Vous êtes sur la page 1sur 14

Computers and Geotechnics 33 (2006) 15–28

www.elsevier.com/locate/compgeo

FEM models including randomness and its application to the


blasting vibrations prediction
Javier Toraño *, Rafael Rodrı́guez, Isidro Diego, José M. Rivas, Marı́a D. Casal
School of Mines, Oviedo University, Independencia 13, 33004, Oviedo, Spain

Received 19 April 2005; received in revised form 23 November 2005; accepted 23 January 2006
Available online 15 March 2006

Abstract

This paper shows the development of a FEM model that predicts ground vibrations due to blasting, a very complex problem involving
factors very difficult to quantify: ground anisotropy, sequencing time of the blasting shots, action of each shot, etc. From the vibration
records in several blasts in a limestone quarry, we have approximated the referred factors obtaining a model that simulates with enough
accuracy the complex waves originated in real blasts. Afterwards, we detail a work methodology which was developed including random-
ness. This methodology allows the practical use of the models by the engineers who design the blasts.
 2006 Elsevier Ltd. All rights reserved.

Keywords: Ground vibrations; Blasting; Vibration monitoring; FEM modelling

1. Introduction only in the seismic phenomena but also recently in the


explosion and blasting analysis (e.g., Jia et al. [2], Chen
The only profitable method nowadays available for the and Zhao [3], Toraño and Rodrı́guez [4], Wu et al. [5],
hard rock excavation is the use of explosives. In mining Liu et al. [6]).
or civil works projects, the usual procedure is making This paper is part of a research project granted with
blasts in areas that are highly close to inhabited places; public funds by the Spanish Ministerio de Fomento (Minis-
so buildings can be damaged by the vibrations transmitted try of Development) whose objective is to develop FEM
through the ground if the blast has not been correctly models that can predict the negative effects of blasting in
designed. During the last 40 years, many researchers have buildings and structures and where the ground vibration
studied this problem in all its different viewpoints. The first prediction is essential. A key aspect considered during
patterns of vibration transmission, empiric formulations, our research is the intentional use of general purpose soft-
were developed in the past sixties, e.g., Edwards and ware and standard PC hardware, so the calculation meth-
Northwood [1]. These studies were completed in the follow- ods and results can be easily applied by any small and
ing years using research campaigns in order to determine middle-sized companies.
the damage produced by vibrations in structures or the During the research development and after an initial
behaviour of structures against seismic phenomena. phase where the modelling fundamentals were settled, we
Finally, in the 1990s, the development of powerful and moved to another phase where we were looking for the
low-cost software and hardware allowed to study the vibra- most exact replication of the vibration waves through the
tion problems using numerical simulation methods, not measuring of vibrations in a high number of blasts of var-
ious kinds. Work started based on the results obtained
from the sensing of several blasts in a limestone quarry
*
Corresponding author. Tel.: +34 985 104 254; fax: +34 985 104 245. using this installation as a real scale laboratory. We chose
E-mail address: jta@uniovi.es (J. Toraño). a quarry not only because it gathers several interesting

0266-352X/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compgeo.2006.01.003
16 J. Toraño et al. / Computers and Geotechnics 33 (2006) 15–28

Fig. 1. Quarry bench very near to buildings.

characteristics for ground modelling as a very well known material behaviour and movements and forces created by
geology and a blast design with low number of shots, but these movements. The simulation, which considers vibra-
also because it reflects a typical problem in many areas of tion, impact and movement, is guided by the widely known
Spain: the need of excavating rocks (mining or civil works) equation:
in mountainous terrain with closely by inhabited buildings ½Mfag þ ½Cfvg þ ½Kfdg ¼ fF g ð1Þ
(see Fig. 1).
where [M] is the mass matrix, {a} is the acceleration vector,
2. Elaboration and calibration of the previous model [C] is the damping matrix, {v} is the velocity vector, [K] is
the stiffness matrix, {d} is the displacement vector and {F}
The ground vibration appears because of the instanta- is the external nodal forces vector.
neous pressure caused by the explosive makes during the The geometric model (Fig. 2) is a ground block measur-
detonation. One of the models previously developed by ing 2000 m long and 500 m high, with a bench height of
the authors (Toraño and Rodrı́guez [4]) reproduces the 20 m in its upper zone which is the area where the blasting
vibration generated by a single shot inside a granite rock is simulated. The model mesh has 10,754 elements and
mass, and consists of a ground block with a bench where 10,767 nodes. This geometry is highly oversized in order
a pressure pulse lasting for a few milliseconds actuates to study the vibrations at high distances and also to sim-
against the bench face. In the present study, we repeat plify the boundary conditions: null horizontal displace-
the process for a single shot in a limestone rock mass. This ments at lateral boundaries and null vertical
model was developed using commercial software based on displacements at the bottom of the model (considering
the finite element method (Anon [7], Belles and Vicente [8]) the wave damping, the results are not affected by the pos-
that allows a non-linear dynamic analysis, using 2D model sible wave reflections in the model limits). Smaller models
design and supposing plain deformations. The software would require more complex boundary conditions as in
employed allows analyzing mechanical events including Liu et al. [6]. The element sizes through the first 500 m vary
simultaneously wide deformations, non-linearity of the from 2 · 2 m2 near the pressuring point to 16 · 16 m2 in the

Fig. 2. Geometric model for the vibration analysis due to blasting.


J. Toraño et al. / Computers and Geotechnics 33 (2006) 15–28 17

border and where the results are more interesting the ele- strated that in some cases the impulse that is the form of
ment size is limited to a maximum of 8 · 8 m2, thus obtain- this time function is more important.
ing the inexistence of wave distortions according to the The model calibration was done according to a general
studies of Kuhlemeyer and Lysmer [9]. vibration transmissivity law in limestone obtained from
The ground is a limestone rock mass characterized by numerous field measurements (Balsa [19], Figueroa and
values based on our experience: density q = 2600 kg/m3, González [20]). Such law allows estimating the peak parti-
Young modulus E = 12,000 MPa and Poisson coefficient cle velocity, v, in a point of the ground surface (expressed in
t = 0.20. This value of the Young modulus can be esti- mm/s), against the scaled distance D/Q0.5 where Q is the
mated from geomechanics characteristics of the rock mass operating load (explosive load detonated on each micro-
using the expression of Serafim and Pereira [10] modified delay) and D the distance from the point where the explo-
by Hoek and Brown [11]: sion is caused. The curve that represents this law and also
rffiffiffiffiffiffiffiffi the points cloud taken from experience are shown in Fig. 3.
rci GSI10
E¼  10 40 ð2Þ This empirical law has the following mathematical
100 expression:
For a compressive strength of the intact rock rci = 70 MPa
v ¼ 3085  Q0:757  D1:651 ð4Þ
and an average value GSI = 57 (it varies between 50 and
65), the Young modulus is near E = 12 GPa. As will be seen below, the explosive load in the quarry
Taking into account that only in the vicinity of the blast blasting was approximately Q = 100 kg, so (4) turns to:
holes the rock gets yielded (Kutter and Fairhurst [12], v ¼ 100;752  D1:651 ð5Þ
McHugh [13], Donzé et al. [14], Essen et al. [15]), and in
some meters the blast wave has no energy to break the The model has been adjusted to allow the peak particle
rock, its long distance transmission is like an elastic wave. velocity variation against the distance and the applied pres-
Then, we will assume the elastic behaviour of the rock mass sure in a similar way as in reality. Thus, (4) changes to:
(this is q, E and t remain constant). v ¼ 3573  P  D1:611 ð6Þ
The method of introducing damping in the model is
done through the Rayleigh classical approach, by making If we consider P = 28 MPa, we obtain a law that approxi-
the damping matrix equal to a lineal combination of the mates the general empirical law given by (5):
mass and stiffness matrix: v ¼ 100;044  D1:611 ð7Þ
½C ¼ a1 ½M þ a2 ½K ð3Þ The limestone transmissivity law for Q = 100 kg and the
The used values in our case are a1 = 6.30 and a2 = 0.05. law obtained from the model that uses P = 28 MPa are
The shape of the pressure pulse that hits the bench face also shown in Fig. 3. Analyzing the figure, we can check
has been developed from the curve of volume increase in a the adequate fit of the model.
hole subject to an inner explosion, supposing that the rela- It is worth mentioning that little changes in the model
tion between pressure and volume follows an expression of (mesh, dimensions, shape of the pressure pulse . . . etc.)
the kind PVc = constant (Mortazavi and Katsabanis [16]); cause changes in the results, but the model can
the pulse length is less than 10 ms and initially an arbitrary always be easily readjusted. On the other hand, this var-
maximum pressure value of 25 MPa is taken. Some iability in the model is avoided in the second part of the
authors, Ding and Zheng [17], Ding et al. [18], have demon- study.

Limestone vibrations transmisivity law Limestone vibrations transmisivity patterns


10000 1000
Peak particle velocity (mm/s)

Empirical transmisivity Law (Q=100kg)


Peak particle velocity (mm/s)

1000 FEM model (P=28 MPa)


Dominant frequency: 25 Hz 100

100
10
10
1
1

0,1
0.1

0.01 0,01
1 10 100 1000 10000 10 100 1000 10000
Scaled distance (D/Q 0.458 ) Distance (m)

Fig. 3. Vibrations transmissivity patterns for limestone (real and model).


18 J. Toraño et al. / Computers and Geotechnics 33 (2006) 15–28

Fig. 4. Overview of the quarry.

3. Measurement campaign Blasting numbers 2, 3 and 4 are similar. The numbers of


shots were 11, 12 and 13, all 18 m long and with approxi-
In this research, the vibrations produced by four blasts mate loads of 109, 108 and 106 kg of explosive per hole.
in a quarry have been measured. In each blast, the vibra- The blast design is more or less constant: 4 m of burden
tions were simultaneously measured in two points placing and 5 m between holes.
seismographers at two different distances, always perpen- Blast number 1 is the only one which is different, as it is
dicular to the blasting line. A photograph of the quarry done in an access road to the bench; thus, the 13 shots have
is shown in Fig. 4 and a plan with the location of the blast- different lengths and loads: 12.5 m and 63 kg, 13 m and
ing benches and the position of the seismographers is 67 kg, 13.5 m and 70 kg, 14 m and 73 kg, 14.5 m and
shown in Fig. 5. 77 kg, 15 and 80 kg, 15.5 m and 83 kg and finally six shots
of 18 m and 100 kg.
In all cases, the connection between the different shots is
done through a non-electric shot line (by means of shock
wave transmission tube). In order to sequence the shots,
a micro-delayed detonator of 25 ms is introduced between
consecutive shots, so the maximum charge per delay load is
limited only to one shot. The clay stopper length is always
around 3 m.
Seismographers were installed at 185–240 m in blast no.
1, and 100–180–240 m in the other blasts. In all cases seis-
mographers were in the quarry concession, several hundreds
of meters apart from the nearest houses. All of the records
gave information about the transmissivity law of the lime-
stone. Nevertheless, the records of the seismographers
located 240 m away from blasting were not representative
of the waveform because of the low magnitude of the veloc-
ity. On the other hand, in order to simplify the study, four
records have been chosen: two of them corresponding to
100 m distance and the other two corresponding to 180 m.
The records of the four blasts are shown in Figs.6–9. At
Fig. 5. Location of the blasting lines and seismographers. a first sight, one of the characteristics of this phenomenon

Blasting No. 1 (185 m)


10
Vertical velocity (mm/s)

7,5
5
2,5
0
-2,5 0 0,1 0,2 0,3 0,4 0,5 0,6
-5
-7,5
-10
Time (s)

Fig. 6. Vertical velocity record of blasting no. 1 (185 m).


J. Toraño et al. / Computers and Geotechnics 33 (2006) 15–28 19

Blasting No. 2 (180 m)


Vertical velocity (mm/s) 10
7,5
5
2,5
0
-2,5 0 0,1 0,2 0,3 0,4 0,5 0,6
-5
-7,5
-10
Time (s)

Fig. 7. Vertical velocity record of blasting no. 2 (180 m).

Blasting No. 3 (100 m)


20
Vertical velocity (mm/s)

15
10
5
0
-5 0 0,1 0,2 0,3 0,4 0,5 0,6
-10
-15
-20
Time (s)

Fig. 8. Vertical velocity record of blasting no. 3 (100 m).

Blasting No. 4 (100 m)


20
Vertical velocity (mm/s)

15
10
5
0
-5 0 0,1 0,2 0,3 0,4 0,5 0,6
-10
-15
-20
Time (s)

Fig. 9. Vertical velocity record of blasting no. 4 (100 m).

is clearly seen: its randomness. The big amounts of influ- 4. Re-calibration of the model and its use in reproducing the
encing variables make vibrations highly unpredictable waves
and even when sensing similar blasts, the results are often
very different. On the other hand, the transmissivity law 4.1. Using the previous model
is always accomplished: for the same distance and for the
same blasting loads the peak particle velocity is always in As was referred above, after an initial phase where the
a short range. modelling fundamentals were settled we moved to another
In this study, we have only taken into account vertical phase where the goal is to obtain the most exact reproduc-
velocities. The measurement of the longitudinal and the tion of the vibration waves. This new calibration process
cross velocities depends on the seismograph geophone will be illustrated from the blasting no. 1 data (vibration
orientation and the relative position in relation to the record at 185 m) because it is the one that presents most
blasting line of the three available velocity sensors. different influencing factors. In Table 1, the length of each
Taking into account the uncertainty that the cross shot, L, the explosive charge, Q, and the nominal delay
component has and also the two-dimensional characteris- between shots, r, are resumed.
tics of the model, the study was simplified on this A first attempt using the model is done by producing 13
sense. identical pressure pulses, separated by 25 ms in time and
20 J. Toraño et al. / Computers and Geotechnics 33 (2006) 15–28

Table 1
Data of blasting no. 1
Blasthole 1 2 3 4 5 6 7 8 9 10 11 12 13
L (m) 12.5 13 13.5 14 14.5 15 15.5 18 18 18 18 18 18
r (ms) 0 25 25 25 25 25 25 25 25 25 25 25 25
Q (kg) 63 67 70 73 77 80 83 100 100 100 100 100 100

Table 2
Pressure and delay for each shot (1st simulation)
Blasthole 1 2 3 4 5 6 7 8 9 10 11 12 13
P0 (MPa) 28 28 28 28 28 28 28 28 28 28 28 28 28
r0 (ms) 0 25 25 25 25 25 25 25 25 25 25 25 25

30 FEM model wave Real wave

20
Vertical velocity (mm/s)

10

0
0,0 0,1 0,2 0,3 0,4 0,5 0,6
-10

-20

-30
Time (s)

Fig. 10. Real velocity record and model result (1st simulation).

with a 28 MPa pressure, corresponding to the maximum ground, which depends on the quarry localization and
charge per delay, around 100 kg of explosives (18 m holes). sometimes on the different quarry zones, because of differ-
The pressure and the delay introduced in the model for ent geological structures, faults, terrain elevations, . . . etc.
each shot are shown in Table 2. The result of the model From the results of this experience (the peak particle veloc-
vs. recorded wave is shown in Fig. 10. ity vs. distance for all of the recorded data is shown in
Fig. 10 shows that the obtained results are not satisfac- Fig. 11), it can be inferred that in this zone, and in the
tory at all, as the simulated wave is not close to the mea-
sured one. Peak velocity is almost 30 mm/s (five times
higher) and the dominant frequency is around 35 Hz, big- Limestone vibrations transmisivity patterns
ger than the real one, which is 25 Hz. The only good 1000
approximation is the length of the wave, around 0.5 s in General transmisivity Law (Q=100kg)
both cases. Quarry transmisivity law (Q=100 kg)
Peak particle velocity (mm/s)

These differences between both waves can be explained 100


as follows: the initial model has been executed using a
transmissivity law that is over the point cloud, that is, it
10
predicts a peak particle velocity prevailing over the normal
limestone. In the real world, the ground response can be
below that transmissivity law. 1

4.2. First adjustment factor: local ground behaviour


0,1
Since the first works about this subject (Langefors and
Khilström [21] refers a number of them), it is known that
0,01
both vibration amplitude and frequency depend on the
10 100 1000 10000
characteristics of the rock mass that transmits the wave.
Distance (m)
Therefore, to calibrate the model better we have to know
the transmissivity curve corresponding to such specific Fig. 11. General and local vibrations transmissivity laws.
J. Toraño et al. / Computers and Geotechnics 33 (2006) 15–28 21

blasting and sampling conditions used, we have to use a fragmentation but also diminishes the ground vibration
reducing coefficient of around 0.4. level. To take this into account, a 50% factor is arbitrarily
The best possibility is to repeat the preceding reasoning introduced in the pressure that simulates the explosive
but using the real transmissivity curve of the quarry (that effect on the rock mass from the second shot.
we know at those points), which will guide us to suppose On the other side, the bigger is the distance from the
that the 100 kg of explosives correspond in the model not blast hole to the free surface of the bench (burden) the less
to 28 MPa but to 0.4 · 28 = 11.2 MPa. rock is excavated by the blasting and bigger are the vibra-
tions transmitted to the ground. In the blasting that is
4.3. Second adjustment factor: explosive charge on each hole being studied, the dip angle of the bench was steeper in
the sides of the blasting line than in the central part. As
Another factor that has great influence on vibrations, the holes inclination was the same, the real burden was less
specifically in their maximum particle velocity, is the in the sides than in the centre. To take this into account, a
amount of explosives that detonates at a certain moment diminishing factor was used in both extreme holes.
(also analyzed in the classical studies, Langefors and Khil- The result of the modified model after these corrections
ström [21]). Analyzing each blast hole separately, we can is shown in Fig. 12. With these new adjustments, we have a
establish that each of them will affect the surrounding better fit of the model and the real sample, as we can
ground with an action proportional to the charge value. observe a diminishing in the peak values in the first and
As the distance of the holes to the wave measuring point final cycles.
is approximately the same, the methodology expressed in
Toraño and Rodrı́guez [4] can be followed and then Eqs. 4.5. Fourth adjustment factor: firing sequence
(4) and (6) can be combined as follows:
The other factor that has to be taken into account in the
P ¼ 0:86  Q0:757 ð8Þ
simulation of a real wave is the time interval existing
thus we can estimate approximately the pressure P (MPa) between the arriving of the different waves produced by dif-
that has to be introduced in the general model taking into ferent blast holes and the measuring point. Two variables
account the explosive load Q (kg) of one hole. Continuing have great influence on this: the sequence of the shot firing
with the model referred above, it was already explained and the real path of the wave from its origin to the measur-
that the value to introduce is around 40%, so Eq. (8) ing point. The first one is studied here as it can be somehow
becomes: be estimated.
As was explained, in order to optimize the blasting we
P ¼ 0:34  Q0:757 ð9Þ
introduce a sequence in the shot firing. This is done
through the use of conventional delay detonators. This sys-
4.4. Third adjustment factor: explosive confinement within tem introduces an error that could be, following field
the rock mass research (Winzer [22]), around 5–10 ms when using 25 ms
detonators (although manufacturers give minor values).
In order to get a good rock fragmentation, the good By this, the theoretical interval of 25 ms between shots
practice says that the explosive load in the different holes could really be between 15 and 35 ms.
cannot be detonated simultaneously but in a planned To check this in this study, video tapes of the shot
sequence. By this we can arrive at a conclusion that when sequence were used to have a reference of the effective time
a blast hole detonates, it encounters a pair of free surfaces of the shot initiation. Four photographs (at 40, 80,120 and
that eases its work: the bench and the void generated by the 320 ms from the beginning of the blasting) are shown in
previous blast hole detonation. This not only helps the rock Fig. 13 where the light corresponding to the transmission

8 FEM model wave Real wave

6
Vertical velocity (mm/s)

0
0,0 0,1 0,2 0,3 0,4 0,5 0,6
-2

-4

-6

-8
Time (s)

Fig. 12. Real velocity record and model result after three corrections.
22 J. Toraño et al. / Computers and Geotechnics 33 (2006) 15–28

Fig. 13. Photographs taken at times 40, 80, 120 and 320 ms.

line and the little detonator explosion can be seen crystal Taking all the information referred above into account,
clear. From their analysis it can be established: at 40 ms in order to define the moment to introduce the pressure
shot 1 is already initiated and shot 2 is reached, at 80 ms pulses in our model the following hypotheses were estab-
shot 3 is already initiated and shot 4 is reached, at lished: shots 2, 4, 5 and 11 detonate approximately from
120 ms shot 5 is reached and at 320 ms shot 10 is initiated the 40, 80, 120 and 320 ms, and the time interval between
and shot 11 is reached. It is clear that the real time delay pressure pulses could range between 15 and 35 ms. Com-
between shots is greater than 25 ms because if it would paring the model and the real waves (see Fig. 14) it can
be right, shot 10 would be initiated at 225 ms and not at be easily observed that we have improved the adjustment,
320 ms. mainly the position of the wave peaks.

8 FEM model wave Real wave


Vertical velocity (mm/s)

0
0,0 0,1 0,2 0,3 0,4 0,5 0,6
-2

-4

-6

-8
Time (s)

Fig. 14. Real velocity record and model result after 4th correction.
J. Toraño et al. / Computers and Geotechnics 33 (2006) 15–28 23

4.6. Other adjustment factors: factors difficult to evaluate these are the factors that we have still not taken into
account. The pressure and the delay introduced in the
Finally, other factors that have influence on the created model for each shot are shown in Table 3.
vibrations have to be taken into account.
For example, the relative position of the blast hole that 4.7. Re-calibration of the model for the blasting nos. 2, 3
detonates and the measuring point, or the anisotropy of the and 4
rock mass (that could make the wave to go through several
types of materials, faults . . . etc.), have influence mainly on Following the same methodology, the model has been
the wave velocity and the moment when the different waves re-calibrated. Then, the vertical velocity records of blasting
reach the seismographer. nos. 2, 3 and 4 at 185, 100 and 100 m have been reproduced
Other factors, such as the clay stopper, the contact of with good results as it can be seen in Figs. 16–18.
the explosive to the drill walls or the presence of water in
the hole, mainly influence the wave amplitude. The pres- 5. Advances obtained in the wave simulation through FEM
ence of these factors that exist but cannot be quantified is modelling
what allows us to continue adjusting the model, but now
purely in an empirical manner trying to get a better fit of As an example, Figs. 19 and 20 show the outputs of two
the model to the reality. On this model, we have adjusted of the elaborated models for blasting no. 2: one with the
specifically the pressure value on some shots with very minimum possible adjustment (after one correction) and
good results (see Fig. 15), which guide us to think that other with all the possible adjustments (after six correc-

8 FEM model wave Real wave

6
Vertical velocity (mm/s)

0
0,0 0,1 0,2 0,3 0,4 0,5 0,6
-2

-4

-6

-8
Time (s)

Fig. 15. Real velocity record and model result (blasting no. 1 at 185 m).

Table 3
Pressure and delay for each shot (final simulation)
Blasthole 1 2 3 4 5 6 7 8 9 10 11 12 13
P6 (MPa) 3.4 3.4 4.5 4.4 6.6 3.3 3.3 3.3 3.3 8.9 7.2 5.6 1.1
r6 (ms) 0 35 33 37 36 32 14 28 30 37 33 34 39

10 FEM model wave Real wave


8
6
Vertical velocity (mm/s)

4
2
0
-2 0,0 0,1 0,2 0,3 0,4 0,5 0,6
"
-4
-6
-8
-10
Time (s)

Fig. 16. Real velocity record and model result (blasting no. 2 at 180 m).
24 J. Toraño et al. / Computers and Geotechnics 33 (2006) 15–28

20 FEM model wave Real wave


15

10
Vertical velocity (mm/s)

5
0
0,0 0,1 0,2 0,3 0,4 0,5 0,6
-5
-10
-15
-20
Time (s)

Fig. 17. Real velocity record and model result (blasting no. 3 at 100 m).

20 FEM model wave Real wave

15

10
Vertical velocity (mm/s)

0
0,0 0,1 0,2 0,3 0,4 0,5 0,6
-5 "
-10

-15

-20
Time (s)

Fig. 18. Real velocity record and model result (blasting no. 4 at 100 m).

10 FEM model wave


8
Vertical velocity (mm/s)

6
4
2
0
-2 0,0 0,1 0,2 0,3 0,4 0,5 0,6
"
-4
-6
-8
-10
Time (s)

Fig. 19. FEM model wave of blasting no. 2 at 180 m (first simulation).

10 FEM model wave


8
Vertical velocity (mm/s)

6
4
2
0
-2 0,0 0,1 0,2 0,3 0,4 0,5 0,6
"
-4
-6
-8
-10
Time (s)

Fig. 20. FEM model wave of blasting no. 2 at 180 m (final simulation).
J. Toraño et al. / Computers and Geotechnics 33 (2006) 15–28 25

tions). By comparing them to the real waves, better results It is known that there is some degree of uncertainty
can be observed in the second one. about each shot detonation instant and how much energy
The more elaborated models simulate real waves with all fraction is transmitted to the ground as a seismic wave.
its complexity and even simulate the bigger vibration time In order to know which random variation can be intro-
length or the fact of a frequency diminishing when we add duced, we will make use of our experience. From the ana-
the effect of several shots. It is obvious that in order to pre- lyzed waves it can be seen that the mean delay between
dict the effect of blasting in buildings or surrounding struc- shots is 32 ms, but in order to adjust the waves as much
tures (our last main objective), it is much better to use the as possible there could be variations around ±7 ms. It
more elaborated models as an opposite to the simple one can also be checked from the models that in order to get
that just gives us simple waves with frequency and ampli- a better adjustment with the real wave, the pressure to be
tude almost constant. introduced in the model can vary around ±30% from the
one estimated with Eq. (10).
6. Practical use of the model In order to produce random results, the model inputs
can be randomized by adding to each delay a random num-
The availability of FEM models capable of reproducing ber between 7 and 7 and multiplying each of the pressures
complex waves similar to the ones found in reality has a by a random number between 0.7 and 1.3.
great interest from the scientific and academic point of Thus, five possible random waves generated by the
view, but has the inconvenience of its difficult direct appli- model appeared. But, from the practical point of view
cation by the engineers who are designing the blasts, at a the interest is not specifically in the waves but in the pos-
quarry in this case. In fact, the model replicates a wave that sible negative wave effects over the structures. These pos-
we know can be close to the real, but as all the real waves sible negative effects are generally defined by limiting the
are different, we know in advance that the wave reproduced peak particle velocity for different frequencies (BS 7385
by the model will not be the same as the one that can be [23], DIN 4150-3 [24], UNE 22-381-93 [25] or RI 8507
measured in the blasting. That is, models as have been seen [26]). In order to show this, the positions of the pairs
are useful in the analysis of historical cases, blasts already of values, frequency–maximum particle velocity of all
done, but are not so useful in the prediction work, when the the semi-cycles that compose the wave, are shown
goal is to know the future results of a blast. in the graphic of the damage criteria with the limit
In order to achieve that these tools can also be used in curves.
the daily work, a methodology that gets more practical Those points clouds corresponding to the five simulated
results has been developed. It uses the basic design param- waves of blasting no. 1 at 185, represented in the graphic,
eters of the blast as initial inputs and then introduces a ran- that define the damage criteria of Spanish standard UNE
domness factor that make possible reaching several 22.381-1993 are shown in Fig. 21. Group I are industrial
solutions. buildings and bays with concrete or metallic structures,
The working method is as follows. The method starts Group II are houses, offices buildings, shopping mal-
from the calibrated model that follows the general law of ls, . . . etc. or very strong architectonic structures and finally
transmissivity and the design parameters of the blast: num- Group III is formed by those archaeological, historic or
ber of shots of the blast (n), explosive charge on each shot special architectonic structures that are very sensitive to
(Qi) and the delay among shots, value depending on the vibrations.
detonators used. In this case, we will suppose a mean delay Rejecting the 5% of the points most separated from the
between shots tr = 32 ms, which is the measured mean (we cloud, a surrounding area is defined that includes the rest
can also use the detonators nominal value, 25 ms). of the characteristics points of the random waves. These
The blast is simulated through ‘‘n’’ pressure pulses, areas are defined by a minimum frequency and a maximum
equally spaced at time tr. The maximum pressure on each velocity of the semi-cycle that in the case of blasting no. 1
pulse is calculated, as was seen, from the explosive charge are 18 Hz and 6 mm/s at 185 m. And this is the result that
on each shot: is really interesting for the engineer in his day to day work:
he will not be interested in the wave but in the area of the
P i ¼ k 1  k 2  ðk 3  Q0:757
i Þ ð10Þ
damage criteria graphic where the representing points of
where k1 is the local conditions factor, in our case k1 = 0.40 his blast will be included. This result can immediately eval-
(k1 = 1 can be used but higher peak particle velocity values uate the blast. Taking into account how the models have
are obtained), k2 is a factor equal to 1 in the first shot and been resolved, it can be inferred that all the blastings sim-
0.5 for the other ones, k3 is the adjustment factor between ilar to the no. 1 (13 shots with their different loads) will give
the model and the general transmissivity law, in this case waves that, initially, cannot be predicted but we can say
k3 = 0.86. We have to insist in the fact that these values that at 185 these will not cause damage to buildings below
used in this model might be modified for another model this criterion, as the area defined by the models is below the
with different geometry, dimensions, number of ele- limit curves. Representing the real wave points, it can be
ments, . . . etc. Nevertheless, it is not difficult to follow the checked that they are effectively inside those estimated
methodology in order to calibrate a different model. areas (see Fig. 21).
26 J. Toraño et al. / Computers and Geotechnics 33 (2006) 15–28

1000
1000

Max. cycle velocity (mm/s)


Max. cycle velocity (mm/s)

Group I Group I
100 100
Group II Group II

Group III Group III

10 10 FEM modelcurve
FEM model curve

Real wave points


FEM model points

1 1
1 10 100 1000 1 10 100 1000
Frequency (Hz) Frequency (Hz)

Fig. 21. Definition of the area through FEM model and real wave points cloud (blasting no. 1 at 185 m).

The estimated areas by this random technique and the If we intend doing a blast analysis in this quarry (blasts
real point cloud on each of the remaining blast nos. 2, 3 nos. 2, 3 and 4), we will conclude that using this blast
and 4 are shown in Figs. 22–24. We must say that using five scheme we can only have problems at distances around
random inputs, we already have a point cloud wide enough 100 m, as the estimated areas cut the limit curves of Groups
to define perfectly each case area. II and III. It can also be concluded that at 180 m it is very

1000 1000
Max. cycle velocity (mm/s)
Max. cycle velocity (mm/s)

Group I Grupo I
100 100
Group II Grupo II

Group III Grupo III

10 10
FEM modelcurve FEM model curve

FEM model points


Real wavepoints
1 1
1 10 100 1000 1 10 100 1000
Frequency (Hz) Frequency (Hz)

Fig. 22. Definition of the area through FEM model and real wave points cloud (blasting no. 2 at 180 m).

1000 1000
Max. cycle velocity (mm/s)
Max. cycle velocity (mm/s)

Group I Group I
100 100
Group II Group II

Group III Group III

FEM model curve FEM model curve


10 10

FEM modelpoints Real wave points

1 1
1 10 100 1000 1 10 100 1000

Frequency (Hz) Frequency (Hz)

Fig. 23. Definition of the area through FEM model and real wave points cloud (blasting no. 3 at 100 m).
J. Toraño et al. / Computers and Geotechnics 33 (2006) 15–28 27

1000 1000
Max. cycle velocity (mm/s)

Max. cycle velocity (mm/s)


Group I Group I
100 100
Group II Group II

Group III Group III

FEM model curve FEM model curve


10 10

Real wave points


FEM model points

1 1
1 10 100 1000 1 10 100 1000
Frequency (Hz) Frequency (Hz)

Fig. 24. Definition of the area through FEM model and real wave points cloud (blasting no. 4 at 100 m).

unlikely that these blasts can produce any damage, as the Acknowledgement
estimated areas are below the building protection curves
of Group II. The authors acknowledge the financial support given to
We must say that due to the quarry size the nearest the project by the Spanish Ministerio de Fomento (Ministry
buildings are 500 m far from the actual working benches, of Development) within the Priority Action Framework
so the blasts are correctly designed to avoid any damage ‘‘New Technologies and Constructive Systems’’, of the Sec-
to the surrounding buildings. torial Area ‘‘Building and Cultural Heritage Conservation’’
(National R+D Plan 2000–2003).

7. Conclusions References

In this paper, the development of a non-complex FEM [1] Edwards AJ, Northwood TD. Experimental studies of the effects of
blasting on structures. The Engineer 1960;210.
model is summarized that allows the prediction of the com-
[2] Jia Z, Chen G, Huang S. Computer simulation of open pit bench
plex ground vibrations due to blasting. All the factors that blasting in jointed rock mass. Int J Rock Mech Min Sci
are known to have influence on the vibrations have been 1998;35(4–5):476–86.
introduced in the model trying to evaluate and quantify [3] Chen SG, Zhao J. A study of UDEC modelling for blast wave
their effect as much as possible. The more elaborated mod- propagation in jointed rock masses. Int J Rock Mech Min Sci
1998;35(1):93–9.
els simulate the real complex waves and even simulate the
[4] Toraño J, Rodrı́guez R. Simulation of the vibrations produced during
fact of frequency diminishing when we add the effect of sev- the rock excavation by different methods. In: Proceedings of the XI
eral shots. It is obvious that in order to predict the effect of international conference on comp meth and exp measures. London:
blasting in buildings or surrounding structures (our last WIT Press; 2003. p. 343–9.
main objective) it is much better to use the more elaborated [5] Wu C, Lu Y, Hao H. Numerical prediction of blast-induced stress
wave from large-scale underground explosion. Int J Numer Anal
models as an opposite to the simple one that just gives us
Metho Geomech 2004;28:93–109.
simple waves with frequency and amplitude almost [6] Liu YQ, Li HB, Zhao J, Li JR, Zhou QC. UDEC simulation for
constant. dynamic response of a rock slope subject to explosions. Int J Rock
A simple statistical study will guide to an approximated Mech Min Sci 2004;41(3). CD-ROM.
solution of the same type as the one exposed, as shown in [7] Anon. Algor12. Finite element analysis and event simulation
software. DocuTehc (user’s guide). USA: Algor Inc; 2000.
the damage criteria. Now then, this statistical analysis can-
[8] Belles P, Vicente P. Ensayo sı́smico en un laboratorio virtual.
not allow reproducing ‘‘real waves’’, which is our new Ingenierı́a Civil 2001;123:53–9.
model contribution. The model works as a virtual labora- [9] Kuhlemeyer R, Lysmer J. Finite element method accuracy for wave
tory that generates possible waves and that can be used propagation problems. J Soil Mech Foundations Div. ASCE 1973;
with other rock types, as the rock characterization in order 99:411–7.
[10] Serafim JL, Pereira JP. Consideration of the geomechanical classifi-
to use FEM models is something that is widely studied and
cation of Bieniawski. In: Proceedings of the international symposium
that can benefit the engineer working with these studied. on engineering, geology and underground construction, Lisbon 1(II),
In order to achieve that these tools can also be used in 1983. p. 33–44.
the daily work wit, a methodology has been developed that [11] Hoek E, Brown ET. Practical stimates of rock mass strength. Int J
gets more practical results. It uses the basic design param- Rock Mech Min Sci 1997;34(8):1165–86.
[12] Kutter HK, Fairhurst C. On the fracture process in blasting. Int J
eters of the blast as initial inputs and then introduces a ran-
Rock Mech Min Sci 1971;8:181–202.
domness factor that makes reaching several solutions [13] McHugh S. Crack extension caused by internal gas pressure
possible. These different solutions help us to define the risk compared with extension caused by tensile stress. Int J Fract
of existence of negative effects due to a real blasting. 1983;21:163–76.
28 J. Toraño et al. / Computers and Geotechnics 33 (2006) 15–28

[14] Donzé FV, Bouchez J, Magnier SA. Modeling fractures in rock [20] Figueroa A, González E. Resultados del tratamiento estadı́stico de los
blasting. Int J Rock Mech Min Sci 1998;34(8):1153–63. estudios de vibraciones producidos por voladuras. Canteras y
[15] Essen S, Onederra I, Bilgin HA. Modelling the size of the crushed Explotaciones 1989;263:104–15.
zone around a blast hole. Int J Rock Mech Min Sci 2003;40:485–95. [21] Langefors U, Kihlström B. The modern technique of rock blasting.
[16] Mortazavi A, Katsabanis PD. Modelling burden size and strata dip Gebers Förlag AB, Stockholm: Almquist & Wiskell; 1963.
effects on the surface blasting process. Int J Rock Mech Min Sci [22] Winzer SR. Indiciator firing times and their relationship to
2001;38:481–98. blasting performance. In: 20th US symposium on rock mechanics,
[17] Ding H, Zheng ZM. Source model of blasting vibration. Sci China, 1979.
Series E–Technol Sci 2002;45(4):395–407. [23] Standard BS 7385. Evaluation and measurement for vibration in
[18] Ding H, Labbas R, Zheng ZM. Features of blast-induced vibration buildings. United Kingdom.
source and identification of geostructures. J Sound Vib [24] Standard DIN 4150-3. Erschütterungen im Bawessen. Germany.
2005;288(1–2):91–106. [25] Standard UNE 22-381-93. Control de vibraciones producidas por
[19] Balsa J. Leyes estadı́sticas de transmisividad en distintos tipos de voladuras. Spain.
roca. Canteras y Explotaciones 1989;272:61–73. [26] US Buerau of Mines Researching Inform RI 8507. USA.