INTERNATIONAL JOURNAL FOR NUMERICAL A N D ANALYTICAL METHODS IN GEOMECHANICS. VOL. 5.

97-113 (1981)

FINITE ELEMENT MODELLING OF GROUND SURFACE

DISPLACEMENTS DUE TO UNDERGROUND BLASTING
D. P. BLAIR
C.S.I.R.O.,Division of Applied Geomechanics, Syndal, Victoria, Australia

SUMMARY
Elastic wave radiation from an equivalent cylindrical cavity due to an underground explosion is considered
in detail. Both static and steady state dynamic finite element models are used to investigate the surface
displacements due to the interaction of such wave radiation with both the free surface topography and
stope formation. For a blast 430 m below the surface the models predict a negligible displacement effect
due to the irregular topography but a significant effect due to the stope. A comparison between the models
and a site experiment verified the cylindrical wave nature of the problem. Due to experimental uncertain-
ties however it was difficult to clearly detect the surface displacement effect due to the large underground
open stope.

INTRODUCTION
A mass of rock affected by a blast can be broadly classified into three main regions. The inner
zone near the blast hole is a region of completely crushed rock where the blast pressure, which
can be as large as 10" Nm-2, exceeds the compressive strength of the rock. The pressure, a, in
the wavefront decreases with increasing distance, r, from the source according to the law
-
c,OC r-", where n 2 for spherical charges (see, for example, Zhukov and Kotenko') and is
somewhat less for cylindrical charges. Thus there is a distance, ra say, from the source at which
the blast stress equals the rock compressive strength. This is the outer limit of the crushed zone.
Furthermore, there is an annular region around the blast source bounded by radii r, and rb
(ra< rb) within which the stress is less than the rock compressive strength yet greater than its
tensile strength. For this annular zone radial cracking can occur (Jaeger and Cook2). The
boundary described by r = rb is often called the boundary of the 'equivalent cavity', and it is
assumed that the material outside such a cavity behaves in an elastic manner when affected by a
blast stress wave (Shave3).
There have been various attempts to employ finite element models for blasting in regions
within the equivalent cavity. To date it appears that such attempts have not been very successful.
For example, Bhandari4 used a static, linear elastic finite element model, involving pressures far
exceeding the compressive strength of the rock, to investigate the highly non-linear region near
the charge hole. Dynamic finite element models of blasting have been reported by Santich' who
used both linear elastic and viscoelastic theory to model regions close to the blast. His main
consideration was for simple one-dimensional (line element) models and so could only discuss
axisymmetrical examples. A simple two-dimensional model of bench blasting in which the blast
was modelled as a plane loading on the side of a borehole neglecting the geometric spreading of
the radiated stress wave was also included in his investigations.
However, the purpose of the present study is not to attempt to model regions within the
equivalent cavity but rather to investigate the elastic behaviour of the material affected by a blast
0363-9061/81/020097-17$01.70 Received 3 October 1979
@ 1981 by John Wiley & Sons, Ltd. Revised 18 December 1979
98 D. P. BLAIR

stress wave outside the bounds of the equivalent cavity. In this aspect, there has been quite a
significant amount of work done on the finite element dynamic modelling of two-dimensional
structures influenced by an elastic stress wave produced by a remote nuclear blast. For example,
Costantino er d 6modelled elastic wave propagation from a crater formed on the surface by a
nuclear blast and examined the interaction of such radiation with a near surface structure in a
layered medium. More recently Heller' has used a dynamic finite element method to model the
response of a large surface dam to a plane wave excitation at the dam base due to a deep
underground nuclear detonation. A dynamic finite element code to handle such problems has
been developed by Farhoomand and Wilson.'
Two-dimensional dynamic finite element modelling of elastic wave propagation has also
gained appeal in the study of ground behaviour due to an earthquake source. In many cases the
source is considered to be remote from the region of interest and hence the finite element model
is loaded with a plane wave disturbance on the boundary (Drake', McCowan er ~ 1 . " ) . Bolt and
Smith'' used this technique to produce theoretical surface seismic anomalies due to buried
structure of arbitrary shape and so they may perform a useful function in exploration seis-
mology. Smith'* produced similar anomalies due to a buried point source. His source, a
representation of a deep earthquake, consisted of a single nodal point displacement and hence
would produce both distortional and dilational waves. There were no available experimental
results to check out the model proposed by Smith.
In all models discussed, that of Bhandari4has been the only one employing a source producing
a cylindrical wavefront similar to that expected from an underground blast. A similar wave field
acting in an axisymmetric dynamic finite element model has been given by Shipley er al.13
However, their model was essentially one-dimensional and hence could not examine the effects
of the radiated stress wave on a free surface. In fact, to the author's knowledge, the dynamic
finite element modelling of a cylindrical blast wave radiating to a free surface from an equivalent
cavity underground has, so far, not been reported. This study considers, in particular, the surface
ground displacements due to such wave radiation with special reference to both surface
topography and underground structure. Since both attributes are irregular there is no analytical
solution for the displacements, hence a numerical solution is sought in finite element modelling.
Both static and steady state dynamic models are presented and experimentally recorded ground
surface displacements are compared with the dynamic solutions.
Two-dimensional plane strain models are considered in which the equivalent cavity is
assumed to be a long cylindrical hole. This is expected to be a reasonable model for regions
moderately close to the charged drill hole. At great distances from the charge it might be
expected that all charges (whether cylindrical or spherical) would appear to be point spherical
charges. However, from an experimental analysis of the charge weight scaling law, N i ~ h o l l s ' ~
found that even at great distances from a cylindrical charge the stress wave maintained its
cylindrical nature. Hence the greatest limitation of a two-dimensional model is not in assuming
cylindrical wave propagation from the equivalent cavity, but rather the two-dimensional
approximations made for both the surface topography and underground structure.
The particular topography and structure considered in this study are those of the Mt.
Charlotte gold mine in Kalgoorlie, Western Australia. However, the treatment is general
enough to be applied to any underground mining environment providing reasonable models for
both topography and underground structure can be employed.
The particular event considered in the Mt. Charlotte operations was the blasting in the E3
stope, the centre of which is 430 m below the surface. Open stoping methods were used to mine
this region and surface seismograms taken for both single and double ring blasts. The
experimental results were obtained in the latter part of 1978, the blasting in the E3 stope being
 FINITE ELEMENT MODELLING 99

completed in March 1979 with the firing of the crown and rib pillars. For each blast event
monitored, the maximum volume of broken rock had a cross section of approximately
10 m x 30 m in the plane of the two-dimensional models considered, hence an equivalent cavity
of diameter approximately 25 m was employed.

T H E STATIC MODEL
Figure 1 shows the geometry used for the initial static blast model, where a is the radius of the
equivalent cavity, d the depth of burial and x the horizontal distance.

7-
d
-

Figure 1. Geometry used for the static model of blasting

Since the surface displacements and stresses are to be determined due to an internal pressure
acting on the cavity walls it is necessary to have the fixed boundary of the model far away in
relation to the depth, d, of cavity burial. This boundary condition requires that d / R << 1; a value
of 0.017 was chosen. Horizontal displacement on the axis of symmetry was fixed at zero.
Figure 2(a) shows a comparison between the analytical solution (see, for example Savin”) and
the finite element predictions for the horizontal stress a,.The distance, x , is given in units of the
hole radius. The symmetric half plate finite element model consisted of 223 plane strain
triangular elements in a uniformly graded mesh.
Figure 2(b) shows the horizontal (U,)and vertical (U,,)displacements predicted by the finite
element method. These displacements are normalized to unit vertical displacement directly
above the cavity. From these results it is immediately obvious that although the stress a, is
negligible for x > 250 or so, the displacements are not; in fact they become negligible only for
100 D. P. BLAIR

T 1 I I

I 1 I I

lN3H33VldSIO 3 A t l V l W
 FINITE ELEMENT MODELLING 101

x > 1,000 or so. Since d = 33 units, this indicates that the fixed boundary should be placed at
least 30 hole depths from the source if correct modelling of a semi-infinite body is to be ensured.
In attempting to employ a static model of a blast hole, Bhandari4 placed fixed zero displacement
boundaries at only three hole depths from the source, where, according to Figure 2 the
displacements at the surface would still be about 50 per cent of the maximum surface
displacement, and the stresses 10 per cent or so of the maximum surface stress. Such incorrect
modelling of the boundary by Bhandari must cast serious doubt over the subsequent stress
analysis and strain-energy density calculations based on his model.
The purpose of the present large plate model is twofold. Firstly, to compare a finite element
solution of the stated problem with an analytic solution, since to the author’s knowledge, such a
comparison has not been made to date. Secondly, to determine the extent of the boundary of a
smaller, more detailed, static model incorporating both the stope geometry and the surface
topography. This latter model is required since present computer costs forbid the combination
of a large model with high detail. Such detail was achieved by employing 295 plane strain
triangular elements in the smaller model which had an area about 300 times less than the larger
model.
Figure 3 shows the surface topography in an exaggerated form of a plot of vertical height in
units of the hole depth, d, of cavity burial as a function of the horizontal distance, x . It should be
appreciated then that the maximum vertical relief is only 6 per cent of the hole depth. Figure 3
also shows a comparison of the analytical solutions with the finite element stress predictions for
the smaller model with irregular topography. Since the elements of this model are quite small,
they should accurately describe the rapid stress change for regions directly above the cavity.
Hence the small discrepancy depicted is considered to be due to the irregular topography alone.
The boundaries of this model were placed at a distance of 5.5 hole depths from the source and
given the fixed appropriate displacements predicted by the initial large model.
A simplified two-dimensional construction of the stope geometry was also incorporated in the
smaller model. This is a gross simplification of the real situation in which the stope is more like a
vertical cylinder than an infinitely long slot. A more accurate solution would be provided by a
three-dimensional finite element analysis; however, this would be prohibitive in cost if the
required fine detail of the structure were to be modelled (this would be especially so in the case of
a dynamic model). Two-dimensional models, though, can give a rough idea of the measure of
importance of the effect of the stope on surface displacements. The approximate effect of the
stope was achieved by simply assigning a Young’s modulus, E,, to the stope material and a
Young’s modulus, E, to the surrounding rock. In all models a Poisson’s ratio of 0-25 was
employed.
Figure 4 shows a plot of both the horizontal and vertical displacements, in the model with
irregular topography, for the cases of E J E = 1.0 and 0.7; the stope structure is also indicated b y
the shaded region. The plot clearly indicates that the stope has a far more pronounced effect on
surface displacements than does the topography. A detailed discussion of the properties of the
stope material is more relevant in a dynamic study of the blast stress wave and hence is treated in
the section on dynamic modelling.

MEASUREMENT OF GROUND MOTION

The equipment
A seismic refraction system (ABEM type 5352) was used to determine the amplitude of
ground velocity (and hence displacement) as a function of frequency.
 ANALYTICAL SOLUTION
FOR FLAT SURFACE
0 FINITE ELEMENT SOLUTION

1 o r

0
0
- 00
0 P
bn a
W
F
-' O r
z

1
O2t
ook .-.-
SURFACE TOPOGRAPHY
0 09,t

0.91
0'g6r
0.92l I 1 I I 1 J
-50 -100 -50 0 50 100 150
x

Figure 3. Effect of surface topography on ground surface displacements

I
 FINITE ELEMENT MODELLING 103
1 m
I
I I I 1

Figure 4. Combined effect of topography and stope formation on ground surface displacements

The geophones employed were type SM2 which had a natural frequency of 10 Hz and a coil
resistance of 385 R. A shunt resistance of 1.0 kR was placed across the coil giving the geophone
65 per cent of critical damping and a flat frequency response above 10 Hz. For this flat region,
the velocity calibration of the geophones was given as 21.7 mV per unit of velocity (1 mm sec-').
The ABEM multichannel amplifier system employed high speed galvanometer recording
onto photographic paper. The galvanometer response was flat from 10 Hz up to 200 Hz, and the
overall system produced a deflection of 1.53 cm on the photographic record when a ground
velocity of 1 mm sec-' was experienced by the geophone.

Experimental results
For each blast monitored the six-member array was laid out in a straight line along the strike
of the stope, i.e. along the surface in the plane of the two-dimensional model shown in Figure 4.
No attempt was made to obtain the exact time of blast initiation; however, in a few cases the
velocity of the first arrival (p-wave) at the surface could be estimated from the differences
observed in the first arrivals recorded at several geophones. This procedure gave two separate
estimates of 5.4 km sec-' and 6.7 km sec-' for the p-wave velocity; the relevance of these values
is discussed in the section on dynamic modelling where the rock material properties play an
important role.
Figure 5 shows a typical set of ground velocity records. For these records the maximum
deflection is about 2 cm, which yields a maximum instantaneous ground vertical velocity of
1.3 mm sec-I. In fact, for all the surveys conducted, this variable never exceeded 4.5 mm sec I .
Even though this is the magnitude for the vertical component alone, it is well below the level of

Figure 5 . Ground surface velocity records obtained using the ABEM seismic system
104 D. P. BLAIR

19 mm sec-' set by the Australian Standards Association. Preliminary experiments with the
three-component Sprengnether device indicated that the vertical component of velocity was
about the same order of magnitude as both the horizontal components, thus the maximum
instantaneous ground total velocity recorded in the present surveys is estimated at
7 . 8 mmsec-'.
It is instructive to estimate the maximum total ground velocity predicted by the empirical
charge weight scaling law (Hagan''):
ut = 1 1 4 3 ( J ~ / r ) " ~ (1)
where ut is the maximum instantaneous total ground velocity in mm sec-', w is the maximum
charge per delay in kg, and r is the distance in metres between the measurement and blast points.
-
For the present study r = 430 m and w 250 kg, hence ut = 5 . 8 mm sec-'. This equation is valid
only for delay blasting in which the delay is greater than 8 m sec. Even though the equation is
based on experimental results for surface blasting alone, it does yield reasonable order of
magnitude results for the present underground blasting situation in which an average nominal
delay time of 25 m sec was employed.
Figure 6 shows the relative amplitude spectrum of both the ground velocity and displacement
for the lowest record of Figure 5 . If u ( t ) is the measured ground velocity as a function of the time
t, and V ( f )its Fourier transform as a function of the frequencyf, then integrating u ( f ) yields the
time dependent displacement d ( t ) . However, it is easier to obtain displacements by operating in
the frequency domain since it can be readily shown that
Wf) = -iV(f)Drf (2)
where i = J-1 and D ( f ) is the amplitude spectrum of the displacement (see, for example,
B r a ~ e w e l l ' ~Equation
). (2)was used to obtain the displacements shown in Figure 6; the relative
scale for each plot is different. The spectra were obtained by discrete Fourier techniques using
the author's Fast Fourier Transform (FFT)based on the Cooley-Tukey algorithm. In sampling
the record, a Nyquist frequency, fn, of 50 Hz was employed since a preliminary Fourier analysis
revealed that there was an insignificant contribution to the ground motion above this frequency.
The Nyquist frequency is the highest frequency capable of being detected with discrete Fourier
techniques and is a function only of the interval, St, between samples. This frequency is given by
f,, = (26t)-'. In the present study 6t is then 10 m sec; since 128 sample points were used in the
investigation a total time duration of 1.28 sec of ground motion was analysed.
From Figure 6 it is clear that two main frequencies of ground motion are present, one at about
13 Hz, the other at about 21 Hz. However, not all recorded ground motion had such clear
frequency resolution, and hence the spectra for a more noisy record than those shown in Figure 5
are depicted in Figure 7. Both Figures 6 and 7 are representative of all the recorded ground
motion insofar as there is a dominant contribution to the ground motion in the frequency band
10 Hz to 24 Hz. The delay pattern of about 25 m sec employed in the blasting caused the small
contribution visible at about 40 Hz or thereabouts. The total relative displacement in the
dominant frequency band of 10 Hz to 24 Hz was then determined for each of the blast records,
and the results are presented and discussed in the section on dynamic modelling where a
comparison is made between such experiments and a dynamic model.

DYNAMIC MODELLING
The static solution discussed can, at best, give only a rough approximation to the first motion
relative response of the structure to a step function for the pressure load on the borehole. Since it
 FINITE ELEMENT MODELLING 105

i
i
W
a

FREOUENCY (Hz 1

Figure 6. Relative amplitude spectrum of both ground velocity and displacement

W
2
I-
5
W
a

FREOUENCY (Hz)
Figure 7. Relative amplitude spectrum of a reasonably noisy record
106 D. P. BLAIR

cannot yield information on the time history of the structure it cannot predict phenomena such
as resonance built-up, etc. The limitations of the static model can be overcome by use of dynamic
finite element models.
Dynamic models can be classified into two broad categories, transient and steady state. In
transient modelling, any general time-dependent forcing function, F ( t ) ,can be used to excite the
structure, hence the structure can experience the simultaneous application of forces covering a
range of frequencies. In steady state modelling F ( t )is a single frequency function specified by its
phase and amplitude.
It is not an easy matter to obtain the forcing function F ( t )that should be applied to the walls of
the equivalent cavity in order to accurately simulate the repetitive nature of delay blasting.
Furthermore, F ( t ) may have to be defined over a considerable time duration in order to detect
the possible built up of resonance of a particular structural component; the longer the time
duration of the solution, the greater the expensive use of computer time. With these limitations
in mind, it was decided to restrict the present study to steady state modelling alone in which the
frequency of the forcing function was dictated by the experimental results, i.e. in the frequency
range 10 Hz to 24 Hz. Modelling at two particular frequencies 14 Hz and 20 Hz was decided
upon.

The material properties

The two basic materials of interest in this study are the intact rock and the dry fill in the stope.
The intact rock is basically dolerite and the relevant dynamic elastic properties of such are given
by Lama and Vutukuri." They examined two sets each of 25 samples and obtained an averaged
value of 9.5x 10" Nm-2 for the dynamic Young's modulus (E), 3.8 x 10" Nm-' for the
dynamic shear modulus (G) and 3.1 x lo3kg m-' for the density p.
The shear wave velocity, V,, is given as
V ,= (G / P ) ~ ' ~ (3)
hence V, = 3.50km sec-'.
The compressional wave velocity, V,, is given as
v,= V J S (4)
where S2 = (1- 2v)/2( 1 - Y ) and v is Poisson's ratio.
If v = 0.25 then V , = 6.00 km sec-', which is within the range of the experimentally deter-
mined in situ values for the present investigations.
The dry fill is basically a mixture of sand and crushed rock, but is predominantly sand.
The basic dynamic properties of dry sand can be inferred from Richart et al.I9 as E = 2.1 x
lo9 NmP2, G = 8.5 x 10' Nm-2 and p = 2.0 x lo3 kg m-3. These are assumed to be the average
fill properties.

The finite element model

The results of the static model indicate that the effect of the stope on surface displacements is
far more significant than the effect of irregular surface topography. This negligible topographic
contribution is still expected for a dynamic model since the surface irregularities, some 20 m at
most, are a lot less than the smallest wavelengths (175 m) of interest, due to the shear waves at a
frequency of 20 Hz. Hence a flat topography was assumed for the dynamic model.
The elements used must be of a size sufficiently small to accurately model the high frequency
response of the structure. The dynamic model employed 514 elements of the 4-node quadrila-
 FINITE ELEMENT MODELLING 107

teral type. The maximum linear dimension of any element was A/4 for the smallest wavelength
( A ) propagated through the structure. According to White et al." such modelling will yield a
mean error of about 7 per cent in the solution, since the finite dimensions of each element cause
the total finite element grid to act as a low pass filter. However, analysis of the smallest
wavelength is a worst case estimate, and a best case estimate is given for the p-waves at a
frequency of 14 Hz. If this is the case, the maximum linear dimension of any element is A/8,for
which the mean solution error is less than 2 per cent. The lowest frequency response of the
structure is governed by the fact that the energy absorbing boundaries must be placed at least
A/2 from the source for the longest wavelength propagated. For the present model the closest
boundary was placed about 800 m from the source, which gave a low frequency cut off of about
4 H z . Figure 8 shows the finite element mesh employed. Both the stope and the circular
equivalent cavity are outlined; energy absorbing conditions were prescribed for all boundaries
of the model except the upper (free surface) boundary.

Figure 8. Finite element mesh used in the dynamic analysis. The equivalent cavity is placed in the centre of the grid and
the stope elements are indicated by a central dot

The method of internally loading the equivalent cavity is similar to that discussed for the static
model with the exception that here the amplitude, phase and frequency of the applied nodal
loads must be specified.

Dynamic solutions
In order to simulate an infinite body using a finite dynamic model, the unified boundary of the
standard viscous type described by White et aL2' has been used.
For the displacements, Figure 9 shows a comparison of the static solution with the lowest
frequency (4Hz) dynamic solution possible with the present finite element grid. These plots
were obtained for a uniform half space and hence only the symmetric solution is shown for
 DYNAMIC SOLUTION
- - - - - STATIC SOLUTION
0 HORIZONTAL DISPLACEMENT
0 VERTICAL DISPLACEMENT

-0.2 I I I I I
0 10 20 30 40 50 60
X

Figure 9. L o w frequency surface displacements compared with the static solutions for a uniform half-space
 FINITE ELEMENT MODELLING 109

positive values of x (the horizontal distance in units of hole radii). Even at such low frequencies
the wave nature of the dynamic solution is apparent especially for the vertical displacements,
where the horizontal distance is about one wavelength. The results have been normalized to unit
vertical displacement directly above the explosion for both static and dynamic solutions. It
should be appreciated that the plots for the dynamic case represent the amplitude of the steady
state surface oscillations (the phases, too, are output from the solution but are not shown).
The difference between the static and dynamic solutions is an indication of the important role
played by the formation of the Rayleigh wave. Directly above the explosion the surface
experiences only a vertically incident p-wave, hence there can be no Rayleigh wave contribution
at this point. However, for surface locations away from this point the cylindrical wavefront
produces both p-waves and reflected sv-waves resulting in the formation of the Rayleigh wave.
A similar situation has been analysed by Wilkins22who used a finite difference scheme to analyse
theoretical transient surface displacements caused by cylindrical waves emanating from an
underground explosion. H e theoretically considered a high energy nuclear detonation located
50 m below the surface of a uniform half space, and quite clearly illustrated the formation and
propagation of the Rayleigh pulse along the surface. For this shallow depth explosion, the
Rayleigh mode contained a large proportion of the total energy of the incident stress wave. For
the deeper source relevant to the present study, the Rayleigh wave mode is still expected to
contain a significant amount of the surface wave energy and thus significantly contribute to the
surface displacements.
Figure 10 shows the normalized vertical displacement amplitudes predicted by the uniform
half space dynamic model for both frequencies 14 Hz and 20 Hz. For these frequencies the
horizontal distance shown is about eight wavelengths or more, thus the wave nature of the
solution is quite apparent. The circled values show the total normalized displacements obtained

1.B
I I I I I 1
- - - - - FREOUENCY 14 H z
FREOUENCY 20 H I
0 EXPERIMENTAL SOLUTION I
I-
$ 1.0
I
w
0
4
P
-I

s0
w
1
I-
<
-I
w
0.9 ,---.
V,'
a

0.0 I 1 I 1 I I
4

Figure 10. Dynamic model solutions for the surface displacements at frequencies of both 14 and 20 Hz. The circled
values are the normalized experimental results
110 D. P. BLAIR

by a simple addition of the Fourier amplitudes in the frequency band 10 Hz to 24 Hz. For all
geophones this total amplitude was then normalized to the unit value chosen for the detector
located at x = 4.8 (Figure lo). The uncertainty bars are estimates in the amplitude fluctuations
due to the uncertainty in both the geophone output (*5 per cent or less) and in estimating the
Fourier amplitude. The latter uncertainty is quite high in some cases and is due to the fact that a
few of the more noisy records had some amplitude contribution outside the frequency band of
interest (this is clearly illustrated in Figure 7) and the ratio of such contribution to the total
Fourier amplitude is defined, here, as the uncertainty.
The approximate effect of the stope was then achieved by assigning elastic properties for the
stope material slightly different to those for the intact rock, in asimilar manner to that used in the
static model. Figure 11shows the approximate effect of the stope for the parameters E J E = 0.7
and GJG = 0.7, where the subscript refers to the stope dynamic elastic property. It was decided
to keep the stope density the same as that of the intact work since a decrease in both E, and
G,would produce the desired decrease in both the p- and s-wave velocities as is expected for
dry sand.
Figure 11 clearly illustrates the significant effect of the stope under dynamic loading.
Although the static model, too, predicted a significant effect due to the stope it should be noted
that the actual magnitude of the present dynamic solution is quite different to that of the less
realistic static solution.
Unlike the uniform half space, under dynamic loading, the presence of the stope now allows
for the possibility of Rayleigh waves at any point on the surface including that directly above the
blast. For a frequency of 20 Hz, the approximate model predicts high displacement amplitudes
on the hanging wall side of the stope. It is possible that the measured high displacements in this
region could be, in part, due to the effect of the stope. However, in view of the experimental
uncertainties involved, it is difficult to clearly detect the stope effects on the measured
amplitudes for the vertical motion.

DISCUSSION
The reasonable agreement between theory and experiment, as indicated by Figures 10 and 11,
reinforces the justification for using cylindrical wave models to represent the blast. For example,
if the measured surface displacements were due to a buried spherical source then they would be
expected to decay, with distance x, at a faster rate than that indicated in Figure 10. The
experimental results thus justify the use of the square root charge weight scaling law given by
equation (1) since it is applicable to cylindrical waves (the spherical source problem is governed
by a cube root charge weight scaling law, i.e. one involving "w).
The experimental justification of this law for the particular underground blasting operations
at Mt. Charlotte has significant practical implications. For example, in the firing of the rib and
crown pillars of the E3 stope a maximum charge weight of 908 kg per delay was employed. Even
though this blast was not monitored, equation (1) can be used to estimate a maximum total
ground velocity, ut, of 16.3 mm sec-'.
In the role of mine planning for future blasts, the equation can be used to predict the increase
in charge per delay that can be employed for firing on a deeper level while maintaining constant
surface ground velocities. Of direct relevance to the mine is the firing in the proposed F stope
development, whose centre is 77.5 m below that of the E3 stope. If r of equation (1)is increased
to 507.5 m and the maximum surface velocity, ut, maintained at 5 - 8 mm sec 1 then the equation
predicts that the maximum charge weight per delay can be increased from 250 kg to 350 kg.
Alternatively expressed, for blasting in the proposed F stope a maximum charge weight of less
 I I I I I
.FF?EOUENCY 14 Hz
----- F R E Q U E N C Y 20 H z

/I-. \
I \

\
8

\
\
\
\

Figure 1 1. Stope effect on surface displacements due to dynamic loading. The stope location is also shown
112 D . P. BLAIR

than 1530 kg per delay would not violate the safety level imposed on the total surface ground
velocity by the Australian Standards Association.

CONCLUSIONS
Experimental measurements of ground surface vibrations were obtained and compared with the
predictions of two-dimensional finite element models, yielding reasonable agreement between
theory and experiment. This led to significant practical implications for the particular mine
environment considered. However, the blasting vibration models proposed are general enough
to be applied to any situation involving underground blasting providing that reasonable
two-dimensional models can be employed for both the topography and structure. In this aspect
the models could be readily utilized to determine, say, the vibrations set up in a tall surface
structure due to underground blasting.
Only static and steady state dynamic models are considered, and thus any spectral contribu-
tion related to the delay time of the blasting pattern could not be directly modelled. In this
particular investigation the delay time was about 25 m sec, hence the spectral contribution
of 40 Hz could not be studied. The only complete method of modelling such a situation
would be to employ a transient dynamic model. One method of solution would involve
obtaining the structural response transfer function by use of a Dirac delta input forcing function.
The structural response to any desired input forcing function could then be obtained from the
transfer function. However, owing to the infinitely high frequencies involved in the delta
function, the low pass filtering action of even the finest finite element mesh could render the
results suspect. An alternative method of solving the transient problem is to employ a lower
frequency forcing function representing the blast stress wave action on the interior walls of the
equivalent cavity. Some estimate of the pressure pulse near the source of a single hole blast has
been obtained by several researchers (see, for example, Santich’). The change of shape of this
pulse as it approaches the walls of the equivalent cavity and the overall effect of delay blasting
would be more difficult to assess. However, such a model would be capable of examining the
effects of structural resonances as a function of the delay time between blasts and hence predict
what value of delay time between blast holes should be employed in order to minimize
underground and surface vibrations. Since this has significant practical applications, work on
such a model is to begin in the near future.

ACKNOWLEDGEMENTS

The technical services of J. Foxton, B. Verhoef and B. Eaton in conducting field experiments
and the assistance given by G. Blackford in data analysis are greatly appreciated.
The author gratefully acknowledges the invaluable discussions he had with W. White, S.
Valliappan, A. Ross and other members of the Department of Civil Engineering Materials,
University of New South Wales, as well as use of their finite element programs for the dynamic
analysis.
The author is grateful for the encouragement given by 0. Jones and B. White, the Dean and
Head of Mining Engineering, respectively, of the School of Mining and Mineral Technology,
Kalgoorlie.
The author is also indebted to Kalgoorlie Lake View Pty. Ltd. for the freedom to research a
practical mining problem on their site and permission to publish this paper.
 FINITE ELEMENT MODELLING 113

REFERENCES
1. V. V. Zhukov ana V. F. Kotenko, ‘Deformation of solid rock when a concentrated charge is exploded, with
allowance for the screening surface’, Fiziko-Teknicheskie Problemy Razrabotki Poleznykh Iskopaemykh, 1, 33-38
( 1974).
2. J. C. Jaeger and N..G. W. Cook, Fundamentals of Rock Mechanics, 2nd edition, Chapman and Hall, 1976.
3. J. A. Sharpe, ‘The production of elastic waves by explosion pressures. I. Theory and empirical field observations’,
Giophysics, 3, 144-165 (1942).
4. S. Bhandari, ‘Studies on rock fragmentation in blasting’, Ph.D. thesis, University of New South Wales (1975).
5. J. R. Santich, ‘Dynamic models of rock blasting’, Ph.D. thesis, University of New South Wales (1976).
6. C. J. Costantino. A. Wachowski and U. L. Barnwell, ‘Finite element solution for wave propagation in layeied media
caused by a nuclear detonation’, Proc. Int. Symp. Wave Propagation and Dynamic Properties of Earth Materials,
Univ. of New Mexico, 59-70 (1967).
7. L. W. Heller, ‘Finite element analysis of the response of Rifle Gap dam to the effects of Rulison underground
detonation’, Proc. Symp. U S . Army Engineers Waterways Experiment Station, Vicksburg (1972).
8. I. Farhoomand and E. Wilson, ‘A non-linear finite element code for analysing the blast response of underground
structures.’ U.S. Army Engineers Waterways Experiment Station, Vicksburg C . R . N-70-1 (1970).
9. L. Drake, ‘Rayleigh waves at a continental boundary by the finite element method’, Bull. Seis. SOC.A m . 62,
1259-1268 (1972).
10. D. W. McCowan, P. Glover and S. S. Alexander, ‘A static and dynamic finite element analysis of the 1971 San
Fernando, California, earthquake’, Geophys. J.R. Astr. SOC.48, 163-1 85 (1977).
11. B. A. Bolt and W. D. Smith, ‘Finite element computation of seismic anomalies for bodies of arbitrary shape’,
Giophysics, 41, 145-150 (1976).
12. W. D. Smith, ‘The application of finite element analysis to body wave propagation problems’, Geophys. J.R. Astr.
SOC.42,747-768 (1975).
13. S. A. Shipley, H. G. Leistner and R. F. Jones, ’Elastic wave propagation-a comparison between finite element
predictions and exact solutions’, Proc. Int. Symp. Wave Propagation
. - and Dynamic Properties of Earth Materials,
Univ. of New Mexico, 509-519 (1967).
14. H. R. Nicholls, ‘Blasting vibrations and their effects on structures’, U S . Dept. of Interior Bureau ofMines Bull. 656.
15. G . N. Savin, Stress Concentration around Holes, Pergamon Press, 1961.
16. T. N. Hagen, private communication.
17. R. M. Bracewell, The Fourier Transform and its Applications, McGraw-Hill, 1965.
18. R. D. Lama and V. S. Vutukuri, Handbook on Mechanical Properties of Rocks-Vol. 2, Testing Techniques and
Results, Trans. Tech. Publications, 1978.
19. F. E. Richart, J. R. Hall and R. D. Woods, Vibrations of Soils and Foundations. Englewood Cliffs, N.J., Prentice
Hall, 1970.
20. W. White, S. Valliappan and 1. K. Lee, ‘Finite element mesh constraints for wave propagation problems’, Proc. 3rd
Int. Conf. on Finite Element Methods, Univ. N.S.W., Sydney, Australia (1979).
21. W. White, S. Valliappan and I. K. Lee, ‘A unified boundary for finite dynamic models’, J. Engng Mech. Din, Proc.
A m . SOC.Civ. Eng. 103,949-964 (1977).
22. M. L. Wilkins, ‘Calculation of surface and ground waves from above-ground and underground explosions’,
Astronautica Acra, 17, 399-404 (1972).