서브메뉴
검색
Characterizing Multiscale Structures and Developing Hybrid Low-Order Models of Chaotic Flows
Characterizing Multiscale Structures and Developing Hybrid Low-Order Models of Chaotic Flows
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202105652
- ISBN
- 9798270298906
- DDC
- 620
- 저자명
- Guo, Alex.
- 서명/저자
- Characterizing Multiscale Structures and Developing Hybrid Low-Order Models of Chaotic Flows
- 발행사항
- [Sl] : The University of Wisconsin - Madison, 2026
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2026
- 형태사항
- 211 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-07, Section: B.
- 주기사항
- Advisor: Graham, Michael D.
- 학위논문주기
- Thesis (Ph.D.)--The University of Wisconsin - Madison, 2026.
- 초록/해제
- 요약Efficient and accurate predictive models are highly sought after in applications such as weather forecasting, chemical processes, and turbulent flows. Wall-bounded turbulence-which is the focus of this work-is a notoriously difficult system to predict due to its chaotic nature and presence of multiscale structures (i.e., eddies or vortices) that must all be resolved in simulations. This thesis contains two main goals. The first is to gain physical insight into wall-bounded turbulence by characterizing its structures. The second goal is to develop and benchmark low-order models of chaotic dynamical systems such as wall-bounded turbulence, which would be instrumental in iterative design and control of these systems due to the many model evaluations required. We start in Chapter 1 with a background on the theory of physical structures embedded in turbulent flow-their appearance and whether they are self-similar across scales-in addition to background on low-order modeling of dynamics on manifolds. In Chapter 2, we present data-driven wavelets as an ideal modal basis for representing and extracting spatially-localized, multiscale structures from data; we use this basis to discover self-similarity of structures in one-dimensional signals of turbulent pipe flow. In Chapter 3, we develop a physics-informed, data-driven reduced-order model (ROM) for forecasting spatiotemporally chaotic dynamical systems. In Chapter 4, we discuss how to combine time-delay embeddings and attention-based neural networks to reconstruct fields from partial observations of the state, which addresses a real-world challenge where sensors (e.g., those measuring atmospheric data) do not always have full coverage. In Chapter 5, we return to using wavelets to characterize structures, but this time in three-dimensional turbulent channel flow. Finally, we conclude in Chapter 6 with a general summary of the chapters and an overview of future work towards developing a high-fidelity, low-order model of wall-bounded turbulence.
- 일반주제명
- Fluid mechanics
- 일반주제명
- Applied physics
- 일반주제명
- Bioengineering
- 일반주제명
- Chemical engineering
- 키워드
- Machine learning
- 키워드
- Physics-informed
- 키워드
- Wavelets
- 기타저자
- The University of Wisconsin - Madison Chemical Engineering
- 기본자료저록
- Dissertations Abstracts International. 87-07B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2026 us c eng d■001000017361015
■00520260202105652
■006m o d
■007cr#unu||||||||
■020 ▼a9798270298906
■035 ▼a(MiAaPQ)AAI32403426
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a620
■1001 ▼aGuo, Alex.
■24510▼aCharacterizing Multiscale Structures and Developing Hybrid Low-Order Models of Chaotic Flows
■260 ▼a[Sl]▼bThe University of Wisconsin - Madison▼c2026
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2026
■300 ▼a211 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-07, Section: B.
■500 ▼aAdvisor: Graham, Michael D.
■5021 ▼aThesis (Ph.D.)--The University of Wisconsin - Madison, 2026.
■520 ▼aEfficient and accurate predictive models are highly sought after in applications such as weather forecasting, chemical processes, and turbulent flows. Wall-bounded turbulence-which is the focus of this work-is a notoriously difficult system to predict due to its chaotic nature and presence of multiscale structures (i.e., eddies or vortices) that must all be resolved in simulations. This thesis contains two main goals. The first is to gain physical insight into wall-bounded turbulence by characterizing its structures. The second goal is to develop and benchmark low-order models of chaotic dynamical systems such as wall-bounded turbulence, which would be instrumental in iterative design and control of these systems due to the many model evaluations required. We start in Chapter 1 with a background on the theory of physical structures embedded in turbulent flow-their appearance and whether they are self-similar across scales-in addition to background on low-order modeling of dynamics on manifolds. In Chapter 2, we present data-driven wavelets as an ideal modal basis for representing and extracting spatially-localized, multiscale structures from data; we use this basis to discover self-similarity of structures in one-dimensional signals of turbulent pipe flow. In Chapter 3, we develop a physics-informed, data-driven reduced-order model (ROM) for forecasting spatiotemporally chaotic dynamical systems. In Chapter 4, we discuss how to combine time-delay embeddings and attention-based neural networks to reconstruct fields from partial observations of the state, which addresses a real-world challenge where sensors (e.g., those measuring atmospheric data) do not always have full coverage. In Chapter 5, we return to using wavelets to characterize structures, but this time in three-dimensional turbulent channel flow. Finally, we conclude in Chapter 6 with a general summary of the chapters and an overview of future work towards developing a high-fidelity, low-order model of wall-bounded turbulence.
■590 ▼aSchool code: 0262.
■650 4▼aFluid mechanics
■650 4▼aApplied physics
■650 4▼aBioengineering
■650 4▼aChemical engineering
■653 ▼aMachine learning
■653 ▼aPhysics-informed
■653 ▼aReduced-order modeling
■653 ▼aTime-series prediction
■653 ▼aWavelets
■690 ▼a0204
■690 ▼a0202
■690 ▼a0542
■690 ▼a0215
■71020▼aThe University of Wisconsin - Madison▼bChemical Engineering.
■7730 ▼tDissertations Abstracts International▼g87-07B.
■790 ▼a0262
■791 ▼aPh.D.
■792 ▼a2026
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17361015▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


