1 of 25

Enabling full-waveform inversion to recover salt bodies in challenging conditions: A field data application ��

Abdullah Alali and Tariq Alkhalifah

2 of 25

Motivation

Enabling FWI to recover salt bodies in challenging condition

2

(Duffy et.al, 2022)

Duffy, Oliver B., et al. "The role of salt tectonics in the energy transition: An overview and future challenges." (2022).

geothermal

3 of 25

Full-waveform inversion (FWI) with salt bodies

Enabling FWI to recover salt bodies in challenging condition

3

(Kalita et.al, 2019)

FWI

BP salt model

Initial model

Inversion result

Kalita, Mahesh, et al. "Regularized full-waveform inversion with automated salt flooding." Geophysics 84.4 (2019): R569-R582.

4 of 25

Full-waveform inversion (FWI) with salt bodies

Enabling FWI to recover salt bodies in challenging condition

4

Top to bottom approach

1.6 Hz

Long offset ~ 20 km

FWI

(Shen et al., 2017)

Shen, Xukai, et al. "Salt model building at Atlantis with full-waveform inversion." SEG Technical Program Expanded Abstracts 2017. Society of Exploration Geophysicists, 2017. 1507-1511.

5 of 25

Objectives

Enabling FWI to recover salt bodies in challenging condition

5

Initial model

Imaging -Picking ToS

Flooding

Imaging – Picking BoS

Unflooding

Top-to-bottom:

Objectives:

    • Automate salt identification

Utilize deep learning(U-net) for flooding and unflooding

​

2. Reduce the imaging cost

Reconstruct the salt in multi-scale frequency FWI

​

    • Mitigate the low frequencies and long offset requirement

Depends on the velocity estimation by U-net

​

6 of 25

1D Training workflow

Enabling FWI to recover salt bodies in challenging condition

6

Network Input

(FWI)

Network Output

(Flooding/unflooding)

7 of 25

Generating training data

Enabling FWI to recover salt bodies in challenging condition

7

 

Random 1D models

Adding salts

FWI

  • Randomly added a salt layer of 4.5 km/s to 50% of the models

​

  • 10000 Random 1D models
  • Minimum velocity 1.5 km/s and it is generally increasing with depth

8 of 25

Validation Samples

Enabling FWI to recover salt bodies in challenging condition

8

Flooding 1

Initial for FWI

FWI

Target

Prediction

9 of 25

Validation Samples

Enabling FWI to recover salt bodies in challenging condition

9

Flooding 2

Initial for FWI

FWI

Target

Prediction

10 of 25

Validation Samples

Enabling FWI to recover salt bodies in challenging condition

10

Unflooding

Initial for FWI

FWI

Target

Prediction

11 of 25

BP 2004 salt model (central part)

Enabling FWI to recover salt bodies in challenging condition

11

Max. Offset 6 km

Min. Frequency 3 Hz

12 of 25

BP inversion – FWI

Enabling FWI to recover salt bodies in challenging condition

12

Flooding 1

Flooding 2

Unflooding

FWI

FWI

Initial

model

Freq < 7 Hz

Freq < 10 Hz

Freq < 15 Hz

13 of 25

BP inversion - flooding 1

Enabling FWI to recover salt bodies in challenging condition

Flooding 1

Flooding 2

Unflooding

FWI

FWI

Initial

model

Freq < 7 Hz

Freq < 10 Hz

Freq < 15 Hz

14 of 25

BP inversion – FWI

Enabling FWI to recover salt bodies in challenging condition

14

Flooding 1

Flooding 2

Unflooding

FWI

FWI

Initial

model

Freq < 7 Hz

Freq < 10 Hz

Freq < 15 Hz

15 of 25

BP inversion - flooding 2

Enabling FWI to recover salt bodies in challenging condition

15

Flooding 1

Flooding 2

Unflooding

FWI

FWI

Initial

model

Freq < 7 Hz

Freq < 10 Hz

Freq < 15 Hz

16 of 25

BP inversion – FWI

Enabling FWI to recover salt bodies in challenging condition

16

Flooding 1

Flooding 2

Unflooding

FWI

FWI

Initial

model

Freq < 7 Hz

Freq < 10 Hz

Freq < 15 Hz

17 of 25

BP inversion - Unflooding

Enabling FWI to recover salt bodies in challenging condition

17

Flooding 1

Flooding 2

Unflooding

FWI

FWI

Initial

model

Freq < 7 Hz

Freq < 10 Hz

Freq < 15 Hz

18 of 25

BP inversion – Final FWI

Enabling FWI to recover salt bodies in challenging condition

18

Flooding 1

Flooding 2

Unflooding

FWI

FWI

Final

FWI

Initial

model

Freq < 7 Hz

Freq < 10 Hz

Freq < 15 Hz

19 of 25

Gulf of Mexico (Mississippi Canyon) data

Enabling FWI to recover salt bodies in challenging condition

19

Shot gather

Frequency spectrum

Max. offset: 4.8 Km

Min. frequency: ~5 Hz

20 of 25

GOM inversion

Enabling FWI to recover salt bodies in challenging condition

20

Flooding 1

Flooding 2

Unflooding

FWI

FWI

Initial model

Freq < 7.5 Hz

Freq < 10 Hz

Freq < 15 Hz

21 of 25

GOM Final inversion and RTM image

Enabling FWI to recover salt bodies in challenging condition

21

Final Inversion

RTM image

FWI and RTM overlay

22 of 25

GOM Final inversion and RTM image

Enabling FWI to recover salt bodies in challenging condition

22

FWI and RTM overlay

Common Angle Gathers

23 of 25

Conclusions

Enabling FWI to recover salt bodies in challenging condition

23

  • A fully automatic salt reconstruction workflow

​

- Efficient 1D training

​

- Prevent adding cost courtesy of the multiscale FWI

​

- Challenging synthetic and field dataset

    • a poor initial velocity model
    • limited offset
    • No low frequencies

​

​

​

24 of 25

  • KAUST for supporting this research
  • SWAG group members for insightful discussions
  • The KAUST Supercomputing Laboratory for the supercomputing resources
  • The sponsors of the DeepWave consortium for their support

​

​

Acknowledgments

25 of 25

Abdullah Alali

Tariq Alkhalifah

08-06-2023

Thank you for listening!