This article was downloaded by: [The UC Irvine Libraries]

On: 25 October 2014, At: 12:52

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954
Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Fragblast: International Journal

for Blasting and Fragmentation
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/nfgb20

Charge Weight Scaling Laws

and the Superposition of Blast
Vibration Waves
a
D.P. Blair
a
Orica Australia Pty. Ltd, George Booth Drive , Kurri
Kurri, NSW 2327, Australia
Published online: 02 Feb 2010.

To cite this article: D.P. Blair (2004) Charge Weight Scaling Laws and the Superposition
of Blast Vibration Waves, Fragblast: International Journal for Blasting and Fragmentation,
8:4, 221-239, DOI: 10.1080/13855140412331291610

To link to this article: http://dx.doi.org/10.1080/13855140412331291610

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information
(the “Content”) contained in the publications on our platform. However, Taylor
& Francis, our agents, and our licensors make no representations or warranties
whatsoever as to the accuracy, completeness, or suitability for any purpose
of the Content. Any opinions and views expressed in this publication are the
opinions and views of the authors, and are not the views of or endorsed by
Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and
Francis shall not be liable for any losses, actions, claims, proceedings, demands,
costs, expenses, damages, and other liabilities whatsoever or howsoever caused
arising directly or indirectly in connection with, in relation to or arising out of the
use of the Content.

This article may be used for research, teaching, and private study purposes.
Any substantial or systematic reproduction, redistribution, reselling, loan, sub-
licensing, systematic supply, or distribution in any form to anyone is expressly
 forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Downloaded by [The UC Irvine Libraries] at 12:52 25 October 2014
 Fragblast
2004, Vol. 8, No. 4, pp. 221–239

Charge Weight Scaling Laws

and the Superposition of Blast Vibration Waves

D.P. BLAIR1

ABSTRACT
Downloaded by [The UC Irvine Libraries] at 12:52 25 October 2014

A detailed discussion is given on the relationship between the vibration waves from a single blasthole and
the vibration waves from a full-scale blast. The relationship is not necessarily simple. In this regard, the full-
scale blast may be considered as a linear or non-linear superposition of blast vibration waves from
representative single blastholes. If linear superposition is assumed, then standard Monte Carlo techniques
may be used to predict the complete vibration time history due to the entire blast. However, for non-linear
superposition the problem is much more complex, and in many aspects remains unsolved. It is also shown
that the traditional charge weight scaling laws used to predict the peak vibration levels of a blast are plagued
with many problems. Two problems of particular concern are that such laws are dimensionally awkward, if
not unsound, and they are also inconsistent with any notion of superposition. In light of the superposition
problem associated with traditional charge weight scaling, a new, weighted scaled distance is proposed for
use in vibration prediction of full-scaled blasts. Although this weighted scaled distance remains dimen-
sionally awkward, it is more consistent with the notion of waveform superposition since it takes account of
all the blastholes within a blast and also all the distances from each hole to a specified monitoring location.
The current and ad-hoc method for selecting a single value of charge weight and distance to calculate a
scaled distance for the entire blast makes no sense, especially for monitoring locations close to a blast. Use
of the weighted scaled distance completely avoids this problem, and when it replaces the traditional scaled
distance, less scatter is observed in the charge weight scaling plots.

Keywords: blast vibration, charge weight scaling, waveform superposition.

1. INTRODUCTION

Predicting the vibration wave due to just a single blasthole remains an unsolved and
very difficult problem, even if that blasthole is perfectly contained and the geology
completely uniform between the blasthole and monitor location. One reason for the
difficult nature of this problem is that the material close to the explosive responds in a
highly non-linear manner, and its non-linearity is not well understood. In other words,
although the total chemical energy of the explosive may be known, the partitioning of
this energy into forms such as material deformation, heat and wave propagation is not.

1
Address correspondence to: D.P. Blair, Orica Australia Pty. Ltd, George Booth Drive, Kurri Kurri, NSW
2327, Australia. Tel.: þ61-24939-5288; Fax: þ61-24939-5299; E-mail: [email protected]

10.1080/13855140412331291610$22.00 # 2004 Taylor & Francis Ltd.

 222 D.P. BLAIR

In realistic geology, the problem of waveform prediction is compounded by the

fact that even a single, undetected crack somewhere in the wave path from the
blasthole to the monitor can completely distort the received waveform. In this regard,
it might be considered that predicting the blast vibration wave from even a single
blasthole will remain an insoluble problem. Evidence of this is shown by the fact that
a transverse vibration component is always seen for the monitoring of single
blastholes. However, in uniform material, symmetry demands that only a radial and
vertical component should exist. There is currently no way to predict the observed
transverse component, and in fact it has some hallmarks of randomness [1].
Due to this difficulty of vibration prediction, it is natural to look for simple
alternatives of estimating some property of the complete vibration wave such as its
Downloaded by [The UC Irvine Libraries] at 12:52 25 October 2014

peak value for the separate components or for the vector sum. In this regard, for a
single blasthole it is reasonable to expect that the peak vibration will increase as the
charge weight, W, increases. It is also reasonable to expect that the peak vibration will
decrease as the distance, d, between blasthole and monitor increases. And this is the
basis of simple charge weight scaling. Thus for a single blasthole, in light of the noted
difficulties, we may have no alternative but to assume that the vector peak particle
velocity (vppv) can be estimated by the following typical charge weight scaling law:

vppv ¼ aðSDÞb ð1Þ

where a and b are adjustable parameters, dependent upon the local ground conditions,
and the scaled distance, SD, is defined as:
d
SD ¼ pffiffiffiffiffi ð2Þ
W
However, even with the simple Equation (1) there is a fundamental problem, and it is
associated with its dimensions. Considering the fundamental units of mass (M),
length (L) and time (T), then both sides of Equation (1) must have the dimension of a
velocity, i.e. a dimension LT1. Although it makes no physical sense for the constant b
to have a dimension, the scaled distance, SD, must always have a dimension LM1=2 ,
and so the constant a must always have a dimension. As an example, if b ¼ 2, then a
must have dimensions M1 L3T1. However, if b ¼ 1, then a must have dimensions
M1=2 L2T1. In fact the dimensions of the site constant a continually change,
depending upon the numerical value of b that is usually found (along with a) by curve-
fitting the observed data. This is quite unsatisfactory from a physics viewpoint, and
clearly demonstrates the awkward, if not unsound, nature of Equation (1).
Perhaps the best way to view Equation (1) is that it provides an assessment of the
numerical value, only, of the peak vibration level and only from a single blasthole,
given the charge weight and distance. In a strict sense, this law should not be
construed as a fundamental equation for vibration prediction, despite its popular
 CHARGE WEIGHT SCALING LAWS 223

exposition given in [2], and its various modifications given in [3, 4]. Furthermore, it is
highly questionable to use such laws as a foundation for any near-field vibration
model as was done by Holmberg and Persson [5], whose model was subsequently
shown to be invalid [6]. It has also been shown [7] that the charge weight scaling laws
are not consistent with the accepted theories of wave attenuation.
The main aim of the present work is to discuss the role of charge weight scaling for
a single blasthole and show how, despite the difficulties, it may be used within a viable
superposition model for the prediction of blast vibration waveforms from full-scale
blasts. Another aim of this work is to show that it is not meaningful to use the current
charge weight scaling method of selecting a single value of charge weight and
distance to predict the vibration from any full-scale blast. Unfortunately, much of the
Downloaded by [The UC Irvine Libraries] at 12:52 25 October 2014

blast vibration community continues to use this questionable predictive technique. It

is for this reason, that an alternative, more scientifically sound, technique for scaling is
presented that gives various weightings to all blastholes and all their distances to a
specific monitoring location.
A method is also given for a more rational estimation of the charge weight that is
likely to have caused the peak vibration observed on any given vibration waveform.
However, this new method is not a vibration predictive technique because it relies
upon analysis of the experimental data, in particular by measuring the arrival time
region of the peak vibration level of the blast.

2. THE SUPERPOSITION OF WAVEFORMS

FROM SINGLE BLASTHOLES

As stated earlier, there is, as yet, no model capable of predicting the waveform from
even a single blasthole. However, if we are prepared to measure the vibration from a
single blasthole (or a collection of separate blastholes) then it is possible to construct a
viable model for the vibration waves produced by a full-scale blast under the
assumption of linear superposition. The complete description of this model is given in
[1] and will not be repeated here except for that detail relevant to the present work. In
this aspect, it is very pertinent to note that, despite the inherent problems of Equation
(1), it is still used in the model. However, this equation is only used to numerically
scale the seed waveform for each blasthole such that it has the expected vppv for a
blasthole of given charge weight, W, and distance, d, to the monitor. In other words,
the waveform representing each blasthole in the full-scale blast has units of velocity
(i.e. units of LT1 ). Thus the physics of the model remains valid since a linear
superposition of such waveforms obviously also has the units of velocity.
However, it is also possible for the vibration waves from each blasthole to super-
pose in a non-linear fashion. Aspects of non-linearity have been given previously in
[7] but will now be expanded somewhat. One such aspect of non-linear superposition
 224 D.P. BLAIR
Downloaded by [The UC Irvine Libraries] at 12:52 25 October 2014

Fig. 1. The separation of identically charged blastholes used to illustrate linear and non-linear superposition.

can be illustrated by considering the total vibration due to identically charged pairs of
blastholes whose separation, S, changes while both remain an equal distance, d, from
a monitor as shown in Figure 1. It is also assumed that the distances from each
blasthole to the monitor are equal, i.e. S is a lot less than d.
Since blastholes 1 and 2 are a relatively large distance apart, the vibration waves
from each will add in a linear manner at the monitor, and this is the basis of typical
Monte Carlo vibration models. However, as the holes get closer together (3, 4) they
will interact in a non-linear fashion, with each hole influencing the radiation of its
neighbour. In the limit of virtual hole overlap (5, 6), the total vibration waveform is
not the linear sum of two separate waveforms, but rather the single vibration
waveform due to one blasthole having twice the charge weight. Thus the charge
weight scaling law itself implies a non-linearity, even for uniform ground in which the
site constants a and b of Equation (1) remain unaltered.
For example, if vppvi,j is defined as the monitored value of vppv due to the
instantaneous initiation of holes i and j in identical ground, then the charge weight
scaling law implies that:
 
d b
vppv1;2 ¼ 2a pffiffiffiffiffi ;
W
  ð3Þ
d b
vppv5;6 ¼ a pffiffiffiffiffiffiffi
2W
If Nvppv is the vppv normalised to that expected for a single blasthole, then
Nvppv1,2 ¼ 2 and Nvppv5,6 ¼ 2b/2. These two vppv values are only identical if b ¼ 2.
 CHARGE WEIGHT SCALING LAWS 225

For b < 2, vppv1,2 > Nvppv5,6, and for b > 2, vppv1,2 < Nvppv5,6. Unfortunately, the
allocation of charge weights for use in the scaling law (1) only makes sense in these
two extreme cases (i.e. holes widely separated, and holes overlapping) and gives no
insight for the case of instantaneous initiation of the close-neighbour holes 3 and 4.
However, in the spirit of charge weight scaling, it is reasonable to assume that the
vibration in this case should lie somewhere between the two extremes. It is a trivial
exercise to extend this discussion to the case of simultaneous initiation of blastholes
having unequal charge weights, and this is left to the interested reader.
If the two holes are now initiated with a time delay greater than the waveform
duration of each, then the vppv will be determined by the charge weight law (1), and
will be the same as that produced by one hole alone if they have equal charge weights.
Downloaded by [The UC Irvine Libraries] at 12:52 25 October 2014

If the charge weights are unequal, then the vppv is obviously determined by that hole
with the greater charge weight. However, if the time delay is less than the waveform
duration (but not zero), then the waveforms from each blasthole will overlap (and thus
superpose, either linearly or non-linearly) and then equations such as (3) cannot be
used to estimate the vppv due to both holes.
In this regard, an interesting vibration application arises for the case of a single
column of explosive compared with multiple explosive decks, even assuming linear
superposition. The full column can be considered as a series of elemental charge
weights distributed continuously along the column, each time delayed according to
the velocity of detonation, VoD. Indeed, this is precisely the basis of the Dynamic
Finite Element Model (DFEM) and the analytical model used in [6] to determine
the vibration from a cylindrical column of explosive under the assumption of linear
superposition. Because of the continually varying distance from each borehole
element to the monitor, there is a continual superposition of elemental waveforms
reaching the monitor. It is very important to note that this superposition must allow for
partial cancellation as well as reinforcement of elemental waveforms to form the total
monitored vibration due to an extended column of charge. One consequence of the
partial cancellation is observed in both the analytical model and DFEM which show
that, dependent upon the VoD, the total vibration may be due primarily to the end
portions of the charge column rather than the central portions. With regard to the
waveform produced by each single small deck, this partial cancellation could be less
because of the shorter length of each deck, but there is now an extra superposition of
each separate deck waveform to produce the total vibration.
The result of all this linear superposition complexity is that replacing a full column
of explosive by a series of decks does not always guarantee a reduction in blast
vibration, even though the total mass of the decks is less than the full column mass. It
is possible that the total vibration due to decked charges may be greater than the
vibration due to a full column, depending upon the rock properties (p-wave in
particular), the explosive VoD, the length of each deck, and its spatial and temporal
separation from neighbouring decks. This state of affairs has been observed by the
 226 D.P. BLAIR

author and his colleagues in modelling exercises and some limited field trials of
decked and undecked columns in single blastholes (unpublished data).
The non-linearity implied by the charge weight scaling law as the blasthole
separation changes, and discussed in connection with Figure 1, can be incorporated
within any Monte Carlo scheme. All that is required is a curve describing the non-
linear variation of peak level with blasthole separation, S., and this level can then be
used to re-scale the vibration waveform from each blasthole. However, the exact non-
linear form of this function remains as yet unknown, and will obviously be related to
the ground plasticity. Another non-linear aspect of vibration is related to the changing
ground conditions in space and time over the region of a blast. This aspect is now
considered.
Downloaded by [The UC Irvine Libraries] at 12:52 25 October 2014

3. LIMITATIONS OF LINEAR SUPERPOSITION

The main limitation with a standard implementation of the Monte Carlo scheme is the
assumption of linear superposition of each blasthole waveform in which the site
constants a and b are obtained from isolated blastholes fired in virgin ground. Such
blastholes are expected to produce an abnormally large vibration when compared to
holes fired in damaged ground. Experimental evidence for this phenomenon has been
reported in [8].
There are two aspects of ground damage under blasting, both of which effectively
reduce the vibration and also imply a non-linearity with regard to superposing the
vibrations from each blasthole to form a total. Firstly, each hole is fired in ground with
a spatial variation in damage that is due to nearby previous blasts. Secondly, during
any one blast the ground deteriorates further due to dynamic alteration of the material
as the blast proceeds. With regard to the first aspect, it is always best to obtain single
blasthole (seed) information in the most representative ground conditions, i.e. holes
fired to a working free face, or in any region close to a previous blast. However, this is
often not possible. With regard to the second aspect, the dynamic alteration during the
blast is not understood, and cannot, as yet, be successfully modelled.
It is because of these two aspects that any linear superposition model (such as that
described in [1]) will invariably overestimate the measured ground vibration. Indeed,
it is the author’s experience that such models can overestimate the true vibration
by a factor of approximately 3 if the seed information is obtained for well-spaced
blastholes in virgin ground. This overestimation is reduced to about 1.5 if the seeds
are obtained in damaged ground (such as single holes fired to a free working face).
However, such an overestimation in ground vibration is not necessarily a severe
disadvantage, as it does, at least, provide a conservative estimate. Furthermore, unlike
the alternatives (such as charge weight scaling) the Monte Carlo superposition model
does provide a means of predicting vibration as a function of delay sequence and its
 CHARGE WEIGHT SCALING LAWS 227

scatter and so opens up the possibility of blast design optimisation for the reduction of
ground vibration.
Nevertheless, the predictive power of any superposition scheme ultimately relies
on a rational selection of the site parameters a and b. In a strict sense, the foregoing
discussion on non-linearity implies that these parameters are not constant for each
blasthole. Although it is not possible to determine how these parameters may vary, it
is possible to estimate their average values for ground during its dynamic alteration.
This can be achieved by the near-field monitoring of a blast in which blastholes
initiate, unimpeded, in a direction towards the monitor, and result in a vibration that
increases significantly with time as blasthole scaled distances generally decrease. The
mean vibration envelope as a function of time [8] will then increase smoothly and
Downloaded by [The UC Irvine Libraries] at 12:52 25 October 2014

quite unambiguously, and encoded in this increase is the information on average

dynamic values of the parameters a and b. A Monte Carlo model of this same blast can
then be run for various input values of a and b until a best fit (in some sense) is
obtained between the observed envelope and the modelled envelope.
In this regard, Figure 2 shows the vppv measured for single blastholes fired in
regions close to a planned open pit blast. There is significant scatter in this data and it
is probably due to the variable damage induced by previous blasts. A least squares
power curve fit to this data yielded a ¼ 247 and b ¼ 0.53. It is interesting to note that
typical values of the parameter b generally lie in the region 1 < b < 2 for single
blastholes, and so the present value may have a considerable uncertainty in its
estimation which is not surprising in view of the scatter.

Fig. 2. Power curve fit to single hole data, yielding a ¼ 247, b ¼ 0.53.
 228 D.P. BLAIR
Downloaded by [The UC Irvine Libraries] at 12:52 25 October 2014

Fig. 3. The observed mean vibration envelope compared with various Monte Carlo predictions.

Figure 3 shows the mean vibration envelope function calculated from the observed
data of 9 near-field monitors. The experimental and analysis details are given
elsewhere [8]. This figure also shows a Monte Carlo model of an identical experi-
mental arrangement to simulate the mean vibration envelope. It is quite obvious that
using the values of site parameters a and b as measured from the single blastholes
considerably overestimates the true vibration. These parameters, alone, were then
varied in the model, and the results also shown. This figure also clearly illustrates why
it is possible to obtain realistic values of the parameters uniquely, since a mainly
controls the amplitude and b mainly controls the shape. The values a ¼ 336 and
b ¼ 2.0 are considered to be representative of the average behaviour of the single
blastholes throughout this blast.
Thus the Monte Carlo model can be refined and calibrated for the particular
mine region of interest. Of course, if predictions are required for another region of
different geology then the process must be repeated, possibly including the firing
of single blastholes to get coarse estimates of a and b, and then near-field moni-
toring of a blast to get refined estimates of these parameters. With regard to site
calibration it is very important to realise that the Monte Carlo superposition model,
perhaps more so than most models, is subject to the GIGO principle (garbage in –
garbage out). The use of inappropriate seed waveforms, or unrealistic estimates
of a and b, or, most importantly, the extrapolation of results outside the range
of distances for which the seeds were collected, will probably invoke the GIGO
principle.
 CHARGE WEIGHT SCALING LAWS 229

4. A WEIGHTED SCALED DISTANCE

It is not always possible to measure the vibrations from a single blasthole, which then
precludes the use of Monte Carlo predictions for blast vibration. In this regard, mines
are often wary of firing a single hole for fear that it might unduly disturb the surrounding
ground. This could be especially so for ring blasting in underground operations. In this
case the only data that can be collected are the vibration waveforms due to the entire
blast. Nevertheless, it is still of significant practical importance to plot the peak
vibration levels at least, and in some meaningful way. However, as noted earlier, it is not
entirely clear how this data is to be scaled, i.e. which charge weight and distance should
be used to calculate an appropriate scaled distance for the entire blast.
Downloaded by [The UC Irvine Libraries] at 12:52 25 October 2014

This uncertainty has given rise to some ad-hoc methods such as using that distance
from the monitor to the centre (or centroid) of the blast and using the maximum
charge weight in any one hole, or the maximum charge weight lying within an 8 ms
time window. However, these traditional methods are questionable, particularly in the
near-field of a blast which is that region especially pertinent to studies of blast
damage. In this region it is very likely that the blasthole, or group of blastholes,
having the smallest scaled distance will primarily determine the blast vibration. Thus
it makes little sense to choose a separate distance based upon the blast centre, or a
maximum charge weight that may be in a hole far removed from the monitor.
There are two approaches that can be used to make a more rational estimate of
charge weights and distances that govern the peak vibration. The first approach is not
a vibration predictive technique because it relies upon analysis of the experimental
data. In this approach, a time–frequency distribution is calculated from the triaxial
vibration components, and the mean frequency estimated from this distribution. The
details of this estimation are given in [9]. A mean period is then defined as the inverse
of the mean frequency, and its value calculated for the time of arrival of the peak
vibration, vppv. The blastholes initiating within this period of time are then
considered to have produced the peak level. Their combined charge weight and mean
distance to the monitor could then be taken as estimates for the values of W and d to be
used in Equation (1) to calculate a scaled distance for the blast.
The second approach is a predictive technique and involves an alteration of
Equation (1) to incorporate the charge weights of all blastholes and the distances from
all holes to each monitor. An obvious solution is to classify each blast-monitor case by
an average scaled distance, SDA, say, that is defined as:
1X N
dj
SDA ¼ pffiffiffiffiffiffi ð4Þ
N j¼1 WJ
where N is the number of blastholes, Wj is the charge weight of the jth blasthole, and dj its
distance to the monitor. The advantage of using a sum of blastholes is that it reduces the
influence due to fluctuations in charge weight, and also implies that the peak vibration is
 230 D.P. BLAIR

invariably due to a number of blastholes. However, the average scaled distance, SDA, is
not an ideal measure since blastholes at large distances and/or having small charge
weights will not significantly influence the vppv yet will significantly influence
(increase) the value of SDA. At the other extreme, in the near field of a blast, it is also
possible to use the minimum scaled distance, SDM, say, which is simply defined as:
 
dj
SDM ¼ min pffiffiffiffiffiffi ð5Þ
WJ
where min implies the minimum value of the function taken over all j holes within a blast.
However, SDM is not necessarily a good measure either, as it depends only on that single
blasthole which will, in an ideal sense, produce the largest vibration. Thus SDM might
Downloaded by [The UC Irvine Libraries] at 12:52 25 October 2014

not be a reasonable parameter to use in realistic cases where any number of blastholes
could contribute to the peak. Furthermore, geological variations could also result in the
vppv being due to a blasthole that might not have the minimum scaled distance. In this
regard it seems better to use a scaled distance that involves a number of blastholes.
Equations (4) and (5) are members of a more general class of weighted scaled
distance, SDk, say, defined as:
2 pffiffiffiffiffiffi!k 31=k
1 XN
Wj
SDk ¼ 4 5 ð6Þ
N j¼1 dj

for non-zero integer values of k  1. The average scaled distance, SDA, is given by
k ¼ 1, and the minimum scaled distance, SDM, is the limit of SDk for large k.

Fig. 4. Log of vppv vs log of traditional scaled distance.

 CHARGE WEIGHT SCALING LAWS 231
Downloaded by [The UC Irvine Libraries] at 12:52 25 October 2014

Fig. 5. Log of vppv vs log of average scaled distance.

It is important to note that Equations (4), (5) and (6)

pall
ffiffiffiffiffiffihave the same dimensions.
However, SDk is strongly influenced by the value of Wj =dj whereas SDA and SDM
are strongly influenced by the inverse of this parameter.
Figure 4 shows the results for 17 coal overburden shots, each monitored at various
locations, and scaled according to the traditional method. In particular, the distance, d,
of Equation (1) is given by that between the monitor and the centre of the blast
pattern, and the charge weight, W, is given by the maximum charge weight initiating
in an 8 ms time window. Figure 5 shows the same vppv data plotted using the average
scaled distance (Eq. (4)).
Figure 6 shows the data plotted using the weighted scaled distance (Eq. (6), with
k ¼ 2) and Figure 7 shows the data plotted using the minimum scaled distance (Eq. (5)).
In each case the scatter can be calculated as the residual sum of squares, i.e. the
sum of squares of the deviations from each regression line. These Residual Sums of
Squares (RSS) can then be normalised to the RSS resulting from the use of the
traditional scaled distance. Figure 8 shows the normalised RSS for the average,
weighted and minimum scaled distance techniques applied to this coal mine data.
Each of the weighted scaled distance measures is classified by the integer k of
Equation (6). The normalised RSS for the average scaled distance is given by the open
circle placed at the k ¼ 1 location for convenience only. In this particular case there
was an insignificant difference in the scatter when calculated using the weighted
scaled distance with k ¼ 1 and the average scaled distance, and both these measures
provided the least amount of scatter. There was also an insignificant difference
 232 D.P. BLAIR
Downloaded by [The UC Irvine Libraries] at 12:52 25 October 2014

Fig. 6. Log of vppv vs log of weighted (k ¼ 2) scaled distance.

Fig. 7. Log of vppv vs log of minimum scaled distance.

between the normalised RSS of the minimum scaled distance and that of the weighted
scaled distance with k ¼ 300. This is shown as the last point on the plot. Figure 8 thus
shows that use of either the average or weighted scaled distance for these observed
 CHARGE WEIGHT SCALING LAWS 233
Downloaded by [The UC Irvine Libraries] at 12:52 25 October 2014

Fig. 8. Normalised residual sum of squares for coal mine data.

vibration results produces less scatter than that resulting from use of the traditional
scaled distance for which the normalised RSS ¼ 1.0.
For this particular coal mine data set, the smallest monitor to blasthole distance
was approximately 100 m, and this does not constitute the very-near-field of any blast.
Since the very-near-field is of particular practical interest for blast damage assess-
ment, it is worthwhile investigating this region using the various scaled distance
measures. Unfortunately, we do not have the required detail for the blast design
parameters (such as hole coordinates, delay sequence etc) for much of the near-field
monitoring data that has been collected. Nevertheless, it is still instructive to examine
a practical design of a blast pattern for the small-scale shot shown in Figure 9.

Fig. 9. Blast design used for the small-scale shot.

 234 D.P. BLAIR
Downloaded by [The UC Irvine Libraries] at 12:52 25 October 2014

Fig. 10a. Log random distribution of monitors.

Fig. 10b. Log Sobol distribution of monitors.

Figures 10a and 10b show 10000 hypothetical monitor locations scattered in a
zone around the blast. Since the results are to be analysed by a linear regression on
log(vppv) versus log (scaled distance) it is more efficient to have monitor locations
 CHARGE WEIGHT SCALING LAWS 235

distributed logarithmically with distance. In Figure 10a the coordinates of each

monitor are first generated randomly and then distributed logarithmically with dis-
tance from the blast centre. However, it is clear that such locations are also clustered
and to avoid this it is better to use locations whose coordinates are first generated from
the sub-random Sobol number sequence [10] and then logarithmically distributed.
Figure 10b shows the results for this latter case.
The remaining results are shown for those monitor locations given in Figure 10b,
since they obviously cover the monitoring region more uniformly than those
shown in Figure 10a. For each monitoring location in Figure 10b, the Monte Carlo
model was run for one simulation only in which delay scatter was ignored and each
blasthole was assumed to emit identical radial and vertical seed waveforms, but had
Downloaded by [The UC Irvine Libraries] at 12:52 25 October 2014

a randomised transverse component as demanded by reality [1]. This particular

arrangement is termed Model 1 (see Fig. 14 later). Figure 11 shows the vppv
predicted by this superposition model and plotted using the traditional scaled
distance, in which W is given by either the total charge weight initiating in an 8 ms
time window or the maximum charge weight (MIC) in any single blasthole. Figure
12 shows the results plotted using the weighted scaled distance with k ¼ 4, and
Figure 13 shows the data plotted using the minimum scaled distance compared to
the average scaled distance. All plots have identical scales in order to simplify
comparisons.
Figure 12 clearly shows that the weighted scaled distance (with k ¼ 4) is a
superior predictor of peak vibration. Although both the weighted and minimum

Fig. 11. Log of vppv vs log of traditional scaled distance.

 236 D.P. BLAIR
Downloaded by [The UC Irvine Libraries] at 12:52 25 October 2014

Fig. 12. Log of vppv vs log of weighted (k ¼ 4) scaled distance.

Fig. 13. Comparison of minimum and average scaled distances.

scaled distance predictors maintain their linearity in the near-field of the blast,
the weighted scaled distance exhibits less scatter and improved behaviour in the
far field.
 CHARGE WEIGHT SCALING LAWS 237
Downloaded by [The UC Irvine Libraries] at 12:52 25 October 2014

Fig. 14. Normalised residual sum of squares for various models.

Figure 14 shows the normalised RSS for three Monte Carlo models of the blast for
which the scaled distance predictors are given by Equation (6) for selected values of k.
As in Figure 8, the normalised RSS for the average scaled distance is given by the open
circles placed at the k ¼ 1 location for convenience only. Since these average scaled
distance predictors do not vary significantly they are not distinguished for each model.
Model 1 has been described previously with some results given in Figures 11 to 13.
Model 2 is identical to Model 1 except that both delay scatter and fluctuating seed
waveforms are employed. Model 3 is identical to Model 2 except that a 20 percent
random variation is added to the charge weight in each blasthole. Thus models 2 and 3
are expected to be a closer approximation to reality than that implied by Model 1.
Figure 14 shows that the influence of delay scatter and fluctuating seed waveforms is
more significant than fluctuations in the charge weight of individual blastholes.

5. DISCUSSION AND CONCLUSIONS

It has been demonstrated with regard to Figure 1 that the charge weight scaling law,
itself, implies a non-linearity in the peak vibration even for two single blastholes as
their separation decreases. Irrespective of charge weight scaling ideas, it is reasonable
to expect that vibration waves from widely-separated holes will superpose linearly
and those from close holes will superpose non-linearly since each hole then lies in
the large-strain, non-linear region of the other. Thus for closely spaced blastholes, the
peak vibrations, too, will superpose in a non-linear fashion. In this regard, at least, the
 238 D.P. BLAIR

scaling law has some consistency with accepted notions of large strain non-linear
attenuation, even though the exact nature of this non-linearity is not yet understood. In
many other aspects, however, the charge weight scaling laws are inadequate, and, in
particular, are fundamentally inconsistent with superposition. For example, if the
vibration waveforms from two or more blastholes are displaced in time yet still
overlap, then the charge weight scaling law cannot be used to predict the peak
vibration even for this simple case. In fact, for any collection of blastholes, it has been
demonstrated that the resulting peak vibration invariably cannot be meaningfully
ascribed to any single distance and any single charge weight as is traditionally done.
Of course, in the very near field, use of the minimum scaled distance may have some
meaning. However, this particular parameter is not traditionally used.
Downloaded by [The UC Irvine Libraries] at 12:52 25 October 2014

It was because of weaknesses in the traditional approach that a new weighted

scaled distance was introduced. In this regard, Figures 4 to 12 clearly show that the
traditional scaled distance is an inferior predictor of peak vibration levels. It is also
worthwhile noting that a plot of log(vppv) versus log(scaled distance) is linear in log-
log space for each individual blasthole, as implied by Equation 1. But when the
vibration is due to the superposition of a collection of blastholes this linearity is not
necessarily maintained. This lack of linearity is clearly seen in Figure 11 when using
the traditional scaled distances, and in Figure 13 when using the average scaled
distance. This finding also questions the current practice of using linear regression
analysis on log(vppv) versus the log(traditional scaled distance), especially in the
near-field of full scale-blasts. However, use of the weighted scaled distance
(especially with moderate values of k) does result in linearity of log(vppv) versus
log(scaled distance) as shown in Figure 12. Thus this finding suggests that it is
meaningful to use linear regression analysis on near field (and far field) data provided
that it is scaled according to Equation (6).
Although the minimum scaled distance shows good properties in the very-near-
field, it is not as robust as the weighted scaled distance in the more-distant-field as
shown by the direct comparison in Figure 13. Furthermore, in the practical case
considered (Figs. 6, 7) the weighted scaled distance out-performed the minimum
scaled distance. This is not surprising in view of the limitations of using the minimum
scaled distance as already noted in connection with Equation (6). It is for these
reasons that the weighted scaled distance approach is considered to be the most useful
in any practical situation.

ACKNOWLEDGEMENTS

My Orica colleague, Les Armstrong, provided much assistance in the monitoring and
analysis of the coal shots (data shown in Figs. 4 to 7). I also thank Mike Croucher of
Bayswater Coal Mine for his general assistance and also for making the site available.
 CHARGE WEIGHT SCALING LAWS 239

REFERENCES

1. Blair, D.P.: Statistical Models for Ground Vibration and Airblast. Int. J. Blasting Fragmentation 3
(1999), pp. 335–364.
2. Dowding, C.H.: Construction Vibrations. Prentice-Hall, New Jersey, 1996, 610 pp.
3. Muller, B. and Hohlfeld, Th.: New Possibilities of Reducing Blasting Vibrations With an Improved
Prognosis. Int. J. Blasting Fragmentation 1 (1997), pp. 379–392.
4. Wu, Y.K., Hao, H., Zhou, Y.X. and Chong, K.: Comparison Study of Coupling Effects of Explosive
Charge on Ground Vibrations. Int. J. Blasting Fragmentation 2 (1998), pp. 25–37.
5. Holmberg, R. and Persson, P.A.: Design of Tunnel Perimeter Blasthole Patterns to Prevent Rock
Damage. In: Proceedings of the Tunnelling’79, 1979, pp. 2870–283 (IMM, London).
6. Blair, D.P. and Minchinton, A.: On the Damage Zone Surrounding a Single Blasthole. In: Proceedings
Downloaded by [The UC Irvine Libraries] at 12:52 25 October 2014

of the Fifth International Symposium on Rock Fragmentation by Blasting, Montreal, Canada, 1996,
pp. 121–130.
7. Blair, D.P.: Some Problems Associated With Standard Charge Weight Vibration Scaling Laws.
In: Proceedings of the Third International Symposium on Fragmentation by Blasting, Brisbane, 1990,
pp. 149–158.
8. Blair, D.P. and Armstrong, L.W.: The Influence of Burden on Blast Vibration. Int. J. Blasting
Fragmentation 5 (2001), pp. 108–129.
9. Blair, D.P.: The Frequency Content of Ground Vibration (2003) (In press).
10. Press, W.H., Teukolsky, S.A., Vetterling, W.T. and Flannery, B.P.: Numerical Recipes. The Art of
Scientific Computing, 2nd edition. Cambridge University Press, New York, 1992, 963 pp.