Technical Papers

Fragmentation modeling
using the Multiple Blasthole Fragmentation
(MBF) model at an openpit mine
by R. Yang, C. McAllister, J. Berendzen and D. Preece

Abstract ■ The Multiple Blasthole Fragmentation (MBF) model models multiple explosive charge
contributions and the effect on fragmentation of delay timing with its associated scatter for each
blasthole. The model uses near-field blast vibration attenuation parameters and the ground
p-wave velocity as inputs for part of the in situ rock property to model rock fragmentation. It
models most blast design parameters explicitly and simulates the effect of wave reinforcement
due to the interaction of simultaneously arriving waves or diminishing cooperative contribution
from long delay intervals between charges within a blasthole or among blastholes. The
fragmentation size is calculated at three-dimensional grid points within a blast, and the fines
and oversized blocks are treated explicitly. The model takes a surveyed irregular geometry of
the free face of a blast as the calculation boundary.
This paper presents a case study on applying the MBF model at an openpit mine. Near-field
vibration measurements from signature hole blasts were conducted to obtain the stress-wave
magnitude and attenuation parameters as well as ground sonic velocity. A production blast was
then monitored with the corresponding fragmentation measured, serving as site-specific inputs
to the MBF model. Various blast design scenarios were then simulated to develop ones that
provide better fragmentation to improve mill throughput for the mine.

Mining Engineering, 2016, Vol. 68, No. 10, pp. 49-54.

Official publication of the Society for Mining, Metallurgy & Exploration Inc.
http://dx.doi.org/10.19150/me.6801

Introduction multiple blastholes with varying delay lay time of each charge and (4) broken
The semiempirical fragmentation intervals. Most of these models are not ground screening, which accounts for
models currently found in the literature three-dimensional. Furthermore, these blasthole confinement and interac-
(Cunningham, 1987; Lownds, 1995) can- models predict an average fragmenta- tion. The fragmentation distribution is
not explicitly model contributions from tion size within a volume surrounding calculated at up to several millions of
a blasthole and use the average size to three-dimensional grid points within a
R. Yang and D. Preece, members generate a size distribution over the blast. A grid point in a detached rock
SME, are principal research fellow volume. The percentage of fines and mass due to fragmentation by earlier
and global technology manager oversize blocks is included in the distri- firing holes ceases to receive energy
of blast modelling development bution. Typically, such models are inad- from later firing blastholes. Fines and
and support for the Americas, equate for predicting fines and oversize oversized blocks are treated explicitly.
respectively, at Orica USA Inc., blocks (Cunningham, 1987). The model accepts inputs from
C. McAllister is mine engineer/ The Multiple Blasthole Fragmenta- blast design parameters, such as: (1)
blasting supervisor at Simplot tion (MBF) model, whose concepts and location and orientation of each blast-
Phosphates LLC and J. Berendzen site-specific applications are presented hole, (2) stemming length, (3) blast-
is technical service engineer in Yang (2014, 2015a and 2015b), mod- hole diameter, (4) multiple decking,
at Southwest Energy Inc. Paper els rock fragmentation mechanisms (5) bench height, (6) initiation points
number TP-15-053. Original and size distributions based on the in a blasthole and initiation sequence
manuscript submitted November work of Seaman, Curran and Shockey within a blast pattern and (6) explo-
2015. Revised manuscript (1976) and calculates peak particle ve- sive strength. It models the effect on
accepted for publication May 2016. locities (PPVs) according to Yang and fragmentation of delay timing with its
Discussion of this peer-reviewed Scovira (2008). Included are: (1) the associated scatter for each blasthole.
and approved paper is invited nonlinear charge weight superposition, The model also has statistical model-
and must be submitted to SME (2) the effect of waveform broadening ing capabilities for geological random
Publications by Jan. 31, 2017. using a variable width time window variation of the attenuation of particle
for each explosive charge, (3) the de- velocities, ground sonic velocity scat-
www.miningengineeringmagazine.com Mınıng engıneerıng    OCTOBER 2016 49
Technical Papers
tering, and timing scatter of blasthole delays (Yang and interest to the charge, and d is the function of the location of
Lownds, 2011). dl along the charge length in meters.
If a distance is defined, the charge-weight-scaled dis-
PPV calculation. Persson, Holmberg and Lee (1994) tance can be calculated with the effective charge weight us-
used integration of the linear superposition of PPVs from ing Eq. (4) and the scaling method above to scale the charge
charge increments along a charge length to calculate the weight to the defined distance. The nearest distance from the
PPV in the vicinity of a blasthole. Such linear superposition point of interest to the dominant charge may be used, as for
may be suitable if the material response is close to linear the charge weight scaling and the scaled distance calculation.
elastic. However, in most cases, in the vicinity of blastholes, The dominant charge is defined as that having the minimum
rock behavior is significantly nonlinear. scaled distance among all charges in the blast (Yang and
For the MBF model, the effective charge weight from Scovira, 2008).
a single charge to a point of interest is calculated by inte- If the stress wave from an explosive charge passes
gration of nonlinear charge weight superposition along the through broken ground created by earlier firing charges, the
charge length. The PPV at a calculation grid point is based amplitude and frequency of the stress wave will attenuate
on nonlinear charge weight superposition from different more than when traveling the same distance through intact
contributing charges accounting for the delay timing be- rock. This additional attenuation due to earlier firing charg-
tween them, the same approach as described in Yang and es is termed “broken ground screening” (Yang and Scovira,
Scovira (2008). A few concepts are described below. 2008) and is used to modify the resultant PPV. Also, earlier
firing blastholes in the vicinity of a firing blasthole will re-
Nonlinear charge weight superposition. Figure 1 shows a duce the confinement to the hole and consequently reduce
point of interest A — a calculation grid point — and an explo- the fragmentation potential (Yang and Kay, 2011).
sive charge demonstrating how the effective charge weight is For modeling rock fragmentation by multiple delayed
calculated using nonlinear charge weight superposition. blastholes, any model — even a full waveform superposition
The charge weight of the segment Δl is: model — must account for the case that fragmented and de-
tached rocks cannot receive further energy input from later
(1)
firing blastholes. For example, the rocks that are fragmented
and moved by blastholes in the first row may not get any
where d is the charge diameter and ρ is the density of the further input from the blastholes in the second row if the
explosive. The effective charge weight, Δwe, at the point A delay between the rows is sufficiently long. Similarly, if a vol-
from the segment Δl is defined as the charge weight that gen- ume of rock is fragmented and displaced by an earlier firing
erates the same PPV contribution as the segment Δl but is blasthole or blastholes it will not receive any further energy
at the nearest distance, dn, from the charge to the point of from a later firing blasthole or blastholes if the delay time is
interest A: sufficiently long.
The MBF model uses two concepts, described in Yang
(2)
and Scovira (2008), to model this phenomenon and the in-
tercharge timing delay effects: (1) a dominant charge at each
(3)
grid point and (2) a variable time window for each charge.

Consequently, the cumulative effective charge weight of the Fragmentation at a point of interest. Seaman, Curran
whole charge at point A is: and Shockey (1976) developed a computational model of
fracture and fragmentation for ductile and brittle materi-
als based on projectile impact experiments and theoretical
(4)
analysis. The model favorably compares fragment size dis-
tributions with measurements and thus appears to be highly
where dn is the nearest distance in meters from the point of relevant and fundamental research work for rock blast frag-
mentation modeling. Aspects of Seaman’s theories on frag-
ment size distribution and crack nucleation are adopted as
the basis for the MBF model.
Figure 1 PPV induced from detonation of explosive charges was
shown in Yang (2014, 2015a and 2015b) to be proportional to
A point of interest A – a calculation grid point – and an
pressure or strain and stress. Based on Seaman, Curran and
explosive charge. Shockey (1976), it is assumed that the number of fractures
that produce rock fragments during blasting is governed by:
(5)

where N0 and η are rock property constants, PPV is the peak

particle velocity in meters per second at the calculation grid
point and PPV0 is the threshold PPV for crack nucleation.
The average fragment size is inversely related to the number
of fractures:

50 OCTOBER 2016    Mınıng engıneerıng www.miningengineeringmagazine.com

 Technical Papers

Figure 2 Figure 3
Five signature holes and a production blast were monitored Recorded waveforms at different distances from a signature
with eight vibration monitors. blasthole.

(6)

where x0 is the initial fragment size in meters. It may be the

in situ block size due to joints.
The rock fragmentation distribution derived by Seaman,
Curran and Shockey is that the cumulative volumetric frac-
tion for fragment sizes smaller than x has the distribution:
(7)
where xc is the critical size of the size distribution in meters,
in direct proportion to the average size of the distribution.

Size distribution of a blasted volume. The average frag-

ment size or any percentage passing size can be displayed
as a volumetric plot, that is, the size can be shown as a func-
tion of the location within the blast. The results can be dis-
played using various contour maps in a cross section cutting
through the blast. A fragmentation size distribution can also
be calculated over a given volume of the blast using:

hole fired, the production blast shown in Fig. 2 was initiated.

(8) The ground p-wave velocity was measured from the signa-
ture-hole blast. The information from the blast vibrations of
where n is the number of grid points in the blast volume as- the signature-hole blast was used as input to the MBF model,
suming that the size of the grid mesh is constant over the while the rock fragmentation measurement from the produc-
volume, xi is the average size in meters at the ith grid point tion blast was used to compare with the MBF modeling re-
calculated from Eq. (6), and x is the passing size in meters. sults and analyze the validity of the MBF prediction.
The passing percentage of a particular size can be calculated In each seismic monitor, three channels recorded the
from Eq. (8). blast vibration from the signature hole blast and the produc-
tion blast, while the fourth channel recorded the detonation
Case study: Phosphate mine time of the first signature blasthole, which is used with the
Five signature holes were blasted, with eight vibration vibration of the blasthole to calculate the ground p-wave ve-
monitors placed in different locations near the blastholes locity. Forty sets of signature hole vibration waveforms and
to record the blasts (Fig. 2). The distances between the eight sets of production blast vibration waveforms were re-
blastholes and monitors ranged from 12 to 67 m (39 to 220 ft). corded. The p-wave velocities of the ground were obtained
The five signature holes were connected with long, two-sec- across different distances. The average of the p-wave veloci-
ond interval delays. Two seconds after the last signature blast- ties was determined to be 1,750 m/s (5,742 ft/s). It was found

www.miningengineeringmagazine.com Mınıng engıneerıng    OCTOBER 2016 51

Technical Papers
that the ground p-wave velocity is low, which is consistent 97.5 percent of measurement data fall below the line. The
with a soft layer of formation of the ground from the bottom bottom curve is the best fit to the measurement data, with 50
of the blast to the monitors mounted on the ground surface. percent of data falling below it or above it. The parameters
From Fig. 3, which shows recorded waveforms at differ- of the two regression curves are used for the Monte Carlo
ent distances to a signature blasthole, it can be seen that the simulation of the PPV variation (Yang and Lownds, 2011).
waveform changes with recording distances. The general
trend is that the farther the recording distance, the longer Rock fragmentation measurement
the duration time — more spread in time — of the waveform Rock fragmentation was measured using digital photo-
and the more complex the waveform becomes. This demon- graphs. The cap rock and interior of the blast were measured
strates the necessity of using multiple seed waveforms to separately. In each region, more than 20 pictures were taken
model the contribution from different blastholes to blast vi- at different time instances during the mucking to obtain the
bration or rock fragmentation at a point of interest. best representative rock fragmentation distribution. Some
Figure 4 shows the regression of the PPVs of signature of the largest blocks were more than 1.5 m (5 ft) in diameter,
hole blasts versus scaled distance. The parameters from the mainly because the rock in the stemming region is much
regression are used as part of the input to the MBF frag- harder than the rock below the stemming region. Figure 5
mentation model or to the Multiple Seed Waveform (MSW) shows the rock fragments under the cap rock or below the
blast vibration model (Yang, Patterson and Scovira, 2009), if stemming region. The rock fragmentation from the interior
blast vibrations are to be predicted. The top curve represents rock is much smaller in size than the cap rock.
the 97.5 percent upper-bound limit line, which means that
MBF modeling of the monitored blast
The MBF model uses the inputs from the signature hole
Figure 4 blast vibrations to predict the fragmentation. Figure 6 shows
the modeled average size of the rock fragmentation distri-
Regression of the peak particle velocity of signature hole bution within the blasted volume, and Figure 7 shows the
blast versus scaled distance (1 in./s = 25.4 mm/s, 1 ft/lb0.5 = model prediction agrees well with measured fragmentation,
0.4526 m/kg0.5). validating the model for the site.

Model predictions of design scenarios

After the model has been calibrated with the model
parameters suitable to the site against the measurement of
fragmentation, various design scenarios can be modeled to
explore favorite design options for better mining economics,
such as optimal fragmentation distribution to improve the
mill throughput.
Figure 8 shows the result of modeling the original design
but with the ANFO explosive mixture replaced with Orica’s
AmexLD 0.6, and with 1.2-m (4-ft) stemming instead of
2.4-m (8-ft) stemming, while maintaining the same bench
height and blast pattern. It indicates that halving the stem-
ming to 1.2 m (4 ft) and using AmexLD, which is low-en-
ergy ANFO, will improve the fragmentation. The scenario
of ANFO with 2.4-m (8-ft) stemming generates more of the
oversized blocks than that of AmexLD with 1.2-m (4-ft)
Figure 5 stemming. The scenario of ANFO with reduced stemming of
1.2 m (4 ft) would generate the finest fragmentation shown
Rock fragments from the interior. in Fig. 8, but fly rock may be an issue due to the proximity
of the mill.
Figure 9 shows a comparison of different delay timings
of interhole and interrow, with 1 ms × 1 ms as the best and
17 ms × 25 ms as the worst, and 2 ms × 5 ms unable to pro-
vide significant improvement compared with 17 ms × 25 ms.
These are consistent with stress-wave enhancement being
more significant if the waveform reinforcement occurs when
the two waves are both closer to the source without much
attenuation. The 1 ms × 1 ms delay pattern may provide
significant fragmentation improvement. However, the blast
vibration and blast overpressure must be examined for spe-
cific site constraints.
Figure 10 shows a comparison of four different blast pat-
terns and loading configurations on the fragmentation re-
sult, including the test production blast — test blast MBF

52 OCTOBER 2016    Mınıng engıneerıng www.miningengineeringmagazine.com

 Technical Papers

Figure 6
The modeled average size of the rock fragmentation distribution within the blasted
volume.

and test blast measurement — from Fig. 7 for reference. cases. This is due to its having the highest powder factor and
The 3.6 × 3.6 m (12 × 12 ft) pattern with pocket charges more even energy distribution in the overburden among the
yields better fragmentation than the test production blast four cases.
even though the former has substantially lower powder fac-
tor of 652 g/m3 (1.1 lb/yd3) than the latter’s 890 g/m3 (1.5 lb/ Discussion
yd3). This could be attributed to the 3.6 × 3.6 m (12 × 12 ft) The MBF model inputs the parameters of the near-field
pattern with pocket charges providing better energy distri- blast vibration attenuation and the p-wave velocity of the
bution in the stemming region than the 2.4 × 2.4 m (8 × 8 ft) in situ rock mass to model the rock fragmentation from a
test production blast. The 1.8 × 1.8 m (6 × 6 ft) pattern with blast design. The rock mass properties, such as strength, stiff-
satellite holes and longer stemming provides better fragmen- ness, density, joint spacing, filling materials between joints
tation than the test production blast even though the former and joint frequency, are encoded in the p-wave velocity and
has slightly lower powder factor than the latter. Again, the vibration attenuation parameters. In addition, the param-
former provides more uniform energy distribution but costs eters x0, N0, PPV0 and η in Eq. (6) are rock properties for
more for drilling and loading. Finally, the pattern of 2.4 × 2.4 fragmentation by blasting. The values of these four param-
m (8 × 8 ft) B × S with 11-cm (4.5-in.) hole and 1.2-m (4-ft) eters could be determined in reference to rock mass proper-
stemming results in the finest fragmentation among the four ties, such as rock strength, joint frequency, spacing and joint

Figure 7 Figure 8
Modeled rock fragmentation distribution for the whole Modeled size distribution: AmexLD 0.6 with 4-ft stemming
blast compared with the measured fragmentation from the versus ANFO with 8-ft stemming.
production blast.

www.miningengineeringmagazine.com Mınıng engıneerıng    OCTOBER 2016 53

Technical Papers

Figure 9 Figure 10
Comparison among different delay timings of interhole and Comparisons of different blast patterns, blasthole diameters
interrow, showing the 1 ms × 1 ms. and stemming region charges.

properties. After determining the initial values from the rock By comparing the modeled scenarios, favorable blast de-
mass properties, they are further adjusted during the model signs were identified and implemented. The percentage of
calibration with the rock fragmentation measurement from oversize material, defined as having particle size of 1.5 m (60
a test blast. in.) or more, was 0.05 percent for the new design versus 4.3
An average size is calculated at each grid point of calcu- percent for the original design. Productivity was improved
lation. A size distribution is assumed as described in Eq. (7). to 18,000 tons of ore per shift, from 10,000 to 11,000 tons
At grid points of calculation near a blasthole or in a region previously. Before the pattern adjustments, the mine would
where shock waves collide, large PPV and small average send 40 loads to the oversize fragment piles but now average
size of fragmentation are calculated. With the assumed size about two loads per shot.
distribution in Eq. (7), the fines of the fragmentation from The case study indicates that the MBF model is useful as
blasting are assumed to be modeled by the MBF. The fines a tool for blast fragmentation optimization with the capabil-
that are generated from material handing, such as digging ity to model most blast design parameters explicitly. ■
and hauling, are not included in the MBF modeling.
It is recognized that the present fragmentation measure- References
ment by the image process could underestimate the fines Cunningham, C.V.B., 1987, “Fragmentation estimations and the Kuz Ram Model – four
years on,” 2nd International Symposium on Rock Fragmentation by Blasting, Society
from blasting. In Fig. 7, the discrepancy between the MBF for Experimental Mechanics, Keystone, CO, pp. 475-487.
prediction and the measured fragmentation could be due to Lownds, C.M., 1995, “Prediction of Fragmentation Based on Distribution of Explosives
underestimation of the fines by the image process. Although Energy,” Research Symposium of International Society of Explosives Engineers
Meeting, Nashville, TN, February 1995.
the image process of rock fragmentation has limitations, it Persson, P., Holmberg, R., and Lee, J., 1994, Rock Blasting and Explosive Engineering,
remains the most widely used technique for rock fragmen- CRC Press LLC, Boca Raton, FL, 560 p.
tation measurement because there is no other method that Seaman L., Curran D.R., and Shockey D.A., 1976, “Computational models for ductile
and brittle fracture,” Journal of Applied Physics, pp. 4814-4826, http://dx.doi.
can provide better results and is also faster, more economi- org/10.1063/1.322523.
cal and accurate. Yang, R., 2014, “A 3D Multiple Blasthole Fragmentation (MBF) Model Based on Peak
Particle Velocity, Nonlinear Charge Weight Scaling, Time Window of Contribution,
Conclusions and Dynamic Rock Fracture Models,” Orica internal report.
Yang, R., 2015a, “A Multiple Blasthole Fragmentation (MBF) Model Based on Nonlinear
Near-field vibration measurements from signature hole Charge Scaling, Delay Time Contribution, and Rock Dynamic Fracture,” 41st Annual
blasts were conducted to obtain the stress-wave magnitude Conference on Explosives & Blasting Technique, New Orleans, LA, ISEE.
attenuation parameters as well as ground sonic velocities. Yang, R., 2015b, “A Multiple Blasthole Fragmentation (MBF) Model – Its Concept,
Formulation, Capability, and Field Comparison Examples,” 11th International
This information along with the blast design parameters are Symposium on Rock Fragmentation by Blasting, Sydney, Australia, Aug. 24-30, 2015.
inputs for the MBF model. By monitoring a production blast Yang, R., and Kay, D.B., 2011, “Multiple seed blast vibration modeling for tunnel blasting
and measuring the corresponding fragmentation, the MBF in urban environments,” Blasting and Fragmentation, Vol. 5, No. 2, pp. 109-122.
Yang, R., and Lownds, M., 2011, ”Modeling effect of delay scatter on peak particle
model was tested for validation. velocity of blast vibration using a multiple seed waveform vibration model,” Blasting
Various blast design scenarios were then simulated to and Fragmentation, Vol. 5, No. 1, pp. 31-46.
develop one that provides better fragmentation so as to im- Yang, R., Patterson, N., and Scovira, D.S., 2009, “An Integrated Approach of
Signature Hole Vibration Monitoring and Modeling for Quarry Vibration Control,”
prove mill throughput for the mine. Parameters varied in 9th International Symposium on Rock Fragmentation by Blasting, Spain, Sept.
this study included (1) blasthole diameter, (2) blast pattern, 13-17, 2009.
(3) strength and density of explosives, (4) pocket charges in Yang, R., and Scovira, D.S, 2008, “A model of peak amplitude prediction for near field
blast vibration based on nonlinear charge weight superposition, time window
the stemming region, (5) satellite holes, and (6) delay timing broadening, and delay time modeling,” Blasting and Fragmentation, Vol. 2, No.
between blastholes. 2, pp. 91-115.

54 OCTOBER 2016    Mınıng engıneerıng www.miningengineeringmagazine.com