Rock Fragmentation by Blasting Sanchidrin (ed)

2010 Taylor & Francis Group, London, ISBN 978-0-415-48296-7

Towards a complete validation of the lattice scheme

in the Hybrid Stress Blasting Model (HSBM)
I. Onederra & G.P. Chitombo
The University of Queensland, Sustainable Minerals Institute, W. H. Bryan Mining and Geology Research
Centre, Brisbane, Qld, Australia

P.A. Cundall & J.K. Furtney

Itasca Consulting Group, Minneapolis, MN, USA

ABSTRACT: The Hybrid Stress Blasting Model (HSBM) is a numerical blast modelling framework
which links an ideal and non-ideal detonation code to a geomechanical rock model. The particle flow code
originally implemented in the geomechanical rock model (PFC3D) has now been replaced by a lattice
scheme. Results from controlled blasting experiments have shown that this scheme is able to realistically
produce breakage patterns from cylinder and cube tests. Predicted versus measured fragment size distributions have shown slight differences attributed in part to the influence of the assumed cubic shape of
fragments to calculate size. Tests have also identified the finest fragment size in the predicted distribution
to be approximately twice the lattice model resolution. With regards to fragment velocities, results underestimate measured values but follow the trend of increasing fragment velocity as a function of explosive
energy. A full scale consistency test has shown the ability of the lattice scheme to realistically define the
extent of breakage and blast damage envelopes.
1

INTRODUCTION

The Hybrid Stress Blasting Model (HSBM) can

be described as a sophisticated blast modelling
research tool, which provides results that can still
be used implicitly for practical blast design. In its
current form it is not a tool that can be used for
day to day blast design and analysis.
The code is being developed through an international collaborative research project funded by a
consortium of companies acknowledged and listed
in this paper and comprised of explosive and equipment suppliers and major mining houses. Over the
course of its development several improvements
and modifications have been made to both the detonation and geomechanical modelling components
in order to improve the calculation speed and the
size of the problems that could be modelled. The
particle flow code (PFC3D) originally implemented
as the platform of the geomechanical model (Ruest
et al. 2006) has now been replaced by a lattice
scheme (Furtney et al. 2009). A crucial task during
the development of the HSBM code is model validation using data from the literature as well as consistency tests designed to evaluate the behavior and
sensitivity of the model. This paper describes the
results of preliminary validation tests of this newly
implemented geomechanical modelling scheme.

THE HSBM FRAMEWORK

The description of the original HSBM framework

has been given by Ruest et al. (2006) and has now
been updated by Furtney et al. (2009). The main
components and software links have not changed
since the inception of the project and these include:
Ideal and non-ideal detonation models to determine explosive detonation characteristics and
outputs from fully coupled and decoupled conditions. This has been implemented in a software interface designated as Vixen2009 which
effectively superseded the original detonation
model described by Cunningham et al. (2006).
A breakage engine or kernel to model the
explosive/rock interaction process through a
hybrid numerical scheme. This has been implemented in an interface designated as Blo-Up.
Existing design and analysis software tools such
as JoinStats and JKSimBlast that support the
building process of larger and more complex
rock mass models and blast layouts.
The Blo-Up interface provides the platform for
model input parameters as well as the graphical
display of specific outputs. These include items
such as 3D graphical representations of micro
and macro fracturing, fragment size distributions,

343

internal fragment damage, pressure histories and

velocity histories.
The original geomechanical model adopted in
the breakage engine relied on the particle flow code
PFC3D developed by the Itasca consulting group
(Itasca 2008). Although this numerical scheme
was shown to be valid and capable of achieving
the main objectives of the project, computational
speed and the model building process became a significant limitation. As a result, the lattice scheme
was introduced, tested and implemented in the
HSBM framework. It is now the basis for the geomechanical (explosive/rock interaction) platform
of the HSBM code.
A detailed description of the lattice scheme is
given by Furtney et al. (2009). In general, the lattice scheme consists of an assembly of point masses
(or nodes), each of which is connected by springs
to its neighbours. In this approach each spring
operates in both the shear and normal directions.
The lattice method applies forces to point masses,
which have only translational degrees of freedom,
and the connecting springs have a tensile breaking
strength. Through this simplified coding structure,
model building and calculation speeds have been
significantly improved.
3

THE HSBM CODE EVALUATION

AND VALIDATION APPROACH

Over the life of the HSBM project, a consistent

evaluation and validation approach has been
adopted which consists of three distinct stages:
1. Evaluations using consistency tests (simple simulations to evaluate the overall behaviour of the
code).
2. Qualitative and quantitative validation tests by
direct comparison with published laboratory
and full scale information.
3. Quantitative validation tests by direct comparison with new controlled and fully instrumented
blasting experiments.
Results discussed in this paper correspond to
both qualitative and quantitative validation tests.
The validation criteria adopted in this particular
work include assessing that:
the scheme is able to reproduce realistic fracture
patterns (e.g. radial fractures, bifurcations and
tensile fractures from boundary reflections);
fragment formation and size calculation algorithms in the HSBM are able to reproduce fragmentation distributions from blasted rock or
rock-like materials;
fragment velocities are within and follow the
trends of values measured under controlled conditions; and

near field rock mass damage profiles in large

scale blasts are consistent with results observed
in production environments.
As part of the HSBM project, several numerical
experiments have been conducted using both small
and large scale experimental data. The results of
a selected number of tests are presented and discussed in this paper and include:
Blasting experiments of concrete blocks and
cylinders by the University of Leoben, Austria
(Reichholf 2003).
Blasting experiments of concrete blocks by the
Norwegian Geotechnical Institute (Kristiansen
et al. 1990).
Consistency test to evaluate the damage profile
of a production scale single hole blast.
4

THE LEOBEN BLASTING

EXPERIMENTS

One of the objectives of these set of experiments

was to determine the characteristics of blast fragmentation curves in order to test the hypothesis of
the natural breakage characteristics of rock (NBC)
described by Steiner (1991) and experimentally
supported by Reichholf (2003).
In the first instance, blasting results from cube and
cylinder shaped concrete block samples were used
in the evaluation of the lattice scheme. In particular,
the experiments correspond to the concrete 2 series
described by Reichholf (2003). The mechanical
properties of the concrete 2 sample which have been
used as input into the HSBM numerical experiments are summarised in Table 1.
With regards to the definition of properties and
calibration of the lattice concrete 2 material, the
current scheme allows for the definition of a damping coefficient. This coefficient has been introduced
to account for the attenuation characteristics of a
given material. Evaluations have shown that this
coefficient may be used as a calibration factor to
take into consideration the complexities of rock
or rock like material textural characteristics. In the
case of concrete for example, this may include the
influence of aggregate distribution, packing and
porosity. As a result the determination of this value

344

Table 1.

Concrete 2 mechanical properties.

Mean values
Uniaxial compressive strength (MPa)
Density (kg/m3)
Tensile strength (MPa)
Elastic modulus (GPa)
Poisson ratio

45.3
2100
2.7
29.6
0.2

has become an active subject of research within the

HSBM project In the Leoben numerical experiments, the damping coefficient was determined
using data from other numerical experiments in
concrete samples and further confirmed with the
modelling results of cube test B20_2 discussed in
this paper. Fragmentation comparisons are made
between experimental and modelling results with
both cube and cylinder samples other than the
B20_2 sample.
Cube and cylinder shaped models were blasted
using a PETN charge that was loaded into a central 5 mm diameter hole drilled axially through the
sample. Ideal detonation input data files were generated for PETN at two different densities using
the Vixen 2009 detonation code. The parameters of
these PETN charges are summarised in Table 2.
The geometry and charging conditions used in the
models are given in Table 3 and the graphical output
of the models are shown in Figure 1. It should be
noted that with the exception of the bottom face,
all faces of the cube samples are assumed to be free.
The bottom is defined as a quiet boundary which is
a viscous boundaries implemented in the scheme to
prevent the stress waves from reflecting off model
boundaries. In the case of Cylinder models all
faces of the cylinder were assumed to be free face
boundaries. The implementation of these boundary conditions is a relatively simple task through the
HSBM/Blo-Up software interface.
The model building process in the HSBM code
requires the definition of some specific model
building and execution parameters such as resolution, the time step increase activation time and
the run time. The model resolution defines the
Table 2.

minimum particle size in a model, in this case a

5 mm (0.5 cm). These and other evaluation studies using the lattice scheme have shown that the
model resolutions should be kept at values of
equal or less than one borehole diameter. The
computational time step increase is activated after
this value to decrease model run times. It is generally estimated by two wave transits across the
model. This is assumed to be a time in which the
generated dynamic stress field has already played
a role in creating the main fracture network and
therefore fragment formation. For the Leoben
experiments, model specific parameters are summarised in Table 4.
It is important to document that cube experiments which consisted of approximately 79,000
nodes at the assigned lattice resolution of 5 mm,
took approximately 30 minutes to reach the required
run time of 10 ms. This time was achieved with a
2.66 GHz CPU running one core and using Windows 64 bit XP Professional as the operating system. At the time of writing this paper improvements

Explosive parameters.

PETN from ideal detonation analysis

Density (g/cc)
Velocity of detonation (m/s)
Heat of reaction (MJ/kg)

Table 3.

Cubes

Cylinder

1.1
6030
5.802

1.25
6618
5.868

Model geometry and charging conditions.

Sample
Length (m)
Width (m)
Height (m)
Diameter (m)
Hole diameter (mm)
Hole length (m)
Approx. charge length (m)

Cubes

Cylinder

B20_2
0.2
0.2
0.2

B20_3
0.2
0.2
0.2

5
0.193
0.186

5
0.195
0.189

B50_8

0.480
0.251
5
0.465
0.450

Figure 1. Cube and cylinder models built in the HSBM

Blo-Up interface.

345

Table 4.

Model specific parameters.

Model (lattice) resolution (mm)

Computational time step increase (ms)
Run time (ms)

5
0.3
10

to the code were being implemented in order to take

advantage of multi-thread (multi core) CPU capabilities. This is expected to improve model building
and model run times.
4.1

Fracture patterns and fragment formation

of cube shaped samples

Fracture formation patterns of sample B20_2 are

shown in Figure 2. The model output shows the
radial fractures both at the top and bottom side
of the cube sample and identifies differences given
by the adopted boundary conditions as well as
the location of the initiation point of the explosive charge. The zones in blue indicate regions in
which bonds have been broken implying that damage has occurred. Figure 2 shows the presence of
major and minor radial fractures as well as fracture
bifurcations.
Figure 3 shows the results of a fragment colouring algorithm which allows the identification of
areas that are expected to contribute to fine and
coarse fragmentation. The image is a slice view
taken at the centre of the explosive charge and
shows in blue the potential source of fine fragments. As has been shown experimentally, modelling results show the extent of intense fracturing
and therefore fine fragmentation in the vicinity of
the borehole (Svahn 2002, Reichholf 2003).

Figure 2.

4.2

Fracture patterns and fragment formation

of cylinder shaped samples

Fracture formation patterns of sample 50_8E are

shown in Figure 4. Fracture intensity is greater on
the right hand side of the model or towards the direction of the detonation front. The left side of the cylinder or the face nearest to the initiation point shows
a well defined set of radial fractures and fragments.
Figure 5 shows sections taken at the centre and along
a section of the explosive charge. These images show
the extent of damage caused by the explosive charge,
in particular in the vicinity of the borehole.

Figure 4.

Fracturing pattern of cylinder sample 50_8.

Figure 5.

Fracturing pattern of cylinder sample 50_8.

Fracturing pattern of cube sample B20_2.

Figure 3. Fracturing patterns and fragment shapes of

cube sample B20_2.

346

4.3

Fragmentation output of cubed

and cylinder shaped samples

The current HSBM code uses a very simple

approach to determine the size distribution of
fragments. A fragment is numerically defined as
a group of nodes linked together by bonds that
have not broken. The total volume of a fragment
is determined by adding the volume from all the
nodes forming that fragment, noting that the volume of a node is directly defined by the resolution
of the model. The size of a fragment is simply
estimated by assuming that all fragments are cubic
in shape and hence the cube root is taken to obtain
a size value. At the time of writing this paper, the
sampling of the cumulative % passing curve generated by the implemented algorithm was a function
of the model resolution and did not follow a standard screen size sampling regime.
Figure 6 shows the cumulative percent passing versus screen size for cube sample B20_3.
A preliminary analysis shows that the modelled
fragmentation follows a realistic trend with an
acceptable prediction of the expected top size. The
model predicts the 50% passing fraction (P50) to be
approximately 47 mm in comparison to a measured
value of 38 mm or a difference of 23%. Due to the
model overestimation of sizes in the intermediate
to coarse end, the coefficient of uniformity defined
as the ratio between P80 and P30 is higher than the
experimentally measured value (i.e. 2.8 versus 2.0).
Differences between modelled and measured values can be explained in part by the influence of the
assumed cubic shape of fragments to calculate size.
It is interesting to note that there is a marked change
in slope or inflection point at approximately 10 mm
which is approximately twice the lattice model
resolution. This may be an indicator of the limit at
which size distributions may be reliably predicted as
a function of model resolution.
Figure 7 shows the cumulative percent passing
versus screen size for the cylinder sample 50_8.

Figure 6.
B20_3.

Fragmentation of concrete cube sample

There is a close match between the predicted and

measured 50% passing fraction (P50) with values
of 47 mm and 46 mm respectively. The difference in
top size is approximately 11% with a predicted value
of 140 mm versus a measured value of 126 mm.
As shown, the uniformity of fragmentation in the
coarse end is significantly different to the measured
uniformity, modelling results estimate a coefficient
of uniformity of 4 which is approximately twice
that of the value of 1.8 measured experimentally.
Overall differences in the uniformity of fragmentation may be explained by a more pronounced effect
of the cubic particle shape assumption in the estimation of size. As illustrated in Figure 8, results
from the cylinder model indicate more elongated
and triangular shaped fragments bounded by well
defined radial fractures. The colour grading ranges

Figure 7.
50_8.

Fragmentation of concrete cylinder sample

Figure 8. Fragmentation regions and fragment shape

of cylinder sample.

347

from blue to red to differentiate between regions

of finer and coarser fragmentation respectively.
As with the cube sample, larger discrepancies are
observed at the size fraction below 10 mm, which
as mentioned earlier, it is approximately twice the
lattice model resolution.
5

THE NGI BLASTING EXPERIMENTS

The objective of the NGI experiments was to

develop a small scale test method to predict how
different explosive parameters affect heave and
fragmentation. These experiments were conducted
and reported by Kristiansen et al. (1990). The
available data consisted of results from small scale
blasts of concrete blocks of 0.5 m 0.5 m 0.5 m
in size, blasted with different explosive types
which were loaded into a 22 mm diameter hole. As
described by Kristiansen et al. (1990), the explosive
charge was placed at the centre of the block and
the samples placed in a steel chamber.
The above information was used to evaluate the
implementation of a gas flow logic in the lattice
model by modelling fragment velocities and comparing it to values measured under controlled conditions. Three explosive types were chosen in these
tests, they included ANFO, Emulsion and PETN.
The ideal detonation characteristics of these explosives calculated using the Vixen 2009 code are given
in Table 5.
Details of the concrete cube model built in the
HSBM Blo-Up interface are shown in Figure 9.
It should be noted that with the exception of the
bottom face, all faces of the cube were assumed
to be free. The bottom face was defined as a quiet
boundary. In this case velocity histories used to
track the displacement of particles where located at
the corners and centre of the explosive charge. The
properties of the concrete samples and model specific conditions are summarised in Tables 6 and 7.
5.1

NGI modelling results

Figure 10 shows a typical displacement versus time

plot for one of history points located at the corner
Table 5.

Explosive parameters of NGI tests.

Ideal detonation analysis

Density (g/cc)
Velocity of
detonation (m/s)
Heat of reaction
(MJ/kg)

ANFO

Emulsion

PETN

0.8
4844

1.0
5247

1.4
7171

3.792

2.205

5.952

Figure 9. NGI model geometry and charging conditions.

Table 6.

Concrete model parameters.

Intact properties
Unconfined compressive strength (MPa)
Density (kg/m3)
Tensile strength (MPa)
Elastic modulus (GPa)
Poisson ratio

55
2480
5
30
0.2

Table 7. Model specific parameters.

Model (lattice) resolution (mm)
Computational time step increase (ms)
Run time (ms)

22
1
50

of a sample. As shown, after approximately 4.2 ms,

the fragment being tracked reaches the end of the
outer boundary or bounced plane located 0.5 m
away. This outer boundary is currently used to
minimize the calculation space of a model. From
this plot, the average velocity of the tracked fragment is of the order of 12 m/s.
The estimated maximum velocity of fragments for all three explosive types is summarised
in Table 8. This data is also plotted against the
available energy of each explosive and shown
in Figure 11.
The trend of increasing fragment velocity against
available explosive energy is consistent with what
has been shown experimentally by Kristiansen
et al. (1990) for similar explosives. However, in
comparison with experimental measurements,
modelling results appear to underestimate the

348

At the time of writing this paper, a new set of

controlled blasting experiments in large concrete
blocks (e.g 3 m 3 m 3 m) were being conducted and analysed in order to further evaluate
and improve the current implementation of both
the borehole and gas flow logic. The advantages of
these new set of experiments carried out in South
Africa, is that the trajectory of fragments were
being directly tracked using high speed video cameras and hence direct comparison could be made
with particle tracking in the numerical model.
Work is continuing in this area.
6
Figure 10. Displacement versus time for particle tracking at the corner of a sample blasted with a PETN
charge.
Table 8.

Maximum fragment velocity.

Explosive

m/s

ANFO
Emulsion
PETN

10.4
9.3
12.2

NEAR FIELD DAMAGE PROFILE

OF PRODUCTION SCALE SINGLE
HOLE BLAST

This consistency test was designed to evaluate the

ability of the lattice scheme to represent a discontinuous rock mass and be able to realistically determine a damage profile behind a production scale
single hole blast. As explained earlier, its evaluation is initially qualitative and relies only on what
has been previously observed under the assumed
conditions and hence relies only in engineering
experience and judgement.
The parameters adopted in this particular consistency test included:

Blasthole diameter = 270 mm

Hole length = 17.7 m
Length of charge = 10.2 m
Explosive type = ANFO (ideal)
Stemming length = 7.5 m
Burden = 5.5 m
Blast type: Free faced.

The modelled rock mass consisted of sandstone

with one horizontal and one sub vertical set of
discontinuities. Modelling results after 500 ms are
shown in Figure 12. This isometric view from the
Figure 11.

Fragment velocity versus available energy.

expected maximum fragment velocities. It is important to mention that the NGI data was the only
sourced available in which velocity measurements
were taken under controlled conditions and the
data was limited to a specific set of procedures. In
these experiments, fragment velocities were indirectly measured by identifying the arrival time of
particles hitting the a steel cylinder surrounding
the sample. An accelerometer attached to the wall
of this cylinder was used for this purpose. This
method differs from the direct tracking of a point
in a sample as measured in the numerical experiments. In this analysis, it has been assumed that
these two techniques are comparable.

Figure 12.
model.

349

Isometric view from the back of the bench

back of the model shows the degree of fracturing,

and in particular the damage accumulated along
discontinuity planes. Front and side section views
taken at the plane of the blasthole are shown
in Figure 13. Results show the modelled extent of
damage or profiles from this single hole blast as
well as the extent of damage along discontinuities. The side view slice defines a damage extent of
approximately 6 m at the top and 2.8 m at the bottom; fracturing however is also shown to extend
along the discontinuity planes.
From a practical perspective, having a better
understanding of the extent of damage along
planes of weaknesses such as weak discontinuities may provide a way of evaluating and quantifying the risk of associated with the potential
loosening of in situ blocks behind the boundaries
of a blast.

Figure 13. Front and side sections view along the plane
of the blasthole.

CONCLUSIONS

Several tests have been conducted to evaluate the

newly implemented lattice scheme in the HSBM
code. Direct comparisons are made between modelled and experimental results with respect to the
ability of the lattice scheme to realistically reproduce fracturing patterns, fragment distributions,
fragment velocities and overall large scale blast
damage profiles.
Results from small scale concrete cube and
cylinder experiments have shown that the lattice scheme, at resolutions equal or less than the
borehole diameter, is able to realistically reproduce
radial fractures, bifurcations and fracturing patterns that are directly influenced by the direction
and propagation of the detonation front as well
as boundary conditions such as free faces. The
fragmentation predicted in the cube and cylinder
experiments follow realistic trends. In the cylinder
test a close match was found between the predicted
and measured 50% passing fraction (P50). The difference in top size is approximately 11%. The uniformity of fragmentation is significantly different
to the measured uniformity and this was attributed
to the impact of the cubic particle shape currently
assumed in the estimation of size. In the cube samples modelling results show an overestimation of
the mean passing fraction of the order of 23%,
however overall fragmentation uniformity appears
to be well matched. In all cases larger discrepancies were observed at the size fraction below 10 mm
which is twice the lattice model resolution. This
may be an indicator of the limit at which size distributions may be reliably predicted as a function of
model resolution. The source and location of fine
and coarse fragmentation appears to be adequately
captured by the models.
With regards to fragment trajectories and
velocities, preliminary analysis has indicated that
the trend of increasing fragment velocity against
available explosive energy is consistent with what
has been shown experimentally. However, modelling results appear to underestimate the indirectly
measured maximum fragment velocities reported
in the NGI data set. A new set of controlled blasting experiments in large concrete blocks have been
conducted in order to further evaluate and improve
the current implementation of both the borehole
and gas flow logic.
A simple consistency test of a production scale
single hole blast has confirmed the ability of the
HSBM code to realistically define the extent of
breakage and post blast damage, of particular
interest is the ability to capture breakage envelopes and the damage concentrated along weak
discontinuity planes.

350

Results from this paper should be viewed as

work in progress as at the time of writing this
paper, further analysis was being conducted and
improvements made in several fronts as part of the
final year extension of the HSBM project. In addition, work is being conducted to better understand
the sensitivities of key input parameters which
forms part of the on-going validation work.
ACKNOWLEDGEMENTS
The authors wish to gratefully acknowledge the
sponsors of the HSBM project who are De Beers,
Debswana Diamond Company, Anglo Base Metals
(Chile), Codelco IM2, LKAB, Sandvik Mining
and Construction, Dyno Nobel Asia Pacific, Rio
Tinto and African Explosives Limited (AEL). In
addition the authors wish to thank the Executive
Steering Committee (ESC) members for allowing
the publication of this paper. Finally, the authors
would like to acknowledge the contributions of
Ewan Sellers of AEL to the evaluation and validation efforts of this project.
REFERENCES
Cunningham, C., Braithwaite, M. & Parker, I. 2006.
Vixen detonation codes: Energy input for the HSBM.
Proc. 8th Int. Symp. on Rock Fragmentation by Blasting
(Fragblast 8), Santiago, Chile, 711 May, pp. 169174.
Santiago: Editec.

Furtney, J., Cundall, P. & Chitombo, G. 2009. Developments in numerical modelling of blast induced rock
fragmentation: Updates from the HSBM project. In
J.A. Sanchidrin (ed.), Proc. 9th Int. Symp. on Rock
Fragmentation by Blasting (Fragblast 9), Granada,
Spain, 1317 September. Rotterdam: Balkema.
Itasca. 2008. PFC3D (Particle Flow Code in 3
Dimensions), Version 4.0. Minneapolis: Itasca Consulting Group, Inc.
Kristiansen, J., Kure, K., Vestre, J. & Bergqvist, I. 1990.
An investigation of heave and fragmentation related
to explosive properties. Proc. 3rd Int. Symp. on Rock
Fragmentation by Blasting, Brisbane, Australia, 2631
August, pp. 8390. Parkville, Vic: Australasian Institute of Mining and Metallurgy.
Reichholf, G. 2003. Experimental investigation into
the characteristics of particle size distributions of
blasted material. PhD Thesis. Montanuniversity
Leoben, Austria.
Ruest, M., Cundall, P., Guest, A. & Chitombo, G. 2006.
Developments using the particle flow code to simulate rock fragmentation by condensed phase explosives. Proc. 8th Int. Symp. on Rock Fragmentation by
Blasting (Fragblast 8), Santiago, Chile, 711 May,
pp. 140151. Santiago: Editec.
Steiner, H.J. 1991. The significance of the Rittinger
equation in present day comminution technology.
Proc. XVII. Int. Min. Proc. Congr., Dresden, 2328
September, pp. I: 177188.
Svahn, V. 2002. Generation of fines around a borehole:
a laboratory study. Proc. 7th Int. Symp. on Rock Fragmentation by Blasting (FRAGBLAST 7), Beijing,
China, 1115 August, pp. 122127. Beijing: Metallurgical Industry Press.

351