國立虎尾科技大學 |

Gradient-Free Active Subspace Construction and Model Calibration Techniques for Complex Models.

Record Type:	Language materials, manuscript : Monograph/item
Title/Author:	Gradient-Free Active Subspace Construction and Model Calibration Techniques for Complex Models./
Author:	Lewis, Allison Leigh.
Description:	1 online resource (130 pages)
Notes:	Source: Dissertation Abstracts International, Volume: 78-08(E), Section: B.
Contained By:	Dissertation Abstracts International78-08B(E).
Subject:	Applied mathematics. -
Online resource:	click for full text (PQDT)
ISBN:	9781369621358

Gradient-Free Active Subspace Construction and Model Calibration Techniques for Complex Models.
Lewis, Allison Leigh.

Gradient-Free Active Subspace Construction and Model Calibration Techniques for Complex Models. - 1 online resource (130 pages)

Source: Dissertation Abstracts International, Volume: 78-08(E), Section: B.

Thesis (Ph.D.)

Includes bibliographical references

As physical models used in large-scale applications become increasingly complex and computationally demanding, numerous issues related to the field of uncertainty quantification (UQ) must be addressed. Knowledge of topics such as model calibration, parameter selection, uncertainty propagation, and surrogate modeling have become essential to researchers in a wide array of fields. In this dissertation, we focus on two aspects of UQ: (i) model calibration in a high-to-low-fidelity framework, and (ii) gradient-free construction of active subspaces for emulation via response surfaces.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2018

Mode of access: World Wide Web

ISBN: 9781369621358Subjects--Topical Terms:

1069907
Applied mathematics.
Index Terms--Genre/Form:

554714
Electronic books.

Gradient-Free Active Subspace Construction and Model Calibration Techniques for Complex Models.
LDR:03673ntm a2200361Ki 4500 001 909842
005 20180426091048.5
006 m o u
007 cr mn||||a|a||
008 190606s2016 xx obm 000 0 eng d
020 $a 9781369621358
035 $a (MiAaPQ)AAI10583456
035 $a AAI10583456
040 $a MiAaPQ $b eng $c MiAaPQ
099 $a TUL $f hyy $c available through World Wide Web
100 1 $a Lewis, Allison Leigh. $3 1180813
245 1 0 $a Gradient-Free Active Subspace Construction and Model Calibration Techniques for Complex Models.
264 0 $c 2016
300 $a 1 online resource (130 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertation Abstracts International, Volume: 78-08(E), Section: B.
500 $a Adviser: Ralph C. Smith.
502 $a Thesis (Ph.D.) $c North Carolina State University $d 2016.
504 $a Includes bibliographical references
520 $a As physical models used in large-scale applications become increasingly complex and computationally demanding, numerous issues related to the field of uncertainty quantification (UQ) must be addressed. Knowledge of topics such as model calibration, parameter selection, uncertainty propagation, and surrogate modeling have become essential to researchers in a wide array of fields. In this dissertation, we focus on two aspects of UQ: (i) model calibration in a high-to-low-fidelity framework, and (ii) gradient-free construction of active subspaces for emulation via response surfaces.
520 $a We first develop an information-theoretic approach to calibrating low-fidelity codes using simulated data from validated high-fidelity models, which are prohibitively expensive to evaluate repeatedly. Our objective is to employ a minimal number of high-fidelity code evaluations as synthetic data for Bayesian calibration of the low-fidelity code under consideration. We employ the mutual information between low-fidelity model parameters and experimental designs to determine input values to the high-fidelity code, which maximize the available information. For computationally expensive codes, surrogate models may be used to approximate the mutual information. We illustrate this framework using a comprehensive set of numerical examples, including several relevant to nuclear power plant design.
520 $a Recent developments in the field of reduced-order modeling---and in particular, active subspace construction---have made it possible to efficiently approximate complex models by constructing loworder response surfaces based upon a small subspace of the original high-dimensional parameter space. These methods rely upon the fact that the response tends to vary more prominently in a few dominant directions defined by linear combinations of the original inputs, allowing for a rotation of the coordinate axis and a consequent transformation of the parameters. In the second portion of this dissertation, we analyze existing gradient-based methods for active subspace construction, and propose a gradient-free active subspace algorithm that is feasible for high-dimensional parameter spaces where finite-difference techniques are impractical. These algorithms are illustrated with several examples from aerospace and nuclear engineering, with input spaces of up to 7700 dimensions.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2018
538 $a Mode of access: World Wide Web
650 4 $a Applied mathematics. $3 1069907
650 4 $a Nuclear engineering. $3 655622
655 7 $a Electronic books. $2 local $3 554714
690 $a 0364
690 $a 0552
710 2 $a ProQuest Information and Learning Co. $3 1178819
710 2 $a North Carolina State University. $3 845424
773 0 $t Dissertation Abstracts International $g 78-08B(E).
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10583456 $z click for full text (PQDT)