東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

Data-Driven Modeling of Nuclear Syst...

Chang, Chih-Wei.

FindBook

Google Book

Amazon

博客來

Data-Driven Modeling of Nuclear System Thermal-Hydraulics.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Data-Driven Modeling of Nuclear System Thermal-Hydraulics./
作者:	Chang, Chih-Wei.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2018,
面頁冊數:	184 p.
附註:	Source: Dissertations Abstracts International, Volume: 80-02, Section: B.
Contained By:	Dissertations Abstracts International80-02B.
標題:	Aerospace engineering. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10969768
ISBN:	9780438282834

Data-Driven Modeling of Nuclear System Thermal-Hydraulics.
Chang, Chih-Wei.

Data-Driven Modeling of Nuclear System Thermal-Hydraulics. - Ann Arbor : ProQuest Dissertations & Theses, 2018 - 184 p.

Source: Dissertations Abstracts International, Volume: 80-02, Section: B.

Thesis (Ph.D.)--North Carolina State University, 2018.

This item must not be sold to any third party vendors.

The goal of this work is to develop a methodology to enhance predictive power of datadriven nuclear system thermal-hydraulics (NSTH) simulation using machine learning. NSTH simulation is instrumental for reactor design, safety analysis, and operator training. Traditionally, it takes extensive research efforts to develop insights and mechanistic understanding of physical processes in reactor system through analysis of experimental data and capture the data in a compact model form. The long time and large resources required for model development constrain the simulation code applicability in dealing with newly designed systems involving new geometries and new coolants. As an alternative to mechanistic and semi-analytical models, some machine learning methodologies, especially deep learning, can effectively capture underlying correlations behind multi-scale data using nonparametric models, or so-called data-driven models. Such approach is referred to as data-driven modeling. The technical approach of the dissertation consists of three components. First, the technical background overview navigates the essential knowledge from related disciplines, including thermal-hydraulics models, system simulation, and machine learning. Second, a methodology is developed to accomplish data-driven modeling of NSTH. The development includes a system that classifies machine learning frameworks for NSTH based on data and knowledge requirements. Finally, framework demonstration focuses on the use of deep learning, which has demonstrated the capability of a universal approximator. Synthetic examples are formulated to investigate technical challenges of using deep learning to achieve data-driven modeling of NSTH. Five machine learning frameworks for NSTH have been introduced in the dissertation including physics-separated ML (PSML or Type I ML), physics-evaluated ML (PEML or Type II ML), physics-integrated ML (PIML or Type III ML), physics-recovered (PRML or Type IV ML), and physics-discovered ML (PDML or Type V ML). The framework classification is based on knowledge and data requirements. Type III ML framework is formulated for the first time in this study. The insights obtained from synthetic examples indicate that Type III ML has the highest potential in leveraging the value from "big data" in thermal fluid research while ensuring datamodel consistency. Various numerical experiments are formulated ranging from system-level simulation to computational fluid dynamics (CFD) to exhibit the advantage of deep learning (DL) for model development. The case studies of system-level simulation using Type I, Type II, and Type III ML frameworks ensure that simulation results satisfy conservation laws with a moderate amount of data. The results indicate that system-level two-phase mixture models can be solved with DLbased closure relations without interference of numerical instability. The CFD case study exhibits that the DL-based Reynolds stress model can assimilate millions of data points to reduce forecast error. Performance of the DL-based stress can be quantified by flow features coverage mapping. The results show that Reynolds-averaged turbulence modeling with the DL-based Reynolds stress model can replicate the transient flow prediction by Reynolds-averaged Navier-Stokes simulation with the k-ϵ model.

ISBN: 9780438282834Subjects--Topical Terms:

1002622
Aerospace engineering.

Data-Driven Modeling of Nuclear System Thermal-Hydraulics.
LDR:04460nmm a2200337 4500 001 2207982
005 20190929184023.5
008 201008s2018 ||||||||||||||||| ||eng d
020 $a 9780438282834
035 $a (MiAaPQ)AAI10969768
035 $a AAI10969768
040 $a MiAaPQ $c MiAaPQ
100 1 $a Chang, Chih-Wei. $3 3434984
245 1 0 $a Data-Driven Modeling of Nuclear System Thermal-Hydraulics.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2018
300 $a 184 p.
500 $a Source: Dissertations Abstracts International, Volume: 80-02, Section: B.
500 $a Publisher info.: Dissertation/Thesis.
500 $a Advisor: Dinh, Nam T.
502 $a Thesis (Ph.D.)--North Carolina State University, 2018.
506 $a This item must not be sold to any third party vendors.
506 $a This item must not be added to any third party search indexes.
520 $a The goal of this work is to develop a methodology to enhance predictive power of datadriven nuclear system thermal-hydraulics (NSTH) simulation using machine learning. NSTH simulation is instrumental for reactor design, safety analysis, and operator training. Traditionally, it takes extensive research efforts to develop insights and mechanistic understanding of physical processes in reactor system through analysis of experimental data and capture the data in a compact model form. The long time and large resources required for model development constrain the simulation code applicability in dealing with newly designed systems involving new geometries and new coolants. As an alternative to mechanistic and semi-analytical models, some machine learning methodologies, especially deep learning, can effectively capture underlying correlations behind multi-scale data using nonparametric models, or so-called data-driven models. Such approach is referred to as data-driven modeling. The technical approach of the dissertation consists of three components. First, the technical background overview navigates the essential knowledge from related disciplines, including thermal-hydraulics models, system simulation, and machine learning. Second, a methodology is developed to accomplish data-driven modeling of NSTH. The development includes a system that classifies machine learning frameworks for NSTH based on data and knowledge requirements. Finally, framework demonstration focuses on the use of deep learning, which has demonstrated the capability of a universal approximator. Synthetic examples are formulated to investigate technical challenges of using deep learning to achieve data-driven modeling of NSTH. Five machine learning frameworks for NSTH have been introduced in the dissertation including physics-separated ML (PSML or Type I ML), physics-evaluated ML (PEML or Type II ML), physics-integrated ML (PIML or Type III ML), physics-recovered (PRML or Type IV ML), and physics-discovered ML (PDML or Type V ML). The framework classification is based on knowledge and data requirements. Type III ML framework is formulated for the first time in this study. The insights obtained from synthetic examples indicate that Type III ML has the highest potential in leveraging the value from "big data" in thermal fluid research while ensuring datamodel consistency. Various numerical experiments are formulated ranging from system-level simulation to computational fluid dynamics (CFD) to exhibit the advantage of deep learning (DL) for model development. The case studies of system-level simulation using Type I, Type II, and Type III ML frameworks ensure that simulation results satisfy conservation laws with a moderate amount of data. The results indicate that system-level two-phase mixture models can be solved with DLbased closure relations without interference of numerical instability. The CFD case study exhibits that the DL-based Reynolds stress model can assimilate millions of data points to reduce forecast error. Performance of the DL-based stress can be quantified by flow features coverage mapping. The results show that Reynolds-averaged turbulence modeling with the DL-based Reynolds stress model can replicate the transient flow prediction by Reynolds-averaged Navier-Stokes simulation with the k-ϵ model.
590 $a School code: 0155.
650 4 $a Aerospace engineering. $3 1002622
650 4 $a Mechanical engineering. $3 649730
650 4 $a Nuclear engineering. $3 595435
690 $a 0538
690 $a 0548
690 $a 0552
710 2 $a North Carolina State University. $b Nuclear Engineering. $3 2099591
773 0 $t Dissertations Abstracts International $g 80-02B.
790 $a 0155
791 $a Ph.D.
792 $a 2018
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10969768