TU Darmstadt / ULB / TUprints

Simulation reaktiver Mehrphasenströmungen mit einer Momentenmethode und tabellierter Chemie

Synek, Benjamin (2018)
Simulation reaktiver Mehrphasenströmungen mit einer Momentenmethode und tabellierter Chemie.
Technische Universität Darmstadt
Ph.D. Thesis, Primary publication

[img]
Preview
Text
Dissertation.pdf - Accepted Version
Copyright Information: CC BY-NC-ND 4.0 International - Creative Commons, Attribution NonCommercial, NoDerivs.

Download (5MB) | Preview
Item Type: Ph.D. Thesis
Type of entry: Primary publication
Title: Simulation reaktiver Mehrphasenströmungen mit einer Momentenmethode und tabellierter Chemie
Language: German
Referees: Sadiki, Prof. Dr. Amsini ; Lang, Prof. Dr. Jens ; Janicka, Prof. Dr. Johannes
Date: 8 December 2018
Place of Publication: Darmstadt
Publisher: tuprints
Date of oral examination: 24 October 2017
Abstract:

Vor dem Hintergrund des Inkrafttretens strengerer Emissionsgesetze, sowie der Entwicklung neuer Triebwerkstechnologien besteht ein hohes Interesse am Einsatz numerischer Simulationen zur Auslegung und Optimierung von Verbrennungsprozessen. Die angewendeten Modelle beinhalten Ansätze zur Beschreibung der komplexen physikalischen Phänomene wie z.B. dem Zerfall des Brennstoffstrahls, der Dispersion der Tropfen, der Interaktion der Tropfen mit der Gasphase, sowie der Massen- und Wärmetransfermechanismen. Ausgehend von der Beschreibung der Gemischbildung können für reaktive Konfigurationen somit Aussagen über die Verbrennung, bzw. deren Effizienz, sowie die Schadstoffbildung gemacht werden.

Aufgrund des hohen numerischen Aufwands sind detaillierte Betrachtungsweisen sehr aufwendig und derzeit noch nicht kommerziell anwendbar. Numerische Modelle zur Beschreibung von Spray-Simulationen basieren daher zumeist auf Verteilungsfunktionen, die zeitliche und räumliche Veränderung der Verteilungsfunktion wird mit einer kinetischen Boltzmann-Gleichung beschrieben. Zur Lösung dieser Gleichung stehen zwei Ansätze zur Verfügung: Stochastische Lagrange und Euler-basierte Momentenmethoden. Stochastische Lagrange-Methoden beinhalten hierbei zwar eine akkurate Beschreibung der auftretenden physikalischen Phänomene, der zu erwartende Rechenaufwand und die optimale Parallelisierung sind jedoch bisher herausfordernde Fragestellungen. Momentenmethoden stellen eine effiziente Alternative dar und bieten Vorteile hinsichtlich der effizienten Parallelisierung, sowie der Kopplung der Phasen.

Im Rahmen dieser Arbeit wird ein hybrider Ansatz angewendet, der eine quadraturbasierte Momentenmethode, die DQMoM (Direct Quadrature method of Moments), mit einer sektionalen Formulierung der Verteilungsfunktion kombiniert. Der Ansatz basiert hierbei auf einer Aufteilung der Verteilungsfunktion auf Basis eines operator-splitting-Ansatzes in einen physikalischen und phasenbasierten Raum mit der entsprechenden separaten Betrachtung der auftretenden Phänomene. Zur Beschreibung der Massen- und Wärmetransfermechanismen innerhalb der Sektionen wird die EMSM-Methode (Eulerian Multi-Size Moment), sowie ein realistisches Gleichgewichtsmodell von Abramzon und Sirignano angewendet.

Sobald ein zünd- und brennbares Gemisch aus Oxidator und verdampftem Brennstoff vorliegt, finden chemische Reaktionen statt. Zur numerischen Lösung der chemischen Reaktionen stehen drei grundlegende, unterschiedlich komplexe Ansätzen zur Auswahl. Diese Ansätze basieren auf der Lösung weniger globaler Gleichungen, der Lösung mit Zugriff auf vorab tabellierte, chemische Reaktionen und deren Zwischenreaktionen, sowie der Lösung einer Vielzahl detaillierter Reaktionsgleichungen.

Bei einer detaillierten Betrachtung wird für jede vorkommende Spezies eine Transportgleichung aufgestellt, die Lösung dieser Mechanismen ist somit sehr rechenzeitaufwendig. Die Lösung weniger globaler Gleichungen wiederum deckt viele Aspekte von Interesse nicht ab - aus diesem Grund werden oftmals tabellierte Ansätze angewendet. Die Reaktionsmechanismen werden dabei in Tabellen abgespeichert, auf die mithilfe der sog. Kontrollvariablen während der Simulation zugegriffen wird. Die Erzeugung der Tabellen erfolgt im Vorfeld der eigentlichen Simulation und basiert auf dem Ansatz, dass eine sich ausbreitende Flamme anhand einer Vielzahl kleiner, eindimensionaler Flammen dargestellt werden kann. Die Interaktion zwischen turbulenten Strukturen und chemischen Reaktionen wird mit Wahrscheinlichkeitsdichtefunktionen realisiert.

Die Validierung der Kopplung erfolgt im Rahmen eines polydispersen, reaktiven Sprays. Die erzielten numerischen Resultate werden mit experimentellen Resultaten von drei ausgewählten Konfigurationen des sog. Sydney-Spray-Brenners für flüssige Brennstoffe unter atmosphärischen Umgebungsbedingungen verglichen.

Alternative Abstract:
Alternative AbstractLanguage

In liquid-driven combustion systems, the fuel is sprayed as a cloud of droplets into the combustion chamber. These droplets evaporate and promote the preparation of a suitable fuel-air mixture prior to combustion. Together with the injection process, the evaporation of droplets influences the evolution of droplet distribution that, in turn, strongly affects the flame structure, combustion efficiency and emissions. Therefore, it is of great interest to study how the droplet evaporation and combustion modeling, that includes heat, mass and momentum transfer processes in both gas and liquid phase, would influence the numerical prediction of spray properties in practical spray applications. On a mesoscopic level, numerical tools for the spray simulation are based on a kinetic description given in terms of a number density function (NDF). The evolution of this NDF is described by the Williams spray equation. In general, there are two approaches to solve the Williams equation: the stochastic Lagrangian and the Eulerian moment methods. In the present thesis the second one is used to solve this equation. The general approach is to calculate the evolution of the moments of the NDF with a set of Navier-Stokes-like transport equations.

The stochastic Lagrangian methods approximate the solution of the Williams equation by individually tracking numerical particles (parcels) within a Lagrangian approach. They are considered as the most robust, reliable and accurate methods for the spray simulation, allowing a straightforward description of the occurring physical phenomena. Thereby, as each parcel often represents several real droplets of same properties in the spray system, the computational cost are being held at a reasonable level. However, some major drawbacks are associated with the stochastic Lagrangian methods. First, even though a parcel approach is considered, the computational cost is high and strongly dependent on the unsteadiness of the flow and mass loading of the system. In order to avoid a high level of statistical noise and achieve an accurate and smooth continuous field of droplet averaged properties, a sufficient amount of sampled parcels is required within each relevant grid cell during each numerical time-step. This required amount of parcels increases significantly for unsteady simulations with fine time and spatial discretization. Especially in the case of Large Eddy Simulation (LES), the necessary amount of sampled parcels is considerably higher in comparison to RANS models. Indeed, LES requires a strong coupling between continuous and disperse phases, significantly increasing the implementation complexity of the Lagrangian method. Nevertheless, a RANS-approach is followed in this thesis. Finally, since most parcels are located in a small region of the computational domain, an optimal parallelization of the solver is hardly demanding. Due to these drawbacks, a great interest for alternative methods has emerged. Recently, the Eulerian moment methods have gained special attention. Their general strategy is to calculate the evolution of the moments of the NDF with a set of Navier-Stokes-like transport equations. These transport equations are directly derived from the Williams equation. Their structure and, consequently, their computational cost, do not depend on the droplet mass loading or the unsteadiness of the system. Since the equations of both disperse and continuous phase share the same structure, the coupling between both phases is straightforward and solver parallelization can be easily optimized. Therefore, the Eulerian moment methods are promising alternatives to the Lagrangian approach.

Numerous Eulerian approaches and their variations have been proposed, each with its advantages and drawbacks. Among prominent methods are the Quadrature Method of Moments (QMOM), the Direct Quadrature Method of Moments (DQMOM) and the Sectional Methods (SM). The DQMOM is a robust approach for describing the transport in phase space. However, a major challenge involves the accurate description of convective transport and drag effects. The general DQMOM approach is to approximate the NDF by a few quadrature points and assume that all droplets of a certain size locally share the same velocity (mono-kinetic assumption). Hence, regarding convective transport and drag, these quadrature points are handled as ordinary monodisperse phases of a multi-fluid method. Indeed, the range of droplet sizes and velocities is substantial in most spray systems. Thus, a significant error associated with convective transport and drag is observed if only a few droplet size classes are considered. This issue has been recently tackled with a so-called Coupled Size-Velocity Moments (CSVM) model, which addresses droplet size-velocity correlations in drag and convectional transport processes within a moment method framework. Within a DQMOM framework, the most immediate strategy to reduce the error related to convective transport and drag is to increase the amount of quadrature points. However, the amount of quadrature points cannot be arbitrarily increased without compromising the well conditioning of the DQMOM solver. Another challenge associated to DQMOM is the accurate prediction of the evaporative flux, since the NDF is not known. Recently, the DQbSMOM, which combines the Direct Quadrature Method of Moments (DQMOM) with a SM, was introduced. The strategy of the DQbSMOM is to split the droplet size domain into so-called sections and to approximate the NDF in each section using DQMOM. Including the droplet evaporation process, the evaporative flux between two adjacent sections is described by applying the Eulerian Multi-Size Moment (EMSM) method using the Ambramzon-Sirignano model, as well as using using the Abramzon-Sirignano model - an equlibrium evaporation model - and the Langmuir-Knudsen model, that numbers among the non-equilibrium evaporation models.

The aim of this thesis is to extend the DQbSMOM to the Flamelet Generated Manifold (FGM) model in order to describe reactive spray systems. For simulating reactive spray systems, a modified FGM model is used to account for multiphase combustion phenomena. Due to the combustion-induced strong decrease of the droplets' diameter and volume fraction, two additional, surface-based approaches have been implemented. The first approach concerns a dynamic subiteration within the evaporation model in case of exceeding a CFL-like criterion, based upon the ratio of evaporation rate and timestep, while the other approach deals with an adapted formulation of the fixed sections in each computational cell. The complete model is successfully evaluated in terms of droplet mean diameters, droplet number concentration and droplet flux, as well as droplet velocity components and temperature of the continuous phase using the so-called Sydney spray burner, with respect to a comparison between a configuration for evaporation and evaporation with subsequent combustion, in order to emphasize the capabilities of the DQbSMOM. The Sydney spray burner consists of a piloted jet in coflow configuration, whereas experimental results are available for non-reactive and reactive acetone, as well as reactive ethanol configurations. The presented numerical results show three configurations with variing boundary conditions. The numerical results have been achieved using \textit{Precise/UNS}, a Finite-Volume-Method-based (FVM) in-house code of \textit{Rolls-Royce Germany}.

English
URN: urn:nbn:de:tuda-tuprints-76086
Classification DDC: 600 Technology, medicine, applied sciences > 620 Engineering and machine engineering
Divisions: 16 Department of Mechanical Engineering > Institute for Energy and Power Plant Technology (EKT)
Date Deposited: 31 Jul 2018 08:37
Last Modified: 09 Jul 2020 02:10
URI: https://tuprints.ulb.tu-darmstadt.de/id/eprint/7608
PPN: 434456675
Export:
Actions (login required)
View Item View Item