Research On Parallel Forward Modeling Of Seismic Wavefield Modeling In 3-D Viscoelastic Random Medium

Traditional seismic prospecting is based on theory of homogeneous medium or layered homogeneous medium, but in accurate seismic exploration, especially in reservoir research, medium which related usually is local heterogeneous. This local heterogeneous medium contains some distributed irregularly small-scaled anomaly geologic bodies, which will effect seismic propagation obviously, so this effect must be considered in oil gas pre-estimate and reservoir research which need study the exact structure of reservoir. In addition, seismic waves will be scattered by way of small-scaled anomaly geologic bodies, and scattering will attenuate direct wave, which caused seismograms be influenced, so conventional seismic prospecting theory and method will not get a good result. In order to solve this problem, we need to do some fundamental research such as description of underground medium and propagation of seismic wave.Because there are more scattered elements in 3-D random medium, seismic wave scatting and attenuation in 3-D and 2-D random medium make a great difference, so it is necessary to construct and forward model 3-D random medium. 3-D random medium forward modeling needs more memory and higher computing speed, single PC or workstation can't meet the requirement, but this problem can be solved by connecting PCs into a cluster and dividing memory requirement into every PC, and let them parallel compute. Using this method, we can forward modeling larger medium model and save computing time.By constructing and forward modeling 3-D viscoelastic random medium model, seismic response in inhomogeneous complex medium will be better modeled, this will be useful in understanding the seismic wave field of karst cave reservoir and helpful in guiding data processing and interpretation.Main research contents and conclusions in the article are following.First, complex heterogeneous medium was described using statistical method. Many kinds of two-dimensional (2-D) random medium model and 3-D random-homogeneous included solution cavity medium model were constructed. There independent autocorrelation lengths (a, b and c) in random medium described the mean scale of heterogeneous anomaly in x, y and z directions respectively. The corresponding relationship between autocorrelation lengths and the scale of geologic bodies was studied by velocity variation and variance.Second, influence of viscoelasticity in forward modeling was reflected by using GSLS (generalized standard linear solid), GSLS shows no dispersivity in both high frequency and low frequency, so a GSLS can be used to model Q of every frequency. .Based on this,3-D viscoelastic wave equation and its discretization by finite-difference method using staggered grid were introduced. The explicit finite difference equations can be derive by using a second-order centred difference scheme to approximate the time derivatives, and a fourth-order staggered scheme with centred differences to approximate the spatial derivatives.Third, a PC 1inux cluster based on MPI was constructed. Based this platform, 3-D random medium model can be divvied into sub volumes ,the number of subvolumes equals to pc numbers, each PC just compute wave field of subvolumes owned to itself. For the communication at the internal edges a two grid thick padding layer has to be introduced, data exchange at the internal edges has to be performed after each time step. After wave field updating, each nodes submitted results to master, master received the data and terminated the program. Using parallel algorithm forward modeling larger scale 3-D random medium model, the problem of high running speed and large memory, which were required in forward modeling 3-D medium, can be easily solved, meanwhile dispersed computing resource will be well used and wealth can be saved.Fourth, 3-D single layered random medium model and random- homogeneous multi-layered medium model were forward modeled. For3-D single layered random medium model, the results showed that scattering wave produced by random perturbations was obviously displayed in seismograms, with autocorrelation lengths becoming larger, scattering event in section became longer, and with dominant frequency of source becoming higher events in section became fine and short. For 3-D random-homogeneous multi-layered medium model, scattering wave produced by random perturbations was obviously displayed in manner of wide and short events in seismograms in middle random medium layer as in single layered random medium, scattering wave appeared under the second reflecting event which was caused by upper layer random perturbations. With dominant frequency of source becoming higher, backward scatting becomes weak and forward scatting becomes strong, forward scatting event also becomes coherent. With autocorrelation lengths becoming larger, backward scatting becomes weak but forward scatting also strong, which influences reflecting event of boundary greatly.Fifthly, 3-D random-homogeneous included solution cavity medium mode was forward modeled. It is obviously to see many scatting wave around cavities in wave field snaps, the wavelength of scatting wave and source wavelet were different. Scatting waves stacked and interfered each other, and formed elementary wave front, perturbations observed from new wave front was the results of each point in elementary wave front stacked and interfered in observing point. Seismic wave field of karst cave reservoir was composed of differences scatting waves, which produced by variety inhomogeneous distributed low speed bodies. So, processing method researcher based on scatting theory should be intensified in order to solve the problem of karst cave reservoir.