JOURNAL OF PROPULSION AND POWER

Vol. 26, No. 2, March–April 2010

Effect of Swirl on Shock Structure in Underexpanded

Supersonic Airflow

A. Abdelhafez∗ and A. K. Gupta†

University of Maryland, College Park, Maryland 20742
DOI: 10.2514/1.45716
This study examines the effect of imparting swirl to underexpanded supersonic-nozzle airflow on shock struc-
ture under matched mass flow conditions. A convergent nozzle with swirling capabilities is used to generate the
underexpanded airflow. Fuel is injected coaxially at the nozzle throat. Nonreacting conditions are considered,
wherein fuel is simulated by mixtures of helium, argon, and krypton inert gases. It was found that the effects of swirl
and nozzle reservoir pressure interfere destructively from the point of view of shock-structure axial compactness.
Increasing reservoir pressure stretches the shock structure axially, whereas swirl shrinks it. On the other hand,
Downloaded by UNIVERSITY OF MARYLAND on September 7, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.45716

constructive interference was observed from the point of view of radial jet expansion; both result in greater jet
diameter. The application of swirl was found to weaken the shock structure at matched reservoir pressure but to
strengthen it at matched mass flow. It was also found that fuel injected at low subsonic Mach numbers into the
supersonic airflow has to propagate initially with a negative shear angle; i.e., the cross-sectional area of fuel-rich core
flow converges first, before this core flow reaches a throat after which it propagates supersonically. This behavior was
found to be advantageous, as it results in reduced shock-structure strength.

Nomenclature The structure of highly underexpanded nozzle jets was investi-

a = speed of sound gated experimentally and analytically by Adamson and Nicholls [2],
D = nozzle exit diameter (11 mm) who presented a method for calculating the position of the first Mach
DR = air/fuel density ratio at injection plane disk. In their calculation, the axial pressure distribution on the flow
Mc = convective Mach number centerline downstream of the nozzle exit (calculated by method
Mrel = air/fuel relative Mach number at injection plane of characteristics) was used to define a fictitious nozzle extension.
p = pressure The Mach disk was then assumed to exist at the point at which
S = swirl number atmospheric pressure would be attained upstream of the disk; i.e., the
v = velocity disk was assumed to exist at the end of a fictitious nozzle extension.
z = axial coordinate Physical arguments were employed to extend the analysis to nozzles

= ratio of specific heats with supersonic exit Mach numbers. An approximate method for
computing the jet boundary, up to the point of maximum jet area, was
Subscripts also given. The analytical results compared favorably with experi-
mental data at relatively low nozzle pressure ratios.
a = axial Lewis and Carlson [3] experimentally determined the distance
t = tangential from the nozzle exit plane to the first Mach disc in gas-only and gas–
particle jets issuing from underexpanded supersonic nozzles. An
empirical correlation of the data was presented that is valid for both
I. Introduction jet types and incorporates the effect of gas specific-heat ratio. In a
WIRLING flow in nozzles occurs in a number of important
S propulsion applications, including turbofans and turbojet
engines, spin-stabilized rockets, and integral rocket/ramjets. In the
relevant experimental investigation, Crist et al. [4] studied the
structure of underexpanded jets with stagnation pressures up to
15,000 psia, ambient pressures down to 100 Hg, and stagnation
first two cases, the tangential velocity component is induced by the temperatures up to 4200 K. The location of the first Mach disk was
motion of turbine blades and by the rocket spin, respectively. For found to be insensitive to the ratio of specific heats, nozzle-lip
ramjets, experimental studies [1] have demonstrated that swirl geometry, and absolute pressure. For overall pressure ratios up to
generated by fixed vanes located in the dump combustor inlet can about 3
105 (i.e., ratio of reservoir to freestream static pressures),
lead to significantly improved combustor performance. Clearly, the the location of Mach disk was found to vary as the square root of
generated swirl in each of those propulsion systems will persist at overall pressure ratio. The diameters of the Mach disk, jet boundary,
some level to the inlet of the exhaust nozzle. Therefore, it is important and intercepting shock were found to increase with decrease in
to examine the effect of the tangential velocity component on nozzle specific-heat ratio and to decrease at high stagnation density, where
flowfield, so that design parameters such as thrust and mass flow rate intermolecular forces become important. At high pressure ratios, the
can be accurately determined. ratio of Mach-disk diameter to Mach-disk position appeared to be
constant for a given gas. It was also found that the properties along the
Received 30 May 2009; revision received 2 November 2009; accepted for jet axis can be approximated by the properties of a flow through a
publication 10 November 2009. Copyright © 2010 by the authors. All rights hypothetical conical nozzle for which the half-angle is given as a
reserved. Published by the American Institute of Aeronautics and function of specific-heat ratio. A simplified expression for the distri-
Astronautics, Inc., with permission. Copies of this paper may be made for bution of Mach number along the jet axis was given to good
personal or internal use, on condition that the copier pay the $10.00 per-copy approximation as a function of specific-heat ratio.
fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers,
Gostintsev et al. [5] studied an underexpanded supersonic swirling
MA 01923; include the code 0748-4658/10 and $10.00 in correspondence
with the CCC.
gas jet issuing from a convergent nozzle. They showed that the
∗
Graduate Student, Department of Mechanical Engineering, 2181 Glenn effect of rotation on the wave structure of an axisymmetric jet is
Martin Hall. Student Member AIAA. qualitatively analogous to the effect of reduction in overpressure
†
Distinguished University Professor, Department of Mechanical ratio. Using formulas for spiral isentropic flow, an approximate
Engineering, 2181 Glenn Martin Hall. Fellow AIAA. expression was obtained for estimating the location of the first Mach
215
 216 ABDELHAFEZ AND GUPTA

disc in the swirling flow downstream of the nozzle exit. Batson and The flow of supersonic swirling jets in a stagnant atmosphere was
Sforzini [6] also studied the structure of swirling flow through further investigated by Cutler et al. [20], Cutler and Levey [21], and
a convergent nozzle, with emphasis on the effect of swirl on the Levey [22]. The swirling jets were created by tangential injection into
flowfield, thrust, and mass flow produced by nozzled devices, such as a swirl chamber and accelerated through a convergent–divergent
jet engines and spin-stabilized rockets. It was reported that the axial nozzle. The researchers observed higher peak helix angles than
velocity component increases, whereas the tangential one decreases, previous studies, as well as lower densities and pressures along the jet
as the flow passes through the nozzle throat. axis. They found that the growth rates of mixing layer increased
Vortex enhancement of supersonic mixing was studied experi- considerably with swirl. Moreover, when the swirling jets were
mentally by Settles [7]. Swirl was used to enhance shear-layer growth operated overexpanded, unstable shock interactions produced vortex
and mixing. It was concluded that swirl enhances compressible breakdown.
mixing; the degree of enhancement increases with increasing swirl. This work provides an experimental/numerical investigation, in
Settles also reported that the effects of convective Mach number and which the effect of imparting swirl to underexpanded supersonic-
density ratio on the enhancement effect of swirl are still unknown and nozzle airflow on shock structure is examined. Matched mass
were thus recommended for future work. flow conditions are considered here. The effect of swirl has not
In an experimental investigation, which is very pertinent to this been fully quantified in the literature yet, due to the inherent three-
current study, Lee et al. [8] examined the near-field flow structure dimensionality of the problem. A convergent nozzle with swirling
of underexpanded coaxial swirl jets. Swirl streams were issued from capabilities is used to generate the underexpanded airflow. Fuel is
a secondary annular nozzle, and a primary inner nozzle provided injected coaxially at the nozzle throat. Nonreacting conditions are
Downloaded by UNIVERSITY OF MARYLAND on September 7, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.45716

the underexpanded free jets. The interactions between the annular considered, wherein fuel is simulated by mixtures of helium, argon,
swirl and the underexpanded core jets were examined to quantify the and krypton inert gases. Analyses are made of the effects of relative
effects of the former on the latter. It was shown that the presence of an Mach number and density ratio across the air/fuel shear layer. The
annular swirl stream causes the core-flow Mach disc to move further effects of these parameters on shock structure are investigated under
downstream, with an increased diameter. In another pertinent study, both nonswirling and swirling conditions. Select cases have been
Lee et al. [9] investigated the effect of nozzle inlet configuration on chosen for a further analysis of the effect of convective Mach number
underexpanded swirling jets, which were generated by a convergent on shear-layer growth.
nozzle with four tangential inlets at the supply chamber. The nozzle
inlet configuration was modified by using different plugs, holes, and
needles, which were also used for measuring the flow properties at II. Experimental Setup
the nozzle inlet. The experimental results showed that the presence The experimental investigation of this present work has been
of a coaxial needle inside the nozzle supply chamber controls the performed at the University of Maryland supersonic facility. The
properties of generated underexpanded swirling jets. The structures used supersonic-nozzle assembly is shown schematically in Fig. 1. A
of these jet flows are highly dependent on the detailed configuration convergent nozzle with an effective inlet-to-exit area ratio of 25 is
of the nozzle supply chamber. used to generate an underexpanded supersonic airflow. Reservoir
Murakami and Papamoschou [10] examined the flow structure pressures of up to about 9 atm (abs) are available, yielding near-field
and mixing enhancement in 2-D and axisymmetric supersonic jets Mach numbers of up to 2.2 under nonswirling conditions. The nozzle
surrounded by secondary annular subsonic coaxial jets. The super- has swirling capabilities, wherein the axial-tangential-entry tech-
sonic jets were issued from a convergent–divergent nozzle operated at nique with four tangential inlets is used to accurately control the
offdesign (i.e., underexpanded or overexpanded) conditions. It was degree of swirl imparted to airflow. This technique has been proven in
shown that the mixing enhancement using secondary parallel injec- previous research to be an efficient method for generating supersonic
tion (referred to as MESPI by the researchers) halves the length of the swirling jets [23–25]. Thermal flow meters/controllers are used to
potential core in both round and 2-D jets. A short distance past the meter the flow rates of axial and tangential air components with an
potential core, mixing enhancement caused a reduction in center- accuracy of 1:5% of full scale.
line Mach number by 30% in round jets and 20% in 2-D jets. The The nozzle was machined out of a single aluminum rod. Aluminum
corresponding reduction in peak molar concentration of a scalar was preferred to stainless steel because the former has higher thermal
injected in the primary flow was 65% in round jets and around 40% conductivity, which prevents overheating of the nozzle walls during
in 2-D jets. combustion experiments. The higher conductivity allows radiated
Increased molecular mixing between fuel and oxidizer is essential heat to be dissipated effectively through the thick nozzle walls. The
for efficient combustion, which can be accomplished by increasing dissipated heat is removed by forced convection of the entrained
the turbulent mixing. It is now well known that large-scale coherent ambient cold air through the large surface area of nozzle external
structures play an important role in incompressible turbulent mixing walls. The thickness of the nozzle lip was optimized to allow for
layers [11]. These structures engulf surrounding unmixed fluid and significant entrainment of ambient air while having adequate rigidity
carry it into the mixing layer. Furthermore, it is possible that two for machining the exit section to the desired surface finish and
adjacent vortical structures roll up upon one another, creating one dimensional tolerance.
larger structure, which leads to spreading of the mixing layer [12]. A coaxial fuel-injection system is used to inject fuel along the axis
These two processes are believed to be fundamental for turbulent of the air nozzle. A support flange upstream of the nozzle ensures
mixing in incompressible mixing layers. and maintains concentricity of the fuel injection system with respect
Studies of compressible shear layers have shown that large-scale to the air nozzle, especially under swirling conditions. This flange
structures are similarly important in compressible turbulent comprises a conical sleeve that embraces the injection system. The
mixing, changing its nature with convective Mach number [13– sleeve wall thickness decreases in the direction of flow to provide
16]. It was found that the visual growth rate of a compressible streamlined performance and prevent any blockage close to the
shear layer with almost unity convective Mach number was only nozzle exit. The sleeve is held in place by three spokes extending to
one-fourth of the growth rate of an incompressible mixing layer the support flange. Their thicknesses have been optimized to provide
having the same freestream velocity and density ratios, regardless rigidity with minimum blockage to the incoming axial component
of the values of these ratios [13]. Moreover, it is known that of airflow. It should be noted here that those spokes are located
supersonic shear layers are highly stable [17], limiting the desired physically upstream of the air tangential inlets and do not affect the
mixing of fuel and oxidizer. It was predicted that a supersonic flowfield of the tangential air component. Some wakes are expected to
shear layer will be completely stable if the Mach number based on exist in the axial-component flowfield behind the spokes, but the
the relative speed exceeds 2.83, assuming an infinitely thin vortex supersonic flow exiting the nozzle was found to be fully axisymmetric
sheet. This predicted stable nature of supersonic mixing layers has with and without swirl.
been later observed both in experiments [13,18] and in numerical The nanosecond schlieren diagnostic technique was used to
simulations [19]. visualize the shock structure in this study. schlieren imaging yields
 ABDELHAFEZ AND GUPTA 217

The light source is a 532-nm Q-switched Nd:YAG laser with pulse

duration of only 5 ns, which explains the term nanosecond schlieren
and makes this setup unique compared with conventional schlieren
setups. The nanosecond light duration allows for capturing instan-
taneous images of the flowfield. No fluctuations are thus accu-
mulated or averaged on the image, which allows for accurate
visualization of the shock structure.
Because of the fact that the intensity of laser light is too high for
safe camera operation, the laser was equipped with neutral-density
Downloaded by UNIVERSITY OF MARYLAND on September 7, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.45716

Fig. 1 Schematic of the University of Maryland’s supersonic-nozzle

assembly.

the first derivative of the refractive index in the test region, thus
giving an estimate of the density gradients within the flowfield. The
intensity of captured light can be further processed using image-
processing techniques to obtain more quantitative information of
the shock structure. Figure 2 shows a schematic of the used setup.

Fig. 3 Illustrations of a) three-dimensional schematic of the numeri-

cally simulated geometry and b) zoom-in cutaway highlighting the nozzle
Fig. 2 Schematic of nanosecond schlieren system. details.
 218 ABDELHAFEZ AND GUPTA

that the side boundaries are far enough from the jet to eliminate any
interference of both and to maintain constant near-stagnation
atmospheric properties at the boundaries. Consequently, the bottom
and side enclosure surfaces were assigned the fixed-pressure
boundary condition, which matches the constant actual atmospheric
ambient pressure. The top side of the enclosure, on the other hand, is
an extrapolated outlet located 55D away from the nozzle exit (78%
of the 70D enclosure length). This guarantees that the flow leaves
the simulated geometry shock-wave-free, since it was observed
experimentally that complete transition to subsonic speeds occurs
about 30D downstream of the nozzle exit.
Axisymmetry was enforced; i.e., only one quadrant of the
geometry depicted in Fig. 3 was simulated. Special emphasis was
placed on the level of cell skewness. The simulated geometry was
subdivided into individual volumes, each meshed separately, to keep
the skewness level of the most skewed cell below 0.5 (see Fig. 4a). A
variable-size grid was generated with tetrahedral cells and a grid
spacing ranging from 0.01 to 2.0 mm (see Fig. 4b). Grid spacing is
Downloaded by UNIVERSITY OF MARYLAND on September 7, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.45716

defined here as the longest edge of the cell. Tighter meshing was
implemented near and at the critical geometry locations: e.g., the
exits of the nozzle and fuel-injection system. Mesh dependence was
carefully examined through testing multiple levels of mesh tightness
(see Fig. 4c). A total of 7,166,860 nodes per quadrant yielded the
desired accuracy. Higher tightness levels did not result in significant
accuracy enhancement and were thus not considered, to optimize the
computational time.
Four subgrid RANS turbulence models from the FASTRAN
library were tested for their capabilities to accurately predict a free
swirling supersonic jet: namely, the k-", k-!, Spalart–Almaras, and
Baldwin–Lomax [26] models. The parameters of each model were
optimized to yield the least rms error, when compared with experi-

3.0
Centerline static pressure (p/patm)

k-omega

2.5 k-epsilon
Spalart-Almaras

2.0 Baldwin-Lomax
Experimental
1.5
Fig. 4 Computational geometry, grid, and numerical validation:
a) individual subvolumes of the simulated geometry, b) grid spacing 1.0
within the center plane, and c) mesh dependence and validation of
numerical code (case 0s1).
0.5

filters that reduce its light intensity. The divergence of the collimated 0.0
beam is then increased by means of a plano-concave lens. The 0.5 0.6 0.7 0.8 0.9
divergent beam fully illuminates a concave mirror, which reflects the Axial distance downstream of nozzle exit (z/D)
light in a collimated fashion through the test section. This is essential a)
for avoiding a skewed perspective of the flowfield. After penetrating
the flow, the light is focused by another concave mirror. A knife-edge
aperture intercepts the light at the focal point of the second mirror to
fulfill the schlieren principles. The resulting images are then captured Location of
at a resolution of 1024
1024 pixels using a high-speed camera that static-pressure
is synchronized with the laser.
measurement

III. Numerical Simulation Type and Assumptions

The commercial CFD-FASTRAN‡ 2008 hybrid large eddy
simulation (LES) and Reynolds-averaged Navier–Stokes (RANS)
code, provided by ESI-Group, was used for all the simulations
conducted in this study. Since a free supersonic flow is involved,
special emphasis was placed on the choice of boundary conditions
that represent the flow surroundings. The entire nozzle assembly was
surrounded by a cylindrical enclosure of 40D diameter and 70D
length, where D is the nozzle exit diameter [11 mm, a good repre- b)
sentation of jet size (see Fig. 3)]. The 40D enclosure diameter ensures Fig. 5 Choice of turbulence model and location of static pressure
measurement: a) choice of RANS turbulence model (case 0s1) and
‡
Data available online at http://www.esi-group.com/products/Fluid- b) schematic presentation of how the centerline static pressure was
Dynamics/cfd-fastran [retrieved May 2009]. measured experimentally for the sake of code validation.
 ABDELHAFEZ AND GUPTA 219

mental data within the near-field supersonic flowfield up to 5D The initial conditions of simulation were set for all cases at 1 atm
downstream of the nozzle exit. Figure 5 compares the obtained static pressure, 300 K static temperature, 9:7 m=s axial velocity,
centerline static-pressure trace of each model to the corresponding and zero radial and tangential velocities. Consequently, the simu-
experimental data. It can be clearly seen that the Baldwin–Lomax lation incorporated the transient behavior as the high-pressure air
model offers the best agreement with experimental data among the expands and marches from the geometry inlet to the exit. An initial
available models within the FASTRAN code. Although several Courant–Friedrichs–Lewy number [28] of 0.1 was chosen that
modifications of the Baldwin–Lomax model have been published, in increases to unity as convergence is approached. Time integration
an attempt to enhance its prediction capabilities, it should be noted is implicit; a point Jacobi scheme was used, and a backward
here that FASTRAN is not an open-source code and none of those Euler discretization was implemented. Each case included 20,000
modifications are available. Nevertheless, the obtained degree of iterations. Convergence to 106 residuals was usually attained after
accuracy with the basic Baldwin–Lomax model was considered 18,500–19,500 iterations.
acceptable for the scope of this study.
Calculation of the viscosity and conductivity was based on the
kinetic theory of gases. A turbulent Schmidt number of 0.9 was used,
and the mass diffusivity was calculated based on Fick’s law with a IV. Test Matrix
Schmidt number of 0.5. A turbulent Prandtl number of 0.9 was used The effect of swirl is investigated here by forwarding the entire
for calculating the turbulent conductivity. Similar to the experimental airflow to nozzle tangential entries. This allows for examining a
conditions, the total temperature at the air inlets was kept fixed at single degree of swirl: namely, the maximum attainable one.
Downloaded by UNIVERSITY OF MARYLAND on September 7, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.45716

300 K, and the total pressure was maintained at 7.91 bar for the Following a definition used for incompressible swirling jets [23,29],
nonswirling cases and 8.82 bar for the swirling cases. The 8.82 bar a nozzle-based geometrical swirl number Sg is defined for air as
value was carefully chosen to ensure a common air flow rate of
175 g=s. The need for higher nozzle reservoir pressure with swirl

ro Ro mt
was explained in detail in the work of Abdelhafez and Gupta [27]. Sg  (1)
The total pressure and temperature at the air inlets were preserved At ma  mt
throughout the iteration process in each examined case until con-
vergence was attained. Owing to the relatively large cross-sectional where ro Ro =At   0:68 for the geometry of used nozzle and its
areas of the air inlets, the entrance velocity of air was only 9:7 m=s, tangential entries, and ma and mt are the axial and tangential com-
resulting in almost identical inlet stagnation and static conditions. ponents of airflow, respectively. Consequently, all swirling cases of
The nozzle walls were set to be isothermal at 280 K, based on this study have the same nozzle-based geometrical swirl number of
multiple temperature measurements of the nozzle interior and 0.68. The term nozzle-based refers to nozzle operation in the absence
exterior walls. This is attributed to the aforementioned fact that the of coaxial injection system, as the presence of this system reduces the
nozzle is made of aluminum, which has a high thermal conductivity geometrical swirl number down to 0.36 [30]. This significant
and thus allows the nozzle to act as a near-isothermal body. The reduction made the examination of the effect of swirl feasible only
walls of the fuel-injection system, on the other hand, were set to be at its maximum attainable degree.
adiabatic, because the injection system is immersed almost totally In addition to an extensive examination of the effect of swirl on the
into the nozzle and conditioning chamber, which allows for neg- shock structure, the effects of two flow parameters are investigated
ligible amounts of heat to be conducted axially upstream through the here under both nonswirling and swirling conditions: namely, the
thin walls of fuel system. Moreover, it is made of stainless steel that relative Mach number Mrel and air/fuel density ratio DR. The former
has a much lower thermal conductivity (relative to aluminum). is defined here as

Table 1 Test matrixa

Caseb Injected gas Mrel DR Experimental Numerical
0, 0s1 c
—— —— —— —— ——
Effect of relative Mach number
1, 1s Helium 0.44 35.50 ✓ ✓
2, 2s Helium 0.41 35.50 ✓ ——
3, 3s Helium 0.39 35.50 ✓ ——
4, 4s Helium 0.37 35.50 ✓ ——
5, 5s Helium 0.35 35.50 ✓ ✓
6, 6s Helium 0.32 35.50 ✓ ——
7, 7s Helium 0.30 35.50 ✓ ——
8, 8s Helium 0.28 35.50 ✓ ——
9, 9s Helium 0.26 35.50 ✓ ✓
10, 10s1 Helium 0.21 35.50 —— ✓
11, 11s Helium 0.00 35.50 —— ✓
12, 12s Helium 0:21 35.50 —— ✓
13, 13s Helium 0:48 35.50 —— ✓
Effect of density ratio
14, 14s 100% helium 0.21 35.50 —— ✓
15, 15s 80% helium/20% argon 0.21 12.68 ✓ ✓
16, 16s 70% helium/30% argon 0.21 9.60 ✓ ——
17, 17s 60% helium/40% argon 0.21 7.72 ✓ ——
18, 18s 50% helium/50% argon 0.21 6.46 ✓ ✓
19, 19s 40% helium/60% argon 0.21 5.55 ✓ ——
20, 20s1 30% helium/70% argon 0.21 4.86 ✓ ——
21, 21s 20% helium/80% argon 0.21 4.33 ✓ ✓
22, 22s 50% helium/50% krypton 0.21 3.24 —— ✓
23, 23s 50% argon/50% krypton 0.21 2.29 —— ✓
a
Constant parameters are air total temperature of 300 K at the inlet and nozzle reservoir pressure of 7.91 bar (nonswirling) and
8.82 bar (swirling).
b
Geometrical swirl number for swirling cases is Sg  0:36.
c
Air only; no fuel injection.
 220 ABDELHAFEZ AND GUPTA

vair  vfuel numerically. The same applies for case pairs 15, 18, and 21 within the
Mrel  (2)
0:5aair  afuel  DR analysis.
3) The numerical simulations extend beyond the experimental
ranges to broaden the scope of the analysis and qualitatively examine
This definition relates the difference in freestream velocities how accurate the experimental trends would be when extrapolated
between fuel and air to the average speed of sound. It should be noted beyond the ranges in which they were obtained.
that the fuel simulant is injected here at velocities smaller than the Note that case pairs 10 and 14 (Table 1) are identical, as they have
sonic velocity of airflow in most examined cases. Therefore, vfuel is the same fuel simulant (helium), Mrel , and DR. These two case pairs
subtracted from vair in the above definition, in order for Mrel to link the analyses of Mrel and DR. Also note that a letter s next to a case
have positive values. Nevertheless, in a few extreme cases, the fuel number denotes a swirling case.
simulant is injected at velocities greater than those of the airflow,
and the corresponding values of Mrel are indicated here without
neglecting their negative signs, to highlight the unique nature of
V. Results and Discussion
those cases: i.e., vfuel > vair . A. Shock Structure (Nonswirling, No Fuel Injection)
Each examined case in this study has a nominal value of Mrel that The shock structure of a simple underexpanded supersonic flow is
describes the injection conditions within that case. Since all cases to shown schematically in Fig. 6. As can be seen, the structure
be presented here use no recess (i.e., fuel is injected at the throat of the comprises a shock-cell unit that gets repeated periodically to form a
air nozzle), the nominal value of Mrel was calculated for each case shock-cell train. This unit can be described as follows. Axial under-
expanded flow undergoes an expansion fan and turns outward. The
Downloaded by UNIVERSITY OF MARYLAND on September 7, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.45716

using the fuel-injection velocity and the sonic (throat) value of vair . A
comparison based on nominal Mrel thus allows for examining the free-jet boundary adapts accordingly and turns outward as well.
effect of fuel-injection conditions on shock structure. Such analysis Passing again through the expansion fan, the outward flow turns back
should not be confused with the shear-layer-specific analysis of the to axial. As the expansion fan meets the boundary, it reflects into a
effect of convective Mach number Mc on shear-layer properties (to be compression fan that coalesces later into the intercepting shock
presented in this study as well). Under nonswirling conditions, sonic wave. The annular flow adjacent to boundary turns inward through
vair was found to be 323 m=s, based on isentropic ideal-gas relations. the compression fan, and the boundary again adapts by turning
No similar simple calculations of vair could be carried out for the inward as well. For slightly underexpanded nozzles, this intercepting
swirling cases, due to the intrinsic three-dimensionality of swirling shock reflects directly into a reflected shock at the centerline, forming
flows. Nevertheless, the results of numerical simulations revealed the familiar diamond configuration. However, as the pressure ratio
that the magnitude of sonic vair is 329 m=s with swirl, which is across the nozzle is increased, this reflection no longer takes place
almost equal to the nonswirling value. This fact allowed for at the centerline, and a Mach disk is formed. The reflected shock
examining the same nominal values of Mrel under both nonswirling turns the inward annular flow back to the axial direction. Since the
and swirling conditions in this study. Mach disk maintains the axial direction of core flow, the entire flow
The shock structure and all properties of the airflow, including the is now axial again. As the reflected shock impinges on the flow free
aforementioned values of vair , depend on air total pressure and
temperature. Both were kept constant at 7.91 bar and 300 K,
respectively, for all nonswirling cases presented in this study, which
resulted in a fixed air flow rate of 175 g=s. It was noticed, however,
that imparting swirl to the airflow at the same nozzle reservoir
pressure of 7.91 bar results in a reduced mass flow rate through the
nozzle. This observation agrees with the findings of many previous
studies (refer to [27] for an extensive review). It was proven that
imparting swirl to the airflow results in additional choking of the
nozzle: i.e., a lower mass flow rate compared with the corresponding
nonswirling conditions at the same reservoir pressure. A theoretical
limit of no flow was even predicted at an infinitely large swirl number.
Fig. 6 Schematic of shock structure of nonswirling highly under-
Therefore, a higher reservoir pressure is necessary to maintain the expanded nozzle flow [2].
same flow rate through the nozzle. It was found in this study that a
value of 8.82 bar yields identical air flow rates of 175 g=s in the
nonswirling and swirling cases.
Table 1 lists the test matrix for the results presented here. A total of
48 cases are examined (24 nonswirling cases plus their swirling
counterparts). Case pair 0 uses no fuel injection and serves for
quantifying the effect of swirl on the supersonic flowfield and shock
strength. Case pairs 1–13 study the effect of Mrel , wherein the
injectant is helium. The injection velocity of helium is changed to
induce different values of nominal Mrel . The effect of DR is studied
through case pairs 14–23, wherein the injectant comprises different
inert-gas mixtures. The mixture composition is varied to change
mixture density and, consequently, DR. To maintain constant Mrel
throughout the DR analysis, the injection velocity was adjusted
to account for the changes in afuel due to the varying injectant
composition. The values to be examined in both analyses of Mrel and
DR were carefully selected according to the following criteria:
1) The experimentally attainable ranges are spanned with narrow
intervals, to quantify the examined effects accurately. For example,
Mrel is examined over the range 0.44–0.26 with eight intervals, and
DR is covered in the range 12.68–4.33 with six intervals.
2) The numerical simulations span the experimental ranges with
wide intervals: i.e., one simulation at the beginning of the range, Fig. 7 Shock structure of nonswirling (case 1) underexpanded nozzle
one in the middle, and one at the end. Case pairs 1, 5, and 9 within the airflow in presence of nonrecessed coaxial injection system with no fuel
Mrel analysis, for example, were examined both experimentally and injection.
 ABDELHAFEZ AND GUPTA 221

boundary, it reflects into an expansion fan, starting another shock- scope of analysis of the latter and provide the desired quantification
cell unit. The repetition of units is continued until viscous effects of certain flowfield parameters. For example, the schlieren tech-
become predominant and this structure is no longer observed. nique is incapable of quantifying the different components of Mach
In the presence of a coaxial injection system, the shock structure number. Therefore, the numerical results are used to provide the
differs significantly from the simple one described above. Figure 7 missing data needed for examining the air/fuel shear layer and
shows a schlieren image as well as a schematic of the shock structure propagation of swirl throughout flowfield.
of free nozzle flow in the presence of a nonrecessed coaxial injection Code validation was performed by comparing the numerical
system with no fuel injection. Two distinct substructures are iden- centerline static-pressure trace within the near-field supersonic flow-
tifiable from the schlieren image and highlighted in the schematic. field of case 0s1 to corresponding experimental data (see Figs. 4c and
The first substructure is the simple nozzle-rim structure discussed 5a). The centerline static pressure was measured experimentally
above. A new substructure is generated due to the existence of the within the near-field supersonic flowfield by inserting a knife-edge
coaxial injection system. It should be noted here that both sub- circular disc vertically inside the flow (see Fig. 5b). The disc has an
structures are not fully independent of each other. The presence of outer diameter of 0:9D with a 0.5 mm internal channel for
each affects the other. This interaction is not indicated on the transferring the static-pressure signal to a 100 psi pressure transducer
schematic in Fig. 7, however, for easier understanding of the newly of 0.15 psi (0.01 bar) full-scale accuracy. The knife edge of the
introduced substructure of the injection system. Indicated here is circular disc serves for cutting through the supersonic flow with
how each structure would propagate if fully independent of the minimum disturbance on the flat side, which is aligned with the
other. From this point forward, the nozzle-rim and injection-system nozzle centerline. It can be seen from Figs. 4c and 5 that the
Downloaded by UNIVERSITY OF MARYLAND on September 7, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.45716

substructures will be denoted as primary and secondary shock numerical code generally overpredicts the static pressure. This trend
structures, respectively, in this study. was observed to persist over the region of interest, which is composed
The secondary structure starts with the airflow generating an inner of the subsonic flow inside the nozzle and the near-field supersonic
conical boundary that completes the cone-frustum shape of the fuel flow up to 5D downstream of the nozzle exit. A maximum error of
system tip. At the centerline, the flow collapses into itself, generating 7% was observed, which shows good agreement and was considered
a conical shock wave that turns the flow back to parallel. This shock acceptable for the scope of this study.
wave impinges on the outer flow boundaries shortly downstream of In light of the aforementioned code-validation comparison it can
the impingement location of the nozzle-rim expansion fan. The outer be concluded the Baldwin–Lomax turbulence model is capable of
boundaries are altered by the impingement of that conical shock as predicting free supersonic swirling flows with good accuracy. This
observed from Fig. 7. The shock reflects into an expansion fan that negates the common generic belief that this model poorly predicts
creates its own compression fan, intercepting shock, Mach disk, and swirling flows, which might be true under subsonic conditions but
reflected shock, similar to the primary structure. Both Mach disks of not under supersonic conditions, based on the findings of this study.
the primary and secondary structures appear distinctly in Fig. 7. Had the modified versions of the Baldwin–Lomax model been
The effect of coaxial fuel injection is shown in Fig. 8. Helium is available in the FASTRAN-code library, better prediction accuracy
used as the fuel stimulant. As observed, the secondary shock struc- might have been achieved. The reader is referred to previous studies
ture is altered slightly. A shear layer develops in place of the former conducted by the authors [31,32], in which the Baldwin–Lomax
inner conical boundaries of airflow. Because of the presence of model was successfully implemented in simulating confined super-
helium, the shear layer does not converge to a sharp point at the sonic flows using the hybrid LES/RANS FASTRAN code.
centerline. Moreover, due to the curved shape of this shear layer, the
airflow undergoes a gradual compression through a compression fan,
which collapses later into a shock wave that generates the secondary C. Supersonic Flowfield
shock substructure. The main objective of this current study is to analyze how swirl
affects the supersonic flowfield and shock strength in underexpanded
B. Validation of Numerical Code airflow with coaxial fuel injection. Before beginning the analysis, an
Numerical simulations of the flowfield have been performed in this important fact should be recalled here pertaining to the subsonic
study, in addition to the experimental investigation, to broaden the flowfield inside nozzle. This flow is blind to the atmospheric back-
pressure outside, due to the sonic barrier at the throat. In other words,
the subsonic flow does not adapt to the backpressure by fixing the
throat static pressure at a certain value, regardless of nozzle reservoir
pressure. However, once the flow exits the nozzle, it is highly
sensitive to the backpressure. If the flow pressure is different from the
backpressure, the former has to be matched to the latter, which results
in the formation of a shock structure in the supersonic flowfield
outside the nozzle. Convergent nozzles always generate under-
expanded jets if the backpressure is atmospheric. The superatmo-
spheric throat static pressure dictates how much expansion is still
needed and consequently controls the geometry and strength of the
formed shock structure, as explained earlier in the analysis of Fig. 6.
Based on the findings of Adamson and Nicholls [2], Lewis and
Carlson [3], and Crist et al. [4], the axial position of the first Mach
disk (relative to nozzle exit) increases with reservoir pressure. In light
of the present study, this statement should be corrected to state that
the axial position of the first Mach disk is affected by both throat
static pressure and degree of swirl. Before explaining why this cor-
rection is necessary, consider a new case, to be called 0s2, which
is a swirling case like 0s1 but has the lower reservoir pressure of
case 0. All three cases have the same inlet total temperature of 300 K.
Figure 9 shows the schlieren images of cases 0, 0s2, and 0s1. The
horizontal white line highlights the differences in positions of the first
Mach disk in cases 0s2 and 0s1 relative to case 0. It can be observed
that the reservoir pressure is not the sole parameter that controls
Fig. 8 Effect of fuel injection on shock structure of underexpanded the position of the first Mach disk. Although cases 0 and 0s2 have
nozzle flow. the same reservoir pressure, the latter has an axially compact shock
 222 ABDELHAFEZ AND GUPTA

Fig. 9 Schlieren images of underexpanded nozzle airflow with no fuel injection for case 0 (left, nonswirling, reservoir pressure of 7.91 bar), 0s2 (middle,
swirling, reservoir pressure of 7.91 bar), and 0s1 (right, swirling, reservoir pressure of 8.82 bar).

structure. This does not apply for cases 0 and 0s1, which have numbers exist inside this region, which explains why its centerline
different reservoir pressures yet similar positions of the first Mach static pressure is almost equal to the total pressure of surrounding
disk. Only the comparison of cases 0s2 and 0s1 reveals the expected supersonic airflow. The presence of the nozzle-rim expansion fan is
Downloaded by UNIVERSITY OF MARYLAND on September 7, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.45716

trend, in which increasing reservoir pressure results in an axial noticed downstream of the inner region. Subatmospheric pressures
stretching of shock structure. The effect of swirl explains those are reached as the flow passes through the expansion fan twice (refer
unexpected behaviors. Swirling flowfields have two unique features back to the analysis of Fig. 7 for more details). Note that case 0s1
that are absent in nonswirling flowfields. First, the flow is pushed expands from 5.4 to 0.4 atm, whereas case 0 expands only from
radially outward by the Coriolis effect. Second, a swirling jet has two 4.0 to 0.5 atm. This shows that the combination of swirl and higher
velocity gradients (axial and tangential) across its shear layer with reservoir pressure results in a greater expansion fan in case 0s1,
the ambient air, which contributes to increasing the jet diameter, as which is in agreement with the findings of a previous study con-
compared with a nonswirling jet with an axial velocity gradient only. ducted by the authors [27]. The considerably larger dark region
The fact that swirl increases the jet diameter is confirmed in Fig. 9, in observed immediately downstream of the nozzle exit in the schlieren
which it can be observed that the jet diameter within the first shock image of case 0s1 is, in fact, a greater nozzle-rim expansion fan.
cell is slightly larger in case 0s2 than it is in case 0. The larger The realm of the first primary and secondary shock cups is
diameter is accompanied by an axially compact shock structure, identified in Fig. 10 by the recovery from the minimum sub-
which appears to be a common feature of supersonic and subsonic atmospheric pressure to a local maximum of superatmospheric
swirling jets. Thus, it can be concluded that the effects of swirl and pressure. It can be seen that the first primary shock cup is approached
nozzle reservoir pressure interfere destructively from the point of at z=D  1:45 in case 0, 1:4 in case 0s1, and 1:25 in case 0s2.
view of shock-structure axial compactness. Increasing reservoir This agrees with the qualitative analysis made earlier in Fig. 9. The
pressure stretches the shock structure axially (case 0s2 to 0s1), effect of swirl is evident in the upstream shift from 1.45 to 1.25, and
whereas swirl shrinks it (case 0 to 0s2). This explains why the axial the effect of nozzle reservoir pressure is obvious in the down-
positions of the first Mach disk in cases 0 and 0s1 are almost stream shift from 1.25 to 1.4. Both effects interfere destructively,
identical. It should be noted here that the effects of swirl and nozzle yielding almost the same axial compactness of shock structure in
reservoir pressure interfere constructively from the point of view of cases 0 and 0s1.
jet radial expansion. Note from Fig. 9 that the flowfield of case 0s1 The strength of the shock structure can also be quantified from
has a slightly larger diameter than that of case 0s2. Fig. 10. Although the structure of case 0 recovers 0.70 atm, that of
To quantify the effects of swirl and nozzle reservoir pressure in case 0s2 recovers only 0.64 atm. The structure of case 0s1, on the
more detail, Fig. 10 shows the axial distributions of computed other hand, recovers 0.89 atm. Thus, it can be concluded here that the
centerline static pressure in the supersonic flowfields of cases 0, 0s2, application of swirl at matched nozzle reservoir pressure weakens the
and 0s1. Multiple observations can be made from Fig. 10. The flat shock structure as expected. It was shown in the work of Abdelhafez
segments of nearly constant static pressure immediately downstream and Gupta [27] that both nonswirling and swirling flowfields start
of the nozzle exit represent the confined inner region observed in off with the same level of energy upstream of the nozzle, but the
Fig. 9 and discussed in the analysis of Fig. 7. Very low subsonic Mach latter dissipates more energy than the former in friction losses inside
the nozzle. Consequently, the swirling throat flow has a smaller
potential of pressure energy to dissipate through shock structure.
9.0
Centerline static pressure (p/patm)

8.0 Case 0
0.25
7.0 Case 0's

6.0 Case 0s 0.20

Swirl number, S

5.0
4.0 0.15

3.0
0.10
2.0
1.0
0.05
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Axial distance downstream of nozzle exit (z/D) 0.00
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Fig. 10 Axial variation of centerline static pressure within first two
shock cells in cases 0 (nonswirling, reservoir pressure of 7.91 bar), 0s2
Axial distance downstream of nozzle exit (z/D)
(swirling, reservoir pressure of 7.91 bar), and 0s1 (swirling, reservoir Fig. 11 Axial variation of supersonic swirl number within first two
pressure of 8.82 bar). shock cells in case 0s1.
 ABDELHAFEZ AND GUPTA 223

The application of swirl at matched mass flow, on the other hand, number and air/fuel DR with and without swirl. Another very
results in a stronger structure, because of the greater potential of important detail to be pointed out here is that swirl is imparted to the
pressure energy at the throat. air at matched mass flow from this point forward; i.e., all swirling
Having analyzed the effect of swirl on the shock structure, the cases have the same mass flow rate of the nonswirling cases
analysis extends now to examine the supersonic swirl number. (175 g=s). This implies that the swirling cases have the elevated
Figure 11 shows the axial variation of supersonic swirl number within nozzle reservoir pressure of 8.82 bar. Matching of air mass flow
the first two shock cells of case 0s1. The swirl number was computed allows for a fair comparison between the different case pairs,
at select sections inside the nozzle using the correlation [23] primarily from the point of view of mixture fractions and mixedness.
R Ro The effect of Mrel is examined in case pairs 1–13 given in Table 1.
2
R va vt r dr Case pairs 1–9 span the Mrel range 0.44–0.26 experimentally. Case
S  i R Ro 2 (3) pairs 1 (range begin), 5 (range middle), and 9 (range end) were
Ro Ri va r dr
selected to be examined both experimentally and numerically, to
where va and vt are the numerically obtained axial and tangential optimize the parameters of the numerical code for best agreement
velocity components, respectively. It can be observed that the with the experimental results. The experimental range is extended
supersonic swirl number decreases in the axial direction. This is numerically to 0:48 in case pairs 10–13. Case pair 11 represents a
attributed to a substantial increase in axial momentum as the flow unique condition, in which fuel is injected at the throat velocity of air,
expands and accelerates. A minimum swirl number of 0.07 is resulting in Mrel of zero. Case pairs 12 and 13 represent the extreme
observed at z=D of about 1.4, before the swirl number experiences a situations, in which fuel is injected at velocities higher than the throat
Downloaded by UNIVERSITY OF MARYLAND on September 7, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.45716

slight local increase between z=D of 1.4 and 2.7. Recall from Fig. 10 velocity of air. The corresponding values of Mrel are indicated here
that the realm of the first primary and secondary shock cups in without neglecting their negative signs, to highlight the unique nature
case 0s1 starts at z=D  1:4. To understand why the supersonic swirl of those two case pairs.
number experiences a slight local increase within the shock cups, it Keeping all airflow properties constant, the flow rate of fuel
should be noted that the axial and tangential velocity components (simulated by helium) was changed to induce different fuel velocities
behave very differently through shock cups. Figure 12 helps explain and thus multiple values of Mrel . The injection Mach number of
this statement. Depicted is a three-dimensional schematic of shock helium was kept below 0.3 (except in case pairs 12 and 13), to
structure, showing the orientation of va and vt with respect to the main maintain a constant helium density and to avoid compressibility
features: i.e., Mach disk and intercepting and reflected shocks. It can effects on the helium side of the air/helium shear layer. The resulting
be easily visualized how va is always perpendicular to the Mach disk. DR was about 35.5 for most cases. The supply pressure of helium
Moreover, both intercepting and reflected shocks are oblique with was carefully selected for each case pair to match the total pressures
respect to va . For these two reasons, the axial Mach number and of helium and air at injection.
momentum undergo significant reductions through shock structure. Figure 13 shows the effect of Mrel at constant DR. The experi-
The tangential velocity component, on the other hand, is always mental results (schlieren images) are depicted in Fig. 13a, and
parallel to all features of the shock structure. According to the Fig. 13b shows the numerical results in the form of Mach number
fundamentals of gas dynamics, the velocity component parallel to the profiles. The nonswirling cases are depicted in the top row of each
plane of a shock wave is preserved through the shock and experiences figure, and the bottom row contains the swirling cases. The values of
no change. Combining the behaviors of va and vt through shock Mrel and injection Mach number of each case pair are indicated at the
structure, one can easily explain the small local increase in swirl top of its column.
number within the realm of shock cups, especially when one recalls The most remarkable observation to be made from Fig. 13 is that
that the swirl number is proportional to the ratio of average tangential the air/fuel shear layer initiates with a negative angle that transforms
to axial momenta. later to positive. In other words, the cross-sectional area of core
flow converges initially to a minimum value before propagating
divergently as expected. Figure 14 helps explain this observation.
D. Effect of Relative Mach Number Mrel
Shown are the axial variations of computed centerline Mach number
Having attained a good understanding of the effect of swirl on the for different values of Mrel under nonswirling conditions. Note that
supersonic flowfield, the analysis proceeds to examine the effect of subsonic injection is implemented throughout the analysis of Mrel .
fuel injection at different relative Mach numbers. Recall that Mrel is However, the centerline Mach numbers are observed to increase from
defined here as the ratio of vair  vfuel to the average speed of sound, the subsonic injection values to supersonic maxima of 1.75–2.10.
where vair is the throat velocity of air, and vfuel is the injection velocity The only possible way for the subsonic core flow to expand to
of fuel, since fuel is injected coaxially at the nozzle throat (i.e., no supersonic speeds is through the fuel/air shear layer resembling a
recess) in all of the following analyses. Also recall that the close convergent–divergent nozzle. This can only be achieved if the cross-
values of sonic air velocity under nonswirling and swirling con- sectional area of the core flow initially converges to a throat before
ditions allowed for examining the same values of relative Mach diverging again. The creation of a throat allows the core flow to
transition from subsonic to supersonic speeds. To attain further
understanding of the location of the core-flow throat within the
flowfield, Fig. 15 shows how the axial position of this throat varies
with Mrel . Note that at high Mrel (i.e., low injection velocities) the
core flow propagates axially for about 0:5D with a negative shear
angle. The throat of core flow, however, approaches the nozzle exit
at low Mrel (high injection velocities), and negative-shear-angle
propagation is confined to an axial distance of 0:25D at the highest
injection velocity examined here.
It can be concluded here from the analysis of the core-flow throat
that fuel injected at low subsonic Mach numbers (high Mrel ) has
to propagate for longer axial distances with a negative shear angle
before the fuel-rich core flow reaches a throat after which it pro-
pagates supersonically. This is advantageous from two aspects. First,
increasing the distance between the core throat and injection point
allows for more mixing to take place across a supersonic/subsonic
Fig. 12 Three-dimensional schematic of shock structure showing shear layer, which is significantly more effective than the fully
orientation of axial and tangential velocity components with respect to supersonic one downstream of the core throat. The second advantage
Mach disk and intercepting and reflected shocks. of low injection Mach numbers is that a negative-angled shear layer
 224 ABDELHAFEZ AND GUPTA
Downloaded by UNIVERSITY OF MARYLAND on September 7, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.45716

Fig. 13 Effect of Mrel : a) schlieren images (constant DR
35:50) and b) numerical results (constant DR
35:50) in the form of Mach number profiles.

propagates with a radially inward component, which allows it to core flow approaches the injection point, as mentioned earlier.
confine the fuel-rich core flow more effectively. This core is thus Consequently, the negative shear angle and its compression fan
consumed more rapidly by the growing shear layer. diminish gradually. At the extreme of sonic injection a positive shear
Another notable advantage of the presence of a negative shear angle exists right from the start, accompanied by a strong shock at the
angle downstream of the injection point can be observed in the injection point. This results in a stronger shock structure. Figure 13b
strength of the shock structure. As shown schematically in Fig. 16, confirms this discussion. Note that the average Mach number within
the inner conical boundaries and conical shock wave, which initiate the second shock cell is significantly reduced as the injection Mach
the secondary shock substructure in the absence of fuel injection, are number increases and Mrel decreases.
replaced by a compression fan that is generated by the curved profile
of the negative shear angle at subsonic injection Mach numbers. This
fan allows for a gradual compression of the airflow. Although the fan E. Effect of Air/Fuel Density Ratio
eventually collapses into a shock wave, the strength of this wave is Having analyzed the effect of relative Mach number, the analysis
significantly lower than the conical one with no fuel injection. If Mrel proceeds to examine the effect of air/fuel DR at constant Mrel . DR is
is decreased (by increasing injection Mach number), the throat of defined here as the ratio of throat density of air to injection density of
 ABDELHAFEZ AND GUPTA 225

Centerline Mach Number 2.5

2.0

1.5
Mrel = -0.48
Mrel = -0.21
1.0 Mrel = 0.00
Mrel = 0.21
Mrel = 0.26
0.5
Mrel = 0.35
Mrel = 0.44
0.0
0.00 0.25 0.50 0.75 1.00 1.25
Axial distance downstream of nozzle exit (z/D)
Fig. 14 Axial variations of centerline Mach number for different values
of Mrel under nonswirling conditions.
Downloaded by UNIVERSITY OF MARYLAND on September 7, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.45716

fuel. Also recall that the close values of sonic air velocity under
nonswirling and swirling conditions allowed for examining the same
values of DR with and without swirl. Another very important detail to
be pointed out again here is that swirl is imparted to air at matched
mass flow; i.e., all swirling cases have the same mass flow rate of the
nonswirling cases (175 g=s). This implies that the swirling cases
have the elevated nozzle reservoir pressure of 8.82 bar.
The effect of DR is examined in case pairs 14–23 given in Table 1.
Case pairs 15–21 span the DR range of 12.68–4.33 experimentally.
Case pairs 15 (range begin), 18 (range middle), and 21 (range end)
were selected to be examined both experimentally and numerically,
to optimize the parameters of numerical code for best agreement
with the experimental results. The experimental range is extended
numerically from both sides. Case 14 examines a DR of 35.5, and
case pairs 22 and 23 have DRs of 3.24 and 2.29, respectively. Note
that case pairs 10 (from Mrel analysis) and 14 are identical, as they
have the same fuel simulant (helium), Mrel , and DR.
Keeping all airflow properties constant, fuel was simulated by
different inert-gas mixtures (helium, argon, and krypton). The
mixture composition is varied to change mixture density and, con-
sequently, DR. To maintain a constant Mrel of 0.21 throughout Fig. 16 Schematic presentation of how the secondary shock
this analysis, the injection velocity was adjusted to account for the substructure initiates; no fuel injection (left), high Mrel (middle), and
changes in afuel due to the varying injectant composition. The supply low Mrel (right).
pressure of fuel was carefully selected for each case pair to match the
total pressures of fuel and air at injection. Compressible injection is
used throughout this analysis, except for case pair 14, which is copied Fig. 17b shows the numerical results in the form of Mach number
over from the Mrel analysis. Case pair 23 represents the extreme profiles. The nonswirling cases are depicted in the top row of each
conditions, in which fuel is injected at its sonic velocity: i.e., the figure, and the bottom row contains the swirling cases. The values of
injection system is choked. DR and injection Mach number of each case pair are indicated at the
Figure 17 shows the effect of DR at constant Mrel . The experi- top of its column, together with the composition of fuel simulant.
mental results (schlieren images) are depicted in Fig. 17a, and In light of the comprehensive Mrel analysis, the effect of DR will be
analyzed here in a concise fashion. It can be observed from Fig. 17
that the air/fuel shear layer again initiates with a negative angle
0.55 that transforms later to positive. Consequently, the cross-sectional
Axial Position of Core-Flow Throat

Case 1 Case 5
area of core flow converges initially to a throat before propagating
0.50
divergently. This allows the core flow to accelerate from the subsonic
Case 10
Case 9 Mach numbers of injection to supersonic maxima of 1.84–2.13 in
Case 11
Fig. 18. The variation of axial position of core throat with DR is
0.45 shown in Fig. 19. Note that at high DR the core flow propagates
axially for about 0:5D with a negative shear angle. However, at a DR
0.40 of 2.29, the throat of core flow is exactly at the nozzle exit and
Case 12
negative-shear-angle propagation is completely absent. The effect of
0.35 DR on shock-structure strength is observed in Fig. 17b. Note again
that the transition from negative to positive shear angles results in
0.30
stronger shock structure at low DR, as evidenced in the decreasing
average Mach number within the second shock cell.
Case 13
0.25
0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 -0.4 -0.5 F. Shear-Layer Growth
Relative Mach Number The analyses of Mrel and DR revealed how the angle of the air/fuel
Fig. 15 Variation of axial position of core-flow throat with Mrel . shear layer critically affects both the shock structure and mixing. The
 226 ABDELHAFEZ AND GUPTA
Downloaded by UNIVERSITY OF MARYLAND on September 7, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.45716

Fig. 17 Effect of DR: a) schlieren images (constant Mrel
0:21) and b) numerical results (constant Mrel
0:21) in the form of Mach number profiles.

following analysis is thus dedicated to attaining a better understanding of reference to the stationary (traditional) one. Note that the former
of the shear-layer growth, which is another very important parameter. moves with the flow at the convective wave velocity vc. The main-
An analysis of the shear-layer thickness is conducted here, following stream Mach numbers on both sides of the shear layer transform to this
the same roadmap of Papamoschou and Roshko [13]. They defined a new frame of reference as follows:
convective frame of reference that incorporates the effect of flow
compressibility and accounts for different speeds of sound on both vair  vc vc  vfuel
Mcair  and Mcfuel  (4)
sides of the shear layer. Figure 20 compares this convective frame aair afuel
 ABDELHAFEZ AND GUPTA 227

Centerline Mach Number 2.5

2.0

1.5

DR = 2.29
1.0 DR = 3.24
DR = 4.33
DR = 6.46
0.5
DR = 12.68
DR = 35.50

0.0
0.00 0.25 0.50 0.75 1.00 1.25
Axial distance downstream of nozzle exit (z/D)
Fig. 18 Axial variations of centerline Mach number for different values
of DR under nonswirling conditions.
Downloaded by UNIVERSITY OF MARYLAND on September 7, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.45716

Fig. 21 Definition of the shear-layer pitot thickness (Papamoschou and

Roshko [13]).
0.5
Axial Position of Core-Flow Throat

Case 14
Case 15 local velocity is vc on an absolute scale but zero (stagnation) on
0.4 the convective one. This stagnation point forces the equilibrium of
Case 18
total pressures on both sides of the shear layer; that is,
0.3 Case 21

air

fuel

 1 2
air 1
1 2
fuel 1
pair 1  air Mcair  pfuel 1  fuel Mcfuel (5)
2 2
0.2
Case 22
Substituting from Eq. (4) into Eq. (5), the convectivewavevelocitycan
0.1
be solved for by trial and error. Backsubstitution in Eq. (4) then yields
the individual convective Mach numbers of air and fuel. The effect
Case 23
of convective Mach number on shear-layer thickness can thus be
0.0 quantified, which is a direct indication of how the degrees of
40 35 30 25 20 15 10 5 0 compressibility of mainstreams on both sides of the shear layer affect
Air-Fuel Density Ratio its growth rate. Shear-layer thickness is determined as follows.
Consider the shear layer shown schematically in Fig. 21. A pitot
Fig. 19 Variation of axial position of core-flow throat with DR.
thickness pit is defined after Papamoschou and Roshko [13] as the
width of total-pressure profile from 5 to 95% of the difference of
where Mc stands for convective Mach number. Both values are mainstream values. The parameters p,
, v, and a of air and fuel
considerably close, and they always have the same sign, since vc mainstreams are thus calculated at the end points of pit .
always lies somewhere between vair and vfuel . If a streamline is traced The analysis of the shear-layer thickness is carried out here on all
across the shear layer, as seen in Fig. 20, a point has to be met, where the the numerical cases given in Table 1, since their simulations contain
all of the necessary data. The axial positions z=D  0:1 and 1.0 were
selected in each case. Before constructing the final plot of pit versus
Mc , 24 additional simulations were conducted to replicate the 12
numerical case pairs of Table 1 under incompressible conditions
while maintaining their individual velocity and density ratios. The
analysis of the shear-layer thickness was applied to the attained
incompressible flowfields at z=D  0:1 and 1.0 as well. The ratio of

1.0
Papamoschou & Roshko [13]
Pitot-thickness of shear layer

Non-swirling
0.8
Swirling

0.6

0.4

0.2

0.0
0.0 0.5 1.0 1.5 2.0
Fig. 20 Schematic presentation of the stationary (top) and convective
(bottom) frames of reference with sketches of streamlines (Papamoschou Mc,air
and Roshko [13]). Fig. 22 Normalized pitot thickness of the shear layer versus Mc;air .
 228 ABDELHAFEZ AND GUPTA

compressible-to-incompressible shear-layer thicknesses was finally doi:10.2514/3.8141

calculated for all of the 24 numerical cases at hand. Figure 22 is the [2] Adamson, T. C., Jr., and Nicholls, J. A., “On the Structure of Jets from
fruit of this effort. The normalized pitot thickness of the shear layer is Highly Underexpanded Nozzles into Still Air,” Journal of the Aero/
Space Sciences, Vol. 26, No. 1, Jan. 1959, pp. 16–24.
plotted versus air convective Mach number. All data points fall in
[3] Lewis, C. H., Jr., and Carlson, D. J., “Normal Shock Location in
the Mach number range of 0.25–0.83. The findings of Papamoschou Underexpanded Gas and Gas-Particle Jets,” AIAA Journal, Vol. 2,
and Roshko [13] are also included. Good qualitative agreement is No. 4, April 1964, pp. 776–777.
observed, since most data points of this current study fall within the doi:10.2514/3.2409
dashed lines that represent bounding envelopes of the results of [4] Crist, S., Sherman, P. M., and Glass, D. R., “Study of the Highly
Papamoschou and Roshko. A very important conclusion to be made Underexpanded Sonic Jet,” AIAA Journal, Vol. 4, No. 1, 1966, pp. 68–
from Fig. 22 is that the normalized shear-layer pitot thickness 71.
decreases with the application of swirl. It should be noted, however, doi:10.2514/3.3386
that this reduction occurs in the normalized thickness and not the [5] Gostintsev, Y. A., Zelentsov, V. V., Ilyukhin, V. S., and Pokhil, P. F.,
absolute one. As a matter of fact, it was noticed throughout the “Structure of Underexpanded Supersonic Swirling Gas Jet,” Fluid
Dynamics, Vol. 4, No. 5, Sept. 1969, pp. 158–162.
computations of Fig. 22 that the absolute value of the shear-layer [6] Batson, J. L., and Sforzini, R. H., “Swirling Flow Through a Nozzle,”
thickness increases slightly with swirl, which agrees with the Journal of Spacecraft and Rockets, Vol. 7, No. 2, 1970, pp. 159–163.
findings of Cutler et al. [20], Cutler and Levey [21], and Levey [22]. doi:10.2514/3.29892
The controversy is caused here by the fact that the thickness of an [7] Settles, G. S., “Supersonic Mixing Enhancement by Vorticity for High-
incompressible shear layer was found to increase more with swirl Speed Propulsion,” NASA Langley Research Center, NASA-CR-
Downloaded by UNIVERSITY OF MARYLAND on September 7, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.45716

than that of a compressible shear layer. 188920, Hampton, VA, Oct. 1991.
[8] Lee, K. H., Setoguchi, T., Matsuo, S., and Kim, H. D., “Proceedings of
the Institution of Mechanical Engineers, Part C (Journal of Mechanical
VI. Conclusions Engineering Science), Vol. 218, No. 1, 2004, pp. 93–103.
This work provided an experimental/numerical investigation, in doi:10.1243/095440604322786974
[9] Lee, K. H., Setoguchi, T., Matsuo, S., and Kim, H. D., “Influence of the
which the effect of imparting swirl to underexpanded supersonic-
Nozzle Inlet Configuration on Underexpanded Swirling Jet,”
nozzle airflow on the shock structure was examined. Matched mass Proceedings of the Institution of Mechanical Engineers, Part C
flow conditions were considered. A convergent nozzle with swirling (Journal of Mechanical Engineering Science), Vol. 220, No. 2, 2006,
capabilities was used to generate the underexpanded airflow. Fuel pp. 155–163.
was injected coaxially at the nozzle throat. Nonreacting conditions [10] Murakami, E., and Papamoschou, D., “Experiments on Mixing
were considered, in which in the fuel was simulated by mixtures of Enhancement in Dual-Stream Jets,” 39th AIAA Aerospace Sciences
helium, argon, and krypton inert gases. Analyses were made of the Meeting & Exhibit, Reno, NV, AIAA Paper 2001-0668, Jan. 2001.
effects of relative Mach number and density ratio across air/fuel shear [11] Brown, G. L., and Roshko, A., “On Density Effects and Large Structure
layer. The effects of these parameters on the shock structure were in Turbulent Mixing Layers,” Journal of Fluid Mechanics, Vol. 64,
No. 4, 1974, pp. 775–816.
investigated under both nonswirling and swirling conditions. The
doi:10.1017/S002211207400190X
following conclusions were made: [12] Winant, C. D., and Browand, F. K., “Vortex Pairing: The Mechanism of
1) The effects of swirl and nozzle reservoir pressure interfere Turbulent Mixing-Layer Growth at Moderate Reynolds Number,”
destructively from the point of view of shock-structure axial com- Journal of Fluid Mechanics, Vol. 63, No. 2, 1974, pp. 237–255.
pactness. Increasing reservoir pressure stretches the shock structure doi:10.1017/S0022112074001121
axially, whereas swirl shrinks it. On the other hand, both effects [13] Papamoschou, D., and Roshko, A., “The Compressible Turbulent Shear
interfere constructively from the point of view of radial jet expansion; Layer: An Experimental Study,” Journal of Fluid Mechanics, Vol. 197,
both result in greater jet diameter. 1988, pp. 453–477.
2) The application of swirl at matched reservoir pressure weakens doi:10.1017/S0022112088003325
[14] Clemens, N. T., and Mungal, M. G., “Two- and Three-Dimensional
the shock structure. The swirling throat flow is less underexpanded
Effects in the Supersonic Mixing Layer,” 26th AIAA/SAE/ASME/
and has a smaller potential of pressure energy to dissipate through ASEE Joint Propulsion Conference, Orlando, FL, AIAA Paper 90-
shock structure. The application of swirl at matched mass flow, on the 1978, July 1990.
other hand, results in a stronger structure, because the throat flow is [15] Hermanson, J. C., and Winter, M., “Imaging of a Transverse Sonic Jet in
more underexpanded and has a greater potential of pressure energy at Supersonic Flow,” 27th AIAA/SAE/ASME/ASEE Joint Propulsion
the throat. Conference, Sacramento, CA, AIAA Paper 91-2269, June 1991.
3) Fuel injected at low subsonic Mach numbers has to propagate [16] Hall, J. L., Dimotakis, P. E., and Rosemann, H., “Experiments in Non-
initially with a negative shear angle. In other words, the cross- Reacting Compressible Shear Layers,” 29th AIAA Aerospace Sciences
sectional area of fuel-rich core flow converges first, before this core Meeting, Reno, NV, AIAA Paper 91-0629, Jan. 1991.
[17] Miles, J. W., “On the Disturbed Motion of a Plane Vortex Sheet,”
flow reaches a throat after which it propagates supersonically. This
Journal of Fluid Mechanics, Vol. 4, No. 5, 1958, pp. 538–552.
behavior is advantageous, as it results in reduced shock-structure doi:10.1017/S0022112058000653
strength. [18] Chinzei, N., Masuya, G., Komura, T., Mukrami, A., and Kudou, K.,
“Spreading of Two-Stream Supersonic Turbulent Mixing Layers,”
Physics of Fluids, Vol. 29, No. 5, 1986, pp. 1345–1347.
Acknowledgments doi:10.1063/1.865698
This work was supported by the Space Vehicle Technology [19] Menon, S., “Numerical Simulations of Supersonic Flows Past Generic
Institute, jointly funded by NASA, U.S. Department of Defense, and SCRAMJET Flameholders,” U.S. Air Force, Rept.. 434, Wright-
U.S. Air Force within the NASA Constellation University Institutes Patterson AFB, OH, 1988.
Project (CUIP), with Claudia Meyer as the Project Manager. This [20] Cutler, A. D., Levey, B. S., and Kraus, D. K., “An Experimental
support is gratefully acknowledged. The simulation package CFD- Investigation of Supersonic Swirling Jets,” 24th AIAA Fluid Dynamics
Conference, Orlando, FL, AIAA Paper 93-2922, July 1993.
GEOM, CFD-FASTRAN, and CFD-VIEW was provided by ESI- [21] Cutler, A. D., and Levey, B. S., “Vortex Breakdown in a Supersonic Jet,”
Group and CFD Research Corporation. This support is gratefully AIAA Paper 91-1815, June 1991.
acknowledged. Assistance provided by Adam Kareem in data acqui- [22] Levey, B. S., “An Experimental Investigation of Supersonic Vortical
sition and analysis is much appreciated. Flow,” M.S. Thesis, School of Engineering and Applied Science,
George Washington Univ., Washington, D.C., Sept. 1991.
[23] Gupta, A. K., Lilley, D. G., and Syred, N., Swirl Flows, Abacus Press,
References Kent, England, U.K., 1984.
[1] Buckley, P. L., Craig, R. R., Davis, D. L., and Schwartzkopf, K. G., [24] Cutler, A. D., Levey, B. S., and Kraus, D. K., “Near-Field Flow of
“The Design and Combustion Performance of Practical Swirlers for Supersonic Swirling Jets,” AIAA Journal, Vol. 33, No. 5, 1995, pp. 876–
Integral Rocket/Ramjets,” AIAA Journal, Vol. 21, No. 5, 1983, pp. 733– 881.
740. doi:10.2514/3.12362
 ABDELHAFEZ AND GUPTA 229

[25] Linck, M., “Spray Flame and Exhaust Jet Characteristics of a Combustion, Combustion Inst., Pittsburgh, PA, 1981, pp. 81–89.
Pressurized Swirl Combustor,” Ph.D. Thesis, Univ. of Maryland, [30] Abdelhafez, A., and Gupta, A. K., “Swirl Effects on Free
College Park, MD, May 2006. Underexpanded Supersonic Airflow,” 47th AIAA Aerospace Sciences
[26] Baldwin, B. S., and Lomax, H., “Thin Layer Approximation and Meeting and Exhibit, Orlando, FL, AIAA Paper 2009-1557, Jan. 2009.
Algebraic Model for Separated Turbulent Flows,” 16th AIAA [31] Abdelhafez, A., and Gupta, A. K., “Numerical Investigation of Oblique
Aerospace Sciences Meeting, Huntsville, AL, AIAA Paper 78-257, Fuel Injection in a Supersonic Combustor,” 46th AIAA Aerospace
Jan. 1978. Sciences Meeting and Exhibit, Reno, NV, AIAA Paper 2008-0068,
[27] Abdelhafez, A., and Gupta, A. K., “Swirling Airflow Through a Nozzle: Jan. 2008.
Choking Criteria,” Journal of Propulsion and Power (submitted for [32] Abdelhafez, A., Gupta, A. K., Balar, R., and Yu, K. H., “Evaluation of
publication). Oblique and Traverse Fuel Injection in a Supersonic Combustor,” 43rd
[28] Courant, R., Friedrichs, K., and Lewy, H., “On the Partial Difference AIAA/ASME/SAE/ASEE Joint Propulsion Conference & Exhibit,
Equations of Mathematical Physics,” IBM Journal of Research and Cincinnati, OH, July 2007, AIAA Paper 2007-5026.
Development, Vol. 11, No. 2, Mar. 1967, pp. 215–234.
[29] Claypole, T. C., and Syred, N., “The Effects of Swirl Burner K. Frendi
Aerodynamics on NOx Formation,” 18th Symposium (International) on Associate Editor
Downloaded by UNIVERSITY OF MARYLAND on September 7, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.45716
 This article has been cited by:

1. A. Abdelhafez, A. K. Gupta. 2011. Effect of Swirl on Mixing in Underexpanded Supersonic Airflow. Journal of Propulsion
and Power 27:1, 117-131. [Citation] [PDF] [PDF Plus]
Downloaded by UNIVERSITY OF MARYLAND on September 7, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.45716