東華大學圖書館 |

Data-Driven Modeling and Estimation of Unsteady Flow Past Aerodynamic Surfaces.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Data-Driven Modeling and Estimation of Unsteady Flow Past Aerodynamic Surfaces./
作者:	Graff, John.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2023,
面頁冊數:	114 p.
附註:	Source: Dissertations Abstracts International, Volume: 85-03, Section: B.
Contained By:	Dissertations Abstracts International85-03B.
標題:	Aerospace engineering. -
電子資源:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30637904
ISBN:	9798380349307

Data-Driven Modeling and Estimation of Unsteady Flow Past Aerodynamic Surfaces.
Graff, John.

Data-Driven Modeling and Estimation of Unsteady Flow Past Aerodynamic Surfaces. - Ann Arbor : ProQuest Dissertations & Theses, 2023 - 114 p.

Source: Dissertations Abstracts International, Volume: 85-03, Section: B.

Thesis (Ph.D.)--State University of New York at Buffalo, 2023.

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

Estimation of unsteady flow-fields around flight vehicles may improve flow interactions and lead to enhanced vehicle performance. Recent progress in computing has enabled the development of data-driven techniques for developing models of dynamical systems such as fluid flows. Although flow-field representations can be very high-dimensional, their dynamics can have low-order representations, leading naturally to the question of model reduction. Further, many low-order models can be estimated using a few, appropriately placed measurements. This dissertation examines the construction of reduced-order models through sparse regression and the construction of sparse sensor arrays for unsteady flow estimation. An algorithm from machine learning, known as Least Angle Regression (LARS), is applied to two problems in flow modeling: 1) the selection of modes resulting from Dynamic Mode Decomposition (DMD), and 2) the discovery of nonlinear equations from data that govern the motion of coherent structures where LARS is used as the sparse regressor in the Sparse Identification of Nonlinear Dynamics (SINDy) process.DMD builds a linear, approximate model of a system's dynamics from data and this work reduces the order of this model by identifying a reduced set of modes that best fit the output using LARS. LARS is modified to be complex-valued to allow for the selection of complex-valued DMD modes. The resulting algorithm is referred to as Least Angle Regression for Dynamic Mode Decomposition (LARS4DMD). Sparsity-Promoting Dynamic Mode Decomposition (DMDSP), a popular mode-selection algorithm, serves as a benchmark for which to compare LARS4DMD. The sparsity parameter required for DMDSP is not needed in LARS4DMD allowing the user to more easily select a desired model cardinality. Numerical results from a Poiseuille flow test problem show that LARS4DMD yields reduced-order models that have comparable performance to DMDSP. Use of the LARS4DMD algorithm on Particle Image Velocimetry (PIV) data of a rotating fin confirms this conclusion on experimental data. Correlation between covariates in LARS can lead to inefficient selection of covariates when building sparse models. A Singular Value Decomposition (SVD) based covariate orthogonalization method is presented for removing correlation between covariates. The use of LARS with the orthogonalization step for building a system of nonlinear equations from data using the SINDy process is referred to as LARS4SINDy. The three-variable Lorenz system and two-vortex flow models provide test cases with known dynamics. LARS4SINDy successfully recovers the Lorenz equations from data and identifies the governing equations of motion for two vortices advecting in a background potential flow.This dissertation also presents a sensor-selection framework for the intended application of data-driven, flow-field estimation. This framework combines data-driven modeling, steady-state Kalman Filter design, and a sparsification technique for sequential selection of sensors. The sensor selection framework is also used to design sensor arrays that can perform well across a variety of operating conditions in 2-D numerically generated flow. Flow estimation results on the 2-D data show that the proposed framework produces arrays that are highly effective at flow-field estimation for the flow behind and an airfoil at a high angle of attack using embedded pressure sensors. Analysis of the 2-D flow-fields reveals that paths of impinging stagnation points along the airfoil's surface during a shedding period of the flow are highly informative locations for placement of pressure sensors.The sensor placement framework is also applied to 3-D unsteady flow data obtained via Particle Tracking Velocimetry (PTV). The flow is generated from a pitching NACA 0010 wing that is towed at a constant velocity. Five ensemble-averaged runs of PTV data were used to train a flow model and construct sparse sensor arrays. The arrays were then used with the associated flow model to estimate the flow-field from a test run of PTV data. Successful estimation of the large flow structures, including the wingtip vortex and Leading Edge Vortex (LEV) was achieved.

ISBN: 9798380349307Subjects--Topical Terms:

1002622
Aerospace engineering.
Subjects--Index Terms:

Vehicle performance

Data-Driven Modeling and Estimation of Unsteady Flow Past Aerodynamic Surfaces.
LDR:05390nmm a2200373 4500 001 2397266
005 20240617111359.5
006 m o d
007 cr#unu||||||||
008 251215s2023 ||||||||||||||||| ||eng d
020 $a 9798380349307
035 $a (MiAaPQ)AAI30637904
035 $a AAI30637904
040 $a MiAaPQ $c MiAaPQ
100 1 $a Graff, John. $0 (orcid)0009-0009-2542-8808 $3 3767028
245 1 0 $a Data-Driven Modeling and Estimation of Unsteady Flow Past Aerodynamic Surfaces.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2023
300 $a 114 p.
500 $a Source: Dissertations Abstracts International, Volume: 85-03, Section: B.
500 $a Advisor: Lagor, Francis.
502 $a Thesis (Ph.D.)--State University of New York at Buffalo, 2023.
506 $a This item must not be sold to any third party vendors.
520 $a Estimation of unsteady flow-fields around flight vehicles may improve flow interactions and lead to enhanced vehicle performance. Recent progress in computing has enabled the development of data-driven techniques for developing models of dynamical systems such as fluid flows. Although flow-field representations can be very high-dimensional, their dynamics can have low-order representations, leading naturally to the question of model reduction. Further, many low-order models can be estimated using a few, appropriately placed measurements. This dissertation examines the construction of reduced-order models through sparse regression and the construction of sparse sensor arrays for unsteady flow estimation. An algorithm from machine learning, known as Least Angle Regression (LARS), is applied to two problems in flow modeling: 1) the selection of modes resulting from Dynamic Mode Decomposition (DMD), and 2) the discovery of nonlinear equations from data that govern the motion of coherent structures where LARS is used as the sparse regressor in the Sparse Identification of Nonlinear Dynamics (SINDy) process.DMD builds a linear, approximate model of a system's dynamics from data and this work reduces the order of this model by identifying a reduced set of modes that best fit the output using LARS. LARS is modified to be complex-valued to allow for the selection of complex-valued DMD modes. The resulting algorithm is referred to as Least Angle Regression for Dynamic Mode Decomposition (LARS4DMD). Sparsity-Promoting Dynamic Mode Decomposition (DMDSP), a popular mode-selection algorithm, serves as a benchmark for which to compare LARS4DMD. The sparsity parameter required for DMDSP is not needed in LARS4DMD allowing the user to more easily select a desired model cardinality. Numerical results from a Poiseuille flow test problem show that LARS4DMD yields reduced-order models that have comparable performance to DMDSP. Use of the LARS4DMD algorithm on Particle Image Velocimetry (PIV) data of a rotating fin confirms this conclusion on experimental data. Correlation between covariates in LARS can lead to inefficient selection of covariates when building sparse models. A Singular Value Decomposition (SVD) based covariate orthogonalization method is presented for removing correlation between covariates. The use of LARS with the orthogonalization step for building a system of nonlinear equations from data using the SINDy process is referred to as LARS4SINDy. The three-variable Lorenz system and two-vortex flow models provide test cases with known dynamics. LARS4SINDy successfully recovers the Lorenz equations from data and identifies the governing equations of motion for two vortices advecting in a background potential flow.This dissertation also presents a sensor-selection framework for the intended application of data-driven, flow-field estimation. This framework combines data-driven modeling, steady-state Kalman Filter design, and a sparsification technique for sequential selection of sensors. The sensor selection framework is also used to design sensor arrays that can perform well across a variety of operating conditions in 2-D numerically generated flow. Flow estimation results on the 2-D data show that the proposed framework produces arrays that are highly effective at flow-field estimation for the flow behind and an airfoil at a high angle of attack using embedded pressure sensors. Analysis of the 2-D flow-fields reveals that paths of impinging stagnation points along the airfoil's surface during a shedding period of the flow are highly informative locations for placement of pressure sensors.The sensor placement framework is also applied to 3-D unsteady flow data obtained via Particle Tracking Velocimetry (PTV). The flow is generated from a pitching NACA 0010 wing that is towed at a constant velocity. Five ensemble-averaged runs of PTV data were used to train a flow model and construct sparse sensor arrays. The arrays were then used with the associated flow model to estimate the flow-field from a test run of PTV data. Successful estimation of the large flow structures, including the wingtip vortex and Leading Edge Vortex (LEV) was achieved.
590 $a School code: 0656.
650 4 $a Aerospace engineering. $3 1002622
650 4 $a Mechanical engineering. $3 649730
650 4 $a Fluid mechanics. $3 528155
653 $a Vehicle performance
653 $a Data-driven techniques
653 $a Model reduction
690 $a 0538
690 $a 0548
690 $a 0204
710 2 $a State University of New York at Buffalo. $b Mechanical and Aerospace Engineering. $3 1017747
773 0 $t Dissertations Abstracts International $g 85-03B.
790 $a 0656
791 $a Ph.D.
792 $a 2023
793 $a English
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30637904