Professional Documents
Culture Documents
Wavelet Estimation For Surface Multiple Attenuation Using A Simulated Annealing Algorithm - Dr. Arthur Weglein
Wavelet Estimation For Surface Multiple Attenuation Using A Simulated Annealing Algorithm - Dr. Arthur Weglein
org/
SP2.4
Cambridge Research, UK
THE SURFACE MULTIPLE ATTENUATION METHOD
SUMMARY
with
D,(m) = O(w)
=2
(2)
where the data without surface multiples, is obtained by
the sum of a series of terms,
1 being the highest order
multiple existing in the input data. For example, the series
with five terms is free of all multiples from first to fourth
order. is just the pre-stack seismic data with source and
receiver array and ghost effects removed. The higher order
terms remove the multiples, order by order. They are
obtained by the recursive equation (2).
and
are
two-dimensional matrices corresponding to frequency slices
of three-dimensional data cubes where the other two
dimensions correspond to the positions of the source and the
receiver of each particular trace. 0 is a matrix operator
obtained from D1 by applying a frequency dependent obliquity
factor, an extrapolation from the depths of sources and
receivers to the free surface datum and the wavelet
deconvolution. In certain applications, when dealing with
small to moderate angles of propagation, which roughly
corresponds to short offsets compared to depth of reflectors,
angle dependent effects such as the removal of array and
ghost efects from and the obliquity factor applied to 0 may
be neglected. In this case the extrapolation should not be
applied to the operator 0. Since is the input data with no
modification, 0 is just the deconvolved version of When
the obliquity and other factors are neglected the wavelet to
be used in the deconvolution is an effective wavelet
containing the frequency dependence of these factors.
INTRODUCTION
The inverse scattering series aproach to surface
multiple attenuation does not rely on periodicity or differential
moveout, nor does it require a model of subsurface structure
or velocities. However, the effectiveness of this approach
depends on the satisfaction of certain prerequisites, for
example, an estimated source wavelet, compensation for
missing near traces, and source and receiver deghosting.
Experience indicates that estimating the source wavelet is a
particularly important precondition. This experience has been
shared by others, e.g., Verschuur et al. (1988), Fokkema and
Van den Berg (1990) and Dragoset (1992), using similar or
closely related surface multiple attenuation methods.
The idea is to use the sensitivity of the multiple
attenuation method to source signature to turn the situation
around, that is, when the multiples have been attenuated the
wavelet has been well estimated. A criteria is needed to
judge when multiple suppression is working. The criteria is
based on the assumption that when multiples are removed
from the data, there is less energy than when multiples are
present. Verschuur et. al. (1989) first introduced this criteria
in association with their surface multiple suppression
rnethod. In their work, a gradient aproach was used to search
for the minimum of energy. In this work the same criteria,
that is, object function, is used, but instead of the gradient
approach, a simulated annealing
algorithm (Kirkpatric
et. al., 1983) is used to search for the global minimum.
Downloaded 05/05/14 to 129.7.16.11. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
Figures 6.a and 6.b show the energy of the data at each
temperature level for the first and second steps of the
optimization process, respectively. In the first steps 45 levels
were used with 45 moves per level. The best wavelet was
found in the last level although a reasonably stable state was
reached after level 33. In the second step 140 temperature
levels were used with 140 moves per level. The best estimate
corresponds to level 128 but a stable state was reached
around level 65. The need for more levels with more moves
in the second step is due to the fact that in this case there are
more parameters to be optmized so there are more points in
the parameter space to be visited. Also, since the second
step is the final one, it is pushed toward lower levels of
temperature to assure that the estimate goes as close as
possible to the optimal point.
Figure 7 shows a shot from marine real data and Figure
8 shows the same shot after the final step of the wavelet
estimation. Figure 9 shows the amplitude spectrum of the
best wavelet. The phase spectrum was again estimated as a
straight line in both steps leading to a final time shift of 3.4
ms and a phase rotation of 233.8 degrees. Figure 10 shows
the wavelet. It should be mentioned that it still has the efects
of source and receiver ghosts plus a 90 degrees phase
rotation that should be corrected together with the application
of the obliquity factor. In this real data example no angle
dependent correction was applied. All of the variables that
control the optimization process such as initial temperatures,
decay rates and number of temperature levels, were the
same in both examples.
Downloaded 05/05/14 to 129.7.16.11. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
ACKNOWlEDGEMENTS
The authors would like to acknowledge Petrobras and
Schlumberger for support and permission to publish this
work.
REFERENCES
Carvalho, P.M, Weglein, A.B., and Stolt, R.H, 1991, Examples
of a nonlinear inversion method based on the T matrix of
scattering theory: Application to multiple suppression: 61st
Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded abstracts.
Carvalho, P.M., 1992, Metodo de
de
relacionadas corn a superficie livre
em
linear de dados sismicos: Tese de doutorado,
Universidade Federal da Bahia.
Dragoset, W.H., 1992, Surface multiple attenuation: Thecnical
report, Western Geophysical.
Fokkema, J.T., and Van den Berg, P.M., 1990, Removal of
surface-related wave phenomena: the marine case: 60th Ann.
Internat. Mtg., Soc. Expl. Geophys., Expanded abstracts.
Kirkpatric, S., Gelatt, C.D., Jr., and Vecchi, M.P., 1983,
Optimization by simullated annealing: Science, 220, 671-680.
Rothman, D.H., 1986, Automatic estimation of large residual
statics correction: Geophysics, 51, 332-346.
Sen, M.K., Stoffa, P.L., 1991, Nonlinear one-dimensional
seismic waveform inversion using simulated annealing:
Geophysics, 56, 1624-I 638.
Verschuur, D.J., Herrmann, P., Kinneging, N.A., Wapenaar,
C.P.A., and Berkhout, A.J., 1988, Elimination of surface related
multiple reflected and converted waves: 58th Ann. Internat.
Mtg., Soc. Expl. Geophys., Expanded abstracts.
125
0
Frequency (Hz)
Figure 3: Amp litude spectra of the wave ets after the first
step (dashed line) and the second step (( otted line) of the
estimation process was applied to data from Figure 1, shown
togheter with the exact spectrum (solid line).
125
Frequency (Hz)
Figure 4: Phase spectra corresponding to the amplitude
spectra shown in Figure 3.
Downloaded 05/05/14 to 129.7.16.11. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/