AVA analysis as a machine learning problem€¦ · hydrocarbon*reserves*from*seismic* images.*...
Transcript of AVA analysis as a machine learning problem€¦ · hydrocarbon*reserves*from*seismic* images.*...
-
University of British ColumbiaSLIM
Ben Bougher, SEG 2016
AVA analysis as a machine learning problem
Released to public domain under Creative Commons license type BY (https://creativecommons.org/licenses/by/4.0).Copyright (c) 2016 SLIM group @ The University of British Columbia.
-
Reflection seismology as an unsupervised learning problem
3
Approaches: Physics driven (conventional) Data driven (thesis contributions)
Motivation: Inability to discover hydrocarbons directly from seismic data.
Problem: Automatically segment potential hydrocarbon reserves from seismic images.
-
Angle domain common image gathers
8
θProblem: Need angle dependent reflectivity responses
Solution: Angle domain common image gather migration
x1
x
z1
dept
h
✓
refle
ctiv
ity
Earth Model R(𝜃) at (x1,z1)
-
Scattering theory (Zoeppritz)
9
θ1θ1
θ2
φ1
φ2
Vp1, Vs1, ρ1
Vp2, Vs2, ρ2
Rpp Rps
Tpp
Tps
2
664
RPP
RPS
TPP
TPS
3
775 =
2
6664
� sin ✓1 � cos�1 sin ✓2 cos�2cos ✓1 � sin�1 cos ✓2 � sin�2sin 2✓
VP1VS1
cos2�1⇢2VS22VP1⇢1V 2S1VP 2
cos 2✓1⇢2VS2VP1⇢1V 2S1
cos 2�2
�cos�2 VS1VP1 sin 2�1⇢2VP2⇢1VP1
⇢2VS2⇢1VP1
sin 2�2
3
7775
�1 2
664
sin ✓1
cos ✓1
sin 2✓1
cos 2�1
3
775
Problem: Relate angle dependent reflectivity to rock physics.
Assumption: Ray theory approximation.
Solution:
Problem: Non-linear, not useful for inversion.
R(✓) / VP , VS , ⇢
-
Scattering theory (Shuey)
10
Limitations: Small perturbations over a background trend valid < 30 degrees
Benefits: linear for i and g invert using simple least squares
0 10 20 30 40 50 60 70 80 90
theta [deg]
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Rpp
Zoeppritz2-term Shuey
Rpp(✓) = i(�VP ,�⇢) + g(�VP ,�VS ,�⇢) sin2 ✓
R(𝜃) at (x1,z1)
Shuey approximation
-
Shuey term inversion as a projection
11
✓
Rpp
Rpp(✓) at x1, z1
✓
ADCIG at x1
θ
ADCIGs
Image for every scattering angle. An angle gather for every slice. A reflectivity curve for every point.
-
Shuey term inversion as a projection
11
✓
Rpp
Rpp(✓) at x1, z1
Fit the reflectivity curve using a linear combination of the Shuey vectors. Each reflectivity curve is projected down to two coefficients.
-
Shuey term inversion as a projection
11
i g
i,g at x1
i
g
Reduced the dimensionality of the angle gathers to two coefficients. Projection coefficients can be plotted to analyze the multivariate relationships.
-
Relation to hydrocarbons
12
�0.4 �0.2 0.0 0.2 0.4I
�0.4
�0.2
0.0
0.2
0.4
G
mudrock linesand - shalegas sand - shale
g =i
1 + k
h1� 4 hVSihVP i
⇣ 2m
+ khVSihVP i
⌘i
*m, c, k are geological parameters determined empirically from well logs/laboratory measurements
Hydrocarbon reserves are found from outliers of a crossplot!
Brine-‐saturated sands and shales follow a mudrock line:
Hydrocarbon saturated sands deviate from this trend.
-
Reality bites
13
5 10 15 20 25 30✓ [deg]
�1.6
�1.4
�1.2
�1.0
�0.8
�0.6
�0.4
�0.2
Rpp
⇥104
0 5 10 15 20 25 30 35 40✓ [deg]
0.00
0.05
0.10
0.15
0.20
0.25
Rpp
0 5 10 15 20 25 30✓ [deg]
0
1
2
3
4
5
6
Rpp
0 5 10 15 20 25 30✓ [deg]
0.0
0.1
0.2
0.3
0.4
0.5I
G
Problem: Shuey components can’t explain the features in real data.
Solution: Use unsupervised machine learning to find better projections.
BG dataset Penobscot dataset Mobil Viking dataset
-
Unsupervised learning problem
14
X 2 Rn⇥d
n is number of samples in the image, d is the number of angles
-
Principle component analysis (PCA)
Eigendecomposition of the covariance matrix:
Project onto the eigenvectors with the two largest eigenvalues. Maximizes the variance (a measure of information).
15
C = XTX =1
n
nX
i
xixTi
-
Marmousi II Earth model
17
0 1000 2000 3000 4000x [m]
0
500
1000
1500
2000
z[m
]
vp
12001400160018002000220024002600
0 1000 2000 3000 4000x [m]
0
500
1000
1500
2000z
[m]
vs
0
150
300
450
600
750
900
1050
0 1000 2000 3000 4000x [m]
0
500
1000
1500
2000
z[m
]
rho
105012001350150016501800195021002250
0 1000 2000 3000 4000x[m]
0
500
1000
1500
2000
z[m
]
zero-o↵set rpp
�0.32�0.24�0.16�0.080.000.080.160.240.32
Specifically made for testing amplitude vs. offset analysis
Contains gas-‐saturated sand embedded in shales and brine-‐sands.
-
Seismic modeling
18
0 500 1000 1500 2000 2500 3000 3500 4000x [m]
0
500
1000
1500
2000
z[m
]
0 5 10 15 20 25 30✓ [deg]
500 1000 1500 2000 2500 3000 3500 4000 4500 5000x [m]
0
500
1000
1500
2000
2500
z[m
]
�5 0 5 10 15 20 25 30✓ [deg]
Physically consistent with the Zoeppritz equations
Migrated visco-‐acoustic survey
True reflectivity Migrated seismic
-
Kernel PCA
Problem: Find a non-‐linear projection that provides better discrimination of trends and outliers.
Solution: Use the “kernel-‐trick” to compute PCA in a high-‐dimensional non-‐linear feature space
20
-
Kernel trickPCA can be calculated from the Gramian inner product matrix:
Replace with a kernel
Example c=2, b=1
21
hxi,xji (xi,xj)
�(x) = [1,p2x1,
p2x2, x
21, x
22,p2x1x2]
XXT =
0
BBB@
hx1,x1i hx1,x2i . . . hx1,xnihx2,x1i hx2,x2i . . . hx2,xni
......
. . ....
hxn,x1i hxn,x2i . . . hxn,xni
1
CCCA
(xi,xj) = (xTi xj + b)
c = h�(xi),�(xj)i
-
RecapSituation: Exploring seismic data for anomalous responses. Problem: Physical model can not explain real world data. Solution: Learn useful projections directly from the data. Assessment:
• PCA is equivalent for physically consistent data, but more robust to processing/acquisition artifacts.
• Kernel PCA makes outliers linearly separable from the background.
Next: Automatic segmentation (clustering)
23
-
Each point is initially considered a cluster. Each iteration merges the closest clusters into a larger cluster. Builds a hierarchy until a defined number of classes is reached.
Hierarchical clustering
24
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9�0.1
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0
1
2
3 4
5
6
7
8
9
5 3 1 2 47 96 80
A
BB
C
Unclassified data Classification hierarchy
-
Results - PCA
25
Clustered physically consistent data Clustered migrated data
-
Results - Kernel PCA
26
Clustered physically consistent data Clustered migrated data
-
Summary
Successes: Each projection could segment the reservoir. Kernel PCA provided advantageous multivariate geometries (linearly separable).
Challenges: Clustering is highly sensitive to user chosen parameters. Kernel PCA is computationally expensive and lacks interpretation.
27
-
Robust PCA
Problem: Find a sparse set of outlying reflectivity responses against a background trend. Assumption: The background trend of similar curves is highly redundant, which forms a low rank matrix. Solution:
28
minL,S
kLk⇤ + �kSk1,1 s.t. L+ S = X
-
Results - physically consistent data
29
0 1000 2000 3000 4000x [m]
0
500
1000
1500
z[m
]
-
Results - migrated seismic
30
-
Summary
Successes: Segmentation of reservoir in both images. Physically interpretable segmentation without clustering.
Challenges: Requires tuning of one optimization trade off parameter. Convergence sensitive to the rank of outliers, not well understood.
31
-
Epilogue
Outcome: Generalized a conventional analysis approach using unsupervised learning models. Successful in segmented potential hydrocarbon reserves from seismic data Future: More data, standardized datasets Quantitative benchmarks
44
-
References[1]Zoeppritz, K., 1919, VII b. Über Reflexion und Durchgang seismischer Wellen durch Unstetigkeitsflächen: Nachrichten von der Gesellschaft der Wissenschaften zu Göttingen, Mathematisch-Physikalische Klasse, 1919, 66–84 [2]Shuey, R. T., 1985, A simplification of the Zoeppritz equations: Geophysics, 50, 609–614. [3]Gardner, G. H. F., L. W. Gardner, and A. R. Gregory, 1974, Formation velocity and density; the diagnostic basics for stratigraphic traps: Geophysics, 39, 770–780. [4]Castagna, J. P., M. L. Batzle, and R. L. Eastwood, 1985, Relationships between compressional-wave and shear-wave velocities in clastic silicate rocks: Geophysics, 50, 571–581. [5]Castagna, J. P., H. W. Swan, and D. J. Foster, 1998, Framework for AVO gradient and intercept interpretation: Geophysics, 63, 948–956. [6]Edgar, J., and J. Selvage, 2013, Dynamic Intercept-gradient Inversion
46