## Search

Now showing items 1-10 of 28

JavaScript is disabled for your browser. Some features of this site may not work without it.

Author

Alkhalifah, Tariq Ali (28)

Sun, Bingbing (6)Song, Chao (5)Kazei, Vladimir (4)Guo, Qiang (3)View MoreDepartment
Earth Science and Engineering Program (28)

Physical Sciences and Engineering (PSE) Division (28)King Abdullah University of Science & Technology (3)Extreme Computing Research Center (1)KAUST (1)View MoreJournalSPE Middle East Oil and Gas Show and Conference (1)KAUST Acknowledged Support UnitSupercomputing Laboratory (5)SWAG (4)SWAG group (4)members of SWAG (2)Shaheen (2)View MorePublisherSociety of Exploration Geophysicists (15)EAGE Publications BV (12)Society of Petroleum Engineers (SPE) (1)Type
Conference Paper (28)

Year (Issue Date)
2019 (28)

Item AvailabilityOpen Access (16)Metadata Only (12)

Now showing items 1-10 of 28

- List view
- Grid view
- Sort Options:
- Relevance
- Title Asc
- Title Desc
- Issue Date Asc
- Issue Date Desc
- Submit Date Asc
- Submit Date Desc
- Results Per Page:
- 5
- 10
- 20
- 40
- 60
- 80
- 100

Enhancing the Near-Surface Image Using Duplex-Wave Reverse Time Migration

Sindi, Ghada; Alkhalifah, Tariq Ali; Fei, Tong; Luo, Yi (SPE Middle East Oil and Gas Show and Conference, Society of Petroleum Engineers (SPE), 2019-03-15) [Conference Paper]

Reverse time migration (RTM) involves zero-lag cross-correlation of forward extrapolated source function wavefields and backward extrapolated receiver wavefields. For a near surface with complex structures and velocity anomalies, forward propagating the source wavelet generates wavefields containing reflections, near-surface multiples, and scattered direct arrivals. The wavefields are recorded as upgoing arrivals contaminated by the same reflections, near-surface multiples, and scattered signals, which can be critical for imaging near-surface structures and scatterers.
Here, we develop a new depth migration, duplex reverse time migration (DRTM) technique to improve imaging of complex near-surface structures. DRTM uses the direct arrival as a source to forward propagate and generate source wavefields, and reversely extrapolated recorded data in a zero-lag cross-correlation imaging condition to generate the final section. The interaction between the data components during cross- correlation can use primaries and multiples to image the near-surface structure correctly. Cross-talk artifacts may exist, but they are comparatively weak.
DRTM is demonstrated on both synthetic and field data examples showing an enhanced image in areas with complex near-surface structures compared to conventional RTM imaging methods. The new algorithm can significantly enhance shallow imaging without additional computation costs compared with conventional RTM. It can produce an image with higher resolution and signal-to-noise (S/N) ratio by replacing the source wavelet with the recorded direct arrivals, which include near-surface information necessary to boost the image in areas with near-surface complexity. Since the direct arrivals are one of the most energetic events recorded, the resultant image is typically of high S/N. The wave can also illuminate shallow zones better than primaries in marine environments.

Adaptive traveltime inversion with information entropy regularization

Sun, Bingbing; Alkhalifah, Tariq Ali (Society of Exploration Geophysicists, 2019-08-10) [Conference Paper]

A recently introduced Adaptive Traveltime Inversion (ATI) provided us with a robust misfit function for reducing cycle skipping in Full-Waveform Inversion (FWI). Unlike the conventional L2-norm approach, ATI computes a matching filter first by deconvolution of the predicted data with the measured ones. If the velocity model is relatively accurate, the resulting matching filter is close to a Dirac delta function. Its traveltime shift, which characterizes the defocusing of the matching filter, is computed by minimization of the cross-correlation between a penalty function like t2 and the matching filter. ATI is constructed by minimization of the least square errors of the calculated traveltime shifts. It has been shown that the resulting traveltime shift corresponds to a first-order moment, which corresponds to the mean value of the resulting matching filter distribution. In order to accelerate the convergence and improve the robustness of the ATI approach, in this abstract, we propose to constraint the variance of the resulting matching filter using an information entropy function. We further demonstrate that, in comparison to AWI which tries to minimize the sum of the mean and the vairance of the resulting matching filter, the misfit of ATI with entropy minimizes the sum of the mean and the logarithm of the variance instead. We use the Marmousi example and an offshore field dataset to demonstrate the effectiveness of the proposed method.

Regularized full-waveform inversion for large 3-D salt bodies

Kalita, Mahesh; Ghazali, Ahmad Riza; Xin, Kefeng; Dzulkefli, Farah Syazana; Alkhalifah, Tariq Ali (Society of Exploration Geophysicists, 2019-08-10) [Conference Paper]

Our objective is to invert for large 3-D salt models using full-waveform inversion (FWI), especially in the absence of good starting models and low frequencies in the seismic data. This objective incurs a couple of critical issues. First, the presence of salt geobodies aggravates the non-linearity and ill-posedness of FWI. Second, a 3-D FWI is computationally very expensive, even more so when the initial model is in hindsight very far from the target one. To mitigate the ill-posedness, we propose to utilize model regularization in the FWI framework to promote a limited variation in the inverted model followed by a post-processing step of FWI to penalize sharp velocity drops in the model. Next, to reduce the computational overburden, we propose to utilize a multi-excitation assumption (MExA) of source wavefields in the FWI gradient calculation step. This assumption, due to the simplistic nature of the source wavelet, approximates the source wavefield wiggle at a gridpoint by a series of its energetic arrivals. As a result, the gradient evaluation using MExA requires us neither to store the entire source wavefield nor to include an additional extrapolation step to propagate the source wavefield from its temporary storage at the boundary. The versatility of the proposed method is demonstrated on a synthetic dataset of the modified SEG/EAGE salt model in which the lowest available frequency is 3 Hz.

Near-surface S-wave velocity estimation using ambient noise from fiber-optic acquisition

Zhang, Zhendong; Alajami, Mamdoh; Alkhalifah, Tariq Ali (Society of Exploration Geophysicists, 2019-08-10) [Conference Paper]

Distributed acoustic sensing (DAS) acquisition measures the ground motion using fiber-optic cables. Unlike conventional sensors, DAS can cost effectively provide dense seismic arrays and long-term operations, which is good for monitoring ambient noise. We propose a similarity-weighted stacking of randomly selected short-time duration noise to generate virtual common-shot-gathers (CSG). The similarity-weighted stacking only counts the primary contributions of coherent events, while a short-time correlation can suppress the crosstalk usually presents in late arrivals. Then, we use the wave-equation Rayleigh-wave dispersion-spectrum inversion, which utilizes all the dispersion modes available and avoids picking the dispersion curve, estimating the shallow S-wave velocities. We use a field DAS data set collected in Saudi Arabia to demonstrate the proposed method.

Reflection waveform inversion in acoustic VTI media

Li, Yuanyuan; Alkhalifah, Tariq Ali (Society of Exploration Geophysicists, 2019-08-10) [Conference Paper]

Full waveform inversion (FWI) in transversely isotropic media with vertical symmetry axis (VTI) provides the opportunity to better match the data at the near and far offsets. However, multi-parameter FWI in general suffers from a serious cycle-skipping and trade-off problem. Reflection waveform inversion (RWI) can help us build a background model by minimizing the reflection data residuals. Thus, we apply RWI to acoustic VTI media. According to the radiation patterns analysis, the acoustic VTI media should be described by a combination of the normal-moveout (NMO) velocity vn and the anisotropic parameters ? and d in the RWI applications. To reduce the trade-off, we first invert for the background vn, and then update the background vn and ?, simultaneously to fit the far-offset reflections. We apply Born modeling to produce the reflections for the two stages of the RWI method. For a follow up FWI applications, we use the background vn and ? to calculate the horizontal velocity vh and the parameters ? and e. The acoustic VTI FWI will utilize the diving waves to improve the background, as well as utilize the reflections for high resolution information. We test the inversion algorithm on the modified VTI Sigsbee 2A model (a salt free part). The results show that the approach can converge to a reasonable result starting from an isotropic model with a linearly increasing vn, even in absence of low frequencies.

Direct Greens function retrieval with internal multiples: An alternative to Marchenko focusing

Guo, Qiang; Vasconcelos, Ivan; Alkhalifah, Tariq Ali (Society of Exploration Geophysicists, 2019-08-10) [Conference Paper]

Subsurface Green's functions provide crucial information for the seismic imaging and redatuming. A complete Green's functions containing the primary reflections and all orders of multiples can be utilized to mitigate artifacts and improve the resolution of seismic imaging. To fulfill this goal, some data-driven approaches using one-sided recorded data from the Earth's surface and a smooth migration velocity of the reference medium were developed. Among these approaches an iterative scheme was proposed using the multidimensional Marchenko equation based focusing functions. The iterative Marchenko approach is intrinsically designed to retrieve the coda of the focusing functions, which is supposed to handle all the internal multiples. The estimated focusing functions are then utilized to calculate the Green's functions by a crosscorrelation step. Inspired by the generalized internal multiple imaging (GIMI), we propose an approach that directly retrieves the Green's functions, instead of solving for the focusing functions. In the GIMI process, the reflection data are projected into the subsurface using the transmission information, followed by an interferometric step, which is similar to the multidimensional crosscorrelation of the Marchenko implementation. Thus, we derive a projected Marchenko equation from the relation between the Green's functions and the focusing functions, which reveals a clear connection to the GIMI. The new formulation offers an opportunity to solve for the Green's functions using an iterative scheme or by dealing with different orders of scattering, separately (a hierarchic approach). We introduce these two schemes and the corresponding adjoint operations, which enable us to adopt an optimization for data fitting. The basic performance of the two schemes are demonstrated on synthetic examples for the purpose of redatuming.

Target-oriented inversion with least-squares waveform redatuming

Guo, Qiang; Alkhalifah, Tariq Ali (Society of Exploration Geophysicists, 2019-08-10) [Conference Paper]

Thanks to the rapid growth in high-performance computing technology, full waveform inversion (FWI) has been successfully implemented in many field data applications. Nevertheless, it is still extremely expensive to perform a multi-parameter FWI over the whole subsurface model space that often needs to be discretized consistently using a fine grid, to delineate for example reservoir scale features. Building on the recent development of target-oriented imaging and inversion, we split the subsurface space into the overburden, above a datum level, and the target zone beneath the datum. Our objective is to retrieve the virtual data at a target level and then estimate a high-resolution model of the critical, possibly reservoir, zone. We first build an overburden velocity model using FWI with the data containing frequencies up to 20 Hz and then retrieve a virtual dataset at the datum survey from the data recorded at the Earth's surface. A least-squares optimization of the waveform redatuming is used for the virtual data retrieval. We finally invert for the target zone using the estimated highly reduced in size, but containing high-frequency, dataset. It will lead to an obvious boost in the convergence rate and bring down the memory and computational cost, even though a finer grid is used for the redatuming and the following inversion of the target zone. The Chevron 2014 blind test dataset is used to demonstrate the effectiveness of this strategy.

Near-surface S-wave velocity estimation using ambient noise from fiber-optic acquisition

ML-descent: An optimization algorithm for FWI using machine learning

Sun, Bingbing; Alkhalifah, Tariq Ali (Society of Exploration Geophysicists, 2019-08-10) [Conference Paper]

Full-Waveform Inversion is a nonlinear inversion problem, and a typical optimization algorithm such as nonlinear conjugate-gradient or LBFGS would iteratively update the model along gradient-descent direction of the misfit function or a slight modification of it. Rather than using a hand-designed optimization algorithm, we trained a machine to learn an optimization algorithm which we refer to as”ML-descent” and applied it in FWI. Using recurrent neural network (RNN), we use the gradient of the misfit function as input for training and the hidden states in the RNN uses the history information of the gradient similar to an BFGS algorithm. However, unlike the fixed BFGS algorithm, the ML version evolves as the gradient directs it to evolve.The loss function for training is formulated by summarization of the FWI misfit function by the L2-norm of the data residual. Any well-defined nonlinear inverse problem can be locally approximated by a linear convex problem, and thus, in order to accelerate the training speed, we train the neural network using the solution of randomly generated quadratic functions instead of the time-consuming FWI gradient. We use the Marmousi example to demonstrate that the ML-descent method outperform the steepest descent method, and the energy in the deeper part of the model can be compensable well by the ML-descent when the pseudo-inverse of the Hessian is not incorporated in the gradient of FWI.

Micro-seismic event estimation using an efficient wavefield inversion method

Song, Chao; Alkhalifah, Tariq Ali (Society of Exploration Geophysicists, 2019-08-10) [Conference Paper]

Micro-seismic event estimation results depend highly on the velocity accuracy. Full waveform inversion (FWI) has been employed to invert for the velocity and micro-seismic source image, simultaneously. However, conventional FWI suffers from the infamous cycle-skipping problem, which is even more serious when the source location is unknown. To mitigate this issue, we formulate an optimization problem to linearly reconstruct the wavefield in an efficient matter using the background model information and allow an enhanced source function to absorb the secondary (perturbation) source information. This reconstructed wavefield is then used to update this enhanced source function using the same background wave equation modeling operator without any inversion or update process. We then use the reconstructed wavefield to extract from the enhanced source function the parts corresponding to the micro-seismic source image and those corresponding to secondary sources (velocity perturbations), which can be used to update the model. In the outer loop iterations we repeat the processes of inverting for the source and updating the model until we achieve convergence. This process and its effectiveness is demonstrated on a complicated synthetic model and a field dataset.

The export option will allow you to export the current search results of the entered query to a file. Different formats are available for download. To export the items, click on the button corresponding with the preferred download format.

By default, clicking on the export buttons will result in a download of the allowed maximum amount of items. For anonymous users the allowed maximum amount is 50 search results.

To select a subset of the search results, click "Selective Export" button and make a selection of the items you want to export. The amount of items that can be exported at once is similarly restricted as the full export.

After making a selection, click one of the export format buttons. The amount of items that will be exported is indicated in the bubble next to export format.