- Home
- Conferences
- Conference Proceedings
- Conferences
81st EAGE Conference and Exhibition 2019
- Conference date: June 3-6, 2019
- Location: London, UK
- Published: 03 June 2019
1 - 20 of 1010 results
-
-
Innovative Inversion Schemes for Model Building and Reflectivity Estimation: A Deep Water West African Case Study
Authors T. Martin, M. Barbaray, G. Venfield and V. ChavdaSummaryLate Cretaceous channel systems create structural uncertainty and impact amplitude fidelity of both Late and Early Cretaceous plays in deep water Côte d'Ivoire seismic data. We present a case study using a combination of Full Waveform Inversion (FWI) and image domain Least-squares migration (LSM) to resolve the impact of complex Late Cretaceous channel systems on deeper targets. A full wavefield FWI approach using a velocity kernel that eliminates the reflectivity imprint, created an accurate velocity model. This removed the structural uncertainty when used in the imaging step. A reflectivity estimate was then determined using LSM. The final dataset had improved amplitude fidelity by compensating for the loss of bandwidth and illumination associated with the Late Cretaceous channels.
-
-
-
A Deblending, Demultiple and High-Resolution Velocity Model Building Workflow Across the Senja Ridge
Authors S. Stokes, D. Manns, M. Romanenko, B. Kjølhamar, R. Myklebust and E. HendenSummaryTriple-source, continuous-recording acquisition is used to acquire a modern, high density 3D survey across the Senja Ridge and the surrounding salt provinces of the Tromsø and Sørvenstsnaget basins, in the Western margin of the Norwegian Barents Sea.
A robust imaging workflow is developed and presented, combining deblending technology with extensive demultiple methodology to precondition the data for velocity model building, utilising Full Waveform Inversion and high-resolution image guided tomography.
The final migrated image delivers a significant uplift in imaging across the area, with the advanced velocity model building techniques adding confidence to the positioning of events both within the Senja Ridge and the surrounding salt basins.
-
-
-
Application of Viscoelastic Pre-stack Time Migration in Thin Sand Reservoir Imaging
Authors H. H. Zeng, Q. Su, L. Lv, G. Yuan, H. Liu, H. J. Meng and B. SongSummaryWith the development of oil and gas fields in eastern China towards the stage of fine exploration, higher seismic resolution is required. How to improve the imaging accuracy of thin sand bodies and small faults is an important factor in identifying favorable reservoirs. Because of the narrow frequency band and small azimuth, conventional seismic exploration is difficult to meet the exploration needs of thin reservoirs and small faults of 3–5 meters. On the basis of broadband, wide-azimuth and high-density(BWH) data acquisition and processing, the viscoelastic pre-stack time migration can effectively compensate the attenuation of high frequency seismic wave absorption caused by viscous and thin layer scattering of the earth medium, restore the attenuated high frequency components, and further obtain high resolution migration imaging results. The application of actual data shows that the method can obtain higher resolution migration imaging results than conventional pre-stack time migration, and the result data effectively support the exploration deployment of horizontal wells for tight oil in this area.
-
-
-
Adaptive Traveltime Inversion: Mitigating Cycle-Skipping by Minimization of the Moment of the Matching Filter Distribution
Authors B. Sun and T. AlkhalifahSummaryWe present a method to obtain a misfit function for robust waveform inversion. This method, designated as adaptive traveltime inversion (ATI), computes a matching filter that matches the measured data to the predicted one. If the velocity model is accurate, the resulting matching filter reduces to an (approximate) 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 $t$ and the matching filter. The ATI misfit function is constructed by the minimization of the least square error of the calculated traveltime shifts. Further analysis shows that the resulting traveltime shifts correspond to the first-order moment, the mean value, of the resulting matching filter distribution. We extend ATI to a more general misfit function formula by computing different order moment of the resulting matching filter distribution. Choosing the penalty function in adaptive waveform inversion (AWI) as $t$, the misfit function of AWI can be interpreted as the second-order moment. We use the Marmousi model to verify the effectiveness of the proposed method.
-
-
-
Elastic Full Waveform Inversion with Convolution-Based Crosstalk-Free Source-Encoding Algorithm
More LessSummarySimultaneous-source method can dramatically improve the efficiency of FWI but faces the problems of crosstalk noise and data mismatch. Besides, additional source wavelet estimation, which is critical for successful FWI, will counteract the efficiency of simultaneous-source algorithm. In this paper, different from conventional solutions, we propose the source-independent elastic FWI with the convolution-based crosstalk-free source-encoding algorithm. Time-domain blended simultaneous-source wavefield can be deblended and expressed as a few frequency-domain snapshots so that this algorithm can be naturally applied to unfixed-spread acquisition system, avoid the crosstalk noise and reduce the great memory requirement to a large degree. Synthetic examples verify the efficiency and accuracy of the propose algorithm with a strong robustness to different incorrect wavelets.
-
-
-
Fluid-Solid Coupled Elastic Full Waveform Inversion in the Curvilinear Coordinates for OBC Data
More LessSummaryThe marine seismic exploration with the ocean bottom cable (OBC) technology could record the compressional (P-) wave and shear (S-) wave information simultaneously. Elastic full waveform inversion (EFWI) uses both P- and S-waves of OBC data to invert multi-parameters with adequate amplitude information and complete illumination of the subsurface. We use different formats of wave equation in fluid and solid mediums, and the appropriate boundary condition applied enables the energy exchange on the interface, which is more stable and efficient than the traditional integrated simulation scheme. However, if the fluid-solid coupled medium has an extremely irregular seabed interface, the conventional rectangular-grid-based finite difference (RFD) method cannot obtain accurate source and receiver wavefields. We introduce the curvilinear coordinates to overcome the problem of the RFD method on handling the irregular fluid-solid interface. To reduce the crosstalk of inverted P- (Vp) and S-velocities (Vs), we derive the direction formulas of Vp and Vs based on P- and S-wave modes separation in the curvilinear coordinates, and finally propose a Curvilinear-grid-based Fluid-Solid Separated- wavefield EFWI (CFS-SEFWI) method. In the numerical examples, a modified Marmousi II model with irregular fluid-solid interfaces is used to test the proposed CFS-SEFWI method.
-
-
-
Robust FWI Updates in the Presence of Cycle Skipping
Authors J. Ramos-Martinez, A. Valenciano, X. Jiang and N. CheminguiSummaryFull Waveform Inversion (FWI) success depends on producing seamless short- and long-wavelength model updates while avoiding cycle skipping. In its traditional implementation, FWI risks converging to an inaccurate result if the data lacks sufficient low frequencies or the starting model is far from the true one. Additionally, the model updates may display a reflectivity imprint before the long-wavelength features are fully recovered. A solution to these fundamental challenges combines the quadratic form of the Wasserstein distance (W2-norm) for measuring the data misfit with a robust implementation of a velocity gradient. The W2-norm reduces the risk of cycle skipping whereas the velocity gradient effectively eliminates the reflectivity imprint and emphasizes the long-wavelength model updates. We illustrate the performance of the new solution on a field survey acquired offshore Brazil. There, we demonstrate how FWI successfully updates the earth model and resolves a high-velocity carbonate layer that was missing from the starting model.
-
-
-
Bayesian Uncertainty Estimation in the Presence of Tomographic Model Error
Authors A. Vlassopoulou, R. Felicio, C. Hagen, I. Jones, M. Ackers and S. SchjelderupSummaryWhether or not we build a parameter field model, or deliver a subsurface image, our industry has been sadly lacking in attempting to assign ‘error bars’ to any of the products created. Given that we can never obtain a “correct” model based on measured data, we need to assess how suitable the derived approximate model or resultant image, is. It transpires that this is an extremely difficult task to undertake in a quantitative manner. There are certain minimum acceptance criteria, which tell us that at least the derived model explains the observed data, namely, flat image gathers following migration with the obtained model, which also match all available well data (at least to within some specified acceptance threshold), but these criteria do not tell us how good the model or image is. Here we adopt a Bayesian analysis of tomographic model error so as to quantify image positioning uncertainty, but more specifically, in this work we consider the effect of the quality of the initial model on the final uncertainty estimation, demonstrating quantitatively how prior model uncertainty affects final posterior image positioning uncertainty estimates.
-
-
-
A Comparison Between Two Elitism-Based Global Optimization Methods for Estimating a Starting Velocity Model for FWI
Authors O.F. Mojica, P. Nogueira, R. Santana and D.E. ReveloSummaryThe performance of two global optimization methods (GOMs) with embedded elitism strategies for solving the 2D acoustic seismic inverse problem is compared here using synthetic acoustic data of the Marmousi model. A real-coded elitist genetic algorithm (GA) that had been used successfully in the past to this end is compared with a newly developed elitist-mutated particle swarm optimization (EMPSO) technique to estimate acoustic macro models of the P-wave velocity field (Vp). We find that EMPSO seems to have higher performance than GA with respect to the final attained value of the cost function when the completion of specified number of iterations is chosen as stopping criterion. The results further show that while both EMPSO and the GA obtain high-quality solutions, the computational effort required by PSO to arrive at such high-quality solutions is less than the effort required to arrive at the same high-quality solutions by the GA. Finally, multiple runs of descent-based full waveform inversion (FWI) started from either final GA or EMPSO models produce high-resolution final models.
-
-
-
Robust Wave Equation Migration Velocity Analysis Using a Normalized Differential Semblance Misfit Function
Authors A. Alali, B. Sun and T. AlkhalifahSummaryThe wave equation migration velocity analysis (WEMVA) techniques try to estimate an accurate subsurface velocity model for migration purposes by relying on the full-wave equation. A popular method in WEMVA is the differential semblance optimization (DSO). The objective function of DSO applies a penalty operator to extended subsurface images to minimize the energy residing in the non-physical extension. The penalty operator used in conventional DSO actually tries to reduce the energy instead of focussing it toward the subsurface zero offset. We introduce a normalization term to the DSO method in which we use a pseudo inverse Born operator to calculate the extended image and show that the new objective is more efficient and robust in focusing the energy and eliminating artifacts in the process of estimating the macro-velocity model.
-
-
-
Wavefront Tomography with Enforced Diffraction Focusing
Authors A. Bauer, B. Schwarz, L. Diekmann and D. GajewskiSummaryWavefront tomography is an efficient and stable tool for the generation of smooth velocity models based on first and second-order attributes, which describe slope and curvature of the measured wavefronts. While slopes are relatively stable to determine, curvatures can become unreliable in the case of strong lateral heterogeneity. Since wavefront tomography is mainly driven by the misfit of modeled and measured wavefront curvatures, its convergence may be compromised by curvatures of bad quality. A possible solution to overcome this problem are diffractions that have a unique property: all measurements belonging to the same diffraction are connected to the same subsurface region. In recent work, we introduced an event-tagging scheme that automatically assigns a unique tag to each diffraction in the data. We propose to use this information to constrain the inversion by enforcing all diffracted measurements with the same tag to focus in depth, thus overcoming the sole dependency of wavefront tomography on second-order attributes. Results for diffraction-only data with vertical and lateral heterogeneity confirm that it is possible to obtain depth velocity models for zero-offset data without using curvature information and that the suggested approach may help to increase the stability of wavefront tomography in complex settings.
-
-
-
First-Arrival Traveltime Tomography Based on the Adjoint State Method with Independence of Surface Normal Vectors
More LessSummaryThe first-arrival traveltime tomography based on adjoint state method has gained widespread attention in recent years. Compared with the traditional method, this method does not need to perform any ray tracing. Only the eikonal equation and the adjoint equation need to be solved in order to obtain the inversion gradient. However, the result of commonly-used adjoint state traveltime tomography is related to the boundary normal vector of the computational domain. In the cases of complex surface, the calculation of adjoint field via the boundary normal vector makes the traveltime residual distributed according to structure-irrelevant parameters. As a result, the tomography gradient deviates from the true gradient and consequently contributes to a untrue tomographic model. In need of solving this issue, this paper derives a new form of adjoint equation that is independent from the boundary normal vector. This tomography method adapts to any kind of observation system. In situations of tough topography, it produces better results compared to the commonly-used adjoint state tomography method. The gradient calculated by this method is similar to that of the traditional ray-based method, but it offers better continuity and smoothness with significantly improved inversion efficiency proved by the synthetic experiments.
-
-
-
Inversion Velocity Analysis by One-Way Wave Equation
More LessSummaryMigration velocity analysis (MVA) is an image domain technique to determine large-scale structure of the subsurface velocity model. However, due to limited surface acquisition geometry and uneven illumination, migration smiles and spurious oscillations around the reflector positions may lead to local minimal in the inversion process. For more stable results, in this paper, we have developed new expressions based on true-amplitude, one-way wave-equation operators and used those expressions to direct velocity analysis which is called inversion velocity analysis (IVA). The differential semblance optimization (DSO) objective function is adopted to evaluate the inverted images quality. Compared to traditional MVA method, the modified inversion method attenuates migration smiles by compensating for geometric spreading and uneven illuminations. Also, the direct inversion has influence on the gradient, with slightly modified objective function, the spurious oscillations around reflectors in the gradient can be attenuated. Numerical examples illustrate the effectiveness of the IVA which offers new perspective to large-scale model construction.
-
-
-
A Robust Differential Semblance Optimization by Quantitative Migration
More LessSummaryWave equation migration velocity analysis is an image domain method which can be implemented to estimate the large-scale model. For classic migration and velocity analysis, the inversion results or stacked image sections often suffer from kinematic artefacts due to the limit of the acquisition geometry and uneven illumination or mathematically due to non-unitary of the wavefield extrapolation operator. In this paper, the modified differential semblance optimization (DSO) objective function with root mean squared image amplitudes, which can reduce the amplitudes sensitivity of DSO, is adopted to evaluate the inverted images quality. To obtain more stable results, we introduce a weight to the traditional migration and velocity inversion without ray tracing calculation. To some extent, the weighting function can automatically compensate for uneven illumination and remove migration artefacts. Thus the gradient of MVA has a smoother behavior and gives an optimal convergence. The results of numerical examples illustrate the robustness of the weighting function.
-
-
-
Optimal Surface and Subsurface Parameters for a Successful Migration Velocity Analysis
Authors T. ZHOU, H. Chauris and F. AudebertSummaryMigration velocity analysis is a method to estimate both reflectivity model and macromodel. We study here the sensitivity of the final result with respect to the selection of input surface offsets and subsurface parameters. We show that the velocity gradient used to update the macromodel highly depends on these parameters. We provide rules to ensure proper gradient by avoiding the dominance of the migration smiles. These rules are developed in the case of simple configurations. We apply them to the Marmousi model and show that final velocity model contains artefacts when the optimal parameters are not selected.
-
-
-
A High-Order Temporal and Implicit Spatial Staggered-Grid Finite-Difference Scheme for Modelling Scalar Wave Propagation
More LessSummaryFinite difference (FD) is a commonly used tool for modelling seismic wave propagation. Temporal high-order FD methods can enhance the accuracy and stability compared with the methods with second-order accuracy in time. The implicit calculation of spatial derivatives also contributes to the improvement in accuracy. We propose a new implicit staggered-grid FD (SFD) scheme based on a combined stencil, which is the combination of pyramid and cross stencils in 3D case. Our scheme calculates the temporal and spatial derivatives by using high-order temporal and implicit spatial FD operators, respectively. Based on the dispersion relations, we optimize the temporal and implicit spatial FD coefficients by using Taylor series expansion (TE) and least squares (LS). We formulate four implicit SFD operators: TE-TE, TE-LS, LS-TE and LS-LS operators. 2D and 3D modelling examples determine that our implicit scheme has greater accuracy than other schemes. Additionally, our implicit SFD scheme allows for a larger grid spacing, which can increase the computational efficiency.
-
-
-
3D Frequency-Domain Finite-Difference Acoustic Wave Modeling on Discontinuous Grids
More LessSummaryThe frequency-domain finite-difference (FDFD) method is an effective tool for implementing frequency-domain seismic modeling and full waveform inversion. However, the computational cost for the FDFD method dealing with 3D large models is prohibitive, limiting its application. As a common strategy to improve computational efficiency, a nonuniform grid is usually adopted in the time-domain finite-difference (TDFD) method instead of the FDFD method. We propose a strategy to implement 3D frequency-domain acoustic wave modeling on discontinuous grids efficiently. In the whole model area discreted by discontinuous grids, we apply the 3D average-derivative optimal scheme (3D-ADOS) to simulate wave propagation in each subregion and use the conventional second-order or rotated second-order finite-difference scheme to calculate the wavefield in the transition area. In this way, the accuracy of the wavefield in the global area remains the same as that obtained by the conventional 3D-ADOS while the computational cost is reduced significantly. The numerical example is shown to verify the feasibility and efficiency of our strategy.
-
-
-
Acoustic Wave Simulation by Lattice Boltzmann Method with D2Q5 and D2Q9 of Different Relaxation Times
More LessSummaryIn this abstract, the lattice Boltzmann method (LBM), a kind of mesoscopic method for modeling waves, is introduced to simulate acoustic wave propagation, which is usually simulated by finite difference method (FDM) in seismology. By choosing different discrete velocity models of D2Q5, D2Q9 and different relaxation times of LBM, with the help of the variable-controlling approach, various seismic wavefields can be obtained. Numerical experiments demonstrate that the wavefields of acoustic wave propagation simulated by LBM with D2Q5 of relaxation time as 0.51 can better coincide with those by acoustic FDM.
-
-
-
Seismic Traveltime Calculation Using Fast Marching Methods with a Rotated Staggered Grid Finite-Difference Scheme
More LessSummaryFast marching methods (FMM), with unconditonal stability and high efficiency, has become one of the most prevalent algorithms for traveltime computation. However, finite-difference (FD) schemes adopted by the existing FMMs define the velocity and traveltime on the same grid system, which doesn’t conform to the physical process of ray propagation.
In this article, by defining the traveltime and velocity on two sets of staggered grid system, we construct a new rotated staggered-grid FD schemes to approximate the eikonal equation. FMM with this staggered-grid FD schemes (RSFMM) can archive higher accuracy than classical FMM with 1st-order and 2nd-order FD schemes. Furthermore, RSFMM calculates the traveltime with the velocity at the centre of grid points involved in the FD operator, which perfectly conforms to the physical process of ray propagation. Therefore, RSFMM has better applicability in complex medium. Considering the Spherical wave characteristics in the vicinity of the source point, we apply a combination scheme of spherical approximation method (SAM) and RSFMM (SAM+RSFMM) to further improve the accuracy. Numerical experiments on homogeneous, layered and Marmousi model approve the high accuracy and applicability in complex media of RSFMM and SAM+RSFMM.
-
-
-
An Efficient Wave-Equation Based AVP Gather Simulation Using the Precondition Least Square Method
More LessSummaryA stable and fast numerical approach to synthesize the seismic prestack AVP gather based on the wave-equation in the frequency domain is presented in this abstract. As a powerful method to extract rock property information, wave-equation based seismic amplitude-variation-with-offset (WEB-AVO) inversion relies much on the corresponding forward modelling engine. To solve the state equations of Lippman-Schwinger type numerically, in the new method, we firstly convert the two integral formulas into a large linear system of equations, in which the total wavefields (pressure and particle velocity) are treated as the unknown. Then, the least-square method is used to solve the total wavefields. Further, introducing the pre-conditioner is essential, and the robustness of this scheme is great improved. For efficiency, parallel computing method is also used to solve the wavefield at different frequency points simultaneously. Two numerical examples demonstrate the effectiveness and superiority of the method.
-