Bulletin of the Seismological Society of America; February 2006; v. 96; no. 1;
p. 272-287; DOI: 10.1785/0120050068
© 2006 Seismological Society of America
Characteristics of Regional Seismograms Produced by Delay-Fired Explosions at the Minntac Iron Mine, Minnesota
Tom T. Goforth1,
Claus H. Hetzer2 and
Brian W. Stump3
1 Department of Geology
Baylor
University
One Bear Place, #97354
Waco, Texas
76798
tom_goforth{at}baylor.edu
(T.T.G.)
2 Infrasound Laboratory
University
of Hawaii, Manoa
Kailua-Kona, Hawaii
96740
chetzer{at}isla.hawaii.edu
(C.H.H.)
3 Department of Geological
Sciences
Southern Methodist University
Dallas, Texas
75275
stump{at}passion.isem.sum.edu
(B.W.S.)
 |
Abstract
|
|---|
This study utilized seismograms from the U.S. Geological Survey seismological
station near Ely, Minnesota, at which 39 delay-fired explosions at the U.S.
Steel Minntac iron mine in Mountain Iron, Minnesota, were recorded. The
waveforms were analyzed in both the time and frequency domains to characterize
amplitude and energy levels as recorded over the regional travel path. The
measurements were then related to delay-firing parameters as furnished by
Minntac.
The seismograms were filtered in five one-octave bands from 0.5 Hz to 16 Hz,
and peak velocities were measured for the P and the
Rg/Lg arrivals in each frequency band. Peak velocities for
each phase in each frequency band were plotted against total yield, yield/delay
period, yield/hole, and spall. Only highly scattered, poor correlation between
peak time-domain measurements and any blast parameter could be observed.
Spectral amplitude modulations (scalloping) in the 0.3-3.0 Hz range were
observed for all of the recorded mining explosions. Each spectrum also showed an
unusual increase in P-wave spectral amplitude with increasing frequency
beginning at about 3 Hz.
Using blast parameters provided by the mine, a linear superposition model
successfully predicted the major features of the observed spectra, including the
existence and variability of the spectral modulations, which in turn account for
the poor correlation of peak ground measurements with yield. The model results
also indicate that the unusual increase in P-wave spectral amplitude
beginning at about 3 Hz may be due to the high P-wave velocity in the
area being mined.
|
Introduction
|
|---|
The International Monitoring System (IMS) and its associated
arrays are part of an international system designed to record small seismic
events around the world. There are thousands of mining explosions each year that
must be characterized as different from signals associated with nuclear
explosions. Although it has been determined that nuclear explosions and single
chemical explosions are spectrally indistinguishable
(Stump et al., 1999),
large production- mining explosions are most commonly generated by delay-
firing, a technique where the explosive material is distributed among one or
more rows of boreholes. The individual charges are then detonated in a set time
sequence. This technique was developed in order to maximize rock fracturing
while minimizing seismic efficiency through destructive interference. It has
been used very successfully in reducing ground motions associated with the
blasting. However, it has proven to be difficult to relate seismograms produced
by delay-fired sources to the explosive yield of the event. The yield dependency
of peak amplitudes in narrow frequency bands is of particular importance because
of the relationship of these amplitudes to magnitude estimates. We are
interested in using high-quality data recorded along a single path to relate
seismogram characteristics to source parameters in well-constrained delay-fired
detonation sequences.
This study utilized seismograms from the U.S. Geological Survey
(USGS) seismological station near Ely, Minnesota, which recorded 39
delay-fired explosions at the U.S. Steel Minntac iron mine in Mountain Iron,
Minnesota. The Minntac mine is located in the Mesabi Range of Minnesota where
magnetite is extracted from taconite ore in the Precambrian Biwabik Iron
Formation. The explosion sequences are designed to fracture the extremely hard
overlying rock in place, as opposed to coal-mining explosions, which are usually
designed to cast the overburden laterally into a pit. The total amount of
explosive detonated during the firing sequences varied from less than 30,000 to
more than 450,000 kg; the explosive weight per delay period varied from 800 to
10,400kg; and the weight per hole ranged from 285 to 2265 kg. Minntac provided
blast parameters for each event, including total yield, yield per delay period,
mass of rock fractured (which was interpreted to be spall), and the number of
holes used in the detonation sequence. In some cases, the angle between the
perpendicular to the direction of delay- fired progression and the direction to
the recording station was provided.
Table 1 lists the blast
parameters for each production explosion. The USGS station at Ely
(EYMN) is situated in the heavily metamorphosed Precambrian basement
rock that constitutes most of the 112-km path length between the station and
Minntac. The Precambrian terranes consist mainly of belts of high-velocity
metavolcanic and metasedimentary rocks enclosed by granitic rocks.
The study was accomplished in two parts. First, the vertical, broadband
waveforms recorded at EYMN were analyzed in both the frequency and
time domains to characterize phase amplitudes as recorded over the regional
travel path. The measurements were then related to delay-firing parameters such
as total yield, yield/delay, yield/hole, spall, and azimuth to the recording
station. Secondly, the empirical spectra were compared to those predicted by a
linear superposition model created by
Anandakrishnan et al. (1997)
and repackaged into MATLAB by
Yang (1998). This model, called
MineSeis, was used to model individual explosions based on delay-firing
parameters provided by Minntac.
|
Data Analysis
|
|---|
The data set consists of 39 vertical-component velocity seismograms digitally
recorded at 40 samples per second. The P-wave arrival was identified,
and a 120-sec window extending from 20 sec before the P arrival to 100
sec after the P arrival was defined. The mean value and linear trend
were removed from each seismogram, and the instrument response was removed so
that the ground motion was scaled to nanometers per second (nm/sec).
Figure 1 shows examples of
the vertical broadband seismograms for the range of yields available. The
seismograms as shown in this figure have been passed through a 0.3-Hz low-cut
butterworth filter to remove the microseisms.
Figure 2 shows an enlarged view
of an unfiltered vertical component. Six-sec microseisms dominate this
seismogram, but the much higher-frequency signal phases are quite distinct. The
initial P arrival is distinct on all the seismograms and is identified
as Pg based on the expected velocity model and distance. No Pn
arrival is visible due to the extended Pg coda. The phases arriving
almost simultaneously at about 34 sec are Rg, the fundamental-mode
Rayleigh wave, and Lg, representing predominantly transverse motion
through the upper crust. This arrival will be designated Rg/Lg
throughout the article.
View larger version (41K):
[in this window]
[in a new window]
|
Figure 1. Examples of vertical-component, broadband seismograms produced by
delay-firing at Minntac Mine, Mountain Iron, Minnesota, and recorded at Ely,
Minnesota. The amplitude scales are in nm/sec.
|
|
View larger version (24K):
[in this window]
[in a new window]
|
Figure 2. Broadband, vertical seismogram showing the arrival of Pg, Sg, and
Rg/ Lg phases. The amplitude scale is in nm/sec.
|
|
The seismograms were filtered in one-octave bands from 0.5 to 16 Hz.
Figure 3 shows an example of
the filtered seismograms and illustrates the excellent signal-to-noise ratio
that exists in each band. The high-frequency (4–16 Hz) energy in the
P wave persists throughout the signal. The peak values in nm/sec were
measured for the P wave and for Rg/Lg in each
frequency band. A 5-sec window (20–25 sec) was used for the P
wave, and a 10-sec window (33–43 sec) was used for
Rg/Lg.
View larger version (40K):
[in this window]
[in a new window]
|
Figure 3. Broadband, vertical seismogram filtered into five one-octave passbands. The
signal-to-noise ratio is excellent in each band. Peak particle velocities were
measured in each band for both P and Rg/Lg phases.
The amplitude scales are in nm/sec.
|
|
Peak velocities for each phase in each frequency band were first plotted
versus the total explosive yield of each delay-fired sequence.
Figure 4a, showing peak
P-wave velocities in the 4–8 Hz band, is representative of all
the peak measurements versus total yield. There is a very scattered correlation
between yield and amplitude for both the P wave and for
Rg/Lg in each frequency band, with nearly an order of
magnitude scatter in peak amplitude values for any given yield. The peak
measurements were then plotted versus average yield per hole. The expectation of
correlation is based on the assumption that the explosive agent is evenly
distributed among the holes. However, only an extremely scattered (again, nearly
one order of magnitude), low-correlation relationship could be seen in the
plots, examples of which are shown in
Figures 4b and 4c. The peak
values were then plotted versus the average spalled mass (fractured rock
accompanying each detonation) per hole, as shown in
Figure 4d for the P
wave. Spall is known to be a contributor to explosion-generated seismograms, but
little correlation is evident in this plot either. During a delay-fired
sequence, it is common for the explosives in several holes to be detonated
simultaneously at each firing delay interval. Therefore, the peak velocities
were plotted versus yield per delay interval. An example of these plots is shown
in Figure 5a. Again, poor
correlation is observed. Since the peak measurement often was determined by a
solitary excursion, the total energy in the analysis windows was computed as the
sum of the squares of the velocities. Again, there was poor correlation and
large scatter associated with each source parameter, as illustrated for
yield/hole in Figure 5b.
Finally, the dependence of peak velocities on the angle between a perpendicular
to the blast pattern and the direction to EYMN was examined. No
dependency on azimuth was observed. The same degree of scatter shown for the
4–8 Hz frequency band was observed for each of the other one-octave
passbands.
View larger version (28K):
[in this window]
[in a new window]
|
Figure 4. Plots showing the scattered relationship of peak particle velocities to
various expressions of yield. (a) Peak P-wave particle velocity on the
vertical component in the 4–8-Hz band as a function of total explosive
yield. (b) Peak P-wave particle velocity on the vertical component in
the 4–8-Hz band as a function of average explosive per borehole. (c) Peak
Rg/Lg particle velocity on the verticle component in the
4–8-Hz band as a function of average explosive per borehole. (d) Peak
P-wave particle velocity on the vertical component in the 4–8-Hz
band as a function of average spalled mass per borehole.
|
|
View larger version (14K):
[in this window]
[in a new window]
|
Figure 5. (a) Peak P-wave particle velocity on the vertical component in the
4–8-Hz band as a function of average explosive per delay period. (b)
P-wave energy in a 5-sec window on the vertical component in the
4–8-Hz band as a function of yield per hole.
|
|
Despite the large number of parameters tested and the size of the data set,
only highly scattered, poor correlation between peak time-domain measurements
and any blast parameter could be observed on the vertical components. A factor
of 10 in amplitude scatter is common. The most diagnostic source parameter seems
to be yield/hole
(Figs. 4b and 4c), and even in
this case, the data plots show peak particle velocities for events with similar
explosive weights per hole varying by a factor of six or more.
Fourier amplitude spectra were computed for each of the 120-sec, vertical
broadband seismograms. Signal-to- noise ratios were excellent from 0.2 Hz to the
antialiasing cutoff at 18 Hz. A typical example comparing a signal spectrum with
the spectrum of a preceding noise window is shown in
Figure 6.
Figure 7 shows representative
examples of the signal spectra and illustrates two interesting characteristics
that were observed throughout the data set. First, there is a broad spectral
minimum centered near 3 Hz with spectral amplitude increasing toward 18 Hz,
where the antialiasing filter severely attenuates the data in the 18–20-Hz
band. Second, amplitude modulations (scalloping) in the low-frequency band
(0.3-3.0 Hz) are prominent. The interval of modulation is fairly constant on a
given seismogram, but varies from event to event, ranging from about 0.3 to 1.0
Hz. For example, Figure 8 shows
Figure 7a with a linear
frequency scale in which the scalloping has a modulation interval of about 0.52
Hz. Various studies
(e.g., Herrin and Goforth, 1977;
Goforth and Herrin, 1979) have
shown that such modulations can be introduced by multipathing. However, since
the signals from all the events traveled the same path while the modulation
interval varies with event, multipathing seems an unlikely cause. A possible
explanation is provided by
Gitterman and van Eck (1993) and
Stump et al. (2002), who suggest that low-frequency spectral modulations are related to the temporal
finiteness of the source and tend to occur at multiples of the inverse of the
total duration of the source firing sequence. Since the firing-sequence
durations at Minntac range from about 1 to 4 seconds, the resultant scalloping
would appear at frequencies easily observable on USGS seismographs.
On a smaller scale, the subexplosions in the blast pattern can be represented as
a series of distinct, correlated events separated by a constant delay time
(Baumgardt and Ziegler, 1988).
The Fourier transform of this function contains modulations at multiples of the
inverse of the delay time as well as the inverse of the total duration, and
these modulations are sometimes observed at high frequencies (
4 Hz) in the
amplitude spectra of delay-fired events. Since the typical intershot and
interrow delay times for Minntac blasts are on the order of 50 msec, the
high-frequency modulations should appear at multiples of 20 Hz, which is the
Nyquist frequency of USGS seismographs. It is therefore impossible to
determine whether these high-frequency modulations are present in the Minntac
data. Even if the seismographs were capable of detecting them, the
high-frequency modulations depend on a constant delay between subexplosions, and
this factor is very sensitive to random fluctuations in the blasting
procedure.
View larger version (47K):
[in this window]
[in a new window]
|
Figure 6. Comparison of the spectrum of event 108 and the noise in a 20-sec window
preceding the signal.
|
|
View larger version (33K):
[in this window]
[in a new window]
|
Figure 7. Examples of amplitude spectra of broadband, vertical seismograms produced by
delay-firing at Minntac. (a) Event 108: total yield = 195,900 kg,
yield/hole = 680 kg; (b) event 99: total yield = 286,300 kg,
yield/hole = 2272 kg; (c) event 28: total yield = 249,150 kg,
yield/hole = 1143 kg; (d) event 8: total yield = 464,325 kg,
yield/hole = 1300 kg.
|
|
View larger version (26K):
[in this window]
[in a new window]
|
Figure 8. Spectrum of event 108 with a linear frequency scale showing scalloping holes
at intervals of about 0.52 Hz.
|
|
Analysis of the 39 seismograms and their associated spectra indicates three
consistent and interesting observations: (1) there is great scatter in the
yield/ground motion relationship, (2) there is a scalloping of the signal
spectrum that is most pronounced in the 0.3–3.0 Hz frequency band, and (3)
there is a remarkable increase in P-wave spectral amplitude as
frequencies increase above 3 Hz.
|
Theoretically Modeled Seismograms
|
|---|
Synthetic seismograms were generated for comparison with the empirical data
and, if possible, to provide explanations for the observed phenomena. The
theoretical model used was created by
Anandakrishnan et al. (1997).
Yang (1998) reformatted the
model into a MATLAB program called MineSeis and also added
constraints to some parameters in order to ensure conservation of momentum. The
model works on the principle of linear superposition in which the seismogram
resulting from a delay-fired explosion is represented as the seismogram from a
single explosion repeated and superposed at appropriate time delays. This
process was experimentally validated by
Stump and Reinke (1988).
Essentially, the process entails the convolution of a theoretical single-shot
seismogram with a comb function in which the amplitude of each tooth is equal to
the total amount of explosives detonated at that particular time. Both the
explosion and the spall energy are contributors to the single-shot seismogram.
The effects of the travel path are obtained by computing appropriate
Green’s functions that simulate the response of a layered medium to an
impulsive force.
The travel path between Minntac and EYMN is composed of
Precambrian metamorphic basement rock with little intermediate layering
(Greenhalgh, 1980), so a very
simple two-layer and half-space velocity model was used to construct the
Green’s functions. For the computation of Green’s functions,
MineSeis takes as its primary input a one-dimensional layered medium consisting
of compressional and shear-wave velocities, densities and attenuation factors
(Q) for each layer, as well as a range and azimuth. The path model used
in the construction of the Green’s functions is shown in
Table 2. P-wave
velocities and densities were taken from crustal models published by
Weber and Goodacre (1966) and by
Ocola and Meyer (1973). Shear-wave velocities were calculated using a value of
Poisson’s ratio of 0.25. P-wave Q-values were estimated
from the values given by
O’Brien (1967) and
Smith (1989), and Q for
shear waves was estimated to be 4/9 of the P value.
The single-shot source is convolved with the appropriate Green’s
function, and the result is a synthetic seismogram as it would be recorded at
the appropriate range and azimuth. The comb function is constructed by
representing the shot array as a pair of M x N matrices
where M is the number of rows in the shot pattern and N is the
number of shot holes in each row. One matrix contains the detonation times of
the individual subexplosions relative to the first shot at t =
0; the other matrix contains the yield of each subexplosion in kilograms. This
setup allows the easy insertion of time-delay anomalies or nondetonating holes
into the blast pattern. The single-shot seismogram is scaled to unity yield and
convolved with the comb function, resulting in a superposition of yield-scaled,
single-shot seismograms. Additional source parameters that can be specified in
the model include the spall mass, the spall takeoff velocity, spall
impact-pulse width, shot burial depth, burden, azimuth of the working face, and
direction of horizontal spall ejection, if any.
|
Comparison of Theoretical and Empirical Results
|
|---|
Seismograms of two production explosions with similar total yields but
different firing geometries were selected for theoretical modeling. The first
explosion, designated event 8, has a yield of 464,325 kg distributed in 357
holes. The firing sequence was modeled as seven rows of 51 shot holes with 10-m
shot-hole spacing, burden of 10 m, and 0.050-sec intershot and interrow delays.
The explosives were assumed to be evenly distributed among the holes. A 1-m/sec
spall velocity was estimated from low-resolution video data. Spall rise and
impact-pulse widths were estimated to be 0.2 sec. It should be noted that the
specific number of rows and shots per row was not available, but from
information such as number of delays and shots per delay, the geometry can be
inferred. The P-wave velocity of the metamorphosed taconite in which
the explosions were detonated was estimated to be 6.0 km/sec.
Figure 9 shows the
theoretically modeled spectrum, scaled in absolute units, superimposed on the
recorded spectrum of event 8. The model spectrum duplicates the scalloping and
the increase in amplitude in the 3–18-Hz band, although the model
overestimates the amplitudes of the scallop lobes at frequencies of 1 Hz and
greater and overestimates the depths of the holes. Such deviations are not
unexpected since the model assumes perfect timing in the detonation sequence and
perfect correlation between the subexplosions
(Der and Baumgardt, 2001),
assumptions that almost certainly do not exist in practice.
Figure 10 shows the same
comparison on a linear frequency scale. The presence and the periodicity of the
scalloping are directly related to the finite duration (2.8 sec) of the firing
sequence. Model experiments utilizing a range of P-wave velocities
indicate that the high-frequency increase in amplitude is caused by the high
(>5.0 km/sec) P-wave velocity in the immediate vicinity of the
explosions. For example, an increase in modeled P-wave velocity in the
explosion medium from 3.5 km/ sec to 5.0 km/sec will result in a sevenfold
increase in spectral amplitude at 18 Hz.
View larger version (45K):
[in this window]
[in a new window]
|
Figure 9. Comparison of the recorded and theoretically modeled spectra, scaled to
nm/sec/Hz, of event 8.
|
|
View larger version (34K):
[in this window]
[in a new window]
|
Figure 10. Comparison of the recorded and theoretically modeled spectra of event 8
plotted as a linear function of frequency.
|
|
The amplitude spectrum of the recorded vertical seismogram from another
production explosion, designated event 7, has a total yield of 475,650 kg, very
similar to the yield of event 8. However, the firing geometry of event 7 was
somewhat different from that of event 8 in that the total yield was distributed
over 230 holes. Otherwise, the input parameters to the model were the same as
used for event 8. The theoretically modeled spectrum for event 7 superimposed
over the recorded spectrum is shown on a logarithmic frequency scale in
Figure 11 and on a linear scale
in Figure 12. The
correspondence is good in terms of the scalloping and the overall amplitude
level, but the model does not do a good job of predicting the amplitudes and
depths of the narrow-band scallops, probably for the same reasons cited earlier.
Again, the scalloping, different from that in event 8, is directly attributable
to the duration of the firing sequence (3.95 sec), and the high-frequency rise
in amplitude is produced by the high P-wave velocity assigned in the
model.
View larger version (48K):
[in this window]
[in a new window]
|
Figure 11. Comparison of the recorded (dashed) and theoretically modeled (solid) spectra
of event 7 scaled to nm/sec/Hz.
|
|
View larger version (41K):
[in this window]
[in a new window]
|
Figure 12. Comparison of the recorded and theoretically modeled spectra of event 7
plotted as a linear function of frequency.
|
|
By comparing the theoretical spectra
(Figure 13) for events 7 and 8,
we see differences in amplitude in some frequency bands approaching a factor of
10, even thought the total yields of the explosions were nearly identical. Thus,
we might expect scatter of this order in yield versus narrow- band ground-motion
plots.
The yield parameter showing the least scatter in the empirical data was yield
per hole. To determine what order of scatter we might expect theoretically, two
explosion sequences with essentially the same yield per hole, but different
firing parameters, were selected for modeling. An event designated event 101 has
essentially the same yield per hole as event 8; however, event 101 has about
half the total yield of event 8 and about half the firing-sequence duration.
Comparisons of the modeled spectrum and the recorded spectrum of event 101 are
shown on a logarithmic frequency scale in
Figure 14 and on a linear scale
in Figure 15. The general
correspondence is good, with the model showing the scalloping and the
high-frequency content. Again, the model fails to duplicate individual
scalloping lobe amplitudes at frequencies greater that 1 Hz. A comparison of the
recorded spectra of events 8 and 101 in
Figure 16 shows that event 101
has significantly larger amplitudes at frequencies of about 0.33 Hz and at 1 Hz,
while event 8 has larger amplitudes at 1.5–2.0 Hz. A comparison of the
theoretical spectra in
Figure 17 predicts these
differences in amplitude. Differences in amplitude in these frequency bands
approach factors of five or six. This is the order of scatter observed in plots
of yield/hole versus peak particle velocities.
View larger version (42K):
[in this window]
[in a new window]
|
Figure 14. Comparison of the recorded (dashed) and theoretically modeled (solid) spectra
of event 101 scaled to nm/sec/Hz.
|
|
View larger version (26K):
[in this window]
[in a new window]
|
Figure 15. Comparison of the recorded and theoretically modeled spectra of event 101
plotted as a linear function of frequency.
|
|
The model seismograms for events 7, 8, and 101 were filtered into the same
one-octave passbands as the recorded seismograms. The peak particle velocities
for the models show about the same degree of scatter as the recorded data. For
example, the model peak P-wave particle velocities for event 7 and
event 8, events that have comparable total yields, differ by a factor of four in
the 4-8-Hz band. Rg values for the two events in the 4–8-Hz
frequency band differ by a factor of three. However, while the models produce
peak time-domain values with about the same scatter as the observed seismograms,
the model matches to the corresponding recorded spectra are not sufficiently
close in detail to predict peak motions in specific narrow bandwidths.
To investigate possible reasons for the departure of the observed values from
the predicted values, the model was used to determine the effects of unplanned
variations in the yields and timing intervals of the subexplosions. The source
array in this experiment was designed to conform to a typical Minntac shot
sequence, with four rows of 50 shots with a 50-msec intershot delay and a
65-msec interrow delay. A firing sequence was designed in which the time of
detonation of each subexplosion was allowed to vary randomly throughout a range
of ±20% of its nominal value. The effect on the resulting P-wave
and Rg/Lg peak velocities was minimal, introducing a scatter
of less than 3%. However, a quadrupling of the Rg/Lg peak
velocities was achieved by modeling a simultaneous detonation of 25
subexplosions, in this case producing 25 times the maximum yield at any other
time in the explosion sequence. Similarly, a quadrupling of the P-wave
peak velocity was achieved by a simultaneous detonation of 12 subexplosions.
While this is the order of scatter observed in the empirical data, such large
deviations from planned blasting conditions are not common. Another result of
unplanned simultaneous detonations is the mitigation and distortion of spectral
scalloping. A simultaneous detonation of only five holes (less than 3% of the
total) slightly displaces the spectral scalloping at frequencies below 1 Hz and
strongly attenuates and distorts the scalloping at higher frequencies. Such
sensitivity of the scalloping to relatively small departures from a planned
blasting program indicates that an extremely detailed knowledge of the actual
firing parameters would be required to predict the peak particle velocities
that would result from a production explosion.
The model was used to examine the effect of the azimuth of the mine working
face. Goforth and Bonner (1995)
and Bonner et al. (1996) reported large variations in regional Rg waveforms with azimuth,
attributing the differences to the topographic presence of the working pit
rather than the direction of the delay-firing progression. There is no working
pit at Minntac, but the effect of the direction of firing can be modeled and
evaluated. Blast patterns were modeled which varied in 30° increments from
the direction to the recording station at EYMN. Delay-fired
explosions were generated using these blast patterns and peak velocities were
measured. In the case of no horizontal movement of material, as is the case at
Minntac, the peak P-wave and Rg/Lg velocities varied
less than 6% with azimuth. Addition of a horizontal spall term produced a great
deal of azimuthal variation, with both P-wave and
Rg/Lg peak velocities varying by over 200% as the direction of
casting was varied. Maximum values occurred when the horizontal component of the
casting was in the direction of the recording station, and minimum values
occurred when the spall was cast directly away from the recording station.
Explosions detonated at Minntac are designed to fracture material in place, and
horizontal casting is a secondary and minor effect and should not contribute to
azimuthal variations.
|
Conclusions
|
|---|
Seismograms recorded at a distance of 112 km from the Minntac mining
operation show P-wave and Rg/Lg peak particle
velocities that correlate poorly with specified Minntac yield parameters,
including total explosive yield, yield/ hole, yield/delay, and mass spalled. The
total energy in short windows containing the P and
Rg/Lg phases showed no better correlation with yield than did
the peak particle velocities. The amplitude spectra of the seismograms show a
prominent scalloping in the frequency band 0.3–3.0 Hz and an unusual
increase in P-wave amplitude with increasing frequency above about 3
Hz.
Theoretically modeled production explosions based on Minntac firing
parameters also produce seismograms with prominent scalloping and poor
correlation of peak particle velocities with yield parameters. The modeling
simulations suggest that both the observed spectral scalloping and the scatter
in peak particle velocities can be attributed to spatial and temporal
differences in the distributed, delay-fired sources. The periodicity of the
scalloping in the spectra can be directly related to the reciprocal of the
overall duration of the delay-fired sequence. The yield-independent modulations,
varying in character from firing sequence to firing sequence, cause scatter in
the yield/peak particle velocity relationship on the order of a factor of 10 in
some frequency bands. Additional scatter and significant perturbation of
spectral scalloping can be produced by the unintended simultaneous detonation of
a relatively small number of subexplosions in a given firing sequence, as well
as by a lack of correlation between subexplosions. The increase in P-
wave spectral amplitude with increasing frequency above 3 Hz can be
produced by high seismic velocities (>5.0 km/ sec) in the immediate vicinity
of the explosion.
|
Acknowledgments
|
|---|
This work was partially supported by Defense Threat Reduction Agency Contract
DSWA01-98-C-0176 and Department of Energy/National Nuclear Security
Administration (DOE/NNSA) Cooperative Agreement DF-FC52-03NA 99510/A000. Baylor
University and the Baylor University Research Council also contributed to the
financial support of the research. The authors would like to thank Don Thompson
of USX-Minntac for his cooperation in supplying firing-sequence data for each
explosion. Rongmao Zhou of the Southern Methodist University geophysics group
and Xiaoning (David) Yang of Los Alamos National Laboratory were particularly
helpful in implementing the MineSeis software.
Manuscript received April 5, 2005
|
References
|
|---|
Anandakrishnan, S., S. R. Taylor, and B. W. Stump
(1997). Quantification and characterization of regional seismic
signals from cast blasting in mines: a linear, elastic model,
Geophys. J. Int.131
, no. 1,45
–60.[CrossRef]
Baumgardt, D. R., and K. Ziegler (1988). Spectral
evidence of source multiplicity in explosions, application to regional
discrimination of earthquakes and explosions, Bull. Seism. Soc.
Am. 78,1773
–1795.[Abstract/Free Full Text]
Bonner, J. L., E. T. Herrin, and T. T. Goforth (1996).
Azimuthal variation of Rg energy from quarry blasts in central Texas,
Seism. Res. Lett.67
, no. 4,43
–56.
Der, Z. A., and D. R. Baumgardt (2001). Source
directivity, signal decorrelation, spectral modulation and analysis of
spatio-temporal patterns of multiple explosions, Pure Appl.
Geophys. 158,2059
–2076.[CrossRef]
Gitterman, Y., and T. van Eck (1993). Spectra of quarry
blasts and microearthquakes recorded at local distances in Israel,
Bull. Seism. Soc. Am.83
, no. 6,1799
–1812.[Abstract/Free Full Text]
Goforth, T. T., and J. L. Bonner (1995). Characteristics
of Rg waves recorded from quarry blasts in central Texas, Bull.
Seism. Soc. Am. 85, no. 4,1232
–1235.[Abstract/Free Full Text]
Goforth, T. T., and E. T. Herrin (1979). Phase-matched
filters: application to the study of Love waves, Bull. Seism. Soc.
Am. 69, no. 1,27
–44.[Abstract/Free Full Text]
Greenhalgh, S. A. (1980). Effects of delay shooting on
the nature of P- wave seismograms, Bull. Seism. Soc. Am.70
, no. 6,2037
–2050.[Abstract/Free Full Text]
Herrin, E. T., and T. T. Goforth (1977). Phase-matched
filters: application to the study of Rayleigh waves, Bull. Seism.
Soc. Am. 67, no. 1,1259
– 1276.[Abstract/Free Full Text]
O’Brien, P. N. S. (1967). Quantitative discussion
on seismic amplitudes produced by explosions in Lake Superior, J.
Geophys. Res. 72, no. 10,2569
–2575.
Ocala, L. C., and R. P. Meyer (1973). Central North
American Rift System 1—structure of the axial zone from seismic and
gravimetric data, J. Geophys. Res.78
, no. 23,5173
–5194.[Web of Science]
Smith, A. T. (1989). High frequency seismic observations
and models of chemical explosions: implications for the discrimination of
ripple- fired mining blasts, Bull. Seism. Soc. Am.79
, no. 4,1089
–1110.[Abstract/Free Full Text]
Stump, B. W., and R. E. Reinke (1988). Experimental
confirmation of superposition from small-scale explosions Bull.
Seism. Soc. Am. 78,
no. 3, 1059–1073.[Abstract/Free Full Text]
Stump, B. W., M. A. H. Hedlin, D. C. Pearson, and Vindell Hsu
(2002). Characterization of mining explosions at regional
distances: implications with the international monitoring system,
Rev. Geophys. 40,
no. 4, 2-1–2-46.
Stump, B. W., D. C. Pearson, and R. E. Reinke (1999).
Source comparisons between nuclear and chemical explosions detonated at Rainier
Mesa, Nevada Test Site, Bull. Seism. Soc. Am.89
, no. 2,409
–422.[Abstract/Free Full Text]
Weber, J. R., and A. K. Goodacre (1966). A
reconnaissance underwater gravity survey of Lake Superior, in The
Earth beneath the Continents; J. S. Steinhart and T. J. Smith
(Editors), American Geophysical Union, Washington, D.C.,56
–65.
Yang, X. (1998). MineSeis—a MATLAB GUI
program to calculate synthetic seismograms from a linear, multi-shot blast
source model, in Proceedings of the 20th Annual Seismic Research
Symposium on Monitoring a Comprehensive Test Ban Treaty,
21–23 September 1998, Santa Fe, New Mexico,755
–764.
This article has been cited by other articles:
|
|
|
|
 
R.-M. Zhou and B. W. Stump
Frequency-Domain Scaling of Single-Fired Mining Explosions with Different Confinements and Explosive Weights Detonated in Porphyry Granite
Bulletin of the Seismological Society of America,
December 1, 2007;
97(6):
1862 - 1879.
[Abstract]
[Full Text]
[PDF]
|
|
|
Top
Abstract
Introduction
