國立虎尾科技大學 |

Data Driven Surrogate Modeling of Two-Phase Flows.

紀錄類型:	書目-語言資料,手稿 : Monograph/item
正題名/作者:	Data Driven Surrogate Modeling of Two-Phase Flows./
作者:	Ganti, Himakar.
面頁冊數:	1 online resource (130 pages)
附註:	Source: Dissertations Abstracts International, Volume: 85-05, Section: B.
Contained By:	Dissertations Abstracts International85-05B.
標題:	Computational physics. -
電子資源:	click for full text (PQDT)
ISBN:	9798380824194

Data Driven Surrogate Modeling of Two-Phase Flows.
Ganti, Himakar.

Data Driven Surrogate Modeling of Two-Phase Flows. - 1 online resource (130 pages)

Source: Dissertations Abstracts International, Volume: 85-05, Section: B.

Thesis (Ph.D.)--University of Cincinnati, 2023.

Includes bibliographical references

The ready availability of computational resources has enabled scientists and researchers to generate huge volumes of data for multiphase flows. For routine design calculations, detailed numerical simulations of multiphase flows with disparate time and length scales and turbulence resolution (DNS/ LES) are resource intensive and computationally expensive. A data-driven surrogate model built with available simulation and experimental data can be used instead of intensive numerical calculations and in the absence of an available mathematical models. Such an attempt is made in this thesis, where data-driven surrogate models are built and developed with applications to two-phase flows.This study is the first of its kind to have developed data-driven surrogate modelling frameworks and algorithms with applications to spatiotemporally varying, statistically stationary and steady state multiphase flows. This work discusses the frameworks necessary for identifying and conducting the surrogate modeling process and with quantified errors. GPs were selected as the algorithm of choice as they can be used with the relatively fewer number of numerical simulations data for training purposes. The GP algorithm was modified to work with multiple independent variables of a multiphase flow configuration. When multiple ML algorithms are available and applicable for the same multiphase flow configuration, it becomes necessary to know how each of the algorithm will perform in terms of efficiency, accuracy and speedup. In this thesis, performance comparison of GP and NN machine algorithms was conducted for accuracy and speedup for a 2D Rayleigh-Taylor instability configuration. Prediction results were reported for accuracy, speedup and efficiency. This thesis addresses three major issues for building, developing and comparing various learning algorithms for application to multiphase flows, by identifying a set of guidelines to -1. build and develop data-driven surrogate models based on Gaussian processes for spatiotemporally varying and statistically stationary multiphase flows.2. modify Gaussian processes to enhance prediction capability of multiphase flows with more than one independent variable.3. performance comparison of Gaussian Process and Neural Network based Machine Learning frameworks for efficiency, accuracy and speedup with extensions to other possible machine learning algorithms.It is hoped that this research contributed in establishing guidelines and approaches for building computationally inexpensive data-driven surrogate models for multiphase flows that can replace numerical simulations for initial and routine design for power generation devices.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2024

Mode of access: World Wide Web

ISBN: 9798380824194Subjects--Topical Terms:

1181955
Computational physics.
Subjects--Index Terms:

Multiphase flowsIndex Terms--Genre/Form:

554714
Electronic books.

Data Driven Surrogate Modeling of Two-Phase Flows.
LDR:04142ntm a22004217 4500 001 1145204
005 20240617111409.5
006 m o d
007 cr mn ---uuuuu
008 250605s2023 xx obm 000 0 eng d
020 $a 9798380824194
035 $a (MiAaPQ)AAI30866464
035 $a (MiAaPQ)OhioLINKucin1684772398259224
035 $a AAI30866464
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Ganti, Himakar. $3 1470445
245 1 0 $a Data Driven Surrogate Modeling of Two-Phase Flows.
264 0 $c 2023
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: Dissertations Abstracts International, Volume: 85-05, Section: B.
500 $a Advisor: Khare, Prashant.
502 $a Thesis (Ph.D.)--University of Cincinnati, 2023.
504 $a Includes bibliographical references
520 $a The ready availability of computational resources has enabled scientists and researchers to generate huge volumes of data for multiphase flows. For routine design calculations, detailed numerical simulations of multiphase flows with disparate time and length scales and turbulence resolution (DNS/ LES) are resource intensive and computationally expensive. A data-driven surrogate model built with available simulation and experimental data can be used instead of intensive numerical calculations and in the absence of an available mathematical models. Such an attempt is made in this thesis, where data-driven surrogate models are built and developed with applications to two-phase flows.This study is the first of its kind to have developed data-driven surrogate modelling frameworks and algorithms with applications to spatiotemporally varying, statistically stationary and steady state multiphase flows. This work discusses the frameworks necessary for identifying and conducting the surrogate modeling process and with quantified errors. GPs were selected as the algorithm of choice as they can be used with the relatively fewer number of numerical simulations data for training purposes. The GP algorithm was modified to work with multiple independent variables of a multiphase flow configuration. When multiple ML algorithms are available and applicable for the same multiphase flow configuration, it becomes necessary to know how each of the algorithm will perform in terms of efficiency, accuracy and speedup. In this thesis, performance comparison of GP and NN machine algorithms was conducted for accuracy and speedup for a 2D Rayleigh-Taylor instability configuration. Prediction results were reported for accuracy, speedup and efficiency. This thesis addresses three major issues for building, developing and comparing various learning algorithms for application to multiphase flows, by identifying a set of guidelines to -1. build and develop data-driven surrogate models based on Gaussian processes for spatiotemporally varying and statistically stationary multiphase flows.2. modify Gaussian processes to enhance prediction capability of multiphase flows with more than one independent variable.3. performance comparison of Gaussian Process and Neural Network based Machine Learning frameworks for efficiency, accuracy and speedup with extensions to other possible machine learning algorithms.It is hoped that this research contributed in establishing guidelines and approaches for building computationally inexpensive data-driven surrogate models for multiphase flows that can replace numerical simulations for initial and routine design for power generation devices.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2024
538 $a Mode of access: World Wide Web
650 4 $a Computational physics. $3 1181955
650 4 $a Statistical physics. $3 528048
650 4 $a Aerospace engineering. $3 686400
653 $a Multiphase flows
653 $a Gaussian processes
653 $a Machine learning algorithms
653 $a Power generation devices
653 $a Surrogate modeling process
655 7 $a Electronic books. $2 local $3 554714
690 $a 0538
690 $a 0217
690 $a 0800
690 $a 0216
710 2 $a University of Cincinnati. $b Engineering and Applied Science: Aerospace Engineering. $3 1470399
710 2 $a ProQuest Information and Learning Co. $3 1178819
773 0 $t Dissertations Abstracts International $g 85-05B.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30866464 $z click for full text (PQDT)