# US 2022/0035069 A1 · Systems and methods for identifying subsurface features as a function of position in a subsurface volume of interest

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2022/0035069 A1
- Document ID: US20220035069A1
- Status: Published application
- Kind: A1
- Application number: 16/945,517
- Filed: 2020-07-31
- Published: 2022-02-03
- Earliest claimed: 2020-07-31
- Assignees: Chevron U.S.A. Inc.; The Texas A&M University System
- Applicants: Chevron U.S.A. Inc.; The Texas A&M University System
- Inventors: Shuxing Cheng; Zhao Zhang; Kellen Leigh Gunderson; Reynaldo Cardona; Zhangyang Wang; Ziyu Jiang
- Web page: https://patentgrep.com/patent/US20220035069A1

## CPC

- G01V 20/00: Geomodelling in general
- G06V 10/454: {Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]}
- G06V 10/82: using neural networks
- G06F 18/214: Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F 2111/10: Numerical modelling
- G06F 2218/08: Feature extraction
- G01V: GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS (means for indicating the location of accidentally buried, e.g. snow-buried, persons A63B29/02)
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Systems and methods are disclosed for identifying subsurface features as a function of position in a subsurface volume of interest. Exemplary implementations may include obtaining target subsurface data; obtaining a conditioned subsurface feature model; applying the conditioned subsurface feature model to the target subsurface data, which may include generating convoluted target subsurface data by convoluting the target subsurface data; generating target subsurface feature map layers by applying filters to the convoluted target subsurface data; detecting potential target subsurface features in the target subsurface feature map layers; masking the target subsurface features; and estimating target subsurface feature data by linking the masked subsurface features to the target subsurface feature data.

## Claims

### Claim 1

A computer-implemented method for identifying subsurface features as a function of position in a subsurface volume of interest, the method being implemented in a computer system that comprises a physical computer processor and a non-transitory storage medium, the method comprising:
obtaining, from the non-transitory storage medium, target subsurface data comprising a subsurface energy value;
obtaining, from the non-transitory storage medium, a conditioned subsurface feature model comprising a conditioned subsurface feature extraction sub-model, a conditioned subsurface object detection sub-model, and multiple conditioned network sub-models, the conditioned subsurface feature model having been generated by applying an initial subsurface feature model to training subsurface data and training subsurface feature data, wherein the training subsurface data comprises subsurface energy values, and wherein the training subsurface feature data comprises the training structural identification values defined by structural identification parameters and corresponding training categorization values defined by categorization parameters and one of the training structural identification values;
applying, with the physical computer processor, the conditioned subsurface feature model to the target subsurface data by:
generating, with the conditioned feature extraction sub-model, convoluted target subsurface data by convoluting the target subsurface data;
generating, with the conditioned feature extraction sub-model, target subsurface feature map layers by applying filters to the convoluted target subsurface data, wherein a given target subsurface feature map layer corresponds to a given spatial scale of the target subsurface data;
detecting, with the conditioned subsurface object detection sub-model, potential target subsurface features in the target subsurface feature map layers;
masking, with the conditioned subsurface object detection sub-model, the target subsurface features; and
estimating, with the multiple conditioned network sub-models, target subsurface feature data by linking the masked subsurface features to the target subsurface feature data.

### Claim 2 (depends on 1)

The computer-implemented method of claim 1, wherein the computer system comprises a display, and wherein the computer-implemented method further comprises:
generating, with the physical computer processor, a representation of geological structures as a function of position in the subsurface volume of interest using visual effects to depict at least some of the target subsurface feature data; and
displaying the representation via the display.

### Claim 3 (depends on 1)

The computer-implemented method of claim 1, wherein the target subsurface feature data comprises a target structural identification value defined by structural identification parameters and a corresponding target categorization value defined by categorization parameters and the target structural identification value.

### Claim 4 (depends on 3)

The computer-implemented method of claim 3, wherein the target structural identification values indicate whether or not there is a subsurface feature.

### Claim 5 (depends on 3)

The computer-implemented method of claim 3, wherein the target categorization values indicate one of a subsurface feature category and a subsurface feature sub-category.

### Claim 6 (depends on 1)

The computer-implemented method of claim 1, wherein the conditioned subsurface feature model comprises an adversarial loss function defined by
min
f
D
,
f
F
⁢
L
D
⁡
(
f
D
⁡
(
f
F
⁡
(
X
)
)
,
Y
D
)
-
⁢
L
G
⁡
(
f
G
⁡
(
f
F
⁡
(
X
)
)
,
Y
G
)
where fD refers to the initial object detection sub-model, fF refers to the initial feature extraction sub-model, YD refers to a ground truth for detection, LD refers to the cross-entropy loss, fG refers to the initial network sub-models, LG refers to the geometrical location classification loss, YG refers to the ground truth of geometrical locations, and refers to a weight.

### Claim 7 (depends on 1)

The computer-implemented method of claim 1, wherein convoluting the target subsurface data comprises applying a kernel to the target subsurface data, wherein the kernel is a weighted matrix moved over the target subsurface data to generate dot products over each region covered by the weighted matrix.

### Claim 8

A computer-implemented method for training a subsurface feature model to identify subsurface features as a function of position in a subsurface volume of interest, the method being implemented in a computer system that comprises a physical computer processor and a non-transitory storage medium, the method comprising:
obtaining, from the non-transitory storage medium, training subsurface data comprising a subsurface energy value and corresponding training subsurface feature data, wherein the training subsurface feature data comprises a training structural identification value defined by structural identification parameters and a corresponding training categorization value defined by categorization parameters and the training structural identification value;
obtaining, from the non-transitory storage medium, an initial subsurface feature model comprising an adversarial loss function, wherein the initial subsurface feature model comprises an initial subsurface feature extraction sub-model, an initial subsurface object detection sub-model, and multiple initial network sub-models;
generating, with the physical computer processor, a conditioned subsurface feature model by training the initial subsurface feature model using the training subsurface data, the training subsurface feature data, and the adversarial loss function, wherein the conditioned subsurface feature model comprises a conditioned subsurface feature extraction sub-model, a conditioned subsurface object detection sub-model, and multiple conditioned network sub-models; and
storing the conditioned subsurface feature model in the non-transitory storage medium.

### Claim 9 (depends on 8)

The computer-implemented method of claim 8, further comprising:
obtaining, from the non-transitory storage medium, target subsurface data comprising a subsurface energy value;
applying, with the physical computer processor, the conditioned subsurface feature model to the target subsurface data by:
generating, with the conditioned feature extraction sub-model, target subsurface feature map layers by convoluting the target subsurface data, wherein a given target subsurface feature map layer corresponds to a given spatial scale of the target subsurface data;
detecting, with the conditioned subsurface object detection sub-model, potential target subsurface features in the target subsurface feature map layers;
classifying, with the conditioned subsurface object detection sub-model, the detected target subsurface features; and
linking, with the multiple network sub-models, the classified subsurface features to the target subsurface feature data.

### Claim 10 (depends on 9)

The computer-implemented method of claim 9, wherein the computer system comprises a display, and wherein the computer-implemented method further comprises:
generating, with the physical computer processor, a representation of geological structures as a function of position in the subsurface volume of interest using visual effects to depict at least some of the target subsurface feature data; and
displaying the representation via the display.

### Claim 11 (depends on 10)

The computer-implemented method of claim 10, wherein the target subsurface feature data comprises a target structural identification value defined by structural identification parameters and a corresponding target categorization value defined by categorization parameters and the target structural identification value.

### Claim 12 (depends on 11)

The computer-implemented method of claim 11, wherein the target structural identification values indicate whether or not there is a subsurface feature.

### Claim 13 (depends on 11)

The computer-implemented method of claim 11, wherein the target categorization values indicate one of a subsurface feature category and a subsurface feature sub-category.

### Claim 14 (depends on 8)

The computer-implemented method of claim 8, wherein the adversarial loss function is
min
f
D
,
f
F
⁢
L
D
⁡
(
f
D
⁡
(
f
F
⁡
(
X
)
)
,
Y
D
)
-
⁢
L
G
⁡
(
f
G
⁡
(
f
F
⁡
(
X
)
)
,
Y
G
)
where fD refers to the initial object detection sub-model, fF refers to the initial feature extraction sub-model, YD refers to a ground truth for detection, LD refers to a cross-entropy loss, fG refers to the initial network sub-models, LG refers to a geometrical location classification loss, YG refers to a ground truth of geometrical locations, and refers to a weight.

### Claim 15

A system for identifying subsurface features as a function of position in a subsurface volume of interest, the system comprising:
non-transitory storage medium; and
a physical computer processor configured by machine readable instructions to:
obtain, from the non-transitory storage medium, target subsurface data comprising a subsurface energy value;
obtain, from the non-transitory storage medium, a conditioned subsurface feature model comprising a conditioned subsurface feature extraction sub-model, a conditioned subsurface object detection sub-model, and multiple conditioned network sub-models, the conditioned subsurface feature model having been generated by applying an initial subsurface feature model to training subsurface data and training subsurface feature data, wherein the training subsurface data comprises subsurface energy values, and wherein the training subsurface feature data comprises the training structural identification values defined by structural identification parameters and corresponding training categorization values defined by categorization parameters and one of the training structural identification values;
apply, with the physical computer processor, the conditioned subsurface feature model to the target subsurface data by:
generating, with the conditioned feature extraction sub-model, target subsurface feature map layers by convoluting the target subsurface data, wherein a given target subsurface feature map layer corresponds to a given spatial scale of the target subsurface data;
detecting, with the conditioned subsurface object detection sub-model, potential target subsurface features in the target subsurface feature map layers;
classifying, with the conditioned subsurface object detection sub-model, the detected target subsurface features; and
linking, with the multiple network sub-models, the classified subsurface features to the target subsurface feature data.

### Claim 16 (depends on 15)

The system of claim 15, wherein the system further comprises a display, and wherein the physical computer processor is further configured by machine readable instructions to:
generate, with the physical computer processor, a representation of geological structures as a function of position in the subsurface volume of interest using visual effects to depict at least some of the target subsurface feature data; and
display the representation via the display.

### Claim 17 (depends on 15)

The system of claim 15, wherein the subsurface feature model comprises one of a mask R-CNN network, a feature pyramids network, an object detection network, a neural network, a random forest model, a support vector machine, a regression, and a Bayesian network.

### Claim 18 (depends on 15)

The system of claim 15, wherein the adversarial loss function is
min
f
D
,
f
F
⁢
L
D
⁡
(
f
D
⁡
(
f
F
⁡
(
X
)
)
,
Y
D
)
-
⁢
L
G
⁡
(
f
G
⁡
(
f
F
⁡
(
X
)
)
,
Y
G
)
where fD refers to the initial object detection sub-model, fF refers to the initial feature extraction sub-model, YD refers to a ground truth for detection, LD refers to a cross-entropy loss, fG refers to the initial network sub-models, LG refers to a geometrical location classification loss, YG refers to a ground truth of geometrical locations, and refers to a weight.

### Claim 19 (depends on 15)

The system of claim 15, wherein the target subsurface feature data comprises a target structural identification value defined by structural identification parameters and a corresponding target categorization value defined by categorization parameters and the target structural identification value.

### Claim 20 (depends on 15)

The system of claim 15, wherein convoluting the target subsurface data comprises applying a kernel to the target subsurface data, wherein the kernel is a weighted matrix moved over the target subsurface data to generate dot products over each region covered by the weighted matrix.
