Processing of the Field
Data using Predictive
Deconvolution
Yuqing Chen Aydar Zaripov Dias Urozayev
King Abdullah University of Science and Technology
08/12/2015
Outline
1) Experiment Description
2) Objective
3) Processing Methodology
4) Results and Interpretation
5) Conclusions
Outline
1) Experiment Description
2) Objective
3) Processing Methodology
4) Results and Interpretation
5) Conclusions
Area of Study
Seismic profile AB is located along the road between
the stadium and the construction site.
Construction
A
B
Outline
1) Experiment Description
2) Objective
3) Processing Methodology
4) Results and Interpretation
5) Conclusions
Objectives:
Problem : noise (harmonic)
Solution : Use Prediction deconvolution
to eliminate harmonic noise meanwhile
compress wavelet (improve resolution)
Outline
1) Experiment Description
2) Objective
3) Processing Methodology
4) Results and Interpretation
5) Conclusions
Deconvolution
Convolutional model (The forward problem):
x(t )  w  t   r  t 
x(t) – recorded seismogram
w(t) – seismic wavelet
r(t) – reflectivity
Deconvolution (The inverse problem):
e(t )  f  t   x  t 
e(t) – Earth’s impulse respoones
Least Square Deconvolution


    y  t   d  t     a   x  t     d  t 
t 0
t 0   0

m n
2
m n
m
x(t) – Recorded seismogram
a(t) – Filter
y(t) – Actually output
d(t) – Desired output





a   x  t     d  t    0



a  s  a  s  t 0   0

m n
2
m
m
mn
mn
 0
t 0
t 0
  a    x  t    x  t  s    d  t  x  t  s 
2
Predictive Deconvolution
Noise attenuated
Periodic event (multiple,
harmonic noise)
Prediction Filter
10
Predictive Deconvolution


  t  a     x t  a   xˆ t  a     x t  a   a   x t   
t 0
t 0   0

T
2
m n
m
2
: Predicted time-advanced seismogram
: Actual time-advanced seismogram
x(t) : current and past seismogram
a(t) : Predict Filter
We can predict the predictable part of the seismogram
like multiples and harmonic noise, etc.
Outline
1) Experiment Description
2) Objective
3) Processing Methodology
4) Results and Interpretation
5) Conclusions
Results and Interpretation
3 critical parameter in predictive deconvolution:
(1) Predict length
(2) Filter length (3) Whiten noise
A cosine
harmonic
noise
His
autocorrelation
map
Results and Interpretation
Use the second maximum value of autocorrelation
map.
Results and Interpretation
Second
maximum
Results and Interpretation
Shot120 (Raw data)
Results and Interpretation
Shot120 (After bandpass)
Results and Interpretation
Shot120 (After PEF_Predict length: 30)
Results and Interpretation
Shot120 (After PEF_Predict length: 2)
Results and Interpretation
Shot120 (After PEF_Predict length: 60)
Results and Interpretation
Shot 40 (After Bandpass)
Results and Interpretation
Shot 40 (After PEF)
Results and Interpretation
3 critical parameter in predictive deconvolution:
(1) Predict length
(2) Filter length
(3) Whiten noise
1. The filter length should at least longer than the predict
length
2. We chose the filter length also from the spectrum.
Seismic trace
Results and Interpretation
50
200
100
300
150
400
Results and Interpretation
3 critical parameter in predictive deconvolution:
(1) Predict length (2) Filter length
Whitening
0.01%
(3) Whiten noise
Results and Interpretation
3 critical parameter in predictive deconvolution:
(1) Predict length (2) Filter length
Whitening
0.05%
(3) Whiten noise
Conclusion
•
•
•
•
•
Predictable noise is mitigated
Improved temporal resolution
Minimum phase assumption
Random reflectivity assumption
High level of noise restricts the
implementation
Recommendations
• 2D predictive deconvolution
• Compute different PEF’s for different
segments of a seismogram
• Change seismogram to minimum
phase before using PEF