eduzhai > Applied Sciences > Engineering >

Deep-Learning based Inverse Modeling Approaches A Subsurface Flow Example

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 53 pages

Abstract: Deep-learning has achieved good performance and shown great potential forsolving forward and inverse problems. In this work, two categories ofinnovative deep-learning based inverse modeling methods are proposed andcompared. The first category is deep-learning surrogate-based inversionmethods, in which the Theory-guided Neural Network (TgNN) is constructed as adeep-learning surrogate for problems with uncertain model parameters. Byincorporating physical laws and other constraints, the TgNN surrogate can beconstructed with limited simulation runs and accelerate the inversion processsignificantly. Three TgNN surrogate-based inversion methods are proposed,including the gradient method, the iterative ensemble smoother (IES), and thetraining method. The second category is direct-deep-learning-inversion methods,in which TgNN constrained with geostatistical information, named TgNN-geo, isproposed for direct inverse modeling. In TgNN-geo, two neural networks areintroduced to approximate the respective random model parameters and thesolution. Since the prior geostatistical information can be incorporated, thedirect-inversion method based on TgNN-geo works well, even in cases with sparsespatial measurements or imprecise prior statistics. Although the proposeddeep-learning based inverse modeling methods are general in nature, and thusapplicable to a wide variety of problems, they are tested with severalsubsurface flow problems. It is found that satisfactory results are obtainedwith a high efficiency. Moreover, both the advantages and disadvantages arefurther analyzed for the proposed two categories of deep-learning basedinversion methods.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×