IP Library › Granted Patent US 12,560,739
Granted Patent B2
US 12,560,739 · App. 17/574,293 · Granted Feb 24, 2026

Machine learning of geology by probabilistic integration of local constraints

Inventors: Bradley Clark Wallet (Norman, OK); Rajesh Goteti (Katy, TX); Konstantin Osypov (Houston, TX)
Assignee: Saudi Arabian Oil Company
G01V20/00G06F30/27G06F2111/10
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,560,739
App. No.
17/574,293
Granted
Feb 24, 2026
Kind
B2
Abstract

Systems and methods include a computer-implemented method: Seismic data is gathered for a reservoir with unknown fractured and unfractured areas. A structural model is generated. A geomechanical model is built. Geomechanically-estimated fractured areas are determined using the geomechanical model, including: areas where fractures are not likely to exist based on a likelihood lower than a first threshold likelihood, areas where fractures are likely to exist based on a likelihood greater than a second threshold likelihood, and areas where fracturing is unknown based on a likelihood between the first threshold likelihood and the second threshold likelihood. Machine learning-based estimates of a likelihood of a fracture of each area of the reservoir are determined using machine learning based on mathematical calculations of the seismic data. Fractured and unfractured areas are determined based on where fractures are likely to exist or not using the geomechanically-estimated fractured areas, machine learning-based likelihoods, and Bayes' Rule.

Claims (58)

1 . A computer-implemented method, comprising:

gathering seismic data for a reservoir with unknown fractured and unfractured areas;

generating a structural model based on the seismic data for the reservoir;

building a geomechanical model based on the structural model;

determining, using the geomechanical model, geomechanically-estimated fractured areas including: areas where fractures are not likely to exist based on a likelihood lower than a first threshold likelihood; areas where fractures are likely to exist based on a likelihood greater than a second threshold likelihood; and areas where fracturing is unknown based on a likelihood between the first threshold likelihood and the second threshold likelihood;

determining, using machine learning based on mathematical calculations of the seismic data, machine learning-based estimates of a likelihood of a fracture of each area of the reservoir, wherein a likelihood lower than a third threshold likelihood indicates an unfractured area, and wherein a likelihood greater than a fourth threshold likelihood indicates a fractured area; and

determining, using the geomechanically-estimated fractured areas, the machine learning-based likelihoods, and Bayes' Rule, the fractured and unfractured areas based on where fractures are likely to exist and not exist.

2 . The computer-implemented method of claim 1 , wherein determining where fractures are likely to exist comprises determining that a fracture exists when:

the geomechanical model indicates a likelihood greater than the second threshold likelihood; and

the seismic data includes a likelihood between a third threshold likelihood and a fourth threshold likelihood.

3 . The computer-implemented method of claim 1 , wherein determining where fractures are likely to exist comprises determining that a fracture exists when:

the geomechanical model indicates a likelihood within a range of the first threshold likelihood and the second threshold likelihood; and

the seismic data includes a likelihood greater than the fourth threshold likelihood.

4 . The computer-implemented method of claim 1 , wherein determining where fractures are likely to exist comprises determining that an existence of a fracture is unsure when:

the geomechanical model indicates a likelihood less than the second threshold likelihood; or

the seismic data includes a likelihood less than a third threshold likelihood.

5 . The computer-implemented method of claim 1 , wherein determining where fractures are likely to exist comprises using a 3-by-3 look-up table with fracture likelihoods of No, Ambiguous, and Yes.

6 . The computer-implemented method of claim 1 , wherein determining where fractures are likely to exist comprises using a 5-by-5 look-up table with fracture likelihoods of Very Unlikely, Unlikely, Ambiguous, Likely, and Very Unlikely.

7 . The computer-implemented method of claim 1 , wherein determining where fractures are likely to exist comprises using an asymmetric look-up table that bases fracture likelihood combinations on one or both of criteria and probabilities and an application of one or more thresholds.

8 . A non-transitory, computer-readable medium storing one or more instructions executable by a computer system to perform operations comprising:

gathering seismic data for a reservoir with unknown fractured and unfractured areas;

generating a structural model based on the seismic data for the reservoir;

building a geomechanical model based on the structural model;

determining, using the geomechanical model, geomechanically-estimated fractured areas including: areas where fractures are not likely to exist based on a likelihood lower than a first threshold likelihood; areas where fractures are likely to exist based on a likelihood greater than a second threshold likelihood; and areas where fracturing is unknown based on a likelihood between the first threshold likelihood and the second threshold likelihood;

determining, using machine learning based on mathematical calculations of the seismic data, machine learning-based estimates of a likelihood of a fracture of each area of the reservoir, wherein a likelihood lower than a third threshold likelihood indicates an unfractured area, and wherein a likelihood greater than a fourth threshold likelihood indicates a fractured area; and

determining, using the geomechanically-estimated fractured areas, the machine learning-based likelihoods, and Bayes' Rule, the fractured and unfractured areas based on where fractures are likely to exist and not exist.

9 . The non-transitory, computer-readable medium of claim 8 , wherein determining where fractures are likely to exist comprises determining that a fracture exists when:

the geomechanical model indicates a likelihood greater than the second threshold likelihood; and

the seismic data includes a likelihood between a third threshold likelihood and a fourth threshold likelihood.

10 . The non-transitory, computer-readable medium of claim 8 , wherein determining where fractures are likely to exist comprises determining that a fracture exists when:

the geomechanical model indicates a likelihood within a range of the first threshold likelihood and the second threshold likelihood; and

the seismic data includes a likelihood greater than the fourth threshold likelihood.

11 . The non-transitory, computer-readable medium of claim 8 , wherein determining where fractures are likely to exist comprises determining that an existence of a fracture is unsure when:

the geomechanical model indicates a likelihood less than the second threshold likelihood; or

the seismic data includes a likelihood less than a third threshold likelihood.

12 . The non-transitory, computer-readable medium of claim 8 , wherein determining where fractures are likely to exist comprises using a 3-by-3 look-up table with fracture likelihoods of No, Ambiguous, and Yes.

13 . The non-transitory, computer-readable medium of claim 8 , wherein determining where fractures are likely to exist comprises using a 5-by-5 look-up table with fracture likelihoods of Very Unlikely, Unlikely, Ambiguous, Likely, and Very Unlikely.

14 . The non-transitory, computer-readable medium of claim 8 , wherein determining where fractures are likely to exist comprises using an asymmetric look-up table that bases fracture likelihood combinations on one or both of criteria and probabilities and an application of one or more thresholds.

15 . A computer-implemented system, comprising:

one or more processors; and

a non-transitory computer-readable storage medium coupled to the one or more processors and storing programming instructions for execution by the one or more processors, the programming instructions instructing the one or more processors to perform operations comprising:

gathering seismic data for a reservoir with unknown fractured and unfractured areas;

generating a structural model based on the seismic data for the reservoir;

building a geomechanical model based on the structural model;

determining, using the geomechanical model, geomechanically-estimated fractured areas including: areas where fractures are not likely to exist based on a likelihood lower than a first threshold likelihood; areas where fractures are likely to exist based on a likelihood greater than a second threshold likelihood; and areas where fracturing is unknown based on a likelihood between the first threshold likelihood and the second threshold likelihood;

determining, using machine learning based on mathematical calculations of the seismic data, machine learning-based estimates of a likelihood of a fracture of each area of the reservoir, wherein a likelihood lower than a third threshold likelihood indicates an unfractured area, and wherein a likelihood greater than a fourth threshold likelihood indicates a fractured area; and

determining, using the geomechanically-estimated fractured areas, the machine learning-based likelihoods, and Bayes' Rule, the fractured and unfractured areas based on where fractures are likely to exist and not exist.

16 . The computer-implemented system of claim 15 , wherein determining where fractures are likely to exist comprises determining that a fracture exists when:

the geomechanical model indicates a likelihood greater than the second threshold likelihood; and

the seismic data includes a likelihood between a third threshold likelihood and a fourth threshold likelihood.

17 . The computer-implemented system of claim 15 , wherein determining where fractures are likely to exist comprises determining that a fracture exists when:

the geomechanical model indicates a likelihood within a range of the first threshold likelihood and the second threshold likelihood; and

the seismic data includes a likelihood greater than the fourth threshold likelihood.

18 . The computer-implemented system of claim 15 , wherein determining where fractures are likely to exist comprises determining that an existence of a fracture is unsure when:

the geomechanical model indicates a likelihood less than the second threshold likelihood; or

the seismic data includes a likelihood less than a third threshold likelihood.

19 . The computer-implemented system of claim 15 , wherein determining where fractures are likely to exist comprises using a 3-by-3 look-up table with fracture likelihoods of No, Ambiguous, and Yes.

20 . The computer-implemented system of claim 15 , wherein determining where fractures are likely to exist comprises using a 5-by-5 look-up table with fracture likelihoods of Very Unlikely, Unlikely, Ambiguous, Likely, and Very Unlikely.

Assignments (3)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Jun 1, 2022
From: ARAMCO SERVICES COMPANY
To: SAUDI ARAMCO UPSTREAM TECHNOLOGY COMPANY
Reel/Frame 060065/0727 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Jun 1, 2022
From: SAUDI ARAMCO UPSTREAM TECHNOLOGY COMPANY
To: SAUDI ARABIAN OIL COMPANY
Reel/Frame 060065/0776 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Jan 13, 2022
From: WALLET, BRADLEY CLARK; GOTETI, RAJESH; OSYPOV, KONSTANTIN
To: ARAMCO SERVICES COMPANY
Reel/Frame 058639/0455 →
Continuity (2)
Provisional Application 63143712 · Jan 29, 2021
Related Publication 20220244423A1 · Aug 4, 2022
References Cited (44)
US 9310500B2 · Kacewicz et al. · 2016 [cited by applicant]
US 10487649B2 · Tilke et al. · 2019 [cited by applicant]
US 20040073370A1 · Dasgupta · 2004 [cited by applicant]
US 20090303834A1 · Sengupta · 2009 [cited by applicant]
US 20130238304A1 · Glinsky · 2013 [cited by applicant]
US 20140233352A1 · Kacewicz · 2014 [cited by applicant]
US 20160202374A1 · Kainkaryam et al. · 2016 [cited by applicant]
US 20160320512A1 · Zhao · 2016 [cited by applicant]
US 20170177764A1 · Guy · 2017 [cited by applicant]
US 20170285193A1 · Zhang et al. · 2017 [cited by applicant]
US 20180067220A1 · Flemings et al. · 2018 [cited by applicant]
US 20200183047A1 · Denli et al. · 2020 [cited by applicant]
US 20200200931A1 · Mao et al. · 2020 [cited by applicant]
US 20200278465A1 · Salman et al. · 2020 [cited by applicant]
US 20210018638A1 · Goteti et al. · 2021 [cited by applicant]
US 20210125312A1 · Mulchandani et al. · 2021 [cited by applicant]
WO WO2017035104 · 2017 [cited by applicant]
U.S. Appl. No. 63/143,712, Wallet et al., filed Jan. 29, 2021. [cited by applicant]
Bachrach et al., “Using geomechanical modeling and wide-azimuth data to quantify stress effects and anisotropy near salt bodies in the Gulf of Mexico.” In SEG Technical Program Expanded Abstracts 2008, Society of Explor… [cited by applicant]
Daly et al., “Multipoint Statistics in Reservoir Modelling and in Computer Vision,” Conference Proceedings, EAGE Conference on Petroleum Geostatistics, Sep. 2007, 5 pages. [cited by applicant]
De Oliveira et al., “Learning Particle Physics by Example: Location-Aware Generative Adversarial Networks for Physics Synthesis,” Computing and Software for Big Science, Sep. 2017, 1(4): 1-24, 24 pages. [cited by applicant]
D'Elia et al., “A tree-structured Markov random filed model for Bayesian image segmentation,” IEEE Transactions on Image Processing, 2003, 12:10 (1259-1273), 16 pages. [cited by applicant]
Durrant-Whyte et al., “Simultaneous Localisation and Mapping (SLAM): Part I The Essential Algorithms, ” IEEE Robotics & Automation Magazine, 2006, 13:2 (99-110), 9 pages. [cited by applicant]
Eidsvik et al., “Stochastic reservoir characterization using prestack seismic data,” Geophysics, 2004, 69:4 (978-993), 33 pages. [cited by applicant]
Guo et al., “Principal component spectral analysis,” Geophysics, 2009, 74(4), 9 pages. [cited by applicant]
Heidari et al., “Coupling geomechanical modeling with seismic pressure prediction.” Geophysics 83, No. 5, 2018, 54 pages. [cited by applicant]
Liu et al., “Auto-painter: Cartoon Image Generation from Sketch by Using Conditional Generative Adversarial Networks, ” Neurocomputing, Oct. 2018, 311: 78-87, 31 pages. [cited by applicant]
Mottle et al., “Pattern recognition in spatial data: A new method of seismic explorations for oil and gas in crystalline basement rocks,” Proceedings of the 15th International Conference on Pattern Recognition, 2000, 31… [cited by applicant]
Nikolinakou et al., “Pore-pressure prediction based on seismic velocities coupled with geomechanical modeling.” In 50th US Rock Mechanics/Geomechanics Symposium. American Rock Mechanics Association, 2016, 8 pages. [cited by applicant]
Park et al., “Bayesian active learning with localized priors for fast receptive field characterization,” Advances in Neural Information Processing Systems, 2012, 25, 9 pages. [cited by applicant]
Sengupta et al., “Relationship between velocity and anisotropy perturbations and anomalous stress field around salt bodies.” The Leading Edge 28, No. 5 (2009), 598-605, 7 pages. [cited by applicant]
Tavolara et al., “A modular cGAN classification framework: Application to colorectal tumor detection,” Scientific Reports, Dec. 2019, 9:18969, 8 pages. [cited by applicant]
Ulvmoen et al., “Improved resolution in Bayesian lithology/fluid inversion from prestack seisimic data and well observations: Part 1—Methodology,” Geophysics, 2010, 75:2 (R21-R35), 15 pages. [cited by applicant]
Vanpaemel et al., “Constructive informative model priors using hierarchical methods,” Journal of Mathematical Psychology, 2010, 55, 106-117, 12 pages. [cited by applicant]
Wallet et al., “Latent space modeling of seismic data: An overview,” The Leading Edge, 2009, 28(12), 6 pages. [cited by applicant]
Wallet, “3D Modeling of Carbonate Turbidite Channel-forms Using Curvature: A Workflow,” SEG Technical Program Expanded Abstracts, 2015, 2 pages. [cited by applicant]
Wallet, “Using the image grand tour to visualize fluvial deltaic architectural elements in south Texas, USA,” Interpretation, 2013, 1(1), 13 pages. [cited by applicant]
Wang, “Image Denoising Using Deep CGAN With Bi-Skip Connections,” Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2019, pp. 724-729, 6 pages. [cited by applicant]
Woolrich et al., “Variational Bayes inference of spatial mixture models for segmentation,” IEEE Transactions on Medical Imaging, 2006, 25(10): 1380-1391, 12 pages. [cited by applicant]
Xu et al., “Prediction-CGAN: Human Action Prediction with Conditional Generative Adversarial Networks,” Proceedings of the 27th ACM International Conference on Multimedia, Oct. 2019, pp. 611-619, 9 pages. [cited by applicant]
Zee et al., “3D geomechanical modeling of complex salt structures.” In SIMULIA Customer Conference. 2011, 16 pages. [cited by applicant]
Zhang et al., “Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm,” IEEE Transactions on Medical Imagining, 2001, 20:1, 13 pages. [cited by applicant]
PCT International Search Report and Written Opinion in International Appln. No. PCT/US2022/014341, dated Jun. 1, 2022, 18 pages. [cited by applicant]
Lefranc et al., “Fractured Basement Reservoir Characterization for Fracture Distribution, Porosity and Permeability Prediction,” Adapted from extended abstract prepared in conjunction with oral presentation at AAPG Inte… [cited by applicant]