System and method for diagnosing borehole structure variances using independent component analysis

1. A well inspection method, comprising:

transmitting an acoustic signal from a well inspection tool into a well structure;

detecting one or more return signals using at least one receiver of the well inspection tool;

generating, using at least one processor, variable density log (VDL) data from the one or more return signals, wherein the VDL data comprises a plurality of waveforms in a time domain;

determining a number of independent components comprised in the VDL data based at least in part variances identified in the VDL data;

decomposing the plurality of waveforms into multiple components using independent component analysis (ICA) and the number of independent components determined in the VDL data, the multiple components associated with one or more local structure variances of the well structure; and

determining characteristics of the well structure based in part on patterns or features associated with the one or more independent components from the multiple components.

2. The method of claim 1 , wherein the characteristics determined of the well structure comprise one or more of: a cement bonding condition, a free pipe condition, or a casing collar condition.

3. The method of claim 1 , wherein the one or more local structure variances of the well structure comprise variation of values in a time scale corresponding to a cement bonding condition, a free pipe condition, and a casing collar condition.

4. The method of claim 1 , further comprising:

determining one or more energy fractions from the one or more independent components; and

applying energy fraction patterns or features within the one or more energy fractions to confirm the patterns or features identified for the one or more independent components.

5. The method of claim 1 , further comprising:

transforming the one or more independent components from the multiple components into a frequency domain representation; and

determining the characteristics of the well structure based on frequency patterns or features in the frequency domain representation of the one or more independent component.

6. The method of claim 1 , further comprising:

modeling responses associated with the characteristics of the well structure into historical data; and

enabling further independent components and corresponding well structure characteristics to be determined using the historical data.

7. The method of claim 1 , further comprising:

modeling responses associated with the characteristics of the well structure into lab-generated data by variations applied to the patterns or features associated with the one or more independent components to vary the characteristics of the well structure; and

enabling further independent components and corresponding well structure characteristics to be determined using the lab-generated data.

8. The method of claim 1 , further comprising:

analyzing the one or more independent components using a machine learning model trained on training data of historical or lab-generated independent components, the historical or lab-generated independent components corresponding to known characteristics of a plurality of well structures; and

enabling inferences from the machine learning model, the inferences associated with the characteristics of the well structures based in part on one or more known characteristics of a plurality of well structures.

9. The method of claim 1 , wherein the VDL log data comprises raw data from multiple receivers, the raw data stored for use as historical data or modified for use as lab-generated data, the historical data or the lab-generated data to generate further independent components that correspond to further well structure characteristics.

10. The method of claim 1 , further comprising:

identifying the variances in the VDL data using principal component analysis (PCA) of the VDL data.

11. A system to be used in well inspection, comprising:

a transmitter within a well inspection tool to transmit an acoustic signal into a well structure;

at least one receiver of the well inspection tool to detect one or more return signals; and

at least one processor to execute instructions to cause the system to:

generate variable density log (VDL) data from the one or more return signals, wherein the VDL data comprises a plurality of waveforms in a time domain;

determine a number of independent components comprised in the VDL data based at least in part variances identified in the VDL data;

decompose the plurality of waveforms into multiple components associated with one or more local structure variances of the well structure using independent component analysis (ICA) and the number of independent components determined in the VDL data; and

determine characteristics of the well structure based in part on patterns or features associated with one or more independent components from the multiple components.

12. The system of claim 11 , wherein the characteristics determined of the well structure comprise one or more of a cement bonding condition, a free pipe condition, or a casing collar condition.

13. The system of claim 11 , wherein the one or more local structure variances of the well structure comprise variation of values in a time scale corresponding to a cement bonding condition, a free pipe condition, and a casing collar condition.

14. The system of claim 11 , wherein the at least one processor to execute the instructions further causes the system to:

determine one or more energy fractions from the one or more independent components; and

apply energy fraction patterns or features, from the one or more energy fractions, to confirm the patterns or features associated with the one or more independent components.

15. The system of claim 11 , wherein the at least one processor to execute the instructions further causes the system to:

transform the one or more independent components from the multiple components into a frequency domain representation; and

determine the characteristics of the well structure based on frequency patterns or features in the frequency domain representation of the one or more independent component.

16. The system of claim 11 , wherein the at least one processor to execute the instructions further causes the system to:

model responses associated with the characteristics of the well structure into historical data; and

enable further independent components and corresponding well structure characteristics to be determined using the historical data.

17. The system of claim 11 , wherein the at least one processor to execute the instructions further causes the system to:

model responses associated with the characteristics of the well structure into lab-generated data by variations applied to the patterns or features associated with the one or more independent components to vary the characteristics of the well structure; and

enable further independent components and corresponding well structure characteristics to be determined using the lab-generated data.

18. The system of claim 11 , wherein the at least one processor to execute the instructions further causes the system to:

analyze the one or more independent components using a machine learning model trained on training data of historical or lab-generated independent components, the historical or lab-generated independent components corresponding to known characteristics of a plurality of well structures; and

enable inferences from the machine learning model, the inferences associated with the characteristics of the well structures based in part on one or more known characteristics of a plurality of well structures.

19. The system of claim 11 , wherein the VDL data comprises raw data from multiple receivers, the raw data stored for use as historical data or modified for use as lab-generated data, the historical data or the lab-generated data to generate further independent components that correspond to further well structure characteristics.

20. The system of claim 11 , wherein the at least one processor to execute the instructions further causes the system to:

identify the variances in the VDL data using principal component analysis (PCA) of the VDL data.