Classification and control of detected drilling vibrations using machine learning

Patent metadata

Field	Value
Publication number	US-12116879-B2
Application number	US-202017130189-A
Country	US
Kind code	B2
Filing date	Dec 22, 2020
Priority date	Dec 22, 2020
Publication date	Oct 15, 2024
Grant date	Oct 15, 2024

Abstract

A vibrational disfunction machine learning model trainer trains a vibrational disfunction classifier to identify one or more types of vibrational disfunction, or normal drilling, based on measurements of at least one of displacement, velocity, acceleration, angular displacement, angular velocity, and angular acceleration acquired for the drill bit. The vibrational disfunction machine learning model trainer trains the algorithm based on data sets corresponding to characteristic behavior for one or more types of vibrational disfunction and normal drilling. The vibrational disfunction classifier operates in real time, and can operate at the drill bit and communicate vibrational disfunction identification in real time, allowing mitigation of vibrational disfunction through adjustment of drilling parameters.

First claim

The invention claimed is:

1. A method comprising:
generating input vectors across multiple time windows based, at least partly, on data collected from sensors associated with a drill bit,
wherein the data collected from the sensors correspond to at least two axes of the drill bit,
wherein the input vectors comprise rotational velocities across a corresponding time window and movement of the drill bit in each axis across the corresponding time window;
inputting each of the input vectors into a trained machine learning model to classify vibrational behavior represented by the input vector;
determining whether output from the trained machine learning model classifies the vibrational behavior of the drill bit as disfunctional and indicates a cause of the vibrational behavior; and
indicating a first vibrational disfunction based, at least partly, on the output of the trained machine learning model, the first vibrational disfunction including a type of the vibrational behavior and the cause of the vibrational behavior.

2. The method of claim 1 , wherein the data collected from the sensors comprises at least one of displacement, velocity, acceleration, rotational displacement, rotational velocity, and rotational acceleration in at least one of a time domain or a frequency domain for the drill bit.

3. The method of claim 1 , wherein indicating the first vibrational disfunction comprises communicating the first vibrational disfunction.

4. The method of claim 3 , wherein communicating the first vibrational disfunction comprises communicating the first vibrational disfunction via mud pulse telemetry.

5. The method of claim 1 , further comprising mitigating the first vibrational disfunction.

6. The method of claim 5 , wherein mitigating the first vibrational disfunction comprises adjusting at least one of a drilling parameter and a drill bit design parameter.

7. The method of claim 1 , wherein the trained machine learning model has been trained to identify a plurality of vibrational disfunctions that comprises at least two of friction-induced stick-slip, cutting-induced stick-slip, friction-and-cutting-induced stick-slip, drillpipe-induced stick-slip, three-dimensional coupled vibrations, high frequency torsional noise, cutting-induced backward whirl, and friction-induced backward whirl.

8. The method of claim 1 , further comprising determining that output of the trained machine learning model classifies the vibrational behavior represented by the input vector as multiple vibrational disfunctions including the first vibrational disfunction.

9. The method of claim 8 , further comprising indicating a second vibrational disfunction in addition to the first vibrational disfunction based on determining that the trained machine learning model output classifies the vibrational behavior represented by the input vector as multiple vibrational disfunctions.

10. The method of claim 1 , further comprising inputting each of the input vectors into a second trained machine learning model to classify vibrational behavior represented by the input vector, wherein the trained machine learning model has been trained to identify a first set of one or more vibrational disfunctions and the second trained machine learning model has been trained to identify a second set of one or more vibrational disfunctions.

11. The method of claim 10 , further comprising inputting each of the input vectors into a third trained machine learning model that has been trained to identify non-disfunctional vibration behavior.

12. The method of claim 10 , further comprising determining classification of the vibrational behavior represented by the input vector based, at least partly, on outputs of the trained machine learning model and the second trained machine learning model.

13. The method of claim 1 , further comprising:
determining an output from a second trained machine learning model classifies the vibrational behavior of the drill bit as a second vibrational disfunction different than the first vibrational disfunction; and
determining whether to arbitrate classification outputs by the trained machine learning model and the second trained machine learning model.

14. A non-transitory, machine-readable medium having instructions stored thereon that are executable by a computing device, the instruction comprising instructions to:
generate input vectors across multiple time windows based, at least partly, on data collected from sensors associated with a drill bit,
wherein the data collected from the sensors correspond to at least two axes of the drill bit,
wherein the input vectors comprise rotational velocities across a corresponding time window and movement of the drill bit in each axis across the corresponding time window;

input each of the input vectors into a trained machine learning model to classify vibrational behavior represented by the input vector;
determine whether output from the trained machine learning model classifies the vibrational behavior of the drill bit as disfunctional and indicates a cause of the vibrational behavior; and
indicate a first vibrational disfunction based, at least partly, on the output of the trained machine learning model, the first vibrational disfunction including a type of the vibrational behavior and the cause of the vibrational behavior.

15. The non-transitory, machine-readable medium of claim 14 , wherein the instructions further comprise instructions to communicate the first vibrational disfunction.

16. The non-transitory, machine-readable medium of claim 14 , wherein the instructions further comprise instructions to mitigate the first vibrational disfunction.

17. The non-transitory, machine-readable medium of claim 14 , wherein the instructions further comprise instructions to:
determine whether output of the trained machine learning model classifies the vibrational behavior represented by the input vector as multiple vibrational disfunctions including the first vibrational disfunction.

18. The non-transitory, machine-readable medium of claim 14 , wherein the instructions further comprise instructions to,
input each of the input vector into a second trained machine learning model to classify vibrational behavior represented by the input vector,
wherein the trained machine learning model has been trained to identify a first set of one or more vibrational disfunctions and the second trained machine learning model has been trained to identify a second set of one or more vibrational disfunctions.

19. An apparatus comprising:
a processor; and
a computer-readable medium having instructions stored thereon that are executable by the processor to cause the apparatus to,
generate input vectors across multiple time windows based, at least partly, on data collected from sensors associated with a drill bit,
wherein the data collected from the sensors correspond to at least two axes of the drill bit,
wherein the input vectors comprise rotational velocities across a corresponding time window and movement of the drill bit in each axis across the corresponding time window;

input each of the input vectors into a trained machine learning model to classify vibrational behavior represented by the input vector;
determine whether output from the trained machine learning model classifies the vibrational behavior of the drill bit as disfunctional and indicates a cause of the vibrational behavior; and
indicate a first vibrational disfunction based, at least partly, on the output of the trained machine learning model, the first vibrational disfunc

Assignees

• Halliburton Energy Services Inc

Inventors

• Chen Shilin

Classifications

• E21B47/18

  through the well fluid {, e.g. mud pressure pulse telemetry} · CPC title

• G06N20/20

  Ensemble learning · CPC title

• E21B2200/22

  Fuzzy logic, artificial intelligence, neural networks or the like · CPC title

• E21B47/013

  Devices specially adapted for supporting measuring instruments on drill bits · CPC title

• E21B44/00

  Automatic control systems specially adapted for drilling operations, i.e. self-operating systems which function to carry out or modify a drilling operation without intervention of a human operator, e.g. computer-controlled drilling systems; Systems specially adapted for monitoring a plurality of drilling variables or conditions · CPC title