Journal of http://jvc.sagepub.

com/
Vibration and Control

Development of a fuzzy model for predicting ground vibration caused by rock blasting in surface
mining
Ebrahim Ghasemi, Mohammad Ataei and Hamid Hashemolhosseini
Journal of Vibration and Control 2013 19: 755 originally published online 21 February 2012
DOI: 10.1177/1077546312437002

The online version of this article can be found at:

http://jvc.sagepub.com/content/19/5/755

Published by:

http://www.sagepublications.com

Additional services and information for Journal of Vibration and Control can be found at:

Email Alerts: http://jvc.sagepub.com/cgi/alerts

Subscriptions: http://jvc.sagepub.com/subscriptions

Reprints: http://www.sagepub.com/journalsReprints.nav

Permissions: http://www.sagepub.com/journalsPermissions.nav

Citations: http://jvc.sagepub.com/content/19/5/755.refs.html

>> Version of Record - Apr 4, 2013

OnlineFirst Version of Record - Feb 21, 2012

What is This?

Downloaded from jvc.sagepub.com at UNIV OF UTAH SALT LAKE CITY on November 28, 2014
Article
Journal of Vibration and Control
19(5) 755–770

Development of a fuzzy model for ! The Author(s) 2012

Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
predicting ground vibration caused by DOI: 10.1177/1077546312437002
jvc.sagepub.com
rock blasting in surface mining

Ebrahim Ghasemi1, Mohammad Ataei1

and Hamid Hashemolhosseini2

Abstract
Ground vibration is an integral part of the rock blasting process in surface mines, which may cause severe damages to
structures and plants in the nearby environment. Therefore, its prediction plays an important role in the minimization of
environmental impacts. The peak particle velocity (PPV) is an important predictor for ground vibration. In this paper, first
a fuzzy logic model was developed to predict PPV based on collected data from blasting events in Sarcheshmeh copper
mine, located in the southwest of Iran. The predictive fuzzy model was implemented on the fuzzy logic toolbox of
MATLAB using the Mamdani algorithm. Then, the PPV was predicted by conventional empirical predictors used in
blasting practice and also by multiple regression analysis. Finally, a comparative analysis between the results obtained
by the fuzzy model and common vibration predictors was carried out. The results indicated the high predictive capacity
of fuzzy model, which can be used as a reliable predictor of ground vibration for the studied mine.

Keywords
Blasting operation, ground vibration, peak particle velocity, fuzzy logic, empirical predictors, multiple regression
Received: 30 September 2011; accepted: 22 December 2011

1. Introduction
assess and control ground vibration due to blasting
In most surface mines, the blasting operation is the first operations (Blair and Jiang, 1995; Bhandari, 1997;
element of the ore extraction process. The primary Siskind, 2000; Kahriman, 2002; Valdivia et al., 2003;
purpose of blasting is rock fragmentation and displace- Kahriman, 2004; Kuzu and Ergin, 2005; Singh et al.,
ment of the broken rocks. For this purpose, the blasting 2006; Uysal et al., 2007; Kuzu, 2008; Ozer et al., 2008;
operation requires a large amount of explosives. Singh Ak et al., 2009; Mesec et al., 2010; Singh and Verma,
and Singh (2005) indicated that fragmentation and 2010). The intensity of ground vibration arising from
displacement of broken rocks use only 20–30% of the rock blasting depends on various parameters. These
total amount of explosive energy. The remainder of the parameters can be broadly divided into two categories,
energy is wasted away in the form of environmental namely, controllable parameters and uncontrollable
side effects, such as ground vibration. The ground parameters. Controllable parameters can be changed
vibration is usually described as a time-varying
displacement, velocity or acceleration of a particular
1
point (particle) in the ground. Department of Mining, Petroleum and Geophysics Engineering,
Ground vibration caused by blasting operations is Shahrood University of Technology, Iran
2
Department of Mining Engineering, Isfahan University of Technology, Iran
acoustic waves that propagate through the rocks. It
differs from the ground vibrations caused by earth- Corresponding author:
Ebrahim Ghasemi, Department of Mining, Petroleum and Geophysics
quakes in terms of seismic source, amount of available Engineering, Shahrood University of Technology, Daneshgah Boulevard,
energy and travelled distances (Giraudi et al., 2009). P.O. Box 3619995161, Shahrood, Iran
Over years, many studies have been conducted to Email: [email protected]; [email protected]

Downloaded from jvc.sagepub.com at UNIV OF UTAH SALT LAKE CITY on November 28, 2014
756 Journal of Vibration and Control 19(5)

and include blast geometry, type and amount of artificial intelligence is a useful tool for better
used explosive, stemming, priming and initiation, prediction of PPV.
while uncontrollable parameters are natural and In this paper, an effort has been made to predict PPV
cannot be changed and include distance, geological with the help of the fuzzy logic approach using data
conditions, initiation timing errors and meteorological collected from blasting events in Sarcheshmeh copper
conditions. mine, Iran. The same data have been also used for the
Ground vibration traveling through the ground may prediction of PPV using multiple regression analysis
damage the surrounding environment and nearby and empirical vibration predictors. The basic idea is
structures when it reaches a certain magnitude. High to find the scope and suitability of the fuzzy logic for
intensity ground vibration not only creates problems prediction of PPV over the widely used conventional
for the nearby population, but also adversely affects vibration predictors. It should be noted that Fisne
the integrity of the structures in the mine area. et al. (2010) have recently developed a fuzzy model
Sometimes, it provokes the population and can put for predicting PPV based on two parameters, charge
mines into closure. High intensity vibration also dam- weight per delay and distance from blast location, for
ages the ground water and harms the ecology of the quarry operations in Istanbul, Turkey. The main differ-
nearby area. Therefore, the effects of ground vibration ence between their study and the present study is that in
on building structures and human beings need to be addition to the two mentioned parameters, this study
predicted, monitored and controlled by the blasting considers other effective parameters on ground vibra-
engineers as part of optimizing the blasting operation. tion, such as burden, spacing, stemming and number of
In order to predict and evaluate the blast vibration holes per delay. In addition, in this study the perfor-
effects and consequences, different indicators have mance of the fuzzy model is compared with common
been proposed, such as peak particle velocity (PPV), empirical predictor models, whereas Fisne et al. (2010)
peak particle acceleration (PPA) and peak particle dis- have compared their obtained results with only the
placement (PPD). Among these indicators, the PPV has United States Bureau of Mines (USBM) empirical
been used frequently in different standards and it has model.
been found to be a reliable indicator for evaluation and This paper is organized into eight sections. In
prediction of losses associated with ground vibration. A Section 2, details of the case study and collected data
number of investigators have studied ground vibration are given, while in Section 3, the background of fuzzy
resulting from blasting and have developed several logic is described. Details of fuzzy model development
empirical equations for predicting the PPV using statis- are given in Section 4 and in Sections 5 and 6, we
tical techniques (Duvall and Petkof, 1959; Langefors describe the development of the multiple regression
and Kihlstrom, 1963; Davies et al., 1964; Ambraseys model and empirical predictor models, respectively,
and Hendron, 1968; Indian Standard Institute, 1973; for prediction of PPV. In Section 7 the predictive
Ghosh and Daemen, 1983; Gupta et al., 1987; Pal capacity of all mentioned models is compared and in
Roy, 1991; Rai and Singh, 2004). These empirical Section 8, results of this paper are given.
PPV predictor models are basically based on two
parameters, maximum charge per delay and distance
from blast site, and do not include other effective 2. Case study: Sarcheshmeh copper
parameters. Because of the large number of influencing mine
parameters and complex interrelation among these
parameters, empirical methods may not be fully suit-
2.1. Mine description
able for such problems. In order to overcome these Sarcheshmeh copper mine is the largest porphyry
shortcomings, artificial intelligence techniques are now copper mine in Iran, which is situated 160 km south-
being used as alternate statistical techniques. In recent west of Kerman and 50 km south of Rafsanjan city, at
years, some researchers have tried to develop new pre- 31.2 N longitude and 56.1 E latitude (Bakhshandeh
dictive models using artificial intelligence techniques, Amnieh et al., 2010). Figure 1 shows the location of
especially an artificial neural network (ANN) incorpo- Sarcheshmeh copper mine, which is 2600 m above sea
rating two or more parameters affecting the ground level. The mine is located near the center of an elon-
vibration (Singh and Singh, 2005; Khandelwal and gated NNW–SSE mountain belt, which is principally
Singh, 2006, 2007; Iphar et al., 2008; Mohamed, 2009; composed of folded and faulted rocks, and extends
Khandelwal et al., 2009; Khandelwal and Singh, 2009; intermittently from Turkey to the southeast
Bakhshandeh Amnieh et al., 2010; Fisne et al., 2010; Baluchestan of Iran.
Monjezi et al., 2010; Verma and Singh, 2010; Dehghani The geology of Sarcheshmeh porphyry deposit is
and Ataee-pour, 2011; Kamali and Ataei, 2011; very complicated and various rock types can be
Mohamed, 2011). The researchers found that the found. The oldest host rock in this mine is Eocene

Downloaded from jvc.sagepub.com at UNIV OF UTAH SALT LAKE CITY on November 28, 2014
Ghasemi et al. 757

Figure 1. Sarcheshmeh copper mine.

Table 1. Basic descriptive statistics of the data collected from Sarcheshmeh copper mine

Parameters Unit Symbol Min. Max. Mean. Std. dev.

Burden m B 3.00 7.50 7.01 1.13

Spacing m S 4.00 11.00 9.29 1.75
Stemming m St 2.40 6.00 5.51 0.95
No. of holes per delay – N 6.00 32.00 14.08 5.33
Charge per delay kg Q 1332.00 9812.00 5595.21 2079.04
Distance from blast location m D 133.00 2845.00 1037.47 591.04
PPV mm/s PPV 0.49 53.55 9.31 9.00

andesite and other rock is Sarcheshmeh granodiorite vertically in a staggered pattern and drilling cuttings
stock. Waste rocks are mainly granodiorite dykes, are used as stemming materials. The diameter and
including prophyric hornblende, prophyric feldspar depth of blast holes are 215 mm and 15 m, respectively.
and prophyric biotite. The ore zone is in the form After each blast, broken rocks, using hydraulic shovels
of an ellipse with 2300 and 1200 m diameters. and trucks, are sent to the waste dump, oxide dump or
Mineralization in this deposit belongs to the late primary crushers, depending on rock type. With the
Tertiary era. The main minerals of the deposit are present design, mine production is about 40,000 ton
chalcopyrite, pyrite, chalcocite, cuprite and mala- per day.
chite. The proved reserve of the deposit is approxi-
mately 826 Mt, with an average grade of 0.78% of
copper, 0.03% of molybdenum, 0.27 ppm of gold,
2.2. Data collection
1.14 ppm of silver, 1.2 ppm of nickel and 0.9 ppm To do this study, the PPV due to ground vibration
of cobalt. by blasting was measured for 120 blast events at vari-
The mine is extracted by open pit mining and the ous blasting patterns and distances from the blast
blasting operation is performed for rock excavation. face in Sarcheshmeh copper mine in a specific period
The height and slope of the working benches are of time. Burden, spacing, stemming, number of holes
12.5 m and 63.4 , respectively. The angle of the per delay, maximum charge per delay, distance from
overall slope ranges from 30 to 35 . The width and blast location to monitoring point and PPV were
slope of the ramp are 30 m and 5 , respectively. recorded in each blast event. Burden, spacing and stem-
Ammonium nitrate and fuel oil (ANFO) is used as ming were measured by a tape meter and the number of
the main explosive material and dynamite cartridges holes per delay by controlling each blasting pattern.
are used as a primer, with bottom hole positioning. The distance from the blasting location to the monitor-
The blasting system is nonelectric and detonating ing point was measured carefully by means of a hand-
cord is applied for initiation. Blast holes are drilled held GPS (global positioning system) and the amount

Downloaded from jvc.sagepub.com at UNIV OF UTAH SALT LAKE CITY on November 28, 2014
758 Journal of Vibration and Control 19(5)

Figure 2. Input and output variables of the fuzzy model.

Figure 3. Fuzzy representation of burden.

Figure 4. Fuzzy representation of spacing.

Downloaded from jvc.sagepub.com at UNIV OF UTAH SALT LAKE CITY on November 28, 2014
Ghasemi et al. 759

Figure 5. Fuzzy representation of stemming.

Figure 6. Fuzzy representation of number of holes per delay.

of maximum charge per delay was recorded for each 3. Fuzzy logic: basic concepts and
blast by controlling the blast hole charge. Furthermore,
the amount of dynamite used for priming was consid-
definitions
ered for determining the maximum charge per delay. The details of fuzzy logic are available in the literature
PPV was measured by using the PDAS-100 digital (Zadeh, 1965; Ross, 1995), but it is explained briefly in
seismograph. It should be mentioned that in all of the the following. The fuzzy logic is a matter of the fuzzy
recorded blasts, the diameter of blast holes, depth of set theory that in particular is used to deal with subjects
blast holes and delay time between blast holes are con- having ambiguities and uncertainties. Fuzzy set theory
stant and equal to 152 mm, 15 m and 50 ms, respec- was first formulized by Zadeh (1965) as a mathematical
tively. The basic descriptive statistics of collected data way to represent linguistic vagueness. A fuzzy set is an
are summarized in Table 1. extension of a crisp set but does not have any sharp

Downloaded from jvc.sagepub.com at UNIV OF UTAH SALT LAKE CITY on November 28, 2014
760 Journal of Vibration and Control 19(5)

Figure 7. Fuzzy representation of maximum charge per delay.

Figure 8. Fuzzy representation of distance from blast location.

and precise boundaries, unlike a crisp set (Aydin, linguistic information. This is called fuzzification and
2004). A fuzzy logic system consists of four parts: is done by membership functions.
(1) the fuzzification process; (2) knowledge base; (3)
fuzzy inference system (FIS); and (4) defuzzification
3.2. Knowledge base
process. In the following, each one of these parts is
described. The knowledge base includes a data base and a rule
base. The data base defines the membership functions
of the fuzzy sets used in the fuzzy rules, whereas the rule
3.1. Fuzzification process base contains a number of fuzzy if–then rules. The
Fuzzy set performs numerical computation by using if–then rules, also known as the fuzzy rules, provide a
linguistic labels. Therefore, in the first part of the system for describing complex (uncertain, vague) sys-
fuzzy logic system, crisp values of input and output tems by relating input and output parameters using lin-
variables should be converted to fuzzy values or guistic variables. Generally, the fuzzy rules are

Downloaded from jvc.sagepub.com at UNIV OF UTAH SALT LAKE CITY on November 28, 2014
Ghasemi et al. 761

Figure 9. Fuzzy representation of peak particle velocity (PPV).

extracted from experts’ judgments, engineering knowl- science and engineering fields, particularly mining, rock
edge and experience. mechanics and engineering geology, by Fisne et al.
(2010), Aydin (2004), Grima and Babuska (1999),
Gokceoglu (2002), Gokceoglu and Zorlu (2004),
3.3. Fuzzy inference system Iphar and Goktan (2006), Tzamos and Sofianos
The FIS, also known as the decision-making unit, per- (2006), Khademi Hamidi et al. (2010), Acaroglu
forms the inference operations on the rules. In fact, (2011) and Ghasemi et al. (2011). In the next section,
fuzzy inference is the process of formulating an input the proposed fuzzy model for predicting PPV is
fuzzy set map to an output fuzzy set using fuzzy logic. described in detail. Of course, it should be mentioned
The core section of a fuzzy logic system is the FIS part, that all predictive models, including fuzzy, multiple
which combines the facts obtained from the fuzzifica- regression and empirical predictors, are developed
tion with the rule base and conducts the fuzzy reason- based on only 75% of the collected data (i.e. 90 data-
ing process. There are several FISs that have been sets) and the rest (i.e. 30 datasets) are used for testing
employed in various applications; the most commonly the model performances.
used include the Mamdani fuzzy model, the Takagi–
Sugeno–Kang (TSK) fuzzy model, the Tsukamoto
fuzzy model and the Singleton fuzzy model. The differ-
4. Development of the fuzzy model
ences between these FISs lie in the consequents of their In this section, a fuzzy model based on the Mamdani
rules, and thus aggregation and defuzzification proce- algorithm is introduced for prediction of PPV in
dures differ accordingly. Sarcheshmeh copper mine. The fuzzy model was imple-
mented on the fuzzy logic toolbox of MATLAB ver. 7.6
(R2008a) software package. The model includes six
3.4. Defuzzification process
input variables and one output variable. Figure 2
The output generated by the FIS is always in the fuzzy shows input and output variables in the MATLAB
(linguistic) form, but most of the time, the need for a environment where burden, spacing, stemming,
crisp and representative value leads to usage of the number of holes per delay, maximum charge per
defuzzifier. The application of the defuzzifier is to delay and distance from blast location are referred to
receive the fuzzy input and provide crisp output. In as input and PPV is referred to as output.
fact, it works in the opposite way to the fuzzifier. In the model, triangular membership functions were
During the past two decades, fuzzy logic has been adopted for describing input and output variables
successfully applied to many real world problems, espe- because of their simplicity and computational
cially in modeling complex and imprecise systems in the efficiency. The triangular membership function, as

Downloaded from jvc.sagepub.com at UNIV OF UTAH SALT LAKE CITY on November 28, 2014
762 Journal of Vibration and Control 19(5)

Table 2. Representation of membership functions and their parameters

Variables Linguistic variables Linguistic values Parameters

Inputs Burden Low [3 3 4.5]

Medium [3.8 5.5 7.2]
High [6.5 7.5 7.5]
Spacing Low [4 4 5.5]
Medium [4.7 7.5 10.3]
High [9.5 11 11]
Stemming Low [2.4 2.4 3.5]
Medium [2.9 4 5.1]
High [4.8 6 6]
No. holes per delay Very low [6 6 10]
Low [9 11 13]
Medium [12 15 17]
High [16 18 21]
Very high [20 32 32]
Max. charge per delay Very low [1332 1332 3100]
Low [2700 4000 5050]
Medium [4800 5750 6700]
High [6300 7250 8500]
Very high [7800 9812 9812]
Distance from blast location Very near [133 133 600]
Near [400 750 1100]
Far [900 1250 1600]
Very far [1400 2845 2845]
Outputs PPV Very very low [0.49 0.49 1.5]
Very low [1.3 3.5 4.5]
Low [4 7 10]
Medium [9 12 20]
High [18 25 31]
Very high [29 34 44]
Very very high [42 53.55 53.55]

described in Equation (1), is used to convert the linguis- Figures 3–9. In addition, Table 2 shows the linguistic
tic values in the range of 0–1: variables, their linguistic values and associated
parameters.
The next stage of the FIS is the construction of the
8 9 if–then rules, which are used to represent the fuzzy rela-
>
> 0 if x  a >
>
>
> >
> tionships between input and output fuzzy variables. In
>
> xa >
>
>
> > this paper, for constructing the rule base of the fuzzy
< if a  x  b >
=
ba model a total of 229 rules were utilized based on
Triangular ðx ; a, b, cÞ ¼ ð1Þ
>
> cx > experts’ experiences and data collected from the case
>
> if b  x  c >
>
>
>
> cb >
> studied mine. Figure 10 shows a fuzzy if–then rule
>
> >
>
: ; editor including 11 rules of the model in MATLAB
0 if c  x
environment. In this figure, VVL stands for very very
low, VL for very low, L for low, M for medium, H for
high, VH for very high, VVH for very very high, VN
where a, b, c are the parameters of the linguistic value for very near, N for near, F for far and VF for very far.
and x is the range of the input parameters. The graph- In the last stage, each result in the form of a fuzzy set
ical representations of the membership functions of is converted into a crisp (real output) value by the
different input and output variables are shown in defuzzification process. In this model, the centroid of

Downloaded from jvc.sagepub.com at UNIV OF UTAH SALT LAKE CITY on November 28, 2014
Ghasemi et al. 763

Figure 10. Fuzzy if–then rule editor for the proposed fuzzy model.

Figure 11. Fuzzy rule viewer for the proposed fuzzy model.

Downloaded from jvc.sagepub.com at UNIV OF UTAH SALT LAKE CITY on November 28, 2014
764 Journal of Vibration and Control 19(5)

Table 3. Summary of the multiple regression model for area (COA) method, which is a common method of
prediction of PPV defuzzification, was employed for the defuzzification
Independent
process (Grima and Babuska, 1999). The crisp value
variables Coefficient Std. error t-value Sig. level adapting the COA defuzzification method was
obtained by
Constant 8.577 6.360 1.349 0.181
B 4.493 2.138 2.102 0.039 R
 A ðzÞ: z: dz
S 2.394 1.078 2.222 0.029 z ¼ R ð2Þ
A ðzÞ: dz
St 0.419 1.608 0.261 0.795
N 0.250 0.198 1.263 0.210
where z* is the crisp value for the z output and A(z) is
Q 0.001 0.001 1.289 0.201
the aggregated output membership function.
D 0.007 0.001 5.609 0.000
The developed fuzzy model can provide an estima-
tion of PPV when proper input data is entered into
model. For example, as can be seen in Figure 11,
when the input parameters are burden ¼ 7.5 m,
spacing ¼ 9.5 m, stemming ¼ 6 m, number of holes
per delay ¼ 15, charge per delay ¼ 5560 kg and dis-
Table 4. Most common empirical predictors for the prediction tance from blast location ¼ 580 m, the predicted
of peak particle velocity (PPV) output PPV is 22.1 mm/s (whereas measured PPV is
Name Equation 21.05 mm/s).

USBM (Duvall and PPV ¼ K:½D=Q1=2 B

Petkof, 1959) 5. Multiple regression model
Langefors–Kihlstrom PPV ¼ K:½Q1=2 =D3=4 B
(1963)
The general purpose of multiple regression is to learn
General predictor B
PPV ¼ K: D : Q A more about the relationship between several indepen-
(Davies et al., 1964) dent or predictor variables and a dependent or criterion
Ambraseys–Hendron (1968) PPV ¼ K:½D=Q1=3 B variable. The mathematical form of the multiple regres-
Indian Standard (1973) PPV ¼ K:½Q=D2=3 B sion is y ¼ 0 þ 1x1 þ 2x2 þ    þ nxn, where the 
Ghosh–Daemen 1 (1983) PPV ¼ K:½D=Q1=2 B : e:D values are regression coefficients. In this paper, multiple
Ghosh–Daemen 2 (1983) PPV ¼ K:½D=Q1=3 B : e:D regression analysis was carried out on 90 blast vibration
Gupta et al. (1987) PPV ¼ K:½D=Q1=2 B : e:ðD=QÞ cases to determine the mathematical equation for PPV
CMRI predictor PPV ¼ n þ K: ½D=Q1=2 1 using statistical software package SPSS 16. The multi-
(Pal Roy, 1991) ple regression model for predicting PPV is given
Rai–Singh (2004) PPV ¼ K:DB :QA :e:D in Equation (3) and the detailed results of the regression
analysis are presented in Table 3:

PPV½mm=s ¼ 8:577 þ 4:493B½m  2:394S½m

 0:419St ½m  0:250N þ 0:001Q½kg ð3Þ
 0:007D½m
Table 5. Obtained values of site constants
Site constants

Empirical predictor K B A n 
6. Empirical predictor models
USBM 71.810 0.922
Langefors–Kihlstrom 19.547 1.257 As mentioned previously, there are several empirical
General predictor 14.240 0.946 0.668 equations that are used in blasting operations to
Ambraseys–Hendron 238.504 0.891 estimate PPV. The most common equations are
Indian Standard 0.039 1.278 shown in Table 4. In these equations, D is the distance
Ghosh–Daemen 1 76.884 1.006 0.000 between the center of the explosive charge and the mea-
Ghosh–Daemen 2 319.697 1.023 0.000 suring unit in meters and Q is charge per delay in kilo-
Gupta et al. 68.082 0.849 0.724 grams. Values of K, A, B, n and  are site constants that
CMRI predictor 75.994 0.904 can be determined by regression analysis. In this paper,
Rai–Singh 11.556 0.904 0.667 0.000 the site constants for Sarcheshmeh copper mine were
determined from the regression analysis of the

Downloaded from jvc.sagepub.com at UNIV OF UTAH SALT LAKE CITY on November 28, 2014
 Table 6. Comparison between measured peak particle velocity (PPV) and predicted PPV values by different predictive models
Ghasemi et al.

Predicted PPV (mm/s)

No. of blast Measured Fuzzy Multiple Langefors– General Ambraseys– Indian Ghosh– Ghosh– Gupta CMRI Rai–
events PPV (mm/s) model regression USBM Kihlstrom predictor Hendron Standard Daemen 1 Daemen 2 et al. predictor Singh

1 8.44 7.00 12.96 9.18 8.95 8.91 9.36 7.58 8.15 7.77 11.32 9.07 9.39
2 13.45 19.80 13.78 8.45 8.80 8.90 8.10 10.15 7.44 6.58 10.37 8.37 9.50
3 21.05 22.10 14.77 10.83 10.96 10.99 10.66 10.54 9.76 9.01 12.87 10.67 11.55
4 53.55 48.80 15.36 15.61 16.04 16.14 15.06 15.41 14.54 13.40 17.55 15.42 16.70
5 2.27 3.01 6.56 3.33 2.69 2.56 4.11 1.21 2.70 3.02 6.70 3.62 2.77
6 0.91 0.71 7.78 6.49 7.09 7.26 5.97 10.65 5.58 4.63 8.22 6.51 7.90
7 5.47 2.37 11.17 9.79 10.80 11.07 8.88 15.57 8.74 7.31 11.58 9.66 11.82
8 0.50 0.76 0.31 2.65 2.53 2.51 2.79 2.52 2.10 1.94 4.75 3.02 2.81
9 27.80 20.60 14.69 12.79 12.82 12.82 12.66 11.46 11.70 10.99 14.84 12.60 13.35
10 16.90 19.80 14.08 8.77 8.98 9.04 8.54 9.52 7.75 6.99 10.74 8.67 9.61
11 33.90 30.80 15.60 21.61 20.76 20.55 22.11 13.91 20.73 20.84 23.54 21.56 20.75
12 1.23 0.75 4.90 2.10 1.93 1.90 2.31 1.68 1.63 1.56 4.52 2.55 2.13
13 9.18 11.70 2.80 7.31 6.66 6.52 7.98 4.36 6.36 6.47 9.72 7.28 6.89
14 2.39 2.88 7.42 3.87 3.68 3.64 4.09 3.28 3.18 3.00 6.01 4.11 3.99
15 4.33 3.02 12.36 8.46 8.89 9.00 8.05 10.62 7.45 6.53 10.35 8.38 9.62
16 11.10 9.24 12.26 7.31 7.37 7.40 7.24 7.58 6.35 5.78 9.30 7.28 7.92
17 34.10 36.10 16.08 28.58 29.87 30.15 26.91 27.47 28.13 26.11 29.90 28.88 30.37
18 16.50 17.20 14.27 13.20 13.59 13.69 12.73 13.63 12.11 11.06 15.17 13.01 14.29
19 8.81 6.97 19.91 8.76 8.16 8.03 9.34 5.69 7.75 7.75 11.09 8.67 8.44
20 7.57 6.97 13.23 8.22 8.67 8.79 7.80 10.57 7.23 6.30 10.10 8.15 9.41
21 12.55 14.70 13.58 9.01 9.52 9.65 8.53 11.50 7.99 6.98 10.90 8.90 10.29

Downloaded from jvc.sagepub.com at UNIV OF UTAH SALT LAKE CITY on November 28, 2014
22 6.37 6.95 10.88 6.91 6.82 6.81 7.00 6.42 5.98 5.56 8.97 6.91 7.29
23 14.9 18.30 10.64 10.15 10.02 9.99 10.21 8.79 9.09 8.59 12.25 10.00 10.51
24 4.05 2.99 8.74 4.96 5.31 5.40 4.66 7.67 4.16 3.49 6.67 5.09 5.94
25 3.22 3.03 7.23 4.09 3.99 3.97 4.21 3.96 3.37 3.10 6.11 4.30 4.36
26 1.82 3.00 8.31 4.40 3.89 3.79 4.97 2.47 3.65 3.75 6.94 4.58 4.09
27 18.45 19.70 10.25 9.18 8.95 8.91 9.36 7.58 8.15 7.77 11.32 9.07 9.39
28 9.15 11.60 5.43 4.86 4.77 4.75 4.97 4.69 4.07 3.76 6.88 5.00 5.17
29 10.93 9.24 12.27 7.32 7.38 7.40 7.25 7.59 6.36 5.79 9.31 7.29 7.93
30 14.55 19.80 13.47 8.38 8.43 8.45 8.32 8.32 7.38 6.78 10.41 8.30 8.98
765
766 Journal of Vibration and Control 19(5)

8
7
6
5
Residual error of PPV (mm/s)

4
3
2
1
0
–1
–2
–3
–4
–5
–6
–7
–8
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Number of blast events

Figure 12. The difference between the values predicted by the fuzzy model and the measured values.

the predicted PPVs by the fuzzy model are closer to the

Table 7. Performance indices for models measured ones, whereas the results of multiple regres-
sion and empirical predictors vary widely. This indi-
Model R2 (%) VAF (%) RMSE MAPE
cates that the prediction of ground vibration using
Fuzzy 94.59 94.59 2.73 23.25 the fuzzy model is more accurate than that using
Multiple regression 34.55 29.57 9.94 107.07 other models. Fuzzy model errors vary between 6.35
USBM 64.36 52.61 8.84 74.42 and þ7.20 mm/s, as can be seen in Figure 12.
Langefors–Kihlstrom 62.64 52.65 8.82 74.52 The correlation coefficient (R2) between the mea-
General predictor 62.12 52.56 8.81 74.81
sured and predicted values is a good indicator to
check the prediction performance of each model.
Ambraseys–Hendron 65.25 51.58 8.96 75.94
Furthermore, in this paper, variance account for
Indian Standard 45.40 40.20 9.81 93.91
(VAF) (Equation (4)), root mean square error
Ghosh–Daemen 1 63.30 51.87 9.30 63.37 (RMSE) (Equation (5)) and mean absolute percentage
Ghosh–Daemen 2 63.53 50.45 9.69 60.61 error (MAPE) (Equation (6)) indices were calculated to
Gupta et al. 65.00 51.98 8.27 117.02 control the prediction performance of models, as
CMRI predictor 63.38 51.74 8.91 79.63 employed by Grima and Babuska (1999), Gokceoglu
Rai–Singh 62.14 52.57 8.63 81.75 (2002) and Yilmaz and Kaynar (2011). If R2 and
VAF are 100 and RMSE and MAPE are 0, then the
model will be excellent:
previously mentioned 90 cases using SPSS 16 software.
The obtained values of site constants for the various
predictor equations are given in Table 5.  
varðAi  Pi Þ
VAF ¼ 1  100 ð4Þ
varðAi Þ
7. Comparison of model performance
In this section the results obtained from different pre- rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
dictive models are compared and performances of the 1 XN
RMSE ¼ ðAi  Pi Þ2 ð5Þ
models are evaluated. It should be mentioned that the N i¼1
collected data from 30 blast events that were not incor-  
porated in the models were used for testing and vali- 1 XN Ai  Pi 
MAPE ¼ i¼1 
100 ð6Þ
dating the models. Table 6 shows this comparison N Ai 
between the predicted PPV using the fuzzy model, mul-
tiple regression model and empirical predictor models where Ai and Pi are the measured (actual) and predicted
with the measured PPV values. As shown in Table 6, values, respectively, and N is the number of testing

Downloaded from jvc.sagepub.com at UNIV OF UTAH SALT LAKE CITY on November 28, 2014
Ghasemi et al. 767

Figure 13. Measured versus predicted peak particle velocity (PPV) (fuzzy model).

Figure 14. Measured versus predicted peak particle velocity (PPV) (Ambraseys–Hendron and Indian Standard predictors).

samples. The calculated indices are given in Table 7. empirical predictor models is approximately identical,
The comparison of VAF, RMSE and MAPE indices except for the Indian Standard predictor. The correla-
and the correlation coefficient (R2) for predicting PPV tion coefficient of these predictors varies between
between these models indicated that the prediction 45.40% for the Indian Standard and 65.25% for the
performance of the fuzzy model is higher than that of Ambraseys–Hendron model (Figure 14).
the multiple regression and empirical models. Figure 13
shows the relationship between measured and predicted
PPV values, with a good correlation coefficient for the
8. Results and discussion
fuzzy model. The multiple regression model with a cor- Ground vibration is an undesirable and important side
relation coefficient of 34.55% showed lower prediction product of rock blasting in the mining industry and its
performance in comparison with other models. As can prediction is significantly important for controlling and
be seen in Table 7, the prediction performance of eliminating associated environmental problems. In this

Downloaded from jvc.sagepub.com at UNIV OF UTAH SALT LAKE CITY on November 28, 2014
768 Journal of Vibration and Control 19(5)

paper, fuzzy logic, multiple regression and empirical intuition can be effectively used for prediction of
predictors were used for the prediction of PPV (as a PPV, which helps in field applications.
vibration indicator) and their results were compared. 6. It should be noted that the proposed fuzzy model
The models of fuzzy logic and multiple regression can be used only for predicting PPV in
were constructed using six inputs and one output. The Sarcheshmeh copper mine and it should not be
fuzzy model was developed based on the Mamdani used directly for PPV prediction in other surface
algorithm and triangular fuzzy membership functions mines. However, the methodology employed in the
were adopted for describing input and output variables. present paper can also be used in other surface mines
In addition, it was constructed based on 229 if–then that employ blasting operation in order to predict
fuzzy rules and the COA method for defuzzification. PPV.
The results of the present paper can be explained as 7. Finally, it is clear that the proposed fuzzy model can
follows. be improved based on more data, which can be
obtained from blasting operations over time.
1. The results of the models for PPV prediction showed
that the equation obtained from the multiple regres-
sion model has lower prediction performance in Acknowledgments
comparison with other models. The reason for this The authors would like to express their thanks to Mrs Ifa
is that the developed regression model in this paper Mahboobi for her kind help and useful suggestions during
is based on the linear relationship between input and the preparation of manuscript. The authors are also very
output variables, whereas the nature of the relation- much grateful to Mr Hesam Dehghani for providing the
ship is complex. database.
2. The empirical predictor models for PPV prediction
revealed a high predictive capacity in comparison Funding
with the multiple regression model, but lower capac-
This research received no specific grant from any funding
ity than that of the fuzzy model. There are two main
agency in the public, commercial or not-for-profit sectors.
reasons for this; the first is that the empirical predic-
tors are site specific and are not suitable for other
sites and the second is that these predictors are based References
on only two parameters, maximum charge per delay Acaroglu O (2011) Prediction of thrust and torque require-
and distance from blast location, and do not include ments of TBMs with fuzzy logic models. Tunnelling and
other effective parameters. Underground Space Technology 26(2): 267–275.
3. In order to predict the PPV, the fuzzy model was Ak H, Iphar M, Yavuz M, et al. (2009) Evaluation of ground
applied successfully; it exhibited more reliable pre- vibration effect of blasting operations in a magnesite mine.
dictions than the regression and empirical models. Soil Dynamics and Earthquake Engineering 29(4): 669–676.
4. The practical outcome of the proposed fuzzy model, Ambraseys NR and Hendron AJ (1968) Dynamic behavior
of rock masses. In: Stagg KG and Zienkiewicz OC (eds)
with acceptable accuracy, can be considered as a
Rock Mechanics in Engineering Practices. New York:
preliminary estimation of PPV and based on it
Wiley, pp. 203–207.
many environmental impacts due to ground vibra- Aydin A (2004) Fuzzy set approaches to classification of rock
tion can be controlled and reduced. It should be masses. Engineering Geology 74(3–4): 227–245.
noted that the most important measure for control- Bakhshandeh Amnieh H, Mozaianfard MR and Siamaki A
ling ground vibration is proper design of controlla- (2010) Predicting of blasting vibrations in Sarcheshmeh
ble blasting parameters. Therefore, the main copper mine by neural network. Safety Science 48(3):
recommendations for ground vibration control are: 319–325.
(1) reducing the maximum charge per delay using Bhandari S (1997) Engineering Rock Blasting Operations.
proper delay interval, reduced blast hole diameter Rotterdam: A.A. Balkema.
and/or deck loading; (2) changing the blast geometry Blair DP and Jiang JJ (1995) Surface vibration due to a ver-
(including burden, spacing, stemming length and tical column of explosive. International Journal of Rock
Mechanics and Mining Sciences and Geomechanics
blast hole inclination) and/or type of explosive; (3)
Abstracts 32(2): 149–154.
using a minimum of practical subdrilling length; (4)
Davies B, Farmer IW and Attewell PB (1964) Ground vibra-
minimizing the degree of confinement by increasing tions from shallow sub-surface blasts. The Engineer 217:
the free face. 553–559.
5. Based on these results, fuzzy logic is a useful and Dehghani H and Ataee-pour M (2011) Development of a
powerful means to enhance the efficiency of blasting model to predict peak particle velocity in a blasting oper-
operation in surface mining. The major advantage ation. International Journal of Rock Mechanics and Mining
of the fuzzy model is that human judgment and Sciences 48(1): 51–58.

Downloaded from jvc.sagepub.com at UNIV OF UTAH SALT LAKE CITY on November 28, 2014
Ghasemi et al. 769

Duvall WI and Petkof B (1959) Spherical propagation of Soil Dynamics and Earthquake Engineering 24(11):
explosion generated strain pulses in rock. US Bureau of 887–892.
Mines Report of Investigation 5483. Kamali M and Ataei M (2011) Prediction of blast induced
Fisne A, Kuzu C and Hudaverdi T (2010) Prediction of envi- vibrations in the structures of Karoun III power plant and
ronmental impacts of quarry blasting operation using dam. Journal of Vibration and Control 17(4): 541–548.
fuzzy logic. Environmental Monitoring and Assessment. Khademi Hamidi J, Shahriar K, Rezai B, et al. (2010)
Epub ahead of print 1 May 2010. DOI: 10.1007/s10661- Application of fuzzy set theory to rock engineering classi-
010-1470-z. fication systems: an illustration of the rock mass excava-
Ghasemi E, Ataei M and Shahriar K (2011) Prediction of bility index. Rock Mechanics and Rock Engineering 43(3):
roof fall rate in coal mines using fuzzy logic. In: proceed- 335–350.
ings of the 30th international conference on ground control Khandelwal M, Kumar DL and Yellishetty M (2009)
in mining, Morgantown, West Virginia University, USA, Application of soft computing to predict blasting-induced
pp.186–191. ground vibration. Engineering with Computers. Epub
Ghosh A and Daemen JK (1983) A simple new blast vibra- ahead of print 13 November 2009. DOI: 10.1007/s00366-
tion predictor (based on wave propagation laws). In: pro- 009-0157-y.
ceedings of the 24th US symposium on rock mechanics, Khandelwal M and Singh TN (2006) Prediction of blast
Texas, USA, pp.151–161. induced ground vibrations and frequency in opencast
Giraudi A, Cardu M and Kecojevic V (2009) An assessment mine: a neural network approach. Journal of Sound and
of blasting vibration: a case study on quarry operation. Vibration 289(4-5): 711–725.
American Journal of Environmental Sciences 5(4): 463–473. Khandelwal M and Singh TN (2007) Evaluation of blast-
Gokceoglu C (2002) A fuzzy triangular chart to predict the induced ground vibration predictors. Soil Dynamics and
uniaxial compressive strength of Ankara agglomerates Earthquake Engineering 27(2): 116–125.
from their petrographic composition. Engineering Khandelwal M and Singh TN (2009) Prediction of blast-
Geology 66(1–2): 39–51. induced ground vibration using artificial neural network.
Gokceoglu C and Zorlu K (2004) A fuzzy model to predict International Journal of Rock Mechanics and Mining
the uniaxial compressive strength and the modulus of elas- Sciences 46(7): 1214–1222.
ticity of a problematic rock. Engineering Applications of Kuzu C (2008) The importance of site-specific characters in
Artificial Intelligence 17(1): 61–72. prediction models for blast-induced ground vibrations.
Grima MA and Babuska R (1999) Fuzzy model for the pre- Soil Dynamics and Earthquake Engineering 28(5): 405–414.
diction of unconfined compressive strength of rock sam- Kuzu C and Ergin H (2005) An assessment of environmental
ples. International Journal of Rock Mechanics and Mining impacts of quarry-blasting operation: a case study in
Sciences 36(3): 339–349. Istanbul, Turkey. Environmental Geology 48(2): 211–217.
Grima MA, Bruines PA and Verhoef PNW (2000) Modelling Langefors U and Kihlstrom B (1963) The Modern Technique
tunnel boring machine performance by neuro-fuzzy meth- of Rock Blasting. New York: Wiley.
ods. Tunnelling and Underground Space Technology 15(3): Mesec J, Kovac I and Soldo B (2010) Estimation of particle
259–269. velocity based on blast event measurements at different
Gupta RN, Pal Roy P and Singh B (1987) On a blast induced rock units. Soil Dynamics and Earthquake Engineering
blast vibration predictor for efficient blasting. In: proceedings 30(10): 1004–1009.
of the 22nd international conference on safety in Mines Mohamed MT (2009) Artificial neural network for prediction
Research Institute, Beijing, China, pp.1015–1021. and control of blasting vibrations in Assiut (Egypt) lime-
Indian Standard Institute (1973) Criteria for safety and design stone quarry. International Journal of Rock Mechanics and
of structures subjected to underground blast. ISI Bull Mining Sciences 46(2): 426–431.
IS-6922. Mohamed MT (2011) Performance of fuzzy logic and artifi-
Iphar M and Goktan RM (2006) An application of fuzzy sets cial neural network in prediction of ground and air vibra-
to the Diggability Index Rating method for surface mine tions. International Journal of Rock Mechanics and Mining
equipment selection. International Journal of Rock Sciences 48(5): 845–851.
Mechanics and Mining Sciences 43(2): 253–266. Monjezi M, Ahmadi M, Sheikhan M, et al. (2010) Predicting
Iphar M, Yavuz M and Ak H (2008) prediction of ground blast-induced ground vibration using various types of
vibrations resulting from the blasting operations in an neural networks. Soil Dynamics and Earthquake
open-pit mine by adaptive neuro-fuzzy inference system. Engineering 30(11): 1233–1236.
Environmental Geology 56(1): 97–107. Ozer U, Kahriman A, Aksoy M, et al. (2008) The analysis of
Jang JR (1993) ANFIS: Adaptive-Network-Based Fuzzy ground vibrations induced by bench blasting at Akyol
Inference System. IEEE Transactions on Systems, Man, quarry and practical blasting charts. Environmental
and Cybernetics 23(3): 665–685. Geology 54(4): 737–743.
Kahriman A (2002) Analysis of ground vibrations caused by Pal Roy P (1991) Vibration control in an opencast mine based
bench blasting at an open-pit lignite mine in Turkey. on improved blast vibration predictors. Mining Science
Environmental Geology 41(6): 653–661. and Technology 12(2): 157–165.
Kahriman A (2004) Analysis of parameters of ground vibra- Rai R and Singh TN (2004) A new predictor for ground
tion produced from bench blasting at a limestone quarry. vibration prediction and its comparison with other

Downloaded from jvc.sagepub.com at UNIV OF UTAH SALT LAKE CITY on November 28, 2014
770 Journal of Vibration and Control 19(5)

predictors. Indian Journal of Engineering and Materials extra parameters. International Journal of Rock Mechanics
Sciences 11(3): 178–184. and Mining Sciences 43(6): 938–949.
Ross TJ (1995) Fuzzy Logic with Engineering Applications. Uysal O, Arpaz E and Berber M (2007) Studies on the effect
New York: McGraw-Hill. of burden width on blast-induced vibration in open-pit
Singh PK, Sirveiya AK, Babu KN, et al. (2006) Evolution of mines. Environmental Geology 53(3): 643–650.
effective charge weight per delay for prediction of ground Valdivia C, Vega M, Scherpenisse CR, et al. (2003) Vibration
vibrations generated from blasting in a limestone mine. simulation method to control stability in the northeast
International Journal of Mining, Reclamation and corner of Escondia mine. Fragblast 7(2): 63–78.
Environment 20(1): 4–19. Verma AK and Singh TN (2010) Intelligent systems for
Singh TN and Singh V (2005) An intelligent approach to ground vibration measurement: a comparative study.
predict and control ground vibration in mines. Engineering with Computers. Epub ahead of print 7 July
Geotechnical and Geological Engineering 23(3): 249–262. 2010. DOI: 10.1007/s00366-010-0193-7.
Singh TN and Verma AK (2010) Sensitivity of total charge Yilmaz I and Kaynar O (2011) Multiple regression, ANN
and maximum charge per delay on ground vibration. (RBF, MLP) and ANFIS models for prediction of swell
Geomatics, Natural Hazards and Risk 1(3): 259–272. potential of clayey soils. Expert Systems with Applications
Siskind DE (2000) Vibrations from Blasting. Cleveland, OH: 38(5): 5958–5966.
ISEE Publication. Zadeh LA (1965) Fuzzy sets. Information and Control 8(3):
Tzamos S and Sofianos AI (2006) Extending the Q system’s 338–353.
prediction of support in tunnels employing fuzzy logic and

Downloaded from jvc.sagepub.com at UNIV OF UTAH SALT LAKE CITY on November 28, 2014