Browsing by Author "Shen, Peng"

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    Horizontal contraction in image domain for velocity inversion
    (Society of Exploration Geophysicists, 2015) Shen, Peng; Symes, William W.
    A kinematically correct choice of velocity focuses subsurface offset image gathers at a zero offset. Infinitesimal warping from the current image toward its focus can be approximated by a horizontal contraction. The image residual can then be taken as the difference between the warped and the original image to account for the velocity error. Least-squares fitting of the effect of a velocity perturbation to this image-warping perturbation produces a tomographic velocity update. We evaluated a warping scheme based on the radial image derivative in the subsurface offset. We found that the corresponding gradient was free of the diffraction edge effect. We further enhanced the efficiency of the velocity update procedure via the use of a diagonal Hessian approximation. We developed a sequence of synthetic examples leading to a salt body inversion. The limitation of the method was argued and analyzed in the presence of refracting structures.
  • Thumbnail Image
    Item
    Wave equation migration velocity analysis by differential semblance optimization
    (2005) Shen, Peng; Symes, William W.
    Differential semblance measures the deviation from flatness or focus of image gathers. The differential semblance objective function posed on the sub-surface offset domain responds smoothly to velocity changes. Therefore gradient descent methods are uniquely attractive for velocity updating by differential semblance optimization. Because of their kinematic fidelity, wave equation (depth extrapolation) migration methods are natural platforms for velocity analysis in complex structures. The gradient of the objective function with respect to velocity is fomulated through the adjoint of differential migration. Limited memory BFGS algorithm is used for the velocity optimization. The method for wave equation velocity analysis developed in this thesis study is applied to both synthetic and real data examples.
  • Thumbnail Image
    Item
    Wave-equation based Imaging and Velocity Analysis in the Common-Offset domain
    (2018-04-20) Nandi, Papia; Zelt, Colin A.; Shen, Peng
    Offset information is commonly available after using ray-based imaging, however the ray approximations can fail to produce an accurate image or representation of subsurface acoustic velocities in complex geology. Wave-equation imaging and velocity analysis provides a more accurate representation of the physical propagation of wavefronts, and extending the computation to the common-offset domain adds a new dimension of freedom for inversion, provides a wide aperture of illumination, and can be parameterized to avoid cycle skipping. Here, I develop common-offset reverse time migration (CORTM), a wave-equation based imaging method which splits the input data into common-offset bins prior to migration, and then extend it to least-squares migration in the common-offset domain (COLSM), an iterative inversion which produces image gathers that are closer to predicting recorded data than CORTM. With an appropriate number of offset bins, COLSM can be parameterized to avoid cycle skipping within an offset gather, increase convergence rates, amplitude fidelity and resolution. Maintaining execution in the common-offset domain additionally opens up avenues for commonly utilized offset-based processing post-migraiton. I also extend wave-equation based migration velocity analysis (WEMVA), an iterative method has been used succesfully to provide low-frequency starting models for full waveform inversion (FWI), to the common-offset domain by minimizing an image residual from either CORTM or COLSM offset images. COLSM's increased amplitude fidelity over CORTM helps to ensure that the image residual is largely a function of moveout instead of errors in amplitude, which increases the likelihood of converging to a global minimum in a long-wavelength update. Division into sufficiently small offset bins can circumvent the requirement to start from an accurate starting model and also avoid cycle-skipping and associated local minima. Lastly, I show how it is possible to approximate the tomographic operator used to create a long-wavelength update in COWEMVA by using CORTM. The following chapters detail the results of and the theory and methodology used to apply CORTM, COLSM and COWEMVA with and without the tomographic approximation to several synthetic models and one field data set.