A Time-Domain Preconditioned Truncated Newton Approach to Visco-acoustic Multiparameter Full Waveform Inversion

Yang, Pengliang; Brossier, Romain; Métivier, Ludovic; Virieux, Jean; Zhou, Wei

doi:10.1137/17m1126126

Cited by 77 publications

(80 citation statements)

References 69 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The obtained high resolution velocity model provides a good starting point to study further multiparameter FWI. The truncated Newton algorithm and a multi-parameter preconditioner are tools designed to reduce potential inter-parameter cross-talks and obtain more reliable second-order parameters such as density or attenuation (Yang et al, 2018). We seek to exploit these methods on this 3D real field dataset for assessing feasibility and interest of extracting multiple parameters with mitigation of cross-talks.…”

Section: Discussionmentioning

confidence: 99%

“…where D is a spatial differential operator; m gathers the density ρ, the elastic stiffness tensor C (parameterized by P-wave velocity V p and Thomsen parameters ε and δ under vertical transverse isotropic (VTI) approximation (Yang et al, 2018)), and the attenuation factor Q −1 embedded in the anelastic coefficients Y ≈ y Q −1 ; w,w and R † ∆d are the incident field, the adjoint field and the backward projected data residuals. Away from the source location, the gradient of the misfit function is thus given by :…”

Section: Method: Time-domain Vti Viscoacoustic Fwimentioning

confidence: 99%

See 1 more Smart Citation

3-D Time-Domain VTI Viscoacoustic Full Waveform Inversion: Application to Valhall Field Data

Yang¹,

Brossier²,

Métivier³

et al. 2018

Proceedings

View full text Add to dashboard Cite

This work presents a first 3D field data application to the OBC Valhall dataset of a recently developed vertical transverse isotropic (VTI) viscoacoustic full waveform inversion (FWI) engine, incorporating attenuation as a passive parameter. Our implementation uses a checkpointing-assisted reverse-forward simulation (CARFS) algorithm to efficiently build the FWI gradient, appearing more efficient than standard checkpointing techniques. The reverse propagation of the incident field is performed stably in attenuating medium with low memory cost, thanks to the decimation and interpolation strategy. We follow a frequency-content continuation workflow: starting from low frequencies, each FWI takes the refined velocity model and re-estimates a new source wavelet at given frequency band, combined with a random shots subsampling. The inverted velocity model clearly captures channel features and gas cloud, in agreement with former studies (Sirgue 2010, Operto 2015}. Estimated source wavelets using this final inverted model show good consistency over shots. The synthetic data can match the observed seismograms quite well with an anisotropic visco-acoustic forward problem engine. This encouraging singleparameter inversion for a real case motivates multiparameter inversion in the time domain while keeping cross-talk effects small.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Method: Time-domain Vti Viscoacoustic Fwimentioning

confidence: 99%

3-D Time-Domain VTI Viscoacoustic Full Waveform Inversion: Application to Valhall Field Data

Yang¹,

Brossier²,

Métivier³

et al. 2018

Proceedings

View full text Add to dashboard Cite

show abstract

“…Time-domain finite-difference modelling (4th order in space and 2nd order in time) is used to generate acoustic wavefield in VTI media (Duveneck and Bakker, 2011). In the presence of attenuation, three standard linear solid (SLS) mechanisms (Yang et al, 2018) are employed to compute the wavefield. We use the checkpoint-assisted reverse forward simulation (CARFS) technique to obtain the gradient of the objective function with respect to the model parameters (Yang et al, 2016).…”

Section: Inversionmentioning

confidence: 99%

“…Choosing a suitable parameterization for the inversion by analysing radiation patterns can help mitigate the problem (Kamath et al, 2017). In viscous media, although velocity and the attenuation Q p scatter the wavefield equally in all directions (Yang et al, 2018), the π/2 phase-shift introduced by Q p should help resolve the two parameters (Hak and Mulder, 2011). It is, however, still necessary to choose an inversion algorithm with an appropriate preconditioner to ensure meaningful parameter updates.…”

Section: Introductionmentioning

confidence: 99%

Effect and Reconstruction of Attenuation in Acoustic FWI - Method and Field Data Application

Kamath

Brossier

Métivier

et al. 2019

81st EAGE Conference and Exhibition 2019 Workshop Programme

View full text Add to dashboard Cite

Main objectives Apply FWI to a 3D field dataset and compare joint (V P0 − Q p) inversion with monoparameter (V P0)only visco-acoustic inversion. Investigate influence of pseudo-Hessian on joint inversion and of the Q p-model on the monoparameter inversion. New aspects New dataset on which we are applying a data sub-sampling strategy and checkpoint assisted forward reverse simulation (CARFS) to perform multiparameter FWI.

show abstract

“…The inverse Hessian operator can resolve part of the trade-off problem, at least theoretically (Kamei and Pratt, 2013;Métivier et al, 2013), but it is not practical to use due to its cost. A preconditioned truncated Newton method was proposed to resolve the trade-off issue between parameters (Yang et al, 2018).…”

Section: Introductionmentioning

confidence: 99%

A sequential inversion with outer iterations for the velocity and the intrinsic attenuation using an efficient wavefield inversion

Song

Alkhalifah

2020

GEOPHYSICS

View full text Add to dashboard Cite

Full-waveform inversion (FWI) has become a popular method to retrieve high-resolution subsurface model parameters. It is a highly non-linear optimization problem based on minimizing the misfit between the observed and predicted data. For intrinsically attenuating media, wave propagation experiences significant loss of energy. Thus, for better data fitting, it is crucial sometimes to consider attenuation in FWI. Viscoacoustic FWI aims at achieving a joint inversion of the velocity and attenuation models. However, multiparameter FWI imposes additional challenges including expanding the null space and facing trade-off issues. Theoretically, an ideal way to mitigate the trade-off issue in multiparameter FWI is to apply the inverse Hessian operator to the parameter gradients. However, it is often not practical to calculate the full Hessian and its matrix inverse, as this will be extremely expensive. To improve the computational efficiency and mitigate the trade-off issue, we use an efficient wavefield inversion (EWI) method to invert for the velocity and the intrinsic attenuation. This approach is implemented in the frequency domain, and the velocity, in this case, is complex valued in the viscoacoustic EWI. We propose a sequential update strategy for the velocity and the intrinsic attenuation, and we repeat the separate optimizations, we refer to as outer iterations, until the convergence is achieved. As viscoacoustic EWI is able to recover a good velocity model, the velocity update leakage to the $Q$ model is largely reduced. We show the effectiveness of this approach using synthetic data generated for the viscoacoustic Marmousi and Overthrust models. To further prove the validity of the proposed approach, we generate data in the time domain using a viscoelastic wave equation solver, and obtain reasonable inversion results in the frequency domain using the viscoacoustic approximation.

show abstract

A Time-Domain Preconditioned Truncated Newton Approach to Visco-acoustic Multiparameter Full Waveform Inversion

Cited by 77 publications

References 69 publications

3-D Time-Domain VTI Viscoacoustic Full Waveform Inversion: Application to Valhall Field Data

3-D Time-Domain VTI Viscoacoustic Full Waveform Inversion: Application to Valhall Field Data

Effect and Reconstruction of Attenuation in Acoustic FWI - Method and Field Data Application

A sequential inversion with outer iterations for the velocity and the intrinsic attenuation using an efficient wavefield inversion

Contact Info

Product

Resources

About