JOURNAL OF SPACECRAFT AND ROCKETS

Vol. 48, No. 2, March–April 2011

Parametric Study of Dual-Expander Aerospike Nozzle

Upper-Stage Rocket Engine

J. Simmons∗ and Richard Branam†

Air Force Institute of Technology, Wright-Patterson Air Force Base, Ohio 45433
DOI: 10.2514/1.51534
The Air Force Institute of Technology is studying a new upper-stage rocket engine architecture: the dual-expander
aerospike nozzle. The goal of this research is to provide the maximum thrust-to-weight ratio in an engine that delivers
a minimum of 50,000 lbf vacuum thrust with a vacuum specific impulse of 464 s. Previous work focused on developing
an initial design to demonstrate the feasibility of the dual-expander aerospike nozzle architecture. That work
culminated in a design exceeding the requirements, delivering an estimated 57,000 lbf thrust with a specific impulse of
472 s by using an oxidizer-to-fuel ratio of 7.03, a total mass flow of 121 lbm=s, and an engine length of 38 in. These
results were computed in a numerical model of the engine. Current work expands the model in preparation for
optimizing its thrust-to-weight ratio. The changes to the model are designed to support running automated
Downloaded by BRISTOL UNIVERSITY on January 27, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.51534

parametric and optimization studies. Parametric studies varying oxidizer-to-fuel ratio, total mass flow, and chamber
length show that a dual-expander aerospike nozzle engine can achieve 50,000 lbf vacuum thrust and 489 s vacuum Isp
with an oxidizer-to-fuel ratio of six, a total mass flow of 104 lbm=s (a reduction of 14%), and an engine length of
27.9 in. (a reduction of over 25%), which should equate to a significant weight savings over the original design.

Nomenclature winit = initial half width of channel, in.

a = half spacing of channel, in. xi = normalized location of the ith station
Acv;i = cross-sectional area of cooling volume at ith
station, in:2
Aflow;i = combusted flow cross-sectional area at ith station, in:2 I. Introduction
Ahx = area of heat exchange for chamber, in:2
Ahxcv;i = area for heat exchange to cooling volume at ith
station, in:2
T HE Air Force Institute of Technology (AFIT) is studying the
dual-expander aerospike nozzle (DEAN) rocket engine archi-
tecture to significantly improve upper-stage rocket performance. The
Ahxnozz;i = area of heat exchange to aerospike nozzle at ith specific performance goals of this research are to provide 50,000 lbf
station, in:2 vacuum thrust with a vacuum Isp of 464 s. The DEAN, shown in
aLH2;i = liquid hydrogen Mach number at ith station Fig. 1, is a liquid hydrogen (LH2)/liquid oxygen (LOX) engine using
aLOX;i = liquid oxygen Mach number at ith station separate expander cycles to pump the fuel and oxidizer and an
F = thrust, lbf aerospike nozzle to improve efficiency [1].
g0 = 32:18 ft=s2 Previous work at AFIT by Martin focused on developing an initial
hi = height of channel at ith station, in. design to demonstrate the feasibility of the DEAN architecture by
Isp = specific impulse, s building a computational model using the Numerical Propulsion
lc = chamber length, in. System Simulation (NPSSTM) software package [1]. That work
li = length of station
0:2 lc , in. culminated in a design exceeding the requirements. This paper
m_ = total mass flow, lbm=s outlines the current work at AFIT to optimize the DEAN’s thrust-
m_ LH2 = liquid hydrogen mass flow, lbm=s to-weight ratio (T=W) by expanding Martin’s model through
m_ LOX = liquid oxygen mass flow, lbm=s simplifying the calculations of the DEAN geometry and adding
mplug = slope of aerospike plug parametric variables to the model to support automated trade studies
n = number of channels in cooling volume in ModelCenterTM. The results of those trade studies indicate the
pi = fluid pressure at ith station, psia DEAN can run at the specified performance using a lower total mass
rci = inner radius of chamber, in. flow and a significantly shorter engine than the previous design.
rci;i = inner radius of chamber at ith station, in.
rco ;i = outer radius of chamber at ith station, in.
rn;i = radius of aerospike nozzle at ith station, in. II. Existing NPSSTM DEAN Model
rn;i1 = radius of aerospike nozzle at, i  1th station, in. The original NPSS DEAN model was prepared by Martin as part
rti = inner radius at throat, in. of his masters thesis at AFIT [1]. This model uses NPSS elements to
Ti = fluid temperature at ith station, R represent the pumps, turbines, pipes, valves, chamber, nozzle, and
Vcv;i = volume of cooling volume at ith station, in:3 cooling channels to accurately model the DEAN architecture. The
wi = half width of channel at ith station, in. results of Martin’s analysis converge to a working design, validating
the DEAN concept. The DEAN thrust, Isp , and T=W exceed both its
requirements and the performance of the upper-stage single-
Received 12 July 2010; revision received 1 November 2010; accepted for expander cycle based RL10A-3 rocket engine, as can be seen in
publication 1 November 2010. This material is declared a work of the U.S. Table 1.
Government and is not subject to copyright protection in the United States.
This excess in thrust performance suggests the dual-expander
Copies of this paper may be made for personal or internal use, on condition
that the copier pay the $10.00 per-copy fee to the Copyright Clearance Center,
cycle offers an advantage in thrust over the single-expander cycle,
Inc., 222 Rosewood Drive, Danvers, MA 01923; include the code 0022-4650/ and it indicates room to optimize the DEAN design further. Before
11 and $10.00 in correspondence with the CCC. discussing the changes made to Martin’s model [1], necessary to
∗ facilitate this optimization, it is useful to review the key elements of
Ph.D. Student, Aero/Astro Engineering. Student Member.
†
Assistant Professor, Aero/Astro Engineering. Associate Fellow. the DEAN architecture and Martin’s model.
355
 356 SIMMONS AND BRANAM

second, physically separate cooling loop from the chamber for use in
the second expander cycle. This second cooling loop increases the
surface area inside the chamber used to drive the turbomachinery,
providing for correspondingly increased power to the pumps and
increased chamber pressure and, in turn, increased engine per-
formance [1]. The aerospike nozzle is also shorter and lighter than a
traditional bell nozzle, leading to improved packaging and T=W.
Finally, the DEAN architecture is a forerunner to similar boost stage
architecture, where the aerospike nozzle’s global performance will
result in even greater gains than in an upper-stage engine.

Fig. 1 Original DEAN geometry. B. NPSSTM Model Details

NPSS is a computer simulation tool for modeling aircraft and
rocket engines. Engine simulations built in NPSS provide higher
A. Existing DEAN Architecture fidelity results than engine cycle studies. NPSS has been developed
The DEAN uses two novel design choices. The first is the use of by the NASA Glenn Research Center, with assistance from the
separate expander cycles for the fuel and the oxidizer. In a traditional aerospace propulsion industry. Models built in NPSS consist of a
expander cycle, the fuel is pumped through a cooling jacket for the series of interconnected software objects representing the compo-
chamber and nozzle. The heat picked up by the fuel from cooling nents of the engine under consideration. The object connections are
Downloaded by BRISTOL UNIVERSITY on January 27, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.51534

the chamber and nozzle is then used to drive the turbine that runs both made using NPSS application programming interfaces modeling
the fuel and oxidizer pumps, before the fuel is introduced into the fluid flows, mechanical connections, and thermal flows called ports.
chamber [3]. A built-in solver in NPSS can then be used to drive the model design
Figure 2 shows the DEAN’s two separate expander cycles. The variables to balance the fluid flows, mechanical connections, and
fuel loop is on the left. Here, the LH2 flows from the fuel tank into the thermal flows in the model and converge on a design point [5].
first fuel pump. Half of the LH2 then flows through a bypass, while Martin’s model [1] of the DEAN engine includes elements to
the remainder is fed into a second fuel pump. This flow then simulate the various components of the engine. The combustion
continues into the aerospike nozzle, where it absorbs heat while chamber is modeled using a RocketComb1 element, which requires
cooling the nozzle. The heated flow then drives the fuel turbine before the chamber radius and volume. The RocketComb element also
joining with the bypass flow and entering the combustion chamber includes ThermalOutputPorts to model heat exchange to the fuel and
through the injector. On the right is the oxidizer loop. In this loop, the oxidizer. The ThermalOutputPorts require the radius at the port’s
LOX flows from the oxidizer tank into the single oxidizer pump. This location, the cross-sectional area of the combusting flow in the
flow then enters the cooling jacket for the combustion chamber, chamber, and the surface area of the portion of the chamber in contact
where it absorbs heat while cooling the chamber. The heated oxygen with the combusting flow. The model calculates the heat exchange
then flows toward the oxidizer turbine, with a small amount (roughly between the chamber walls for the oxidizer expander cycle and the
10%) sent around the turbine using a bypass and the remainder used internal portion of the aerospike for the fuel expander cycle. The
to drive the turbine. Finally, the oxygen flows are joined before aerospike nozzle is modeled using a RocketNozzle element, which
entering the combustion chamber through an injector [1]. requires the throat area and the expansion ratio and includes more
The two separate expander cycles present benefits and challenges. ThermalOutputPorts for use in the heat exchange between the
Increased chamber pressures and, in turn, engine performance can be aerospike and the fuel expander cycle.
obtained by separating the power demands between two separate The fuel and oxidizer tanks are modeled using Starter elements.
turbines driven by separate working fluids. The separate expander The plumbing connecting the engine components to the tanks and
cycles also ensure the fuel and oxidizer remain physically separated each other is modeled using Valve04 elements that require the cross-
until entering the combustion chamber, eliminating one of the more sectional area to model the pressure drop in the plumbing and
catastrophic failure modes in traditional expander cycles, namely, customized CoolingVolume elements, which require the cross-
failure of an interpropellant seal [1]. However, the LOX cycle sectional area and fluid volume to model the heat loss in the lines. The
requires a turbine material that will work well in an oxygen cooling jackets around the combustion chamber and the aerospike
environment. Additional research at AFIT showed that Inconel 718 nozzle are modeled using the ThermalOutputPorts in the chamber
provides both satisfactory oxygen resistance and suitable mechanical and nozzle and the plumbing elements connected by Wall2 elements.
performance for use in both the pump and the turbine in the LOX The pumps are modeled using customized Pump elements, which
cycle [4]. require the efficiency, pressure ratio, and mass flow of each pump.
The second novel design choice is the use of an aerospike, or plug, Finally, the turbines are modeled using customized Turbine
nozzle. Traditional bell nozzles are optimized to perform at a single elements, which require the efficiency, pressure ratio, mass flow, and
pressure or altitude. This optimization makes them very efficient at the cross-sectional area of the flow.
their design point but much less so for the remainder of the flight. While the cooling jackets in the DEAN are continuous volumes,
Aerospike nozzles on the other hand, due to their geometry, adapt the NPSS model represents them as a series of eight discrete stations.
their performance to the current altitude. Even though aerospike Figure 3 shows the locations of these stations in the model. The
nozzles perform less efficiently than bell nozzles at the design chamber (24 in. long in Martin’s design [1]) consists of five equally
altitude of a given bell nozzle, when considering the altitudes seen by spaced sections, with stations (represented by stars) at the midpoint
a launch vehicle, the global performance of an aerospike nozzle of each section plus a station at the throat of the rocket. Note that the
exceeds the global performance of a bell nozzle, making the oxidizer loop (the outer wall of the chamber) and the fuel loop (the
aerospike nozzle an attractive design for launch vehicles. More aerospike) are represented by separate sets of stations. The external
important, for the DEAN, the use of an aerospike nozzle provides a portion of the aerospike (14 in. long in Martin’s design) has two

Table 1 Existing DEAN design results versus requirements

DEAN [1] Requirement [1] Percent difference, % RL10A-3-3A [2] Percent difference, %
Vacuum thrust 57,231 lbf 50,000 lbf 14:5 16,500 lbf 247
Vacuum Isp 472 s 464 s 1:7 443 s 6:5
T=W 119 106 12 53 124
 SIMMONS AND BRANAM 357
Downloaded by BRISTOL UNIVERSITY on January 27, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.51534

Fig. 2 DEAN architecture.

NPSS dependent variables and component inputs form the set of

user-specified design choices or design variables. The NPSS inde-
pendent variables and component outputs form the set of user-
specified responses or response variables. In the previous model, the
design variables included the user-specified dependent variables (the
pressure ratios for the three pumps) and the component inputs (for
example, the volume of the combustion chamber and the nozzle
expansion ratio). The response variables include the user-specified
Fig. 3 DEAN geometry (dimensions in inches) (credit Martin [1]).
independent variables (the efficiencies of the first LH2 pump and of
the LOX pump) and the component outputs (for example, the
additional stations in the first half of the nozzle that are only in the vacuum thrust and Isp ). Table 2 summarizes the design and response
fuel loop. variables of the previous model.
NPSS models have four classes of variables: the dependent Follow-on research at AFIT focusing on designs for the LH2 and
variables the NPSS solver will ensure equal specified values, the LOX pumps provides verification for the NPSS model of the DEAN.
independent variables the NPSS solver is free to adjust as necessary In each design study, the respective expander cycle is modeled using
to converge the design, component inputs read by the NPSS solver spreadsheet calculations and Pumpal®/RITAL®. These additional
but not altered by it, and component outputs calculated by the NPSS models agree well with the NPSS model. For example, the Pumpal
solver but not driven toward a particular value. Taken together, the model of the LH2 pumps shows the required power of the first pump

Table 2 Existing DEAN design point [1]

Design variables Response variables
LOX pump pressure ratio, PRp;LOX 103 Vacuum thrust (Fvac 57,000 lbf
LH2 pump 1 pressure ratio, PRp;LH2;1 45 Vacuum Isp 472 s
LH2 pump 2 pressure ratio, PRp;LH2;2 2 Total mass flow, m
_ 121 lbm=s
LH2 pump 2 efficiency,
p;LH2;2 0.83 O=F 7.03
Chamber length, lc 24 in. Chamber pressure, P0 1739 psia
Outer chamber radius, rco 6 in. Chamber temperature, T0 6586 R
Inner chamber radius, rci 2 in. LOX pump efficiency,
p;LOX 0.66
Chamber volume, Vc 2075 in:3 LOX pump power 2587 HP
Throat area, At 15:9 in:2 LH2 pump 1 efficiency,
p;LH2;1 0.67
Nozzle length, ln 14 in. LH2 pump 1 power 2527 HP
Expansion ratio,  125 LH2 pump 2 power 1046 HP
LOX turbine pressure ratio, PRt;LOX 1.82
LOX Turbine efficiency,
t;LOX 0.949
LOX turbine power 2587 HP
LH2 turbine pressure ratio, PRt;LH2 1.84
LH2 turbine efficiency,
t;LH2 0.9
LH2 turbine power 3573 HP
LOX temperatures, TLOX;i 179–617 R
LOX pressures, PLOX;i 3810–4500 psia
LOX max Mach number, Mmax;LOX 0.56
LH2 temperatures, TLH2;i 145–610 R
LH2 pressures, PLH2;i 3670–4000 psia
LH2 max Mach number, Mmax;LH2 0.67
 358 SIMMONS AND BRANAM

is 2523 HP (less than 1% difference from the NPSS value), and the
required power of the second pump is 1079 HP (approximately 3%
difference from the NPSS model). Furthermore, the pressure coming
out of the second pump is 4050 psia versus 4000 psia in the NPSS
model (approximately 1% difference) [6]. The LOX pump values
show similar agreement. For example, the Pumpal model of the LOX
pump shows the required power is 2215 HP (approximately 12%
difference from the NPSS model), and the pressure coming out of the
LOX pump is 4600 psia versus 4500 psia in the NPSS model
(approximately 2% difference) [4].
Given the importance of balancing the heat and mechanical flows
in an expander cycle, it is worth noting the performance of the
turbomachinery in existing DEAN design. Naturally, the converged
design point balances the power required by the pumps with that
provided by the turbines for each of the two expander cycles. While
the performance of the turbomachinery may appear high when
compared with RL10A, which provides only 789 HP from its
turbine, Pratt and Whitney built an LH2 expander cycle that produces
5900 HP as part of an upper-stage demonstrator engine [6]. This puts
the LH2 cycle well within demonstrated capability. As for the LOX
Downloaded by BRISTOL UNIVERSITY on January 27, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.51534

cycle, research at AFIT has proposed designs for the LOX pump and
turbine that provide the required performance as a part of this
architecture [4].

III. Parametric NPSSTM DEAN Model

The goal of the current research at AFIT is to optimize the DEAN
engine to provide the maximum T=W, while maintaining the
required thrust and Isp , and not violating operational limits for the
materials and working fluids. The T=W can be maximized by either Fig. 5 Comparison of simplified geometry to Martin’s [1] original
increasing the thrust or decreasing the weight. Since the DEAN has a geometry.
fixed thrust requirement, the optimization problem can be restated as
optimizing the DEAN engine to provide the minimum weight, responses. To use the model to optimize the design, the NPSS model
subject to the previously mentioned constraints. Identifying the must be altered to use the chamber and nozzle geometries and the
design variables driving the weight of the engine and updating the local mass flows to the turbomachinery as design variables. This
model to use those variables as the design variables enables auto- section details the modifications made to the previous model to
mated searching over the design trade space for optimal designs. convert it into a parametric model based on the design variable needs
The rocket engine chamber, nozzle, and turbomachinery are the and two changes made while the NPSS model was being modified to
elements used to estimate the total mass (weight) of the rocket engine reflect the most recent DEAN architecture being considered.
during conceptual design. The chamber and nozzle masses are driven
by their geometry and material, while the turbomachinery masses are
driven by their local mass flows [3]. The model previously captured A. Updated DEAN Architecture
the geometry of the DEAN as hard-coded values, tying the model to a Before modifying the DEAN model to support parametric trade
single design point. Additionally, the model treated the mass flows as studies, two architecture changes were made to the fuel expander

Fig. 4 Updated DEAN architecture.

 SIMMONS AND BRANAM 359

Table 3 Parametric study over LH2 pump 1 pressure ratio

Design variables Response variables
LOX pump pressure ratio, PRp;LOX 103 Vacuum thrust, Fvac 49,500–60,000 lbf
LH2 pump 1 pressure ratio, PRp;LH2;1 30–60 Vacuum Isp 416–491 s
LH2 pump 2 pressure ratio, PRp;LH2;2 2 Total mass flow, m
_ 119–132 lbm=s
LH2 pump 2 efficiency,
p;LH2;2 0.83 O=F 4.2–11.3
Expansion ratio,  125 Chamber pressure, P0 1500–1940 psia
Chamber temperature, T0 5000–6660 R

cycle to reflect refinements in the design. Figure 4 shows this updated q








































DEAN architecture. The first change was to remove the LH2 bypass, Ahxnozz ;i
  rn;i  rn;i1   rn;i  rn;i1 2  l2i (5)
feeding the cooling jacket for the aerospike with the full mass flow of
the LH2 instead of only half. The second change was to require the
two fuel pumps to run at the same pressure ratio to minimize the n
round  r=winit  a (6)
maximum pressure ratio required for the fuel pumps.

B. NPSSTM Model Details

wi
  r=nchannels  a (7)
The modifications to convert the NPSS DEAN model into a
Downloaded by BRISTOL UNIVERSITY on January 27, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.51534

parametric model involved the chamber and nozzle geometry and the
mass flows through the turbomachinery. The system-level param-
eters chosen to drive the chamber and nozzle geometries are the hi
2  wi  AR (8)
chamber length, the inner and outer chamber radii, and the shape of
the nozzle. A number of intermediate geometric values are deter-
mined by the shape of the nozzle. To facilitate calculating these
values, the geometry of the aerospike and cooling volumes were Ahxcv;i
2  wi  n  li (9)
simplified to linear approximations from their higher-order calcu-
lations (the aerospike radii are calculated using the method of
characteristics [1]). Figure 5 compares the results of these simpli-
fications to the original geometry. Acv;i
2  wi  hi  n (10)
The calculations based on these simplifications include the radius
of the aerospike at each station [Eq. (2)], the combusted flow cross-
sectional area [Eq. (3)], the surface area of the heat exchanges for the Vcv;i
Acv;i  li (11)
chamber [Eq. (4)], the aerospike nozzle [Eq. (5)] and cooling
volumes [Eq. (9)], the cross-sectional area [Eq. (10)], and the volume The system-level parameters chosen to drive the mass flows to the
of the fluid in the cooling volumes [Eq. (11)]. These values are turbomachinery are the O=F ratio and total mass flow. These
calculated from the inner and outer radii of the chamber and the parameters also drive the performance constraints of thrust and Isp .
throat, the chamber and aerospike lengths, the cooling channel half Equation (12) [3] determines the LOX mass flow from these system-
widths and half spacings, and the aspect ratios (ARs): level parameters, and Eq. (13) [3] determines the LH2 mass flow. In
mplug
rti  rci =0:6  lc  (1) the original model, the O=F ratio can only be controlled indirectly by
changing the pressure ratios of the pumps, leading to coupling effects
appearing in the O=F plots. For example, a parametric study,
rci ;i
rci  mplug  xi  lc (2) described in Table 3, varied the pressure ratio of the first LH2 pump:
O=F
_ LOX

m m
_ (12)
1  O=F
Aflow;i
  rc2o ;i  rc2i ;i  (3)

1
_ LH2

m m
_ (13)
Ahx
2    rco ;i  li (4) 1  O=F

Fig. 6 Isp vs O=F, original NPSS model.

 360 SIMMONS AND BRANAM

Fig. 7 Thrust vs O=F, original NPSS model.

Downloaded by BRISTOL UNIVERSITY on January 27, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.51534

Figure 6 shows the relationship between Isp and O=F for the DEAN defined dependent variables were changed to the total mass flow and
based on this parametric study. The peak Isp is at an O=F ratio of an equation setting the two fuel pumps to the same pressure ratio. The
approximately six. Ordinarily, this result would be produced by a user-defined independent variables were changed to the throat area
parametric study as a function of the independent variable, O=F, and to the pressure ratios for all three pumps.
implying the total mass flow was constant. Equation (14) [3] supports One final change was made to the NPSS model. In the model, the
the conclusion that thrust versus O=F would follow the trend in mass flow through the oxidizer bypass was hard coded to a specific
Fig. 6. However, Fig. 7 shows a peak at an O=F of six and then a rise value, resulting in instabilities in the model when the system mass
toward another peak for O=F < 5: flow or the O=F ratio changed. The oxidizer mass flow through the
bypass was set to be 10% of the oxidizer mass flow, and the mass flow
F to the oxidizer turbine was set to 90%. These percentages were also
Isp
(14) made user configurable.
_ 0
mg

The deviation from the describing equations rests with the

assumption of constant total mass flow. Figure 8 is a three- IV. System-Level DEAN Model
dimensional plot showing thrust versus both O=F and total mass flow. A system-level model of the DEAN engine was built around the
The vertical axis is the thrust, the left axis is O=F, and the back axis is parametric NPSS model in order to run automated trade studies. This
the total mass flow. The starred point is O=F equals six. The total system-level DEAN model was built in ModelCenterTM. Model-
mass flow has a sharp turn as O=F decreases beyond that point, Center is a multidisciplinary modeling environment used to study
leading to the second increase in thrust as total mass flow reaches its the trade space of a design and optimize that design. ModelCenter
peak during the parametric study. To eliminate coupling effects like can combine analyses developed in a variety of tools, including
this example, and an instability in the LH2 efficiencies, the user- MATLAB, Mathcad, Excel, and command line executables (for

Fig. 8 Thrust vs O=F and total mass flow, original NPSS model.
 SIMMONS AND BRANAM 361
Downloaded by BRISTOL UNIVERSITY on January 27, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.51534

Fig. 9 System-level model of DEAN.

example) into a single system-level model. Users can then use the system-level design variables. They are the total mass flow, the O=F
built-in trade study and visualization tools to conduct sensitivity ratio, the inner chamber radius, the outer chamber radius, the
studies to gain insight into what aspects of the design are the key chamber length, and a guessed value for the throat area. The second
drivers influencing the desired results. ModelCenter also provides a component calculates the nozzle radius using Eq. (3), a nozzle length
rich set of built-in optimization tools used to find values for the key based on the slope of the aerospike nozzle, and the chamber volume
drivers optimizing the design for a given goal (such as minimizing the using volume equations for cylinders and cones from the design
mass of the DEAN engine) while ensuring constraints are not variables. The third component runs the NPSS DEAN model using
violated (such as maintaining the required thrust and Isp ). These the values of the design variables and the values calculated in the
optimization tools include a gradient optimizer, a genetic optimizer CalculateGeometry component. Since the throat area is an inde-
called DarwinTM, and an optimizer that uses surrogate models of the pendent variable in the NPSS model, the NPSS solver may select a
design space during optimization called DesignExplorerTM [7]. value that does not match the guessed value. If the model chooses
Figure 9 shows the system-level DEAN model in the ModelCenter such a value, the geometry values calculated from the guessed throat
interface. The model consists of six components and two data display area (specifically, the nozzle radius and cooling volume sizes) are
elements. Table 4 summarizes these components. incorrect. The fourth component addresses this problem using a
The first four components in the DEAN system-level model are direct substitution iteration solver. The solver directs the system-
used to run the DEAN NPSS model. The first component stores the level model to be run in a loop, substituting the NPSS calculated
throat area for the guessed throat area until the two are the same value,
closing the design.
Table 4 System-Level DEAN Model Components The remaining two components in the DEAN system-level model
are used for postprocessing the DEAN NPSS model. The fifth com-
Label Name Description ponent is actually four components that are part of a ModelCenter
1 Design A component storing the system-level assembly component. ModelCenter assembly components allow
variables design variables model developers to group related components together to simplify
2 Calculate A script component calculating the geometry the display and management of the model. Three of the four
geometry parameters for the DEAN model from the subcomponents in the geometry component are built-in Model-
design variables Center geometry primitives: two surfaces of revolution to render the
3 DEAN The parametric NPSS DEAN model imported into
the ModelCenter framework using the
chamber, the aerospike, and a 12-in.-long arrow to provide a scale for
embedded Quick Wrap tool the DEAN geometry. The final subcomponent is a ModelCenter
4 Converger A direct substitution convergence loop used to script component transforming the output from the DEAN NPSS
converge the guessed throat area to the model into geometry strings that define the curves for the surface of
calculated throat area revolution components.
5 Geometry A collection of geometry primitives (surfaces of The last analysis component calculates the Mach numbers for the
revolution and an arrow) and a script component LOX and LH2 fluid flows in the cooling volumes. These values are
to render the live view of the DEAN geometry critical constraints on the engine design, as the flows cannot be
seen in Eq. (8) allowed to reach sonic conditions. Previous constraints for these
6 Mach A script component that uses polynomial
numbers interpolation to calculate the Mach numbers for
values were a maximum Mach number in the LOX flows of 0.6 and
the LOX and LH2 fluid flows in the cooling a maximum Mach number in the LH2 flows of 0.9 [1]. Martin [1]
volumes calculated the Mach numbers at each station by looking up the
7 Data monitor A built-in ModelCenter component set to display speed of sound for the fluids in the Thermophysical Properties of
the key performance values from the DEAN Fluid Systems online handbook‡, from the National Institute of
model
8 Geometry A live preview of the DEAN geometry based on
‡
rendering the geometry components in Eq. (5) Available at http://webbook.nist.gov/chemistry/fluid/ [retrieved
27 May 2010].
 362 SIMMONS AND BRANAM

Table 5 DOEs over O=F, total mass flow, and chamber length
Design variables Response variables
O=F 5.5–7.5 Vacuum thrust, Fvac 39,700–69,500 lbf
Total mass flow, m_ 85–140 lbm=s Vacuum Isp 466–499 s
Chamber length, lc 22–30 in. Chamber pressure, P0 1798–3243 psia
Outer chamber radius, rco 6 in. Chamber temperature, T0 6130–6850 R
Inner chamber radius, rci 2 in. Throat area, At 7:2–17:5 in:2
Expansion ratio,  125 Nozzle length, ln 13.7–14.8 in.
LOX bypass, % 90 Chamber volume, Vc 1810–2480 in:3
LOX pump efficiency,
p;LOX 0.773 LOX temperatures, TLOX 162–886 R
LOX turbine efficiency,
t;LOX 0.949 LOX pressures, pLOX 2865–5864 psia
LH2 pump 1 efficiency,
p;LH2;1 0.8 LH2 temperatures, TLH2 73–634 R
LH2 pump 2 efficiency,
p;LH2;2 0.83 LH2 pressures, pLH2 3220–10,750 psia
LH2 pump efficiency,
p;LH2 0.9

Table 6 Data ranges for speed-of-sound tables DEAN model. Table 5 summarizes the design of experiments
Fluid Temperatures, R Pressure range, psia,
(DOEs) trade study used to establish these ranges.
step size
100 psia To create the interpolation polynomials, a set of isothermal speed-
Downloaded by BRISTOL UNIVERSITY on January 27, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.51534

of-sound versus pressure tables for both oxygen and hydrogen were
LOX 150, 275, 400, 525 2800–6000 downloaded from the Thermophysical Properties of Fluid Systems
650, 775, 900
online handbook. Table 6 shows the temperatures and pressures
LH2 70, 165, 260, 355 3200–7200
450, 545, 640 selected for these tables. Note, the LH2 pressure range was truncated
after determining pressures above 7200 psia represented designs no
longer being considered.
The speed-of-sound tables were then imported into ModelCenter
Standards and Technology, based on the fluid temperatures and for surface fitting using the data import plug in. The interpolation
pressures. The Mach number component automates this process by polynomials were then created by running a cubic fit to the data
using third-order polynomials to interpolate the speed of sound for in ModelCenter’s response surface modeling (RSM) toolkit.
LOX and LH2 over the temperature and pressure ranges seen in the Equation (15) shows the polynomial generated for the speed of sound
in LOX, and Eq. (16) shows the polynomial generated for the speed
of sound in LH2. Both equations are good fits to the data, as can be
seen from their adjusted R2 values and a graphical comparison with
the results from the earlier design. Equation (15) has an adjusted R2
value of 0.995, and Eq. (16) has an adjusted R2 value of 0.988.
Figure 10 shows that Mach numbers calculated from both equations
also compare favorably to results from the earlier design of the
DEAN:

aLOX;i
6009:519  25:08448Ti  0:05203138pi

 0:03725712Ti2  0:0004540472Ti pi  1:669077
 1005 Ti3  5:292126  1007 pi Ti2 (15)

aLH2;i
4014:306  15:03106Ti  0:7100687pi  0:04704292Ti2

 2:660187  1005 p2i  0:0009330969Ti pi  3:34523
 1005 Ti3  4:862856  1008 Ti p2i (16)

V. Results and Analysis

The analysis of the system-level DEAN model took place in two
phases. First, parametric studies were run over the chamber length,
the O=F ratio, and the total mass flow to establish the boundaries of
the design trade space. The results of the parametric studies were then
used to scale the original DEAN design to more closely match the
performance goals.

A. Parametric Studies
Table 7 summarizes the first parametric study of the DEAN
system-level model, which varied the O=F ratio from 5.5 to 7.5.
Initial results from this trade study are promising. The plot of Isp
versus O=F in Fig. 11 shows the peak Isp occurs at an O=F ratio of
approximately six, just as in the previous study shown in Fig. 6.
However, the maximum LH2 Mach numbers exceed the constraint
value established by Martin [1] across the entire parametric study, as
Fig. 10 Comparison of fluid Mach number calculations. can be seen in Fig. 12. A review of the LH2 Mach numbers at all of
 SIMMONS AND BRANAM 363

Table 7 Parametric over O=F

Design variables Response variables
O=F 5.5–7.5 Vacuum thrust, Fvac 56,700–60,000 lbf
Total mass flow, m
_ 121:25 lbm=s Vacuum Isp 468–495 s
Chamber length, lc 24 in. Chamber pressure, P0 2135–2415 psia
Outer chamber radius, rco 6 in. Chamber temperature, T0 6180–6705 R
Inner chamber radius, rci 2 in. Throat area, At 12:1–12:8 in:2
Expansion Ratio,  125 Nozzle length, ln 14.2–14.3 in.
LOX bypass, % 90 Chamber volume, Vc 1576–1579 in:3
LOX pump efficiency,
p;LOX 0.773 LOX temperatures, TLOX 166–657 R
LOX turbine efficiency,
t;LOX 0.949 LOX pressures, pLOX 3860–4380 psia
LH2 pump 1 efficiency,
p;LH2;1 0.8 Max LOX Mach number 0.40–0.41
LH2 pump 2 efficiency,
p;LH2;2 0.83 LH2 temperatures, TLH2 85–436 R
LH2 pump efficiency,
p;LH2 0.9 LH2 pressures, pLH2 4590–8340 psia
Max LH2 Mach number 0.99–1.20

the stations reveals the Mach numbers are within acceptable limits for Figure 13 is the plot of Isp versus O=F ratio for the updated inner
nozzle radii greater than 3.3 in. To test this conclusion, the inner chamber radius. The added detail in this plot shows the O=F ratio for
Downloaded by BRISTOL UNIVERSITY on January 27, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.51534

chamber radius was increased from 2.0 to 3.3 in. However, the NPSS maximum Isp is between 5.9 and 6.1, higher than might be expected
model could not converge, because some of the LH2 cooling vol- (just based on the O=F ratio), given the RL10 upper-stage engine runs
umes had become too large. Addressing this result, several cooling at an O=F ratio of 5.0 [3]. However, while the total mass flow and
volume AR calculations were adjusted until the model converged. As most of the geometric parameters are constant in Fig. 13, the chamber
expected, the maximum LH2 Mach number for this new design fell pressure and temperature are not. Figure 14 shows how chamber
within the constraint value of 0.9. Running the parametric study with pressure varies with O=F, and Fig. 15 shows how the chamber
the adjusted ARs resulted in maximum LH2 Mach numbers between temperature varies with O=F. The chamber pressures seen in Fig. 14
0.81 and 0.88, enabling continued exploration of the DEAN trade range from 2625 to 2875 psia. These values are considerably higher
space. than the 640 psia of the RL10B-2 [3], or even the practical limits of a

Fig. 11 Isp vs O=F ratio, system-level model.

Fig. 12 Fluid mach numbers vs O=F ratio, system-level model.

 364 SIMMONS AND BRANAM

Fig. 13 Detailed Isp vs O=F ratio, system-level model.

Downloaded by BRISTOL UNIVERSITY on January 27, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.51534

single-expander cycle LOX/LH2 engine, which Krach and Sutton chamber pressure, and chamber temperature, in a combined plot to
determined to be between 1375 and 2300 psia [8]. This dramatic present a unified view of their interactions. The O=F ratio is on the
difference in chamber pressures suggests higher chamber pressures left axis, the chamber pressure is on the rear axis, and Isp is on the
are possible with the dual-expander cycle of the DEAN architecture vertical axis. Finally, the chamber temperature is shown as the shades
than with single-expander cycles. of the data points. This figure shows the maximum Isp not only
Figure 16 uses a three-dimensional plot with varying shades of happens at an O=F of six but also at the maximum chamber pressure
gray on the data points to present all four of these variables, Isp , O=F, and somewhere in the middle of the chamber temperature range seen

Fig. 14 Detailed chamber pressure vs O=F ratio, system-level model.

Fig. 15 Detailed chamber temperature vs O=F ratio, system-level model.

 SIMMONS AND BRANAM 365
Downloaded by BRISTOL UNIVERSITY on January 27, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.51534

Fig. 16 Isp vs O=F ratio, chamber pressure, and chamber temperature, system-level model.

in this study. Returning to Fig. 15 reveals the chamber temperature Table 8 summarizes the second parametric study run of the DEAN
for the O=F where the maximum Isp occurs is located at a knee in the system-level model, which varied the total mass flow from 85 to
temperature curve. The trends evident in Fig. 16 suggest the solution 122 lbm=s. As one would expect, thrust is linearly related to total
is simply one possible curve on a solution surface space. mass flow (Fig. 17). A total mass flow of only 104 lbm=s is sufficient

Table 8 Parametric over total mass flow

Design variables Response variables

O=F 7 Vacuum thrust, Fvac 40,800–58,400 lbf
Total mass flow, m_ 85–122 lbm=s Vacuum Isp 479–481 s
Chamber length, lc 24 in. Chamber pressure, P0 2525–2710 psia
Outer chamber radius, rco 6 in. Chamber temperature, T0 6770–6790 R
Inner chamber radius, rci 2 in. Throat area, At 7:8–10:4 in:2
Expansion ratio,  125 Nozzle Length, ln 14.5–14.8 in.
LOX bypass, % 90 Chamber volume, Vc 1560–1578 in:3
LOX pump efficiency,
p;LOX 0.773 LOX temperatures, TLOX 165–766 R
LOX turbine efficiency,
t;LOX 0.949 LOX Pressures, pLOX 3650–4850 psia
LH2 pump 1 efficiency,
p;LH2;1 0.8 Max LOX Mach number 0.35–0.39
LH2 pump 2 efficiency,
p;LH2;2 0.83 LH2 temperatures, TLH2 75–645 R
LH2 pump efficiency,
p;LH2 0.9 LH2 pressures, pLH2 3850–5900 psia
Max LH2 Mach number 0.87–0.91

Fig. 17 Vacuum thrust vs total mass flow, system-level model.

 366 SIMMONS AND BRANAM

Table 9 Parametric over chamber length

Design variables Response variables
O=F 7 Vacuum thrust, Fvac 57,100–58,200 lbf
Total mass flow, m_ 121:25 lbm=s Vacuum Isp 471–480 s
Chamber length, lc 14–26 in. Chamber pressure, P0 950–2975 psia
Outer chamber radius, rco 6 in. Chamber temperature, T0 6435–6805 R
Inner chamber radius, rci 2 in. Throat area, At 9:8–28:9 in:2
Expansion ratio,  125 Nozzle length, ln 12.4–14.6 in.
LOX bypass, % 90 Chamber volume, Vc 1000–1700 in:3
LOX pump efficiency,
p;LOX 0.773 LOX temperatures, TLOX 160–780 R
LOX turbine efficiency,
t;LOX 0.949 LOX pressures, pLOX 2300–5200 psia
LH2 pump 1 efficiency,
p;LH2;1 0.8 Max LOX Mach number 0.32–0.42
LH2 pump 2 efficiency,
p;LH2;2 0.83 LH2 temperatures, TLH2 70–610 R
LH2 pump efficiency,
p;LH2 0.9 LH2 pressures, pLH2 2450–6250 psia
Max LH2 Mach number 0.86–0.97

to achieve 50,000 lbf thrust. Optimizing the DEAN design near thrust and 489 s Isp . Table 10 summarizes this new design, and
104 lbm=s will result in another weight savings over the original Fig. 19 shows this new, much smaller design next to the original
Downloaded by BRISTOL UNIVERSITY on January 27, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.51534

design. design.
Table 9 summarizes the final parametric study run on the DEAN The performance values of this design are favorable. The
system-level model, considering the chamber length from 14 to 26 in. 50,900 lbf thrust not only more closely matches the design require-
The curve in Fig. 18 showing thrust versus chamber length has two ments, but it also falls within the demonstrated 50,300 lbf of
transitions: one at 14.25 in. and one at 22.75 in. The shorter of these expander cycle engines, as seen in the upper-stage demonstrator
chamber lengths offers substantial weight savings while producing engine [6]. The lengths of the engine chamber and nozzle are
nearly the same thrust. This result supports selecting a chamber not significantly smaller than the previous design, indicating the weight
longer than 14.25 in., but further optimization is warranted. Selecting is similarly reduced. The converged values for the turbomachinery
a chamber length of 14.25 in. results in a total engine length of are equally promising. The power required by the pumps is properly
27.85 in. The new length is nearly 25% shorter than the original balanced by power generated in the turbines; in the LOX cycle, this
design and 69% shorter than the RL10B-2 [9]. This result suggests power is 40% of original design, while in the LH2 cycle, the power is
that the DEAN architecture can be packaged in a considerably only increased by 8% over the original design. Together with the
smaller volume than existing upper-stage engines. reduced mass flows and the pressure ratios, the turbomachinery of the
scaled DEAN design has both lighter and more robust options over
the original design.
B. Scaling the DEAN Engine The results also indicate opportunities for improvement, both in
The previous parametric studies examined how three parameters, areas where requirements are still exceeded by the design, including
the O=F ratio, the total mass flow, and the chamber length, affect the T=W, Isp , and chamber pressure, and in areas where there are
weight and performance of the DEAN. Those three parameters will concerns including fluid Mach numbers and wall temperatures. The
be the design variables to be modified in scaling the DEAN to the original design of the DEAN had an estimated weight of 479 lbm [1].
50,000 lbf thrust target. The parametric studies suggest values for This gives an upper limit for the scaled DEAN T=W of 106, meeting
each of these parameters. As noted earlier, the engine provides or exceeding the required T=W. The vacuum Isp is now over 5%
the greatest Isp at O=F
6 for these parametric choices. Similarly, higher than the requirement, suggesting the nozzle can be modified to
the study indicates a mass flow of approximately 104 lbm=s save weight and meet the Isp requirement. The chamber pressure is
achieves the design vacuum thrust of 50,000 lbf. This result is for an 25% lower than the previous design, which leads to a more robust
engine operating at O=F
7. With an O=F of six, the thrust will chamber design. Turning to areas of concern, at least one constraint,
likely be higher. Finally, the results from the chamber length the maximum LH2 Mach number, is still higher than desired.
parametric study suggest a chamber length of 14.25 in. is an Additionally, the wall temperatures, which must stay within their
acceptable value. Entering these values into the system-level model material limits, still need to be determined. To determine the wall
results in a significantly smaller engine performing at 50,900 lbf temperatures, notional materials need to be chosen.

Fig. 18 Vacuum thrust vs chamber length, system-level model.

 SIMMONS AND BRANAM 367

Table 10 Scaled DEAN design parameters

Design variables Response variables
O=F 6 Vacuum Thrust, Fvac 50,900 lbf
Total mass flow, m_ 104 lbm=s Vacuum Isp 489 s
Chamber length, lc 14.25 in. Chamber pressure, P0 1310 psia
Outer chamber radius, rco 6 in. Chamber temperature, T0 6413 R
Inner chamber radius, rci 2 in. Throat area, At 18:9 in:2
Expansion ratio,  125 Nozzle length, ln 13.6 in.
LOX bypass, % 90 Chamber volume, Vc 970 in:3
LOX pump efficiency,
p;LOX 0.773 LOX pump pressure ratio, PRp;LOX 58
LOX turbine efficiency,
t;LOX 0.949 LOX pump power 1053 HP
LH2 pump 1 efficiency,
p;LH2;1 0.8 LOX turbine pressure ratio, PRt;LOX 1.6
LH2 pump 2 efficiency,
p;LH2;2 0.83 LOX turbine power 1053 HP
LH2 pump efficiency,
p;LH2 0.9 LOX temperatures, TLOX 160–435 R
LOX pressures, pLOX 2450–2600 psia
Max LOX Mach number 0.37
LH2 pumps pressure ratio, PRp;LH2 9.3
LH2 pump 1 power 417 HP
LH2 pump 2 power 3430 HP
LH2 turbine pressure ratio, PRt;LH2 1.8
Downloaded by BRISTOL UNIVERSITY on January 27, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.51534

LH2 turbine power 3847 HP

LH2 temperaturess, TLH2 73–324 R
LH2 pressures, pLH2 2960–3770 psia
Max LH2 Mach number 0.96

Second, the DEAN architecture offers significant performance

gains over single-expander cycle-based upper-stage engines, such as
the RL10B-2. The DEAN has a chamber pressure that is twice that of
the RL10B-2. This increased chamber pressure results in twice the
vacuum thrust and a 5% higher vacuum specific impulse. Finally, the
DEAN engine is 69% shorter than the RL10B-2, despite its increased
thrust and specific impulse.

Acknowledgments
The authors would like to thank the following institutions for
providing the software used in this research: NASA for the
Numerical Propulsion System Simulation package and Phoenix
Integration for providing ModelCenter. The views expressed in this
paper are those of the authors and do not reflect the official policy or
position of the U.S. Air Force, the U.S. Department of Defense, or the
U.S. Government.

References
[1] Martin, D. F., “Computational Design of Upperstage Chamber,
Aerospike, and Cooling Jacket of Dual-Expander Rocket Engine,” M.S.
Thesis, Air Force Inst. of Technology, Wright-Patterson, AFB, OH,
2008.
[2] Minick, A., and Peery, S., “Design and Development of an Advanced
Liquid Hydrogen Turbopump.” 34th AIAA/ASME/SAE/ASEE Joint
Propulsion Conference and Exhibit, Cleveland, OH, AIAA, Paper 1998-
3681, 13–15 July 1998.
[3] Humble, R. W., Space Propulsion Analysis and Design, McGraw–Hill,
New York, 1995.
[4] Strain, W. S., “Design of an Oxygen Turbopump for a Dual Expander
Fig. 19 Comparison of geometry for the scaled DEAN to the original Cycle Rocket Engine,” M.S. Thesis, Air Force Inst. of Technology,
geometry. Wright-Patterson, AFB, OH, 2008.
[5] “NPSS User Guide,” Ver. 5, NASA Glenn Research Center, Cleveland,
OH, 30 May 2006.
VI. Conclusions [6] Arguello, A. A., “The Concept Design of a Split Flow Liquid Hydrogen
The results of this study support two key conclusions. First, Turbopump,” M.S. Thesis, Air Force Inst. of Technology, Wright-
parametric modeling with automated trade studies is a powerful ap- Patterson, AFB, OH, 2008.
proach in conceptual design of rocket engines. Significant im- [7] ModelCenter 9.0 User Manual, Phoenix Integration, Blacksburg, VA,
provements were made to the design of the DEAN engine by 2010.
[8] Krach, A. E., and Sutton, A. M., “Another Look at the Practical and
modifying the numerical model to support parametric modeling and Theoretical Limits of an Expander Cycle, LOX/H2 Engine,” 35th
exploring the trade space through automated trade studies. These AIAA/ASME/SAE/ASEE Joint Propulsion Conference and Exhibit,
improvements included reducing the size of the turbomachinery by Los Angeles, AIAA Paper 1999-2473, 20–24 June 1999.
reducing the total mass flow, improving the oxidizer-to-fuel ratio for [9] Sutton, G. P., Rocket Propulsion Elements, 8th ed., Wiley, Hoboken,
optimal thrust performance, and reducing the overall length by 25%. NJ, 2010.
All of these savings were realized while maintaining the required
vacuum thrust performance of 50,000 lbf and increasing the vacuum J. Martin
specific impulse from 472 to 489 s. Associate Editor