Granted Patent B2

US 12,729,618 · App. 18/162,599 · Granted Sep 8, 2026

Inventors: Klemens Katterbauer (Dhahran, SA); Abdallah A. Alshehri (Dhahran, SA); Alberto Marsala (Venice, IT); Ali Abdallah Alyousef (Dhahran, SA)

Assignee: SAUDI ARABIAN OIL COMPANY

E21B44/00E21B7/04E21B47/02E21B2200/22

Abstract

Claims (41)

1 . A method for predicting conditions ahead of a drill bit while drilling a well, the method comprising:

performing, using a machine learning model, a classification of formation properties ahead of the drill bit, based on data comprising logging-while-drilling (LWD) data obtained while drilling the well;

applying the classification of the formation properties to perform a geosteering of the drill bit;

further comprising, prior to performing the classification;

training the machine learning model;

prior to training the machine learning model;

weighting, in training data used for the training, different types of data based on quality;

wherein the quality is assessed using a signal-to-noise ratio.

2 . The method of claim 1 , wherein the classification is performed in real-time, while drilling the well.

3 . The method of claim 1 , wherein the classification of the formation properties comprises a classification of at least one selected from a group consisting of lithology and saturation.

4 . The method of claim 1 , wherein the classification of the formation properties comprises a quantification of an uncertainty of the classification.

5 . The method of claim 1 , wherein the data further comprise at least one selected from a group consisting of electromagnetic data and seismic data.

6 . The method of claim 1 , wherein the LWD data comprise at least one selected from a group consisting of sonic data, deep azimuthal resistivity data, porosity data, density data, pressure data, and temperature data.

7 . The method of claim 1 , further comprising, prior to performing the classification:

reconciling the data to eliminate inconsistencies between different types of data in the data.

8 . The method of claim 1 , further comprising, prior to performing the classification:

removing outliers from the data.

9 . The method of claim 1 , wherein the machine learning model is a deep belief network based on Restricted Boltzmann Machines.

10 . The method of claim 1 , wherein training data used for the training originates from one selected from a group consisting of an offset well and the well.

11 . The method of claim 1 , further comprising after training the machine learning model:

evaluating the machine learning model; and

retraining the machine learning model when performance is considered insufficient, based on the evaluation of the machine learning model.

12 . A system for predicting conditions ahead of a drill bit while drilling a well, the system comprising:

a drilling system for drilling the well, the drilling system comprising the drill bit and a drill bit logging tool; and

a control system configured to:

perform, using a machine learning model, a classification of formation properties ahead of the drill bit, based on data comprising logging-while-drilling (LWD) data obtained from the drill bit logging tool while drilling the well using the drill bit;

applying the classification of the formation properties to perform a geosteering of the drill bit;

further comprising, prior to performing the classification:

training the machine learning model;

prior to training the machine learning model:

weighting, in training data used for the training, different types of data based on quality;

wherein the quality is assessed using a signal-to-noise ratio.

13 . The system of claim 12 , wherein the classification is performed in real-time, while drilling the well.

14 . A non-transitory machine-readable medium comprising a plurality of machine-readable instructions executed by one or more processors, the plurality of machine-readable instructions causing the one or more processors to perform operations comprising:

performing, using a machine learning model, a classification of formation properties ahead of a drill bit while drilling a well, based on data comprising logging-while-drilling (LWD) data obtained while drilling the well;

applying the classification of the formation properties to perform a geosteering of the drill bit;

further comprising, prior to performing the classification:

training the machine learning model;

prior to training the machine learning model:

weighting, in training data used for the training, different types of data based on quality;

wherein the quality is assessed using a signal-to-noise ratio.

Assignments (1)

Continuity (1)

References Cited (55)