서브메뉴
검색
상세정보
State estimation of spatio-temporal phenomena- [electronic resource]
State estimation of spatio-temporal phenomena- [electronic resource]
상세정보
- 자료유형
- 학위논문(국외)
- 자관 청구기호
- 기본표목-개인명
- 표제와 책임표시사항
- State estimation of spatio-temporal phenomena - [electronic resource] / Yu, Dan.
- 발행, 배포, 간사 사항
- 발행, 배포, 간사 사항
- 형태사항
- 1 online resource(208 p)
- 일반주기
- Source: Dissertation Abstracts International, Volume: 78-08(E), Section: B.
- 일반주기
- Adviser: Suman Chakravorty.
- 학위논문주기
- Thesis (Ph.D.)--Texas A&M University, 2016.
- 요약 등 주기
- 요약This dissertation addresses the state estimation problem of spatio-temporal phenomena which can be modeled by partial differential equations (PDEs), such as pollutant dispersion in the atmosphere. After discretizing the PDE, the dynamical system has a large number of degrees of freedom (DOF). State estimation using Kalman Filter (KF) is computationally intractable, and hence, a reduced order model (ROM) needs to be constructed first. Moreover, the nonlinear terms, external disturbances or unknown boundary conditions can be modeled as unknown inputs, which leads to an unknown input filtering problem. Furthermore, the performance of KF could be improved by placing sensors at feasible locations. Therefore, the sensor scheduling problem to place multiple mobile sensors is of interest.
- 요약 등 주기
- 요약The first part of the dissertation focuses on model reduction for large scale systems with a large number of inputs/outputs. A commonly used model reduction algorithm, the balanced proper orthogonal decomposition (BPOD) algorithm, is not computationally tractable for large systems with a large number of inputs/outputs. Inspired by the BPOD and randomized algorithms, we propose a randomized proper orthogonal decomposition (RPOD) algorithm and a computationally optimal RPOD (RPOD*) algorithm, which construct an ROM to capture the input-output behaviour of the full order model, while reducing the computational cost of BPOD by orders of magnitude. It is demonstrated that the proposed RPOD* algorithm could construct the ROM in real-time, and the performance of the proposed algorithms on different advection-diffusion equations.
- 요약 등 주기
- 요약Next, we consider the state estimation problem of linear discrete-time systems with unknown inputs which can be treated as a wide-sense stationary process with rational power spectral density, while no other prior information needs to be known. We propose an autoregressive (AR) model based unknown input realization technique which allows us to recover the input statistics from the output data by solving an appropriate least squares problem, then fit an AR model to the recovered input statistics and construct an innovations model of the unknown inputs using the eigensystem realization algorithm. The proposed algorithm outperforms the augmented two-stage Kalman Filter (ASKF) and the unbiased minimum-variance (UMV) algorithm are shown in several examples.
- 요약 등 주기
- 요약Finally, we propose a framework to place multiple mobile sensors to optimize the long-term performance of KF in the estimation of the state of a PDE. The major challenges are that placing multiple sensors is an NP-hard problem, and the optimization problem is non-convex in general. In this dissertation, first, we construct an ROM using RPOD* algorithm, and then reduce the feasible sensor locations into a subset using the ROM. The Information Space Receding Horizon Control (I-RHC) approach and a modified Monte Carlo Tree Search (MCTS) approach are applied to solve the sensor scheduling problem using the subset. Various applications have been provided to demonstrate the performance of the proposed approach.
- 주제명부출표목-일반주제명
- 부출표목-단체명
- 기본자료저록
- Dissertation Abstracts International. 78-08B(E).
- 기본자료저록
- Dissertation Abstract International
- 전자적 위치 및 접속
- 원문정보보기
- 소장사항
-
20180515 2018
MARC
008180601s2016 us esm 001c eng■001MOKWON01260640
■00520180518093021
■007cr
■020 ▼a9781369692792
■035 ▼a(MiAaPQ)0803vireo:18168Yu
■035 ▼a(MiAaPQ)AAI10588187
■040 ▼aMiAaPQ▼cMiAaPQ
■090 ▼a전자도서(박사논문)
■1001 ▼aYu, Dan.
■24510▼aState estimation of spatio-temporal phenomena▼h[electronic resource]▼cYu, Dan.
■260 ▼a[Sl]▼bTexas A&M University▼c2016
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2016
■300 ▼a1 online resource(208 p)
■500 ▼aSource: Dissertation Abstracts International, Volume: 78-08(E), Section: B.
■500 ▼aAdviser: Suman Chakravorty.
■5021 ▼aThesis (Ph.D.)--Texas A&M University, 2016.
■520 ▼aThis dissertation addresses the state estimation problem of spatio-temporal phenomena which can be modeled by partial differential equations (PDEs), such as pollutant dispersion in the atmosphere. After discretizing the PDE, the dynamical system has a large number of degrees of freedom (DOF). State estimation using Kalman Filter (KF) is computationally intractable, and hence, a reduced order model (ROM) needs to be constructed first. Moreover, the nonlinear terms, external disturbances or unknown boundary conditions can be modeled as unknown inputs, which leads to an unknown input filtering problem. Furthermore, the performance of KF could be improved by placing sensors at feasible locations. Therefore, the sensor scheduling problem to place multiple mobile sensors is of interest.
■520 ▼aThe first part of the dissertation focuses on model reduction for large scale systems with a large number of inputs/outputs. A commonly used model reduction algorithm, the balanced proper orthogonal decomposition (BPOD) algorithm, is not computationally tractable for large systems with a large number of inputs/outputs. Inspired by the BPOD and randomized algorithms, we propose a randomized proper orthogonal decomposition (RPOD) algorithm and a computationally optimal RPOD (RPOD*) algorithm, which construct an ROM to capture the input-output behaviour of the full order model, while reducing the computational cost of BPOD by orders of magnitude. It is demonstrated that the proposed RPOD* algorithm could construct the ROM in real-time, and the performance of the proposed algorithms on different advection-diffusion equations.
■520 ▼aNext, we consider the state estimation problem of linear discrete-time systems with unknown inputs which can be treated as a wide-sense stationary process with rational power spectral density, while no other prior information needs to be known. We propose an autoregressive (AR) model based unknown input realization technique which allows us to recover the input statistics from the output data by solving an appropriate least squares problem, then fit an AR model to the recovered input statistics and construct an innovations model of the unknown inputs using the eigensystem realization algorithm. The proposed algorithm outperforms the augmented two-stage Kalman Filter (ASKF) and the unbiased minimum-variance (UMV) algorithm are shown in several examples.
■520 ▼aFinally, we propose a framework to place multiple mobile sensors to optimize the long-term performance of KF in the estimation of the state of a PDE. The major challenges are that placing multiple sensors is an NP-hard problem, and the optimization problem is non-convex in general. In this dissertation, first, we construct an ROM using RPOD* algorithm, and then reduce the feasible sensor locations into a subset using the ROM. The Information Space Receding Horizon Control (I-RHC) approach and a modified Monte Carlo Tree Search (MCTS) approach are applied to solve the sensor scheduling problem using the subset. Various applications have been provided to demonstrate the performance of the proposed approach.
■590 ▼aSchool code: 0803.
■650 4▼aAerospace engineering
■690 ▼a0538
■71020▼aTexas A&M University▼bAerospace Engineering.
■7730 ▼tDissertation Abstracts International▼g78-08B(E).
■773 ▼tDissertation Abstract International
■790 ▼a0803
■791 ▼aPh.D.
■792 ▼a2016
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T14821035▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
■980 ▼a20180515▼f2018
![State estimation of spatio-temporal phenomena - [electronic resource] / Yu, Dan.](/Sponge/Images/bookDefaults/DDbookdefaultsmall.png)


한글
ENG
日本
中文
Việt Nam