American Oil and Gas Reporter - August 2015 - 72

SpecialReport: Hydraulic Fracturing Technology
FIGURE 1
Microseismic Events from Original Source Mechanism Analysis
4,000

3,000

2,000

1,000

0 ft

1,000

2,000

Dip-Slip Class (35, 85, -99)
Strike-Slip Class (160, 90, 180)

3,000

4,000

and the red events that extend away from
the wellbore in a sharp linear path are
the strike-slip class.
The two distinct event classifications
make this well a good candidate to test
automatic moment tensor inversion techniques. In the original processing, these
two pure shear classes were able to describe all events well enough to correct
for the radiation patterns necessary to locate the events and distinguish between
the two very different fracture styles.
When the same data are calculated by
automatic full moment tensor inversion,
it should show the same basic solution,
but with a more robust continuous catalog
of source mechanisms.
A simple k-means clustering technique
was applied to the six components of the
moment tensor solution to match the
original picking method (where only a
few classes of source mechanism were
calculated in the event catalog) to a fully
deterministic source mechanism catalog.
The objective of this clustering technique
was to determine whether the numerous
full moment tensor solutions would group
in the same general pattern as that of the
original method. If the patterns are similar,
it serves to validate the automatic moment
tensor solution as being at least as accurate
as the traditionally accepted hand-picking
method.
Figure 2 displays the results from the
72 THE AMERICAN OIL & GAS REPORTER

same Eagle Ford well using data reprocessed with full automatic moment
tensor inversion, again classifying events
through a simple k-means cluster algorithm. Note that the blue and red colors
have been switched to emphasize dis-

tinctions between the two figures, which
are essentially identical. Accordingly, in
Figure 2, the blue events are the strikeslip class and the red events are the dipslip class.
The results are similar to the original
results from the manual picking method,
showing that the automatic method can
match the accuracy of the manual picking
method in identifying the fault from the
standard fracture events. In this case, the
k-means cluster matched exactly, with
the exception of only two events.
Furthermore, instead of having two
moment tensors to describe these two
classes of events, the results from the automatic method have 4,500 moment tensors describing every event, which makes
it possible to apply additional analysis
techniques for a more in-depth evaluation
of the recorded microseismic data.
The ability to automatically determine
moment tensors in real time can be very
helpful when trying to avoid faults and
large natural features, such as the one in
the Eagle Ford example dataset. Since a
source mechanism is needed for a properly
imaged location in surface microseismic,
it could be difficult to identify this feature
immediately in the dataset because erroneous locations would obscure the linearity
of the feature. The change would be apparent only by looking directly at raw

FIGURE 2
Microseismic Events from Continuous Full Moment
Tensor Inversion Analysis
4,000

3,000

2,000

1,000

0 ft

1,000

2,000

3,000

4,000

k-Means Cluster 1
k-Means Cluster 2
American Oil and Gas Reporter - August 2015

Table of Contents for the Digital Edition of American Oil and Gas Reporter - August 2015
