US 12,189,075 B2
Building scalable geological property models using machine learning algorithms
Mehran Hassanpour, Houston, TX (US); Gaetan Bardy, Artiguelouve (FR); and Genbao Shi, Sugar Land, TX (US)
Assigned to Landmark Graphics Corporation, Houston, TX (US)
Appl. No. 17/585,441
Filed by Landmark Graphics Corporation, Houston, TX (US)
PCT Filed Dec. 3, 2019, PCT No. PCT/US2019/064262
§ 371(c)(1), (2) Date Jan. 26, 2022,
PCT Pub. No. WO2021/040763, PCT Pub. Date Mar. 4, 2021.
Claims priority of provisional application 62/891,740, filed on Aug. 26, 2019.
Prior Publication US 2023/0367031 A1, Nov. 16, 2023
Int. Cl. G01V 20/00 (2024.01); G06N 3/091 (2023.01)
CPC G01V 20/00 (2024.01) [G06N 3/091 (2023.01)] 12 Claims
OG exemplary drawing
 
1. A method of predicting rock properties at a selectable scale, the method comprising:
receiving coordinates of location of respective sample points of a plurality of sample points;
receiving measurement data for each sample point of the plurality of sample points, the measurement data being associated with one or more measurements or measurement interpretations at the location of the sample point;
receiving, for each sample point of the plurality of sample points, a scale that indicates a measurement volume and a resolution used to obtain the one or more measurements, the measurements interpretation associated with the sample point, or a combination thereof, wherein different measurement volumes are received for different sample points of the plurality of sample points;
training a deep neural network (DNN) by applying the received coordinates, the resolution, the measurement data, and the measurement volume associated with each sample point of the plurality of sample points;
associating the sample point with a rock property as a function of the received coordinates, the measurement data, and the measurement volume applied for the sample point; and
configuring the DNN to receive a request point and generate rock property data for the request point, the request point including coordinates and a selectable measurement volume, the rock property data being determined for the request point as a function of the coordinates and the selectable measurement volume.