Simulating Acoustic-Flame Interactions in Rocket Engines Gauss Centre for Supercomputing e.V.


Simulating Acoustic-Flame Interactions in Rocket Engines

Principal Investigator:
Klaus Hannemann

Institute of Aerodynamics and Flow Technology, Spacecraft Department. German Aerospace Center (DLR), Göttingen

Local Project ID:

HPC Platform used:
SuperMUC and SuperMUC-NG of LRZ

Date published:


Combustion instabilities are strong acoustic disturbances that often form spontaneously inside rocket combustion chambers and may lead to catastrophic failure of the launch vehicle. Although all physical processes that may contribute to the onset of such instabilities are well understood, the exact mechanisms of how these processes interact are still largely unknown. It is, however, widely accepted that the interaction between acoustic waves and the flame inside the rocket engine is very important to the formation of combustion instabilities. Even though it is very difficult to investigate the flow field in rocket combustion chambers experimentally due to the harsh flow environment, DLR researchers successfully designed a subscale combustion chamber for specifically studying the interaction between acoustics and the flame.

Combustion chamber H (BKH, see Fig. 1) consists of five primary shear coaxial injectors that are placed in the middle of the combustion chamber through which oxygen and hydrogen are fed into the reaction zone. Besides its large main nozzle in axial direction, BKH contains a secondary nozzle perpendicular to the main flow direction which can be opened and closed periodically by a siren exciter wheel. By controlling the rotation frequency of the exciter wheel, different acoustic oscillations, so called eigenmodes, can be excited inside the combustion chamber. BKH is also equipped with two windows on its side walls allowing to observe the flame dynamics using high-speed cameras and optical measurement techniques.

Even though BKH experiments provided a lot of insight into the nature of acoustic-flame interactions, many other aspects can only be observed by detailed numerical simulations. The purpose of this SuperMUC project is therefore to numerically reproduce BKH experiments for a an operating condition that is vulnerable to combustion instabilities in real flight engines.

Recent experiments on a different subscale rocket engine [2] showed that combustion instabilities occur when an eigenmode of the oxygen injector is in resonance with the main oscillation mode of the combustion chamber, i.e. when both modes have the same frequency. In this project we investigate for the first time such a mode coupling scenario by carefully controlling the eigenfrequency of the oxygen injector and tuning it to the combustion chamber eigenfrequency.

Results and Method

The flow field inside combustion chamber H is simulated using the DLR TAU code. In order to solve the conservation equations for mass, momentum, energy and additional turbulence quantities, the combustor volume is discretized into small nonoverlapping control volumes (so called finite volumes) in which the conserved quantities are calculated. By applying a large number of small time steps consecutively we can accurately simulate the evolution of the complex flow field inside the combustion chamber.

The simulation of rocket engines also requires sophisticated models for the thermodynamics of the propellants and the chemical reactions involved [3]. Because of the high chamber pressure in BKH (60.3 bar), oxygen is a so called supercritical fluid, meaning that its density is similar to liquid water but it behaves otherwise like a densified gas. BKH is simulated using the scale-resolving Detached Eddy Simulation (DES) model which accurately captures small turbulent features of the flow but is computationally very expensive.

The DLR TAU code can be applied efficiently on supercomputing architectures like SuperMUC. Typical BKH simulations are running on 2,520 CPUs in parallel. The project used 13.7 Mio. Core-h in total for the simulation of three different operating conditions of BKH and smaller preparatory studies.

This project provides the first scale-resolving simulation results for combustion chamber H operating with both cryogenic oxygen and hydrogen. Fig. 2 shows a typical flow field highlighting the dense oxygen cores (gray structures) which are surrounded by the hot flame. This flow field shows the undisturbed flame shape when no acoustical excitation is present.

The simulated combustion chamber pressure of 61.3 bar agrees well with the experimental value. In order to determine the combustion chamber eigenmodes, which is the first step to enforce acoustic mode coupling in simulations, an artificial pressure pulse is placed inside the combustion chamber. This pulse excites all chamber eigenmodes whose shapes, frequencies and decay rates can be determined using decomposition techniques, e.g. Dynamic Mode Decomposition (DMD). The resulting mode shapes are shown in the mp4-video below. They consist of regions with higher pressure (red) and lower pressure (blue) oscillating at a single frequency.

Chamber modes consist of longitudinal (L) and transversal (T) modes and combinations of both. The most important chamber eigenmode in terms of combustion instabilities is the 1T mode. Its mode frequency also agrees very well with the experimental value. The 1T mode is most easily excited by the siren and forces the flame to oscillate in the direction of the siren. During this motion, the dense oxygen cores are shortened significantly and they flatten in a plane perpendicular to the excitation direction.

Once the chamber 1T mode frequency is known, we can tune the oxygen injector to have its eigenfrequency matching the chamber 1T mode frequency. We then compare results from two simulations where the injector mode can couple to the chamber mode, and one without mode coupling. In a preparatory study we used computationally cheaper unsteady Reynoldsaveraged Navier-Stokes (URANS) simulations to investigate the coupling scenario. Even though we successfully achieved the desired mode coupling, no flame response was observed in the simpler URANS setup.

This result suggests that the simpler simulations possibly neglect a key feature for the development of instabilities, which we believe is the vortex shedding at the main injectors. This mechanism is resolved in Detached-Eddy simulations that are currently running in the last phase of this project. We expect these highly resolving simulations to clearly show if the vortex shedding plays a crucial role or if we have to focus on other possible mechanisms that might be related to the development of combustion instabilities.

On-going Research / Outlook

The scale-resolving simulations in this project would not have been possible without the resources from SuperMUC-NG even though we still had to limit the simulations to half of the complete combustion chamber. During the project we realized that postprocessing of large amounts of time-resolved data requires a suitable framework that can process larger-than-memory datasets in reasonable time. In the future we plan on extending our simulation activities to more realistic combustion chambers and future fuels, e.g. methane, that are currently of great interest for the aerospace industry.

References and Links

[1] S. Beinke. Dissertation. School of Mechanical Engineering. University of Adelaide, Australia, 2017.
[2] S. Gröning, J. Hardi, D. Suslov and M. Oschwald. Journal of Propulsion and Power, (0):560-573, 2016.
[3] T. Horchler, S. Fechter, S. Karl, K. Hannemann. 8th European Conference for Aeronautics and Aerospace Sciences (EUCASS), Madrid, 2019.

Research Team

Klaus Hannemann (PI), Tim Horchler, Stefan Fechter
Institute of Aerodynamics and Flow Technology, Spacecraft Department. German Aerospace Center (DLR), Göttingen

Scientific Contact

Dipl.-Phys. Tim Horchler
Deutsches Zentrum für Luft- und Raumfahrt e.V., DLR in der Helmholtz-Gemeinschaft
Institut für Aerodynamik und Strömungstechnik
Abteilung Raumfahrzeuge
Bunsenstrasse 10, D-37073 Göttingen (Germany)
e-Mail: tim.horchler [@]

Local project ID: pr27ji

May 2021