You are on page 1of 14

Rock Mechanics and Rock Engineering

https://doi.org/10.1007/s00603-019-01822-8

ORIGINAL PAPER

Influence of Joint Conditions and Blast Design on Pre‑split Blasting


Using Response Surface Analysis
A. K. Raina1 

Received: 20 August 2018 / Accepted: 20 April 2019


© Springer-Verlag GmbH Austria, part of Springer Nature 2019

Abstract
Blasting is a major method of excavation in mining, civil construction and infrastructure projects. The integrity of the
rockmass of the final wall after excavation is important for the stability and economics of the operation. Perimeter blasting
achieves a planned surface of the rockmass. The pre-split is one such technique, deployed when there is no free face and, thus,
involving several interactions of rockmass properties, including strength, joint spacing and joint orientation, with respect
to blastholes, blast design, explosive configuration, blasthole deviations and other variables. The mechanism and models of
the pre-split are perfunctory, particularly in defining interactions of pre-splitting variables. The basis of pre-splitting vis-à-
vis dominant variables affecting performance is, accordingly, discussed here. Since the half cast factor, generally used for
damage assessment, has its limitations, the index of blast damage has been compared while modelling performance. Models
using response surface analysis for pre-splitting, for both blast damage index and half cast factor, were developed using
historical data. The angle of the blasthole with respect to the major joint orientation, spacing of the joints, blasthole spacing,
drill deviation, linear charge concentration and compressive strength has been used to compare the above-mentioned dam-
age criteria. Blast damage index has been found to be a better predictor for determining rockmass damage due to blasting.
The desirability of the variables deployed has been determined using optimization procedures. A fresh impetus to pre-split
studies is expected to provide a ground for future research, particularly studies using advanced computational and numeri-
cal algorithms.

Keywords  Blast damage · Perimeter blasting · Contour blasting · Rock properties · Joint orientation · Response surface
analysis

List of Symbols pb Estimated blasthole pressure


Sj/Sb Ratio of joint spacing to pre-split hole spacing pn Normal component of blasthole pressure
f(Sj/Sb) Resolved values of Sj/Sb pt Tangential component of blasthole pressure
qlc Linear charge concentration (kg/m) vmax Peak particle velocity
θ Angle of inclination of blasthole between ρr Density of rock
blasthole and major joint orientation in radians vp p-Wave velocity of rock
δdd Drill deviation k Rock constant
σc Unconfined uniaxial compressive strength IBD @1m Blast damage index calculated at 1 m from
(weighted average) also referred to as UCS or blasthole and HCF% is half cast factor
uniaxial compressive strength of the rock percentage
σt Tensile strength of intact rock
σtd Dynamic tensile strength estimated from com-
pressive strength 1 Introduction
dh Blasthole diameter
Blasting is the preferred method of rock breakage in civil
* A. K. Raina and other rock engineering projects. There are several rea-
rainaji@cimfr.nic.in sons for the preference. One is the fact that blasting exploits
the tensile strength of the rock during breakage. In addi-
1
CSIR-Central Institute of Mining and Fuel Research, Unit‑I, tion, ease of operation and cost-effectiveness, barring the
Telangkhedi Area, Civil Lines, Nagpur 440 006, India

13
Vol.:(0123456789)
A. K. Raina

handling of explosives, make blasting more popular. There less than 25 ms between blastholes produced good results.
are, however, several side effects of blasting. Apart from Worsey (1984) concluded that quasi-static gas pressure is
vibrations, air overpressure, flyrock and fumes, blasting also a major reason for pre-splitting and that fractures intersect
results in damage to the parent rock that may extend to sev- discontinuities in the rock approximately at right angles,
eral diameters of the blasthole. To avoid damage to the par- indicating that joint orientation results in a geometric effect
ent rock, several methods have been devised that fall into the on final face profiles. Tariq and Worsey (1995) conclude that
category of perimeter blasting. These include pre-splitting, an open joint behaves like a free face and, hence, the spac-
contour blasting and buffer or trim blasting (Jimeno 1995; ing of the blastholes in jointed rocks has a major impact on
Hustrulid 1999). The methods have been used in various the formation of split lines. Szuladzinski and Saleh (1996)
types of excavations, from underground to surface, to pro- stress the role of tensile strength, and favour a quasi-static
duce a smooth wall or to create slopes that are more stable approach to the formation of a split line over the dynamic
than those produced by production blasting. Pre-splitting is approach. However, it may be pointed out that many theo-
a special blasting technique to achieve smooth rock surfaces. ries of pre-split formation are based on two-dimensional
Pre-split blastholes are drilled at predefined spaces, charged static models and, hence, arguments that use the quasi-static
with carefully planned amounts of explosive and fired before mechanism to explain anything in blasting are questionable
the production holes are exploded. (Blair 2015).
Mainstream theories of blasting, as applied to pre-split One of the main reasons for such assumptions may be the
blasting, are based on the premise that there is an initial simplification of analytical results as, in quasi-static theo-
shock loading that creates a fracture between pre-split holes ries, equations of equilibrium are used to replace equations
and that is further extended or widened by gas pressures. of motion to avoid complexity in modelling. Inertial forces,
The debate on the role of shock and gas from explosive deto- rate-dependent material behaviour and reflected stress waves,
nation in rock breakage is still ongoing as summarised by forming parts of the dynamic system, are, thus, ignored in
Minchinton and Lynch (1996). Prima facie, the mechanism such quasi-static approaches (Anderson 1995). Blair (2003)
appears to be simple but things are much more complex in pointed out that the two-dimensional dynamic theory pre-
nature. There are inherent rock properties and blast-induced dicts that the tensile component of hoop stress is of lower
effects that finally define the outcome of the pre-split. magnitude and is unlikely to produce pre-split formation.
As of today, there are around 200 references in the pub- The only way to achieve a tensile stress of the magnitude
lic domain (e.g. Google Scholar) and 270 references in the to achieve a pre-split would be to introduce a nearby reflec-
private domain (ISEE online Reference Database) where tor, preferably parallel to the split line. Minchinton (2011)
pre-split or pre-splitting finds mention in the title or text of used the two-dimensional, dynamic model of Minchinton
published papers that pertain to different aspects and case and Lynch (1996) which included gas flow, crack forma-
studies of pre-split blasting. A comprehensive literature tion and growth, and verified that pre-splits cannot form in
review, including the mechanics of blast damage, methods an extended medium without nearby reflectors. Dyskin and
of measurement, damage prediction methods, classification Galybin (2000), however, arrived at conclusions that are the
systems and advantages of damage control, has been pro- opposite of the findings of Minchinton (2011), and point to
vided by Raina et al. (2000). Verma et al. (2018) reviewed the fact that quasi-static model predictions can be mislead-
damage induced by blasting and developed new models for ing as the method overestimates the tensile component of
predicting the same. Xiao et al. (2018) observed that the sta- hoop stress.
bility of rock structures depends on the method of excavation Damage to rockmass due to blasting is primarily deter-
and is time dependent. mined by charge diameter and, hence, by peak particle
Sun (2013) compiled details of at least 16 models of ana- velocity owing to the strain generated, rock tensile strength,
lytical nature that are currently used to determine the zone rock impedance, blasthole spacing and rockmass being
of damage due to blasting. However, none of these models blasted. The analysis of vibrations by Blair (2015) showed
involves travelling shear waves despite evidence that sug- that, despite scatter in the data, the vibration reduction fac-
gests that damage to rockmass is primarily dependent on tor across the pre-splits was around 40% pointing to the fact
these waves (Blair 2010). that the pre-split was a thin damage zone rather than a plane
Kutter and Fairhurst (1971) concluded that the spacing pre-split as conceived generally.
and firing time of pre-split blastholes determine the per- The spacing, aperture and orientation of joints in a rock-
formance of the split along the centre line connecting such mass in relation to the pre-split line have a significant impact
holes. Konya (1980) observed that pre-split fractures were on the profile of the final wall obtained after the split (Singh
formed as a result of gas pressure and stress waves contrib- 2001). Drilling error is one of the main factors in such cases.
uted little to this process. In addition, he reported that instan- However, owing to the complexity of the problem and model-
taneous firing of blastholes was not necessary and a delay of ling constraints, there are very few publications addressing

13
Influence of Joint Conditions and Blast Design on Pre‑split Blasting Using Response Surface…

the role of rock joints on the pre-split. Damage is more prom- 2.1 Geology of the Dam Area
inent at the open surface due to lack of confinement. Lewan-
dowski et al. (1996) while studying coal mine overburden The rock formation seen in this area revealed composite for-
concluded that, with pre-split, quality decreased, notably with mation of limestone, quartzite and shale bands. The salient
increased joint frequency and orientation of 5°–30° with pre- details of the formations at various workings are as follows:
split holes. The solid angle subtended by the pre-split holes
finds no mention in research as it complicates calculations. 1. The rock formation in the power house area is hard com-
This reduces the problem to a two-dimensional one. posite limestone and shale.
Despite many case studies and a few empirical models for 2. The dam foundation and stilling basin area expose alter-
blast-induced damage to rockmass, a comprehensive model nate layers of thinly bedded shale and quartzite with
incorporating rock, explosive, joints and their disposition moderate joint spacing.
vis-à-vis drill hole orientation, and drilling variables is lack- 3. The lower portions of the power house (50–33 level) are
ing. Moreover, most models for blast-induced damage to the composed of thinly layered shale, olive green shale and
rockmass do not address the interactions of the variables quartzite with mild water seepage.
involved and their behaviour in space. This formed the basis 4. The joints are highly disturbed at the top 10–15 m as was
of this research with an intention to lay a foundation for evident from an excavation done earlier. Solution cavi-
future investigations for evolving a comprehensive model ties and worn-out joint faces were clearly visible during
to predict and optimise procedures. With modern day com- the initial phases of excavation.
putational and instrumentation capabilities, it is incumbent 5. Three major joint sets are exposed in the area, out of
on the research fraternity to deploy more resources for the which the bedding joint is horizontal and the other two
exact understanding of the pre-split process. Accordingly, joints are vertical and intersect each other at around 80°–
this effort is practically an initiative for future research and 90°. The disposition of joints in the quartzite formation
development to fully understand the interaction of variables as observed on the surface is provided in Fig. 1.
during the pre-split process.
The three joints exposed in the area of study have simi-
lar attitude in all the litho-units visible in the study area.
2 Study Area, Variables and Data Quartzite and limestone exposed in the area show moder-
Presentation ately spaced joints while shale is thinly laminated.

To ascertain the role of different rockmass and blast design 2.2 Engineering Properties of Intact Rocks
variables on pre-split blasting results, a rationale was devel-
oped for such analysis on the basis of existing literature and Different engineering properties of the rockmass were
historical data. The data for the study were generated at a observed in the field using standard methods. Data about
hydropower project in India where significant pre-splitting intact rock were acquired from an analysis of core samples
was conducted. The selected site had four major areas of acquired from rocks in the area. The engineering properties
excavation that are defined as under. used in the analysis are provided in Table 1.

1. Power house area on the right bank of a river with a high


wall of 30–35 m height.
2. Shear key trench on the bedrock in the foundation area of
the dam, the walls of which were designed to be inclined
at an angle of 45° with a vertical depth of 3.5 m.
3. Key trenches connected with shear key but with vertical
walls and a depth of 3.5 m.
4. Tail race channel with mildly inclined walls.

The factors involved in the interactions presented here can


be broadly classified into rockmass, blast design and drilling.
These are regarded as independent variables and are taken as
input. Half cast factor and blast damage index are dependent
in nature. Such variables were determined at the above site
during the excavation and are detailed below.
Fig. 1  Disposition of joints in quartzite observed in the area

13
13
Table 1  Engineering properties of the rocks
Serial no. Sample no. Rock type Direc- Uniaxial compres- Average uniaxial P-wave S-wave Density (g/cm3) Shear Young’s Poisson’s ratio
tion w.r.t sive strength compressive strength velocity velocity modulus modulus
sample (MPa) (MPa) (km/s) (km/s) (GPa) (GPa)

1 46 Black shale Across 6.6 1.4 2.14 4.1 12.1 0.5


2 46 Black shale Along 47.1 3.2 1.6 2.14 5.5 14.6 0.3
3 46 Black shale Along 69.1
4 46 Black shale Along 119.4
5 46 Black shale Along 161.8 99.4
6 250 Dark quartzite Across 104.7 6.6 2.1 0.4
7 262 Dark quartzite Across 22.3 5.1 2 0.4
8 2 Dark quartzite Across 66 64.3 5.1 1.3 2.53 4.5 13.0 0.5
9 250 Dark quartzite Along 96.7 5.2 1.2 1.98 2.7 8.1 0.5
10 262 Dark quartzite Along 27.4 4.5 2.2 2.32 11.5 30.8 0.3
11 2 Dark quartzite Along 30.6 2.89
12 262B Dark quartzite Along 42.4 49.3
13 11 Dark red limestone Across 16.3 5.3 2.2 0.4
14 8 Dark red limestone Across 40.9 28.6 7.2 2.5 0.4
15 11 Dark red limestone Along 59.4 6.8 2.1 2.79 12.5 36.2 0.4
16 8 Dark red limestone Along 70.3 7.3 1.7 2.39 7.3 21.3 0.5
17 8 Dark red limestone Along 131.8 87.2 1.9
18 89 Grey shale Across 10.4
19 90 Grey shale Across 12.8 4.2 2.2 0.3
20 88 Grey shale Across 31.9 18.4 5.2 2.1 0.4
21 89 Grey shale Along 152 2.3
22 90 Grey shale Along 28.1 2.8 1.2 2.61 3.5 9.8 0.4
23 88 Grey shale Along 85.7 88.6 2.9 1.4 1.96 4.1 10.8 0.3
24 OG shale Olive green shale Along 98.1 98.1 2.73 0.0
25 61 Yellow shale Across 8.5 8.5 3.4 0.9 0.5
26 61 Yellow shale Along 52.7 2.1 0.4 2.18 0.3 0.8 0.5
27 61B Yellow shale Along 92.1 72.4 2.24
A. K. Raina
Influence of Joint Conditions and Blast Design on Pre‑split Blasting Using Response Surface…

2.3 Definitions of Input, Output Variables the final wall during the pre-split. The proximity of the blast-
and Rationale hole to the joints defining wedges precludes the possibility
of achieving smooth surfaces after blasting. Even if only
The terminology used in the data and analysis is provided the major joint is considered, its orientation vis-à-vis the
below. blasthole orientation is no longer simple but presents a solid
Rock Strength Strength of rock primarily determines angle that is further complicates modelling. To simplify the
failure under dynamic loading conditions. Since the com- analysis, the angle between joint orientation and blasthole
pressive and tensile strengths of a material are related, new (θ) was resolved into a linear parameter by standardising the
fracturing is also determined by the property. Despite the angle in radians.
fact that the in situ strength of the rockmass has varying To understand the role of joints in pre-split and to lay
strengths owing to discontinuities and their conditions, we a foundation for future modelling, simple mechanisms of
rely on intact rock strength to predict pre-split performance. damage to rockmass are defined, keeping in view the main
Pre-split Hole Spacing The blasthole spacing used for focus of the paper. Two main conditions are presented here.
the pre-split line is defined in several texts and can be deter-
mined by the following method after Sanden (1974), Calder Condition 1  Major joint at right angles to pre-split blasthole.
(1977) and Chiappetta (1991):
[ ] In this condition, it is assumed that the major joint is
pb + 𝜎 t
Sb ≤ 2dh , (1) oriented perpendicular to the pre-split blasthole. Two com-
𝜎t ponents of the pressure generated in the blasthole are con-
where Sb is the spacing between the pre-split holes, dh is the sidered as acting at right angles to the blasthole and are
diameter of the blasthole, pb is blasthole pressure (approxi- called the normal (pn) and tangential pressures (pt) as shown
mated from velocity of detonation and density of explosive) in Fig. 2. All other faces are highly confined and the only
and σt is the tensile strength of the rock. unconfinement is on the top face that represents the open
Drill and Charge Diameter A drill with 115 mm diameter surface of the rock.
and charge diameter of 25 mm was used in all the pre-split As can be observed in Fig. 2, the orientation of the joints
blasts and can thus be treated as constant. Drill deviation favours damage and high-pressure gases can penetrate the
(δdd) from planned to actual was calculated from the meas- joints, resulting in unwanted breakage, particularly in near-
ured data and absolute values were used for the analysis. surface operations. Two real-time examples in support of
Joint Spacing Joint spacing (Sj) is a major factor influenc- the above assertion, obtained from the current study area,
ing pre-split. It is generally believed that larger joint spacing are given in Figs. 3 and 4.
yields better results while smaller spacing may not yield A series of interactions of the joints with incipient pres-
good results in pre-splitting. However, the case may be dif- sures can be observed in Figs. 3 and 4. These range from dam-
ferent with change in angle of inclination of blasthole with age to the top surface, mobilisation of primary and second-
respect to joint orientation. Joint spacing and hole spacing ary joints, wedging, block loosening and ploughing of rock
(Sb) data showed strong correlation and, to avoid aliasing
S
during modelling, a ratio of S j was used. In addition, the ratio
b
Sj
Sb
was further resolved into a linear form by deploying a
S S
simple spherical function, S j = 34 𝜋r3 , where S j is the volume
b b
of the equivalent sphere with radius r.
Linear Charge Concentration (qlc) The variable is very
important in the case of pre-splitting as the amount of explo-
sive used per linear metre defines the pressures that are
exerted on the walls of blastholes. A general practice is to
use fully decoupled charges and, hence, the decoupling ratio
plays a major role in defining the final output of blasts. In
the present case, explosive and drill diameters were uniform.
Hence, the decoupling ratio was also constant.
Joint Orientation Joint orientation is a complex issue in
blasting as explosive pressures interact on all planes and in
all dimensions. Thus, pre-splitting outcome can be strongly
influenced by their disposition in space. The orientation of Fig. 2  Normal and tangential pressures acting on joint planes (blast-
minor joints also plays a vital role in defining the shape of hole is oriented perpendicular to the major joint

13
A. K. Raina

Fig. 3  Results of pre-splitting when a major joint is oriented perpen-


dicular to a blasthole (example 1)

Fig. 5  Normal and tangential pressures acting on joint planes (blast-


hole is oriented at 45° to the major joint

Fig. 4  Results of pre-splitting when a major joint is oriented perpen-


dicular to a blasthole (example 2)

fragments about joint intersections. One major observation in


this case is that despite high HCF% percentage, i.e. > 90%, the
damage zone in the rock is deep. This observation also sug-
gests that HCF% may not be a good index of pre-split blasting
Fig. 6  Results of the pre-split blast where blasthole is inclined at 45°
efficiency as is also proved later by the results of analysis.
to the major joint

Condition 2  Major joint at 45° to pre-split blasthole.

In this condition, it is assumed that the major joint is


oriented at 45° to the pre-split blasthole. Again, the two
components of the pressure generated in the blasthole are
considered to act at right angles to the blasthole and are
called the normal (pn) and tangential pressures (pt) as shown
in Fig. 5. All other faces are highly confined and the only
unconfinement is on the top face that represents an open
surface of the rock as shown in Fig. 5.
From Fig. 5, it is observed that tangential pressures are
acting in a compressive manner and the expanding gases
are not able to open up the joints, resulting in damage. The
damage zone in this case is significantly reduced or practi-
cally absent. The effect is also demonstrated and supported Fig. 7  Contrast in results when blasthole is oriented 45° and 90° to
by Figs. 6 and 7. the major joint

13
Influence of Joint Conditions and Blast Design on Pre‑split Blasting Using Response Surface…

A close examination of Figs. 6 and 7 reveals that, despite 3 Pre‑split Blast Data
the presence of closely spaced joints and deviation in drill-
ing, damage to the parent rock is minimal. Figure 7 provides The pre-split design for the proposed final line of excavation
further insight into the role of the orientation of the joints on was made with the help of Eq. 1. The rest of the production
pre-splitting by contrasting the performance of the pre-split holes were designed with the standard blast design used in
in 45°- and 90°-oriented blastholes, particularly when Sj/ open cut or wedge cut designs as per the requirements. Half
Sb < 1. One of the observations from the figure is that joint cast factor (HCF%) for blasted rock was calculated with the
orientation has a major influence on pre-splitting perfor- help of digital image analysis software, Fragalyst 4.2©, in
mance. This is established through the analysis in the fol- which digital images of the blasted surface were acquired,
lowing text, also. calibrated and the lengths of half casts were measured and
Half Cast Factor Pre-split performance is generally logged. Since HCF% is a ratio, scale effects arising from
measured in terms of half cast factor (HCF%), the ratio of image analysis are eliminated. The results of the tests in
visible half casts to total drill length. HCF% is very popu- limestone formation without pre-splitting are given in Fig. 8.
lar, simple and easy to calculate. However, HCF% does not Figure 8 shows two major joints in limestone crossing
work below 40% and does not consider deeper damage to each other at around 90° that have led to the formation of
rockmass leading to under- or over-estimation of the dam- wedges and resulted in damage to the rockmass and the
age. Even a 100% value of HCF% does not mean that the production of irregular surfaces due to blasting as no pre-
rockmass is not damaged at all, as demonstrated above. splitting was designed in the area. The pre-split results of
The observation thus presents a significant gap between the high wall (approx. 30 m high) in the powerhouse are
rockmass damage result assessment and interpretation, and shown in Fig. 9.
model development. As is seen in Figs. 9 and 10, the surface of the final wall is
Index of Blast Damage (IBD; Yu and Vongpaisal 1996) quite smooth and cut along the desired line of interest. Some
was used as a measure of damage assessment. Although the
index requires several variables for calculation, it is worth-
while, considering the importance of blasting near perma-
nent structures. Damage in terms of IBD was calculated at
a distance of 1 m from the blasthole from the following
equation:
Vmax × 𝜌r × vP
IBD = , (2)
k × 𝜎td

where vmax is the peak vector sum of ground vibration


(mm/s) calculated from the vibration regression equation
obtained from field tests, ρr density of rock (g/cm3), vp
weighted p-wave velocity of rock in m/s, k the strength of
rock obtained from rockmass rating and σtd is the dynamic
Fig. 9  Final wall shape with pre-splitting in limestone formation
tensile strength of the rock calculated from uniaxial com-
pressive strength using the method of Mohanty (1987).

Fig. 10  Final wall shape with pre-splitting in limestone and shale for-


Fig. 8  Final wall shape without pre-splitting in limestone formation mations

13
A. K. Raina

Measurements for the input variables of 265 pre-split


blastholes were made and data were compiled for further
analysis. An initial analysis was done with the help of artifi-
cial neural networking (ANN; Raina et al. communicated).
The importance and sensitivity of all the variables measured
in the field were worked out with the help of ANN. The
most important and sensitive variables determined by ANN
were selected for statistical analysis and modelling of blast-
induced damage to rockmass. Averages of the variables thus
selected for further analysis are given in Table 2.

4 Data Analysis
Fig. 11  The shape of shear key as excavated using pre-split blasting Response surface method or analysis (RSA) is a statistical
technique to trace the relationship(s) in space between sev-
eral explanatory variables and one or more response vari-
ables. The method, introduced by Box and Wilson (1951),
aims to use designed experiments to obtain a response that
is optimal. However, with RSA, it is also possible to ana-
lyse historical data from tests. DesignExpert is software that
uses the RSA technique to develop models. The software
helps find relationships between independent and dependent
variables in three-dimensional space through the advanced
analysis of variance method (ANOVA). The software allows
selection of different algorithms such as backward, forward,
stepwise or all hierarchical propagation for the elimination
of unwanted terms during model development.
Response surface analysis (RSA) of the historical data
of Table 2 was used to analyse, deduce the statistics and
Fig. 12  Overall picture of the dam foundation area showing shear key develop the models. RSA was performed with the help of
and key trenches DesignExpert© software using the data of columns 1–5
as independent and that of columns 6 and 7 in Table 2 as
damage to the rock can be seen in the collar region (top) dependent variables. The initial evaluation of the data sug-
owing to least confinement and ploughing of rock across gested a quadratic model. Backward propagation was used to
the major joints. eliminate unwanted terms in the analysis as the technique is
Figure 10 also shows damage to the wall rock in the col- considered more robust. The analyses of ANOVA for HCF%
lar regions of the bench that can be attributed to weathered and IBD were performed independently while evolving sepa-
or disturbed formation at the top and ploughing due to the rate models for both output(s). The ANOVAs of the model
placement of blastholes in the area where vertical joints for HCF% and IBD are presented in Table 3.
intersect. The regression statistic of the analysis thus conducted is
The final shape of the pre-split blast and the formation of presented in Table 4.
a shear key in shale formation are shown in Fig. 11. From Table 4, it is evident that the predictability of blast-
Figure 11 indicates that the pre-split was performed as induced damage to rockmass is better defined in terms of
planned and was controlled particularly owing to favourable IBD than in terms of the HCF% as the calculated, adjusted
blasting and joint conditions. The overall picture of blasting and predicted correlation coefficients in case of IBD con-
in the area showing shear key and key trenches excavated form to the norms. The plot of distribution of externally
with the use of pre-splitting is depicted in Fig. 12. studentized residuals (Fig. 13) and the plot of predicted vs.
Figure 12 shows that the performance of pre-split in the actual IBD (Fig. 14) confirm the results. Although the results
shear key area was better than that in the key trenches despite of calculated and adjusted R2 in case of HCF% are in good
being in the same formation. The damage to the rockmass agreement (Table 3), the predicted R2 is quite low as also
was controlled by the disposition of the joints with respect evident from Figs. 15 and 16. One of the reasons for the
to the blasthole direction. above results is that IBD is a ratio of resisting and breaking

13
Influence of Joint Conditions and Blast Design on Pre‑split Blasting Using Response Surface…

Table 2  Data of the pre-split Sl. no. Sj/Sb f(Sj/Sb) qlc (kg/m) θ δdd σc (MPa) IBD at 1 m HCF%
blast and its outcome for
selected 1 2 3 4 5 6 7

1 0.35 0.44 0.64 1.57 0.97 40.7 1.9 59


2 0.43 0.47 0.93 1.57 0.90 63.7 2.3 51
3 0.31 0.42 0.57 1.57 0.96 40.7 1.5 65
4 0.29 0.41 0.93 1.57 0.95 63.7 2.1 63
5 0.40 0.46 0.93 1.57 0.96 63.7 2.1 63
6 0.31 0.42 0.64 1.31 0.83 63.7 1.3 87
7 0.31 0.42 0.50 1.31 0.89 63.7 2.6 28
8 0.19 0.36 0.93 1.31 0.91 63.7 2.4 51
9 0.18 0.35 0.50 1.31 0.83 63.7 2.3 55
10 0.38 0.45 0.93 1.31 0.90 63.7 1.7 62
11 0.11 0.30 0.71 0.79 0.79 54.5 3.8 80
12 0.19 0.36 0.64 0.79 0.80 54.5 2.2 84
13 0.16 0.34 0.71 0.79 0.82 54.5 2.9 79
14 0.16 0.34 0.86 0.79 0.89 54.1 2.4 68
15 0.16 0.34 0.86 0.79 0.88 54.1 2.0 64
16 0.15 0.33 0.93 0.79 0.83 54.5 3.2 56
17 0.16 0.34 0.71 0.79 0.89 54.5 2.9 83
18 0.19 0.36 0.57 0.79 0.92 54.1 3.1 97
19 0.19 0.36 0.71 0.79 0.55 54.5 2.5 91
20 0.14 0.32 0.57 1.57 0.82 54.5 2.6 80
21 0.14 0.32 0.57 1.57 0.97 54.5 2.6 74
22 0.20 0.36 0.64 1.57 0.94 54.5 1.8 85
23 0.17 0.34 0.71 1.57 0.86 54.5 2.8 79
24 0.12 0.31 0.86 1.57 0.92 54.1 3.1 57
25 0.24 0.39 0.64 1.57 0.97 54.5 1.9 66
Min. 0.11 0.30 0.50 0.79 0.55 40.7 1.25 28.00
Max. 0.43 0.47 0.93 1.57 0.97 63.7 3.78 97.00
Std. dev. 0.09 0.05 0.15 0.35 0.09 6.27 0.57 15.29

Sj/Sb ratio of joint spacing to pre-split hole spacing, f(Sj/Sb) are the resolved values of Sj/Sb, qlc is linear
charge (kg/m), θ is angle between the pre-split hole and the major joint orientation, δdd is the drill devia-
tion, σc is the unconfined uniaxial compressive strength (weighted average), IBD @1  m is blast damage
index calculated at 1 m from blasthole and HCF% is half cast factor, Note joint angle has been normalised
to 0–1, σtd is the dynamic tensile strength estimated from the compressive strength, HCF 0–100% with
maximum value of 100 assumed to indicate best pre-split performance, IBD values ≤ 0.1 indicating least
damage and ≥ 2.0 indicating major failure

Table 3  ANOVA for reduced Source model Sum of squares df Mean square F value p value prob > F Comment
models of HCF% and IBD
IBD 8.02 15 0.53 18.39 < 0.0001 Significant
HCF% 5210.84 12 434.24 8.21 0.0005 Significant

Sj ( )
forces, in contrast to the superficial nature of the HCF%. In IBD = 37.9 − 48.5
Sb
+ 74qlc + 11.4(𝜃) − 79.9𝛿dd − 1.1 𝜎c
addition, IBD is based on the principles of induced strain and, Sj
hence, gives better results. − 1.7 𝜎c + 7.2qlc (𝜃) − 63.8qlc 𝛿dd − 0.5qlc 𝜎c
The final form of the models as obtained from the RSA Sb
]2
is given in the following equations for I BD and HCF%, Sj
[
− 11.1𝛿dd (𝜃) − 0.6𝜎c (𝜃) + 2.5𝛿dd 𝜎c − 183.9
respectively: Sb
+ 11.1(𝜃) + 2
0.01𝜎c2 , (3)

13
A. K. Raina

Table 4  Regression statistic for the two different responses


Statistic IBD HCF% Comment

Standard deviation 0.17 7.27


Mean 2.42 69.08
C.V. % 7.05 10.53 The coefficient of variation for this model. It is the error expressed as a percentage of the mean
PRESS 1.51 4054.06 The predicted residual sum of squares for the model. A measure of how well a particular model fits
each point in the design
R2 0.97 0.89
Adjusted R2 0.92 0.78 R2 adjusted for the number of parameters in the model relative to the number of points in the design
Predicted R-squared 0.82 0.31 Predicted R-squared is a measure of how good the model predicts a response value
Adequate precision 18.06 12.06 Adequate precision (signal to noise ratio) to use model to navigate design space (> 4 is desirable)

Fig. 13  Residual vs. run number for IBD Fig. 15  Residual vs. run number for HCF%

Fig. 14  Predicted vs. actual values of IBD Fig. 16  Predicted vs. actual values of HCF%

13
Influence of Joint Conditions and Blast Design on Pre‑split Blasting Using Response Surface…

Sj The DesignExpert software allows visualisation of the


HCF% = −510.8 − 2145.7 + 956.5qlc − 589.9(𝜃) change in response in three dimensions of the output with
Sb
Sj change in input variables. Such simulations have been car-
ried out for average values of the input variables used in
( )
+ 1684𝛿dd + 10.7 𝜎c − 1399.3 (𝜃)
Sb Eqs. 3 and 4. The interactions of the variables incorporated
Sj in the above models during such simulations for IBD are
+ 4424 𝛿dd − 741.2qlc 𝛿dd + 12.3qlc 𝜎c
Sb HCF% are shown in Figs. 17, 18, 19, 20, 21, and 22.
+ 19.5(𝜃)𝜎c − 50.1𝛿dd 𝜎c − 704.3q2lc . There is an optimal solution to Sj/Sb with respect to
(4)
IBD (Fig. 17) along with the prediction of higher damage
Although there are 21 terms entering into the ANOVA, with higher qlc values but the response surface of a simi-
the back-propagation method eliminated six insignificant lar iteration with HCF% (Fig. 18) shows a reverse trend.
terms during the analysis while maintaining the hierarchy A perfect solution to Sj/Sb and θ is presented by the IBD
of the model. The results of ANOVA and evaluation of model (Fig. 19), while the HCF% model (Fig. 20) does not
model terms are further explained with the help of Table 5 offer such a solution. The results of IBD with qlc and Sj/Sb
that indicates that the model and the terms entering into the (Fig. 21) indicate that, as qlc increases, the damage can be
equation developed are statistically significant. The models of exponential nature even with higher values of σc. The
confirm the earlier assertion that blast-induced damage is results in such cases of the HCF% model (Fig. 22) present
a complex phenomenon that cannot be resolved by simple a linear trend of damage. This further confirms that HCF%
linear regression owing to complex interactions within the is not the best suited index for the determination of damage
independent variables. to rockmass.

Table 5  ANOVA for response Source Sum of squares df Mean square F value p value Prob > F
surface model (partial sum of
squares) Model 8.02 15 0.53 18.39 < 0.0001
A—Sj/Sb 0.044 1 0.044 1.52 0.2482
B—qlc 1.37 1 1.37 46.94 < 0.0001
c—Theta 0.17 1 0.17 5.81 0.0393
D—drill deviation 1.46 1 1.46 50.25 < 0.0001
e—UCS (σi) 0.019 1 0.019 0.64 0.4456
A × e 0.49 1 0.49 16.68 0.0027
B × c 0.99 1 0.99 33.99 0.0002
B × D 1.37 1 1.37 47.04 < 0.0001
B × e 1.26 1 1.26 43.39 0.0001
c × D 0.74 1 0.74 25.45 0.0007
c × e 0.054 1 0.054 1.86 0.2054
D × e 1.89 1 1.89 65.08 < 0.0001
A2 1.19 1 1.19 40.80 0.0001
c2 0.053 1 0.053 1.84 0.2085
e2 0.31 1 0.31 10.71 0.0097
Residual 0.26 9 0.029
Cor. total 8.28 24

Where df is degrees of freedom attributed to the blocks, generally equal to one less than the number of
blocks, F value is test for comparing model variance with residual (error) variance. If the variances are
close to the same, the ratio will be close to one and it is less likely that any of the factors have a significant
effect on the response. Calculated by model mean square divided by residual mean square; p value is the
probability value, usually relating the risk of falsely rejecting a given hypothesis. Prob > F is probability of
seeing the observed F value if the null hypothesis is true. Small probability values call for rejection of the
null hypothesis. The probability equals the proportion of the area under the curve of the F-distribution that
lies beyond the observed F value
The Model F value of 18.39 implies the model is significant. There is only a 0.01% chance that an F value
this large could occur due to noise. Values of “Prob > F” less than 0.0500 indicate model terms are signifi-
cant. In this case, B, c, D, Ae, Bc, BD, Be, cD, De, A2, and e2 are significant model terms. Values greater
than 0.1000 indicate the model terms are not significant

13
A. K. Raina

Fig. 17  IBD response with Sj/Sb and qlc interaction Fig. 20  HCF% response with Sj/Sb and theta interaction

Fig. 18  HCF% response with Sj/Sb and qlc interaction Fig. 21  IBD response with qlc and UCS interaction

Fig. 22  HCF% response with qlc and UCS interaction


Fig. 19  IBD response with Sj/Sb and theta interaction

Based on the results from the RSA, an optimization rou- process. Several solutions for different combinations of vari-
tine was run for 20 solutions to ascertain the desirability of ables can be seen to give a desirability of 1.0. The table can
various independent variables using the IBD model and is be used to optimise the results of pre-split blasting in similar
presented in Table 6. situations and thus provides a basis for the development of a
Table 6 also shows the constraints applied and the results comprehensive model.
of desirability identified with the numerical optimization

13
Influence of Joint Conditions and Blast Design on Pre‑split Blasting Using Response Surface…

Table 6  Desirability and IBD Solution Sj/Sb qlc θ δdd σc IBD Desirability


for different simulations based number
on Eq. 2
1 0.14 0.51 0.79 0.55 61.09 − 1.78 1.00
2 0.33 0.56 0.79 0.55 62.32 − 2.76 1.00
3 0.22 0.57 0.79 0.55 60.22 − 1.85 1.00
4 0.32 0.58 0.79 0.55 54.20 0.35 1.00
5 0.25 0.70 0.79 0.55 63.53 − 0.80 1.00
6 0.12 0.50 0.79 0.55 52.12 − 0.05 1.00
7 0.19 0.61 0.79 0.55 61.92 − 1.31 1.00
8 0.28 0.64 0.79 0.55 56.19 0.48 1.00
9 0.15 0.61 0.79 0.55 55.77 0.58 1.00
10 0.29 0.77 0.79 0.55 62.56 0.29 1.00
11 0.14 0.51 0.79 0.55 50.72 0.09 1.00
12 0.17 0.56 0.79 0.55 51.21 0.65 1.00
13 0.32 0.55 0.79 0.55 63.44 − 3.24 1.00
14 0.27 0.55 0.79 0.55 62.23 − 2.89 1.00
15 0.43 0.73 0.79 0.55 61.22 0.69 1.00
16 0.13 0.50 0.79 0.55 48.36 0.65 1.00
17 0.41 0.50 0.79 0.56 63.72 − 3.60 1.00
18 0.12 0.50 0.81 0.55 48.71 0.70 0.99
19 0.43 0.66 0.82 0.55 63.72 − 1.61 0.99
20 0.43 0.78 1.01 0.55 60.32 0.70 0.92

5 Conclusions Acknowledgements  The author is thankful to the Director CSIR-


Central Institute of Mining and Fuel Research, India, for his permis-
sion to publish the paper. The sponsorship of APGENCO is gratefully
Models for HCF% and IBD have been developed for pre- acknowledged. Thanks to all my colleagues for their help and coop-
dicting blast-induced damage to the rockmass using histori- eration during data generation. My profound thanks to Sri Ramesh
cal data from a hydropower project and response surface for his critical views all along the study. Thanks to Suraj for help with
references and Geeta and KP Madhu for critically examining the text
analysis. The dimensions of the model were initially reduced
for errors. Thanks to anonymous reviewers whose initial criticism has
using artificial neural networks (not part of this paper) and helped in improving the paper significantly.
relationships were established using multivariate regres-
sion. The performance of HCF% and IBD models has been
evaluated through such analysis. The index of blast damage References
proved to be a better predictor than the widely used half cast
factor. The interactions of various variables have been pro- Anderson TL (1995) Fracture mechanics, fundamentals and applica-
vided. The role of various variables on pre-split performance tions, 2nd edn. CRC Press, Boca Raton
Blair DP (2003) A fast and efficient solution for wave radiation from a
is thus defined through comprehensive equations. There are,
pressurized blasthole. Int J Blast Fragment 7:205–230
however, shortcomings as the results are based on constant Blair DP (2010) Seismic radiation from an explosive column. Geophys-
drill diameter, explosive diameter and, hence, decoupling. ics 75(1):E55–E65
Nature of explosive and explosive diameter to drill diameter Blair DP (2015) Wall control blasting. In: Eleventh Int Symp Rock
Fragmentation by Blasting, 24–26 Aug., Sydney, pp 13–26
ratio will need to be looked into. Moreover, it is observed
Box GEP, Wilson KB (1951) On the experimental attainment of opti-
that, if Sj/Sb < 1, then Sj plays a major role in defining the mum conditions (with discussion). J Roy Stat Soc B 13(1):1–45
final condition of the wall and the angle of the blasthole with Calder P (1977) Pit slope manual, Chapter 7, Perimeter blasting. CAN-
respect to the major joint orientation assumes a greater role. MET, Energy, Mines and Resources Canada, Report 77–14, p 82
Chiappetta RF (1991) Pre-splitting and controlled blasting techniques
However, results for Sj/Sb > 1 will need evaluation. Since the
including air decks and dimension stone criteria. In: Chiappetta
data presented used only values of 45° to 90° of inclination RF (ed) Proc blast technology instrumentation and explosives
of the blasthole to the orientation of the major joint of the applications seminar, San Diego
rockmass, there is further scope for defining the relation- Dyskin AV, Galybin AN (2000) Crack interaction mechanism of
pre-split rock blasting. In ECF13, San Sebastian 2000. http://
ships over a 180° angle. Resolution of the solid angle of the
www.grupp​ofrat​tura.it/ocs/index​.php/esis/ECF13​/paper​/downl​
pre-split blasthole also presents a complex task for future oad/8426/4874. Accessed 17 Nov 2018
investigations.

13
A. K. Raina

Hustrulid W (1999) Blasting principles for open pit mining. Vol. 1 Sanden BH (1974) Pre-split blasting. M.Sc. Dissertation. Queen’s
general design concepts. AA Balkema, Rotterdam, pp 293–312 University
Jimeno CL, Jimeno EL et al (1995) Drilling and blasting of rocks. AA Singh SP (2001) The influence of geology on blast damage. CIM Bull
Balkema, Rotterdam, pp 252–270 94(1048):121–127
Konya CJ (1980) Pre-split blasting, theory and practice. In: AIME Sun C (2013) Damage zone prediction for rock blasting. Ph.D. Thesis
Annual Meeting Las Vegas, Nevada, Feb 24–28, pp 80–97 University of Utah
Kutter HK, Fairhurst C (1971) On the fracture process in blasting. Int Szuladzinski G, Saleh A (1996) Mechanism of smooth blasting and its
J Rock Mech Mining Sci Geomech Abst 8(3):181–202 modelling. In: Proceedings of the Fifth International Symposium
Lewandowski T, Luan Mai VK, Danell RE (1996) Influence of dis- on Rock Fragmentation by Blasting, 25-29 August, Montreal, pp
continuities on presplitting effectiveness. In: Proceedings of the 195–200  
Fifth International Symposium on Rock Fragmentation by Blast- Tariq SM, Worsey PN (1995) An investigation into the effect of varying
ing, 25–29 August, Montreal, pp 217–225 joint aperture and nature of surface on pre-splitting. In: 22nd Ann
Minchinton A (2011) The importance of understanding detonics and Conf on Expl and Blasting Tech, Feb. 4–8, Orlando, pp 186–193
rock response under explosive loading. In: First Int Blasting Geol- Verma HK, Samadhiya NK, Singh M et  al (2018) Blast induced
ogy Workshop, Melbourne, Australia rockmass damage around tunnels. Tunn Undergr Space Technol
Minchinton A, Lynch P (1996) Fragmentation and heave modelling 71:149–158
using a coupled discrete element gas flow code. In: Proceedings Worsey PN (1984) The effect of discontinuity orientation on the suc-
of the Fifth International Symposium on Rock Fragmentation by cess of pre-split blasting. In: Proc 10th Ann Conf on Expl and
Blasting, 25–29 August, Montreal, pp 71–80 Blasting Tech, Jan 29–Feb 2, Lake Buena, Vista, pp 197–213
Mohanty B (1987) Strength of rock under high strain rate loading Xiao XH, Xiao PW et al (2018) Large deformation characteristics and
conditions applicable to blasting. In: Proceedings of the 2nd reinforcement measures for a rock pillar in the Houziyan under-
International Symposium on Rock Fragmentation by Blasting, ground powerhouse. Rock Mech Rock Eng 51(2):561–578
FRAGBLAST-2, Keystone, Colorado, August 23–26, pp 72–79 Yu TR, Vongpaisal S (1996) New blast damage criteria for under-
Raina AK, Chakraborty AK et al (2000) Rock mass damage from ground blasting. CIM Bull 89(998):139–145
underground blasting, a literature review, and lab-and full-scale
tests to estimate crack depth by ultrasonic method. Fragblast Publisher’s Note Springer Nature remains neutral with regard to
4(2):103–125 jurisdictional claims in published maps and institutional affiliations.

13

You might also like