41st AIAA/ASME/SAE/ASEE Joint Propulsion Conference & Exhibit 10 - 13 July 2005, Tucson, Arizona

AIAA 2005-4428

Dynamics Combustion Characteristics in Scramjet Combustors with Transverse Fuel Injection

Jeong-Yeol Choi* Pusan National University, Pusan 609-735, Korea and Fuhua Ma, Vigor Yang The Pennsylvania State University, University Park, PA 16802, U.S.A.
Abstract A comprehensive DES quality numerical analysis has been carried out for reacting flows in constant-area and divergent scramjet combustor configurations with and without a cavity. Transverse injection of hydrogen is considered over a broad range of injection pressure. The corresponding equivalence ratio of the overall fuel/air mixture ranges from 0.167 to 0.50. The work features detailed resolution of the flow and flame dynamics in the combustor, which was not typically available in most of the previous studies. In particular, the oscillatory flow characteristics are captured at a scale sufficient to identify the underlying physical mechanisms. Much of the flow unsteadiness is related not only to the cavity, but also to the intrinsic unsteadiness in the flowfield. The interactions between the unsteady flow and flame evolution may cause a large excursion of flow oscillation. The roles of the cavity, injection pressure, and heat release in determining the flow dynamics are examined systematically. flight conditions. The fuel must be injected, mixed with air, and burned completely within such a short time span. A number of studies have been carried out worldwide and various concepts have been suggested for scramjet combustor configurations to overcome the limitations given by the short flow residence time. Among the various injection schemes, transverse fuel injection into a channel type of combustor appears to be the simplest and has been used in several engine programs, such as the Hyshot scramjet engine, an international program leaded by the University of Queensland [1]. For the enhancement of fuel/air mixing and flame-holding, a cavity is often employed. For example, the CIAM of Russia introduced cavities into its engines [2] and U.S. Air Force also employed cavities in the supersonic combustion experiments [3]. From the aspect of fluid dynamics, transverse injection of fluid into a supersonic cross flow and flow unsteadiness associated with a cavity are of significant interest topics due to their broad applications in many engineering devices. Extensive efforts have been applied to study these phenomena, and much of the results have great relevance relevant to scramjet combustors. Papamoschou and Hubbard[4] observed the fluid dynamic instability of injector flow. BenYakar et al.[5] also observed essentially the same unstable injection jet in their supersonic combustion experiment and the showed that supersonic combustion is overlaid with the large eddy motions of the unstable injection jet. The unsteady nature of transverse injector flow has been first studied numerically by von Lavante et al.[6] but its physical nature has been discussed less importantly. A comprehensive study directly applied to combustor dynamics, however, is rarely found. The obstacles lie in the difficulties in conducting highfidelity experiments and numerical simulations to

Introduction The success of future high-speed air transportation will be strongly dependent on the development of hypersonic air-breathing propulsion engines. Although there exist many fundamental issues, combustor represents one of the core technologies that dictate the development of hypersonic propulsion systems. At a hypersonic flight speed, the flow entering the combustor should be maintained supersonic to avoid the excessive heating and dissociation of air. The residence time of the air in a hypersonic engine is on the order of 1 ms for typical
*

Corresponding Author, Associate Professor, Department of Aerospace Engineering, aerochoi@pusan.ac.kr Post-doctoral Research Fellow, Department of Mechanical Engineering, mafuhua@psu.edu Distinguished Professor, Department of Mechanical Engineering, vigor@psu.edu -1American Institute of Aeronautics and Astronautics

characterize the flow transients at time and length scales sufficient to resolve the underlying mechanisms. Choi et al.[7] has studies a two-dimensional scramjet combustor configuration with transverse fuel injection and a cavity flame holder with highly refined computations. They found that RichtmyerMeshkov shear layer instability or cavity-driven instability, those are unavoidable in supersonic combustor,[8,9] triggers the injector flow instability, and the disturbed injector flow greatly enhanced the fuel/air mixing and combustion. However, stabilized combustion has not been observed and overall combustion characteristics have not been justified. It is because of the constantarea combustor configuration and ambiguously defined combustor inlet. Thus, the present study attempts to achieve improved understanding of the unsteady flow and flame dynamics in a realistic scramjet combustor configuration employing a modified combustor configuration and Reynolds averaging of the flow field. Theoretical Formulation and Numerical Treatment Governing Equations The flowfield is assumed to be two-dimensional for computational efficiency, and can be described with the conservation equations for a multi-component chemically reactive system. The coupled form of the species conservation, fluid dynamics, and turbulent transport equations can be summarized in a conservative vector form as follows.
Q F G Fv G v + + + = +W x t x y y

where, i=1~N. Numerical Methods The governing equations were treated numerically using a finite volume approach. The convective fluxes were formulated using Roe's FDS method derived for multi-species reactive flows along with the MUSCL approach utilizing a differentiable limiter function. The spatial discretization strategy satisfies the TVD conditions and features a high-resolution shock capturing capability. The discretized equations were temporally integrated using a second-order accurate fully implicit method. A Newton sub-iteration method was also used to preserve the time accuracy and solution stability. Detailed descriptions of the governing equations and numerical formulation are documented in a previous work [11]. Chemistry Model and Turbulence Closure The present analysis employs the GRI-Mech 3.0 chemical kinetics mechanism for hydrogen-air combustion [12]. The mechanism consists of eight reactive species (H, H2, O, O2, H2O, OH, H2O2 and HO2) and twenty-five reaction steps. Nitrogen is assumed as an inert gas because its oxidation process only has a minor effect on the flame evolution in a combustor. Turbulence closure is achieved by means of Mentor's SST (Shear Stress Transport) model derived from the k- two-equation formulation [13]. This model is the blending of the standard k- model that is suitable for a shear layer problem and the Wilcox k- model that is suitable for wall turbulence effect [14]. Baridna et al.[15] reported that the SST model offers good prediction for mixing layers and jet flows, and is less sensitive to initial values in numerical simulations. Code Verification The overall approach has been validated against a number of steady and unsteady flow problems including shock-induced combustion oscillation. Good agreement has been obtained with experimental data [11,16,17]. In addition, numerical study was carried out to validate the present turbulence modeling and to justify the grid resolution for simulating transverse gas injection across the supersonic flow over a flat plate. The analysis simulates the experiment described in Ref. [18] with a static pressure ratio of 10.29, for which several numerical studies have been previously carried out [19, 20]. In this case, choked nitrogen flow is vertically injected through a 1-mm-wide slot locating 33 cm behind the leading edge into a supersonic airflow with a Mach number of 3.75. The present study used the same computational domain as that of

(1) where the conservative variable vector, Q, convective flux vectors, F and G, diffusion flux vectors, Fv and Gv, and reaction source term W are defined in Eq. (2). Details of the governing equations and thermo-physical properties are described in Ref. 10.
i v i u i uv u 2 + p u v 2 + p uv v Q= F= G= Hv Hu e kv ku k v u
i u id i u id wi 0 xy xx yy xy 0 Fv = W= Gv = y x 0 k y k x sk k k y x s

(2a)

(2b)

-2American Institute of Aeronautics and Astronautics

Chenault and Beran [20]. Computations were carried out for various combinations of grid systems having 71 to 351 points near the injection port in the streamwise direction and 41 to 251 points clustered near the wall in the transverse direction. Furthermore, a parametric study was performed on the effects of numerical and turbulence modeling parameters. The numerical parameters were optimized to maintain numerical stability and solution convergence. The turbulence parameters of the SST model [13] have negligible effects on the solutions for the grid systems employed herein.
4.0
Aso et. al., Experiment(1991) Rizzetta (1992) Chenault & Beran (1998) 141x101 211x151 281x201 351x251

3.0

2.0

1.0

0.0 0.8

0.9

1.0

1.1

1.2

x/L Fig. 1 Wall pressure distribution of the twodimensional transverse injection across the supersonic flow over a flat plate.
Figure 1 compares the wall-pressure distributions between the numerical and experimental results. A coarse grid results in a longer separation distance ahead of the injection port and cannot predict the pressure picks near the injection port, although the solution seems to better match the experimental result. The 281201 and 351251 grids have nearly identical results within 5% relative error range over the entire wall. In comparison with previous results, the present turbulence model predicts the same separation distance and peak pressure as the k- model while maintaining smooth pressure increase in the front separation region. Also, pressure variation behind the injector is more closely predicted by the SST model. The 281201 grid was then applied to the scramjet combustor simulation. The minimum vertical spacing is y+ 5, and 21 grid points are employed in the injector port. Issues on Turbulence and Chemistry Present computational formulation constitutes an unsteady Reynolds averaged Navier-Stokes analysis (URANS), but showed highly refined combustion characteristics overlaid with large eddy motions of

unstable injector flow, using a massive computational grid for two-dimension.[7] These kinds of results were comparable to the quality of DES (Detached Eddy simulation), a seamless hybrid approach of LES and RANS, easily attainable by present SST turbulence model by using local grid size instead of wall distance. Another important issue is the closure problems for the interaction of turbulence and chemistry in supersonic conditions. Recently, there were many attempts to address this issue using LES methods, PDF approaches, and other combustion models extended from subsonic combustion conditions. Although much useful advances were achieved, the improvement was insignificant in comparison with the results obtained from laminar chemistry and experimental data, as discussed by Mbus et al [21]. A careful review of existing results, such as Norris and Edwards [22] suggests that the solution accuracy seems to be more dependent on grid resolution than the modeling of turbulence-chemistry interaction. In view of the lack of reliable models for turbulence-chemistry interactions, especially for supersonic flows, the effect of turbulence on chemical reaction rate is ignored in the present work. Highly refined turbulent flows were post-processed by two kinds of time averaging techniques, Reynolds average and discrete time average. Reynolds average is a continuous time average of the flow variables at a given location and its visualization result is equivalent to the image taken with long-time exposure. The discrete time average is an average of several intermediate results within a given time frame, and its image is equivalent to the averaged image of motion pictures. Presently the continuous average was taken for final 20,000 iterations of computation equivalent to final 1.2 ms physical time period, and discrete average was taken from 20 intermediate results among the same time period. Configuration of Scramjet Combustor Combustor Configuration The supersonic combustor considered in this study is shown in Fig. 1. The channel type combustor of 10 cm height and 131 cm length is composed of transverse fuel injection and a cavity. This combustor configuration is quite similar to the Hyshot test model, except for the cavity, in which a swallowing slot is employed to remove the boundary layer from the inlet and the combustor starts with a sharp nose [1]. A cavity of 20 cm length and 5cm depth, having an aspect ratio, L/D of 4.0, is employed at 20 cm downstream of the injector. In the present computation, intake region was also considered for the physical assessment of the thermal choking condition,

pw /p

-3American Institute of Aeronautics and Astronautics

slip wall Air 14cm H2 20cm 20cm 10cm 5cm 131cm no-slip supersonic exit adiabatic

x = 59cm, reference pressure probing point

Fig. 2

Scramjet combustor configuration Summary of Results Numerical simulations were carried out for 32 cases of different configurations and fuel injection pressure ratios: 1) divergent nozzle-type combustor and constant area combustor, 2) with and without cavity, 3) reacting and non-reacting flows and 4) fuel injection pressure ratios of 5.0, 7.5, 10.0 and 15.0. All the cases were run for 12 ms starting from the initial condition, which is longer than the typical test time of the ground based experiments. The plots of the instantaneous flowfields shown in the followings were taken at 12 ms, and averaged images were taken from next 1.2 ms. Figure 4 to 7 shows the instantaneous and averaged results of reacting flows for different configurations with dependency on fuel injection pressure. The images are overlaid temperature and pressure contours to account for the combustion and shock structures. Black curve is the sonic line to show the flow choking. Fig. 8 is the pressure histories of each case at reference pressure probing point at x=59 cm along the bottom wall, and Fig. 9 is its frequency spectrum obtained by FFT (fast Fourier transform) analysis. Only 4 to 12 ms time period was considered in FFT analysis to exculde the transient effects at initial phase. In the following only the overall characteristics will be discussed in this paper since the detailed feature of unsteady combustion was discussed in the previous paper. In Fig. 4, the computational results for divergent nozzle configuration without cavity, the flow is nearly undisturbed and close to steady state. It is though that the present grid resolution is not enough to capture the flow instability and eddy motions for this flow conditions. It is also considered that RANS may not sufficient to account for the supersonic fuel-air mixing. The eddy motions are captured for large injection pressure of 10.0 and 15.0. Averaged results for these cases are quite different to that of undisturbed low injection pressure cases by showing greatly enhanced fuel-air mixing and combustion. The averaged field shows that most of the combustor is maintained at supersonic condition, and there is a significant delay in active combustion. The active combustion begins roughly at 80 cm from cowl

Operating Conditions The incoming air flow to the combustor is set to Mach number 3 at 600 K and 1.0 MPa. This combustor inlet condition roughly corresponds to a flight Mach number 5-6 at an altitude of 20 km, although the exact condition depends on the inlet configuration. Gaseous hydrogen is injected vertically through a slot of 0.1 cm width to the combustor through a choked nozzle. The fuel temperature is set to 151 K. The injector exit pressures are 0.5, 0.75, 1.0 and 1.5 Mpa, and the overall equivalence ratios are from 0.167 to 0.5. Combustor Conditions A total of 800160 grids are used for the maincombustor flow passage, 5050 for inlet cowl and 159161 grids for the cavity. The grids are clustered around the injector and the solid surfaces and the injector. 54 grid points are included in the injector slot and the minimum grid size near the wall is 70 m. All the solid surfaces are assumed to be no-slip and adiabatic, except for the upper boundary. For convenience and reduction of the number of grid points required to resolve the boundary layer, the upper boundary is assumed to be a slip wall, which is equivalent to the flow symmetric condition in the present configuration. Extrapolation is used for the exit boundary. Time step is set to 6 ns according to the minimum grid size and the CFL number of 2.0. Four sub-iterations are used at each time step. Fig. 2 is a magnified plot of the computational grid around the injector, cavity and the fore part of the combustor.

H2 Fig. 2 Magnified plot of computational grid around the injector and the fore part of combustor.

-4American Institute of Aeronautics and Astronautics

(a) Instantaneous Field

(a) Instantaneous Field

(b) Averaged Field of Discrete Frames

(b) Averaged Field of Discrete Frames

(c) Continuously Averaged Field Fig. 4 Nozzle configuration without cavity

(c) Continuously Averaged Field Fig. 5 Nozzle configuration with cavity

-5American Institute of Aeronautics and Astronautics

(a) Instantaneous Field

(a) Instantaneous Field

(b) Averaged Field of Discrete Frames

(b) Averaged Field of Discrete Frames

(c) Continuously Averaged Field Fig. 6 Channel configuration without cavity Fig. 7

(c) Continuously Averaged Field Channel configuration with cavity

-6American Institute of Aeronautics and Astronautics

Fig. 8

Bottom wall pressure histories at x=59cm

Fig. 9

Frequency spectrum of bottom wall pressure at x=59cm -7American Institute of Aeronautics and Astronautics

lip for injection pressure ratio of 10.0 and 50 cm for injection pressure ratio of 15.0. Both cases showed the unsteady but stabilized combustion. However, pressure build-up for the case of 10.0 was insignificant but pressure at reference probing point was maintained at about 0.3MPa for the case of 15.0. It is considered combustion efficiency would be low for these cases, especially for the case of 10.0. In Fig. 5, flow is disturbed for all injection pressure ratios due to the presence of cavity, and pressure histories in Fig. 8 also show higher levels than the cases of same flow conditions without cavity. Thus combustion is considered to be enhanced greatly. Especially at pressure ratio 15.0, the active combustion region is anchored over the cavity where flow is choked locally. Fig. 6 and 7 show the results of the constant area channel type combustors with and without cavity. Except for the low injection pressure ratio case of 5.0, where unsteady but stabilized combustion is observed, pressure is continuously build-up due to the heat addition in constant area combustor. Although there is a certain amount of time delay, the cases of injection pressure ratio of 7.5 also showed the pressure build-up. Therefore, most of the cases go eventually to thermal choking condition, though there are time differences. For the cases of injection pressure of 10.0, detached bow shocks are shown ahead of the cowl lip, and the bow shock ran against the inflow boundary. The role of cavity, enhancing the combustion, is also observed in these results. For low injection pressure conditions, where combustion is stabilized, the pressure level is higher for the case with cavity. However, the cavity makes little differences for the higher injection pressure condition except the time difference in thermal choking. To account for the turbulence frequencies that dominate the unsteady combustion phenomena, FFT analysis is carried out and plotted in Fig. 9. For the cases of divergent nozzle type combustor, two dominant frequencies are observed at around 2 kHz and 10 kHz for the cases without cavity, and one more dominant frequency at 5 kHz for the cases with cavity. The frequency of 5 kHz is considered as the second mode of cavity oscillation, as predicted by Rossiters semi empirical formula in the previous works. For the cases of constant area combustor, very low frequencies are dominant. Major oscillation frequencies are spread over 2 to 10 kHz bad, but they are mixed-up and are not easily identified, due to its transient nature of continuous pressure build-up. Conclusion The reacting flow dynamics in a scramjet combustor

was studied by means of a comprehensive numerical analysis. The present results show a wide range of phenomena resulting from the combustor area effects, the interactions among the injector flows, shock waves, shear layers, and oscillating cavity flows. The overall results were also discussed by means of time averaging the unsteady flow fields. Further analyses are expected from these results to clarify and quantify the overall combustion characteristics.

Acknowledgements For the present study, the first author was sponsored by Agency for Defense Development and National Research Laboratory program of Korea Science and Engineering Foundation. The supports are acknowledged greatly.

References [1] Centre for Hypersonics - HyShot Scramjet Test Programme, http://www.mech.uq.edu.au/hyper/hyshot/ [2] McClinton, C., Roudakov, A., Semenov, V. and V. Kopehenov, Comparative flow path analysis and design assessment of an axisymmetric hydrogen fueled scramjet flight test engine at a Mach number of 6.5, AIAA Paper 96-4571, VA, Nov. 1996 [3] Mathur, T., Gruber, M. Jackson, K., Donbar, J., Donaldson, W., Jackson, T and Billig, F., Supersonic Combustion Experiments with a Cavity-Based Fuel Injector, Journal of Propulsion and Power, Vol.17, No.6, 2001, pp.1305-1312. [4] Papamoschou, D., and Hubbard, D.G., "Visual Observations of Supersonic Transverse Jets," Experiments in Fluids, Vol. 14, May 1993, pp. 468-471., http://supersonic.eng.uci.edu/scramjet.htm [5] Ben-Yakar, A., Kamel, M .R., Morris, C. I. and Hanson, R. K., "Experimental Investigation of H2 Transverse Jet Combustion in Hypervelocity Flows," AIAA Paper 97-3019, 1997. [6] Von Lavante, E., Zeitz, D. And Kallenberg, M., Numerical Simulation of Supersonic Airflow with Transverse Hydrogen Injection, Journal of Propulsion and Power, Vol.17 No.6, 2001, pp.1319-1326. [7] Choi, J.-Y., Yang, V. and Ma., F., "Combustion Oscillations in a Scramjet Engine Combustor with Transverse Fuel Injection," Proceedings of the Combustion Institute, Vol. 30/2, Dec. 2004, pp. 2851-2858, or AIAA Paper 2003-4515.

-8American Institute of Aeronautics and Astronautics

[8] Papamoschou, D., and Roshko, A., "The Turbulent Compressible Shear Layer: An Experimental Study," Journal of Fluid Mechanics, Vol. 197, 1988, pp. 453-477. [9] Ben-Yakar, A. and Hanson, R. K., "Cavity FlameHolders for Ignition and Flame Stabilization in Scramjets: An Overview," Journal of Propulsion and Power, Vol. 17, No. 4, 2001, pp.869-877. [10] Choi, J. Y., Jeung, I. S. and Yoon, Y., Numerical Study of SCRam-Accelerator Starting Characteristics, AIAA Journal, Vol. 36, No. 6, 1998, pp. 1029-1038. [11] Choi, J.-Y., Jeung, I.-S. and Yoon, Y., "Computational Fluid Dynamics Algorithms for Unsteady Shock-Induced Combustion, Part 1: Validation," AIAA Journal, Vol. 38, No. 7, July 2000, pp.1179-1187. [12] Smith, G. P., Golden, D. M., Frenklach, M., Moriarty, N. W., Eiteneer, B., Goldenberg, M., Bowman, C.T., Hanson, R.K., Song, S., Gardiner Jr., W.C., Lissianski, V.V., and Qin, Z., GRIMech, http://www.me.berkeley.edu/gri_mech/ [13] Menter, F. R., Two-Equation Eddy-Viscosity Turbulence Models for Engineering Application, AIAA Journal, Vol. 32, No. 8, 1994, pp.15981605. [14] Wilcox, D. C., Turbulence Modeling for CFD, DCW Industries, La Caada, CA, 1993. [15] Bardina, J. E., Huang, P. G., and Coakly, T. J., Turbulence Modeling Validation, AIAA 97-

2121, 1997. [16] Choi, J.-Y., Jeung, I.-S. and Yoon, Y., "UnsteadyState Simulation of Model Ram Accelerator in Expansion Tube," AIAA Journal, Vol. 37, No. 5, 1999, pp.537-543. [17] Choi, J.-Y., Jeung, I.-S. and Yoon, Y., "Scaling Effect of the Combustion Induced by Shock Wave/ Boundary Layer in Premixed Gas," Proceedings of the Combustion Institute, Vol. 27, 1998, pp. 21812188. [18] Aso, S., Okuyama, K., Kawai, S. and Ando, Y., Experimental Study on Mixing Phenomena in Supersonic Flows with Slot Injection, AIAA Paper 1991-0016, 1991. [19] Rizzetta, D., Numerical Simulation of Slot Injection into a Turbulent Supersonic Stream, AIAA Paper 1992-0827, 1992. [20] Chenault, C.F. and Beran, P.S., K and Reynolds Stress Turbulence Model Comparisons for TwoDimensional Injection Flows, AIAA Journal. Vol.36, No. 8, 1998, pp.1401-1412. [21] Mbus, M., Gerlinger, P. and Brggermann, Scalar and Joint scalar-Velocity-Frequency Monte Carlo PDF simulation of Supersonic Combustion, Combustion and Flame, Vol. 132, 2003, pp.3-24. [22] Norris, J. W. and Edwards, J. R., Large-Eddy Simulation of High-Speed Turbulent Diffusion Flames with Detailed Chemistry, AIAA Paper 970370, 1997.

-9American Institute of Aeronautics and Astronautics