Geomech. Geophys. Geo-energ. Geo-resour.

(2017) 3:121–130
DOI 10.1007/s40948-016-0048-9

ORIGINAL ARTICLE

Modelling of primary fragmentation in block caving mines

using a finite-element based fracture mechanics approach
A. Paluszny . R. W. Zimmerman

Received: 17 November 2016 / Accepted: 28 December 2016 / Published online: 24 April 2017
Ó The Author(s) 2017. This article is an open access publication

Abstract The growth of fractures around an under- in the direct overburden stress case, fractures tend to
cut of a block cave is simulated. A finite element based grow close to the cave. Over-arching fractures
approach is used, in which fractures are represented as concentric to the undercut continue to grow as the
non-planar 3D surfaces that grow in response to cave progresses.
boundary stresses and interaction. A new mesh is
recreated at each step to compute the displacement Keywords Fracture propagation  Primary
field. Stress intensity factors are computed around fragmentation  Caving  Finite element  Stress
fracture tips using a technique that computes the intensity factors
interaction integral over a virtual disk. Fracture
geometry is updated using Paris and Schöllmann
propagation laws, and a geometric fracture pattern
ensues from the simulation. The growth of fractures is 1 Introduction
examined in the lower 20 m of a mine at 812 m depth.
The growth of 30, 60 and 90 fractures is examined. Underground mining methods extract large volumes
Realistic extraction schedules for over 100 draw of rock directly from the subsurface by undercutting
points control the rate of mass extraction. The effect rocks that contain minerals of interest, and relying on
of rock bridges as overburden stress shields is blasting and gravity-driven fragmentation to extract
investigated. Bridges are modelled by constraining ore at low cost (Brown 2003). Block caving, in situ
the vertical displacement of the top boundary. This leaching, and in situ fragmentation, all aim to
case is compared to a Neumann-type overburden stress decimate the rock whilst still underground so as to
boundary condition in which the overburden is felt reduce the environmental footprint of the mining
throughout the top of the cave. In both cases, fractures activity, whilst maintaining low production costs.
grow to form a dome shape above and around the cave Fracture growth during tunnelling, caving and blasting
during extraction. For the case of a fixed top boundary, is a key controlling parameter in designing extraction
fracture growth is observed away from the cave, while procedures, in order to attain specific fragment size
distributions and avoid excessive fines production
(Rance et al. 2007). Numerical modelling can be
useful in determining how fractures grow under
A. Paluszny (&)  R. W. Zimmerman
specific mine scenarios, and may conduce to large
Department of Earth Science and Engineering, Imperial
College, London, UK economic benefits to the industry if energy and mass
e-mail: [email protected] conservative methods are systematically applied as

123
122 Geomech. Geophys. Geo-energ. Geo-resour. (2017) 3:121–130

part of subsurface mining procedures. In addition, Drucker-Prager based plastic or de-bonding pseudo-
high-accuracy numerical methods can be instrumental fracture models. The standard DEM method utilises
in defining safety limits and identifying mine-specific arbitrary dampening penalty values that affect the
risk scenarios. ability of the method to conserve energy. Furthermore,
Numerous numerical methods have been developed this method requires the definition of micro-properties
to model fracture and fragmentation. These are using laboratory calibration tests that are rock type and
discussed here in the context of subsurface mining, scale dependent (Lisjak and Grasselli 2014), and are
and their suitability to model primary fragmentation. specific to a fixed mesh size—and therefore cannot be
In particular, a main focus of the discussion is finite readily translated to the field scale without additional
element derived techniques that model multiple frac- empirical assumptions. De-bonding fracturing is arbi-
tures explicitly and account for variable matrix trary and is not based on rigorous fracture growth
material heterogeneities. principles. In contrast, a DEM impulse approach uses
The finite element method (FEM) is suitable for the directly measurable material properties, such as
analysis of fracture interaction during growth, and for Young’s modulus and Poisson’s ratio, and ensures
combination with flow and poro-elastic analysis for energy conservation during fragmentation by relying
hydraulic fracture growth. FEM requires the genera- on impulses instead of penalties to resolve contact
tion of meshes to discretise the solution domain. The (Tang et al. 2014).
extended finite element method (XFEM) is a special- The present paper investigates primary fragmenta-
isation of FEM, in which elements have enriched tion around a growing cave, by modelling fracture
shape functions which model a partition of unity propagation during caving. As opposed to XFEM or
function, allowing for discontinuities to exist within DEM approaches, the current paper directly applies
each element sub-domain (Moës et al. 1999). The FEM to simulate the growth of multiple fractures. The
main advantage of XFEM is that it can represent finite element method is combined with a geometry-
discontinuities within a pre-existing mesh, and can driven approach in which fracture geometry evolves
model discrete fracture propagation without re-mesh- independently from its discretisation. This method
ing. The enrichment of the shape functions can be used provides three key ingredients to running multiple
to model fracture growth and friction, as well as fluid fracture growth and interaction simulations: (1) it
flow through fractured media. However, by avoiding allows fractures to be represented discretely, as sub-
re-meshing, XFEM cannot ensure that the mesh is dimensional entities; (2) it allows for the definition of
sufficiently refined ahead of the crack tip as growth geologically realistic, site-specific matrix hetero-
progresses, leading to large errors in stress and energy geneities; (3) it uses input properties that are measur-
estimations around the tips. It is ideally suited for the able and well-understood rock properties, such as
problem of moving interfaces, and for fracture anal- Young’s modulus and Poisson’s ratio; (4) by applying
ysis cases for which fracture paths can be predicted. adaptive re-meshing, the mesh is always suited for the
Furthermore, enrichment functions become unsta- current geometry; (5) it can be directly validated
ble for fracture interaction, and when fractures share against analytical solutions for stress intensity factors
nodes. as well as against laboratory experiments for single
In contrast, the discrete element method (DEM) and multiple fractures under tension, compression and
models the movement of volumetric objects and their shear loading. Furthermore, this approach represents
interaction based on the solution of dynamical equa- fracture and matrix domains explicitly, and allows
tions, using a penalty-based damping to resolve modelling coupled fluid flow, as well as thermal and
collisions. Agglomerates of smaller-scale spherical poro-elastic effects.
or polygonal/polyhedral elements represent bodies.
Extensions, such as Discontinuous Deformation Anal-
ysis (DDA), solve for displacements instead of forces 2 Methodology
(Shi 1988). DEM is simple to implement, and efficient
to run in both 2D and 3D. It can be combined with The method presented here utilises a mesh to compute
FEM to model non-linear deformation of the bodies the displacement field, and replaces that mesh as the
(FEMDEM; Munjiza et al. 1995), usually using geometry progresses (Paluszny and Zimmerman

123
Geomech. Geophys. Geo-energ. Geo-resour. (2017) 3:121–130 123

2011, 2013). Thus, the mesh can be substituted by a representation of the fracture (Paluszny and Zimmer-
mesh-free point cloud based computation of the stress man 2013).
field, or by a boundary element based computation of
the stress intensity factors. The method aims to 2.3 Discretisation
remain geometrically focussed, with the sole aim of
growing multiple interacting fractures in an accurate The volumetric domain is discretised using isopara-
manner. metric quadratic tetrahedra. Fracture surfaces are
discretised using isoparametric quadratic triangles.
2.1 Constitutive model Side- and corner quarter-point tetrahedral elements are
placed along the crack front. These are a type of
The rock is assumed to behave in a linear elastic quadratic elements in which the centre node is shifted
manner, and so the stress–strain relation is given by towards the tip; they are placed in order to better
r ¼ D[e  ^e, where e and r are the infinitesimal strain capture the stress singularity at the fracture tip (Nejati
and stress vectors, D is the elasticity matrix, and e^ is et al. 2015a). Displacements are computed at the
the vector of initial strains. The finite element is nodes, material properties are defined at Gaussian
applied to solve the deformation field, taking into integration points, and stresses and strains are com-
account surface tractions and body forces. puted at the integration point locations. Tips are
discretised into nodes and segments, and stress
2.2 Geometry intensity factors are computed at each node along
the tips.
The model geometry, including boundaries and frac-
tures, is defined using three-dimensional cubic Non- 2.4 Stress intensity factors and growth criteria
Uniform Rational B-Spline (NURBS) surfaces
(Fig. 1). The model boundaries form a watertight As the rock deforms, energy release rates are com-
domain. Multiple fractures grow within the model puted at the crack tips, after which energy-based
domain, in a planar or curved manner, depending on methods are used to predict the onset and direction of
the effects of the boundary conditions and the fracture growth. These serve as local approximations
interaction with other fractures. At each step, fracture of energy in each modality of growth (I, II, and III) that
geometry is updated by extending the NURBS can be used to determine propagation.

Fig. 1 Fracture
representation schematic. A
surface is extended during
growth by applying
deformation to the tips of the
fracture and extending the
underlying NURBS
representation of the
fracture using propagation
vectors at each tip

123
124 Geomech. Geophys. Geo-energ. Geo-resour. (2017) 3:121–130

Fig. 2 Stress intensity

factor J- and interaction
integral computation
techniques. Initially based
on structured meshes, new
methods allow for
unstructured meshes, whilst
reducing integration domain
to a virtual disk

Specifically, the interaction integral (Yau et al. technique is used to compute the interaction integral
1980; Nakamura and Parks 1988) is computed along [see details and equations in Nejati et al. (2015b)],
discrete locations of the fracture tip in order to which allows the independent computation of KI, KII,
approximate model stress intensity factors, KI, KII, and KIII along the fracture tips.
and KIII (Fig. 2). Classic stress intensity factor com- The Paris-law is used to govern fracture extension
putations require structured meshes to be defined (Lazarus 2003; Paluszny and Zimmerman 2011) at
around the tips. These are generally created by hand, every tip, and a 3D growth angle models the direction
and can be generated automatically into integrated in which each tip extends (Schöllmann et al. 2002).
hybrid meshes (Bremberg and Faleskog 2015). How- The calculation of local Ks allows the modelling of
ever, as the density of fractures increases, the growth as a function of both interaction and boundary
complexity of the generation of such a structured conditions simultaneously (Fig. 3).
mesh also increases. A virtual integration technique The direction of crack growth is not constrained to
avoids the need of a structured mesh by using an follow the pre-existing mesh; instead, a new mesh is
artificial ‘virtual’ mesh to integrate around the tips, created at each step to conform to the new fracture
using a ‘virtual’ cylinder (Cervenka and Saouma geometry. The quality and refinement of the mesh is
1997; Paluszny and Zimmerman 2011) or a ‘virtual’ adjusted to the new geometry. As a result, discrete
disk (Nejati et al. 2015b). When aided by isoparamet- fractures are allowed to grow, and coalesce, leading to
ric quadratic elements (Daimon and Okada 2014) and the formation of primary fragmentation.
with the quarter-point optimisation (Nejati et al.
2015a), this technique allows swift and accurate 2.5 Implementation
computation of stress intensity factors on unstructured
tetrahedral meshes. The integration yields high quality The presented methods have been implemented into
solutions when the integration domain radius approx- the Imperial College Geomechanics Toolkit (ICGT;
imates the mesh size around the fracture (Nejati et al. Paluszny and Zimmerman 2011), which interacts with
2015b). Thus, the use of quarter-point quadratic CSMP?? (Complex Systems Modelling Platform,
tetrahedra around the fracture front in combination C??), an object-oriented finite-element based library
with the virtual technique allows the use of coarse, that is specialised in simulating complex multi-
unstructured tetrahedral meshes for growth computa- physics processes (Matthäi et al. 2001). It has been
tions. In this work, the disk virtual integration extensively validated, against both experimental and

123
Geomech. Geophys. Geo-energ. Geo-resour. (2017) 3:121–130 125

Fig. 3 Double notch

validation tests show that
fracture interaction is
influenced by boundary
conditions as well as
spacing. In both cases
compression is followed by
an increase in tension. In
a tension increase occurs
over five steps while for
b tension increase occurs
over twenty steps

Fig. 4 Displacement field

when applying overburden
boundary conditions, case B

field experiments, and can run on workstations as well undercut following a realistic draw point extraction
as on high-performance computing systems. schedule. The mine has 120 draw points that are
uniformly distributed over the undercut area. The rock
is assumed to form a muck pile, which is modelled as a
3 Results and discussion continuum of lesser strength (one tenth of the original
strength) directly below the cave. Details of the
The growth of fractures embedded in a extraction schedule and the draw point have been
1000 9 200 9 200 m cave is modelled. Fractures presented by Paluszny et al. (2015). Fracture propa-
are initially disk-shaped, and propagate in response to gation is modelled in response to gravity-driven
the growth of the mined cave. The models contain deformation of the cave (Fig. 4).
between thirty and ninety fractures. The rock is Material properties correspond to an underground
assumed to contain many more fractures, and only mine. The rock is assumed to be primarily composed
the largest of those fractures are modelled here. It is of Biotite Quartz Monzonite, with a Young’s modulus
assumed that the radii of the fractures of this scale of 61 GPa, an average volume-weighted UCS of
follow a Gaussian distribution, with a mean of 10 m, 114 MPa, and a Poisson’s ratio of 0.26. Tensile
and standard deviation of 0.1 m. Initially, fractures do strength is taken as UCS/10, i.e., rt = 11.4 MPa, and
not intersect each other or the boundaries. Growth is KIC is taken as 1.4 MPa m0.5.
expected to be a function of the boundary conditions, Two boundary conditions are modelled. In case A,
and of the effect of the interaction of multiple the top surface is fixed in the z direction. This
fractures. A 200 9 200 9 20 m rectangular area corresponds to a shielded overburden stress caused
undercuts the mine at the centre of the model, and by the effect of local heterogeneities in the rock
the rock fragments are assumed to migrate through the matrix, such as a local bridge. In case B, an overburden

123
126 Geomech. Geophys. Geo-energ. Geo-resour. (2017) 3:121–130

Fig. 5 Case A. Sixty orthogonal fractures propagating as a function of cave growth, for a fixed top boundary. The undercut is located at
the bottom centre of the model. a Front view, b front view of shaded surfaces and c side view

Fig. 6 Case B. Ninety originally orthogonal fractures grow due to undercutting, for a loading top boundary. The undercut is located at
the bottom centre of the model. a Front view, b side view of rendered fracture surfaces and c rotated side view

123
Geomech. Geophys. Geo-energ. Geo-resour. (2017) 3:121–130 127

Fig. 7 Case B. Growth of 30 fractures at (1) 10% and (2) 30% margin away from the boundary

stress of 22 MPa (815 m depth) is applied at the top all datasets have between thirty and ninety initial disk-
(Fig. 4). In both cases, all model sides are fixed in all shaped fractures, these have all slightly different radii
directions except the vertical, and the bottom bound- and locations for each run. (2) Initial fractures are
ary (with the exception of the undercut) is fixed. placed directly above the undercut, occupying 50% of
The two cases are not identical, as they have several the area across length of the model (x-axis, from left to
key differences in addition to the boundary conditions. right for the front views). (3) In addition, in the y and
(1) cases are made up of different distributions on z directions, initial fractures are placed respecting a
initial fracture locations and orientations. Even though margin of 10 and 30% to the boundaries. (4) Fractures

123
128 Geomech. Geophys. Geo-energ. Geo-resour. (2017) 3:121–130

Fig. 8 Case B. Growth of 60 fractures around the cave over 18 months of extraction

are grown with a beta propagation exponent of 2, and patterns for case A have been discussed by Paluszny
maximum extension length ranges between 10 and et al. (2015).
20 m per month (see Paluszny and Zimmerman
2011, 2013). 3.2 Case B: direct overburden

3.1 Case A: shielded overburden Fractures around an undercut grow in response to

caving without any shielding of the overburden
Figure 5 shows sixty orthogonal fractures that prop- stresses. The modality also conduces to the systematic
agate due to the growth of the cave. In general, a growth of fractures around the cave. Specifically,
shielded overburden will lead to the growth of primary fractures grow around the undercut and pronounce the
fractures close to the cave, in an arching manner. formation of the cave by growing in a dome shaped
However, it also leads to massive growth away from manner. Fractures interact and form a curved concen-
the cave, directly influencing the shape of the ensuing tric discontinuity that further induces caving and
cave, due to the intersection and coalescence of the accelerates the fragmentation of the rock immediately
interacting fractures. More instances of fracture above the cave.

123
Geomech. Geophys. Geo-energ. Geo-resour. (2017) 3:121–130 129

Figure 6 shows a rendered view of ninety fractures Acknowledgements The authors thank Rio Tinto for their
growing during extraction. Fractures are coloured sponsorship of early versions of this work.
differently to enhance their domains. In Fig. 7, four Open Access This article is distributed under the terms of the
steps of growth of two different thirty-fracture sets are Creative Commons Attribution 4.0 International License (http://
shown. On the left, fractures leave a 10% margin to the creativecommons.org/licenses/by/4.0/), which permits unre-
boundary, whereas on the right they leave a 30% stricted use, distribution, and reproduction in any medium,
provided you give appropriate credit to the original
margin. In both cases, the resulting pattern is compa- author(s) and the source, provide a link to the Creative Com-
rable. Fractures closest to the extraction pattern grow mons license, and indicate if changes were made.
first. The same behaviour is observed in Fig. 8, where
sixty fractures grow around the undercut. The dome
shape is even more pronounced, and interaction plays
a role in creating a swarm of fractures that line the top References
region of the cave. Fractures undergo a complete
reorientation caused by local stresses, and little Bremberg D, Faleskog J (2015) A numerical procedure for
interaction integrals developed for curved cracks of general
influence of the initial orientation of the fractures is shape in 3-D. Int J Solids Struct 62:144–157
noticed. Fractures are concentric to the undercut, and Brown ET (2003) Block caving geomechanics. Julius
grow along the regions of the rock that suffers the most Kruttschnitt Mineral Research Centre, Isles Road,
differential displacement. Fractures influence the Indooroopilly, Queensland 4068, Australia
Cervenka J, Saouma VE (1997) Numerical evaluation of 3-D
shape of the cave in the case of direct stress boundary SIF for arbitrary finite element meshes. Eng Fract Mech
conditions, as observed before for shielded boundary 57(5):541–563
conditions. Daimon R, Okada H (2014) Mixed-mode stress intensity factor
All simulations run on a Dell Precision Workstation evaluation by interaction integral method for quadratic
tetrahedral finite element with correction terms. Eng Fract
with two Intel Xeon CPUs X5460 @ 3.16 GHz each. Mech 115:22–42
Their duration varies between 2 and 6 h, depending on Lazarus V (2003) Brittle fracture and fatigue propagation paths
the amount of fractures and amount of extracted years of 3D plane cracks under uniform remote tensile loading.
(up to 2 years of extraction). Int J Fract 122:23–46
Lisjak A, Grasselli G (2014) A review of discrete modeling
techniques for fracturing processes in discontinuous rock
masses. J Rock Mech Geotech Eng 6(4):301–314
Matthäi SK, Geiger S, Roberts SG (2001) The complex systems
platform CSP3D3.0: Users Guide. Technical report, ETH
4 Conclusions Zurich Research Reports
Moës N, Dolbow J, Belytschko T (1999) A finite element
A finite element based method to model primary method for crack growth without remeshing. Int J Numer
Methods Eng 46(1):131–150
fragmentation driven by fracture growth has been Munjiza A, Owen DRJ, Bicanic N (1995) A combined finite-
applied to examine the influence of boundary conditions discrete element method in transient dynamics of fractur-
on fracture growth patterns during caving. The method ing solids. Eng Comput 12:145–174
represents fractures explicitly, and utilises stress inten- Nakamura T, Parks DM (1988) Three-dimensional stress field
near the crack front of a thin elastic plate. J Appl Mech
sity factors and energy-based criteria to model growth. 55:805–813
The model grows limited numbers of fractures that can Nejati M, Paluszny A, Zimmerman RW (2015a) On the use of
capture the overall behaviour of fracture growth during quarter-point tetrahedral finite elements in linear elastic
caving. Fractures are observed to influence the shape of fracture mechanics. Eng Fract Mech 144:194–221
Nejati M, Paluszny A, Zimmerman RW (2015b) A disk-shaped
the resulting cave. Results show that when overburden domain integral method for the computation of stress
stresses are shielded, such as in the case of the formation intensity factors using tetrahedral meshes. Int J Solids
of a bridge, fractures tend to grow away from the cave. Struct 69–70:230–251
When stresses are not shielded, fractures tend to grow Paluszny A, Zimmerman RW (2011) Numerical simulation of
multiple 3D fracture propagation using arbitrary meshes.
closer to the cave. In both cases, fractures grow to form a Comput Methods Appl Mech Eng 200:953–966
dome shape around and above the undercut. These Paluszny A, Zimmerman RW (2013) Numerical fracture growth
results can be useful when adjusting extraction sched- modeling using smooth surface geometric deformation.
ules in response to cave growth. Eng Fract Mech 108:19–36

123
130 Geomech. Geophys. Geo-energ. Geo-resour. (2017) 3:121–130

Paluszny A, Nejati M, Zimmerman RW (2015) A numerical Shi G (1988) Discontinuous deformation analysis—a new
model for fracture propagation leading to primary frag- numerical model for the statics, dynamics of block sys-
mentation in block caving mines. In: 13th International tems. PhD thesis, University of California, Berkeley, USA
Congress of Rock Mechanics ISRM13, Montreal, 10-13 Tang XH, Paluszny A, Zimmerman RW (2014) An impulse-
May 2015. based energy tracking method for collision resolution.
Rance JM, van As A, Owen DRJ, Feng YT, Pine RJ (2007) Comput Methods Appl Mech Eng 278:160–185
Computational modeling of multiple fragmentation in rock Yau JF, Wang SS, Corten HT (1980) A mixed-mode crack
masses with application to block caving. U.S. Rock analysis of isotropic solids using conservation laws of
Mechanics/Geomechanics Symposium (ARMA), 27–31 elasticity. J Appl Mech 47:335–341
May, Vancouver, Canada
Schöllmann M, Richard HA, Kullmer G, Fulland M (2002) A
new criterion for the prediction of crack development in
multiaxially loaded structures. Int J Fract 117(2):129–141

123