Multi-injector modeling of transverse combustion instability experiments

Kevin J Shipley, Purdue University

Abstract

Concurrent simulations and experiments are used to study combustion instabilities in a multiple injector element combustion chamber. The experiments employ a linear array of seven coaxial injector elements positioned atop a rectangular chamber. Different levels of instability are driven in the combustor by varying the operating and geometry parameters of the outer driving injector elements located near the chamber end-walls. The objectives of the study are to apply a reduced three-injector model to generate a computational test bed for the evaluation of injector response to transverse instability, to apply a full seven-injector model to investigate the inter-element coupling between injectors in response to transverse instability, and to further develop this integrated approach as a key element in a predictive methodology that relies heavily on subscale test and simulation. To measure the effects of the transverse wave on a central study injector element two opposing windows are placed in the chamber to allow optical access. The chamber is extensively instrumented with high-frequency pressure transducers. High-fidelity computational fluid dynamics simulations are used to model the experiment. Specifically three-dimensional, detached eddy simulations (DES) are used. Two computational approaches are investigated. The first approach models the combustor with three center injectors and forces transverse waves in the chamber with a wall velocity function at the chamber side walls. Different levels of pressure oscillation amplitudes are possible by varying the amplitude of the forcing function. The purpose of this method is to focus on the combustion response of the study element. In the second approach, all seven injectors are modeled and self-excited combustion instability is achieved. This realistic model of the chamber allows the study of inter-element flow dynamics, e.g., how the resonant motions in the injector tubes are coupled through the transverse pressure waves in the chamber. The computational results are analyzed and compared with experiment results in the time, frequency and modal domains. Results from the three injector model show how applying different velocity forcing amplitudes change the amplitude and spatial location of heat release from the center injector. The instability amplitudes in the simulation are able to be tuned to experiments and produce similar modal combustion responses of the center injector. The reaction model applied was found to play an important role in the spatial and temporal heat release response. Only when the model was calibrated to ignition delay measurements did the heat release response reflect measurements in the experiment. While insightful the simulations are not truly predictive because the driving frequency and forcing function amplitude are input into the simulation. However, the use of this approach as a tool to investigate combustion response is demonstrated. Results from the seven injector simulations provide an insightful look at the mechanisms driving the instability in the combustor. The instability was studied over a range of pressure fluctuations, up to 70% of mean chamber pressure produced in the self-exited simulation. At low amplitudes the transverse instability was found to be supported by both flame impingement with the side wall as well as vortex shedding at the primary acoustic frequency. As instability level grew the primary supporting mechanism shifted to just vortex impingement on the side walls and the greatest growth was seen as additional vortices began impinging between injector elements at the primary acoustic frequency. This research reveals the advantages and limitations of applying these two modeling techniques to simulate multiple injector experiments. The advantage of the three injector model is a simplified geometry which results in faster model development and the ability to more rapidly study the injector response under varying velocity amplitudes. The possibly faster run time is offset though by the need to run multiple cases to calibrate the model to the experiment. The model is also limited to studying the central injector effect and lacks heat release sources from the outer injectors and additional vortex interactions as shown in the seven injector simulation. The advantage of the seven injector model is that the whole domain can be explored to provide a better understanding about influential processes but does require longer development and run time due to the extensive gridding requirement. Both simulations have proven useful in exploring transverse combustion instability and show the need to further develop subscale experiments and companions simulations in developing a full-scale combustion instability prediction capability.

Degree

M.S.A.A.

Advisors

Anderson, Purdue University.

Subject Area

Aerospace engineering

COinS