49th AIAA Aerospace Sciences Meeting including the New Horizons Forum and Aerospace Exposition AIAA 2011-803

4 - 7 January 2011, Orlando, Florida

Thermodynamic Modeling of a Rotating Detonation Engine

Craig A. Nordeen 1 and Douglas Schwer2

University of Connecticut, Storrs, CT 06269-3139
Naval Research Laboratory, Washington, DC 20375

Fredrick Schauer3 and John Hoke4

Air Force Research Laboratory, Wright-Patterson AFB, OH, 45433
Innovative Scientific Solutions Inc., Dayton, OH, 45440

and
Downloaded by UNIVERSITY OF TENNESSEE on October 26, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2011-803

Baki Cetegen5 and Thomas Barber6

University of Connecticut, Storrs, CT 06269-3139, USA

In pursuit of greater thermal and propulsive efficiencies in rockets or gas turbines, a one-
dimensional thermodynamic model of a Rotating Detonation Engine (RDE) is compared to a
numerical simulation model with good results. A ZND detonation model is modified to
include stagnation properties and account for the velocity vectors that occur upstream of the
detonation. Features of the RDE and their impact on the model are discussed. Velocity
triangles, commonly used in the gas turbine industry, are shown to be an effective tool for
understanding energy transfer in RDE’s.

I. Introduction

A continuous detonation or rotating detonation engine (RDE) utilizes

supersonic combustion phenomena that may deliver as much as a
20% increase in propulsive or thermal efficiency over conventional
methods.1 A RDE contains a detonation wave that propagates
continuously around an annular combustion chamber, while a fuel and
oxidizer mixture is delivered continuously as shown in Fig. 1. The
exhaust is continuous supersonic flow with a rotating pressure wave. The
increase in efficiency is expressed as a higher kinetic energy, rather than
an increase in pressure. Possible applications may be found in rockets or
gas turbine combustion chambers.
The idea and first embodiment of a continuous detonation device was
conceived in the 1950’s. Early work at the Lavrent’ev Institute of
Hydrodynamics in Siberia2 and the University of Michigan3 explored the Figure 1. 3-D RDE simulation.
possibilities of such a device. Today experimental and theoretical work
continues at more than a dozen institutions worldwide.4,5
The continuous nature of the flow is more compatible with existing propulsion technology when compared with
pulse detonation engines (PDE). The simplicity of the device eliminates several issues inherent in typical PDEs.
Except for the initial start sequence, ignition and transition to detonation issues are eliminated, and detrimental
vibration effects are reduced. Mechanical valves are eliminated, and complexity is reduced to nonmoving parts.

1
Graduate Assistant, Mechanical Engineering, 191 Auditorium Rd, UConn, Storrs, 06269, AIAA Member.
2
Mechanical Engineer, Center for Reactive Flow and Dynamical Systems, Code 6410, AIAA Member.
3
Research Engineer, Propulsion Directorate, 1790 Loop Rd., Dayton, OH 45433, AIAA Senior Member
4
Senior Engineer, 2766 Indian Ripple Rd., Dayton, OH 45440, AIAA Senior Member
5
Professor and Department Head, Mechanical Engineering, 191 Auditorium Rd, UConn, Storrs, 06269
6
Professor-in-Residence, Mechanical Engineering, 191 Auditorium Rd, UConn, Storrs, 06269, AIAA Associate Fellow

1
American Institute of Aeronautics and Astronautics

Copyright © 2011 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.
 Understanding the operation of a RDE requires a basic thermodynamic model. The requirements for this model
are driven by its suitability as an initial analysis tool of a RDE in much the same way that a Brayton cycle model is
used for preliminary analysis of gas turbines. The model must be one-dimensional and independent of flow
geometry. There must be means to account for the first order effects of thermodynamic states and an accounting of
loss mechanisms. An assessment of efficiency and performance must be made with a reasonable degree of fidelity.
Common thermodynamic equations of state should be used and the chemistry of combustion should be manifest
only as heat added and appropriate gas constants. Above all, the model must be understandable at a fundamental
level.
A thermodynamic assessment is made of a rotating detonation wave engine for the purpose of creating a
parametric model. This model is based on a ZND (Zeldovitch-von Neumann-Doring) 6 analysis modified by the use
of the Rankine-Hugoniot equations and the application of a vector analysis of the upstream conditions. This model is
compared to the thermodynamic cycle based on data from a computational simulation of an RDE.
With some adjustments, the modified ZND model approximates many features of the computational model.
Further refinements should improve the predictability of the model. This model provides a reasoned thermodynamic
Downloaded by UNIVERSITY OF TENNESSEE on October 26, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2011-803

basis for theoretical understanding, design and testing of RDE’s.

II. Numerical Simulation

The simulation method is documented in a separate paper by Schwer and Kailasanath7 and will not be discussed
in detail. In summary, a premixture of hydrogen-air is injected through micro-nozzles along the inlet wall. The
model is a two-dimensional Euler computation without heat or viscous diffusion. The chemistry of combustion is an
induction parameter model.
The modeled chamber is 14 cm in diameter by 17.7 cm long and is modeled on a 0.2 mm x 0.2 mm grid. The
heat added is 3.5500e10 erg/gm. The molecular weight of the reactants is 20.9167. Specific heats were extracted
from the simulation are 1.4256 for the reactants and 1.2412 for the products. The gas constants are 3.975e6
erg/gm/K for reactants and 3.477e6 erg/gm/K for products.

III. RDE General Features

A proper model of the thermodynamic cycle requires an understanding of the transfer of energy in an RDE.
There are many processes involved, and only the most significant will be discussed. The wave will be conceptually
treated as a shock wave with heat addition, as in the traditional ZND analysis. The transfer of energy through the
wave can be followed through a series of vector diagrams along streamlines of relative flow in the rotating frame of
reference, and the corresponding path lines in the fixed frame of reference. These same streamlines form the basis
for an enthalpy-entropy cycle analysis. For a number of reasons, the streamlines exhibit distinct thermodynamic
cycles. However, the streamline cycles are not so different as to exclude a generalized RDE cycle that will be the
basis of the one-dimensional model. Before the streamlines are discussed, a description of the basic features of the
RDE will create a useful vocabulary. Investigators including Hishida8 have explored many of these features.

Figure 2. Unrolled RDE contour of stagnation enthalpy and major features.

2
American Institute of Aeronautics and Astronautics
 The rotating detonation wave can be viewed in a moving frame of reference tied to the azimuthal motion of the
wave. If the wave features are essentially unchanged across the radial chamber gap, it may be reduced to a two-
dimensional phenomena as shown in Fig. 2, and “unrolled” onto a Cartesian coordinate system. Axial dimensions
are parallel to the Y-axis, and azimuthal dimensions are plotted along the X-axis. Reactants enter the chamber from
the lower boundary fed by a constant pressure plenum (not shown). Products exhaust from the top boundary.
Periodic boundary conditions exist at the left and right borders. This model shows a single wave. We shall assume a
single wave for the purposes of this discussion.
Fig. 2 is a three-dimensional contour snapshot from a time-accurate numerical simulation. The vertical axis
(perpendicular to page) and the color map represent stagnation enthalpy. A light source highlights many subtle
features that are not apparent in two-dimensional contour plots. A fill zone of uncombusted reactants (A) feeds the
detonation wave (B) which is moving from left to right.
In the wake of the detonation, transverse waves (C) form a cellular pattern familiar to students of detonation.
Ripples from the transverse waves continue to spread and fill the entire expansion region (D), which includes all
combustion products outside of the detonation and fill zone. The ripples move downstream from the detonation and
Downloaded by UNIVERSITY OF TENNESSEE on October 26, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2011-803

appear to be synchronous with the shedding vortices of the shear layer (E). The shear layer marks a boundary
between freshly detonated products and older combustion products that have circled back around the annulus. This
shear layer forms an effective deflection ramp for the older products and defines one boundary of the expansion
geometry. Velocities in the rotating frame of reference tend toward the supersonic. Therefore, the deflected older
products result in an oblique shock wave (F) originating at the intersection (G) of the fill zone boundary (H) and the
detonation wave (A). The shock shows evidence of separation in this area forming a protruding normal shock at the
origin (G). The angles of the shock and shear layer are controlled by the dynamic balance between the two zones.
Preliminary studies also show a dependence of the angles on the local gas constants.
The rotating wave creates a blocked inlet region (J) from the high, but decaying, pressure field of the detonation.
This blocks injection flow from the inlet boundary from entering the chamber. Some experimental investigators9
have reported injector flow reversing into the injectors and supply plenums. The lower boundary is numerically
modeled effectively as isentropic micro-jets governed only by the difference of pressure across the boundary. If the
chamber pressure is higher than the plenum chamber, flow is zero and backflow is prevented. This represents the
most elemental method of injection that can capture RDE phenomena. Other methods of injection modeling are
beyond the scope of this paper.
Once the wave pressure has decayed below the supply plenum pressure, flow will recover and begin to flow
from the recovery zone (K). The recovery flow (L) is, however, exposed to high pressure and temperature and is
ignited. This flow does not pass through the detonation, and subsequently follows a distinctly different
thermodynamic cycle. The recovery flow is marked by the absence of the transverse wave ripples.
The stopping and restarting of flow is a necessary condition of stable operation of the RDE. The passing
detonation wave acts as a fluidic valve shutting off its own fuel supply and by its passing, restarts the flow. The
approximately triangular fill zone shape is a consequence of the restarting flow and, because of the circular duct,
allows the combustion products to act as a fourth boundary for their own detonation wave. The other three
boundaries are the inlet, inside and outside annular walls. Together these form a rectangular channel for the
propagation of the detonation.
Several expansion waves (M) emanate from the start of the recovery zone. The effect on the thermodynamics
will be discussed below.
A faint secondary shock (N) originates from the oblique shock origin (G). It wraps around and re-enters from the
right periodic boundary before penetrating and reflecting off the fill zone boundary (H). The trajectory of the
secondary shock is partly dependent on the gas constants.
The lower portion of detonation flow (O) will pass through the oblique wave (F). This is a strong shock and has
a visible effect on the thermodynamic cycle of the lower flow. The upper portion of the detonation flow (P) does not
pass through the oblique wave and follows a different cycle path. The percentage of flow through these different
areas has dependency on geometry. In particular, the axial length of the annulus creates a cutoff for the oblique
shock.

3
American Institute of Aeronautics and Astronautics
 IV. Nomenclature
 = density Uwave = azimuthal velocity vector of rotating wave

= specific heat ratio V = velocity vector in fixed frame of reference
 =1/M02 W = velocity vector in rotating frame of reference
Cp = specific heat at constant pressure x = 0/ =specific volume ratio
C-J = Chapman-Jouget y = p/p0 = pressure ratio
D = detonation velocity
h = enthalpy Subscripts:
Isp = specific impulse 0,1 = upstream and downstream conditions
M = Mach number amb = ambient
p = pressure ref = reference state
s = static property
Downloaded by UNIVERSITY OF TENNESSEE on October 26, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2011-803

q = heat added
Rgas = Specific gas constant t = stagnation or total property
s = entropy x = azimuthal direction
S = Rankine-Hugoniot function y = axial direction
T = temperature

V. Numerical Simulation Streamlines

Streamlines are created in the rotating frame of reference by first subtracting the azimuthal wave velocity (U)
from the absolute velocity (V) using the vector sum Eq. 1.


 


W = V
U (1)
This produces a velocity field (W) in the rotating frame of reference. Twenty streamlines are started along the
inlet boundary spaced equal distances apart. The starting positions as a group are indexed such that streamline 5
straddles the fill boundary and runs near the center of the shear layer. When the streamlines reach the left periodic
boundary, they are restarted on the right. This is repeated until the streamlines terminate at the exit boundary.
Streamline flow starts at the bottom inlet boundary and moves up and to the left in the rotating frame of reference.
The streamlines are then used as the basis for an interpolation of the numerical simulation flow states. Pressure,
temperature, velocity vectors, and other values may then be obtained along the streamline. Figure 3 is a two-
dimensional color contour plot of stagnation enthalpy, with streamlines showing flow in the rotating frame of
reference.

Figure 3. Stagnation enthalpy contour with streamlines.

4
American Institute of Aeronautics and Astronautics
 The difference between Fig. 2 and Fig 3 should be noted. Many of the subtler features, such as the secondary
shock, and the regular nature of the transverse wave ripples, are not apparent in Fig. 3. Even with a narrower color
field, the magnitudes of many features are small and are not resolved using this plotting method.
Several issues arise from this approach. The use of a velocity vector field assumes that the flow field is static.
The description of the shear layer instabilities and the existence of detonation transverse waves indicate that the field
is far from static. Furthermore, the assumption of a steady rotational velocity is undermined by the knowledge that
the rotational velocity in this simulation appears to undergo a subtle limit cycle. Simulation of planar detonation
waves and RDE laboratory experiments confirm the existence of the phenomena.10,11
The streamlines shown in Fig. 3, therefore, do not represent a true steady state process. Even though broad trends
are evident, thermodynamic properties taken from the time-accurate flow field show excessive noise and spurious
values in derived quantities such as enthalpy and entropy. Therefore, a time-averaging approach is taken. The field
values are averaged over ten rotations of the detonation wave. The transient features are suppressed and a steady
state field is acquired. Streamlines are then created from the result as shown in Fig. 4.
Downloaded by UNIVERSITY OF TENNESSEE on October 26, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2011-803

Figure 4. Time-averaged stagnation enthalpy with streamlines.

VI. Simulation Pathlines

Pathlines are the locus of fluid particles in the laboratory frame of reference, and can be computed from the
streamlines calculated above. The y values of the streamline points are the same as the pathlines, since rotation is
only in the x direction. The fixed frame velocities are known along the streamline and an integration is performed to
compute the x values starting at the origin of the streamlines. The computed pathlines are shown in Fig. 5.

Figure 5. Time-averaged pathlines in fixed frame of reference.

5
American Institute of Aeronautics and Astronautics
 An Archimedes screw may be seen as a mechanical analog to the RDE. The streamline threads are formed in the
rotating frame by the action of the detonation wave. As the wave moves circumferencially to the right, the fluid
particles are driven up the pathlines. Unlike the Archimedes screw, the fluid is compressible and the path is not
straight. Turning of the flow is necessary to do work on the fluid. As will be shown, this is identical to the turning
and acceleration of gases in turbomachinery and is essential to understanding the cycle and the energy transfer.
The deviation of the pathlines from a pure axial motion can be interpreted as local swirl and implies work may
be done on or by the gases. This is evident in the portion of the lines corresponding to the fill zone flow from the
inlet. The injected inlet flow at the boundary enters as a pure axial vector. Swirl is induced as the pathlines traverse
the pressure gradients in the fill zone. The fill zone will be discussed in detail in a future paper.
While the streamlines do not cross by definition, the pathlines do cross. Particles with the same time stamp enter
the crossings at different times as the wave passes through the fixed field. Because of the lack of easy visualization,
the rotating field is the preferred choice for studying the process.
It is noted that not all twenty streamlines are represented. Streamline 10 originates in the region of blocked flow
and is suppressed. An examination of Fig. 4 shows its tag located on the inlet border. Streamlines 9 and 10 are
Downloaded by UNIVERSITY OF TENNESSEE on October 26, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2011-803

supressed in Fig. 3 because the starting points are indexed to a different location.

VII. Streamline Thermodynamic Cycle

Pressure, temperature and velocities in the fixed frame are extracted from the field data interpolating along the
streamlines. The fixed frame velocities along the streamline are used to compute the kinetic velocity head. Thus, a
stagnation enthalpy may be computed from Eq. 2.12

2 
2
V =Vx +Vy
2

1 2
ht = hs + V = C pTt (2)
2
ht = C p (Tt
Tamb )
It is noted that the fixed frame velocities are used in the computation of enthalpy, and the rotating frame
velocities used to establish the streamline. The fixed velocity head is used in stagnation calculations because that
represents the real work on the gas in the fixed frame.
Stagnation or total entropy may now be computed in Eq. 3 from the static temperature and pressures.
Alternately, it can be shown that entropy can also be computed from the stagnation pressure and temperature. In
addition, standard compressible flow formulas are used to compute stagnation for pressure and volume.
 T   p   T   p 
s = Cp ln  s 
Rgas ln  s  = Cp ln  t 
Rgas ln  t  (3)
 Tref   pref   Tref   pref 
Cp,
and Rgas are typically functions of combustion chemistry, temperature and fluid composition. For this study,
gas properties are assumed to have different, but constant, values for the reactants and the products of combustion.
During combustion, specific heat ratio (
) and the gas constant (Rgas) are computed as a linear function of reaction
progress, which is available from the simulation flow field data. Cp is then computed from
and Rgas. These gas
constants are not computed directly in the numerical simulation, but are derived values from the computational
chemistry.
Total and static enthalpy for each streamline may now be plotted against entropy in standard H-S diagrams, and
pressure and volume are plotted in P-V diagrams. Fig. 6 is a plot of static pressure and volume. Fig. 7 contains
stagnation pressure and volume values. Fig. 8 shows static values of enthalpy vs. entropy. Fig. 9 details stagnation
enthalpy against entropy. The difference between the stagnation and static values is the kinetic energy in the fixed
frame of reference.
All streamlines are from the time-averaged numerical simulation data. The modified ZND model has not been
discussed yet, but is plotted along with the streamline data for comparison.
In general, the numerical streamline thermodynamic states show very similar structure to the classic ZND
models in both static, stagnation, H-S and P-V diagrams. Obviously, there is a difference between the models, which
may be inaccurately described as a scaling issue due to the similarities of the curves. These differences will be
discussed and future work will narrow the difference. In future studies, the models will be validated against
experimental data.

6
American Institute of Aeronautics and Astronautics
Downloaded by UNIVERSITY OF TENNESSEE on October 26, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2011-803

Figure 6. Static P-V diagram of streamlines & ZND cycle

Figure 7. Stagnation P-V Diagram of streamlines & modified ZND cycle

Figure 8. Static H-S diagram with streamlines & ZND cycle.

7
American Institute of Aeronautics and Astronautics
Downloaded by UNIVERSITY OF TENNESSEE on October 26, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2011-803

Figure 9. Stagnation H-S diagram with streamlines & modified ZND cycle

A. Detonation Streamline Cycles

As shown in Fig. 6-9 each streamline has a unique thermodynamic cycle. The circles indicate an approximate
start of the detonation shock wave, a similar rise in pressure for the recovery flow or the start of a suppressed
streamline. Conditions in the fill zone create unique upstream conditions for each streamline. These precise
conditions will not be discussed in detail and are the subject of future work.
In brief, the fill zone possesses a pressure gradient that accelerates and turns the flow. The longer streamlines
have more time to accelerate and enter the detonation with a velocity greater than the shorter streamlines. The fill
zone pathlines in Fig. 5 show an increasing deviation from the initial axial velocity imparted by the injectors. This
induced swirl is a result of the acceleration. As a result, Mach numbers range from approximately 0.5 near the inlet
boundary to near sonic conditions near the fill zone border. The increasing upstream velocity represents an increase
in entropy as the increased velocity head is carried across the detonation shock. As a result, the enthalpy and entropy
of each streamline upstream of the shock varies considerably and contributes to the difference between each
streamline trace.
A Brayton cycle is shown for reference in Fig. 8 and 9. The compressor discharge of the Brayton cycle and the
start of the ZND model are tied to a H-S point based on a density average of enthalpy and entropy just upstream of
the start of the numerical shock. This point will be referred to as station 3, consistent with common turbomachinery
nomenclature for compressor discharge. Thus in Fig. 8, station 3 is tied to the average static conditions and in Fig. 9,
positioned at the corresponding stagnation conditions. For this reason the Brayton cycle in Fig. 8 shows as a very
low pressure cycle and is barely covered by the thickness of the plot lines. Station 3 conditions in the fill zone for
this model range from 1-1.2 atmospheres static pressure and 1.8-2.0 atmospheres total pressure. Future studies will
include sensitivity to inlet chamber pressures.
The detonation wave shows only as a thin line in Figures 2-4 and is about two dozen computational cells thick.
In the HS diagrams this process covers the streamline trace from the start of detonation to the thermal choke point.
The streamline maximum pressure occurs before the thermal choke is reached. Maximum enthalpy or temperature
occurs at or before the start of the isentropic expansion.
Once past the thermal choke the streamlines are expanded isentropically on average. Deviations from the
isentropic are due to standing waves in the rotating flow field. Those streamlines that pass through the oblique shock
are pumped up to a higher enthalpy before expanding again isentropically. However, this moves their expansion
lines as entropy is increased by the oblique shock. This suggests that the chamber length should be as short as
possible to minimize the losses associated with the oblique shock.

B. Non-detonation Streamline Cycles

Streamline 5 displays the most thermodynamic variation of those that deviate from the general detonation path of
SL 1-4 and 11-20. The streamline was positioned quite deliberately so that it would straddle the fill zone boundary
and run close to the middle of the shear layer. Referring to Fig. 9, SL 5 starts at the very beginning of the fill zone
(5a) where fresh unburned reactants are first injected after the recovery flow zone has moved on. It follows a
8
American Institute of Aeronautics and Astronautics
 decaying pressure trajectory heading in the general direction of station 3. Before it gets to station 3 it reverses and
heads back up close to the upper constant pressure line of the Brayton cycle. This reversal is the crossing SL 5
makes across the fill zone boundary, where it is ignited and burns at a near constant pressure. At 5c, it enters the
oblique shock origin (G) and is compressed (5d). From there it is expanded through the midline of the shear layer
(5e).
Streamline 5 represents a narrow band of flow that starts as an unburned mixture in the fill zone. It traverses the
fill boundary and combusts in a constant pressure process. In other words, this narrow band of flow is not parallel to
the fill zone boundary, but cross the boundary at a very shallow angle.
Streamlines 6-9 are recovery flow. They are combusted along a constant pressure line before passing through the
expansions waves (M) shown in Fig. 2. They are then compressed by the oblique shock before completing their final
expansion.
Streamline 10 shows as a single point on the H-S diagram (Fig. 8). It is part of the suppressed flow from the Fig.
2 blocked region (J).
Downloaded by UNIVERSITY OF TENNESSEE on October 26, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2011-803

VIII. Modified ZND model

The Zeldovitch-von Neumann-Doring detonation model is the most commonly used conceptual model of a
planar detonation. The ZND cycle used in this work is common to many combustion texts and will be briefly
described. The reader is referred to Lee6 for derivations of the version used for this analysis. The typical one-
dimensional analysis of detonations starts with the assumption of a planar wave traveling through a mixture capable
of detonation. The mixture is at rest in the laboratory reference. The detonation consists of a normal shock, an
induction zone, reaction zone and a thermal choke. The normal shock raises pressure and temperature to levels
which start an auto-ignition sequence in the mixture. The combustion chemistry is assumed to require an induction
period, which delays the release of heat until a period after the max pressure is attained.
Figure 6 traces the traditional Hugoniot analysis. The leading normal shock starts at station 3 where pressure and
volume are at the upstream conditions. The normal shock follows the shock Hugoniot curve up to the von Neumann
peak. The controlling assumptions are important to recall. These curves are dependent on the static properties of the
upstream fluid. These are the curves obtained for a fluid at rest and contain no kinetic energy. The kinetic energy of
the downstream fluid is implicitly accounted for in the Hugoniot curve of all possible states for a control volume
analysis with heat addition. The upstream static pressures used for Fig. 6 and 8 are the static pressures obtained for
the Mach conditions of the fill zone.
Figures 7 and 9 plot the stagnation properties of the RDE using the modified ZND model. To understand this an
examination of the RDE velocity vectors is required.
A. Vector Sum and Stagnation Properties.
The velocity vectors surrounding the detonation are shown in Fig. 10. The use of this diagram is not limited to
the detonation wave. Variations on this theme will explain the transfer of energy throughout the RDE. Equation 1 is
repeated for reference. The vector sum follows typical turbomachinery convention,13 except that V is substituted for
C in the commonly used variation: W=C-U.

Figure 10. Vector sum of rotating detonation.

9
American Institute of Aeronautics and Astronautics
 An understanding of this diagram starts with the vector V0, which is the fill zone velocity in the fixed frame of
reference and presents a crosswind vector to the approaching detonation. V0x is the azimuthal component and is
initially zero when the mixture is injected normal to the inlet boundary. As the inlet flow makes its way through the
fill zone, the decaying pressure wave from the previous detonation creates a pressure gradient forcing the inlet flow
to the left, thus inducing negative swirl (turbomachinery convention). The swirl angle in this simulation is small,
about 3 degrees on average. However, with higher inlet velocities, inlet swirl vanes and at off design points, swirl
could become more significant. As seen in Fig. 10, swirl has an impact on V0W, the projected component normal to
the shock. V0W is significant because it is the kinetic energy that is carried across the leading shock of the detonation.
This velocity head is largely converted to heat as it crosses the shock and represents an efficiency loss to the system.
V0p is parallel to the shock and its kinetic energy is carried across isentropically. A designer of a RDE will be
motivated to reduce V0W to a minimum.
To solve this velocity triangle, it is necessary to find the detonation velocity, D. If V0 and D are known, Uwave and
W0 may be determined. Uwave is the azimuthal velocity of the detonation wave and under the steady state assumption
is common to all velocity triangles. W0 is the upstream velocity that carries fluid particles across the detonation
Downloaded by UNIVERSITY OF TENNESSEE on October 26, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2011-803

shock. Values of Uwave computed by this method are found to be within 5% of the numerical simulation wave
velocity.
The detonation velocity D may be found from Eq. 4, in the Rankine-Hugoniot equation set.6

p1 

+ *

S

= 0 1
0D
2
(

+1
0 1 )

p 



S

+ * (4)
y= 1= 0 1
p0 (
+1 
1
*
)
u 
 W1 

+ * ± S

x= 1 = = 0= 1 1
( )
  
u 
0 Lee
 W 0 Fig.10 1

+1
0 1
 is equal to 1/M0 at the Chapman-Jouget (C-J) condition. S is a function of the specific heat ratios and the heat
* 2

added, but is equal to zero at the C-J condition. The downstream pressure, p1, of the detonation is found by Eq. 4.
The variables x and y in this context are the axis of the static PV diagrams (Fig. 6).
The value in Eq. 4 of pressure and density are static values and dependent on the stagnation pressures and
temperatures. W0 and W1 are the velocities of fluid particles relative to the wave and associated with the relative
stagnation enthalpy, which is constant across the shock. Since Uwave is known, V may be computed using the vector
sum Eq. 1 and indicated in Fig. 10. It is the fixed frame velocity, V, which is used to compute the stagnation
enthalpy of the fluid. This is true of V1 as well as the intermediate values of the shock Hugoniot VSH and the
Rayleigh line VRL.
The Hugoniot equation (5) can be used to compute intermediate state values along the shock Hugoniot where the
heat added q equals zero.
 +1
0
x + 2 * q'

1 q
y= 0 where q' =  (5)
 +1 p 0
1 x
1 0

1
1
The von Neumann peak pressure point is found when q equals zero in the Rankine-Hugoniot equations for the
Rayleigh line. Intermediate values along the Rayleigh line are computed as q is incremented from zero to the heat of
combustion.
The Rayleigh line has a subsonic and a supersonic portion as shown in Fig. 6. The subsonic Rayleigh line is used
in this model since the heat release occurs subsonically relative to the wave coordinates (W). The same velocity (V)
in the fixed frame can be supersonic and is used for the stagnation properties. A portion of the argument for the
subsonic Rayleigh line is discussed in section IX. The reader is referred to Lee6 or Kuo15 for more details.
The same method of computing intermediate values of Cp,
and Rgas are used for Shock Hugoniot and Rayleigh
lines as were used for the stagnation properties of the streamlines. Please see the above discussion on the streamline
construction.

10
American Institute of Aeronautics and Astronautics
 B. Energy Transfer Through Pathline Turning
An examination of Fig. 5 and 10 shows a turning of the flow in the fixed frame of reference as V0 becomes V1
when the detonation passes. This is the same phenomena seen in one-dimensional detonations where the
downstream velocity follows, but does not catch, the detonation wave. In this way, the detonation transfers energy to
the axial flow by turning the flow in addition to the compression and heat addition. For this flow field a leftward
turning indicates work may be done by the fluid, and for a rightward turning, work is done on the fluid. The reader is
referred to turbomachinery texts, such as Hill and Peterson,13 for more information. Turning, such as occurs in the
fill zone or at the crossing of the oblique shock, can be seen operating all through the flow field and will be the
subject of future work.

C. Rayleigh Line Sensitivity To Gas Constants

The shape or local slope of the Rayleigh line is governed by the variability of the gas constants. The end points
are fixed by the known gas constants for reactants and products. The Rayleigh lines in most texts are computed with
Downloaded by UNIVERSITY OF TENNESSEE on October 26, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2011-803

constant values of
and Rgas. These shapes of both subsonic and supersonic lines show a tangency to each other and
to the upper C-J point on the Hugoniot (Saad12). This tangency has not disappeared in Fig. 8 and 9, but has moved to
a very small region near the upper C-J point.
In the usual discussion of stable planar detonations, temperature is normally shown to reach a maximum at the
thermal choke at the exit plane of the detonation. Typically, this is considered the end of the detonation proper and
identical with the upper C-J point. From this argument, it would be expected that the slope of the Rayleigh line
would be zero at the C-J point. As discussed, with constant properties it is not. The drop in enthalpy from the
maximum down to the C-J point also represents a drop in temperature contradicting the thermal choke assumption.
In most situations, the ultimate dominance of the Fanno line makes this issue moot. For this detonation model,
viscosity and friction losses are assumed negligible and the issue remains.
An examination of the Euler simulation streamline behavior shows a tendency to follow a subsonic Rayleigh line
with a drop in enthalpy before the C-J point and subsequent expansion. It is known that heat release in the
simulation continues after the maximum enthalpy. The stream undergoes a small non-isentropic expansion in this
region and is consistent with the general shape of the subsonic Rayleigh line.

D. Gas Constant Dependence On Temperature.

The assumption in this model is that the gas constants are linearly dependent on the progress of reaction. Since
the gas constants are known to be non-linearly sensitive to temperature, and the shock Hugoniot is computed with
constant
and Rgas, it is implied that the current model constants are not a function of temperature. This was a
conscious choice. Modeling dependence on temperature typically requires an integration loop to converge on the
properties. Avoiding this complication insured a stable model and isolated the sensitivity to reaction progress. Future
models will incorporate property dependence on temperature.

E. Discontinuity Of The Stagnation Hugoniot

The stagnation Hugoniot as shown in Figures 7 and 9 is discontinuous with the lower Chapman-Jouget point.
The upstream velocity along the stagnation Hugoniot was necessarily held constant. As a result, the Hugoniot
equation shows no solution for the deflagration section of the Hugoniot. A related investigation into the effects of
the upstream velocity normal to the flow has been conducted by Wintenberger and Shepherd.14
Wintenberger has compared a detonation ramjet with a conventional deflagration ramjet. In these two cases, both
the detonation and the deflagration have flame fronts normal to the chamber flow. The upstream velocity for the
deflagration approaches zero, recovering any dynamic pressure from the air stream. The upstream velocity for the
detonation is necessarily equal to the detonation velocity to hold the wave stable in one position. Under these
conditions, the detonation ramjet is not as efficient as a deflagration ramjet. The entropy in these cases is greater for
the detonation ramjet, reversing the normal “S” shaped curve of the static Hugoniot. The stagnation Hugoniot shows
a discontinuity between the detonation and deflagration sections indicating no real solution for these areas.
The reason that a practical detonation engine must be a PDE, RDE or OWDE lies in this discussion. These are
all machines that reduce the upstream velocity in order to reduce the dynamic head losses through the detonation
shock. In the case of the PDE, pulsing the detonation allows the upstream velocity to come to rest, recovering the
dynamic head as static pressure. Both the RDE and the OWDE reduce this dynamic head through an oblique wave.

11
American Institute of Aeronautics and Astronautics
 IX. Numerical Simulation And Modified ZND Model Comparison

The streamlines that pass through the detonation (1-4,11-20 in Fig. 9) are tightly grouped and follow a pattern
that is similar to the ZND detonation model. The similarities and differences between the streamlines and the ZND
are still under study. Entropy prediction at the isentropic expansion is quite accurate, but enthalpy and pressures
show lower values than the ZND model, and the maximum values occur at different points in the cycle. Similarities
in the curve shape imply that similar processes are at work.
From the start of detonation, the streamlines follow a shock pressure curve that does not show the high values of
total enthalpy or velocity head indicated by the ZND shock Hugoniot. An examination of the streamlines shows that
the release of heat in the numerical model occurs along the entire length of the enthalpy rise, rather than after the
max pressure of the ZND model. The heat release starts with a short induction delay and reaches a maximum at the
streamline max pressure point. A small amount of heat addition to the streamlines actually continues past the
thermal choke into the expansion, effectively reducing the combustion efficiency of the detonation. The
thermodynamic effect alone would suppress the maximum temperatures predicted by the simulation and account for
Downloaded by UNIVERSITY OF TENNESSEE on October 26, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2011-803

some of the difference between the ZND model and the simulation. As described, this will reduce pressures and
enthalpy is reduced relative to the ZND model.
However, when we compare Fig. 6 with Fig. 11 taken from Kuo15, we see that the rate of chemical reaction
could be affecting the path of streamline reaction. Faster chemical kinetics produce curves with a lower max
pressure. Thus, a thermodynamic model that would accurately mimic the simulation streamlines must account for
the speed of chemical reaction. In this case, the hydrogen-air mixture of the simulation is one of the faster chemical
reactions, but not generally considered fast enough to release heat during the initial shock. The original goal of this
study is to produce a simple one-dimensional thermodynamic model, but it is possible that future models may need
to factor in the speed of chemical kinetics.

Figure 11. ZND detonation structure and effect of chemical kinetics.

A possible explanation of the difference between the ZND model and the simulation may be found in the nature
of the numerical simulation. The detonation is approximately two dozen computational cells thick from the start of
the shock to the C-J thermal choke. The initial pressure rise and heat release is only four computational cells thick.
This may not provide enough resolution to model a smooth and continuous rise in pressure, and still satisfy the
conservation equations. The leading shock is several mean free paths thick and will not be easily approximated by
the current numerical grid. As noted previously, this is a time averaged numerical solution. The averaging will also
smear out the pressure and heat release waves, making the shock thicker than the unsteady time-accurate solution. A
finer grid study in progress may help resolve what is numerical and what is physical. .
The difference between the one-dimensional planar ZND model and simulation will likely never disappear
entirely. For example, the effects of the transverse waves and the multi-dimensional structure in the detonation cells
have no counter part in the ZND model. A first step in closing this gap would be the creation of empirical
proportionality constants. A program of measured sensitivity studies plus experimental data would be required to
determine the range and nature of these factors.

12
American Institute of Aeronautics and Astronautics
 X. Specific Impulse and RDE Thermodynamic Losses.
Comparisons between various cycles generally compare thermal efficiencies. Thermal efficiency does not
always provide a fair metric. In a Brayton cycle, the transfer of work from the turbine to the compressor is not
covered by the overall thermal efficiency, yet this work is subtracted from the turbine expansion. The remaining
expansion in the turbine is what provides useful work. For this reason, this study uses specific impulse as a measure
of comparison.
The numerical study generates 4970 seconds of fuel-based specific impulse. For comparison, the method used
for PDE’s by Heiser and Pratt.16. Using the same parameters, this yielded 5420 seconds and a thermal efficiency of
32% for a cycle that is operating with an overall pressure ratio of 1.4 over atmospheric pressure. A third
calculation12 using an ideal expansion to atmospheric yielded 5230 seconds of optimum specific impulse. All three
results are within ±5%. These results are in rough agreement with other research results8,17.
These results would seem to put performance in the same league with some low bypass turbofans and turbojets.
This is not yet a fair comparison, since many of the parasitic losses common to propulsion systems have not been
accounted for in this modeling. The most obvious omissions are the work necessary for compression at the inlet and
Downloaded by UNIVERSITY OF TENNESSEE on October 26, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2011-803

non-isentropic losses in the expansion.

Shock losses and a mixed thermal cycle account for much of the losses inherent in a RDE. The ideal expansion
Isp was based on an estimation of representative streamline stagnation pressure and temperature at the start of the
isentropic expansion. Until the difference between the ZND model and the simulation are reconciled, the streamlines
present a conservative representation of the expansion properties.
The RDE is a mixed thermodynamic cycle consisting of at least four distinct cycles. The primary cycle consists
of those streamlines which pass through the detonation and expand without encountering the oblique shock. The
ideal expansion for this cycle starts at the upper C-J point and expands isentropically to atmospheric pressure. If this
ideal expansion were the only process representative of the RDE, specific impulse would be closer to 7600 seconds
and hints at the potential for an optimized cycle. However, the major complicating factor is that the primary flow
percentage of total flow is approximately 68%. The expansion regions for the four cycles are shown in Fig. 12.

Figure 12. Stagnation H-S diagram expansion region with adjusted ZND model
If the ZND model in Fig. 6-9 were used as a basis, the primary flow Isp would be 8600 seconds. The sensitivity
of the ZND model is demonstrated by the amount of adjustment required to bring the C-J point in line with the
simulation. Combustion efficiency was reduced from 100% to 89%. Station 3 Mach number was increased from a
streamline average of 0.85 to 1.36, effectively increasing V0 (Fig. 10). Lowering the combustion efficiency cover the
enthalpy discrepancy between the ZND model and simulation. This lowers the max temperature, but also reduces
the entropy generation. Increasing V0 increases the entropy generation by increasing V0W. An increase in the local
swirl would accomplish the same thing.
A portion of the detonation flow passes through the oblique shock wave and emerges with a substantially lower
total pressure and total enthalpy. This flow is approximately 18% of the total. The start of expansion of this flow
13
American Institute of Aeronautics and Astronautics
 happens to coincide with the expansion of the recovery flow, which accounts for 10% of the total flow. Together,
these two flows account for 28% of the total producing an Isp of 72. The boundary flow is a constant pressure heat
addition and is 4% of the total flow. The boundary flow contributes only 61 seconds to the Isp. The total mass
fraction adjusted specific impulse, therefore, is 5230 seconds.
The boundary flow presents a unique challenge. Only one streamline was presented as representative of this
region. Streamlines placed to either side of the midline have shown a wide range in thermodynamic behavior. The
time-averaging approach to the streamline masks its true nature as it moves through the shear layer. The shedding
Kelvin-Helmholtz vortices create a great deal of noise in what is, only on average, an isentropic expansion. It will be
a challenge to fully capture the nature of this stream. Fortunately, it accounts for a small, but not insignificant,
portion of the total flow.
Other challenges are presented. The mass fraction of each flow is not found entirely from the fluid
thermodynamics, since there is dependence on geometry for the oblique detonation and recovery flow dependence
on injection method. The entropy contribution by compression of the inlet gases, injection losses and fill zone
dynamics have not been fully explored. The shear layer behavior is synchronized to the transverse waves of the
Downloaded by UNIVERSITY OF TENNESSEE on October 26, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2011-803

detonation. These ripples create noise in the expansion regions that are, on average, isentropic, but leave other
unresolved issues.
It is clear that the combined effects of the four major flow streams must be included before the ZND model can
be an effective representative of the RDE. A geometry independent model may be constructed from the ZND model.
However, empirical loss factors based on the above discussion must be used to account for the shock and flow
losses. Sensitivity studies are required to explore the limits of any empirical factors.

XI. Conclusion
A simple one-dimensional thermodynamic model of an RDE has been constructed using an approach common to
other cycles. It is clear that the ZND model when combined with the vector sum offers an approximate, but likely
optimistic, model for an RDE. In spite of the simplicity of the modified ZND model and the complexity of the
simulated behavior, the ZND model describes most of the major characteristics of the RDE and is still a good
foundation for a thermodynamic model. Further work is required to close this gap.

14
American Institute of Aeronautics and Astronautics
 References
1
Falempin, F., and LeNaour, B., "R&T Effort on Pulsed and Continuous Detonation Wave Engines," 16th
AIAA/DLR/DGLR International Space Planes and Hypersonic Systems and Technologies Conference American Institute of
Aeronautics and Astronautics, Bremen, Germany, 2009, pp. 1.
2
Voitsekhovskii, B.V., "Maintained detonation," Soviet Physics, Doklady, Vol. 4, No. 6, 1959, pp. 1207-1209.
3
Nicholls, J.A., Cullen, R.E., Adamson Jr., T.C., "The feasibility of a rotating detonation wave rocket motor : final report."
University of Michigan, ORA Report 05179, Ann Arbor, Michigan, 1964, pp. 1-176.
4
Roy, G.D., and Frolov, S.M., "Pulse and continuous detonation propulsion," Torus Press, Russia, 2006, pp. 1-338.
5
Roy, G.D., Frolov, S.M., and Sinibaldi, J., "Pulsed and continuous detonations," Fifth International Colloquium on Pulsed
and Continuous Detonations, Torus Press, Moscow, 2006, pp. 1-349.
6
Lee, J.H.S., "The detonation phenomenon," Cambridge University Press, Cambridge ; New York, 2008, pp. 26-46.
7
Schwer, D.A., and Kailasanath, K., "Numerical Investigation of Rotating Detonation Engines," 46th
Downloaded by UNIVERSITY OF TENNESSEE on October 26, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2011-803

AIAA/ASME/SAE/ASEE Joint Propulsion Conference & Exhibit, AIAA 2010-6880, Nashville, TN, 2010, pp. 1-15
8
Hishida, M., Fujiwara, T., and Wolanski, P., "Fundamentals of rotating detonations," Shock Waves, Vol. 19, No. 1, 2009,
pp. 1-10.
9
Bykovskii, F.A., Zhdan, S.A., and Vedernikov, E.F., "Continuous spin detonations," Journal of Propulsion and Power,
Vol. 22, No. 6, 2006, pp. 1208.
10
Henrick, A.K., Aslam, T.D., and Powers, J.M., "Simulations of pulsating one-dimensional detonations with true fifth
order accuracy," Journal of Computational Physics, Vol. 213, No. 1, 2006, pp. 311-329.
11
Hayashi, A.K., Kimura, Y., Yamada, T., "Sensitivity Analysis of Rotating Detonation Engine with a Detailed Reaction
Model," 47th AIAA Aerospace Sciences Meeting Including The New Horizons Forum and Aerospace Exposition, American
Institute of Aeronautics and Astronautics, El Segundo, CA 90245, 2009, pp. 9.
12
Saad, M.A., "Compressible Fluid Flow, Second Edition," 2 ed., Prentice Hall, Upper Saddle River, New Jersey 07458,
1993, pp. 20, 125, 250.
13
Hill, P.G., and Peterson, C.R., "Mechanics and thermodynamics of propulsion," Addison - Wesley Publ. Co., Readin
Massachusetts [etc.], 1992, pp. 284, 286.
14
Wintenberger, E., and Shepherd, J.E., "Stagnation Hugoniot analysis for steady combustion waves in propulsion
systems," Journal of Propulsion and Power, Vol. 22, No. 4, 2006, pp. 841.
15
Kuo, K.K., "Principles of combustion," Wiley, New York, 1986, pp. 262.
16
Heiser, W.H., and Pratt, D.T., "Thermodynamic cycle analysis of pulse detonation engines," Journal of Propulsion and
Power, Vol. 18, No. 1, 2002, pp. 69-71.
17
Braun, E.M., Lu, F.K., Wilson, D.R., "Detonation Engine Performance Comparison Using First and Second Law
Analyses," 46th AIAA/ASME/SAE/ASEE Joint Propulsion Conference & Exhibit, AIAA, Nashville, TN, 2010, pp. 8.

15
American Institute of Aeronautics and Astronautics