Bertrand Denel

Implementation of the Full Waveform Inversion on GPU Cluster

For more than 30 years, seismic reflection has been the main technology used in the Oil & Gas industry. Physics is well known and is based on solving approximations of the wave equation. In addition, geophysicists are now able to find numerical implementations which are the best suited to the computer hardware. Improving velocity models is still very challenging. It is the key to get an accurate seismic depth image. Mainly based on asymptotic approximations, seismic velocity analysis tools have some limitations when dealing with complex velocity models and complex wave propagation paths. Recent progress in high performance computing gives the ability to revive the Full Waveform Inversion formulation for refining velocity models. This technology, known since the mid 80s, is highly challenging due to the non linearity and the non uniqueness of solution. It is based on the minimization of an objective function measuring the difference between computed and observed data. The minimization process is very CPU intensive. Indeed, this is an iterative method which requires the computation of the derivative of the objective function with respect to the model parameters at each iteration by cross correlating the back propagated residual wavefield with the corresponding forward source propagated wavefield.

The reverse time migration technique coupled with finite element methods

The Reverse Time Migration (RTM) is a technique for imaging the subsurface based on successive solutions of the full wave equation which must be computed by fast numerical methods to perform well in three-dimensional domains. In this framework, we propose to solve the full wave equation by the Interior Penalty Discontinuous Galerkin method (IPDG). We analyse the performances of IPDG by comparing it to classical spectral element method and we show that IPDG is more adapted to the RTM. We present 2D migration results to illustrate the ability of the new method to recover the location of the interfaces of stratified media.