Geosteering inverse problems become challenging since the new generation of extra-deep EM geosteering tools with a much larger depth of detection has emerged on the market. The earth model to be inverted will have more detailed layer structures. Hence, the associated inverse problems become more complicated, and traditional deterministic methods are easily stuck in the local optima. Bayesian inversion arises as an alternative approach. This kind of stochastic optimizations are in general better at searching for global optimal solutions and handling uncertainty quantification. In this abstract, we propose an innovative approach, the Hamiltonian Monte Carlo (HMC) sampling method, to solve the statistical geosteering inverse problems. HMC uses the Metropolis accept/reject rule on the search of update, which introduces the Hamiltonian functions. The examples in the simulations demonstrate the efficacy of the HMC inversion.
Presentation Date: Wednesday, September 27, 2017
Start Time: 3:05 PM
Location: 362D
Presentation Type: ORAL
