International Journal of Rock Mechanics and Mining Sciences 103 (2018) 267–274

Contents lists available at ScienceDirect

International Journal of
Rock Mechanics and Mining Sciences
journal homepage: www.elsevier.com/locate/ijrmms

Empirical and probabilistic analysis of blast-induced ground vibrations T

a a,⁎ b
Sunny Murmu , Priti Maheshwari , Harsh Kumar Verma
a
Department of Civil Engineering, Indian Institute of Technology Roorkee, Roorkee, India
b
Central Institute of Mining and Fuel Research, Regional Centre, CBRI Campus, Roorkee, India

A R T I C L E I N F O A B S T R A C T

Keywords: An attempt has been made to include an additional blast design parameter, burden, in obtaining the vector peak
Blast-induced ground vibrations particle velocity, VPPV. A large set of about 640 blast data pertaining to different rock types from ten different
Vector Peak particle velocity (VPPV) sites in India and Turkey has been collected from the literature. Analysis of these data has been carried out
Blast design parameters resulting in the proposal of an empirical model for peak particle velocity with due consideration to the burden
Burden
along with monitoring distance and maximum charge per delay. The performance of the proposed model has
Probabilistic analysis
been compared with existing models, and the proposed model has been found to serve the purpose of predicting
Monte-Carlo simulations
VPPV with greater accuracy. Further, probabilistic analysis of VPPV has been conducted by performing Monte-
Carlo (MC) simulation on proposed empirical model. Typical results corresponding to Chittorgarh limestone
mines in India are presented. The input parameters namely monitoring distance and maximum charge per delay
have been assumed as lognormally distributed random variables, while the burden has been assumed as discreet
variable. The analysis of results of MC simulations revealed that output or the state variable, VPPV follows a
lognormal distribution. It was possible to take into account the variability in the blast parameters and therefore
to study its influence on VPPV.

1. Introduction Vppv = kD−b (1)

The advent of drill and blast method (DBM) in the mining and civil
engineering industry, brought about a significant reformation in the
rock breaking technology across the globe. Even though other tech-
nology like Rock Breakers, Tunnel Boring Machines, Road Headers,
surface miners etc., have evolved with the passage of time however the
explosives are still used due to its flexibility to deal with any geo-
mining condition. It also brings along other benefits such as, low initial
investment, cheap explosive energy, easy availability, faster rate of
advancement in excavation. On the contrary, it also has many side ef-
fects such as ground vibrations, fly rock, noise etc. Amongst all the side
effects of blasting, ground vibrations play a major role in affecting the
stability of the existing surface structures, be they residential, com-
mercial or structures of historical importance. It is an inevitable and
integral aspect of blasting. The most intriguing portion of blasting is
that approximately only 20–30% of explosive gets utilized in frag-
mentation and throw of rock mass, whereas the remaining 70–80% gets
dissipated in the form of ground vibrations, fly rock, noise, air blasts,
back break etc.1 Literature shows that peak particle velocity (PPV) is
commonly a good index of damage to structures. The most widely
adopted empirical model to predict peak particle velocity is scaled
distance relationship:
where Vppv is peak particle velocity (PPV) in (mm/s); D , the scaled
m m
distance in ( 0.5 or 0.33 ) which is the ratio of distance of monitoring
kg kg
point from the geometric center of the blasting area, R in (m) to max-
imum charge per delay, W in (kg); k and b are site specific constants
which describe the characteristic of propagating media, blast design
and geology.
Many research workers have contributed towards the in-depth un-
derstanding of some of the controllable and non-controllable para-
meters pertaining to blast induced ground vibration (BIGV). There are
various other empirical models similar to Eq. (1). A summary of some
empirical models is listed in Table 1.
An overview of the literature revealed that various factors govern
the intensity of BIGV which can be termed as controllable parameters
and non-controllable parameters,13 as mentioned in Table 2. It was
agreed that the maximum charge per delay, delay period, blast geo-
metry and distance of monitoring location are the governing factors
pertaining to PPV at any point of interest.6,14 The effect of burden was
first studied by Bergmann et al.15 by blasting the granite rock and the
resulting vibrations were measured with the help of pressure gauges.
However, no solid conclusion could be drawn from this study. Blair16
illustrated how the delay interval between blast holes can be chosen to

⁎
Corresponding author.
E-mail address: pritifce@iitr.ac.in (P. Maheshwari).

https://doi.org/10.1016/j.ijrmms.2018.01.038
Received 20 May 2017; Received in revised form 12 July 2017; Accepted 20 January 2018
1365-1609/ © 2018 Elsevier Ltd. All rights reserved.
S. Murmu et al. International Journal of Rock Mechanics and Mining Sciences 103 (2018) 267–274

Table 1 Kounuk10 quantified the effect of discontinuity frequency (λ ) in the

Summary of various PPV prediction models. attenuation of PPV. Elevli and Arpaz20 evaluated the contribution of
different parameters on PPV using relation diagram method (RDM).
Sl. No. Reference Empirical models
They concluded that the effective parameters for the ground vibration
1. Duvall and Petkof2 R −b at the point of blast were explosive amount per delay, burden and
Vppv = k ( )W stemming and in order to reduce the amount of vibration the same
2. Langefors and Kihlstrom3 b
⎛ W ⎞2
parameters should be modified. Blair21 extensively studied the de-
Vppv = k ⎜ 2 ⎟
pendency of PPV on charge length analytically. It was concluded that
⎝ R3 ⎠
3. Ambraseys and Hendron4 −b the peak vibration measured underground due to underground blasting
R
Vppv = k ⎛ 3 ⎞ was not significantly dependent on charge weight and this was con-
⎝ W⎠
4. IS: 69225 2 −b sistent with model predictions for an infinite visco-elastic medium.
⎛ R3 ⎞
Vppv = k
⎜W ⎟ However, the peak vibration measured at the surface and due to surface
⎝ ⎠ blasting was found to strongly depend upon charge weight, due to
5. Ghosh and Daemen6 R −b −αR
Vppv = k ( )W
e geological influences such as near-surface layering. Görgülü et al.22
6. Ghosh and Daemen6 R
−b investigated the effect of blast design parameters and rock properties
Vppv = k ⎛ 3 ⎞ e−αR
⎝ W⎠ using artificial neural network (ANN) and concluded that artificial in-
7. Roy7 R −1 telligence methods give better prediction vis-à-vis the empirical models.
Vppv = n + k ( ) W
Simangunsong and Wahyudi11 estimated the effect of number of coal
8. Bilgin et al.8 −b
Vppv = k ⎛
R
⎞ *B γ layer (Nc) between blasting area and monitoring point and incident
⎝ Q ⎠
angle on PPV. Kumar et al.12 considered the effect of rock mass prop-
9. Rai and Singh9 Vppv = kR−bW ae
−αR

10
erties UCS (fc), unit weight (γ), RQD, GSI on PPV.
10. Ak and Kounuk −b
Vppv = k ( )R
W
λα A close look at the literature suggests that various efforts have been
11. Simangunsong and Wahyudi 11 −b invested to study the parameters that affect blast induced ground vi-
R
Vppv = k ⎡ (1 + cosθi + Nc ) ⎤
brations and utilize them to predict peak particle velocity. However
⎣ Q ⎦
12. Kumar et al.12 fc 0.642 D−1.463 miniature effort have been taken to quantify the effect of blast design
Vppv = ,
γ parameters even though it is a controllable parameter and it governs
0.642 −1.463
(0.3396 * 1.02GSI GSI1.13) D ground vibrations markedly. This has been found to be still debatable
Vppv =
γ
whether ground vibrations, in terms of peak particle velocity, is influ-
enced by burden, one of the blast design parameters, or not. Therefore,
Table 2
the need has been felt to study the influence of burden as blast design
Controllable parameters and non-controllable parameters affecting BIGV. parameter (if any) in evaluating vector peak particle velocity (VPPV).
Further, in order to account for the variability of parameters influen-
Controllable Parameters Uncontrollable Parameters cing VPPV, a need has also been felt for a probabilistic analysis. In view
of this, in the present work, an attempt has been made first to propose a
Blast Design Explosive parameters Geotechnical and Geo-
Parameters mechanical parameters model for prediction of VPPV employing multiple regression analysis of
10 different mining sites in India and Turkey. The parameters like
• Hole diameter • Explosive type • Rock mass strength burden, maximum charge per delay and monitoring distance have been
• Hole depth • Maximum charge • Ground water condition
• Bench height per delay • Discontinuity frequency
used to develop the VPPV predictive equation. Subsequently, a prob-
• Burden • Charge per hole • Bedding plane abilistic analysis has been carried out with the help of Monte-Carlo
• Spacing • Total charge simulations on proposed model.
• Stemming • Powder factor
• Sub drilling • VOD
• No. • Delay
2. Data acquisition
of holes and time
rows • Direction of
• Hole inclination initiation For the development of an empirical model, blast data have been
collected from the literature pertaining to ten different open cast
mining sites: seven from India and three from Turkey. Details of the
control and minimize the vibration energy within the structural re- data are summarized in Table 3. The data set for Chittorgarh Mines and
sponse band of most houses. Induced vibrations were found to be de- six other mine sites22 in India are shown in Figs. 1–7 as VPPV vs. scaled
pendent upon the accuracy of the delay initiators as well as the level of distance plot for various values of the burden. The data pertaining to
random fluctuations between each blast hole signature. Bilgin et al.8 Turkey sites can be obtained from respective publications as mentioned
carried out an extensive study in a lignite mine in order to eliminate in Table 3 and therefore have not been presented here.
environmental influences as a result of blasting. The authors made a
reassessment by adding burden (B ) as a parameter to classical particle 3. Deterministic analysis: empirical model
velocity prediction models and came to a conclusion that there was
1–20% recovery in the regression coefficient. Blair17 depicted that how To develop an empirical model for determination of vector peak
the separate influence of variables, like weight and type of explosive particle velocity (VPPV), first of all an attempt has been made to
used per delay, the delay time sequence, scatter in that sequence, identify the parameters influencing VPPV. Critical review of literature
spatial pattern of blastholes and properties of the transmitting medium, suggests that maximum charge per delay (W) and monitoring distance
may be analyzed using a Monte Carlo model that has a predictive power (R) are the two basic parameters significantly affecting the vector peak
superior to that of the traditional charge weight scaling laws. Blair and particle velocity. As mentioned above, there are both the thoughts exist
Armstrong18 pointed out that the ground vibrations are affected by the in literature. Few research workers state that VPPV is independent of
condition of rock mass surrounding the blast hole and the vibrations burden and some of the research workers concluded that burden also
were found to be independent of burden. Uysal et al.19 exclusively in- influences VPPV. In view of this, the data has been analyzed to obtain
vestigated the effect of burden on blast induced vibration on open pit the expressions for VPPV in two manners: once by considering the
mines and discovered that burden width had a significant impact on burden (model 1) and another by not taking the burden into account
PPV. It was proved that vibration reduces as burden increases. Ak and (model 2) while analyzing the data.

268
S. Murmu et al. International Journal of Rock Mechanics and Mining Sciences 103 (2018) 267–274

Table 3
Summary of blast data of ten different sites.

Reference Project site No. of observations Range of parameters

Burden, B (m) Monitoring distance, R Charge per delay, W PPV (mm/sec)

(m) (kg)

Verma et al.23 Chittorgarh Limestone Mines 233 3–6 33–1370 25.6–414.09 1.01–96.5
Singh et al.24 West Mudidih Colliery Seam (WMCS) IV, India 22 2–5 28–122 125–990 2.5–3.9
West Mudidih Colliery Seam (WMCS) III, India 44 2.6–5 23–350 9–3892 1.34–52.6
Kusunda Coal mine, India 37 3–5 90–422 25–400 1.08–28.2
Bansra Coal mine, India 43 2.75–4.5 125–426 13–300 1.7–52.2
Lajkura Coal mine, India 204 3–6.5 110–446 30–270 0.6–243
Samaleshwari Coal mine, India 54 3–7 139–563 37–2778 1.1–373
Uysal et al.17 Seyitomer Lignite Enterprise Open Pit Mine (SLE), 20 4–14 25–150 100 12.1–123
Turkey
Garp Lignite Enterprise open Pit Mine (GLE), 16 3–6 25–100 62.5 13.2–97.9
Turkey
Ak et al.25 Magnesite Incorporated Company Open Pit Mine 43 2.5–4 198–1280 74.5–1208 2–28.5
(MIC), Turkey

Fig. 3. Blast data for West Mudidih Colliery Seam (WMCS) III, India (VPPV vs. scaled
distance).
Fig. 1. Blast data for Chittorgarh Mines site, India (VPPV vs. scaled distance).

Fig. 2. Blast data for West Mudidih Colliery Seam (WMCS) IV, India (VPPV vs. scaled Fig. 4. Blast data for Kusunda coal mine, India (VPPV vs. scaled distance).
distance).

the increase in monitoring distance along with burden (B), VPPV has
3.1. Model 1: VPPV as a function of burden (B), maximum charge per
been found to reduce exponentially.
delay (W) and monitoring distance (R)
The generalized relationship between VPPV (Vppv), distance of blast
area from monitoring point (R) and maximum charge per delay (W) as
While considering the burden as one of the parameter influencing
deduced from the literature was suggested by Rai and Singh9 as given
VPPV, the study revealed that when burden alone was correlated with
below:
VPPV, no relationship could be established between these. However,
when monitoring distance (R) was normalized with burden (B), it has Vppv = K . R−b . W c . e−αR (2)
been observed that it showed an inverse relationship with VPPV. With
where, K, b and c were site specific coefficients and α was defined as

269
S. Murmu et al. International Journal of Rock Mechanics and Mining Sciences 103 (2018) 267–274

this site was 237, out of which 150 blast data has been used for de-
velopment of the model. The parameters k, b, and γ worked out to be
953.37, 1.45 and 0.001 respectively. The regression coefficient has
been obtained as
150 2
∑i = 1 (Vˆi . PPV − Vi, PPV )
Rc2 = 1 − 150 2
∑ (Vˆi . PPV − VPPV )
i=1 (4)
th
where V̂i, PPV is i observed value of vector peak particle velocity
(VPPV), Vi, PPV , the corresponding model computed (as per Eq. (3)) value
of VPPV and Vi, PPV , the mean value of VPPV.
With the above mentioned values of parameters k, b, and γ, the
value of Rc2 worked out to be 0.81. Further, the model has been vali-
dated employing remaining eighty-seven blast data with the help of
another coefficient defined as
237 2
Fig. 5. Blast data for Bansra coal mine, India (VPPV vs. scaled distance).
∑i = 151 (Vˆi . PPV − Vi, PPV )
Rv2 = 1 − 237 2
∑ (Vˆi . PPV − VPPV )
i = 151 (5)

The developed model stands validated provided ≤ Rv2

Fig. 6. Blast data for Lajkura coal mine, India (VPPV vs. scaled distance).
i.e., theRc2 ,
model is said to perform well even on the data which has not been used
for its development purpose. The parameter, Rv2 worked out to be 0.71
and therefore, it builds up the validation of developed model given by
Eq. (3) with parameters k, b, and γ as 953.37, 1.45 and 0.001 respec-
tively.
After validating the proposed model, the site specific parameters, k,
b, and γ have been obtained for other sites under consideration. This
has been presented in Table 4 along with respective values of coeffi-
cient R2.
In order to check for the credibility of the proposed empirical
model, the performance has been compared with other existing models
using the data sets aforementioned. Since the empirical models pro-
posed do not contain specific constant values, rather they are general-
ized equations. The coefficient of determination (R2) , variance ac-
counted for (VAF ) , standard error (SE ) , root mean square error (RMSE )
have been determined for the performance check of the proposed
model. The performance of the proposed model has been presented as
compared to other models and is presented in Table 5. A close look at
the table suggests that proposed model provides better prediction of
vector peak particle velocity (VPPV) compared to other models except
in the case of Garp Lignite Enterprise (GLE) Open Pit Mine, Turkey. In
this case, Roy7 has been found to predict VPPV better. The proposed
model takes into account the burden in prediction of VPPV which other
models do not.

3.2. Model 2: VPPV as a function of maximum charge per delay (W) and
monitoring distance (R)

As mentioned above, regression analysis of Chittorgarh mines data

has also been carried out to obtain the expression for vector peak
particle velocity (VPPV) without taking burden into account and the
following relation has been obtained:

Fig. 7. Blast data for Samaleshwari coal mine, India (VPPV vs. scaled distance).
Table 4
Values of site specific parameters, k, b, and γ (Model 1).
energy factor. Complying with the observation that VPPV exponentially
Project Site k b γ R2
reduces with monitoring distance normalized by burden, a factor R
B
has been introduced resulting in a modification of Eq. (2), as follows: Chittorgarh Limestone Mines, India21 953.37 1.45 0.001 0.80
WMCS IV, India22 4181.66 2.18 −0.015 0.81
−b
R γ R WMCS III, India22 251.52 1.08 −0.005 0.80
VPPV = k ⎛ 2/5 ⎞ e B
⎝W ⎠ (3) Kusunda Coal mine, India22 5916.45 1.94 −0.004 0.82
Bansra Coal mine, India22 10,976.78 1.97 −0.003 0.79
where k, b, and γ are site specific model parameters obtained by mul- Lajkura Coal mine, India22 3145.912 1.104 −0.007 0.80
tiple regression analysis of the observed data. Samaleshwari Coal mine, India22 6656.1 1.58 −0.005 0.91
SLE, Turkey23 915.985 1.17 0.002 0.73
In view of the large data size pertaining to Chittorgarh Mines, India,
GLE, Turkey23 1160.957 1.95 0.03 0.83
these data has been employed for the development of an empirical MIC, Turkey23 3327.578 1.86 0.0009 0.68
model using the multiple regression analysis. Total observed data for

270
S. Murmu et al. International Journal of Rock Mechanics and Mining Sciences 103 (2018) 267–274

Table 5 Table 6
Performance of proposed model as compared to few existing models. Values of site specific parameters, k and b (Model 2).

Project Site Models R2 VAF SE RMSE Project Site k b R2

Chittorgarh Limestone Proposed model 1 0.80 82.14 0.34 3.65 Chittorgarh Limestone Mines, India21 566.8 1.26 0.80
Mines, India Proposed model 2 0.80 75.55 0.34 4.27 WMCS IV, India22 5209.7 2.619 0.80
Duvall and Petkof2 0.79 70.33 0.35 4.71 WMCS III, India22 295.77 1.27 0.78
Langefors and 0.43 17.89 0.58 7.86 Kusunda Coal mine, India22 11,890 2.284 0.81
Kihlstrom3 Bansra Coal mine, India22 12,971 2.11 0.78
IS: 69225 0.43 17.89 0.58 7.86 Lajkura Coal mine, India22 21,444 1.904 0.77
Roy7 0.75 74.91 4.33 4.30 Samaleshwari Coal mine, India22 11,721 2.004 0.89
WMCS IV, India Proposed model 1 0.81 77.70 0.60 45.45 SLE, Turkey23 614 1.12 0.73
Proposed model 2 0.80 72.77 0.60 50.28 GLE, Turkey23 478.186 1.19 0.78
Duvall and Petkof2 0.75 60.23 0.66 60.84 MIC, Turkey23 1495.177 1.58 0.66
Langefors and 0.31 1.98 1.10 95.80
Kihlstrom3
IS: 69225 0.31 1.78 1.10 95.55 In a similar manner, the validation of this proposed model has been
Roy7 0.64 63.89 57.96 55.13
carried out and the condition Rv2 ≤ Rc2 has been found to be valid in this
WMCS III, India Proposed model 1 0.80 75.48 0.42 6.20
Proposed model 2 0.78 68.93 0.43 6.85 case as well thus validating the proposed model. Analysis of data re-
Duvall and Petkof2 0.72 56.79 0.48 8.05 levant to other mining sites has also been carried out to obtain the site
Langefors and 0.42 32.43 0.70 10.30 specific parameters, k and b and the same has been presented in Table 6
Kihlstrom3 along with respective R2 values.
IS: 69225 0.42 32.43 0.70 10.30
Roy7 0.58 58.12 8.02 7.84
Performance of this model has also been checked for other mine
Kusunda Coal mine, India Proposed model 1 0.82 69.26 0.36 4.0 sites and have been depicted in Table 5 with respect to other mine sites.
Proposed model 2 0.81 66.68 0.37 4.17 Table 5 shows that this model also predicts VPPV better than other
Duvall and Petkof2 0.76 59.48 0.43 4.61 models existing in the literature but for some of the sites, Model 1
Langefors and 0.38 34.64 0.66 6.02
predicts the VPPV in better manner. Although the difference in the
Kihlstrom3
IS: 69225 0.38 34.64 0.66 6.02 quantities employed for performance check, i.e., coefficient of de-
Roy7 0.62 62.11 4.53 4.41 termination (R2) , variance accounted for (VAF ) , standard error (SE ) ,
Bansra Coal mine, India Proposed model 1 0.79 72.01 0.42 6.94 root mean square error (RMSE ) have not been found to be very sig-
Proposed model 2 0.78 70.82 0.43 7.07 nificant.
Duvall and Petkof2 0.74 67.95 0.47 7.45
In view of the above analysis, it can be remarked that although
Langefors and 0.46 45.96 0.67 9.97
Kihlstrom3 burden has some influence on VPPV but it is not significant.
IS: 69225 0.46 46.10 0.67 9.94
Roy7 0.68 67.75 7.53 7.35
Lajkura Coal mine, India Proposed model 1 0.80 70.26 0.59 29.21
4. Probabilistic analysis
Proposed model 2 0.77 70.94 0.63 28.39
Duvall and Petkof2 0.74 68.79 0.68 28.57
Langefors and 0.40 48.07 1.02 40.29 Rock blasting is a complex phenomenon resulting in various un-
Kihlstrom3 desirable after effects, ground vibration being one of them. These ef-
IS: 69225 0.40 48.02 1.02 40.33 fects are inevitable and cannot be entirely circumvented owing to the
Roy7 0.67 67.40 29.72 29.58
uncertainty associated with the blast design and explosive parameters.
Samaleshwari Coal mine, Proposed model 1 0.91 78.60 0.51 44.34
India Proposed model 2 0.89 83.08 0.55 39.68 In order to cater this problem in a simple manner, empirical approaches
Duvall and Petkof2 0.86 78.56 0.62 44.91 have been developed to predict and prevent peak particle velocity
Langefors and 0.66 50.56 0.97 68.55 (PPV). These deterministic models help the practicing engineers to
Kihlstrom3 design the blast pattern so that PPV can be minimized. The benefit of
IS: 69225 0.66 50.87 0.97 68.45
Roy7 0.83 82.77 39.19 38.45
these deterministic models is that they directly predict PPV via single
SLE, Turkey Proposed model 1 0.73 56.34 0.42 23.42 and simple equation. The major disadvantage of this approach is that
Proposed model 2 0.73 57.18 0.41 23.19 they are site dependent and have been developed by statistical fits
Duvall and Petkof2 0.73 57.35 0.41 23.17 pertaining to the site measured range of data. Many assumptions have
Langefors and 0.73 57.22 0.41 23.17
to be made derived by intuition, past experience and sound engineering
Kihlstrom3
IS: 69225 0.73 56.75 0.41 23.26 judgement. Therefore it is deemed necessary to carry out the prob-
Roy7 0.63 63.16 22.56 21.41 abilistic analysis which will lead to better prediction of hazards, so that
GLE, Turkey Proposed model 1 0.83 91.58 0.32 9.14 effective mitigation measures can be implemented. In order to cater to
Proposed model 2 0.78 87.45 0.35 11.20 these inadequacies, effort has been made in this paper to predict vector
Duvall and Petkof2 0.78 87.43 0.35 11.22
Langefors and 0.78 87.40 0.35 11.27
peak particle velocity (VPPV) with a probabilistic analysis approach.
Kihlstrom3 There are various probabilistic methods to account for the variability in
IS: 69225 0.78 87.50 0.35 11.20 parameters employed for determination of VPPV. Blair17 applied Monte
Roy7 0.90 89.91 10.48 9.81 Carlo waveform superposition models which includes probabilistic as-
MIC, Turkey Proposed model 1 0.68 66.26 0.37 2.98
sessment of VPPV for any given blast design. In this model, Monte Carlo
Proposed model 2 0.66 63.32 0.37 3.12
Duvall and Petkof2 0.63 60.68 0.39 3.25 variations are imposed on parameters such as delay sequence times and
Langefors and 0.02 2.27 0.64 5.17 the seed waveforms. Another approach, as employed by Blair26, in
Kihlstrom3 which the observed scatter in measured VPPV values is used in con-
IS: 69225 0.02 2.28 0.64 5.17 junction with simple charge weight scaling to give a probabilistic as-
Roy7 0.63 62.99 3.17 3.09
sessment of peak vibration levels. This approach does not involve either
waveform superposition or Monte Carlo simulations. It is completely
R −b based on charge weight scaling laws and observed scatter in VPPV. Next
VPPV = k ⎛ 2/5 ⎞ approach is conventional Monte Carlo method which essentially re-
⎝W ⎠ (6)
quires a deterministic model for the prediction of state variable. The

271
S. Murmu et al. International Journal of Rock Mechanics and Mining Sciences 103 (2018) 267–274

variability in input parameters is considered and accordingly, realiza-

tions of these are generated. These realizations are then fed to chosen
deterministic model and realizations of state variable is generated via
Monte Carlo simulations. Variability in state variable can then be stu-
died by means of its structure in terms of type of distribution, mean,
variance etc. There are many other approaches for probabilistic analysis
which can be chosen appropriately. To analyze the uncertainty arising
from variability in the blast design and explosive parameters for pre-
diction of VPPV, Monte-Carlo (MC) simulation has been adopted.

4.1. Monte-Carlo simulations

Monte-Carlo simulation has been proved to be an important sam-

pling technique in probabilistic analysis allowing the variability or
uncertainty to be accounted for. Basically it uses the deterministic
model and associates to it random variables of the input variables in the
form of frequency histogram or probability density function within
their range of estimate. The result is recorded and this process is re-
Fig. 8. Frequency histogram of monitoring distance (R): Chittorgarh Mines site, India.
peated number of times, each time associating different randomly se-
lected values. When the simulation is complete, a large number of re-
sults are obtained from the model, based on random input values. These
results depict the likelihood or probability of the occurrence of an event
in the model.
The following steps have been adopted to conduct MCS for a specific
site: (i) Data of explosive parameters and burden has been compiled for
the site. (ii) Deterministic model has been developed as presented be-
fore and represented by Eq. (3) with parameters k, b, and γ determined
corresponding to the site. (iii) Distribution functions which resemble
the probability and range of values associated with the explosive and
blast design parameters that could be expected in blast have been de-
fined. (iv) The number of simulations have been decided. (v) The rea-
lizations for input blast parameters have been generated following the
predefined probability distribution functions. These realizations have
been fed to the deterministic model for decided number of simulations.
(vi) Output as peak particle velocity (PPV) in the form of its probability
density and distribution functions has been generated pertaining to the
number of simulations.

4.2. Input parameters Fig. 9. Frequency histogram of maximum charge per delay (W): Chittorgarh Mines site,
India.
Monte-Carlo simulations (MCS) have been conducted for
Chittorgarh Limestone Mines site typically employing Model 1 (Eq. (3))
as the deterministic model. However, the approach is general enough
and can be conducted for any other site once the appropriate de-
terministic model for the site is available. The deterministic model has
been taken as per Eq. (3) with parameters k, b, and γ as 953.37, 1.45
and 0.001 respectively. To decide the type of probability distribution
function of input parameters, first the frequency histogram of input
variables, i.e., monitoring distance, maximum charge per delay and
burden has been obtained from the experimental data. This is depicted
in Figs. 8–10 respectively. Owing to the non-negative values and fre-
quency histogram of monitoring distance (R) and maximum charge per
delay (W), it has been assumed that these follow lognormal distribu-
tion. However, burden has been assumed as a discreet variable.
The goodness of fit determines how well the distribution fits the
sample data. In order to determine the goodness of fit for the selected
probability distribution functions corresponding to each input vari-
ables, Chi-square (χ2) tests have been conducted. The computed value
of χ2 has been found to be 20.39 and 17.61 for monitoring distance (R)
Fig. 10. Frequency histogram of burden (B): Chittorgarh Mines site, India.
and maximum charge per delay (W) respectively as against the corre-
sponding tabulated values of 21.67 and 18.31 for a significance level of
0.05. This depicted that choice of the probability distribution function
is appropriate. In view of this, monitoring distance has been assumed to
follow lognormal distribution with mean = 392.28 m and standard
deviation = 200.48; maximum charge per delay as lognormal dis-
tribution with mean = 80.05 kg and standard deviation = 55.35 and

272
S. Murmu et al. International Journal of Rock Mechanics and Mining Sciences 103 (2018) 267–274

burden as discreet variable as ({3,3.5,4,5.5,6},

{0.33,0.27,0.33,0.05,0.02}).

4.3. Probabilistic Analysis of Chittorgarh Limestone Mines, India

To generate realizations for the parameters, R and W, first of all

random numbers (uniformly distributed random variables between 0
and 1) have been generated. Corresponding mean and standard de-
viation for normal distribution have been obtained and then by em-
ploying Central Limit Theorem, the normally distributed random vari-
ables have been generated. Finally, corresponding realizations for
lognormally distributed random variables R and W have been generated
by taking the exponential of realizations of normally distributed
random variables.
The analysis has been done for different number of simulations like
100, 200, 500, 1000 and 1250 and probability distribution functions
(PDF) for VPPV corresponding to every number of simulations have Fig. 12. Probability distribution functions: measured, predicted and simulated peak
been studied. It has been observed that after 1000 simulations, there particle velocity.
was no difference in the PDF of VPPV. In view of this, it was decided to
have 1000 simulations for the probabilistic analysis employing Mote-
Safety), India, the peak particle velocity (PPV) at this particular range
Carlo simulations. This depicts that probability of occurrence of any
of frequency should not exceed 15 mm/s for domestic houses, 25 mm/s
value of PPV can be obtained with an accuracy of 0.001 or 0.1%.
for industrial buildings and 10 mm/s for structures of historical im-
As explained above, the probabilistic analysis for Chittorgarh
portance27. Since Chittorgarh is known for its famous Chittorgarh Fort
Limestone Mines, India has been carried out employing 1000 Monte-
which is honored as a world heritage site, it has threat due to the
Carlo simulations. This resulted into 1000 realizations of state variable
mining practices. From Fig. 12, probability of occurrence of peak par-
which is peak particle velocity (PPV). The frequency histogram of PPV
ticle velocity assuming a value greater than 10 mm/s has been obtained
has been depicted in Fig. 11 which gives an idea about the type of
as 6.1%.
distribution function that PPV can follow. Accordingly, it was assumed
One can obtain the probability of occurrence of any value of PPV
that PPV follows the lognormal distribution with mean as 4.10 mm/s
from results of probabilistic analysis which can be very useful for as-
and standard deviation as 5.10. To check the goodness of fit, Chi-square
suming safe charge per delay and burden, so that the adverse effect of
(χ2) test has been conducted and for a significance level of 0.05, the
PPV can be minimized upon the domestic, industrial or historical
tabulated value of χ2 (27.60) has been found to be greater than its
structures in the vicinity of mining area. Also the blast design can be
computed value (27.10). This check assures about the appropriateness
implemented accordingly to minimize the phenomenon of blast induced
of the chosen distribution for peak particle velocity as lognormal dis-
ground vibrations.
tribution.
Probability distribution function of PPV has been obtained from
5. Concluding remarks
experimental or measured data and also from the simulated data.
Further, the same can be obtained from developed empirical model
There has always been hue and cry between the mines management
represented by Eq. (3) with parameters k, b, and γ as 953.37, 1.45 and
and the locality owing to the threat mining activity holds because of
0.001 respectively. All these have been plotted together for comparison
ground vibrations. Researchers have been trying to predict PPV so as to
purpose and are depicted in Fig. 12. It can be observed that the simu-
mitigate this adverse effect. There are umpteen numbers of prediction
lated PPV is in close proximity to the measured PPV adding more sce-
models but the inclusion of blast design parameters has been lacking
narios predicting an extensive range of PPV. This proves that prob-
and somewhat debatable. Furthermore, uncertainty is associated with
abilistic analysis is a good tool to simulate PPV. The dominant
the blast design and explosive parameters. Therefore, an attempt has
frequency pertaining to blasting at Chittorgarh was mostly > 25 Hz23.
been made to cater to these uncertainties in the present work. The aim
In Indian context as per DGMS (Directorate General of Mines and
of this study was to provide a empirical prediction model using blast
design parameters. Further to carry out a probabilistic analysis of
ground vibrations and to understand the effect of various parameters on
VPPV. Salient features of the present work are summarized below.
A total of 637 published blast data from ten project sites having
different physic-mechanical properties associated with the rock mass
were collected. Two empirical equations to predict VPPV using ex-
plosive parameters like monitoring distance and maximum charge per
delay were developed with and without consideration of blast design
parameter i.e. burden.
An empirical model incorporating the influence of burden on VPPV
for blast due to surface mining was developed, as follows:

R −b γ R
VPPV = k ⎛ 2/5 ⎞ e B
⎝W ⎠ (7)

Whereas without the consideration of burden, the empirical model is

R −b
VPPV = k ⎛ 2/5 ⎞ .
⎝W ⎠ (8)
Fig. 11. Frequency histogram of peak particle velocity.
It was observed that the proposed models, in general, provides

273
S. Murmu et al.
better prediction of VPPV. The influence of burden was not found to be
significant. This proved that the proposed empirical models are in ac-
cordance with the literature and comply by the rules of the physical
phenomenon pertaining to blast induced ground vibration. Owing to
the fact VPPV attenuates exponentially with increase in monitoring
distance and burden. The proposed empirical models were found to be
useful for the prediction of VPPV for any rock types with accuracy and
credibility. However, this aspect needs the conduct of some more study,
keeping rock mass quality in consideration.
Probabilistic analysis was conducted considering monitoring dis-
tance and maximum charge per delay as lognormally distributed
random variables and burden as discreet variable. The peak particle
velocity exhibited lognormal distribution. A comparison between
measured, predicted and simulated results displayed that Monte-Carlo
simulation can adequately predict PPV. The probability of occurrence
of threat from blasting operation implementing a particular blast design
could be determined from results obtained from the probabilistic ana-
lysis.
References
1. Parida A, Mishra MK. Blast vibration analysis by different predictor approaches – a
comparison. Procedia Earth Planet Sci. 2015;11:337–345.
2. Duvall WI, Petkof B. Spherical propagation of explosion generated strain pulses in rock.
US Bureau of Mines; 1959:21 [Report of Investigation 5483].
3. Langefors U, Kihlstrom B. The modern techniques of rock blasting. New York: John
Wiley and Sons Inc; 1963.
4. Ambraseys NR, Hendron AJ. Dynamic behavior of rock masses: rock mechanics in
engineering practices. In: Stagg KG, Zienkiewiez AC, eds. Rock mechanics in en-
gineering practices. London: John Wiley and Sons Inc; 1968:203–207.
5. IS: 6922. Criteria for safety and design of structures subjected to underground blast. N.
Delhi, India: Bureau of Indian Standards; 1973.
6. Ghosh A, Daemen JJK. A simple new blast vibration predictor (based on wave pro-
pagation laws). In Proceedings of the 24th US Symposium on Rock Mechanics. College
Station, Texas: 151–161; 1983.
7. Roy P Pal. Vibration control in an opencast mine based on improved blast vibration
predictors. Min Sci Technol. 1991;12(2):157–165.
8. Bilgin A, Esen S, Kilic M. TKI Bell Lignite Plant research for solving environmental pro-
blems caused by blasting. Ankara: Turkish Coal Enterprise; 1998 [Final Report].
274
International Journal of Rock Mechanics and Mining Sciences 103 (2018) 267–274
9. Rai R, Singh TN. A new predictor for ground vibration prediction and its comparison
with other predictors. Indian J Eng Mater Sci. 2004;11:178–184.
10. Ak H, Kounuk A. The effect of discontinuity frequency on ground vibrations pro-
duced from bench blasting: a case study. Soil Dyn Earthq Eng. 2008;28:686–694.
11. Simangunsong GM, Wahyudi S. Effect of bedding plane on prediction blast-induced
ground vibration in open pit coal mines. Int J Rock Mech Min Sci. 2015;79:1–8.
12. Kumar R, Choudhury D, Bhargava K. Determination of blast-induced ground vibra-
tion equations for rocks using mechanical and geological properties. J Rock Mech
Geotech Eng. 2016;8:341–349.
13. Siskind DE. Vibrations from blasting. Cleveland, OH: International Society of
Explosives Engineers; 2000.
14. Nicholls HR, Johnson CF, Duvall WI. Blasting vibrations and their effects on struc-
tures. USBM Bull No 656. 1971:105.
15. Bergmann OR, Riggle JW, Wu FC. Model rock blasting-effect of explosives properties
and other variables on blasting results. Int J Rock Mech Min Sci Geomech Abstr.
1973;10:585–612.
16. Blair DP. Blast vibration control in the presence of delay scatter and random fluc-
tuations between blastholes. Int J Numer Anal Methods Geomech. 1993;17:95–118.
17. Blair DP. Statistical models for ground vibration and airblast. Int J Blasting Fragm.
1999;3:335–364.
18. Blair DP, Armstrong LW. The influence of burden on blast vibration. Fragblast.
2001;5(1–2):108–129.
19. Uysal Ö, Arpaz E, Berber M. Studies on the effect of burden width on blast-induced
vibration in open pit mines. Environ Geol. 2007;53:643–650.
20. Elevli B, Arpaz E. Evaluation of parameters affected on the blast induced ground
vibration (BIGV) by using relation diagram method (RDM). Acta Montan Slov.
2010;15(4):261–268.
21. Blair DP. Blast vibration dependence on charge length, velocity of detonation and
layered media. Int J Rock Mech Min Sci. 2014;65:29–39.
22. Görgülü K, Arpaz E, Demirci A, Koçaslan A, Dilmaç MK, Yüksek AG. Investigation of
blast-induced ground vibrations in the Tülü boron open pit mine. Bull Eng Geol
Environ. 2013;72(3–4):555–564.
23. Verma HK, Pandey Y, Roy PPal. Mining near world heritage site – a case study of
Limestone mining near Chittorgarh Fort complex. In Proceedings 6th Asian Mining
Congress organized by The Mining Geological and Metallurgical Institute of India (MGMI),
Kolkata, India, Feb. 2016. 23-26: 2016: 265-274.
24. Singh PK, Roy SK, Sinha A. A new blast damage index for the safety of coal mine
openings. J Min Technol. 2003;112(2):97–104.
25. Ak H, Iphar M, Yavuz M, Konuk A. Evaluation of ground vibration effect of blasting
operations in a magnesite mine. Soil Dyn Earthq Eng. 2009;29:669–676.
26. Blair DP A probabilistic analysis of vibration based on measured data and charge
weight scaling. In EFEE Proceedings of the Sixth World Conference on Explosives and
Blasting Techniques, Lisbon, Portugal: 319-337.
27. 〈https://elibrarywcl.files.wordpress.com/2015/02/dgms-cir_1997-all-dgms-
circulars-1997.pdf〉. July 01, 2017.