Rigorous Study of Entrainment Using DNS and LES of Turbulent Premixed Jet Flames Gauss Centre for Supercomputing e.V.

COMPUTATIONAL AND SCIENTIFIC ENGINEERING

Rigorous Study of Entrainment Using DNS and LES of Turbulent Premixed Jet Flames

Principal Investigator:
Luis Cifuentes

Affiliation:
Chair of Fluid Dynamics, University of Duisburg-Essen

Local Project ID:
pr53fa

HPC Platform used:
SuperMUC of LRZ

Date published:

The characterization of the interactions between the flame and turbulence is essential in combustion systems. In the last decades, the wrinkling of a flame due to turbulence has been envisioned as the result of vortices interacting with the flame front. However, this preconceived notion should be substantiated through theoretical studies and numerical simulations to accurately understand the nature of the flame-turbulence interactions that are responsible for flame strain and local quenching events.

For turbulent premixed flames, such as the one studied in this research, the flame is characterized by a scalar field c, which ranges from 0 in the fresh gases to 1 in the fully reacted products (see Figure 1). The vortices interacting with the flame front are represented by the vorticity field ωi (where ωi is a vector quantity and it gives us the tendency of a fluid particle to rotate or circulate at a particular point; i.e., it gives us the information about the rotation of the fluid). On the other hand, the strength of those vortices can be illustrated by the enstrophy field E, which is used to identify the turbulent/non-turbulent interface or enstrophy interface. Fluid from the non-turbulent region becomes turbulent through the propagation of the enstrophy interface. This dynamics of the turbulent entrainment affects the flame structure, but it is still to be precisely discerned. This research aims at understanding the physics of entrainment in turbulent premixed flames and addresses the important, but actually poorly understood question of “How does entrainment of hot products work in turbulent premixed flames?”. This is an essential issue in reactive turbulent flows, because a better understanding of the dynamics of the enstrophy interface and the flame front would lead to better predictions of flame instabilities and scalar structures.

In the present project, a turbulent jet experiment has been reproduced by state-of-the-art numerical tools. The aim has been to create a database of the time-resolved evolution of the flame front and the enstrophy interface of reacting and non-reacting turbulent jets in co-flow, which can be used to analyze the entrainment process in great detail – by ourselves and by other researchers. This research required HPC resources to investigate the structure of the flame front and the enstrophy interface that have to be captured with high-resolution simulations.

Figure 1 shows the instantaneous scalar field c (left), the enstrophy field (middle) and a three-dimensional purple iso-surface c = 0.5 (right) with vortical structures (blue), for both reactive and inert cases of the turbulent jet. Analyzing the scalar field (left), it can be seen that the scalar-turbulence interaction is apparently weak upstream for both cases, with the progress variable propagating in a low turbulence environment close to the nozzle, which does not affect the inner structure of the scalar field much. On the other hand, for the enstrophy field (middle), it can be seen that the scalar-turbulence interaction becomes stronger downstream for both reactive and inert cases. It is also noticeable that for the reacting jet some flame fingers and pocket formation with high enstrophy values and fresh gases occur, whereas the thickness of the scalar interface increases and spans downstream for the inert case. The three-dimensional figure (right) shows how the vortical structures affect the scalar interface downstream, increasing the extent of scalar wrinkling.

All simulations in this project were run with the in-house LES/DNS code PsiPhi originally developed by Prof. Andreas Kempf at the Imperial College London. The computational time estimated for this project was completely used. This project required long/running high resolution simulations to capture the fine details of large and small structures, which are critical to explore the interaction of the flame with the small-scale features of the enstrophy interface and to characterize the entrainment processes. Therefore, a long flow initialization process for the development of all the structures within the computational domain was required.

These turbulent jet-flow simulations also required a long period of computing time to initialize the flow field due to the formation of slow recirculation zones in the corners of the nozzle exit. Only after reaching a statistically steady/state of the flow field and combustion, the sampling of the data was started.

Ten numerical cases (5 inert and 5 reactive) with different grid sizes were simulated on HPC system SuperMUC (80 million core hours), in order to systematically asses the effects of turbulent entrainment in reacting and non-reacting flows. Three grid resolutions were initially evaluated to get confidence in the results and to establish a grid sensitivity study. The first grid resolution was of 200μm, the inert and reactive cases consisted of 800×300×300 grid points uniformly spaced with 72 million cells. The second grid resolution was of 100μm, which increased the number of cells by a factor of two in every direction, hence, 1600×600×600 were the grid points resulting in a LES mesh of 576 million cells. The high fidelity DNS mesh consisted of 3200×1200×1200 equidistant cells and a grid of 4608 million cells with a resolution of 50μm. The database was mined and important observations were extracted.

Finally, this project generated preliminary data that will be available for future post-processing – by ourselves and by other researchers. These preliminary results can help us understand how the interaction between the scalar and enstrophy interfaces strongly influences the local flame geometry.

A comprehension of these physical mechanisms should better guide the formulation of sound and accurate mixing and combustion models. This can eventually help generating clean, save and cost-effective combustion systems, for burning fossil and sustainable fuels to provide power, heat and transportation.

As the principal investigator, Dr.-Ing. Luis Cifuentes, and as project member, Prof. Dr.-Ing. Andreas Kempf, Institute for Combustion and Gas Dynamics, Chair of Fluid Dynamics at University Duisburg-Essen, we acknowledge the grant of high-performance computing resources at HPC system SuperMUC Phase 1 with grant number pr53fa.

The project is supported through a Marie Skłodowska-Curie Action for Dr.-Ing. Luis Cifuentes (EU-Project 706672 – ITPF: www.uni-due.de/itpf/) at the Institut für Verbrennung und Gasdynamik – Fluiddynamik, Universität Duisburg-Essen.

Dissemination of results

  • Cifuentes L, Kempf A. Flow and scalar topologies in turbulent premixed flames. Biannual meeting of the German Section of the Combustion Institute ‘Deutscher Flammentag’. Bochum – Germany (17.09.2019).
  • Cifuentes L, Kempf A. Flame-turbulence interactions in turbulent premixed flames. European Researchers' Night. Kiel – Germany (28.09.2018).
  • Cifuentes L, Kempf A, Dopazo C. Mini-symposium on Flame- and Flow-Topologies (Part I and Part II). ERCOFTAC Symposium ETMM12, Montpellier – France (27.09.2018).
  • Cifuentes L, Kempf A, Dopazo C. Small-scale flow topologies in the neighborhood of enstrophy and scalar interfaces. ERCOFTAC Symposium ETMM12, Montpellier – France (27.09.2018).
  • Cifuentes L, Kempf A. DNS of turbulent premixed flames: Flow and added small-scale topologies. Sixteenth International Workshop on Premixed Turbulent Flames PTF-16, Dublin – Ireland (27.07.2018).

Project Member and Scientific Contact

Dr.-Ing. Luis Cifuentes
Chair of Fluid Dynamics
University of Duisburg-Essen
Carl-Benz-Str. 199, D-47057 Duisburg (Germany)
e-mail: luis.cifuentes [@] uni-due.de

Local project ID: pr53fa

Dember 2019

Tags: CSE Computational Fluid Dynamics University of Duisburg-Essen LRZ