Geoscience Reference
In-Depth Information
pressure change during an event was observed during the
transition from the constant pressure period to the con-
stant flow period (E0p) at time 1796 s (Figure 5.20). This
event had a pressure increase of 13.3 kPa, resulting in an
SNR of 9.8. All other events had higher SNRs, ranging
from about 22 at E3p to 108 at E5p. This process results
in the conclusion that all of the observed pressure changes
were due to a physical change in the system and were not
due to noise. Combined with the electrical data, it is clear
that the pressure changes were caused by seal breakage
events that lead to a burst-like fluid movement.
A noise analysis of the electrical potential signals was
also performed. The noise baseline was established after
DC offset and trend removal, reference channel subtrac-
tion, and after the constant flow was initiated, but before
the onset of rapid electrical activity. Note that when the
reference channel was subtracted, all residual correlated
noises were removed, but the uncorrelated noise added
in quadrature. This resulted in a net reduction of overall
noise in all channels but channel 13 where the uncorre-
lated noise was dominant over the correlated noise com-
ponent; the reference channel removal caused the noise
in this channel to increase slightly. The mean RMS noise
level for all channels was computed, resulting in the
observation that the noise level on channel 13 was more
than 9 times greater than the mean of the noise of all
the other channels. The noise level on channel 13 is
computed to be 0.121 mV RMS, and the mean noise level
of all channels (with the exception of channel 13) is
0.0134 mV RMS. The inspection of the waveforms
shown in Figure 5.17 indicates that the main events of
interest have high voltages relative to the noise level.
Since each channel represents a spatial measurement
point on the cement block, only the channels that contrib-
ute to the peak voltage response are relevant to the SNR
calculations. Events E2 and E3 have peak channel SNRs
of over 200 and over 1900, respectively.We conclude that
the SNR of these channels does not contribute signifi-
cantly to dipole location uncertainty.
Positional uncertainty analysis of the E2 GA current
dipole solution was performed by adding Gaussian noise
to the computed forward solution using the E2 dipole
moment model vector. This new noisy measurement
vector was used as the input to the GA, where a new
dipole solution would be computed. The new solution
would be compared with the initial solution. Three levels
of noise were used in this analysis: 1, 5, and 10%
noise levels were used. These noise levels were computed
at as a random Gaussian additive voltage computed from
the mean voltage of each channel. Solutions for the three
noise cases were found at the following dipole point
positions: 2 for 1%, 62, for 5%, and 61 for 10% noise.
These solution points have a displacement from the
initial solution of one radial dipole point in the position
Thresholded dipole position for top array
Thresholded dipole position for back array
0.12
0.12
0.11
0.11
0.10
0.10
0.09
0.09
0.08
0.08
Initial solution = 42
1% Noise = 2
5% Noise = 62
10% Noise = 61
0.07
0.07
0.06
0.06
0.20
0.21
0.22
0.23
0.24
0.25
0.26
-0.08
-0.09
-0.1
-0.11
-0.12
-0.13
X
-position (m)
Z
-position (m)
(a)
(b)
Figure 5.35
Localization of the causative source of current and noise analysis.
a)
and
b)
These figures show the spatial positions
of the dipoles found during the localization uncertainty test. Note that the solutions cluster near the initial solution found during
the inversion process. The bias in the +
y
and
z
directions may indicate that the true solution may be between the solutions
with noisy data and the solution found initially. (
See insert for color representation of the figure
.)
−
