Chaiwoot Boonyasiriwat 3D Multisource Full-Waveform Inversion using Quasi-Monte Carlo Phase Encoding Chaiwoot Boonyasiriwat Jan. 7, 2010
Outline Introduction to Multisource Technology Phase Encoding Multisource Full-Waveform Inversion Quasi-Monte Carlo Phase Encoding Numerical Results 3D SEG/EAGE Overthrust Model Summary Future Work Acknowledgment 1
Introduction: Multisource Tech. Migration Delayed-shot migration (Zhang et al., 2005) Random phase encoding (Romero et al., 2000) Plane-wave migration (Liu et al., 2006; Shan and Biondi, 2008; Shan et al., 2009) Full-Waveform Inversion (FWI) Plane-wave FWI (Vigh and Starr, 2008) Random phase encoding (Krebs et al., 2009; Zhan et al., 2009) 2
Single-Source Method 3
Linear Phase Encoding—Planewave Decomposition Multisource Method Linear Phase Encoding—Planewave Decomposition 4
Multisource Method Random Phase Encoding 5
CSG vs SSG CSG SSG 6
Outline Introduction to Multisource Technology Phase Encoding Multisource Full-Waveform Inversion Quasi-Monte Carlo Phase Encoding Numerical Results 3D SEG/EAGE Overthrust Model Summary Future Work Acknowledgment 7
Evaluate misfit function and compute gradient Conventional FWI Model No Observed Data Evaluate misfit function and compute gradient Evaluate misfit function Perturb Model No Search criterion Convergence criterion Yes Done Yes 8
Evaluate misfit function and compute gradient Multisource FWI Model No Encoded Data Evaluate misfit function and compute gradient Evaluate misfit function Perturb Model No Search criterion Convergence criterion Yes Done Yes 9
2D Multisource FWI Krebs Method (ExxonMobil): one SSG, random source polarity, dynamic encoding 8
2D Multisource FWI Zhan Method (UTAM): multiple SSGs, random time shift, static encoding, deblurring filter 9
Krebs Source Configuration 3D Multisource FWI Krebs Source Configuration 10
Closely Packed Source Configuration 3D Multisource FWI Closely Packed Source Configuration 11
Loosely Packed Source Configuration 3D Multisource FWI Loosely Packed Source Configuration 12
Quasi-Monte Carlo Source Configuration 3D Multisource FWI Quasi-Monte Carlo Source Configuration 13
Static vs Dynamic Configuration Iteration 1 2 14
Outline Introduction to Multisource Technology Phase Encoding Multisource Full-Waveform Inversion Quasi-Monte Carlo Phase Encoding Numerical Results 3D SEG/EAGE Overthrust Model Summary Future Work Acknowledgment 15
3D SEG/EAGE Overthrust Model 16
Numerical Results True Velocity Model 17
Initial Velocity Model Numerical Results Initial Velocity Model 18
Velocity Model from Static QMC Numerical Results Velocity Model from Static QMC 19
Velocity Model from Dynamic QMC Numerical Results Velocity Model from Dynamic QMC 20
Velocity Model from Krebs Method Numerical Results Velocity Model from Krebs Method 21
Outline Introduction to Multisource Technology Multisource Full-Waveform Inversion Numerical Results 3D SEG/EAGE Overthrust Model Summary Future Work Acknowledgment 22
Summary 3D multisource FWI using multiple SSGs is compared with multisource FWI using one SSG. Using multiple SSGs and a dynamic QMC phase encoding provides a better-quality velocity model than a static QMC phase encoding and Krebs method. Theoretical speedups need to be verified. More reliable timing results will be presented later. 23
Outline Introduction to Multisource Technology Multisource Full-Waveform Inversion Numerical Results 3D SEG/EAGE Overthrust Model Summary Future Work Acknowledgment 24
Future Work Compare various multisource configurations. Compare random phase encoding with plane-wave encoding. Apply random time shifts and deblurring filter. Apply 3D multisource FWI to field data. 25
Acknowledgment Sponsors of 2009 UTAM consortium HPC: Aron Ahmadia and Mark Cheeseman Shaheen: Iain Georgeson and Jonathan Anderson Multisource: Ge Zhan and Wei Dai Workstation: Benoit Marchand KAUST: Jerry Schuster 26