Seitz, Tobias (2018)
Geometry identification and data enhancement for distributed flow measurements.
Technische Universität Darmstadt
Dissertation, Erstveröffentlichung
Kurzbeschreibung (Abstract)
The measurement of fluid motion is an important tool for researchers in fluiddynamics. Measurements with increasing precision did expedite the development of fluid-dynamic models and their theoretical understanding. Several well-established experimental techniques provide point-wise information on the flow field. In recent years novel measurement modalities have been investigated which deliver spatially resolved three-dimensional velocity measurements. Note that for methods such as particle tracking and tomographic particle imaging optical access to the flow domain is necessary. For other methods like magnetic resonance velocimetry, CT-angiography, or x-ray velocimetry this is, however, not the case. Such a property and also the fact that those methods are able to provide three-dimensional velocity fields in a rather short acquisition time makes them in particular suited for in-vivo applications.
Our work is motivated by such non-invasive velocity measurement techniques for which no optical access to the interior of the geometry is needed and also not available in many cases. Here, an additional difficulty is that the exact flow geometry is in general not known a priori. The measurement techniques we are interested in, are extensions of already available medical imaging modalities. As a prototypical example, we consider magnetic resonance velocimetry, which is also suited for the measurement of turbulent fluid motion. We will also discuss computational examples using such measurement data.
General purpose.
Our main goal is a suitable post-processing of the available velocity data and also to obtain additional information. The measurements available from magnetic resonance velocimetry consist of several components given on a fixed field of view. The magnitude of the MRT signal corresponds to a proton density and thus e.g. the density of water molecules. Those data typically give a clear indication of the position and size of the flow geometry. The velocity data, on the other hand, are substantially perturbed outside the flow domain. This is a typical feature of measurements stemming from magnetic resonance velocimetry. Note that the surrounding noise usually has a notably higher magnitude than the actual measurements.
Thus, a first necessary step will be to somehow separate the domain containing valuable velocity data from the noise surrounding it. For this reason, we apply some kind of image segmentation where we make use of the given den-sity image. Since the velocity values are given on the same field of view the segmentation directly transfers to those data.
Due to the measurement procedure also the segmented velocity data are contaminated by measurement errors. Therefore, besides segmentation, additional post-processing is necessary in order to make the flow measurements available for further usage. In a second step, we propose a problem adapted data enhancement method which is able to provide a smoothed velocity field on the one hand, and also provides additional information on the other hand, like for instance the pressure drop or an estimate for the wall shear stress.
The two main steps will therefore be:
(i) The identification of the flow geometry, where we make use of the available density measurements.
(ii) The denoising and improvement of the segmented velocity data, by using a suitable fluid-dynamical model.
Outline.
In part I of this thesis, we introduce our basic approach to the geom- etry identification and velocity enhancement problems described above. Both problems are formulated as optimal control problems governed by a partial differential equation and we shortly discuss some general aspects of the analysis and the solution of such problems in section 4.
In part II, we thoroughly discuss and analyze the geometry identification problem introduced in section 2. The procedure is formulated as an inverse ill-posed problem and we propose a Tikhonov regularization for its stable solution. We show that the resulting optimal control problem has a solution and discuss its numerical treatment with iterative methods. Finally, a systematic discretization can be realized using finite elements which is also demonstrated by numerical tests.
The velocity enhancement problem is introduced in part III. We propose a linearized flow-model which directly incorporates the available measurements. The resulting modeling error can be quantified in terms of the data error. The reconstruction method is then formulated as an optimal control problem subject to the linearized equations. We show the existence of a unique solution and derive estimates for the reconstruction error. Additionally, a reconstruction for the pressure is obtained for which we derive similar error estimates. We discuss the systematic discretization using finite elements and show preliminary computational examples for the verification of the derived estimates.
In order to verify the applicability of the proposed methods to realistic data, we consider an application using experimental data in part IV. We use measurements of a human blood vessel stemming from magnetic resonance velocimetry obtained at the University Medical Center in Freiburg. After a suitable pre-processing of the available data, we apply the geometry identification method in order to obtain a discretization of the blood vessel. Using the generated mesh, we reconstruct an enhanced velocity field and the pressure from the available velocity data.
Typ des Eintrags: | Dissertation | ||||
---|---|---|---|---|---|
Erschienen: | 2018 | ||||
Autor(en): | Seitz, Tobias | ||||
Art des Eintrags: | Erstveröffentlichung | ||||
Titel: | Geometry identification and data enhancement for distributed flow measurements | ||||
Sprache: | Englisch | ||||
Referenten: | Egger, Prof. Dr. Herbert ; Tropea, Prof. Dr. Cameron ; Notsu, Prof. Dr. Hirofumi | ||||
Publikationsjahr: | 2018 | ||||
Ort: | Darmstadt | ||||
Datum der mündlichen Prüfung: | 19 Dezember 2017 | ||||
URL / URN: | http://tuprints.ulb.tu-darmstadt.de/7254 | ||||
Kurzbeschreibung (Abstract): | The measurement of fluid motion is an important tool for researchers in fluiddynamics. Measurements with increasing precision did expedite the development of fluid-dynamic models and their theoretical understanding. Several well-established experimental techniques provide point-wise information on the flow field. In recent years novel measurement modalities have been investigated which deliver spatially resolved three-dimensional velocity measurements. Note that for methods such as particle tracking and tomographic particle imaging optical access to the flow domain is necessary. For other methods like magnetic resonance velocimetry, CT-angiography, or x-ray velocimetry this is, however, not the case. Such a property and also the fact that those methods are able to provide three-dimensional velocity fields in a rather short acquisition time makes them in particular suited for in-vivo applications. Our work is motivated by such non-invasive velocity measurement techniques for which no optical access to the interior of the geometry is needed and also not available in many cases. Here, an additional difficulty is that the exact flow geometry is in general not known a priori. The measurement techniques we are interested in, are extensions of already available medical imaging modalities. As a prototypical example, we consider magnetic resonance velocimetry, which is also suited for the measurement of turbulent fluid motion. We will also discuss computational examples using such measurement data. General purpose. Our main goal is a suitable post-processing of the available velocity data and also to obtain additional information. The measurements available from magnetic resonance velocimetry consist of several components given on a fixed field of view. The magnitude of the MRT signal corresponds to a proton density and thus e.g. the density of water molecules. Those data typically give a clear indication of the position and size of the flow geometry. The velocity data, on the other hand, are substantially perturbed outside the flow domain. This is a typical feature of measurements stemming from magnetic resonance velocimetry. Note that the surrounding noise usually has a notably higher magnitude than the actual measurements. Thus, a first necessary step will be to somehow separate the domain containing valuable velocity data from the noise surrounding it. For this reason, we apply some kind of image segmentation where we make use of the given den-sity image. Since the velocity values are given on the same field of view the segmentation directly transfers to those data. Due to the measurement procedure also the segmented velocity data are contaminated by measurement errors. Therefore, besides segmentation, additional post-processing is necessary in order to make the flow measurements available for further usage. In a second step, we propose a problem adapted data enhancement method which is able to provide a smoothed velocity field on the one hand, and also provides additional information on the other hand, like for instance the pressure drop or an estimate for the wall shear stress. The two main steps will therefore be: (i) The identification of the flow geometry, where we make use of the available density measurements. (ii) The denoising and improvement of the segmented velocity data, by using a suitable fluid-dynamical model. Outline. In part I of this thesis, we introduce our basic approach to the geom- etry identification and velocity enhancement problems described above. Both problems are formulated as optimal control problems governed by a partial differential equation and we shortly discuss some general aspects of the analysis and the solution of such problems in section 4. In part II, we thoroughly discuss and analyze the geometry identification problem introduced in section 2. The procedure is formulated as an inverse ill-posed problem and we propose a Tikhonov regularization for its stable solution. We show that the resulting optimal control problem has a solution and discuss its numerical treatment with iterative methods. Finally, a systematic discretization can be realized using finite elements which is also demonstrated by numerical tests. The velocity enhancement problem is introduced in part III. We propose a linearized flow-model which directly incorporates the available measurements. The resulting modeling error can be quantified in terms of the data error. The reconstruction method is then formulated as an optimal control problem subject to the linearized equations. We show the existence of a unique solution and derive estimates for the reconstruction error. Additionally, a reconstruction for the pressure is obtained for which we derive similar error estimates. We discuss the systematic discretization using finite elements and show preliminary computational examples for the verification of the derived estimates. In order to verify the applicability of the proposed methods to realistic data, we consider an application using experimental data in part IV. We use measurements of a human blood vessel stemming from magnetic resonance velocimetry obtained at the University Medical Center in Freiburg. After a suitable pre-processing of the available data, we apply the geometry identification method in order to obtain a discretization of the blood vessel. Using the generated mesh, we reconstruct an enhanced velocity field and the pressure from the available velocity data. |
||||
Alternatives oder übersetztes Abstract: |
|
||||
URN: | urn:nbn:de:tuda-tuprints-72540 | ||||
Sachgruppe der Dewey Dezimalklassifikatin (DDC): | 500 Naturwissenschaften und Mathematik > 510 Mathematik | ||||
Fachbereich(e)/-gebiet(e): | DFG-Graduiertenkollegs > Graduiertenkolleg 1529 Mathematical Fluid Dynamics Exzellenzinitiative > Graduiertenschulen > Graduate School of Computational Engineering (CE) 04 Fachbereich Mathematik 04 Fachbereich Mathematik > Numerik und wissenschaftliches Rechnen DFG-Graduiertenkollegs Exzellenzinitiative > Graduiertenschulen Exzellenzinitiative |
||||
Hinterlegungsdatum: | 22 Apr 2018 19:55 | ||||
Letzte Änderung: | 22 Apr 2018 19:55 | ||||
PPN: | |||||
Referenten: | Egger, Prof. Dr. Herbert ; Tropea, Prof. Dr. Cameron ; Notsu, Prof. Dr. Hirofumi | ||||
Datum der mündlichen Prüfung / Verteidigung / mdl. Prüfung: | 19 Dezember 2017 | ||||
Export: | |||||
Suche nach Titel in: | TUfind oder in Google |
Frage zum Eintrag |
Optionen (nur für Redakteure)
Redaktionelle Details anzeigen |