Method for analyzing cement integrity in casing strings using machine learning

Patent metadata

Field	Value
Publication number	US-10858933-B2
Application number	US-201615575024-A
Country	US
Kind code	B2
Filing date	May 18, 2016
Priority date	May 18, 2015
Publication date	Dec 8, 2020
Grant date	Dec 8, 2020

Abstract

The present disclosure provides methods and systems for analyzing cement integrity in a depth interval of a wellbore having a multiple string casing with an innermost annulus disposed inside at least one outer annulus. The method includes processing ultrasonic data obtained from ultrasonic measurements on the interval of the wellbore to determine properties of the innermost annulus. The method also includes processing sonic data obtained from sonic measurements on the interval of the wellbore to extract features of the sonic data. The features of the sonic data are input to a machine learning processing to determine properties of both the innermost annulus and the least one outer annulus. Additional processing of ultrasonic and sonic data can also be used to determine properties of both the innermost annulus and the least one outer annulus. These properties can be used to analyze cement integrity in the depth interval of the wellbore.

First claim

The invention claimed is:

1. A method for analyzing cement integrity in a depth interval of a cased wellbore traversing a formation, wherein the cased wellbore includes a multiple string casing with an innermost annulus disposed inside at least one outer annulus, the method comprising:
(i) processing ultrasonic data obtained from ultrasonic measurements on the depth interval of the wellbore to determine properties of the innermost annulus;
(ii) processing sonic data obtained from sonic measurements on the depth interval of the wellbore to extract features of the sonic data;
(iii) determining properties of the innermost annulus disposed between an innermost casing and a surrounding casing and the least one outer annulus using machine learning processing, wherein the properties of the innermost annulus and the least one outer annulus include at least one property characterizing state of material in the innermost annulus and at least one property characterizing state of material in the at least one outer annulus, and wherein the features of the sonic data from process (ii) are inputs to the machine learning processing;
(iv) analyzing the cement integrity based on the determined properties.

2. The method of claim 1 , further comprising:
storing the properties of the innermost annulus as determined in process (i) and the properties of the innermost annulus and the at least one outer annulus as determined in process (iii) in computer-readable storage media for analysis of the cement integrity in the interval of the wellbore.

3. The method of claim 1 , wherein the machine learning processing of process (iii) uses the properties of the innermost annulus of process (i) as a constraint in the determination of the properties of the innermost annulus and the at least one outer annulus.

4. The method of claim 1 , wherein the machine learning processing comprises a machine learning classifier.

5. The method of claim 4 , further comprising:
training the machine learning classifier for a variety of anticipated conditions of the innermost annulus and the at least one outer annulus in different formation types and wellbore fluids.

6. The method of claim 5 , wherein the training of the machine learning classifier uses at least one of: (a) synthetic sonic data generated using modelled data with perturbations and noise; and (b) field data.

7. The method of claim 5 , further comprising:
validating the machine learning classifier using field data and synthetic data.

8. The method according to claim 1 , wherein:
the properties of the innermost annulus and the at least one outer annulus include fill states for both the innermost annulus and the at least one outer annulus; and
the fill states represent one of a solid, liquid, and gas phase of the material of the innermost annulus and the at least one outer annulus.

9. The method according to claim 1 , wherein:
the properties of the innermost annulus include bond state for the innermost annulus for the case of solid fill state for the innermost annulus;
the properties of the at least one outer annulus include bond state for the at least one outer annulus for the case of solid fill state for the at least one outer annulus; and
the bond state for the innermost annulus or the at least one outer annulus characterizes interfacial conditions of cement to casing bonding or cement to formation bonding.

10. The method according to claim 1 , wherein the machine learning processing determines properties of both the innermost annulus and the at least one outer annulus that correspond to the features of the sonic data from process (ii).

11. The method according to claim 1 , wherein the features comprise specific attributes of slowness and attenuation dispersions of the sonic data.

12. The method according to claim 11 , wherein the sonic data arises from at least one of a monopole excitation, a dipole excitation, and a quadrupole excitation.

13. The method according to claim 12 , wherein the attributes comprise at least one of:
(a) a number of Stoneley modes arising from a monopole excitation;
(b) a number of casing extensional modes arising from a monopole excitation;
(c) a number of dipole flexural modes arising from a dipole excitation;
(d) a number of cut-off modes arising from a dipole excitation;
(e) presence of formation modes arising from a dipole excitation;
(f) a number of quadrupole modes arising from a quadrupole excitation; and
(g) slowness and attenuation dispersion characteristics of the modes present.

14. The method of claim 1 , wherein the ultrasonic measurements include ultrasonic pulse echo and pitch-catch measurements.

15. The method of claim 14 , wherein process (i) comprises performing an inversion of the ultrasonic data to determine the properties of the innermost annulus.

16. The method of claim 15 , wherein the properties determined by the inversion of the ultrasonic data comprise at least one of:
(a) compressional wavespeed V p as a function of azimuth direction ϕ and axial depth z;
(b) shear wavespeed V s as a function of azimuth direction ϕ and axial depth z;
(c) density ρ as a function of azimuth direction ϕ and axial depth z;
(d) acoustic impedance Z as a function of azimuth direction ϕ and axial depth z;
(e) bond parameters as a function of azimuth direction ϕ and axial depth z;
(f) bond state as a function of azimuth direction ϕ and axial depth z;
(g) annulus fill state as a function of azimuth direction ϕ and axial depth z; and
(h) a measure of casing eccentering as a function of axial depth z.

17. The method of claim 15 , wherein the inversion of the ultrasonic data employs a hierarchical Bayesian graphical model to determine certain properties of the innermost annulus.

18. The method of claim 17 , wherein the certain properties include the fill state and bond state for the innermost annulus.

19. The method of claim 17 , wherein the hierarchical Bayesian graphical model is trained using a forward model relating the properties of the innermost annulus to ultrasonic measurement outputs.

20. The method of claim 17 , wherein the hierarchical Bayesian graphical model incorporates available spatial prior information pertaining to properties of the innermost annulus.

21. The method of claim 1 , further comprising:
(iv) processing sonic data obtained from sonic measurements on the depth interval of the wellbore in order to determine properties of the at least one outer annulus.

22. The method of claim 21 , wherein process (iv) comprises performing an inversion of the sonic data to determine the properties of the least one outer annulus.

23. The method of claim 22 , wherein the properties identified by the inversion of the sonic data comprise at least one of:
(a) compressional wavespeed V p as a function of azimuth direction ϕ and axial depth z;
(b) shear wavespeed V s as a function of azimuth direction ϕ and axial depth z;
(c) density ρ as a function of azimuth direction ϕ and axial depth z;
(e) acoustic impedance Z as a function of azimuth direction ϕ and axial depth z; and
(f) bond parameters as a function of azimuth direction ϕ and axial depth z.

24. The method of claim 22 , wherein the inversion of the sonic data of process (iv) uses the properties of the innermost annulus of process (i) as a constraint in the determination of the properties of the at least one outer annulus.

25. The method of claim 21 , further comprising

Assignees

• Schlumberger Technology Corp

Inventors

• Bose Sandip
• Skataric Maja
• Zeroug Smaine
• Sinha Bikash Kumar

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G01V1/50

  Analysing data · CPC title

• G01V1/44

  using generators and receivers in the same well (G01V1/52 takes precedence) · CPC title

• G06N20/00 Primary

  Machine learning · CPC title

• G01V2210/542

  Casing · CPC title