US 12,140,019 B2
Methods for characterizing and evaluating well integrity using unsupervised machine learning of acoustic data
Lingchen Zhu, Medford, MA (US); Sandip Bose, Brookline, MA (US); and Smaine Zeroug, Lexington, MA (US)
Assigned to SCHLUMBERGER TECHNOLOGY CORPORATION, Sugar Land, TX (US)
Appl. No. 16/973,059
Filed by SCHLUMBERGER TECHNOLOGY CORPORATION, Sugar Land, TX (US)
PCT Filed Jun. 6, 2019, PCT No. PCT/US2019/035774
§ 371(c)(1), (2) Date Dec. 8, 2020,
PCT Pub. No. WO2019/236832, PCT Pub. Date Dec. 12, 2019.
Claims priority of provisional application 62/682,249, filed on Jun. 8, 2018.
Prior Publication US 2021/0270127 A1, Sep. 2, 2021
Int. Cl. E21B 47/107 (2012.01); E21B 47/002 (2012.01); E21B 47/005 (2012.01); G01V 1/48 (2006.01); G06N 3/045 (2023.01); G06N 3/047 (2023.01); G06N 3/08 (2023.01)
CPC E21B 47/107 (2020.05) [E21B 47/0025 (2020.05); E21B 47/005 (2020.05); G01V 1/48 (2013.01); G06N 3/045 (2023.01); G06N 3/047 (2023.01); G06N 3/08 (2013.01)] 19 Claims
OG exemplary drawing
 
1. A method for characterizing well integrity of a cased well, the method including:
i) collecting sonic waveform data for acoustic signals received by a receiver array of a sonic logging tool for a depth interval of the cased well;
ii) determining a high-dimensional representation of the sonic waveform data;
iii) determining an intermediate-dimensional representation of the sonic waveform data from the high-dimensional representation of the sonic waveform data using a first unsupervised dimension-reduction machine learning model of a first type that compresses unlabeled high-dimensional sonic waveform data to an intermediate dimension latent variable space;
iv) determining a low-dimensional representation of the sonic waveform data from the intermediate-dimensional representation of the sonic waveform data using a second unsupervised dimension-reduction machine learning model of a second type that compresses intermediate-dimensional sonic waveform data through a single-layer network architecture to a low-dimensional latent variable space;
v) clustering the low-dimensional representation of the sonic waveform data to identify low-dimensional clusters and corresponding low-dimensional cluster labels directly from the low-dimensional representation of the sonic waveform data;
vi) determining a cementing property based on a geophysical representation of the sonic waveform data at a receiver azimuth and depth of the sonic waveform data indicated by the low-dimensional cluster labels, wherein the cementing property is related to cementing integrity of an annulus of the cased well at the depth interval; and
vii) determining a well integrity characterization of the cased well based on the cementing property.