Methods and systems for processing borehole dispersive waves with a physics-based machine learning analysis

1. A method for processing borehole wave modes, the method comprising:

measuring, by an acoustic logging tool within a borehole passing through a formation, an acoustic wave to generate waveform data of an acoustic wave;

determining wave features from the waveform data, wherein the wave features are associated with a quality of the waveform data;

generating one or more data-driven scale factors by applying a trained model to the wave features, wherein the one or more data-driven scale factors adaptively balance a data-driven process and a model-based process based on the quality of the waveform data;

determining data-driven scale parameters from different wave features by comparing wave features generated by the trained model to directly measured wave features;

determining final parameters of the data-driven scale factor by combining the different wave features;

generating a formation body wave slowness by performing a hybrid process based on the one or more data-driven scale factors, wherein the hybrid process is a combination of the data-driven process and the model-based process, such that the hybrid process is more reliant on the data-driven model when a larger measured frequency length is present;

adjusting a downhole operational parameter based at least in part on the formation body wave slowness, wherein the downhole operational parameter comprises one of a drilling parameter, a logging parameter, a completion parameter, and a production parameter; and

providing the formation body wave slowness to a user interface and/or saving the formation body wave slowness to a log, wherein the formation body wave slowness provides wellbore and/or formation characteristics throughout drilling.

2. The method of claim 1 , further comprising: determining formation characteristics of a borehole environment; and modeling a fluid-filled borehole based on the determined formation characteristics of the borehole environment.

3. The method of claim 2 , further comprising calculating one of a full waveform response of the modeled fluid-filled borehole environment, an Airy-phase frequency, and a cut-off frequency.

4. The method of claim 1 , further comprising calculating one of a semblance value of target waves, an effective data length in frequency, and a minimum effective frequency.

5. The method of claim 1 , further comprising: training a model with synthetic data to generate the trained model; and applying the trained model to field data.

6. The method of claim 5 , further comprising: generating the synthetic data, wherein the synthetic data includes waveform data; adding noise to the generated synthetic data; determining wave features of the synthetic waveform data; determining data-driven scales by inverting the wave features of the synthetic waveform data; and training the model with the wave features of the synthetic waveform data and the data-driven scales.

7. The method of claim 1 , further comprising: determining data-driven scale parameters from different wave features by comparing wave features generated by the model to directly measured wave features; and determining final parameters of the data-driven scale factor by combining the different wave features.

8. The method of claim 1 , wherein the hybrid process comprises performing one of a fully modeling-based process, a fully data-driven process, and a limited data-driven process.

9. The method of claim 8 , wherein the limited data-driven process comprises: generating a set of parameter ranges of the data-driven process from the one or more data-driven scale factors; and performing an inversion processing comprising limiting adjustable parameters in a fixed range.

10. The method of claim 1 , further comprising generating a visualization of one of a waveform of the received acoustic wave, a time-domain semblance map, a misfit, and a fitting quality of the curve.

11. A system for processing borehole wave modes, the system comprising: a processor; and a memory comprising instructions which, when executed, cause the processor to:

measure, by an acoustic logging tool within a borehole passing through a formation, an acoustic wave to generate waveform data of the acoustic wave;

determine wave features from the waveform data, wherein the wave features are associated with a quality of the waveform data; generate one or more data-driven scale factors by applying a trained model to the wave features, wherein the one or more data-driven scale factors adaptively balance a data-driven process and a model-based process based on the quality of the waveform data;

determine data-driven scale parameters from different wave features by comparing wave features generated by the trained model to directly measured wave features;

determine final parameters of the data-driven scale factor by combining the different wave features;

generate a formation body wave slowness by performing a hybrid process based on the one or more data driven-scale factors, wherein the hybrid process is a combination of the data-driven process and the model-based process, such that the hybrid process is more reliant on the data-driven model when a larger measured frequency length is present;

adjusting a downhole operational parameter based at least in part on the formation body wave slowness, wherein the downhole operational parameter comprises one of a drilling parameter, a logging parameter, a completion parameter, and a production parameter; and

providing the formation body wave slowness to a user interface and/or saving the formation body wave slowness to a log, wherein the formation body wave slowness provides wellbore and/or formation characteristics throughout drilling.

12. The system of claim 11 , wherein the memory further comprises instructions which, when executed, cause the processor to: determine formation characteristics of a borehole environment; and model a fluid-filled borehole environment based on the borehole environment.

13. The system of claim 11 , wherein the memory further comprises instructions which, when executed, cause the processor to: calculate one of a semblance value of target waves, an effective data length in frequency, a minimum effective frequency, a full waveform response of a modeled fluid-filled borehole environment, an Airy-phase frequency, and a cut-off frequency.

14. The system of claim 11 , wherein the memory further comprises instructions which, when executed, cause the processor to: generate synthetic data including waveform data; add noise to the generated synthetic data; determine wave features of the synthetic waveform data; determine data-driven scales by inverting the wave features of the synthetic waveform data; train a model with the wave features of the synthetic waveform data and the data-driven scales to generate the trained model; and apply the trained model to field data.

15. The system of claim 11 , wherein the memory further comprises instructions which, when executed, cause the processor to: determine data-driven scale parameters from different wave features by comparing wave features generated by the model to directly measured wave features; and determine final parameters of the data-driven scale factor by combining the different wave features.

16. The system of claim 11 , wherein the hybrid process comprises performing one of a full modeling-based process, a fully data-driven process, and a limited data-driven process, and wherein the memory further comprises instructions which, when executed, cause the processor to: generate a set of parameter ranges of the data-driven processing from the one or more data-driven scale factors; and perform an inversion processing comprising limiting adjustable parameters in a fixed range.