[Link]
au/p/1952_33_Birch/
Seismic Response to hydraulic fracturing in caving mines
D.J. Birch and S.G. Meyer
Institute of Mine Seismology, Australia
We present qualitative observations of the recorded rock mass response to hydraulic-
fracturing in four case studies from three major caving mines and demonstrate usefulness
in the application of advanced seismological methods to describe this response. Three out
of the four cases represent a rare opportunity where the seismic response to hydraulic-
fracturing was fairly isolated. The activation of structures was readily evidenced using
source mechanism analysis. Hydraulic-fracturing applied along the abutment of a cave
was shown to be effective in inducing propagation in a highly stressed, stalled section of
the cave. The dominant mode of failure in all cases was shear. The amount and type of
volumetric change associated with seismicity was dependent on the proximity of the
events to mine voids. Seismic activity rates were estimated to be between 10 and 75 times
higher during fluid injection than in the time prior to treatment.
INTRODUCTION
Hydraulic fracturing (HF) is commonly employed in underground block caving mines as a means to
precondition the rock mass, thereby improving cavability and reducing the seismic hazard Van As
(2000) and Catalan et al (2017). When the pressure of the injected fluid exceeds the sum of the rock
strength and in situ minimum principal stress, failure is induced. Failure may be interpreted as tensile
(hydraulic fracturing), hydro-shearing (frictional strength on a pre-existing plane of weakness is
exceeded, inducing slip) or a combination of both where a new fracture is created and then slips as a
result of shear stress acting on the plane Gischig and Preisig (2015). The end result is an increase in
fragmentation which reduces the capacity of the rock mass to accumulate strain energy, thereby,
reducing seismic hazard.
Mines need to evaluate the effectiveness of preconditioning methods and often rely on production
figures such as draw rates, rock fragmentation, and percentage of hang-ups, for quantification, i.e.,
when the rock is mined through, or look to numerical models to provide answers. Modelling the effect
of hydraulic fracturing in the rock mass remains inherently troublesome due to necessary assumptions
about the medium, in situ stress field, fluid interaction and mode of failure. A thorough analysis of the
immediate recorded seismic response to HF is also useful in determining effectiveness and can be
performed directly after treatment. Seismological tools such as advanced location methods and the
analysis of source parameters and mechanisms can be used to interpret the response in terms of the
mode of failure, assess changes in the rock mass properties and stress field and finally, reevaluate the
effect on seismic hazard.
Deep Mining 2019 Conference
Muldersdrift, 24–25 June 2019
The Southern African Institute of Mining and Metallurgy
445
HF logs from four different campaigns (provided by the respective cave mining operations) were
analysed in conjunction with the recorded seismic response which varied depending on the sensitivity
of the seismic network and rock mass conditions. Information about the individual HF treatments
varied from only summary information, i.e., location of the injection point and volume injected, to time
history borehole logging for individual fracs. At two of the mines the method was applied on a trial
basis and at the third, HF was applied to precondition the rock mass, thereby mitigating severe strain
bursting which affected the development of two tunnels in different parts of the mine Rojas and
Landeros (2017). This study represents a rare opportunity where the seismic response to HF was fairly
isolated, i.e., no production mining taking place nearby, in three out of the four cases. Table I
summarises each case. The seismic event magnitude scale is the Hanks and Kanamori (1979) moment
magnitude scale denoted by MHK. The number of events is shown in Table II.
Table I. Summary of hydraulic fracturing campaigns from three different caving mine operations
Number of Number of injection Average volume of fluid Magnitude range of recorded seismic
Case Study
boreholes episodes injected per frac (kL) events (MHK)
1 1 13 11.21 -2.3 to -0.1
2 3 92 16.13 -1.8 to 0.8
3 7 297 10.15 -3.5 to 1.5
4 12 505 Not provided -2.8 to 1.1
METHOD
There is a vast difference in the scale of the individual case studies; however, each case was treated
similarly from a seismological perspective.
DATA QUALITY CONTROL
Ground motion resulting from the sudden inelastic deformation related to HF was recorded by each
mine’s microseismic monitoring network. These networks mostly used 14 Hz and 4.5 Hz geophones.
The events were selected based on their spatial and temporal association with episodes of HF. The initial
phase of the analysis involves accurate calculation of locations and source parameters of events
attributed to HF. This is achieved by correctly establishing site settings across the microseismic network
and then batch reprocessing the data set. In each case study a controlled database of blasts with known
locations was used to derive an appropriate velocity model in the volume of rock mass where HF took
place and to confirm the correct sensor orientation settings.
In some cases where large impedance contrasts exist, it was necessary to establish site specific velocities.
Relative location, also known as the Arrival Time Difference method or master event location Mendecki
(1997), proved useful in cases where the depths of events and seismic wave velocities varied
significantly. In a solid rock mass, the first arriving P-waves are polarised in a direction parallel to the
direction of propagation at the geophone. This information can be used to further constrain the locations
of events by minimising the misfit between the measured polarity of the P-wave at each site (assuming
correct orientation settings were established and ray paths are straight) and the current source location.
Figure 1 shows an example where the relative location algorithm was applied. The relocated events
(blue spheres) are more clustered and there is a stronger association with an interpolated shape of the
yield zone of the cave (grey mesh). The circled events were located using sensors that formed a planar
configuration, resulting in the initial location algorithm finding a local minimum. First arrival polarities
confirmed that these events occurred closer to the cave and relative location corrected this.
446
Quality control of source parameters is conducted by calculating the difference between the observed
and theoretical source displacement spectra at each site for a controlled subset of events. The result
should be zero across all frequencies, thus, we are able to invert for the anelastic attenuation factors
(constant QP and QS at frequencies much larger than corner frequency) and identify sites showing severe
amplification effects that may be caused by poor grouting in boreholes or damaged sensors and/or
cables. Problematic sites were removed from subsequent source calculations.
Figure 1. Result of the relative location algorithm applied to HF-induced events along a cave back (Case 2). Red
spheres are the original locations and the blue spheres are the relocated events. Blue lines represent the treatment
boreholes
SOURCE MECHANISM ANALYSIS
Following reprocessing of the database of HF related events, the source mechanisms were calculated
using a source mechanism inversion method which inverts for the mechanism that provides the best
agreement to the observed polarities and amplitudes of P- and S-waves at trusted sites. A full moment
tensor solution is provided. The data sets include thousands of events, which necessitates an automated
approach to the calculation. Therefore, selection criteria based on the minimum number of usable P-
and S-wave arrivals and focal sphere coverage are used to ensure reliable results.
Results from the application of this method in each of the four case studies are summarised in Table II.
The minimum requirements for a source mechanism included; 6 P-wave arrivals, 6 S-wave arrivals and
a minimum focal sphere coverage of 50%.
Table 2. Total number of full moment tensor calculations carried out for each case study
Case study Number of HF-related events Number of full moment tensor solutions Percentage successful calculations
1 2064 704 34%
2 1233 251 20%
3 8519 1737 20%
4 2636 518 20%
447
The interpretation of HF-induced seismic events focuses on all aspects of the full moment tensor
solution and makes inferences about the mode of failure and changes in the stress field, and possibly
identifies structures which may have failed suddenly. These aspects include:
i. directions of the principal strain axes
ii. amount of constant-volume shear vs. volumetric change
iii. nodal planes.
The directions of the maximum and minimum principal strain axes are proxies for the directions of
principal stress in the rock mass driving sudden inelastic deformation at a particular location and point
in time. The T-axes (direction of minimum principal strain) indicate extension while P-axes (direction
of maximum principal strain) indicate compression. This information can be visualised in 3D to provide
an indication of how the stress field is oriented in the volume.
Hudson plots Hudson (1989), also referred to as source-type plots, display the range of the obtained
moment tensor types in terms of two parameters T and k, which characterise the relative amounts of
shearing (deviatoric component of the full moment tensor) and volumetric change (isotropic component
of the full moment tensor), respectively. These plots provide a generalised overview of the dominant
source-type/mode of failure. It is common to see a number of events with large volumetric change
components inwards (implosive). This can be attributed to elastic convergence in nearby excavations.
Source mechanisms and event locations can easily be compared with the spatial distribution of mapped
geological structures but often unknown structures are illuminated by the HF. Clusters of events with
significant double-couple components can potentially be attributed to slip on a geological structure if
the distribution of the events is found to be co-planar with one of the nodal planes of the source
mechanisms of the events in question.
RESULTS
The case studies and corresponding results are described individually. It is important to keep in mind
that Case 2 was the only case study where HF was applied in the yield zone of an established cave.
Case 1
Kiruna mine, Sweden
A HF trial was conducted in a single borehole involving 13 fracs. The borehole was located between
two mining blocks far from the cave. The microseismic monitoring array provided excellent coverage
with a high level of sensitivity. This is reflected in the high number of events relative to the number of
fracs and high success rate of moment tensor solutions (See Tables I and II). The relative location method
was applied to the event set. Figure 2 suggests that a correlation exists between the amount of fluid
injected, range of spatial distribution of recorded events and number of recorded events.
The calculation of seismic source mechanisms produced a large majority of predominantly double-
couple (shearing) type events (See Figure 3). A high number of planar clusters were identified. These
were characteristically varied in terms of the planar distribution of locations relative to the associated
seismic source mechanisms. In certain clusters (See Figure 2) the planar distribution of event locations
or event planes are aligned with a common nodal plane from the mechanism solutions, indicating slip
along an existing plane of weakness. In other groups the strike of the event plane was sub-parallel to
the P-axes (~45° offset from the nodal planes). This is not fully understood but one possibility is shear
slip along parallel joints brought about by perturbations in the stress field during fluid injection.
Size distributions of two event sets (Normalised to 30 days) are compared in Figure 4. The events were
recorded in the months before hydraulic-fracturing and during fluid injection in the same volume.
448
Although the time scales are different, normalisation allows us to compare the activity rates based on a
particular magnitude of completeness. A completeness magnitude of Mmin = -1.2 is appropriate in this
case. The activity rate is approximately 75 times higher during injection. The slope appears to be steeper
indicating a higher ratio of small-to-large events.
Figure 2. Correlation of event location planes with source mechanism solutions looking NW. The dots represent
all recorded events coloured according to time. Spheres represent the injection points scaled according to volume
injected and coloured according to time. Stereonets display nodal planes and the P-, T- and B-axes (represented
by red, blue and green markers, respectively) of the source mechanism solutions for planar clusters
Figure 3. Hudson diagram of source mechanism solutions obtained for HF-induced events for Case 1
449
Figure 4. Comparison of normalised cumulative size distribution plots for HF-induced events and events
recorded in the same volume prior to HF between Blocks 30 and 34 at Kiruna mine
Case 2
Malmberget mine, Sweden
Hydraulic fracturing was conducted on a trial basis along the western flank of the Printzsköld cave
involving three boreholes drilled from the surface (~500 m to 700 m in depth). The in situ stress
conditions and shape of the cave provide favourable conditions for large events that are readily felt on
surface. Therefore, a major theme of this trial was to study the effect on cave propagation to re-distribute
stress and reduce the seismic hazard. The relocated events are closely associated with the yield zone of
the cave. The effect of the yield zone is evident in the HF parameters. Breakdown pressure tends to
decrease while volume injected increases as the injection points get nearer the cave. The source
mechanisms when plotted on a Hudson diagram (See Figure 5) tend towards tensile closing. This can
be explained by closure (failure towards the source) within the yield zone of the cave.
Figure 5. Hudson diagram of source mechanism solutions obtained for HF-induced events along the a cave
abutment
450
Directions of the principal strain axes obtained in the source mechanism calculations support the notion
of cave growth, particularly in the bridge between the two sub-caves. The T-axes (Blue vectors in Figure
6) were predominantly sub-orthogonal to the cave indicating extension. The P-axes (Red vectors in
Figure 6) correlate well with the in situ maximum principal stress, which is horizontal and oriented
approximately NW-SE.
σ1
Figure 6. Orientations of P-axes displayed in plan view (left) and T-axes looking North (right) obtained from
source mechanisms of events recorded during treatment. Red arrows show the approximate orientation of the in
situ maximum principal stress
Comparing the spatial distribution of events in the two months before, during and after treatment (See
Figure 7) shows how HF altered the shape of the seismogenic volume around the cave in the short term.
The zone around the deeper sub-cave was responsible for most of the recorded activity prior to HF as a
result of stress concentration in a stalled section (bridge). After treatment, however, the event clusters
shifted, smoothing over this section and associating with the yield zone of the larger, more established
cave.
Figure 7. Comparison of event density clusters before, during and after HF treatment along the cave abutment
A comparison of the size distributions of events (normalised to 30 days) recorded 6 months prior to and
during fluid injection shows a higher activity rate (approximately 10 times for a minimum magnitude
of completeness of MHK -1.1), a higher proportion of smaller magnitude events and a smaller maximum
451
recorded magnitude (See Figure 8). This supports the notion that there is a reduction in the
accumulation of strain energy through an increase in fragmentation.
Figure 8. Comparison of normalised cumulative size distribution plots for HF-induced events and events
recorded in the same volume prior to HF along the Printzsköld cave at Malmberget mine
Case 3
Personnel access tunnel El-Teniente mine, Chile
HF is common practice at El-Teniente mine, being applied regularly to improve production during
block caving through increased fragmentation as well as reduce hazard around the working levels. This
case fell into the latter of the two uses – to precondition the rock mass for the development of a tunnel
with the aim of reducing the number of damaging strain bursts Rojas and Landeros (2017). Treatment
boreholes were drilled on either side of approximately 120 m of planned tunnel.
Development of the tunnel had already started prior to HF. Treatment started 20 – 50 m from the end
of the tunnel moving away in the direction of planned tunnelling. As a result, the first fracs generated
a large seismic response around the tunnel end. P-axes from source mechanisms in this cluster are sub-
parallel to the excavation suggesting that deformation was driven by tunnel-induced stresses (See
Figure 9). It is, therefore, reasonable to conclude that HF successfully redistributed stress in this volume.
The majority of source mechanisms are classified as shear slip (See Figure 10). A large planar
distribution of events was visible after relocation. Source mechanism analysis showed agreement with
one of the nodal planes in this cluster and the event plane (See Figure 11). The tunnel had not yet
intersected this plane.
Figure 9. Distribution of P-axes from HF-induced events around the tunnel
452
Figure 10. Hudson diagram of source mechanism solutions obtained for HF-induced events at TAP
Figure 11. Activation of a structure by HF ahead of the tunnel. Spheres mark the injection points and are
coloured according to time. Nodal planes and poles of the principal strain axes from the selected events (circled)
are projected onto the stereonet (insert)
A comparison of the size distributions of events recorded prior to HF versus events induced by HF in
the same volume is presented in Figure 12. A longer time window (3 years) was required to sample the
volume before HF compared to Case 1 and 2. During this time the majority of events were related to
blasting (early development of the TAP tunnel) and had a larger completeness magnitude due to a lower
system sensitivity at the time. For Mmin -0.4, the activity rate is approximately 30 times higher during
fluid injection. The slopes in the size distributions are similar, unlike Case 1 and 2 where a steeper slope
was observed during HF. However, the maximum recorded magnitude during HF was also lower in
this case.
453
Figure 12. Comparison of normalised cumulative size distribution plots for HF-induced events and events
recorded in the same volume prior to HF around the TAP tunnel, El-Teniente mine
Case 4
XC-22/23 tunnel El-Teniente mine, Chile
This scenario was similar to Case 3, only on a larger scale, and is also discussed by Rojas and Landeros
(2017). HF spanned a section, approximately 340 m long, trending east to west, that would later be
tunnelled through. Brzovic et al (2017) reported that the implementation of HF resulted in an increased
rate of seismic activity in the area and a lower range of magnitudes. The relocated events showed a large
horizontal spatial extent to the north and south of the treatment boreholes. It is likely that this diffusion
was aided by existing geological structures and driven by the in situ maximum principal stress, which
is sub-horizontal and oriented approximately NNW–SSE (See Figure 13). There was good agreement
between the orientation of σ1 and P-axes from the mechanism solutions. The T-axes are mostly sub-
vertical. The majority of seismic source mechanisms were double-couple (See Figure 14). Evidence of
shallow-dipping, sub-parallel structures was found (See Figure 15).
As was observed in Case 3 (TAP), events recorded in the volume prior to HF were scarce. There was,
however, a history of seismic events related to the development of the tunnel. A comparison of the
cumulative size distributions (See Figure 16) yields a similar result. The total injection duration had to
be approximated due to incomplete information. An average injection duration calculated form
partially complete information at TAP was assumed and used in the normalisation of the size
distribution for HF-induced events at XC-22/23. It is reasonable to assume that the same HF procedure
was carried out at both sites. The observations in both cases are similar. Figure 16 shows a higher level
of activity (~30 times for Mmin -0.2) during HF, similar slopes for both size distributions and a lower
maximum recorded magnitude during HF.
454
Figure 13 - Distribution of events recorded during HF in plan view (top) and looking north (bottom). The
events are coloured according to time (left) and mechanism solutions (right) are represented by maximum and
minimum principal axes (P- axes are red and T-axes are blue)
Figure 14. Hudson diagram of source mechanism solutions obtained for HF-induced events at XC-22/23
455
Figure 15. Evidence of shallow-dipping, sub-parallel planes that were active during HF at XC-22/23. Nodal
planes and principal strain axes from selected events are projected onto stereonets
Figure 16. Comparison of normalised cumulative size distribution plots for HF-induced events and events
recorded in the same volume prior to HF around the XC-22/23 tunnel, El-Teniente mine
456
CONCLUSIONS
We present qualitative observations of the recorded rock mass response to HF at three major caving
mines and demonstrate usefulness in the application of advanced seismological methods to describe
this response.
The activation of structures is fairly common and readily evidenced using source mechanism analysis.
Curiously, we observed event locations forming planes aligned parallel to the P-axes. A possible
explanation for this is slip along parallel joints, not caused directly by fluid, but rather through
perturbations in the stress field as a result of fluid injection. One case of treatment applied along the
abutment of an established cave (partially within the yield zone) is fairly unique. A key observation in
this case was the effectiveness of the method to redistribute stress and propagate the cave in a critical
zone, one where large events were being generated as a result of stress concentration. HF was
instrumental in reducing the number and severity of damaging strain bursts during the development
of two tunnels at El-Teniente mine. At XC-22/23, the effect of local geology and in situ stress resulted in
events being recorded at large distances from the injection points (>100 m). Unfortunately, this could
not be statistically evaluated as it was for the other cases (See Table 3) due to missing information.
Characteristics of the recorded seismic response are summarised in Figure 17 and Table III and provide
valuable information on the implications for seismic hazard during fluid injection. Figure 17 is a plot of
the distances between individual events triggered during fluid injection with magnitudes above the
respective Mmin, and injection points. Start and stop times of HF were used to filter the data. The median
and standard deviation was calculated for each set of data (See Table 3). The spatial extent of the
recorded response in Case 2 (Malmberget) was significantly larger than the rest. This can be attributed
to the proximity of the cave yield zone.
Dominant source types were estimated by taking the mean of the parameters k (Volumetric change) and
T (constant-volume shear). The values are given in Table III and plotted in Figure 18. Results suggest
that the observed mode of failure in all cases was predominantly shear. Cases 3 and 4 showed slightly
more explosive mechanisms. Case 2 was the most implosive of them all. This was expected due to the
proximity of the events to the cave.
Increases in the activity rates during fluid injection were estimated from size distribution plots
(normalised to 30 days) of events recorded during fluid injection and prior to HF in the same volume.
Completeness magnitude between the event set pairs varied as system sensitivity had, in most cases
improved by the time HF started. In all cases, the maximum recorded magnitude during HF is lower
while the activity rates are between 10 and 75 times higher.
Table III. Characteristics of the recorded seismic response to HF in deep caving mines
Dominant Spatial range of Maximum recorded magnitude (MHK)
Case Activity rate increase
source-type recorded response
study during fluid injection Before HF During HF
mean (k,T ) (median, STDEV) [m]
1 -0.049, 0.112, 26, 10 ~75 times (Mmin -1.2) 1.8 -0.1
2 -0.150, 0.039 55, 57 ~10 times (Mmin -1.1) 2.4 0.8
3 0.077, 0.041 48, 28 ~30 times (Mmin -0.4) 2.6 1.5
4 0.088, 0.103 –, – ~30 times (Mmin -0.2) 1.9 1.1
457
Figure 17. Concatenated distance versus time plot of the observed seismic response during HF
Figure 18. Dominant source types represented by mean values of k and T from moment tensor solutions of HF-
induced events in each case displayed on a Hudson source-type plot
458
REFERENCES
Brzovic, A., Skarmeta, J., Blanco, B., Dunlop, R. and Sepulveda, M.P. (2017). Sub-horizontal faulting
mechanism for large rock bursts at the El-Teniente mine. Proceedings of the 9th International
Symposium on Rockbursts and Seismicity in Mines. Vallejos, J.A. (ed). Santiago, Chile. pp. 124–132.
Catalan A., Onederra, I. and Chitombo, G. (2017). Evaluation of intensive preconditioning in block and
panel caving – part II, quantifying the effect on seismicity and draw rates. Mining Technology,
DOI: 10.1080/14749009.2017.1319527.
Gischig, V.S. and Preisig, G. (2015). Hydro-fracturing versus hydro-shearing: A critical assessment of
two distinct reservoir stimulation mechanisms. Proceedings of the 13th International Congress of Rock
Mechanics. ISRM, Montréal, Canada.
Hanks, T. C. and Kanamori, H. (1979). A moment magnitude scale. Journal of Geophysical Research, 84 (5),
pp. 2348–2350.
Hudson, J., Pearce, R. and Rogers, R. (1989). Source Type plot for inversion of the moment tensor. Journal
of Geophysical Research, 94(B1), 765–774.
Mendecki, A. J. (1997). Seismic Monitoring in Mines. Chapman and Hall, Cambridge.
Rojas, E. and Landeros, P. (2017). Hydraulic fracturing applied to tunnels development at El-Teniente
mine. Proceedings of the 9th International Symposium on Rockbursts and Seismicity in Mines. Vallejos,
J.A. (ed). Santiago, Chile. pp. 244–250.
Van As, A. and Jeffrey, R.G. (2000). Hydraulic fracturing as a cave inducement technique at NorthParkes
Mines. Proceedings of MassMin 2000. Mang, H.A., Rammerstorfer, F.G. and Eberhardsteiner, J.
(eds). AusIMM, Brisbane, Queensland. pp. 165-172.
459
Denver Julian Birch
Mine Seismologist
Institue of Mine Seismology
Denver is a Mine Seismologist for the Institute of Mine Seismology in Hobart, Australia. He began his
career as a Junior Scientist at the Council for Geoscience, South Africa in 2009 where his work included
studies on site response, seismic disturbances in residential areas, seismic wave attenuation and
modelling fluid-induced seismicity. In 2013 he graduated Cum Laude with a Master of Science degree
from the University of Pretoria Department of Physics and in 2017 joined the Institute of Mine
Seismology. Denver has since been providing routine and advanced seismological services to the
mining industry with a focus on rock mass response to hydraulic fracturing in underground mines.
460
