You are on page 1of 12

Article

Journal of Vibration and Control


2019, Vol. 0(0) 1–12

Synthesis of single-hole signatures by group © The Author(s) 2019


Article reuse guidelines:
sagepub.com/journals-permissions
delay for ground vibration control in rock DOI: 10.1177/1077546319892435
journals.sagepub.com/home/jvc
blasting
Lifeng Li1 and Jhon Silva-Castro2

Abstract
Prediction and control of ground vibrations become essential as with the development of neighborhoods in the proximity of
active mining operations or the need for new infrastructure in urban centers, both requiring the use of blasting. Novel
ground vibration prediction models attempt to reproduce a whole vibration waveform from a blast and are based, in most
cases, on the collection of vibrational information from a single blasthole. A single blasthole should have the same
characteristics (geometry and weights of explosives) as the blastholes used in production shots. In some cases, the
collection of the fundamental information (the signature) is straightforward. In more complex cases, the fundamental
information from ground vibration data is collected from previous production shots. This study presents a novel
methodology to assess the fundamental ground vibration information (the signature) using known information such as one
event waveform (a production shot waveform) and the timing sequence used (the comb function) for the shot. The
methodology is based on the analysis of group delay, a concept widely used in signal processing, and is modified here for the
analysis of ground vibration waveforms. The methodology is developed using real data collected in coal and quarry mining
operations, and at the end of this document, one case study with step-by-step calculations is presented to show the benefits
of the methodology.

Keywords
Ground vibrations, single-hole signature, group delay, synthetic waveforms, rock blasting

1. Introduction based on the convolution operation given in equation (1) to


mathematically represent the vibrations produced by a blast
Millisecond-delayed detonations are used to achieve better
event
fragmentation and to control ground vibration levels pro-
duced by blast events in mining or construction works. y ¼ h∗s (1)
Various methods, including scaled distance law based on
empirical formula (Dowding, 1996; Nicholls et al., 1971; In equation (1), h is the single-hole vibration waveform
Thoenen and Windes, 1942) and signature hole method (the vibration produced by a single hole), which was as-
based on signals and systems (Anderson and Brinckerhoff, sumed by Anderson as identical for every hole. The variable
2008; Blair, 2004; Silva, 2012; Yang and Scovira, 2010) or s is the blasting time function, which can be assumed as
other works (Khandelwal and Singh, 2006; Monjezi et al., a Kronecker comb function. Last, y is the waveform pro-
2010; Oktay, 2017; Verma and Singh, 2011), have con- duced by the blast event.
tributed to the development of fundamental or practical
research on ground vibrations produced by a blast event.
Although most of the available methodologies for 1
Wuhan University of Technology School of Resources and Environmental
ground vibration analysis attempt to predict only the peak Engineering, China
2
values of the ground vibrations produced by the blast event University of Kentucky, KY, USA
(e.g. the peak particle velocity), the more elaborated sig- Received: 22 August 2019; accepted: 12 November 2019
nature hole methods try to replicate the whole ground vi-
Corresponding author:
bration waveform produced by a blast event (Anderson and Lifeng Li, Wuhan University of Technology, School of Resources and
Brinckerhoff, 2008; Blair, 2004; Silva, 2012; Yang and Environmental Engineering, 122 Luoshi Rd, 430070 Wuhan, China.
Scovira, 2010). The original signature hole methodology is Email: lifeng_li@outlook.com
2 Journal of Vibration and Control 0(0)

The assumption of identical vibration waveform for each sets are from a different surface coal mine also in West
hole is indeed not accurate (Silva, 2012), and the straight Virginia. The sample rate used to collect the vibration
use of equation (1) can provide results that can be three waveform information was 1024 samples/s (typical for
times the magnitude than the actual ground vibration mining applications). Some explosion conditions and blast
(Spathis, 2010). patterns are summarized in Table 1.
If the assumption is that individual holes produce dif-
ferent waveforms, it is more appropriate to superpose time-
lagged waveforms into a production blast waveform, as is 3. Fourier transform analysis
shown in the following equation
3.1. Fourier transform of the superposition of
y ¼ h0 ðt  t0 Þ þ h1 ðt  t1 Þ þ / þ hN ðt  tN Þ (2) individual waveforms
where y still represents the waveform produced by the blast The discrete Fourier transform of equation (2) can be ex-
event, hn ðn ¼ 0; 1; …; N Þ are the individual vibration pressed using equation (3). This equation is the frequency-
waveforms from N blastholes, tn ðn ¼ 0; 1; …; N Þ are the domain representation of the superposition of individual
firing times of the N sequential blastholes, and t0 is the waveforms
initial time of the blast event (it is assumed as zero ðt0 ¼ 0Þ
in most of the cases). Y ðωÞ ¼ H0 ðωÞ þ H1 ðωÞ  eiωt1 þ / þ HN ðωÞ  eiωtN
However, the general procedure to assess ground vi- (3)
bration levels requires the measurement of a single-hole
In equation (3), Y ðωÞ is the Fourier transform of the
waveform using equation (1) or even measurement of
waveform produced by the blast event and Hn ðωÞðn ¼
multiple single-hole waveforms by using equation (2), and
0; 1; …; N Þ is the Fourier transform for each individual
then proceed with the superposition of waves produced by
waveform (vibration produced by each hole).
each hole to assess the blast event waveform. Furthermore,
Using equation (3), the blast event can be seen as the
the measurement of a single waveform(s) requires addi-
superposition of phase-delayed Fourier spectra of the
tional operations besides the regular blast designs, which
waveforms produced by each blasthole. Furthermore, using
has possibilities to interrupt the continuity and integrality of
the Fourier transform, a real function of time is transformed
production blasts, and thus increase the costs.
into a complex function of frequency ðωÞ. Using Euler’s
To overcome the aforementioned difficulties, attempts
formula
on synthesizing single-hole signatures by deconvolution
inverse to equation (1) have been made in the past two years eix ¼ cos x þ i  sin x (4)
(Li and Silva-Castro, 2017; Silva-Castro and Li, 2018).
Those methods work well with small-scale blasts with Equation (3) can be expressed in terms of magnitude
a small number of blastholes. However, it becomes more and phase. The expression is as follows
difficult to obtain reasonable single-hole signatures when
the number of blasthole or delays increases. Therefore, new AY eiθY ¼ AH0 eiθH0 þ AH1 eiθH1  eiωt1 þ / þ AHN eiθHN  eiωtN
¼ A eiθH0 þ A eiðθH1 ωt1 Þ þ / þ A eiðθHN ωtN Þ
methods are urgently needed to adapt for blasts with dif-
H0 H1 HN
ferent number of holes (delays).
This study presents a novel methodology inverse to (5)
equation (2) to perform the synthesis of single-hole
where AY is the magnitude of the blast event,
waveforms using the concept of group delay analysis.
AHn ðn ¼ 0; 1; …; N Þ is the magnitude of each signature,
The proposed method can potentially eliminate the in-
θY is the phase of the blast event, and ðθHn 
terruption of single-hole waveform measurement to pro-
ωtn Þ with ðn ¼ 0; 1; …; N Þ is the phase of each individual
duction blasts and thus make the implementation of blast
waveform. This equation indicates that there is a linear
engineering smoother. Now that single-hole waveforms can
phase shift component given by ðωtn Þ added to the original
be obtained mathematically in every round of blasts, sig-
phase ðθHn Þ of each individual waveform.
nature hole method can be used to adaptively predict and
control blast vibrations.
3.2. Magnitude spectrum analysis
2. Field tests and data collection
3.2.1. Effective range of magnitude spectrum. Magnitude is
The collected field data used in this study contain four sets one constituent part of Fourier transform as it appears in
of ground vibration signals. They have different numbers of equation (5), so it is necessary to first investigate the
holes: 6, 47, 62, and 83. The 6-hole data are from a test magnitude spectrum in more detail. In general, the spectral
conducted at Guyan surface coal mine in West Virginia in magnitude of blast vibrations tends to zero after a certain
2011 (Silva Castro and Lusk, 2012). The other three data frequency. In other words, frequency components higher
Li and Silva-Castro 3

Table 1. Measurement condition and collected data.


Number of holes Distance (m) Blast pattern Measured data
6 210

47 150

62 120

83 55

than this certain frequency do not contribute to the amplitude 98% of the cumulative integral of the magnitude with respect
of the ground vibrations. In this article, the maximum fre- to the frequency within a range of 0 Hz–100 Hz. An example
quency value that effectively contributes to the amplitude of is shown in Figure 1, and the value of flim is determined as
the ground vibration is called the “effective frequency limit” 59.5 Hz. Different ground vibration signals will have dif-
or flim and is defined as the frequency corresponding to the ferent effective frequency limits.
4 Journal of Vibration and Control 0(0)

3.2.2. Comparison of magnitude spectrum between a blast magnitude spectrum of the single hole when the blast event
event and a single-hole waveform (a signature). Using ground spectrum is available.
vibration data recorded for several blast events, where the
waveform of the blast event and that of a single hole were 3.2.3. Assessment of the single-hole signature magnitude
recorded, it is possible to compare the magnitude spectrum spectrum from the blast event spectrum. As analyzed pre-
of the blast event ðAY Þ and the magnitude of a single hole viously, each blasthole produces a different vibration
ðAHn Þ according to equation (6). Figure 2 shows four dif- waveform; in other words, in equation (6), if AHn and θHn are
ferent blast events with different number of blastholes (one different for different holes, then equation (6) (Hadamard,
blasthole per delay). 1902) cannot give one unique solution. Without loss of
In Figure 2, the number of delays (blastholes) decreases generality, it can be assumed that AHn and θHn possess
from blast 1 to blast 4. As seen in Figure 2, the magnitude certain randomness, and they are the realization of an en-
spectrum of the blast event is more irregular and with higher semble magnitude AH and phase θH . For the magnitude AHn ,
magnitude values than the less delay (less blastholes) it can be assumed that
magnitude spectrum. Also, in all the cases, the single-hole
spectrum is always of lower amplitude than the blast event AHn ¼ AYs  cHn (6)
magnitude spectrum. Although there are differences be- where AYs is the smoothed magnitude spectrum of a mea-
tween the amplitude of the magnitude spectrum of the blast sured blast event. The smoothing process can be conducted
event and the single hole, they have a similar dominant by computation of moving average of the blast event
frequency and similar distribution of the magnitude with the magnitude spectrum ðAY Þ, and cHn is a random factor which
frequency. This relationship is used to assess the unknown in this article is assumed to follow a normal distribution
N ðμHn ; σ Hn Þ. Because the single-hole magnitude spectrum
is generally of lower amplitude than the event blast mag-
nitude spectrum, as shown in Figure 3, the range of μHn can
be assumed between zero and one. The optimum value of
μHn and σ Hn can be determined by an iterative process.

3.3. Phase spectrum analysis


As mentioned before, because frequencies higher than the
effective frequency limit ðflim Þ do not contribute to the
amplitude of the ground vibration, the analysis of the phase
spectrum is also correspondingly performed for the interval
of frequencies lower than flim . Initially, the principal angle
values (Schafer, 1969) or phase angles are computed as
angle values within the range of ðπ; πÞ. The computed
Figure 1. Example of effective frequency limit. principal phase angle curve is not a continuous function,

Figure 2. Comparison between blast event magnitude spectrum and a single blasthole magnitude spectrum for four blasts in the
transverse vibration direction.
Li and Silva-Castro 5

Figure 3. Comparison of unwrapped phase curves.

although continuity is a necessity for operations on phases, with respect to frequency. It represents a measure of the
such as the computation of complex cepstrum, phase pre- medium phase distortion or the transit time of the vibration
diction, and computation of group delay. The process to signal through the medium versus the frequency. In the
obtain a continuous phase curve from a discontinuous phase following sections, group delay will be used to study the
spectrum is called the unwrapping operation of the principal relationship between the waveform of a blast event and that
phase angles. In this context, the unwrapping algorithm of a single hole.
used in this study originates from Schafer’s research about
a homomorphic system and cepstrum (Schafer, 1969). It
allows positive phase differences between successive phase 4.1. Group delay—phase derivatives
elements by adding multiples of ±2π when absolute jumps The continuous form of group delay is defined as (Gouriet,
between adjacent frequency elements are greater than π. 1957)
This algorithm is widely adopted by many other researchers
(Childers et al., 1977; Thráinsson et al., 2000; Shrikhande dθ
τðωÞ ¼  (7)
and Gupta, 2001). dω
Figure 3 shows the unwrapped phase curves from the
where τðωÞ is the group delay which is a function of fre-
blast events and the single-hole signatures in Figure 2. As
quency, θ is the unwrapped phase, and ω is the angular
seen in Figure 3, the unwrapped phase spectra have a
frequency.
monotonically decreasing trend. In addition, the angle of
The vibrations produced by blasting are measured in
difference between the phase spectrum of the blast event and
a discrete form. Thus, the discrete form of group delay using
a single-hole phase spectrum increases with the increase of the
the symmetric derivative theorem can be expressed as the
number of holes (number of delays) involved in the blast event.
following limit (Aull, 1967)
Despite the observable trend between the blast event and
the single-hole phase spectrum in Figure 3, it is not possible θðω þ ΔωÞ  θðω  ΔωÞ
to establish at this moment a quantitative relationship be- τðωÞ ¼  lim (8)
Δω → 0 2Δω
tween both, so the concept of the group delay to assess
a single waveform (a signature) is introduced next. where Δω is the unit increment of the frequency. If Δω is
small, the central difference (Sheppard, 1899) can be used in
equation (8) as an approximation of the derivative given by
4. Group delay of blast ground vibrations
θðω þ ΔωÞ  θðω  ΔωÞ
All frequency components of a vibration signal are delayed τðωÞ ¼  (9)
2Δω
when propagating through a medium (Blauert, 1978;
Gouriet, 1957; Shulz, 1971). Group delay can be mathe- The unit frequency increment in equations (8) and (9) is
matically defined as the derivative of the phase characteristic calculated by
6 Journal of Vibration and Control 0(0)

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2π Pn
Δω ¼ (10) 0 ðτðmÞ  μτ ðnÞÞ
2
L στ ðnÞ ¼ (13)
nþ1
where L is the length of the discrete Fourier transform. If
equation (9) is written in the discrete form, with ω ¼ nΔω,
3σ range ¼ ½μτ ðnÞ  3σ τ ðnÞ; μτ ðnÞ þ 3σ τ ðnÞ (14)
equation (9) becomes
where μτ ðnÞ is the cumulative moving average of n group
θðn þ 1Þ  θðn  1Þ
τðnÞ ¼  (11) delay values, σ τ ðnÞ is the cumulative standard deviation of n
2Δω group delay values, and 3σ range is the range that covers
99.7% of n group delay values.
4.2. Statistical characteristics of group delay of blast The wide range (dash curves) at the bottom of
ground vibrations Figure 4(a) for values lower than 1 Hz is the result of the
outliers identified at the left bottom of the figure. Figure 4
4.2.1. Frequency-dependent distribution. Group delay can be also shows that the group delay is frequency dependent and
calculated for a blast event and for a single-hole waveform has lower scatter behavior below the flim value. Above the
by applying equation (11). As an example, the group delay flim frequency, the values of group delay start to vary highly.
of both the blast event and the single waveform for Blast 1 The change in the scatter of the group delay with flim is
(83 delays) in Figure 3 is calculated and presented in more evident for the group delay of a single waveform
Figure 4. The horizontal lines in Figure 4(a) and (c) cor- (Figure 4(c)) than for the blast event (Figure 4(a)).
respond with the effective frequency limit ðflim Þ. In the case Figure 4 also shows the magnitude spectrum for the two
of Blast 1, this value is 49.1 Hz. There are also two dash signals ((4b) and (4d)). The relationship between the scatter
vertical curves in Figure 4(a) and (c), respectively, which of the group delay values and the magnitude can be seen in
represent 3-standard-deviation range ð3σÞ for the group the figure. When a frequency has a relatively high mag-
delay data. The 3σ curves indicate the inclusion of the nitude spectrum value, such frequency also has more
99.7% group delay data and were computed using equations contribution to the waveform amplitude of the ground vi-
(12)–(14) brations, and its corresponding group delay results in less
Pn scatter than those frequencies with low contribution to the
τðmÞ
μτ ðnÞ ¼ 0 (12) vibration. This is one explanation of why the group delay
nþ1
scatter presents different behavior above and below the flim
frequency.
Finally, Figure 4 shows that for the range given by
½0; flim , it is possible to represent the group delay versus
frequency relationship by a continuous smooth line, es-
pecially for the signature waveform. This line can be drawn
using a moving average filter for the data group delay versus
frequency in the ½0; flim  range.

4.2.2. Resemblance between group delay distribution and the


waveform. Another significant feature of group delay is its
unit of time, which is the same as that of ground vibration
waveform. Investigations (Boore, 2003; Nigam, 1982;
Sawada et al., 2000) have shown that the histograms (or
distributions) of group delays resemble the envelopes of
ground vibration waveforms. The mean of distribution of
the group delay most of the time corresponds to the location
of the envelope’s peak, and the standard deviation of the
group delay is related to the duration of the waveform. To
illustrate this concept in detail, the group delay histograms
for both the blast event and the single vibration waveform
(signature) for Blast 1 (Figure 4) are included in Figure 5(a)
and (b), respectively. The histograms were prepared using
Figure 4. A typical group delay plot of Blast 1. (a) τ Y denotes the the frequency range given by ½0; flim .
group delay of the production blast, (b) AY denotes the magnitude In this article, the group delay histogram is used to
of the production blast, (c) τ H denotes the group delay of the estimate the length in time of the unknown signature
signature, and (d) AH denotes the magnitude of the signature. waveform. For convenience, the entire duration of the
Li and Silva-Castro 7

Figure 5. Group delay histograms and waveform for the event and signature.

vibration waveform in a particle velocity vs time repre-


sentation can be divided in two parts: the main part and the
tail part (Figure 5). The main part of the signal will contain
the particle velocity peak and the “essential” amplitudes of
the waveform, whereas the tail part will include very small
amplitudes (almost unessential amplitudes). Usually, when
a seismograph returns the raw seismic data, the data are in
most cases longer than what are needed for analysis. So, the
data need truncation to appropriately contain the main part
and the tail part of the vibration record.
After analysis of different vibration waveforms collected
for this research, it was evident that the 0.95 quantile of the
group delay distribution will separate the main and the tail
parts of the record. This is illustrated in Figure 5(a) and (b);
in such figures, the lines of 2.02 s and 0.23 s correspond
with the 0.95 quantile for the event waveform and the
Figure 6. Duration and group delay’s statistics among production
collected signature waveform, respectively. The equation blasts, single-hole signatures, and firing timing sequence.
representing the length of the main part of a general
waveform can be written as in equation (15) which is also a comb function, Lcomb ; (4) length of the entire
lmain ¼ μ þ 1:65 σ (15) production blast waveform, Ly ; and (5) length of the entire
single-hole signature waveform, Lh . There are also four
where lmain denotes the length of the waveform’s main part, statistical parameters labeled in this figure: (a) mean of the
and μ and σ are the mean and standard deviation, re- production blast’s group delay distribution, μy ; (b) standard
spectively, of the group delay distribution for a particle deviation of the production blast’s group delay, σ y ; (c) mean
velocity versus time waveform. of the signature’s group delay distribution, μh ; and (c)
To define the tail part of the unknown signature wave- standard deviation of the signature’s group delay, σ h .
form, it is necessary to plot the three functions under Using those parameters in Figure 6, it is possible to
analysis; they are as follows: the complete event waveform, propose various relationships. The first one is about the
the firing timing sequence, and the signature waveform. duration of the main portion of the production blast and
Such plot is included in Figure 6. the signature. Considering that a production blast waveform
There are five duration parameters in Figure 6: (1) du- ðyÞ and the single-hole signatures ðhn Þ are linked by
ration of the main part of the production blast waveform, equation (3) with the timing sequence ðtn Þ (see Figure 1),
Ly;main ; (2) duration of the main part of the single-hole the length of the main portion of the production blast can be
signature, Lh;main ; (3) duration of the firing time sequence, written as
8 Journal of Vibration and Control 0(0)

Ly;main ¼ Lcomb þ Lh;main (16) as well as the unwrapped phase spectrum is included in
Figure 7.
Equation (16) indicates that the difference between the
duration of the production blast’s main part and that of the Step 2. According to Figure 6, some parameters are to be
signature’s main part is just the duration of firing time calculated.
sequence. When electronic detonators are used, the timing First, compute the group delay histogram and calculate
is accurate enough to determine Lcomb . So, after de- the mean and standard deviation of the group delay data, μy
termining Ly;main by equation (16), the value of Lh;main can and σ y , for the measured production blast in the interval
be estimated. given by ½0; flim . The results are calculated as 0.944 s and
Then, a proportional relation of the statistical parameters 0.579 s, respectively.
can be devised from the ratio of duration parameters Second, determine the duration of the main part of the
μy σ y Ly;main production blast, Ly;main , and the duration of the main part of
¼ ¼ (17) the signature ðLh;main Þ by equations (15) and (16). In
μh σ h Lh;main
equation (16), Lcomb is the duration of the firing sequence of
Equation (17) is used to find the value of μh and σ h , the blast event. In this case study, there were 62 holes
provided all other parameters can be calculated based on detonated at 20 ms intervals, so the total duration of the
collected data. In addition, μy , μh , σ y , and σ h determine the firing sequence Lcomb is 1.191 s. The results are calculated as
relative location and scale between the distributions of the Ly;main ¼ μy þ 1:65 σ y ¼ 0:944 þ 1:65ð0:579Þ s ¼ 1:899 s
production blast and the signature.
Equations (16) and (17) can bridge the distributions of
group delay between the production blast and the signature Lh;main ¼ Ly;main  Lcomb ¼ 1:899  1:191s ¼ 0:708 s
so that it is possible to simulate random group delays of the
signature. However, the duration of the main part cannot Third, compute the mean and standard deviation of
represent the entire waveform, and the tail part is also group delay, μh and σ h , for the signature by equation (17).
necessary to be included in a synthetic signature. If the Thus, μh and σ h are estimated parameters, and the corre-
entire length of production blast and timing sequence are sponding distribution is used as reference of location and
known, then the whole length of the signature, Lh , can be scale for synthetic group delays, which means the distri-
computed by equation bution of synthesized group delays should be within the
reference distribution.
Ly ¼ Lcomb þ Lh (18)
Lh;main 0:708
μh ¼ μy  ¼ 0:944  ¼ 0:351s
Ly;main 1:899
5. Synthesis of group delay and signatures
from production blast Lh;main 0:708
σh ¼ σy  ¼ 0:579  ¼ 0:215 s
As mentioned before, in a regular mining production blast, Ly;main 1:899
there is available information regarding the seismic in-
formation (particle velocity records of the blast event), and
in the case when electronic detonators are used, the firing Step 3. Next step is to randomly synthesize a series of group
timing sequence can be considered as known. Using all the delays for signatures which are within the reference dis-
characteristics of the waveforms discussed in previous tribution ðμh ; σ h Þ. This includes a series of operations on the
sections and the relationships proposed between the event group delay of the blast event.
waveform and the signature waveform when both are (1) Obtain a smoothed group delay spectrum, τ 0y , of the
known, a methodology to assess the single-hole signature blast event (production blast) by a moving average filter.
waveform for vibration prediction and control is introduced The purpose is to make the group delay of the blast event
next. The methodology is illustrated by a case study in less scattered.
which the data were collected in a surface coal mine. The (2) Obtain a seed group delay function τ 0h by the fol-
event waveform was recorded for a production shot of 62 lowing equation so as to compress the seed group delay by
holes, using electronic detonators and a timing given by a scale factor of σ h =σ y and center it at μh
20 ms between charges.   σ
τ 0h ¼ τ 0y  μ0y
h
þ μh (19)
σy
Step 1. Perform fast Fourier transform (FFT) of the
measured blast event waveform and determine the ef- (3) Add random factors to the seed group delay spectrum
fective frequency limit ðflim Þ of the Fourier spectrum. The ðτ 0h Þ, including changes in the sign, the scale ðaÞ, and the
waveform of the blast event and the magnitude spectrum shifting of the delay spectrum ðbÞ. The randomly generated
Li and Silva-Castro 9

Figure 7. Blast event waveform, spectral magnitude, and unwrapped phase.

Figure 8. Randomly generated synthetic group delay functions.

group delay spectrum will be denoted by τ h. The operation 2, the random scale factor ðaÞ can be such that the extent of
is expressed in equation (20) as the synthetic group delay changes between σ 0h and σ h . Fi-
 
τ h ¼ sign  a  τ 0h  μh þ b (20) nally, the shifting factor ðbÞ will move the synthetic group
delay around the seed group delay within the range of the
In equation (20), a negative sign (1) means to flip the reference distribution.
group delay spectrum around μh , the mean of reference Sixty-two (62) random synthetic signatures were repro-
distribution. According to the reference distribution in step duced for this case study, and they are included in Figure 8.
10 Journal of Vibration and Control 0(0)

Step 4. Based on the randomly generated group delays, it is detonators were used in those blast field tests in this re-
possible to synthesize the spectral magnitudes and un- search, which makes sure the actual firing time of each
wrapped phases by using equations (6) and (11). Then, the delay is accurate enough to be the same as the designed
signatures can be synthesized by Fourier series. Figure 9 timing. As a result, the duration of firing time sequence
shows the results. would be close to reality. If accurate actual firing times
cannot be guaranteed when less accurate detonators are
Step 5. Finally, using the synthetic signatures and the used, the reliability of the proposed method may be
linear superposition introduced in equation (5), it is affected.
possible to back-calculate the event waveform. Figure 10
shows the recorded event waveform (waveform produced
by 62 holes) and the back-calculated waveform using the
6. Conclusions
synthetic generated signatures. As seen in Figure 10, the Below are some conclusions according to the findings of
simulated event waveform is similar to the measured this research and the proposed methodology.
waveform.
At this point, the proposed methodology can be sum- 1. To describe a production blast, a linear superposition
marized and presented in the following flowchart: model is more suitable than a strict convolution
Figure 11 shows that the proposed methodology inno- model, especially for complex situations.
vatively integrates systems and signal theory and statistical 2. The inverse problem of synthesizing a single-hole
methods together to solve an inverse problem of blast signature from the measured production blast wave-
vibrations. The key for successfully conducting this form cannot be solved by conventional deconvolution
method is the duration of firing time sequence. Electronic methods.

Figure 9. Spectral amplitude, unwrapped phase, and synthetic signatures using the proposed methodology.

Figure 10. Measured event waveform versus back-calculated event waveform using the synthetic signatures.
Li and Silva-Castro 11

Figure 11. Flowchart of the proposed methodology.

3. The phase contents of Fourier transform have fun- References


damental influence on the time-domain waveform. Anderson D and Brinckerhoff P (2008) Signature hole blast vi-
However, phase angles do not carry much useful bration control-twenty years hence and beyond. Proceedings of
information about synthesis of signature from pro- the Annual Conference on Explosives and Blasting Technique
duction blasts. We have to turn to its derivative— 2(34): 27.
group delay. Aull CE (1967) The first symmetric derivative. The American
4. Group delay has the advantage of time unit, and its Mathematical Monthly 74(6): 708–711.
distribution resembles the envelope of a time-domain Blair DP (2004) Charge weight scaling laws and the superposition
of blast vibration waves. Fragblast 8(4): 221–239.
waveform. Quantitative relationships have been de-
Blauert J and Laws P (1978) Group delay distortions in electro-
veloped in this article so that an ensemble of group
acoustical systems. The Journal of the Acoustical Society of
delay spectra of signatures can be synthesized based America 63(5): 1478–1483.
on the measured production blast data. Boore DM (2003) Phase derivatives and simulation of strong
5. A methodology of synthesizing group delay and ground motions. Bulletin of the Seismological Society of
signatures based on group delay operations has been America 93(3): 1132–1143.
proposed. It takes into account the randomness of each Childers DG, Skinner DP and Kemerait RC (1977) The cepstrum:
blasthole and wave propagation path in the way of a guide to processing. Proceedings of the IEEE 65(10): 1428–1443.
group delays. This methodology successfully syn- Dowding CH (1996) Construction Vibrations. Upper Saddle
thesized an ensemble of signatures, which can be used River, NJ: Prentice Hall, Vol. 610.
to predict future production blasts. Gouriet GG (1957) Two theorems concerning group delay with
practical application to delay correction. Proceedings of the
IEE-Part C: Monographs 105(7): 240–244.
Hadamard J (1902) Sur les problèmes aux dérivées partielles et
Declaration of conflicting interests leur signification physique. Princeton, NJ: Princeton University
The author(s) declared no potential conflicts of interest with re- Bulletin, 49–52.
spect to the research, authorship, and/or publication of this article. Khandelwal M and Singh TN (2006) Prediction of blast induced
ground vibrations and frequency in opencast mine: a neural
network approach. Journal of Sound and Vibration 289(4–5):
Funding 711–725.
The author(s) disclosed receipt of the following financial support Li L and Silva-Castro J (2017) Spectral division deconvolution of
for the research, authorship, and/or publication of this article: The blast vibration signals for signature estimation. ISEE 43rd
study is supported by the Fundamental Research Funds for the annual conference on explosives and blasting technique, Or-
Central Universities (WUT: 193108006). lando, FL, February 2017. Orlando, FL, Princeton, NJ: In-
ternational Society of Explosives Engineers.
Monjezi M, Ahmadi M, Sheikhan M, et al. (2010) Predicting blast-
ORCID iD induced ground vibration using various types of neural networks.
Lifeng Li  https://orcid.org/0000-0002-4554-6563 Soil Dynamics and Earthquake Engineering 30(11): 1233–1236.
12 Journal of Vibration and Control 0(0)

Nicholls HR, Johnson CF and Duvall WI (1971) Blasting Vi- Silva JJ (2012) Blast Vibration Modeling Using Improved Sig-
brations and their Effects on Structures. Washington, D.C.: nature Hole Technique for Bench Blast. Lexington, KY:
U.S. Bureau of Mines, Bulletin, 656. University of Kentucky.
Nigam NC (1982) Phase properties of a class of random proc- Silva-Castro JJ and Lusk B (2012) Modeling of signature waves
esses. Earthquake Engineering & Structural Dynamics 10(5): based on Fourier series for blast vibration prediction. Blasting
711–717. Fragment 6: 179–194.
Oktay T (2017) Output variance constrained bending control of Silva-Castro J and Li L (2018) Deconvolution of blast vibration
rotating Euler–Bernoulli beam. Proceedings of the Institution signals by Wiener filtering. Inverse Problems in Science and
of Mechanical Engineers, Part E: Journal of Process Me- Engineering 26(10): 1522–1538.
chanical Engineering 231(2): 202–211. Spathis AT (2010) A brief review of measurement, modeling and
Sawada S, Morikawa H, Toki K, et al. (2000) Identification of management of vibrations produced by blasting. In:
path and local site effects on phase spectrum of seismic ground Vibrations from blasting: proceedings and monographs in
motion. In: Proceedings of the 12th world conference on engineering, water and earth sciences, Grenada, Spain,
earthquake engineering, Auckland, 2000. Upper Hutt, New September 2009. London: CRC Press.
Zealand: New Zealand Society for Earthquake Engineering, Thoenen JR and Windes SL (1942) Seismic Effects of Quarry
Silverstream. Blasting. Washington: US Department of the Interior, Bureau
Schafer RW (1969) Echo Removal by Discrete Generalized Linear of Mines, Bulletin 442.
Filtering. Cambridge, MA: Massachusetts Institute of Technology. Thráinsson H, Kiremidjian AS and Winterstein SR (2000) Mod-
Sheppard WF (1899) Central-difference formulae. Proceedings of eling of Earthquake Ground Motion in the Frequency Domain.
the London Mathematical Society s1–s31(1): 449–488. Stanford CA: John A. Blume Earthquake Engineering Center
Shrikhande M and Gupta VK (2001) On the characterisation of Report 134.
the phase spectrum for strong motion synthesis. Journal of Verma AK and Singh TN (2011) Intelligent systems for ground
Earthquake Engineering 5(4): 465–482. vibration measurement: a comparative study. Engineering with
Shulz FJ (1971) Envelope delay phase delay group delay chroma Computers 27(3): 225–233.
delay; what does it mean? how is it measured?. In: 19th Annual Yang R and Scovira DS (2010) A model for near and far-field blast
NCTA convention. Washington, D.C.: National Cable Tele- vibration based on multiple seed waveforms and transfer
vision Association. functions. Blast Fragment 4: 91–116.

You might also like