Deep neural network surrogates for inverse problems
Abstract/Contents
- Abstract
- Inverse problems in subsurface flow are generally challenging due to the need for a large number of expensive numerical solutions to partial differential equations (PDEs). Inverse modeling typically consists of generating realizations of the unknown model parameters and matching the corresponding model's prediction to the measurements. Model errors, measurement errors, and data scarcity require one to quantify predictive uncertainty in model predictions, which further exacerbates the computational cost of inverse modeling. The latter can be ameliorated either by devising inversion frameworks that require fewer forward model runs to converge, or by constructing a much more efficient forward surrogate model that replaces the PDE solver. In this dissertation, we pursue these two strategies and apply them to three inverse problems of practical importance in subsurface applications. The common thread in this investigation is the use of deep neural network (DNN) surrogates that accelerate forward modeling by several orders of magnitude. In the three applications, we first present a study on identification of the statistical parameters of a discrete fracture network (DFN); we further focus on a higher-dimensional inverse problem on contaminant source identification; finally, we expand the unknown parameter dimension and tackle a realistic three-dimensional inverse problem, in which both a heterogeneous conductivity field and the contaminant release history are identified from sparse observations.
Description
Type of resource | text |
---|---|
Form | electronic resource; remote; computer; online resource |
Extent | 1 online resource. |
Place | California |
Place | [Stanford, California] |
Publisher | [Stanford University] |
Copyright date | 2021; ©2021 |
Publication date | 2021; 2021 |
Issuance | monographic |
Language | English |
Creators/Contributors
Author | Zhou, Zitong |
---|---|
Degree supervisor | Tartakovsky, Daniel |
Thesis advisor | Tartakovsky, Daniel |
Thesis advisor | Horne, Roland N |
Thesis advisor | Kitanidis, P. K. (Peter K.) |
Degree committee member | Horne, Roland N |
Degree committee member | Kitanidis, P. K. (Peter K.) |
Associated with | Stanford University, Department of Energy Resources Engineering |
Subjects
Genre | Theses |
---|---|
Genre | Text |
Bibliographic information
Statement of responsibility | Zitong Zhou. |
---|---|
Note | Submitted to the Department of Energy Resources Engineering. |
Thesis | Thesis Ph.D. Stanford University 2021. |
Location | https://purl.stanford.edu/bb876gs9072 |
Access conditions
- Copyright
- © 2021 by Zitong Zhou
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...