PDF-simulation of combustion phenomena in realistic combustion chambers

Das Ziel der Arbeit ist die Entwicklung eines neuen Verbrennungsmodells zur Beschreibung der Wärmefreisetzung in realen Brennkammern. Das Modell soll einfach anwendbar, robust und für den praktischen Einsatz tauglich sein. Es soll eine Verbesserung gegenüber den etablierten Modellen darstellen und durch modularen Aufbau und gut definierte Schnittstellen zum einen ausbaubar sein und zum anderen eine Kopplung an verschiedene CFD-Codes ermöglichen. Es wird Wert darauf gelegt, dass das Modul den globalen Verbrennungsprozess beschreiben kann und sich nicht einer bestimmten Teilproblematik widmet. Dennoch soll die Möglichkeit bestehen, das Modell zu erweitern und Spezifikationen einzuführen. Bezüglich des Rechenaufwandes und des Speicherbedarfs soll das Modell eine praktische Berechnung technisch relevanter Probleme ermöglichen. Weiterhin sollen Erfahrungen und Vorteile bereits bestehender Ansätze verwendet werden. Nachteile und Beschränkungen anderer Verbrennungsmodelle sollen bei dem hier zu entwickelnden Modell vermieden werden. In diesem Zusammenhang sind insbesondere Speicherbedarf, Rechenzeit und eine breite Anwendbarkeit zu nennen. Ein Zugriff auf Tabellen soll vermieden werden, um bei dem Modul die Möglichkeit von parallelisiertem Rechnen offen zu halten und eine „Online“-Rechnung zu ermöglichen. Aufgrund dieser Vorgaben wird in der vorliegenden Arbeit ein Verbrennungsmodell entwickelt, das auf gekoppelten Wahrscheinlichkeitsdichtefunktionen und reduzierten chemischen Mechanismen beruht. Das Modell wird an einen kommerziellen Strömungslöser (FLUENT 5) angebunden. Der Aufbau ist modular gestaltet und bietet die Möglichkeit, Teilprogramme (z.B. Varianzberechnung) zu nutzen. In drei Anwendungsfällen werden etablierte Verbrennungsmodelle und das neu entwickelte Modell angewandt und miteinander verglichen. Die untersuchten Verbrennungsprozesse steigern sich in Komplexität und Praxisnähe von Fall zu Fall. Die praktische Relevanz wird durch die Einordnung in das Borghi-Diagramm [Borghi, 1988] belegt. Weiterhin werden zur Validierung der Rechnungen experimentelle Daten benutzt und die Messergebnisse teilweise numerisch analysiert. Die Arbeit ist in verschiedene thematische Abschnitte unterteilt. In Kapitel 2 werden die zum Verständnis dieser Arbeit notwendigen physikalisch-chemischen Grundlagen vermittelt. Kapitel 3 beschreibt Aufbau und Funktion des neu entwickelten Verbrennungsmodells. Im vierten Kapitel werden drei praxisrelevante Anwendungsfälle mit unterschiedlichen Modellen untersucht. Eine Zusammenfassung in Kapitel 5 schließt die Arbeit thematisch ab. In Kapitel 6 sind Anhänge und in Kapitel 7 die Literaturstellen aufgeführt.Over the past decades global air traffic increased continuously [Airbus,2001]. In parallel with an increasing number of airplanes goes the pollution by combustion engines (jet engines and turbo prop engines). One action in order to keep the environmental damage of air traffic at the lowest level possible is to develop economic low emission engines for the upcoming generations of aircraft. Under ideal conditions hydrocarbon fuels can burn with oxygen to water and carbon dioxide. Unfortunately kerosene, the most common fuel for aircraft engines, is a mixture of several complex hydrocarbons, that produce a variety of other combustion products, when burned with air. An overview of this process is given in Figure 1 in Chapter one. In order to reduce the pollution of aircraft engines the fuel consumption and the efficiency of engines have to be improved. Basis of all possible improvements and developments is the knowledge of the processes that occur in engines. A fundamental aspect within this context is the knowledge of the physico-chemical processes in the combustion chamber. In this part of the engine (shown in Figure 2) the internal energy of the fuel is transformed in thrust. The calculation of combustion phenomena is a highly complicated task. This is especially true for practical applications as aircraft engine combustors. Harsh surrounding conditions (high temperatures, high pressure, vibration) and the interaction of different physical and chemical processes pose a challenge for both measurements and simulation. Nevertheless the prediction of combustion has a tremendous importance in several respects and justifies in-depth investigations. Improvements in combustion can lead to economic advantages as an increase of combustion efficiency results in a lower fuel consumption. This also affects the environmental aspect. In general the less fuel an engine burns the less exhaust gases and pollutants it produces. Unfortunately this is not true for all kinds of pollutants. To develop an efficient and environmental friendly combustor various aspects as e.g. air-fuel-ratio, pressure, residence times, temperature etc. have to be considered. As experimental investigations are difficult and expensive, numerical simulations are important in the development and modification of combustors. Velocity and temperature distribution as well as concentrations of major species are of particular interest for the design process. Within this thesis a new combustion model to describe the turbulence-chemistry interaction in industrial relevant combustion devices is developed. The model is based on assumed probability density functions for species mass fractions and temperature and on heavily reduced chemical mechanisms. The combustion model is a separate module. It is coupled to a commercial CFD-code (Fluent 5). Transport equations for all participating species as well as the temperature and the corresponding variances are solved. Double delta Dirac functions are chosen as the shape of the PDF. The location of the peaks is determined by the first and second moments. Respecting several statistical constraints e.g. conservation of variances, the chemical reaction is calculated for all possible combinations of species and temperature. Depending on the fuel, global one step, two step or four step reduced reaction mechanisms are used. These mechanisms are taken from literature (see Annex B). They do not represent realistic chemical reactions but empirically derived overall reactions and are consequently simplified. The Arrhenius reaction rates are weighted by the PDF and returned to the main code. The module therefore is a joint PDF (JPDF) function of species and temperature. The new combustion model uses established techniques as finite rate reduced mechanisms and combines them with a turbulence incorporating PDF approach. The model is tested and verified by comparison of numerical results with experimental results. Three different practical applications of increasing complexity are calculated by the JPDF and other combustion models. The DLR Jetflame (Chapter 4.1) is a lab-scale diffusion flame that has been carefully investigated by different laser measuring techniques. Next to numerical simulation of the flame and comparison of simulation and measurements, experimental data are examined numerically and the deviation from chemical equilibrium conditions is determined. Due to the proximity to the equilibrium and the non-existence of heavily reduced mechanisms for the used fuel this flame is calculated with an PDF-equilibrium chemistry approach. Experimental and numerical results show good agreement for both velocities and temperature distributions. The second application is the DLR Gasfilmdüse (Chapter 4.2), a confined diffusion swirl flame. As the nozzle is based on a MTU AeroEngines design, the geometry of the nozzle and operating conditions are very close to conditions in jet engines. Velocity and temperature measurements are available for comparison with numerical calculations. Simulations are performed with the newly developed JPDF combustion model as well as with an EDC and a PDF equilibrium combustion model. Results are compared with corresponding measurements. Only when using the JPDF model the distribution of temperature and velocity is described correctly. The other models that are used within this work fail to predict even fundamental characteristics of the flame. The third application that was used to test the JPDF-model is a real experimental combustion chamber segment built by MTU AeroEngines GmbH (Chapter 4.3). A photograph of the MTU-E3E combustor is shown in Figure 4.27. This combustion chamber was developed within the Engine E3E program, a national German technology initiative with the scope to design an economic, efficient and environmentally friendly engine. The E3E-combustor was experimentally investigated under real operation conditions. Measurements of velocity, temperature, and species concentration were performed on a test rig at DLR Cologne. These experimental results are compared with numerical results, obtained by using the JPDF-model and other combustion models. For the numerical simulation a three-dimensional calculation grid with about 300000 cells was used. The grid was unstructured and locally refined. Especially the regions around inflows (nozzle, cooling holes,…) was highly refined. In Figure 4.29 the surface grid of the MTU-E3E calculation is shown. Due to resolution constraints the effusion cooling in the first part of the combustion chamber is modeled as a slight constant inflow over the entire affected upper and lower surface. Boundary conditions for the calculation stem from measurements and overall pressure drop calculations [MTU,1999]. Several different combustion models are used to calculate the velocity and temperature distribution in the combustion chamber. Furthermore the air-to-fuel ratio and the fuel type are varied. The JPDF model leads to good results though the influence of the simplifications in chemistry and shape of the PDF is visible. For this application the newly developed model has no significant advantage over the equilibrium chemistry and flamelet model. The JPDF model is a self-contained combustion model that is linked to a commercial CFD-code. The application of the combustion model leads to good results in the investigated cases. Especially for the DLR Gasfilmdüse, only the JPDF model succeeds to predict the velocity and temperature distribution. Due to its modular design the model can be adapted and modified when new reduced chemistry models are available