Underground formation characterization and uncertainty quantification require the inversion of the subsurface sensing data using probabilistic methods. Markov chain Monte Carlo (MCMC) sampling methods have been extensively used to solve nonlinear inversion problems. However, the computational demands can be prohibitively large for high dimensional problems. Transitional Markov chain Monte Carlo (TMCMC), as an efficient sampling algorithm based on MCMC, has better performance than MCMC when the target distribution is high dimensional and multimodal. Given that TMCMC is naturally parallelizable, we parallelize TMCMC using the Message Passing Interface (MPI) to infer the underground formation with noisy electromagnetic well logging data. The experiment results show that the parallel TMCMC algorithm has reliable performance on the inversion of 1D underground formations with up to 7 layers while exhibiting good parallel scalability for such problems.
Presentation Date: Monday, October 12, 2020
Session Start Time: 1:50 PM
Presentation Time: 3:05 PM
Location: Poster Station 2
Presentation Type: Poster
