Wenfeng Du

Improvement on AVO Equations in VTI Media

In weakly anisotropic VTI media, it is assumed that the ray directions of phase and group velocities are consistent, and P wave excites the horizontal interface, we take into account the equations between the anisotropic angle and the isotropic angle with the same horizontal slowness, and re-deduce AVO equations for VTI media again. The new equation shows that it has no effect on AVO equation for P wave, but improves AVO equation for SV wave.

Seismic small-scale discontinuity sparsity-constraint inversion method using a penalty decomposition algorithm

We consider sparsity-constraint inversion method for detecting seismic small-scale discontinuities, such as edges, faults and cavities, which provide rich information about petroleum reservoirs. However, where there is karstification and interference caused by macro-scale fault systems, these seismic small-scale discontinuities are hard to identify when using currently available discontinuity-detection methods. In the subsurface, these small-scale discontinuities are separately and sparsely distributed and their seismic responses occupy a very small part of seismic image. Considering these sparsity and non-smooth features, we propose an effective L 2-L 0 norm model for improvement of their resolution. First, we apply a low-order plane-wave destruction method to eliminate macro-scale smooth events. Then, based the residual data, we use a nonlinear structure-enhancing filter to build a L 2-L 0 norm model. In searching for its solution, an efficient and fast convergent penalty decomposition method is employed. The proposed method can achieve a significant improvement in enhancing seismic small-scale discontinuities. Numerical experiment and field data application demonstrate the effectiveness and feasibility of the proposed method in studying the relevant geology of these reservoirs.

The Trust Region Method for Time-Domain Full Waveform Inversion

Full waveform inversion is an effective inversion method to obtain high resolution and precision seismic imaging. The problem of full waveform inversion aims to solve the optimization problem of the misfit between the synthetic and recorded data. There are a lot of conventional local optimization methods, such as the methods depending on gradient and Hessian matrix, to solve the ill-posed nonlinear problem. Gradient and Hessian matrix information plays an important role in inversion algorithms. However, these methods always lead to local minimum so that the result of inversion could not be accepted. To confront the problem, we propose a new approach called the trust region method, to solve the unconstrained optimization problem. Strong convergence properties are the key for trust region method, and the new step is limited to the neighbourhood of the current step. The gradient and Hessian matrix information has an effect on inversion algorithms. To test the capability of this method, we apply it to the model of salt domes and synthetic Marmousi from the time-domain acoustic full waveform inversion point of view for reconstructing velocity model. Inversion results show that the model inverted by our method fits the true model well and the precision is high. Hence, the trust region method is a promising method for the problem of full waveform inversion.