the crack initiation location and the subsequent
propagation path. (3) Compared with the traditional Mohr-Coulomb and Hoek-Brown strength criterion, the
double elliptic strength criterion is more useful for modelling the fracture occurring in mechanical
fragmentation and can represent the transition from brittle failure to ductile cataclastic failure with
increasing confining pressure. (4) A series of numerical experiments including both intact rock and notched
rock were conducted using the R-T2D code to obtain the physical-mechanical properties and fracture
toughness, and to simulate the crack initiation and propagation, as well as the whole progressive fracture
process. The developed R-T2D code seems to have built a bridge between the physical-mechanical
parameters and fracture toughness. The detailed visually shown stress distribution and redistribution, crack
nucleation and initiation, stable and unstable crack propagation, interaction and coalescence, and
corresponding load-displacement curves can be proposed as benchmarks for numerical programs for crack
propagation. (5) In mechanical fragmentation, a crushed zone is always available near the tool. The crushed
zone has an important influence on the chipping process and energy utilization. The crushed zone is in fact
the zone with a high density of microcracks, and some of the rock in this zone behaves in a ductile manner
with stress satisfying the ductile failure surface of the double elliptic strength criterion. (6) A simple
description and qualitative model of the rock fragmentation process induced by truncated indenters are
summarized as follows. Little damage to the rock is observed at the linear-elastic deformation stage. Then
conical Hertzian cracks are initiated adjacent to both corners of the truncated indenter and propagated in the

III
well-known conical Hertzian manner. As the loading displacement increases, some of the elements in the
high confining pressure zone immediately under the indenter fail in the ductile cataclastic mode, with the
stress satisfying the ductile failure surface of the double elliptic strength criterion. With the cataclastic
failures and tensile conical cracks releasing the confining pressure, the elements in the confining pressure
zone are compressed into failure and the crushed zone comes into being. In the crushed zone, microcracking
is pervasive. The intensity of the microcracking within the zone increases and a re-compaction behaviour
occurs with increasing loading displacement. Associated with this microcracked region there is a volumetric
expansion and a tensile stress field, which drives side cracks to propagate in a curvilinear path. It is thought
that the curvilinear path is caused by the heterogeneity of the rock. With an increased loading displacement,
the side cracks rapidly propagate and intersect with the free rock surface to form rock chips. (7) The
confining pressure has an important influence on the failure mode in indentation tests. As the confining
pressure increases, a small but noticeable increase in the indentation strength is measured. With decreasing
confining pressure, of particular interest is a change in the rock failure mode when the confining stress is
reduced below a critical value. Instead of the usual formation of rock chips adjacent to the punch, vertical
cracks are propagated beneath the punch, causing the specimen to be split in half when the confining
pressure on the sample is less than a critical value. This result is of practical interest. In boring hard rock at
low confining stresses, the creation of such tensile fractures beneath an indenter may serve to fragment the
rock sufficiently to facilitate its removal. (8) The simultaneous loading of the rock surface by multiple
indenters seems to provide a possibility of forming larger rock chips, controlling the direction of subsurface
cracks and consuming a minimum total specific energy. (9) The results simulated by the R-T2D code
reproduce the progressive process of rock fragmentation under mechanical loading: the build-up of the
stress field, the formation of the crushed zone, surface chipping, and the formation of the crater and
subsurface cracks. Therefore, the R-T2D code is indeed a valuable numerical tool for research on rock
fracture which can be utilized to improve our understanding of rock-tool interaction and the rock failure
mechanisms under the action of mechanical tools, which, in turn, will be useful in assisting the design of
fragmentation equipment and fragmentation operations.
On the basis of the above research, a number of interesting problems are discussed and future research is
suggested in the last part of the thesis.

Key words: Rock fracture, numerical simulation, heterogeneity, fracture toughness, strength criterion,
mechanical fragmentation, cutting, indentation.

IV
The thesis comprises a summary and the following four papers.

Paper A Liu HY, Roquete M, Kou SQ, Lindqvist PA. Characterization of rock heterogeneity and its
numerical verification. To be submitted to Engineering Geology.
Paper B Liu HY, Kou SQ, Lindqvist PA. Numerical modelling of the progressive fracture processes of
heterogeneous rock material using various test techniques. To be submitted to Engineering
Fracture Mechanics.
Paper C Liu HY, Kou SQ, Lindqvist PA. Numerical simulation of the fracture process in cutting
heterogeneous brittle material. International Journal for Numerical and Analytical Methods in
Geomechanics 2002; 26: 1253-1278.
Paper D Liu HY, Kou SQ, Lindqvist PA, Tang CA. Numerical simulation of the rock fragmentation
process induced by indenters. International Journal of Rock Mechanics & Mining Sciences
2002; 39: 491-505.

V
 Table of Contents
1 Introduction ·············································································································································1

1.1 R & D for and calibration of the R-T2D code·····················································································1

1.2 Preliminary application of the R-T2D code in mechanical fragmentation ·········································2

2 R & D for and calibration of the R-T2D code ··························································································5

2.1 Characterization of rock heterogeneity ·····························································································5

2.2 Mesoscopic constitutive law, non-linear behaviour and associated seismicity ································ 9
2.3 Mohr-Coulomb and double elliptic strength criterion ······································································10
2.4 Calibration of the R-T2D code by simulating basic rock mechanics experiments ···························· 11

3 Preliminary application of the R-T2D code in mechanical fragmentation ·············································· 16

3.1 Numerical modelling of the fracture process in cutting heterogeneous brittle rock ·························16
3.1.1 Numerical model ······················································································································16
3.1.2 Quasi-photoelastic stress fringe pattern····················································································16
3.1.3 Chipping process ······················································································································17
3.1.4 Force-loading displacement response and associated seismicity ·············································20
3.1.5 Results for rock fragmentation by cutters with different back rake angles ····························· 20
3.2 Numerical modelling of the rock fragmentation process induced by indenters ······························· 24
3.2.1 Numerical model ······················································································································25
3.2.2 Quasi-photoelastic stress fringe pattern ···················································································25
3.2.3 Chipping process ······················································································································27
3.2.4 Force-penetration curve and associated seismicity ··································································27
3.2.5 Comparison between the numerical and the experimental results ···········································28
3.2.6 Results for rock fragmentation induced by double indenters ··················································29

4 Discussions ·············································································································································33

4.1 Influence of heterogeneity on rock behaviour ··················································································33

4.2 Physical-mechanical properties and fracture toughness ···································································34
4.3 Failure modes in pure mode-II and mixed-mode I-II loading ··························································35
4.4 Acoustic emission (AE) and the fracture process zone (FPZ) ··························································36
4.5 Crushed zone in mechanical fragmentation ······················································································37
4.6 Influence of confining pressure on the failure mode in indentation··················································38
4.7 Interaction between neighbouring indenters in multiple indentations···············································39

5 Conclusions··············································································································································42

6 Suggestion for future research·················································································································45

References······················································································································································46

Papers A-D

VI
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

1 Introduction

1.1 R & D for and calibration of the R-T2D code

The fracture and failure of rock have been the subject of extensive research in mechanical fragmentation,
the results of which have led to a number of comprehensive texts (Lindqvist, 1982; Whittaker et al, 1992;
Kou, 1995; Tan, 1996; Mishnaevsky, 1998). This extensive work has provided an understanding of rock as
a heterogeneous brittle material that fractures via the formation, growth and coalescence of microcracks.
Though experimental observations have provided a great deal of insight into the complicated fracture
process, the mechanism of rock fracture under mechanical loading is not clear and the evolution of the
progressive fracture process cannot be visually shown at this moment. Besides, it is too expensive to
conduct a large number of experiments. The use of linear-elastic fracture mechanics (LEFM) to describe
these mechanisms analytically is arduous (Schlangen and Garboczi, 1997), since the fracture patterns
consist of a main crack with various branches, secondary cracks and microcracks. Moreover, analytical
models have to be simplified and sometimes this simplification ignores important factors influencing the
material behaviour. Heterogeneity is an example of one such factor for rocks. During the past few years,
with the rapid development of computing power, interactive computer graphics and topological data
structure, numerical tools have become a good means of gaining some insight into the problem of rock
fracture.
Numerical simulation of the fracture process in rock requires robust numerical methods that can allow
the efficient resolution of multiple interacting cracks and rigorous fracture models that can reflect the
material fabric characteristics. In fact, a large number of numerical methods and fracture models have been
developed for research on the rock fracture process. The numerical methods most widely used for analysis
of the rock fracture process in rock engineering are the finite difference method (FDM), finite element
method (FEM), boundary element method (BEM) and discrete element method (DEM).
FLAC (Itasca, 1995) is the outstanding representative of the finite difference method (FDM). There are
nine basic constitutive models provided in FLAC, Version 3.3 (Itasca, 1995), arranged into null, elastic and
plastic model groups. The plastic model group contains some models, e.g. the Mohr-Coulomb model and
the strain-softening/hardening model, which allow rock fracture to be simulated. McKinnon and Barra
(1998) used the strain-softening constitutive law to model primary fracture initiation and growth with FLAC.
As for the finite element method (FEM), Saouma and Kleinosky (1984) simulated crack initiation and
subsequent crack propagation in the process of major chip formation with their finite element program –
SICRAP (simulation of crack propagation program). Cracker (Swenson and Ingraffea, 1988) is a finite
element code used for modelling the dynamic propagation of a discrete crack. FRAN2D (Wawryznek and
Ingraffea, 1989) is a two-dimensional crack propagation simulator using the finite element method.
Alehossein and Hood (1996) used DIANA (1993) and NUMA (1995), which are also finite element
methods, to simulate crack problems in indentation.
Recent advances in the boundary element method (BEM) have resulted in a new generation of the
method for the solution of crack problems in fracture mechanics. In the boundary element method, the
displacement discontinuity method (DDM) is mainly used to model fracture. Tan et al (1996) used the
DDM coupled with a splitting model to simulate side crack propagation in indentation. Blair and Cook
(1998) coupled DDM in their non-linear rule-based model to analyse compressive fracture in rock using
statistical techniques.
The formulation and development of the distinct element method have been progressing for a period of
more than 30 years, beginning with the initial presentation by Cundall (1971). The method was originally

1
Hongyuan Liu Luleå University of Technology

created as a two-dimensional representation of a jointed rock mass, but has been extended to applications in
studies on microscopic mechanisms in granular material (Cundall and Strack, 1983), and crack development
in rock and concrete (Plesha and Aifantis, 1983; Lorig and Cundall, 1987). The representative codes of the
distinct element method are UDEC (Cundall, 1980; Cundall and Hart, 1985), 3DEC (Cundall, 1988; Hart et
al., 1988), DIBS (Walton, 1980), 3DSHEAR (Walton et al., 1988), PFC (Itasca, 1995) and DDA (Shi, 1988).
Rock fracture models, such as the energy model (Glucklich, 1963; Karihaloo, 1984; Shen and
Stephansson, 1994), the stress model (Murrell and Digby, 1970; Paul and Mirandy, 1976; Tirosh and Catz,
1981), the lattice model (Herrmann and Roux, 1992; Schlangen and Van Mier, 1992; Schlangen and
Garboczi, 1997; Place and Mora, 1999), the RFPA model (Tang, 1997; Tang et al, 1998; Tang et al, 2000),
the cellular automaton model (Wilson et al, 1996; Psakhie et al., 1999) and so on, have been developed to
accompany the numerical methods. Wang and Shrive (1995) have provided a good detailed review and
summary of typical existing models for brittle fracture in compression. These models are versatile and able
to handle complex geometry and material behaviour. In these models, the behaviour of the materials, i.e. the
rocks, is usually assumed to range from simple linear-elastic to non-linear behaviour. The analyses range
from static and quasi-static to dynamic analyses.
The fracture process of heterogeneous rock and rocklike material is so complex that it is not surprising
that different numerical methods and fracture models have been developed. A number of idealizations have
been used in the above numerical methods and fracture models which may limit their correspondence to the
complex fracture processes occurring in the loading of heterogeneous brittle materials. Although all these
models have improved our understanding of rock fracture, the discrepancy between prediction and
experiments is still significant. At this stage of investigation, few of the above studies can visually simulate
the progressive fracture process including crack initiation, propagation, interaction and coalescence. The
most important factor of all is that most of them have not related their work to the uniquely heterogeneous
character of rock. Moreover, none of them can determine input parameters on the basis of the specific rock
studied.
It appears that the success of modelling brittle fracture in heterogeneous material depends on the full
understanding of fracture mechanisms, the soundness of a universal fracture criterion and the effectiveness
of the numerical technique used. At present the finite element method is the most mature method of the
above four numerical methods. Preliminary researches (Tang, 1997; Tang et al, 1998; Kou et al, 1999; Tang
et al, 2000; Liu et al, 2001a-c; Kou et al, 2001; Liu et al, 2002a-c) indicate that the RFPA (Rock Failure
Process Analysis) model (Tang 1997) is appropriate for research on the rock fracture problem. Therefore, in
this thesis a novel numerical program called the rock and tool interaction code (R-T2D) is developed on the
basis of the finite element method (FEM) and the RFPA model to investigate the rock fracture process
under mechanical loading.
This thesis is an extension of previous work conducted in China. The investigation dealing with the
research and development (R & D) for and the calibration of the R-T2D code consists of four parts: 1) the
characterization of rock heterogeneity; 2) mesoscopic constitutive laws, non-linear behaviour and associated
seismicity; 3) the Mohr-Coulomb and the double elliptic strength criterion; and 4) the calibration of the R-
T2D code by simulating basic rock mechanics tests.

1.2 Preliminary application of the R-T2D code in mechanical fragmentation

Rock fragmentation by mechanical tools is the preferred technique of effective rock removal because of
the continuity of the process, the directly fragmenting rock mode involved and the smaller energy loss
incurred. The technique is employed in one way or another in most of today’s mining and tunnelling
machines. As drilling and mining environments are becoming increasingly severe, improved understanding

2
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

of rock behaviour under mechanical loading and optimisation of cutter or drilling design are necessary in
order to hold down costs.
Rock cutting and indentation constitute the fundamental mechanisms for most mechanical excavation
methods. In rock cutting and indentation, a tool is forced into the rock, and the rock near the tool is crushed
into small or large fragments, leaving subsurface cracks in the remaining rock wall. Research analysing the
stress and failure occurring in mechanical fragmentation captures the basics and enables a more detailed
study of the physical process. The results of this type of work are important both for the understanding of
rock fragmentation and for the evaluation of the damage to the remaining rock. However, the analysis of
mechanical fragmentation has mostly concerned the tool penetration and the rock failure at the surface,
mainly because rocks are non-transparent and it is difficult to trace the propagation of the rock fracture and
fragmentation within the rock.
Lindqvist (1982), Cook et al (1984), Howarth and Bridge (1988) were among the first to conduct crack
discrimination. The most informative results were found in the work performed by Lindqvist et al (1994). A
general picture of fractures in rock under mechanical tools was summarized. It was found that rock
fragmentation by mechanical tools was mainly a chipping process. Therefore, it was of great interest to
follow up closely the process of chipping in order to achieve a better understanding of the mechanisms. In
order to conduct research on the chipping process in mechanical fragmentation, the theories of information,
fractals and fuzzy sets were introduced by Mishnaevsky (1998). However, at present the stress state and
crack propagation behaviour involved in the chipping process are too complex to be studied with a
theoretical analysis method. The failure process may, however, be traceable in detail by applying
appropriate numerical methods. Gradually mature computational mechanics theories and high-powered
personal computers have made this possible. From this point of view, there is a clear incentive to develop
numerical methods that can simulate the process of material failure under mechanical tools, e.g. cutters and
indenters. Numerical models that can accurately simulate rock-tool interaction and failure mechanisms,
including chip formation, can be used in parametric studies to determine the optimal tool (cutter or indenter)
designs, reducing much of the expensive and time-consuming experimental work which would otherwise
have to be carried out to assess the performance of a particular tool design.
As a matter of fact, several attempts have been made to simulate the rock fracture process under
mechanical tools with numerical methods. The three main numerical methods, the finite element method
(FEM), the boundary element method (BEM) and the discrete element method (DEM), have been used to
calculate the stress field and simulate the fracture under a mechanical tool by many researchers, such as
Lawn and Swain (1975), Lawn and Wilshaw (1975), Hood (1997), Cook et al (1984), Saouma and
Kleinosky (1984), Chen and Alehossein (1996), Alehossein and Hood (1996), and Tan et al (1997). These
simulations coupled with fracture models have been capable of detecting some mechanisms of rock
fragmentation under mechanical tools. In the simulations using an elastic model, the rock has been assumed
to behave as an ideal, isotropic elastic material. Due to the fact that in-elastic deformation within the
crushed and damaged rock has not been taken into account, the simulations have over-predicted the peak
load at the onset of chip formation (Chen and Alehossein, 1996) and ignored its effects on the indentation
stress field afterwards. In some of the above simulations, a blister model (Yoeff, 1982) has been introduced
to describe the crushed and damaged zone. However, not only is the shape of the crushed and damaged zone
uncertain, but also the blister model needs to be validated further. Moreover, most of the above works have
not taken the heterogeneity of rocks into consideration. The most important drawback of all is that none of
the above simulations has concerned the progressive formation process of the whole crack system in
mechanical fragmentation, and almost none of them has considered the interaction between the stress fields
induced by neighbouring tools in simultaneous loading.
At present numerical methods based on FEM coupled with a fracture model are promising methods for
simulating indentation-induced fracture. In this thesis the progressive failure process of rock under

3
Hongyuan Liu Luleå University of Technology

mechanical loading is simulated using a computer program called R-T2D (the rock and tool interaction code),
developed in the first part of this thesis, which is based on FEM and the RFPA model (Tang, 1997). The
simulated results reproduce all of the fragmentation processes in mechanical fragmentation: the build-up of
the stress field, the formation of the crushed zone, surface chipping, crater formation and the formation of
subsurface cracks. According to the simulated results, some qualitative and quantitative analyses are
performed.

4
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

2 R & D for and calibration of the R-T2D code

The numerical approach called R-T2D (the rock and tool interaction code) has been developed on the
basis of FEM (the finite element method) and the RFPA (rock failure process analysis) model. The basic
elements of the RFPA model can be summarized as follows (Tang, 1997): 1) by introducing the
heterogeneity of rock properties into the model, the RFPA model can simulate the non-linear deformation of
a quasi-brittle rock with an ideal brittle constitutive law for the local material; 2) by introducing a reduction
of the material parameters after element failure, the RFPA model can simulate strain-softening and
discontinuum mechanics problems in a continuum mechanics mode; 3) by recording the event-rate of failed
elements, the RFPA model can simulate seismicities associated with the progressive fracture process.

2.1 Characterization of rock heterogeneity

The actual goal of numerical simulation is to simulate fracture in a real material. Real materials, such as
rock, are in general random and inhomogeneous, and therefore their fracture will be influenced by the
heterogeneous microstructure of the material. Heterogeneity has to be implemented to simulate the fracture
process correctly. Different techniques have been used in the past to implement heterogeneity (Schlangen
and Garboczi, 1997). These techniques all have their own application scale for which they will give good
results. The different options include: 1) Randomly assigning different properties, possibly following some
kind of distribution, to the elements in the numerical model (Batrouni and Hansen, 1988; Tang, 1997); 2)
Using a mesh with a random geometry, but equal properties for the elements (Garboczi and Day, 1995); 3)
Generating a microstructure and projecting this on regular elements, assigning different properties to the
elements depending on their position (Schlangen and Van Mier, 1992; Schlangen and Garboczi, 1997); 4)
Using a combination of a random geometry and a generated grain structure (Bazant et al, 1990). However,
none of these techniques correlate their parameters describing heterogeneity with the practical physical-
mechanical parameters of the specific rock. To solve this problem, a heterogeneous material model is
proposed in Paper A.
A common building and decorative stone, granite, is used in Paper A as a representative of
heterogeneous rock. The specific granite used comes from Umeå in Sweden and is a fine- and even-grained
granite with a massive structure. In order to model the mechanical behaviour and the fracture process of the
real heterogeneous rock material, integrated image analysis and numerical modelling have been applied
fully on-line to simulate the fracture behaviour of the granite specimen: image acquisition, image processing,
meshing, and computation of stress, strains and displacements, and crack initiation, propagation,
coalescence and interaction. Fig. 1 shows the successive steps of image acquisition with a microscope,
automatic image analysis and numerical modelling for the representative volume element (RVE) from the
heterogeneous granite specimen. Then the homogenisation theory is used to construct a numerical model to
simulate the mechanical behaviour and fracture process under a uniaxial compressive load.
However, homogenisation modelling needs detailed microstructure observation. Moreover, because of
limited computer capacity and calculation speed, there is a scale problem when modelling large-scale
macroscopic specimens or rock masses in the field. Therefore, maybe it is better to use the statistical
method to build a numerical model on the basis of the statistical analysis of microstructures. Fig. 2 shows a
comparison between the force-displacement curves resulting from the homogenisation modelling and the
statistical modelling. As can be seen from the figure, the results from the homogenisation modelling and the
statistical modelling show an almost identical curve in the linear-elastic deformation stage and non-linear-

5
Hongyuan Liu Luleå University of Technology

elastic deformation stage before the peak force. Therefore, the two specimens have almost the same
macroscopic elastic modulus and compressive strength.

a) Micrograph of a representative
volume element (RVE) under a
polarized microscope

b) View after image processing,

segmentation and polygons

c) Numerical modelling of an
RVE built by the R-T2D code

Fig. 1 Successive steps of the integrated image analysis and numerical modelling

6
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

Statistical Modelling
1400 Homogenization Modelling
1200

1000

800
600

400

200

0
0 0.04 0.08 0.12 0.16 0.2
Fig. 2 Comparison between homogenization modelling and statistical modelling

On the basis of the comparison between the homogenisation modelling and the statistical modelling, it is
observed that, in order to deal with real random microstructures in numerical simulation, rock heterogeneity
can be characterized better with statistical approaches. Here, the Weibull statistical distribution (Weibull,
1951; Hudson and Fairhurst, 1969) is used to characterize the rock heterogeneity. The two-parameter
Weibull distribution can be expressed as follows:
 mσ m −1   σ m 
 exp −   , σ ≥ 0
p(σ ) =  σ 0m   σ 0   (1)

 0, σ < 0
where,
m = The shape parameter describing the scatter of σ
σ0 = The scale parameter
The mathematical expectation and dispersion of the two-parameter Weibull distribution are respectively:
 1
E (σ ) = σ 0 Γ1 +  (2)
 m
  2  1 
D(σ ) = σ 02 Γ1 +  − Γ 2 1 +  (3)
  m  m 
where Γ is the Gamma function. Fig. 3 shows the relationship between the scaled mathematical expectation
E (σ ) / σ 0 , the dispersion D(σ ) / σ 02 and the shape parameter m. As can be seen from the figure, with the
shape parameter m increasing, the scaled mathematical expectation approaches 1, i.e. the scale parameter σ0
is approximately the mean value of the Weibull distribution. When m trends to infinity, the mathematical
dispersion, i.e. the variance, trends to zero. Therefore, in order to describe the heterogeneity of rock, we can
define the shape parameter m as the homogeneous index and the scale parameter σ0 as the seed parameter.
In numerical simulation, we build a two-dimensional element network model, in which the elemental

7
 Hongyuan Liu Luleå University of Technology

parameters, such as the critical strength σ c , the elastic modulus E c , etc, follow the Weibull distribution law
with the homogeneous index m and seed parameters σ 0 , E 0 , etc. Fig. 4 shows the Weibull distribution with
the typical homogeneous index m and the seed parameter σ0 = 200.

Scaled Dispersion: D(σ)/σ02

Scaled expectation: E(σ)/σ0

1.00 1.0

0.98
0.8

0.96
0.6
0.94
0.4
0.92

0.2
0.90

0.88 0.0

0 50 100 150 200 0 50 100 150 200

Shape parameter: m Shape parameter: m

Fig. 3 Scaled mathematical expectation and dispersion of the Weibull distribution with different
shape parameters
Q(σ)
P(σ)

σ0 m=1
0.01 m=1.1 1
0.008 m=1.5 0.8 m=1
m=2 m=1.1
0.006 m=3 0.6 m=1.5
m=4 m=2
0.004 m=5 0.4
m=3
0.002 0.2 m=4
0 σ 0 m=5
σ
0 100 200 300 400 500 0 100 200 300 400 500

Fig. 4 Probability density for typical homogeneous Fig. 5 Cumulative distribution of the Weibull
indices distribution

Then a sample space is formed after we specify these elemental parameters following the Weibull
distribution. Moreover, even if the parameters of the elements have the same distribution function, the
spatial distribution may be different. Rocks consisting of these elements with the same characteristic
parameters have almost the same macroscopic mechanical behaviour, e.g. peak strength. However, they
may have different failure patterns. The great difference lies in the details of the mesoscopic structure,
which represent the disorder of the mesoscopic structure. The special discrete characteristics of
heterogeneous rock or rocklike brittle material are exactly reflected by the disorder of the mesoscopic
structures. Usually the disorder of the probability distribution in physical space can be achieved by the
Monte Carlo method. The simplest method is to generate a series of random data which have a uniform
distribution between 0 and 1. Because the Weibull distribution is non-monotone, the integral distribution
function is derived as follows:

8
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

σ   σ m 
Q(σ ) = ∫ ϕ ( x)dx = 1 − exp −    (4)
0
  σ 0  
The random data between 0 and 1 which are generated by the Monte Carlo method have certain values in
the y-coordinate of the integral distribution function, as shown in Fig. 5. Then the corresponding values in
the x-coordinate can be found, which are the values that we are looking for. Therefore, a series of random
data map a series of element parameters, which can be given to elements in a finite element network. The
method based on statistics and randomicity satisfies the requirements of heterogeneity and randomicity of
element parameters in a finite element network of rock material.
Therefore, on the basis of the Weibull statistical distribution, we characterize the heterogeneity of rock
as the following two types of characteristic parameters: (1) the homogeneous index m and (2) the elemental
seed parameters, i.e. the mean value of the main physical-mechanical parameters (critical strength, σ0,
elastic modulus, E0, etc) of the elements. For any specific rock, the homogeneous index can be determined
on the basis of the defect distribution of the microstructure and the elemental seed parameters can be
obtained from laboratory tests.

2.2 Mesoscopic constitutive law, non-linear behaviour and associated seismicity

According to the RFPA model (Tang, 1997), heterogeneity is the source of non-linearity. It has been
shown that the global non-linear behaviour observed in brittle rock can be simulated with brittle-elastic
elements if heterogeneity is considered (Tang et al, 2000). Therefore, in R-T2D, linear mesoscopic
constitutive laws have been introduced for all the elements, and the elements are given different strength
and elastic constant parameters depending on the heterogeneity of the rock materials. This makes the model
simple but successful in simulating non-linear rock behaviour and the associated seismic events. During the
loading process, the finite element method is used to compute the stress and deformation in each element,
and the strength criterion is used to examine whether or not the elements undergo a phase transition. An
external load is slowly applied on the constructed model step by step. When in a certain step the stresses in
some elements satisfy the strength criterion, the elements are damaged and become weak according to the
rules specified by the strength criterion. The stress and deformation distributions throughout the model are
then adjusted instantaneously after each rupture to reach the equilibrium state. At positions with an
increased stress due to stress redistribution, the stress may exceed the critical value and further ruptures are
caused. The process is repeated until no failure elements are present. The external load is then increased
further. In this way the system develops a macroscopic fracture. Thus the code links the micromechanical
model to the continuum damage model and ultimately to macrostructure failure. Energy is stored in the
element during the loading process and is released as acoustic emissions (AE) through the onset of
elemental failure. Due to the interaction induced by stress redistribution and long-range deformation, a
single important element failure may cause an avalanche of additional failures in neighbouring elements,
leading to a chain reaction releasing more energies.
Accordingly, R-T2D, a Microsoft Windows application, is a package for analysis of the progress of rock
failure under mechanical loading. It has been developed by considering the deformation of an elastic
material containing an initial random distribution of micro-features. As the load is applied, the fractures will
grow, interact and coalesce. As a result, the behaviour becomes non-linear and in addition macroscopic
fractures are formed. A user-friendly pre- and post-processor is integrated to generate the finite element
mesh and prepare the input data. The analysed results may be displayed graphically as an “animation” to aid
users in understanding the mechanics of the failure process.

9
Hongyuan Liu Luleå University of Technology

2.3 Mohr-Coulomb and double elliptic strength criterion

The Mohr-Coulomb criterion and the Hoek-Brown criterion are the two most popular strength criteria in
mining and geotechnical engineering. The RPFA model coupled with these criteria has successfully
investigated some problems (Tang et al, 1998; Liu, 2000; Fu, 2000; Zhu, 2000; Tang et al, 2000 a-d; Liu et
al, 2001 a-c; Tang et al, 2001; Kou et al, 2001; Yang, 2001). However, the traditional Mohr-Coulomb
criterion and Hoek-Brown strength criterion are only valid for the brittle part of the rock failure envelope
(Verhoef and Ockeloen, 1996; Zhao, 2000). Rock cutting and indentation experiments have shown that very
high confining pressures exist near the tips of the tools, maybe with values above the brittle-ductile
transition stress known from triaxial tests (Lindqvist and Lai, 1983; Verhoef and Ockeloen, 1996). In this
case, a cap strength criterion may be an appropriate choice. In fact, constitutive equations under the
category of cap plasticity models have been formulated by numerous authors (Faruque, 1987). However,
most of them are based on soil or concrete and may not be applicable to rock.

τg

τ g = kσ g
(σ ,τ )
c
g
c
g
(σ g* ,τ g* )

σ gt σ gp 0
σg

Fig. 6 Double elliptic strength criterion with generalized shear and normal stress representation

In the current simulations, as shown in Fig. 6, a double elliptic cap model, proposed by Yu (1998), is
adopted. The double elliptic cap model can be expressed in terms of the generalized shear stress
τ g = (σ 1 − σ 3 ) / 2 , where σ 1 and σ 3 are the maximum and minimum principal stress respectively, and the
generalized normal stress σ g = (σ 1 + σ 3 ) / 2 as follows:
 a σ 2 + 4τ 2 + bσ = c
g g g
 ' (5)
a σ g + 4τ g + b σ g = c
2 2 '

where a, b, a’, b’ and c are constant parameters. The constant parameters can be defined according to the
uniaxial tensile strength σ t , the uniaxial compressive strength σ c , the transition point ( σ g* ,τ g* ) from brittle
to ductile failure and the hydrostatic pressure p0 , as shown in Fig. 6. Hoek and Brown (1980) chose the
relationship σ 1 = 3.4σ 3 ( τ g = kσ g , where k = 6 / 11 , in a generalized shear and normal stress presentation)

10
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

as the best approximation of the brittle-ductile transition. Therefore, the limiting stress conditions can be
represented as the following two groups of equations.
 a σ 2 + 4τ 2 + bσ = c,τ ≥ kσ
g g g g g
 ' (6)
a σ g + 4τ g + b σ g = c,τ g < kσ g
2 2 '

The brittle failure face of the double elliptic criterion will represent rock failure in the shear mode or tensile
mode, just as the modified Mohr-Coulomb (including tensile cut-off) or Hoek-Brown strength criterion at a
low confining pressure. The ductile failure surface will represent rock failure in the ductile cataclastic mode
at a high confining pressure. The major advantages of the proposed model are that the double elliptic
strength criterion avoids the tip angle in the tension area in the linear failure criterion, and the failure
criterion and yield cap model have a unified mathematical form, in which the ductile yield face is
interrelated with the brittle failure surface.

2.4 Calibration of the R-T2D code by simulating basic rock mechanics experiments

In Section 2.1, heterogeneous rock material has been characterized according to a few characteristic
parameters: the elemental seed parameters (the critical strength σ0 or elastic modulus E0 of mesoscopic
elements, etc) and the homogeneous index m. For any specific rock, the seed parameters can be obtained
from the main physical-mechanical parameters in laboratory tests and the homogeneous index can be
determined on the basis of the defect distribution of the microstructure (Orlovskaja et al, 1997; Curtis and
Juszczyk, 1998; Davies, 2001). Therefore, we can use these characteristic parameters as the input
parameters of R-T2D to simulate the fracture behaviour of heterogeneous rock materials.
In Paper B, a series of numerical specimens with the same seed parameters – e.g. critical strength (σ0 =
200 MPa) and elastic modulus (E0 = 60 GPa), etc – but different homogeneous indices, m, which vary
between 1.1 and 200, are numerically uniaxially compressed. The geometry of the constructed specimens
follows that suggested by ISRM (International Society of Rock Mechanics, 1979) for the uniaxial
compressive strength test, i.e. the diameter is 54 mm and the ratio between the height and the diameter is 2.5.
Besides the above characteristic parameters, before performing finite element analysis, some other
parameters are needed to specify the rock, such as the Poisson’s ratio (0.25) and the ratio (10%) between the
tensile strength and the compressive strength of the element.
Dimensionless modulus: Ec/E0
Dimensionless strength: σc/σ0

1.0 1.1

1.0
0.8

0.9
Data: Data1_B Data: Data1_B
0.6 Model: ExpDecay1 Model: ExpDecay1
Chi^2 = 0.00104 Chi^2 = 0.00023
y0 0.85928 ±0.01349 0.8 y0 1.02453 ±0.0052
x0 0 ±0 x0 0 ±0
0.4 A1 -0.80668 ±0.0257 A1 -0.62081 ±0.03225
t1 10.68877 ±1.05482 0.7 t1 2.59074 ±0.19259

0.2
0.6

0.0 0.5
0 50 100 150 200 0 50 100 150 200

Homogeneous index: m Homogeneous index: m

Fig. 7 Scaled macroscopic strength and elastic modulus corresponding to different homogeneous indices

11
Hongyuan Liu Luleå University of Technology

The simulated macroscopic specimen parameters corresponding to the characteristic parameters of

mesoscopic elements are shown in Fig. 7. From the figure, we can observe that the dimensionless uniaxial
compressive strength and dimensionless elastic modulus approach 1 with the homogeneous index increasing.
In other words, the simulated uniaxial compressive strength and elastic modulus of the specimen approach
the specified seed parameters of the elements. As can be seen from the figure, with a high homogeneous
index, m (larger than 200), our heterogeneous model becomes the traditional homogeneous model. If we fit
the curves with the appropriate function, we can obtain the following two empirical formulas representing
the relationship between the macroscopic parameters (compressive strength, σc, elastic modulus, Ec) of the
specimen and the characteristic parameters (homogeneous index, m, critical strength, σ0 and elastic modulus,
E0) of the mesoscopic elements:
σc  m 
= 0.85928 − 0.80668 exp −  (7)
σ0  10.68877 
Ec  m 
= 1.02453 − 0.62081exp −  (8)
E0  2.59074 
where σc and Ec are the simulated uniaxial compressive strength and elastic modulus, respectively, σ0 and
E0 are the characteristic compressive strength and characteristic elastic modulus of the mesoscopic elements,
and m is the homogeneous index. The uniaxial compressive strength and elastic modulus corresponding to
any specific rock in practice can be obtained from laboratory tests, and the homogeneous index can be
determined on the basis of the defect distribution of the microstructure (Orlovskaja et al, 1997; Curtis and
Juszczyk, 1998; Davies, 2001). Therefore, we can use these two empirical formulas to determine the input
parameters (the characteristic parameters of elements) for the R-T2D software package.
In the laboratory, basic rock mechanics experiments are conducted to obtain the physical-mechanical
properties and fracture toughness of rock materials. In Paper B, a great variety of numerical specimens for
the uniaxial compressive strength (UCS) test, the Brazilian tensile strength (BTS) test, diametral
compression of a notched Brazilian disc (NBD), and the three-point bending (3PB) and four-point shearing
(4PS) tests are constructed with the same characteristic parameters as those defined in Section 2.1 to
simulate the progressive fracture process and determine the local physical-mechanical properties and
fracture toughness. In building these models, the following main characteristic parameters of specimens are
used: the homogeneous index is m = 2, the compressive strength is σc = 180 MPa and the elastic modulus is
Ec = 60 GPa. We further assume here that the Poisson’s ratio is 0.25 and a fixed ratio (0.1) exists between
the tensile strength and the uniaxial compressive strength of the rocks. On the basis of these parameters, we
can use granite as the reference rock, whose main physical-mechanical parameters and fracture toughness
are shown in Table 1.

Table 1 The physical-mechanical properties and fracture toughness of the reference granite a

Parameters σc σt E KIC KIIC

(MPa) (MPa) (GPa) (Mpa.m1/2) (Mpa.m1/2)
Rock Type
Reference Granite 175 15.6 62.5 1.72 1.75
a
Newhurst granite in Whittaker et al (1992)

12
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

Table 2 Final failure modes simulated by the R-T2D code and a comparison with the experimental (or
theoretical) results

Test Numerically Experimental (or Theoretical) Results

Type Simulated Results
UCS

BTS

(Gomez et al, 2001)

NBD-I

(Jia et al, 1996) (Chen et al, 1998)

NBD-II

(Jia et al, 1996) (Chen et al, 1998)

13
Hongyuan Liu Luleå University of Technology

Table 2 (Continued)

Test Numerically Simulated Results Experimental (or Theoretical) Results

Type
NBD-
Mixed

S3PB

(Landis, 1999)
4PB

(Rao, 1999)
A3PB

L1
(Xeidakis, 1997)

Table 2 shows the final failure modes in the various test techniques and a comparison between the
numerically simulated results and the experimental (or theoretical) results. To investigate the fracture
problem, fracture mechanics needs a well pre-fabricated notch. There is no need for a pre-fabricated notch
in the R-T2D code, although it also works very well when there is a well pre-fabricated notch. Rock
heterogeneity acts as a stress concentrator under loading. A crack is created at a point where the principal
stresses reach the tensile strength or the compressive strength. Therefore, the code can be applied to un-
notched test geometries such as the geometry of the uniaxial compressive test, the Brazilian test and so on.
Of course, it also works perfectly when there is a pre-fabricated notch. In Paper B we used it to study the
diametral compression of a notched Brazilian disc (NBD) specimen, the bending and shearing loading of a
notched beam. The detailed visually shown stress distribution and redistribution, crack nucleation and
initiation, stable and unstable crack propagation, interaction and coalescence, and corresponding load-

14
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

displacement curves are found to be in good agreement with experimental results in the literature, and can
be proposed as benchmarks for numerical programs for crack propagation.

Table 3 Simulated physical-mechanical parameters and fracture toughness

Parameters Mixed-mode
σc σt E KIC KIIC
Tests (MPa) (MPa) (GPa) (Mpa.m1/2) (Mpa.m1/2) KIC KIIC
(Mpa.m1/2) (Mpa.m1/2)
UCS 183.5 59.5
BTS 18.7 1.74
NBD-I 1.70
NBD-II 2.78
NBD-Mixed 1.68 2.72
3PB-I 3.74
4PB-II 2.40
3PB-Mixed 4.11 0.66

Table 3 summarizes the simulated physical-mechanical parameters and fracture toughness for the
numerical specimens with the same characteristic parameters. The results obtained using the different
methods can be compared with each other. For example, the ratio between the uniaxial compressive strength
and the tensile strength is σc/σt =183.5/18.7 = 9.8, which is consistent with the statement that the ratio
varies between 8 and 12 for most rocks. The mode-I fracture toughness values obtained from the BTS test
(1.74), the NBD test in pure mode I (1.70) and the NBD test in mixed mode I-II (1.68) are close to each
other. The ratio between the mode-II fracture toughness and the mode-I fracture toughness is 2.78/1.70 =
1.64, which is consistent with most of the experimental results, which show that the pure mode-II fracture
toughness is usually larger than the pure mode-I fracture toughness (Ingraffea, 1981; Whittaker et al, 1992;
Al-Shayea et al, 2000). The slight difference was caused by the spatial distribution of heterogeneity, and the
effects of size, shape, thickness, notch length and so on. Moreover, most of the simulated physical-
mechanical parameters and fracture toughness values are close to those of the reference granite, whose
physical-mechanical properties and fracture toughness were previously shown in Table 1. On the basis of
the different test methods and a large number of obtained data, we can now be confident that rock can be
specified according to the characteristic parameters, the built heterogeneous model is reasonable and the
developed R-T2D code is a valuable numerical tool for research on the progressive fracture process of
heterogeneous material. Besides, from Table 3, we have observed that the fracture toughness data from
diametral compression of the NBD specimen show little variance compared with those from the bending
and shearing tests. Therefore, taking the size effects, simple configuration and the loading procedure into
consideration, the disc type specimen may be more useful in measuring fracture toughness.

15
Hongyuan Liu Luleå University of Technology

3 Preliminary application of the R-T2D code in mechanical fragmentation

Mechanical fragmentation is the preferred technique of effective rock removal because of the continuity
of the process, the directly fragmenting rock mode involved and the smaller energy loss incurred. The
technique is employed in one way or another in most of today’s mining and tunnelling machines. In Paper C
and D, cutting and indentation, the fundamental mechanisms in mechanical fragmentation, have been
investigated using the R-T2D code.

3.1 Numerical modelling of the fracture process in cutting heterogeneous brittle rock

Numerical modelling of the fracture process in cutting heterogeneous brittle rock is conducted in Paper
C. Here, just the results are reported.

3.1.1 Numerical model

Rock cutting is simplified to a two-dimensional plane strain problem, as shown in Fig. 8. The gray
scales represent different values of Young’s modulus. The brighter the colour, the higher is the value of the
Young’s modulus of the element. The steel cutter is modelled as a quasi-isotropic homogeneous material
with a high strength compared with the rock, whereas the rock is modelled as a heterogeneous material. An
external displacement in the horizontal direction (0.005 mm/step) is applied to the steel cutter from the left
side, and the upper side is kept at the same horizontal level without any movement in the vertical direction.
The whole model is divided into 250 × 125 (= 31,250) mesoscopic finite elements.

Fig. 8 Model for numerical simulation of rock cutting

3.1.2 Quasi-photoelastic stress fringe pattern

In the initial stage of cutting rock, the rock is deformed as an elastic body. Therefore, before performing
failure analysis, the quasi-photoelastic stress fringe patterns induced in the cutter and homogeneous and
heterogeneous rocks are simulated, as shown in Fig. 9. It is apparent that the stress fields in rock consist of
three zones: the interaction zone, the confining pressure zone and the zone outside the confining pressure
zone. The interaction zone is caused by a mismatch between the elastic moduli of the rock and cutter. The
rock in the confining pressure zone is generally in a state of rather high tri-axial compression. In the post-
failure analyses, it is found that the confining pressure zone defines, to a great extent, the shape and size of

16
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

the so-called crushed zone. The stress contours in the cutter, which are caused by the reaction from the rock,
are also clearly shown in the fringe patterns. The high stress concentration in the corners often causes tools
to break or wear. The stress distribution coupled with the heterogeneity will determine the sites of crack
initiation and the directions of crack propagation in the following failure process.

Fig. 9 Quasi-photoelastic stress fringe patterns in a cutter and rock

3.1.3 Chipping process

Fig. 10 shows the development of the chipping process and the progressive evolution of the stress fields
as the cutting displacement increases. When the cutter is applied on the rock, a high stress concentration is
induced in the rock near the cutter, and the rock far from the cutter is almost unaffected. As shown in Fig.
10 (a), cracks sprout first at the lower edge of the cutter and propagate, dipping into the rock at
approximately 450 in the direction of the cutter displacement. The mechanism of the calculated cracks is
tensile failure. That is because the tensile strength of brittle material is far lower than the compressive
strength. Ahead of the cutter, the rock is highly stressed and it is in a triaxial state due to the high confining
pressure. Therefore, the rock in this zone does not fail at the initial stage of cutting. As the cutter
displacement increases, tensile cracks sprout at the upper edges of the cutter, and at the same time the cracks
initiated from the lower edge of the cutter propagate downward, as shown in Fig. 10 (b). With continuous
loading displacement, compressive failure occurs ahead of the confining pressure zone, as shown in Fig. 10
(c, d). With the tensile failure and compressive failure releasing the confinement for the tri-axial
compressive zone, the rock in the initial confining pressure zone is compressed into failure and the crushed
zone gradually comes into being with a dimension similar to that of the cutter-rock interface, as shown in
Fig. 10 (e). According to experimental results (Verhoef and Ockeloen, 1996), due to the very high confining
pressure, the stresses of the elements in the confining zone may fail in the ductile cataclastic mode with the
stresses satisfying the ductile face of the double elliptic strength criterion. However, in the present
simulation, the stresses of few elements satisfy the ductile failure surface of the double elliptic strength
criterion. Maybe this is because the depth of the cut is not great enough or there is no confinement on the
rock. From the simulated results, it is found that in fact the crushed zone is the zone with a high density of
microcracks, and it is not necessary to consider the crushed zone as some separate body which interacts with
the tool and the rock massif. This kind of investigation has also been conducted by other researchers
(Kutuzov et al, 1969; Simonov and Vyskrebtsov, 1975; Mishnaevsky, 1999).
The crushed zone has an important influence on the subsequent development of the crack system, which
changes the transferring direction of the force applied by the cutter. As shown in Fig. 10 (a-d), before the
formation of the crushed zone, all of the cracks show a tendency to dip down into the rock. Associated with

17
Hongyuan Liu Luleå University of Technology

a. Cutter displacement: 0.07 mm f. Cutter displacement: 0.145 mm

b. Cutter displacement: 0.09 mm g. Cutter displacement: 0.19 mm

c. Cutter displacement: 0.12 mm h. Cutter displacement: 0.26 mm

d. Cutter displacement: 0.125 mm i. Cutter displacement: 0.38 mm

e. Cutter displacement: 0.13 mm j. Cutter displacement: 0.5 mm

Fig. 10 Fracture process in cutting (maximum principal stress distribution)

18
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

the crushed zone there must be a volumetric expansion and a tensile stress field, which result in tensile crack
propagation and the formation of chipping cracks. Therefore, after the formation of the crushed zone, two
main categories of cracks, chipping cracks and subsurface cracks, come into being, as shown in Fig. 10 (e).
As the displacement increases, one of the chipping cracks rapidly intersects with the free surface and small
pieces of rock are chipped, as shown in Fig. 10 (f). At the same time the other main chipping crack
propagates in a curvilinear path but approximately parallel to the free surface of the rock. The subsurface
cracks propagate in a curvilinear path to dip into the rock at a certain angle in the direction of the cutter. It is
thought that the curvilinear path is caused by the heterogeneity of the rock. The mechanism of the chipping
crack is tensile failure mainly induced by tensile stress associated with the crushed zone. With the
continuous increase in the loading displacement, the main chipping crack, which is driven by tensile stress,
propagates and bifurcates as shown in Fig. 10 (g). As the loading displacement increases, the bifurcated
chipping cracks propagate and interact as shown in Fig. 10 (h). At the same time, more and more cracks
form around the crushed zone, which makes the crushed zone enlarge and its boundary become vague. With
a continuous increase in the loading displacement, the bifurcated chipping cracks coalesce and form a single
major chipping crack again, as shown in Fig. 10 (i). As the chipping cracks propagate and intersect with the
free surface, the major chip is expected to come into being as shown in Fig. 10 (j). The cracks which
propagate to dip into the rock form the subsurface cracks after the formation of the major chip.
According to the simulated results, when cutting heterogeneous brittle material, the process of chip
formation is caused by a complicated stress state, with the elements mainly failing in a tensile mode or a
mixed shear and tensile mode and sub-fracture occurring in a dominant tensile mode. The subsurface
fractures are actually the bifurcation of some major tensile fracture. Therefore, in cutting brittle materials, as
differentiated from cutting plastic materials, the mechanism of chip removal is not shear, but tensile or
mixed-mode crack growth. Similar phenomena were also noted by Mishnaevsky (1998).

1500 c e
Normalized Load (N)

b d X direction
Y direction
1000 a
f g j (i) Load-displacement
h i
500

0
0 0.1 0.2 0.3 0.4 0.5
Normalized Displacement (mm)
800

600
AE counts

400
(ii) AE-displacement
200

0
0 0.1 0.2 0.3 0.4 0.5
Normalized Displacement (mm)
Fig. 11 Load-displacement curve and associated seismicity (letters corresponding to plates in Fig. 10)

19
Hongyuan Liu Luleå University of Technology

3.1.4 Force-loading displacement response and associated seismicity

The load displacement response, as shown in Fig. 11 (i), is actually the reflection of the fractures ahead
of the cutter. On the basis of the assumption that there is consistency between the damage and the seismicity
associated with the fracture process, the acoustic emission (AE) phenomenon during the fracture process is
also observed, as shown in Fig. 11 (ii). From the two curves, we know that the load displacement response
is therefore closely related to the fractures in the rock induced by the cutter. This can be made clear when
we compare Fig. 11 (i, ii) with Fig. 10. The increasing load at the very beginning causes quasi-elastic
deformation of the rock surface, as shown in Fig. 10 and 11 marked by (a). At this stage, the load
displacement relation is quasi-linear and, correspondingly, there is almost no acoustic emission in the rock,
except for that caused by a very small amount of damage occurring. This quasi-elastic deformation is then
followed by the deformation of the crushed zone and cracked zones. At this stage, the movement of the rock
fragments in the crushed zone is restrained by the surrounding intact rock. In order to break the rock,
continuous loading is required. This would cause intense comminution of the rock fragments if the
confinement were sufficiently high. However, in the present case the moving tendency of the rock
fragments in the crushed and cracked zone pushes the surrounding rock upwards, due to a smaller
confinement on the upper side, and results in chipping cracks. During the process of rock fracturing the
cutter gradually penetrates the rock. When the chipping cracks reach the rock surface, the surrounding rock,
which offered the upper confinement at the earlier stage, together with part of the rock underneath the cutter,
moves away forming the chips. The movement of the cutter meets less resistance in this case. The force
acting on the cutter is thus reduced to a rather low value, as shown in Fig. 11, and the cutter penetrates the
rock to a greater depth with a decreasing load. After that the rock fragments can be removed with a very
small load.
It is clear from the curve of the AE counts that this load drop is caused by a sudden increase in the AE
counts (or the number of failed elements). Generally, every big increase in the AE counts results in a big
drop. The load-displacement curve for the cutting process obtained from the simulation shows that the
simulated chipping process is similar to the chipping process found through experimental observation of
brittle cutting (Deketh et al, 1998). It is also somewhat similar to the results obtained from indentation
experiments carried out in the laboratory (Kou et al, 1996). In addition, from the above-described simulation,
it is easy to understand that the cutting depth is mainly dependent on the load magnitude, the macro-
mechanical properties of the rock, and the heterogeneity of the rock. However, it is also influenced by the
cutter geometry, which characterises the load distribution.

3.1.5 Results for rock fragmentation by cutters with different back rake angles

The stress fringe patterns induced by cutters with different back rake angles (200, 00 and -200) are shown
in Fig. 12. As can be seen in the fringe patterns, the far field stress distributions are almost the same for the
three cutters. This mechanical behaviour is in accordance with the well-known Saint Venant principle.
However, each of the three cutters exhibits a different stress distribution in an area immediately adjacent to
the cutter. Finally, this difference in the stress distribution causes a completely different cutting efficiency.
This further indicates the importance of stress redistribution during the failure process in cutting and the
importance of the crushed zone.
Fig. 13 shows the final failure patterns induced by cutters with different rake angles. Fig. 14 shows the
corresponding elastic energy release (ENR) and accumulated ENR in the crushed zones and the whole
specimens. According to the load-displacement curves (Fig. 11), the work necessary for fragmenting the
rock can be calculated for the three different rake angles. According to the elastic energy release (ENR-)
displacement curves (Fig. 14), the useful energy can be calculated for the three different rake angles.

20
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

According to the ENR-displacement curves in the crushed zone (Fig. 14), the energy consumed by the
formation of the crushed zone can be calculated. From the final fracture pictures (Fig. 13), the area of rock
spalled by the chipping crack can be calculated. According to the simulated results, the energy utilisation
ratios are 14%, 18% and 21% respectively for the 200, 00 and –200 rake angles, which are greater than the
experimental observation results, because the influences of the loading rate, friction, temperature, etc. are
not considered. The energies consumed by the formation of the crushed zone are 41%, 71% and 63% of the
useful energies respectively, which are close to the experimental observation results. The specific energies
for chipping are 199, 123 and 49 kJ/m3 respectively for the three cutters with rake angles of 200, 00 and –200.
When comparing the three different rake angles, it is reasonable to conclude that a rock cutter with a rake
angle of –20 degrees is the most efficient of the three cutters. This is a reasonable conclusion because the
wedging action of sharp cutters produces tensile stresses in the rock in a fairly direct manner. This
advantage of a sharp cutter begins to disappear, however, as the geometry of these tools changes from a
sharp to a blunt shape because of friction. For applications that require the machining of stronger rocks, it is
often necessary to increase the wedge angle of a drag bit in order to increase the tool strength.

(i) Rake angle: 20 degrees

(ii) Rake angle: 0 degrees

(iii) Rake angle: -20 degrees

Fig. 12 Quasi-photoelastic stress fringe patterns induced by cutters with different rake angles

21
Hongyuan Liu Luleå University of Technology

(i) Rake angle: 20 degrees

(ii) Rake angle: 0 degrees

(iii) Rake angle: -20 degrees

Fig.13 Failure patterns with different rake angles

ENR
0.020 Accumulated ENR 0.10
Accumulated ENR (N)

0.08
0.015
ENR (N)

0.06
0.010 i. ENR-displacement in whole
0.04 specimen (Rake angle: 200)
0.005
0.02

0.000 0.00
0 0.1 0.2 0.3 0.4 0.5
Displacement (mm)

Fig. 14 Elastic energy release (ENR), accumulated ENR and displacement curve

22
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading
ENR
0.020 Accumulated ENR 0.10

Accumulated ENR (J)

0.08
0.015
ENR (J)

0.06
0.010 ii. ENR-displacement in crushed
0.04 zone (Rake angle: 200)
0.005
0.02

0.000 0.00
0 0.1 0.2 0.3 0.4 0.5
Displacement (mm)

ENR
0.020 Accumulated ENR 0.10

Accumulated ENR (J)

0.08
0.015
ENR (J)

0.06 iii. ENR-displacement in whole

0.010 specimen (Rake angle: 00)
0.04
0.005
0.02

0.000 0.00
0 0.1 0.2 0.3 0.4 0.5
Displacement (mm)
ENR
0.020 Accumulated ENR 0.10
Accumulated ENR (N)

0.08
0.015
ENR (N)

0.06 iv. ENR-displacement in crushed

0.010
0.04 zone (Rake angle: 00)
0.005
0.02

0.000 0.00
0 0.1 0.2 0.3 0.4 0.5
Displacement (mm)

ENR
0.020 Accumulated ENR 0.10
Accumulated ENR (J)

0.08
0.015
ENR (J)

0.06
0.010
0.04 v. ENR-displacement in crushed
0.005
zone (Rake angle: -200)
0.02

0.000 0.00
0 0.1 0.2 0.3 0.4 0.5
Displacement (mm)
Fig. 14 (Continued)

23
Hongyuan Liu Luleå University of Technology

ENR
0.020 Accumulated ENR 0.10

Accumulated ENR (J)

0.08
0.015
ENR (J)

0.06
0.010 vi. ENR-displacement in whole
0.04
specimen (Rake angle: -200)
0.005
0.02

0.000 0.00
0 0.1 0.2 0.3 0.4 0.5
Displacement (mm)

Fig. 14 (Continued)

3.2 Numerical modelling of the rock fragmentation process induced by indenters

The rock fragmentation process induced by a single indenter and multiple indenters is investigated in
Paper D. Here, just the results are reported.

Displacement: 0.005mm/step

a
σconf. = 20 MPa

σconf. = 20 MPa

Displacement: 0.005mm/step

b
σconf. = 20 MPa

σconf. = 20 MPa

Fig. 15 Numerical simulation models: a) Single indenter, b) Double indenters

24
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

3.2.1 Numerical model

The indentation problem is simplified to a plain strain condition. A plane passing through the central
axis of the indenter is considered, as shown in Fig. 15 (a) and (b). The grey scales represent different values
of Young’s modulus. The brighter the colour, the higher is the value of the Young’s modulus of the element.
The indenter is simulated as a homogeneous material whose elastic modulus is a few times higher than that
of rock in order to safeguard against permanent deformation of the indenter. The rock specimen is
considered as a heterogeneous material. Both the indenter and the rock are divided into many mesoscopic
elements. In the case of the single indenter, the model is discretized into 150*250 = 37,500 elements. In the
case of simultaneous loading with double indenters, the model is divided into 150*300 = 45,000 elements.
A displacement increment (0.005 mm/step) is applied on the indenters and a confining pressure (20 MPa)
acts on both the lateral sides of the constructed rock specimen.

3.2.2 Quasi-photoelastic stress fringe pattern

The first requirement of any soundly-based simulation of indentation fracture is a detailed knowledge of
the stress field within the loaded system. Our foremost aim here is to investigate the distribution of the
elastic stress component primarily responsible for the operation of fracture processes. At the same time, it is
also important that we should pay special attention to the distribution of the hydrostatic or high confining
pressure zone, for this zone will determine the extent of the irreversible deformation within the field. The
solutions for the stress field arising when an isotropic, linear-elastic half-space is subjected to a normal
point load were first given in 1885 by Boussinesq, who invented the famous Boussinesq elastic field. The
stress field arising when an isotropic, linear-elastic half-space is subjected to normal loading by a smooth
spherical indenter is called the ideal Hertzian elastic field. Fig. 16 shows the simulated quasi-photoelastic
stress fringe pattern induced by a single indenter when the rock is considered as homogeneous material.
According to the simulated results, the indentation stress field is not uniform. The stresses are extremely
large close to the loading point and decrease rapidly with increasing distance from the loading point. The
stress distributions are symmetrical about the axis. A very high stress field is induced in the regions both
immediately underneath the punch and on either side of the punch. Lindqvist (1982) analyzed the one-point
stress field presented by Lawn and Swain (1975) and used it to predict the cone crack and median crack.
However, Lindqvist (1975) pointed out that the Boussinesq elastic stress field gives no evidence of the
evolution of tensile surface (lateral) cracks creating the indentation crater, which is considered to be
responsible for rock fragmentation in indentation. Therefore, it is necessary for us to perform failure
analysis in the following.

Fig. 16 Simulated quasi-photoelastic stress fringe pattern caused by a truncated indenter

25
Hongyuan Liu Luleå University of Technology

A F

B G

C H

D I

E J

Fig. 17 Simulated results for the rock fragmentation process induced by a single indenter (Stress distribution)

26
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

3.2.3 Chipping process

When the indenter acts on the rock, a high stress zone, which corresponds to the highlight zone in Fig.
17, comes into being immediately beneath the indenter. A fan-shaped stress field is radiated outside the
highly stressed zone. Far from the highly stressed zone, the stress field is like water-waves due to the
heterogeneity of the rock and the confining pressure, as shown in Fig. 17 (A). As the stress intensity builds
up with an increasing load, one or more of the flaws nucleates a crack around the two corners of the
truncated indenter. It is interesting to find that, although the rock immediately beneath the indenter is highly
stressed, it does not fail primarily because of the high confining pressure. On the contrary, cracks initiate
first on both corners of the truncated indenter to form cone cracks. The cone cracks lose their symmetrical
shapes because of the rock heterogeneity, as shown in Fig. 17 (B). With the loading displacement increasing,
the cone cracks driven by tensile stresses run downward along the stress trajectories of the maximum
principal stresses in the well-known conical Hertzian mode. At the same time, due to increasing stress, the
elements immediately beneath the indenter fail. Some of them fail, even if there is a high confining pressure,
in the ductile cataclastic mode with the stresses satisfying the ductile failure surface of the double elliptic
strength criterion. Others are compressed into failure because the formation of cone cracks and ductile
cataclastic failure release the confining pressure, as shown in Fig. 17 (C). The crushed zone gradually
comes into being as the elements in the high confining pressure zone fail. The formation of the crushed zone
has an important influence on the direction of the cracks. Before the formation of the crushed zone, all of
the cracks propagate downwards, which is expected to form cone cracks or subsurface cracks, as shown in
Fig. 17 (B, C). During the formation of the crushed zone, some cracks bifurcated from the cone cracks
propagate almost parallel to the free surface and some new cracks are initiated from the crushed zone, as
shown in Fig. 17 (D). This is because the crushed zone has a changeable shape and volume. Yoffe (1982)
provided a blister model to explain this behaviour. With the loading displacement increasing, a re-
compaction of the crushed zone occurs. During the re-compacting process, the crushed zone expands to
both sides, which drives the cracks initiated from the crushed zone or bifurcated from the cone cracks to
propagate approximately parallel to the free surface expected to form side cracks. Because of the
heterogeneity, the side cracks on both sides have no symmetric shapes and propagate in a curvilinear path.
In this simulation, the propagating velocity of the side cracks on the right side is faster than that on the left
side, as shown in Fig. 17 (E). As the loading displacement increases, the side cracks on the left side, which
are dormant at first, begin to propagate forward. At the same time, the side cracks on the right side
propagate stably, as shown in Fig. 17 (F). With the loading displacement increasing, the side cracks on both
sides of the indenter propagate stably and almost parallel to the free surface, but in a curvilinear path. In
addition, some cracks propagate downwards at an angle of about 45 degrees to form cone cracks. Some
bifurcated cracks propagate downwards and are expected to form subsurface cracks, as shown in Fig. 17 (G).
As the penetration displacement increases, the side cracks on both sides of the indenter propagate stably to
form almost symmetrical shapes. At the same time, some discrete cracks initiate under the crushed zone and
are expected to form median cracks, as shown in Fig. 17 (H). With the loading displacement increasing, the
side crack on the right side of the indenter accelerates and propagates unstably to form chips, as shown in
Fig. 17 (I). As the loading displacement increases, the side crack on the left side of the indenter is expected
to form chips also. At the same time, more discrete cracks initiate under the indenter. Some of them
coalesce to form subsurface cracks, as shown in Fig. 17 (J).

3.2.4 Force-penetration curve and associated seismicity

Fig. 18 shows the simulated force-penetration curve and associated seismicity during the rock
fragmentation process induced by a single indenter. The lateral pressure gradually acts on both boundaries

27
Hongyuan Liu Luleå University of Technology

of the rock specimen. At point A1 in the force-penetration curve, the lateral pressure attains its maximum
value: 20 MPa. The letters A-J correspond to the slides marked with the letters A-J in Fig. 17. The
associated seismicity, which is usually called the acoustic emission (AE) phenomenon in rock mechanics, is
obtained by recording the counts of the failure elements. From the picture it can be seen that, although there
are microcracks, as shown by the AE counts, the force-penetration curve has almost a linear shape in the
initial loading stage, i.e. the curve between point (A1) and point (A), which is the linear-elastic deformation
stage of the load displacement curve. As shown in Fig. 17 (A), little damage to the rock occurs until this
point. With the loading displacement increasing, the force-penetration curve attains its first peak value:
point (B). As shown in Fig. 17 (B), at this point a conical Hertzian crack is initiated at the punch corners. As
the elements immediately beneath the indenter fail, the force-penetration curve falls off to its first trough at
point (C), where the crushed zone comes into being. As the loading displacement increases, a re-compaction
behaviour occurs in the crushed zone immediately beneath the indenter, and the loading displacement curve
climbs up. As shown in Fig. 18 from point (D) to point (I), small corrugations are the main characteristic of
the force-penetration curve. The small corrugated variations in the force, which occur prior to a major
breakout (chipping), are indications of the propagation of cracks, the crushing of microstructural grains and
the formation of small chips. The large drop in the force occurs after point (I). At this time, a major chip
forms on the right side of the indenter, which indicates that a substantial part of the rock has now been
separated or unloaded. Further penetration by the indenter repeats the peak-trough behaviour in a cyclic
pattern as shown in Fig. 18 between point (I) and point (J). Another major chip on the left side of the
indenter is expected to form. Accompanying the major chip, there is also a large acoustic emission (AE)
count. As shown in Fig. 18, the small corrugations and large peak-trough are the two main features of the
force-penetration curve. The peak force for the next chip is normally greater, due to an increase in both the
depth of penetration and the contact area of indentation. It is found that this corrugated peak-trough (saw-
tooth) behaviour is typical for most indenter types.

350 6000
AE Counts J
AE Counts (N)

300 Force I 5000

FG
Force (N)

250 E
C 4000
200 B
H 3000
150 D
A 2000
100
50 1000
A1
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Displacement (mm)
Fig. 18 Force-penetration curve during rock fragmentation by single indenter

3.2.5 Comparison between the numerical and the experimental results

Fig. 19 shows a comparison between the experimental and the numerically simulated results. After the
comparison, a general picture of fractures in rock under indentation can be summarized. Underneath the
indenter there is a crater, a disintegrated and sometimes partly re-compacted zone, a crushed but basically
confined zone, and a cracked zone. Outside the cracked zone there are mainly three kinds of cracks: median
cracks, radial cracks, and side cracks according to their positions related to the indenter. The crater is
formed due to tool penetration and rock removal. The re-compacted zone lies immediately under the

28
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

indenter and is caused by the re-compaction of the crushed rock. The crushed zone is in fact the zone with a
high density of microcracks, which make the boundary between the crushed zone and the crack zone
become vague. The median cracks and radial cracks are not obvious in some cases. Side cracks are initiated
in the vicinity of the crushed zone or bifurcated from the cone cracks and propagate nearly parallel to the
free surface, and side cracks are mainly responsible for rock fragmentation in indentation.

Numerical result (R-T2D)

Experimental result

Fig. 19 Comparison between the numerically simulated result and the experimental result

3.2.6 Results for rock fragmentation induced by double indenters

Fig. 20 (A) shows the simulated quasi-photoelastic stress fringe pattern induced by double indenters
when rock is regarded as a homogeneous material. Comparing Fig. 16 with Fig. 20, we will find that there
are some new features in the stress contours induced by double indenters. According to the simulated elastic
stress field caused by a single indenter as shown in Fig. 16, this kind of stress field is expected to form cone
cracks or subsurface cracks. However, there is no evidence of the formation of side cracks, which are
considered to be responsible for the formation of chipping in the rock fragmentation process in indentation.
As shown in Fig. 20 (A), the directions of the maximum principal stresses seem to be influenced to a small
extent immediately near the indenters. The interaction of the principal stress induced by each of the two
indenters is clear between the two indenters. The stress trajectories change their directions in the rock
between the indenters. Rock chips are expected to form between the two indenters. With the distance from
the indenters increasing, the stresses slowly converge to a combined field. In the far region the stress field
appears to originate from a load distributed over the area between the two indenters, giving stress directions
similar to the stress field induced by a single indenter, as shown in Fig. 16.

29
Hongyuan Liu Luleå University of Technology

a)

b)

Fig. 20 Quasi-photoelastic stress fringe patterns induced by double truncated indenters

a) Homogeneous rock specimen; b) Heterogeneous rock specimen

Fig. 21 shows the rock fracture process induced by double indenters. At the first stage of loading, the
respective stress fields induced by the double indenters are equal to those induced by the single indenter.
The rocks immediately under the indenters are highly stressed. The stresses decrease rapidly with increasing
distance from the loading points. However, with increasing loading displacement, the interference of the
two stress fields induced by the two indenters is more and more obvious in the rock between the two
indenters, as shown in Fig. 21 (A). Cone cracks are initiated first around the two corners of both indenters,
as shown in Fig. 21 (B). As shown in Fig. 21 (C), before the formation of the crushed zone, the cone cracks
bifurcate and propagate approximately parallel to the free surface expected to form side cracks. Because of
the heterogeneity, the crack systems under the two indenters are not completely identical. However,
similarities exist between the two crack systems. As the loading displacement increases, the crushed zones
gradually come into being immediately under the indenter, and the side cracks are driven by tensile stresses
associated with the expansions of the crushed zone, as shown in Fig. 8 (D, E). With the loading
displacement increasing, more and more side cracks are initiated from the two crushed zones, as shown in
Fig. 21 (F). With a continuous loading displacement, the side cracks propagate stably forward and the
subsurface cracks are richly developed under the indenters. Because of the heterogeneity, the side cracks
propagate in a curvilinear path. As shown in Fig. 21 (G), the interactions between the side cracks are more
and more obvious as the tips of the side cracks become closer and closer. With the increasing loading
displacement, the side cracks coalesce in a complicated manner and the rock between the two indenters is
chipped, as shown in Fig. 21 (H). The chipped rock has a complicated geometrical shape. As the loading
displacement increases, the crack systems are well developed under the two indenters, as shown in Fig. 21
(I). More cracks are initiated from the crushed zone. Some of them propagate to form side cracks and some
of them dip into the rock forming subsurface cracks. At the same time, some discrete cracks are initiated
under the crushed zones to form radial and median cracks. With increasing loading displacement, more and
more side cracks interact and coalesce. Therefore, after the major chips, some small chips occur as shown in
Fig. 21 (J).

30
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

A F

B G

C H

D I

E J

Fig. 21 Simulated result for the rock fragmentation process induced by double indenters (stress distribution)
31
Hongyuan Liu Luleå University of Technology

Fig. 22 shows the recorded force-penetration curve associated with the fragmentation process induced
by double indenters, and this curve is similar to that for the fragmentation process induced by a single
indenter. This means that almost the same amount of energy is consumed. However, the chipped rock
volume is larger than that in the fragmentation induced by a single indenter. Therefore, simultaneous
loading of the rock surface by multiple indenters with an appropriate line spacing seems to provide a
possibility of forming larger rock chips, controlling the direction of subsurface cracks and consuming a
minimum total specific energy.

350 12000
AE Counts
300 Force 10000
I
AE Counts (N)

250 H 8000

Force (N)
G
DE F
200
B J 6000
150 A
4000
100 C
50 2000
A1
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Displacement (mm)
Fig. 22 Force-penetration curve during rock fragmentation by double indenters (The
force is the resultant force on the two indenters.)

32
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

4 Discussions

4.1 Influence of heterogeneity on rock behaviour

According to the proposed heterogeneous model, the macro specimen parameters and the elemental seed
parameters have been correlated with the homogeneous index through Eqs. 7 and 8. We have defined the
homogeneous index and the elemental seed parameters as characteristic parameters. In loading numerical
specimens with the same characteristic parameters using the same testing method, similar resultant stress-
strain curves are observed. However, the final fracture patterns may be different depending on the spatial
random distribution. For any specific rock, we can conduct a UCS and a BTS test to obtain the uniaxial
compressive strength, elastic modulus and so on. The homogeneous index can be determined on the basis of
the defect distribution of the microstructure. Therefore, the proposed heterogeneous model is reasonable and
applicable.
220

200
m=1.2
180
m=1.5
160 m=2
m=3
140 m=5
Stress (MPa)

120

100

80

60

40

20

0
0 0.1 0.2 0.3 0.4 0.5
Loading displacement (mm)
Fig. 23 Stress-displacement curve induced in specimens with the same mean values but
different homogeneity

In loading a numerical specimen with a lower homogeneous index, there is a big difference between the
macro parameters and the elemental seed parameters. This is reasonable, since elements with a lower
strength will fail earlier during loading. The resultant stress-strain curve has a typical linear-elastic
deformation stage, a non-linear deformation stage, a post-failure stage and a residual strength stage, as

33
Hongyuan Liu Luleå University of Technology

shown in Fig. 23. The higher the value of the homogeneous index is, the higher are the macro parameters of
the specimen. When the homogeneous index is very high (larger than 200), the heterogeneous model will
become the traditional homogeneous material model. As a result, the curve becomes more linear, the
strength loss is sharper and no obvious nucleation sites of failure events are observed. Therefore, the R-T2D
code, through the inclusion of rock heterogeneity, can simulate the complete stress-strain curve and
associated seismicity.
In a series of numerical experiments conducted in Paper B, corresponding to every heterogeneous
specimen, a homogeneous specimen with the same geometry is built to obtain the quasi-photoelastic fringe
patterns. The stress distribution and redistribution in the heterogeneous specimen are shown in the
subsequent fracture process analysis. It is found that rock heterogeneity has a strong influence on the stress
distribution and re-distribution, which results in a tortuous crack propagation path. In front of the crack
propagation direction, if the strength of the element is not so high, the crack will propagate straight, which
corresponds to the transgranular failure observed with SEM in experiments. If the strength of the elements is
very high, the crack will propagate round elements with a high strength and a tortuous crack propagation
path comes into being, which can be considered as the intergranular failure observed with SEM in
experiments. Therefore, the R-T2D code can simulate the tortuous crack propagation path which is caused by
rock heterogeneity.

4.2 Physical-mechanical properties and fracture toughness

Rock failure results from the propagation of one or more cracks. In theoretical studies and engineering
applications, rock failure is investigated from two points of view: strength theory and fracture mechanics. It
has long been known that there is a quantitative relationship between the physical-mechanical parameters
and fracture toughness of rock. For example, Whittaker et al (1992) obtained some approximate relations
between fracture toughness and tensile strength, compressive strength, point load strength, hardness, and the
velocity of acoustic waves in rock on the basis of experimental data from many references. However, those
formulas have very low regression coefficients. Recently, Zhang (2002) obtained an empirical relation
between the mode-I fracture toughness and the tensile strength of rock with a higher regression coefficient.
However, this kind of relation does not help us understand the mechanism occurring in nature.
Usually the notched specimen is studied through the principles of fracture mechanics and the intact rock
specimen is often studied using conventional rock mechanics principles. In the series of numerical
experiments mentioned above, we have simulated the fracture processes of both intact rock and notched rock,
although the R-T2D code uses characterized physical-mechanical parameters as input parameters. It is found
that R-T2D can also simulate very well the fracture toughness and the progressive fracture process of the
notched specimen in terms of fracture mechanics. The simulated physical-mechanical parameters and
fracture toughness for the numerical specimens with the same characteristic parameters in Paper B are
summarized in Table 3. The numerical results obtained using the different methods are comparable with
each other. For example, the ratio between the uniaxial compressive strength and the tensile strength is σc/σt
=183.5/18.7 = 9.8, which is consistent with the statement that the ratio varies between 8 and 12 for most
rocks. The mode-I fracture toughness obtained from the BTS test is 1.74, that from the NBD test in pure
mode I is 1.70, and that from the NBD test in mixed mode I-II is 1.68. The ratio between the mode-II
fracture toughness and the mode-I fracture toughness is 2.78/1.70 = 1.64, which is consistent with most of
the experimental results, which indicate that the pure mode-II fracture toughness is usually larger than the
pure mode-I fracture toughness (Whittaker et al, 1992; Al-Shayea et al, 2000; Ingraffea, 1981). The slight
difference was caused by the spatial distribution of heterogeneity, and the effects of size, shape, thickness,

34
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

notch length and so on. Moreover, most of the simulated physical-mechanical parameters and fracture
toughness values are close to those of the reference granite, whose physical-mechanical properties and
fracture toughness were previously shown in Table 1. Besides, the comparison between the results from
diametral compression of NBD specimens and those from the bending and shearing tests shows that the disc
type specimens are more suitable for measuring the fracture toughness of rock.

Table 4 Predicted results for fracture toughness based on physical-mechanical properties

Parameters Values from Adopted Equationsa Input values

relationships
KIC=1.81 K IC = 0.708 + 0.006σ c
K vs σc KIIC=1.03 K IIC = 0.114 + 0.005σ c σc=183.5
KIC=2.27 K IC = 0.27 + 0.107σ t
K vs σt KIIC=1.66 K IIC = 0.05 + 0.086σ t σt=18.7
KIC=1.88 K IC = 0.336 + 0.026 E
K vs E KIIC=1.32 K IIC = 0.251 + 0.018 E E=59.5

a
Adopted equations in Whittaker et al (1992)

In the following, the simulated results are used to check the empirical relationship between the physical-
mechanical properties and fracture toughness proposed by Whittaker et al (1992), as shown in Table 4,
although these formulas have very low regression coefficients. In the table, the symbols are defined as
follows: KIC is the mode-I fracture toughness, KIIC is the mode-II fracture toughness, σc is the uniaxial
compressive strength, σt is the tensile strength and E is the elastic modulus. It is found that the predicted
fracture toughness data are far more reasonable compared with those of the reference granite and our
numerical simulations. Moreover, all of the predicted values for the mode-II fracture toughness based on the
relations are smaller than the values for the mode-I fracture toughness, which is contradicted by most of the
experimental results, which usually indicate that the mode-II fracture toughness is two or three times higher
than the mode-I fracture toughness (Rao, 1999). Therefore, a prediction based on these empirical
relationships is not reliable. Considering the difficulty in determining the fracture toughness of rock in the
laboratory, maybe it is a good option to obtain the fracture toughness by using a numerical method after we
have determined the physical-mechanical properties. The developed R-T2D code builds a bridge between the
fracture toughness and the physical-mechanical properties of rock.

4.3 Failure modes in pure mode-II and mixed-mode I-II loading

Classical fracture mechanics defines three basic modes of fracture, i.e. mode I (opening), mode II
(sliding) and mode III (tearing). In our research, we find that in mixed-mode I-II loading and even in pure
mode-II loading, tensile failure is the overwhelming failure mode, as shown in Paper B. The same
phenomenon has also been observed in many examples of experimental research (Tapponnier and Brace,
1976; Swartz and Taha, 1990; Whittaker et al, 1992; Schlangen and Van Mier, 1992; Rao, 1999). On the
basis of observations using the scanning electron microscope (SEM), Tapponnier and Brace (1976)
conclude that most microcracks are tensile in nature.

35
Hongyuan Liu Luleå University of Technology

However, Rao (1999) argued that classical fracture mechanics defines three basic modes of fracture from
the loading point of view. It is assumed that the fracture propagates in the pre-fabricated crack plane. That is
true for most metal materials. For heterogeneous rock materials, however, shear loading on a pre-fabricated
crack is not certain to result in mode-II fracture, i.e. shear fracture propagating in the pre-fabricated crack
plane, because of the rock heterogeneity and the high ratio between the compressive strength and the tensile
strength. Traditional testing methods for KIIC (NBD and 4PS tests), which have been developed on the basis
of the shear loading rather than the mode-II fracture mechanism, usually lead to KIIC values of rock smaller
than or close to KIC values (Ingraffea, 1981; Huang and Wang, 1985; Whittaker et al, 1992). Obviously, it is
contradictory to fact that the shear strength of rock should be much larger than its tensile strength. Our
numerical analyses also indicate that, even under pure mode-II loading, tensile stresses inevitably exist
around the notch tips in the same order of magnitude as compressive stresses and cause crack propagation in
the direction parallel to the maximum principal stress. Similar results are found in the literature (Swartz and
Taha, 1990; Schlangen and Van Mier, 1992). The crack develops in the direction perpendicular to the
maximum tensile stress, which is characteristic of mode-I fracture rather than mode-II fracture. Maybe that
is why none of these tests has been accepted as a standard testing method for determining the KIIC of rock.

4.4 Acoustic emission (AE) and the fracture process zone (FPZ)

Linear-elastic fracture mechanics (LEFM) is usually used to describe the fracture extension and the
fracture mechanism under essentially linear-elastic (i.e. ideally brittle) conditions. However, during a crack
propagation process, ahead of the crack tip there is always a non-linear zone which is referred to as the
fracture process zone (FPZ) in rock. LEFM is usually invalid in this case, unless the fracture process zone is
so small that the condition of small-scale yielding (SSY) is satisfied. Therefore, it is important to understand
the mechanisms that develop the fracture process zone and what happens in it. Mihashi (1987) considers
that around the branches of the main crack there are closed microcracks and that in front of them opening
microcracks exist. Wittmann (1992) separated an inner microcracking zone from a surrounding isolated
microcracking zone. Moreover, Otsuka and Date (2000) summarized that these concepts are valuable, but
did not define the fracture process zone completely.
The load-displacement relationships measured in the notched specimen are shown in Fig. 18, Fig. 24,
Fig. 30, Fig. 34, Fig. 41 and Fig. 48 in Paper B. AE inspections were made at the lettered points, A, B, C, etc,
on the curves. In the figures, the number of failure events and the number of accumulated AE events are also
plotted. After the maximum load, the failure events increase rapidly. The measured AE source locations in
Paper B show that AE events are observed around the notch tip even in the first loading period. The AE
events at other places are caused by heterogeneity and the loading configuration. The majority of AE events
are located in the period around the peak load and show a tendency to concentrate around the notch tip.
After this period, the AE map shows a more widely spread distribution. Moreover, the energy released in the
failure of elements can be calculated approximately according to the element strengths, and then the
magnitude of the energy release can be calculated. In the above-mentioned figures in Paper B, the red circles
represent the seismicities caused by tensile failure in the current loading step, the green circles represent the
seismicities caused by compressive failure, and the black circles record the accumulative seismicities in the
previous steps. The diameters of the circles represent the relative magnitudes of the AE released energy. It is
obvious that the higher energy events tend to localize near the notch tip. As a result, it is considered that
these events are related directly to the fracture of rock and that this area associated with most of the energies
could be called an FPZ. In this area, more densely distributed events are observed. Our numerically
simulated results are consistent with the results which were observed by Otsuka and Date (2000) through an

36
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

AE technique and which indicate that the microcrack zone corresponds to an FPZ or at least must be a main
part of an FPZ.

4.5 Crushed zone in mechanical fragmentation

According to the simulated results in Paper C, Paper D and the review by Mishnaevsky (1995), a zone of
highly fractured and inelastically deformed rock, which is usually called the crushed zone, is always
available near the tools in mechanical fragmentation. The crushed zone has an important influence on the
chipping process and energy utilization. In this zone microcracking is pervasive and rocks behave in a
ductile manner. Associated with this microcracked or dilatant region there must be a volumetric expansion,
which results in tensile crack propagation to the surface to form rock chips. The ductile behaviour of brittle
materials under a sufficiently high confining pressure has also been observed by many researchers
(Lindqvist and Lai, 1983; Mishnaevsky, 1995). However, in mechanical fragmentation the very nature of the
inelastic deformation zone remains a highly contentious issue. It is not a simple matter to establish which of
the two basic and competing processes, shear-induced flow (either plastic or viscous) or pressure-induced
densification (compaction of an “open” microstructure), dominates within the small contact zone in rock
materials. Yoffe (1982) provided a blister model to describe the behaviour of the crushed zone. In our
simulations, it can be shown that in the regions immediately underneath the punch, the elements in the rock
specimen are generally in a state of triaxial compression. According to Cook et al’s (1984) simulation using
the Mohr-Coulomb criterion, it is interesting to note that immediately beneath the punch the rock does not
fail because of confining pressure. However, according to the experiments conducted by Lindqvist and Lai
(1983) and Cook et al (1984), the compressive strength of the rocks in the regions beneath the punch has
been reached and therefore failure occurs.
In order to assess the significance of the state of stress at a given point underneath the indenter, the
double elliptic strength criterion is adopted in our simulations. This assessment is achieved by the
computation of the Mohr circle from the principal stresses at that point. When this circle falls outside the
brittle failure envelope of the double elliptic criterion, the rock at that point is considered to have exceeded
the strength of the rock and brittle failure occurs. When this circle falls outside the ductile yield envelope of
the double elliptic criterion, the rock at that point is considered to have exceeded the strength of the rock and
ductile cataclastic failure occurs. According to the simulation, it is found that the stresses of some elements
immediately beneath the indenter satisfy the ductile failure surface and the elements fail in the ductile
cataclastic failure mode, which contributes to the formation of the crushed zone. With increasing loading
displacement, a re-compaction behaviour occurs in the crushed zone immediately beneath the indenter.
However, in rock cutting simulation, the stresses of few elements satisfy the ductile failure surface of the
double elliptic strength criterion. Maybe this is because the depth of the cut is not great enough or there is no
confinement on the rock.
According to previous research in mechanical rock fragmentation, about 70% (Artsimovich et al, 1978)
to 85% (Protassov, 1985) of the energy is consumed by the formation of the crushed zone. Lindqvist (1982)
has shown that the energy loss associated with friction, microcracking and so on concerns 89% of the
transmitted energy in disc cutting, whereas the useful energy is about 2-3%. In rock cutting simulation, the
effective energy utilized to fragment the rock is 21%. The energy consumed by the crushed zone is 63%.
The calculated result for the energy ratio consumed by the crushed zone is close to the results of previous
observations. However, in our calculation, the useful energy is larger than the observed result. Maybe this is
because the simulation does not consider the influence of the loading rate, temperature and friction, etc.

37
Hongyuan Liu Luleå University of Technology

5MPa
6000
10MPa
5000 20MPa
Force (N)
4000
3000
2000
1000
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Displacement (mm)
Fig. 24 Force-penetration curve with different confining pressures
σconf. = 10 MPa

σconf. = 10 MPa

Fig. 25 Failure mode of a rock specimen under low confining pressure

4.6 Influence of confining pressure on the failure mode in indentation

During the research in Paper D, we found that the lateral pressure had an important influence on the
formation of side cracks, which are considered to be responsible for the rock fragmentation in indentation.
The force-penetration curves for the three different lateral pressure values, 5, 10 and 20 MPa, are shown in
Fig. 24. This figure shows a small but noticeable increase in the indentation strength with increased
confinement. With decreasing confining pressure, decreases in the indentation strength are not of great
interest, which can be observed in many tri-axial tests. Of particular interest is a change in the rock failure
process when the confining stress is reduced below a critical value. Instead of the usual formation of rock
chips adjacent to the punch, vertical cracks are propagated beneath the punch, causing the specimen to be
split when the confining pressure on the sample is less than 20 MPa. Fig. 25 shows the simulated result
when the lateral pressure remains 10 MPa.
Cook et al (1984) conducted a series of indentation tests using flat-bottomed circular punches to load the
flat surfaces of the rock sample and presented similar results. In the punching tests at low confining stresses,
the specimens failed in tension across a plane passing through the axis of the punch. Such failures may be a
result of the finite size of the laboratory specimens and may not occur if a semi-infinite surface of rock is
loaded by the punch. However, this result is of practical interest. In boring hard rock at low confining
stresses, the creation of such tensile fractures beneath an indenter may serve to fragment the rock
sufficiently to facilitate its removal. According to the simulation, it is evident that at a low confining stress

38
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

the region of tensile failure extends completely beneath the punch. As the confining stress increases, tensile
failure is restricted to a region adjacent to the punch corners, and with the confining pressure decreasing, this
region extends downwards in an approximately vertical direction as the punch load increases. The numerical
simulation is validated by the experiments conducted by Cook et al (1984).

Fig. 26 Simulated quasi-photoelastic stress fringe patterns induced by double truncated indenters

4.7 Interaction between neighbouring indenters in multiple indentations

During simultaneous loading by multiple indenters, the rock between the indenters can be additionally
chipped owing to the interaction between neighbouring systems of cracks. This means that there is a
possibility of increasing the removing capacity of a multi-indenter tool by using the interaction between the
long cracks. In simultaneous loading by multiple indenters, the separation between the centre-lines of
adjacent indenters is defined as the line spacing (S). This line spacing is very important in the design and
operation of machines for rock drilling because of the interaction between adjacent side splays of chipping
grooves. It seems that in simultaneous loading with an appropriate line spacing, the interaction of the stress
fields induced by two indenters gives a possible combined crack at great depth to form large rock chips. Fig.

39
Hongyuan Liu Luleå University of Technology

26 shows the simulated quasi-photoelastic fringe patterns induced by double indenters with different line
spacings (S=18a, S=12a and S=6a). A comparison of these pictures shows that when the line spacing is
small, as shown in Fig. 26 C, the stress distributions induced by the double indenters are similar to those
caused by a single indenter. As the line spacing increases, the indenters will act independently in the zones
adjacent to the indenters and interact with each other in the zone between the indenters (Fig. 26 B). With the
appropriate line spacing, as shown in Fig. 26 A, the interaction between the two indenters gives a possible
combined crack at great depth to form large rock chips.

d1=18mm

d2=12mm

d3=6mm

Fig. 27 Rock fragmentation induced by double indenters with different line spacings

Fig. 27 shows the final failure modes corresponding to the line spacings shown in Fig. 26. With a small
line spacing, the coalescence occurs at a shallower depth and the fragment will therefore be smaller. With a
large line spacing, there will be no crack coalescence and the multiple indenters act on the rocks separately.
With the appropriate line spacing, the crack systems induced by the two indenters will interact with each
other and the long cracks will finally coalesce to give a possible combined crack at great depth to form large
rock chips. Using the blister model (Yoeff, 1982) to describe the behaviour of the crushed zones, by placing
the initial side crack seeds at specific locations and in specific directions, Tan et al (1997) also simulated the

40
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

coalescence of side cracks in rock fragmentation by two indenters simultaneously. They concluded that only
cracks at specific locations and orientations can generate crack coalescence, while others will have no
development or miss each other. Moreover, the maximum spacing found is about 5d, where d is the diameter
of the crushed zone according to the blister model.
10000
S = 6a
8000 S = 12a
S = 18a
6000
Force (N)

4000

2000

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Loading displacement (mm)
Fig. 28 Force-penetration curve in simultaneous loading by double indenters with different line spacings

Fig. 28 shows the simulated force-penetration curves corresponding to the three different line spacings
(S=6a, 12a and 18a). The curves have almost the same shape, which means that the work performed by the
indenters in the three cases is almost the same. However, the chipped depth and the volumes of the detached
rock in the three cases are different, as shown in Fig. 27. This means that during indentation by multiple
indenters, the volume of detached material can be increased by the interaction of crack systems without an
increase in the loading energy; i.e. the specific energy can be smaller. This effect is caused by the fact that
long cracks in two crack systems induced by the indenters can join together due to the interaction of their
stress fields. Therefore, the simultaneous loading of the rock surface by multiple indenters with the
appropriate line spacing seems to provide a possibility of forming larger rock chips, controlling the direction
of subsurface cracks and consuming a minimum total specific energy.
In simultaneous loading, the line spacing between the indenters can be formulated as follows:
S = 2l + 2 K s a (9)
where S is the line spacing, l is the linear length of the long side crack, a is the indenter diameter and Ks is
the coefficient of the indenter shape: Ks =1 for a blunt indenter (Kou, 1995), with which the cracks are
nucleated near the side faces of the indenter; Ks=0.8 (Mishnaevsky, 1998) for a spherical indenter, with
which the cracks are initiated in the rock near 0.8a; and Ks=0 (Kou, 1995) for a sharp indenter, with which
the cracks are nucleated near the indenter axis. Moreover, the chipped depth d in the rock between the two
indenters is equal to the maximum depth of the side crack propagation.

41
Hongyuan Liu Luleå University of Technology

5 Conclusions
1 In order to model the complicated fracture behaviour of heterogeneous rock under loading, a method
integrating microscopic observation, image analysis and numerical modelling can be used.
Microstructure information concerning a heterogeneous rock can be obtained from a sequence of
physical or optical cuts by microscopic observation. Image analysis can be used to simplify the
microstructure, identify the mineral composition and obtain the morphological information. Then a
numerical model can be constructed by subsequent meshing to study the influence of morphological
details and their evolution during the fracture process.
2 Considering the research scale and computer capacity, rock heterogeneity can be characterized better
with the statistical approach. On the basis of the Weibull distribution, a heterogeneous material model is
proposed to characterize the heterogeneity of rock, and in this model rock is specified according to a few
characteristic parameters: the homogeneous index and the elemental seed parameters. For any specific
rock in practice, the seed parameters, such as the uniaxial compressive strength, elastic modulus, etc,
can be obtained from laboratory tests and the homogeneous index can be determined on the basis of the
defect distribution of the microstructure.
3 Using these characteristic parameters as input parameters, a numerical code called the rock and tool
interaction code (R-T2D) is developed to investigate progressive rock fracture. A series of numerical
experiments including both intact rock and notched rock are conducted to obtain the physical-
mechanical properties and fracture toughness, and to simulate the crack initiation and propagation and
the whole progressive fracture process. It seems that the R-T2D code is indeed a valuable numerical tool
for research on the progressive rock fracture process. Although the R-T2D code uses the characteristic
parameters as input parameters, reasonable values for the fracture toughness are obtained. Therefore, the
developed R-T2D code seems to have built a bridge between the physical-mechanical parameters and
fracture toughness. Moreover, the detailed visually shown stress distribution and redistribution, crack
nucleation and initiation, stable and unstable crack propagation, interaction and coalescence, and
corresponding load-displacement curves are found to be in good agreement with experimental results in
the literature and can be proposed as benchmarks for numerical programs for crack propagation.
4 Rock heterogeneity has an important influence on the crack initiation location and the subsequent
propagation path. It is found that in crystalline heterogeneous materials such as granite, grain boundaries
act as the predominant source of stress concentrating cracks. At the locations of strong mineral grains
such as quartz, the first detectable fracturing starts at the grain boundaries. However, although grain
boundaries are the preferred sites of microcracks, intragranular cracks also occur at the locations of
weak mineral constituents such as feldspar and mica grains, where harder minerals induce a point load
on neighbouring soft minerals. In the localization of cracks, secondary fracturing crosses the strong
mineral grains to form a macroscopic shear band.
5 Compared with the traditional Mohr-Coulomb and Hoek-Brown strength criterion, the double elliptic
strength criterion is more useful for modelling the fracture occurring in mechanical fragmentation in that
it can represent the transition from brittle failure to ductile cataclastic failure with increasing confining
pressure. The brittle failure face of the double elliptic criterion will represent rock failure in the shear
mode or tensile mode, just as the modified Mohr-Coulomb strength criterion (including tensile cut-off)
or Hoek-Brown strength criterion at a low confining pressure. The ductile failure surface will represent
rock failure in the ductile cataclastic mode at a high confining pressure.
6 In mechanical fragmentation, a crushed zone is always available near the tool. The crushed zone has an
important influence on the chipping process and energy utilization. According to our simulations, the
crushed zone is in fact the zone with a high density of microcracks. In this zone microcracking is

42
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

pervasive and rocks behave in a ductile manner with stress satisfying the ductile failure surface of the
double elliptic strength criterion. Associated with this microcracked or dilatant region there must be a
volumetric expansion, which results in tensile crack propagation to the surface to form rock chips.
7 In the optimum design of cutter tools, the shape of the tool is an important factor. In the research
conducted for this thesis, the influence of the geometrical shape of tools with different back rake angles,
–200, 00 and 200, has been investigated. The effective energy utilisation ratio for cutters with different
back rake angles, –200, 00 and 200, is 14%, 18% and 21% respectively. The energy consumed by the
formation of the crushed zone is 41%, 71% and 63% respectively. The specific energies for chipping are
199, 123 and 49 kJ/m3 respectively for the three cutters with rake angles of 200, 00 and –200. When
comparing the three different rake angles, it is reasonable to conclude that a rock cutter with a rake
angle of –20 degrees is the most efficient of the three cutters.
8 According to the simulated results, a simple description and qualitative model of the rock fragmentation
process induced by truncated indenters can be given as follows. Little damage to the rock is observed at
the linear-elastic deformation stage. Then conical Hertzian cracks are initiated adjacent to both corners
of the truncated indenter and propagated in the well-known conical Hertzian manner. As the loading
displacement increases, some of the elements in the high confining pressure zone immediately under the
indenter fail in the ductile cataclastic mode, with the stress satisfying the ductile failure surface of the
double elliptic strength criterion. With the cataclastic failures and tensile conical cracks releasing the
confining pressure, the elements in the confining pressure zone are compressed into failure and the
crushed zone comes into being. In the crushed zone, microcracking is pervasive. The intensity of the
microcracking within the zone increases and a re-compaction behaviour occurs with increasing loading
displacement. Associated with this microcracked region there is a volumetric expansion and a tensile
stress field, which drives side cracks to propagate in a curvilinear path. It is thought that the curvilinear
path is caused by the heterogeneity of the rock. With an increased loading displacement, the side cracks
rapidly propagate and intersect with the free rock surface to form rock chips.
9 In indentation tests, the confining pressure has an important influence on the failure mode. As the
confining pressure increases, a small but noticeable increase in the indentation strength is measured.
With decreasing confining pressure, decreases in the indentation strength are not of great interest, which
can be observed in many tri-axial tests. Of particular interest is a change in the rock failure process
when the confining stress is reduced below a critical value. Instead of the usual formation of rock chips
adjacent to the punch, vertical cracks are propagated beneath the punch, causing the specimen to be split
in half when the confining pressure on the sample is less than a critical value. This result is of practical
interest. In boring hard rock at low confining stresses, the creation of such tensile fractures beneath an
indenter may serve to fragment the rock sufficiently to facilitate its removal.
10 The simultaneous loading of the rock surface by multiple indenters seems to provide a possibility of
forming larger rock chips, controlling the direction of subsurface cracks and consuming a minimum
total specific energy. In the design and operation of machines for rock drilling, the line spacing between
the indenters is a very important factor because of the interaction between adjacent side splays of
chipping grooves. With a small line spacing, the cracks induced by the double indenters will behave in a
manner similar to those caused by a single indenter. With the appropriate line spacing, the side cracks
caused by the two indenters will propagate, interact and coalesce, giving a possible combined crack at
great depth to form large rock chips.
11 As drilling and mining environments are becoming increasingly severe, improvements through drilling
optimisation are needed to hold down costs. Numerical models that can accurately simulate tool-rock
interaction and failure mechanisms, including chip formation and interaction between adjacent indenters,
could be used in parametric studies to determine the optimal indenter designs, reducing much of the
expensive and time-consuming experimental work which would otherwise have to be carried out to

43
Hongyuan Liu Luleå University of Technology

assess the performance of indenter designs. The results simulated by the R-T2D code reproduce the
progressive process of rock fragmentation under mechanical loading: the build-up of the stress field, the
formation of the crushed zone, surface chipping, and the formation of the crater and subsurface cracks.
The R-T2D code can be utilized to improve our understanding of rock-tool interaction and the rock
failure mechanisms under the action of mechanical tools, which, in turn, will be useful in assisting the
design of fragmentation equipment and fragmentation operations.

44
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

6 Suggestion for further research

6.1 Numerical simulation of specific rocks

In this thesis, we have characterized the heterogeneity of rock according to the following two types of
characteristic parameters: (1) the homogeneous index m and (2) the elemental seed parameters. For any
specific rock in practice, the seed parameters can be obtained from the main physical-mechanical
parameters in laboratory tests and the homogeneous index can be determined on the basis of the defect
distribution of the microstructure (Orlovskaja et al, 1997; Curtis and Juszczyk, 1998; Davies, 2001). Further
researches are needed to determine the homogeneous index.

6.2 Consideration of dynamic effects

In the rock fragmentation industry, blasting is another popular method widely employed in the areas of
excavation and construction. As for the simulation of rock blasting, dynamic effects should be considered.
Some researches (Zhang, 2001) have indicated that dynamic factors have an important influence on rock
fracture. When the loading rate is up to and over 106 MPa m1/2 S-1 or when the inertia effects must be
considered, a quasi-static assumption may not be suitable and methods for measuring dynamic effects are
needed. In this case, a dynamic wave propagation module should be developed in the R-T2D code.

6.3 Coupling of the thermo-hydro-mechanical (THM) process

Rock fragmentation is in fact a coupled thermo-hydro-mechanical (THM) process. The simulation of

rock fragmentation will be made more practical if the THM process is combined into the R-T2D code.

6.4 Three-dimensional (3D) numerical model

With the development of computing power, interactive computer graphics and topological data structure,
the two-dimensional model should be extended to create a three-dimensional model, so that the numerical
simulation may come closer and closer to the situation in practice.

6.5 Quantitative relationship between the applied force and the resultant crack length in
mechanical fragmentation

In mechanical fragmentation, there is a big difference between the theoretical crack length and the actual
crack length. In order to improve the present theoretical relationship and make it more apt for practical
applications, it is imperative to develop a series of numerical simulations which would consider rock
confinement, contact conditions and the effect of the loading rate. Moreover, relatively new theoretical
researches, numerical studies and field measurements are needed to determine the indentation force and the
crack length in rock.

45
Hongyuan Liu Luleå University of Technology

References:
Alehossein H, Hood M. State-of-the-art review of rock models for disc roller cutters. In: Rock mechanics,
eds. Aubertin, Hassani and Mintri. Rotterdam: Balkema, 1996. pp. 693-700
Al-Shayea NA, Khan K, Abduljauwad SN. Effects of confining pressure and temperature on mixed-mode
(I-II) fracture toughness of a limestone rock. Int. J. Rock Mech. Min. Sci. 2000; 37: 629-643
Artsimovich GV, Poladko EN, Sveshnikov IA. Investigation and development of rock-breaking tool for
drilling, Nauka, Novossibirsk, 1978
Batrouni GG, Hansen A. Fourier acceleration of iterative processes in disordered systems. J. Stat. Phys.
1988; 52(3/4): 747-773
Bazant ZP, Tabbara MR, Kazemi, MT, Pijaudier-Cabot G. Random particle model for fracture of aggregate
or fibre composites. J. of Engng. Mech. 1990; 116: 1686-1705
Blair SC, Cook NGW. Analysis of compressive fracture in rock using statistical techniques – Part I. A non-
linear rule-based model. Int. J. Rock Mech. Min. Sci. 1998; 35 (7): 837-848
Chen L, Alehossein H. Kerf effects in spherical indentation. In: Rock mechanics, eds. Aubertin, Hassani
and Mintri. Rotterdam: Balkema, 1996. pp. 717-724
Cook NGW, Hood M, Tsai F. Observations of crack growth in hard rock loaded by an indenter. Int. J. Rock
Mech. & Min. Sci. 1984; 21: 97-107
Cundall PA. A computer model for simulating progressive large scale movements in blocky rock systems.
In: Proc. Symp. Int. Soc. Rock Mechanics, Nancy, France, Vol.1, paper no. II-8, 1971.
Cundall PA. UDEC: a generalized distinct element program for modeling jointed rock. Peter Cundall
Associates, Report PCAR-1-80, U.S. Army, European Research Office, London, Contract DAJA37-79-
C-0548, 1980
Cundall PA, Strack ODL. Modeling of microscopic mechanisms in granular material. In: Mechanics of
granular materials: new models and constitutive relations, eds. Jenkins JT and Satake M, Elsevier,
Amsterdam, 1983. pp. 137-149
Cundall PA, Hart RD. Development of generalized 2-D and 3-D distinct element programs for modeling
joint rock, Itasca Consulting Group; U.S. Army Corps of Engineers, Misc. Paper SL-85-1, 1985
Cundall PA. Formulation of a three-dimensional distinct element model – Part I: A scheme to detect and
represent contacts in a system composed of many polyhedral blocks. Int. J. Rock Mech. Min. Sci. &
Geomech. Abstr. 1988; 25: 107-116
Curtis RV, Juszczyk AS. Analysis of strength data using two- and three-parameter Weibull models. Journal
of Materials Science 1998; 33: 1151-1157
Davies IJ. Empirical correction factor for the best estimate of Weibull modulus obtained using linear least
square analysis. Journal of Materials Science 2001; 20: 997-999
Deketh HJR, Alvarez GM, Hergarden IM, Giezen M, Verheof PNW. Towards the prediction of rock
excavation machine performance. Bull. Eng. Geol. Env. 1998; 57, 3-15.
DIANA user’s manual, Release 5.1, TNO Building and Construction Research, The Netherlands, 1993
Faruque MO. A cap type constitutive model for plain concrete. In: Constitutive laws for engineering
materials: theory and applications, Elsevier Science Publishing, 1987, pp. 395-402.
Fu YF. Numerical research on the fracture process of brittle rock. PhD thesis, Northeastern University, 2000
(in Chinese)
Glucklich J. Fracture of plain concrete. J. Engng Mech. 1963; 89: 127-138
Garboczi EJ, Day AR. An algorithm for computing the effective linear elastic properties of heterogeneous
materials: 3-D results for composites with equal phase Poisson ratios. J Mech Phys Solids 1995; 43:
1349-1362

46
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

Hart R, Cundall PA, Lemos J. Formulation of a three-dimensional distinct element model – Part II:
Mechanical calculations for motion and interaction of a system composed of many polyhedral blocks.
Int. J. Rock Mech. Min. Sci. & Geomech. Abstr. 1988; 25: 117-126
Herrmann HJ, Roux S. Statistical models for the fracture of disordered media, Elsevier Science, 1992
Hoek E, Brown ET. Underground excavations in rock. Balkema, Rotterdam, 1980, pp. 215
Hood M. Phenomena relating to the failure of hard rock adjacent to an indenter. J. South Africa Min. &
Metal. 1997; 12: 113-123
Howarth DF, Bridge EJ. Microfracture beneath blunt disc cutters in rock. Int. J. Rock Mech. & Min. Sci.
1988; 25: 35-38
Hudson JA, Fairhurst C. Tensile strength, Weibull’s theory and a general statistical approach to rock failure.
In: The Proceedings of the Civil Engineering Materials Conference held in Southampton 1969; 901-904
Huang JA, Wang SJ. An experimental investigation concerning the comprehensive fracture toughness of
some brittle rocks. Int. J. Rock Mech. Min. Sci. & Geomech. Abstr. 1985; 22 (2): 99-104
Ingraffea AR. Mixed mode fracture initiation in Indiana limestone and Westerly granite. In: Proc. 22nd US
Symp. Rock Mech. 1981; 199-204
ISRM. Suggested methods for determining the uniaxial compressive strength and deformability of rock
materials. Int. J. Rock Mech. Min. Sci. & Geomech. Abstr. 1979; 16 (2): 135-140
Itasca Consulting Group Inc. FLAC (Fast Langrangian Analysis of Continua), Version 3.3, ICG,
Minneapolis, 1995
Itasca Consulting Group Inc. Particle flow code in 2 dimensions and particle flow code in 3 dimensions,
Version 1.1, Minneapolis: ICG, 1995
Karihaloo BL. Compressive fracture of glassy brittle materials. In: Advances in fracture research (Fracture
84). Pergamon Press, Oxford, 1984
Kou SQ. Some basic problems in rock breakage by blasting and by indentation. PhD thesis, Luleå
University of Technology, 1995: 180 D
Kou SQ, Lindqvist PA, Tan XC. Formulation of models for indentation depth and crack length in
mechanical excavation of rock, SKB report, HRL-96-20, 1996.
Kou SQ, Lindqvist PA, Tang CA and Xu XH. Numerical simulation of the cutting of inhomogeneous rocks.
Int. J. Rock Mech. Min. Sci. 1999; 36 (5): 711-717
Kou SQ, Liu HY, Tang CA. Numerical simulation of the failure process when cutting inhomogeneous rocks:
influence of rake angles. In: Proceedings of the 4th Regional Symposium on Computer Applications in
the Minerals Industries, eds. Eloranta P and Särkkä P, APCOM 2001; Tampere, Finland, pp. 39-48
Kou SQ, Liu HY, Lindqvist PA, Tang CA, Xu XH. Numerical investigation of particle breakage as applied
to mechanical crushing – Part II: Interparticle breakage. Int. J. Rock Mech. Min. Sci. 2001; 38: 1163-
1172
Kutuzov BN, Kryukov GM, Tarasenko VP. Process of impact interaction between tool and rock. Moscow:
MGI, 1969
Lawn BR, Swain MV. Microfracture beneath point indentation in brittle solids. J. Mat. Sci. 1975; 10: 113-
122
Lawn BR, Wilshaw TR. Indentation fracture: principles and application. J. Mat. Sci. 1975; 10: 1049-1081
Lindqvist PA. Rock fragmentation by indentation and disc cutting. PhD thesis, Luleå University of
Technology, 1982: 20 D
Lindqvist PA, Lai HH. Behavior of the crushed zone in rock indentation. Rock Mech. Rock Engng., 1983;
16: 199-207
Lindqvist PA, Suarez LM, Montoto M, Tan XC, Kou SQ. Rock indentation database: testing procedures,
results and main conclusions. SKB report: Pr 44-94-023, Sweden, 1994

47
Hongyuan Liu Luleå University of Technology

Liu HY. Virtual research on the overburden rock fracture process induced by mining excavation. MSc thesis,
Northeastern University, 2000 (in Chinese)
Liu HY, Tang CA, Rui YQ. Numerical simulation of the overburden strata failure process as multi-coal is
mined. Chinese Journal of Rock Mechanics and Engineering, 2001; 20 (2): 190-196 (in Chinese)
Liu HY, Tang CA, Yang TH, Li LC. Numerical simulation research on acoustic emission characteristics
under symmetrical and asymmetrical compressive loads. Rock and Soil Mechanics 2001; 22 (4): 383-
387 (in Chinese)
Liu HY, Liu JX, Tang CA. Numerical simulation of the failure process of overburden rock strata caused by
mining excavation. Chinese Journal of Geotechnical Engineering 2001; 23 (2): 201-204 (in Chinese)
Liu HY, Kou SQ, Lindqvist PA. Numerical modelling of the rock fragmentation process induced by
multiple indenters. In: Mining and tunnelling innovation and opportunity, eds. Hammah R, Bawden W,
Curran J and Telesnicki M, NARMS-TAC 2002, Toronto, Canada. pp. 757-764
Liu HY, Kou SQ, Lindqvist PA, Tang CA. Numerical simulation of the rock fragmentation process induced
by indenters. Int. J. Rock Mech. Min. Sci. 2002; 39: 491-505
Liu HY, Kou SQ, Lindqvist PA. Numerical simulation of the fracture process in cutting heterogeneous
brittle material. International Journal for Numerical and Analytical Methods in Geomechanics 2002; 26:
1253-1278
Lorig LJ, Cundall PA. Modeling of reinforced concrete using the distinct element method. In: Fracture of
concrete and rock, eds. Shah SP, Swarty SE, SEM, Bethel, CT, 1987. pp. 459-471
McKinnon SD, Barra IG. Fracture initiation, growth and effect on the stress field: a numerical investigation.
Journal of Structural Geology 1998; 20 (12): 1673-1689
Mihashi H. States of the art and a view of the fracture mechanics of concrete. Journal of JCI 1987; 25 (2):
14-25
Mishnaevsky LL Jr. Physical mechanisms of hard rock fragmentation under mechanical loading: a review.
Int. J. Rock Mech. Min. Sci. 1995; 32, 763-766
Mishnaevsky LL Jr. Damage and fracture of heterogeneous materials: modelling and application to drilling
tool improvement. A.A. Balkema, Rotterdam, 1998
Mishnaevsky LL Jr. Rock fragmentation and optimization of drilling tools. In: Fracture of rock,
Southampton, WIT Press, 1999, pp. 173
Murrell SAF, Digby PJ. The theory of brittle fracture initiation under triaxial stress conditions – I and II.
Geophys. J. R. Astronomical Soc. 1970; 19: 309-334 and 499-512
NUMA: a general purpose finite element code (Alehossein H). CMTE-CSIRO/EM, Australia, 1995
Orlovskaja N, Peterlik H, Marczewski M, Kromp K. The validity of Weibull estimators: experimental
verification. Journal of Materials Science 1997; 32: 1903-1907
Otsuka K, Date H. Fracture process zone in concrete tension specimen. Engineering Fracture Mechanics
2000; 65: 111-131
Paul B, Mirandy L. An improved fracture criterion for three-dimensional stress states. J. Engng Mater.
Technol. 1976; 98: 159-163
Place D, Mora P. The lattice solid model to simulate the physics of rocks and earthquakes: incorporation of
friction. Journal of Computational Physics 1999; 150: 332-372
Plesha ME, Aifantis EC. On the modeling of rocks with microstructure. In: Rock mechanics: theory-
experiment-practice, eds. Mathewson CC, Association of Engineering Geologists, New York, 1983. pp.
27-35
Protassov YI. Theoretical principle of mechanical fragmentation of rocks, Nedra, Moscow, 1985
Psakhie SG, Moiseyenko DD, Smolin AY. The features of fracture of heterogeneous materials and frame
structures: potentialities of MCA design. Computational Materials Science 1999; 16: 333-343

48
 Numerical Modelling of the Rock Fracture Process under Mechanical Loading

Rao QH. Pure shear fracture of brittle rock: a theoretical and laboratory study. PhD thesis, Luleå University
of Technology, 1999: 08 D
Saouma VE, Kleinosky M. Finite element simulation of rock cutting: a fracture mechanics approach. In:
Proc. 25th U.S. Symp. on Rock Mech. 1984. pp. 792-799
Schlangen E, Van Mier JGM. Experimental and numerical analysis of micromechanisms of fracture of
cement-based composites. Cement and Concrete Composites 1992; 14 (2): 105-118
Schlangen E, Garboczi EJ. Fracture simulations of concrete using lattice models: computational aspects.
Engineering Fracture Mechanics 1997; 57 (2/3): 319-332
Shen B, Stephansson O. Modification of the G-criterion for crack propagation subjected to compression.
Engng. Fracture Mech. 1994; 47: 177-189
Shi GH, Goodman RE. Discontinuous deformation analysis: a new method for computing stress, strain and
sliding of block systems. In: Key questions in rock mechanics, eds. Cundall PA et al. Balkema,
Rotterdam, 1988. pp. 381-383
Simonov VV, Vyskrebtsov VG. Work of roller-bit tools and their improvement. Moscow: Nedra, 1975.
Swartz SE, Taha NM. Mixed mode crack propagation and fracture in concrete. Engng. Fract. Mech. 1990;
35: 137-144
Swenson DV, Ingraffea AR. Modelling mixed-mode dynamic crack propagation using finite elements:
theory and applications. Computational Mechanics 1988; 3: 187-192
Tan XC. Modelling of drill string buckling and tool indentation in rock drilling and fragmentation. PhD
thesis, Luleå University of Technology, 1996: 197 D
Tan XC, Kou SQ, Lindqvist PA. Simulation of crack fragmentation by indenters using DDM and fracture
mechanics. In: Proc. of 2nd North American Rock Mech. Conf. Montreal, Canada, June 19-21, 1996. pp.
685-692
Tan XC, Lindqvist PA, Kou SQ. Application of a splitting fracture model to the simulation of rock
indentation subsurface fractures. J. Numer. & Anal. Meth. Geomech. 1997; 21: 1-13.
Tang CA. Numerical simulation of progressive rock failure and associated seismicity. Int. J. Rock Mech.
Min Sci. 1997; 34: 249-262
Tang CA, Fu YF, Kou SQ, Lindqvist PA. Numerical simulation of loading inhomogeneous rocks. Int. J.
Rock Mech. Min. Sci. & Geomech. Abstr. 1998; 35, 1001-1007
Tang CA, Liu H, Lee PKK, Tsui Y, Tham LG. Numerical studies of the influence of microstructure on rock
failure in uniaxial compression – Part I: Effects of heterogeneity. Int. J. Rock Mech. Min. Sci. 2000; 37:
555-569
Tang CA, Tham LG, Lee PKK, Tsui Y, Liu H. Numerical studies of the influence of microstructure on rock
failure in uniaxial compression – Part II: Constraint, slenderness and size effect. Int. J. Rock Mech. Min.
Sci. 2000; 37: 571-583
Tang CA, Liu HY, Qin SQ, Yang ZF. Influence of heterogeneity on crack propagation mode in brittle rock.
Acta Geophysica Sinica 2000; 43 (1): 116-121
Tang CA, Lin P, Liu HY, Liang ZZ. On failure modes and strength characterization of brittle disordered
materials under uniaxial compression and tension. International Journal of Key Engineering Material,
FEOFS 2000, Pohang, Korea. pp. 637-642
Tang CA, Xu XH, Kou SQ, Lindqvist PA, Liu HY. Numerical investigation of particle breakage as applied
to mechanical crushing – Part I: Single particle breakage. Int. J. Rock Mech. Min. Sci. 2001; 38: 1147-
1162
Tapponnier P, Brace WF. Development of stress-induced microcracks in Westerly granite. Int J Rock Mech
Min Sci & Geomech Abstr 1976; 13: 103-112
Tirosh J, Catz E. Mixed-mode fracture angle and fracture locus of materials subjected to compressive
loading. Engng. Fracture Mech. 1981; 14, 27-38

49
Hongyuan Liu Luleå University of Technology

Verhoef PNW, Ockeloen JJ. The significance of rock ductility for mechanical rock cutting. In: Rock
mechanics, Balkema, Rotterdam, 1996. pp. 709-716
Wang EZ, Shrive NG. Brittle fracture in compression: mechanisms, models and criteria. Engineering
Fracture Mechanics 1995; 52 (6): 1107-1126
Walton OR. Particle dynamic modeling of geological materials. Lawrence Livemore National Laboratory,
Report UCRL-52915, 1980
Walton OR, Braun RLM, Mallon RG, Cervelli DM. Particle dynamic calculations of gravity flows of
inelastic fractional spheres. In: Micromechanics of granular material, Amsterdam: Elsevier Science
Publishers, 1988. pp. 153-161
Wawrzynek P, Ingraffea AR. An interactive approach to local remeshing around a propagating crack. Finite
Elements in Analysis and Design 1989; 5: 87-96
Weibull W. A statistical distribution function of wide applicability. J Appl Mech 1951; 293-297
Whittaker BN, Singh RN, Sun G. Rock fracture mechanics: principles, design and applications. Elsevier,
Amsterdam, 1992
Wilson SA, Henderson JR, Main IG. A cellular automaton fracture model: the influence of heterogeneity in
the failure process. Journal of Structural Geology 1996: 18 (2/3): 343-348
Wittmann FH. Fracture process zone and fracture energy: fracture mechanics of concrete structures. In:
Proceedings of FraMCoS-1. Amsterdam: Elsevier, 1992. pp. 391-403
Yang TH. Research on the seepage property and its coupling with stress in the rock fracture process. PhD
thesis, Northeastern University, 2001 (in Chinese)
Yoffe EH. Elastic stress fields caused by indenting brittle materials. Philosophical Magazine A. 1982; 46:
617-628
Yu MH. Double twin strength criterion and its application. Science Press. Beijing, 1998 (in Chinese).
Zhang ZX. Laboratory studies of dynamic rock fracture and in-situ measurements of cutter forces for a
boring machine. PhD thesis, Luleå University of Technology, 2001: 23 D
Zhang ZX. An empirical relation between mode I fracture toughness and the tensile strength of rock. Int. J.
Rock Mech. & Min. Sci. 2002; 39: 401-406
Zhao J. Applicability of Mohr-Coulomb and Hoek-Brown strength criteria to the dynamic strength of brittle
rock. Int. J. Rock Mech. & Min. Sci. 2000; 37: 1115-1121
Zhu WC. Mesoscopic numerical model for the fracture process of concrete and its application. PhD thesis,
Northeastern University, 2000 (in Chinese)

50