Morphing Wing Fighter Aircraft Synthesis/Design Optimization · Morphing Wing Fighter Aircraft...
Transcript of Morphing Wing Fighter Aircraft Synthesis/Design Optimization · Morphing Wing Fighter Aircraft...
Morphing Wing Fighter Aircraft Synthesis/Design
Optimization
Kenneth Wayne Smith Jr.
A thesis submitted to the Faculty of the
Virginia Polytechnic Institute and State University
in partial fulfillment of the requirements for the degree of
Master of Science
In
Mechanical Engineering
Dr. Michael von Spakovsky, Chair
Dr. Walter O’Brien
Dr. Michael Philen
Dr. David Moorhouse
January 16, 2009
Blacksburg, Virginia
Keywords: exergy, morphing wing, optimization, aircraft, decomposition
Copyright © 2009 by Kenneth Wayne Smith Jr.
All other copyrighted items used by permission
ii
Morphing Wing Fighter Aircraft Design/Synthesis and Optimization
by Kenneth Wayne Smith Jr.
Abstract
This thesis presents results of the application of energy-based large-scale optimization of
a two-subsystem (propulsion subsystem (PS) and airframe subsystem-aerodynamics (AFS-A))
air-to-air fighter (AAF) with two types of AFS-A models: a fixed-wing AFS-A and a morphing-
wing AFS-A. The AAF flies 19 mission segments of a supersonic fighter aircraft mission and
the results of the study show that for very large structural weight penalties and fuel penalties
applied to account for the morphing technology, the morphing-wing aircraft can significantly
outperform a fixed-wing AAF counterpart in terms of fuel burned over the mission. The
optimization drives the fixed-wing AAF wing-geometry design to be at its best flying the
supersonic mission segment, while the morphing-wing AFS-A wing design is able to effectively
adapt to different flight conditions, cruising at subsonic speeds much more efficiently than the
fixed-wing AAF and, thus yielding significant fuel savings.
Also presented in this thesis are partially optimized results of the application of a
decomposition strategy for large-scale optimization applied to a nine-subsystem AAF consisting
of a morphing-wing AFS-A, turbofan propulsion subsystem (PS), environmental controls
subsystem (ECS), fuel loop subsystem (FLS), vapor compression/polyalphaolefin loop
subsystem (VC/PAOS), electrical subsystem (ES), central hydraulics subsystem (CHS), oil loop
subsystem (OLS), and flight controls subsystem (FCS). The decomposition strategy called
Iterative Local-Global Optimization (ILGO) is incorporated into a new engineering aircraft
simulation and optimization software called iSCRIPT™ which also incorporates the models
developed as part of this thesis work for the nine-subsystem AAF. The AAF flies 21 mission
segments of a supersonic fighter aircraft mission with a payload drop simulating a combat
situation. The partially optimized results are extrapolated to a synthesis/design which is believed
to be close to the system-level optimum using previously published results of the application of
ILGO to a five-subsystem AAF to which the partially optimized results of the nine-subsystem
AAF compare relatively well.
In addition to the optimization results, a parametric study of the morphing AFS-A
geometry is conducted. Three mission segments are studied: subsonic climb, subsonic cruise,
iii
and supersonic cruise. Four wing geometry parameters are studied: leading-edge wing sweep
angle, wing aspect ratio, wing thickness-to-chord ratio, and wing taper ratio. The partially
optimized AAF is used as the baseline, and the values for these geometric parameters are
increased or decreased up to 20% relative to an established baseline to see the effect, if any, on
AAF fuel consumption for these mission segments. The only significant effects seen in any of
the mission segments arise from changes in the leading-edge sweep angle and wing aspect ratio.
The wing thickness-to-chord ratio shows some effect during the subsonic climb segment, but
otherwise shows no effect along with the taper ratio in any of the three mission segments studied.
It should be emphasized, however, that these changes are made about a point (i.e.
synthesis/design), which is already optimal or nearly so. Thus, the conclusions drawn cannot be
generalized to syntheses/designs, which may be far from optimal. Also note that the results upon
which these conclusions are based may very likely highlight a weakness in the conceptual-level
drag-buildup method used in this thesis work. Further optimization studies using this drag-
buildup method may warrant setting the thickness-to-chord ratios and taper ratios rather than
having them participate in the optimization as degrees of freedom (DOF).
The final set of results is a parametric study conducted to highlight the correlation
between the fuel consumption and the total exergy destruction in the AFS-A. The results for the
subsonic cruise and supersonic cruise mission segments show that at least for the case when the
AFS-A is optimized by itself for a fixed specific fuel consumption that there is a direct
correlation between the fuel burned and total exergy destruction. However, as shown in earlier
work where a three-subsystem AAF with AFS-A, PS, and ECS is optimized, this may not always
be the case. Furthermore, based on the results presented in this thesis, there is a smoothing effect
observed in the exergy response curves compared to the fuel-burned response curves to changes
in AFS-A geometry. This indicates that the exergy destruction is slightly less sensitive to such
changes.
iv
Acknowledgements
This masters thesis is the product of a significant amount of work by myself and my
advisor, Dr. Michael von Spakovsky, as well as cooperation with an industry partner, TTC
Technologies Inc., and the Air Force Research Laboratories, specifically Dr. David Moorhouse
and Dr. Jose Camberos. The work was sponsored through a joint award of an SBIR Phase II
project to Virginia Tech and TTC Technologies.
I would firstly like to thank Dr. von Spakovsky for providing the opportunity to study this
topic, his sponsorship, his mentorship, his guidance, and his passion for his beliefs, all of which I
learned greatly from throughout my time as a graduate student at Virginia Tech.
Secondly, I would like to thank Dr. Ken Alabi of TTC Technologies Inc. for helping me
so much with learning their software and putting up with my seemingly numerous mistakes as
the code progressed through the stages of development and integration.
Thirdly, I would like to thank the members of the defense committee: Dr. Michael von
Spakovsky, Dr. Walter O’Brien, Dr. Michael Philen, and Dr. David Moorhouse for serving on
my committee and providing feedback on my thesis draft.
Fourthly, I would like to thank all my friends who helped me get through graduate school
and pulled me away from my work when necessary (or sometimes more than necessary), those in
Chi Alpha, you know who you are, and my new friends in Maryland who are seeing the
aftermath of finishing this project.
Fifthly, I would like to thank my family who has supported, encouraged, commiserated,
and advised me throughout this project. To my parents, Kenneth Sr. and Danlynne, for the work
ethic, endurance, and stubbornness to finish a project of this length that you imprinted in me, I
thank you as well.
But mostly, I would like to thank my Lord and Savior, Jesus Christ. For the strength,
endurance, ability, opportunity, family, friends, direction, love and direction, Father, I thank
You.
Proverbs 16:9
v
Table of Contents
Abstract ........................................................................................................................................... ii
Acknowledgements ........................................................................................................................ iv
Table of Contents .............................................................................................................................v
List of Figures ............................................................................................................................... IX
List of Tables ............................................................................................................................... XII
Nomenclature ............................................................................................................................... XV
Chapter 1 Introduction .....................................................................................................................1
1.1 Morphing-Wing Aircraft ....................................................................................................... 1
1.2 Aircraft System/Subsystem Synthesis/Design ...................................................................... 3
1.3 Modeling and Simulation ...................................................................................................... 5
1.4 Large-Scale Optimization and Mission Integration .............................................................. 7
1.5 Decomposition for Large-Scale Optimization ...................................................................... 9
1.6 The Use of Exergy Analysis................................................................................................ 10
1.7 Thesis Objectives ................................................................................................................ 11
Chapter 2 Literature Review ..........................................................................................................14
2.1 Benefits and Design Challenges for Morphing Aircraft ..................................................... 14
2.2 Study of Morphing-wing Effectiveness in Fighter Aircraft ................................................ 18
2.2.1 Airframe Subsystem – Aerodynamics (AFS-A) ........................................................... 19
2.2.2 Propulsion Subsystem (PS) .......................................................................................... 21
2.2.3 Most Important Results from Butt (2005) .................................................................... 21
2.3 Decomposition Strategies for Large-scale Aircraft Synthesis/Design Optimization .......... 22
2.4 Effects on Aircraft Synthesis / Design of Different Objective Functions ........................... 29
2.5 Exergy Methods for the Development of High Performance Vehicle Concepts ................ 35
2.6 Integrated Mission-Level Analysis and Optimization of High Performance Vehicle
Concepts ................................................................................................................................ 41
Chapter 3 Model Description and Synthesis/Design Problem Description ...................................47
3.1 Problem Definition .............................................................................................................. 47
3.2 Airframe Subsystem ............................................................................................................ 49
3.2.1 Lift and Drag ................................................................................................................ 50
3.2.2 Mission Analysis .......................................................................................................... 51
vi
3.2.3 Weight Fraction Model ................................................................................................. 54
3.2.4 Calculation of WTO ........................................................................................................ 55
3.2.5 Morphing-wing Considerations .................................................................................... 56
3.2.6 AFS-A Exergy Model ................................................................................................... 58
3.3 Propulsion Subsystem ......................................................................................................... 59
3.3.1 PS Layout and Station Definitions ............................................................................... 59
3.3.2 PS Thermodynamic Model ........................................................................................... 61
3.3.3 Thrust and Performance Calculations ........................................................................... 65
3.3.4 PS Exergy Model .......................................................................................................... 67
3.4 Environmental Controls Subsystem .................................................................................... 68
3.4.1 ECS Layout and Definitions ......................................................................................... 68
3.4.2 ECS Thermodynamic Model ........................................................................................ 69
3.4.3 ECS Exergy Model ....................................................................................................... 74
3.5 Vapor Compression / PAO Subsystem ............................................................................... 75
3.5.1 VC/PAOS Thermodynamic Model .............................................................................. 76
3.5.2 VC/PAOS Exergy Model ............................................................................................. 78
3.6 Fuel Loop Subsystem .......................................................................................................... 79
3.6.1 FLS Thermodynamic Model ........................................................................................ 80
3.7 Oil Loop Subsystem ............................................................................................................ 81
3.7.1 OLS Thermodynamic Model ........................................................................................ 82
3.7.2 OLS Exergy model ....................................................................................................... 84
3.8 Central Hydraulic Subsystem .............................................................................................. 85
3.8.1 CHS Thermodynamic Model........................................................................................ 85
3.8.2 CHS Exergy Model ...................................................................................................... 88
3.9 Electrical Subsystem ........................................................................................................... 89
3.9.1 ES Thermodynamic Model ........................................................................................... 91
3.9.2 ES Exergy Model ......................................................................................................... 92
3.10 Flight Controls Subsystem ................................................................................................ 93
3.10.1 FCS Thermodynamic Model ...................................................................................... 94
3.10.2 FCS Exergy Model ..................................................................................................... 94
vii
Chapter 4 Large-scale System Synthesis/Design Optimization Problem Definition and Solution
Approach ........................................................................................................................................96
4.1 AAF Aircraft System Synthesis/Design Optimization Problem ......................................... 96
4.1.1 System-Level Optimization Problem Definition .......................................................... 97
4.1.2 Need for Decomposition ............................................................................................... 99
4.2 Iterative Local-Global Optimization (ILGO) Approach ................................................... 100
4.2.1 Local-Global Optimization (LGO) ............................................................................. 100
4.2.2 ILGO Approach .......................................................................................................... 104
4.3 System-Level, Unit-Based Synthesis/Design Optimization Problem Definitions ............ 106
4.3.1 Subsystem Integration and Coupling Functions ......................................................... 106
4.3.2 AFS-A System-Level, Unit-Based Synthesis/Design Optimization Problem Definition
.......................................................................................................................................... 110
4.3.3 PS System-Level, Unit-Based Synthesis/Design Optimization Problem Definition . 114
4.3.4 ECS System-Level, Unit-Based Synthesis/Design Optimization Problem Definition116
4.3.5 VC/PAOS System-Level, Unit-Based Synthesis/Design Optimization Problem
Definition .......................................................................................................................... 119
4.3.6 FLS System-Level, Unit-Based Synthesis/Design Optimization Problem Definition 123
4.3.7 OLS System-Level, Unit-Based Synthesis/Design Optimization Problem Definition
.......................................................................................................................................... 125
4.3.8 CHS System-Level, Unit-Based Synthesis/Design Optimization Problem Definition
.......................................................................................................................................... 128
4.3.9 ES System-Level, Unit-Based Synthesis/Design Optimization Problem Definition . 130
4.3.10 FCS System-Level, Unit-Based Synthesis/Design Optimization Problem Definition
.......................................................................................................................................... 132
4.4 Optimization Decision Variables and Variable Constraints ............................................. 134
4.5 Solution Approach............................................................................................................. 140
4.6 iScript™ Scripting Language and Optimization ............................................................... 141
Chapter 5 Results and Discussion ................................................................................................143
5.1 Two-Subsystem Optimization Results .............................................................................. 143
5.2 Nine-Subsystem Results .................................................................................................... 147
5.2.1 Preliminary Synthesis/Design Analysis ...................................................................... 147
viii
5.2.2 Projected Optimum and Comparison ......................................................................... 149
5.3 Parametric Study of the Morphing-Wing AFS-A ............................................................. 151
Chapter 6 Conclusions/Recommendations ..................................................................................159
References ....................................................................................................................................162
Appendix A Fan Performance Map Code ....................................................................................168
IX
List of Figures
Figure 1.1 Lockheed hunter-killer morphing aircraft concept (Bowman, Sanders, Weisshar,
2002). ........................................................................................................................................... 2
Figure 1.2 NASA Oblique-Wing Demonstrator, the NASA AD-1 (Dryden X-Press, 1979) ........ 3
Figure 2.1 Effect of morphing on the synthesis/design space of thrust to weight (T/W) vs wing
loading (W/S). ............................................................................................................................ 15
Figure 2.2 Mechanization of the mission adaptive wing (MAW) trailing edge. ......................... 16
Figure 2.3 Example of the benefit of mission adaptive wing (MAW) technology. ..................... 16
Figure 2.4 Mission Profile by segment or leg (Mattingly, Heiser, and Daley, 1987).................. 18
Figure 2.5 Subsystems and subsystem coupling functions (Rancruel, 2002). ............................. 27
Figure 2.6 Evolution of the gross take-off weight, fuel weight, AFS-A weight, and PS weight at
different points of the iterative local-global optimization (ILGO) approach (Rancruel,
2003)...…………………………………………………………………………………………28
Figure 2.7 Variation of vehicle specific thrust and exergy destruction rate with fan bypass ratio
and turbine inlet temperature for a fixed compressor pressure ratio of 8 for the supersonic
penetration mission segment (Periannan, 2005). ....................................................................... 31
Figure 2.8 Variation of vehicle specific thrust and exergy destruction rate with fan bypass ratio
and compressor pressure ratio for a fixed turbine inlet temperature of 1700° K for the
supersonic penetration mission segment (Periannan, 2005). ..................................................... 32
Figure 2.9 Variation of vehicle specific fuel consumption and exergy destruction rate with fan
bypass ratio and compressor pressure ratio for a fixed turbine inlet temperature of 1700 K for
the supersonic penetration mission segment (Periannan, 2005). ............................................... 33
Figure 2.10 Optimum gross takeoff weight with and without AFS-A DOF for objectives 1, 2,
and 3 (Periannan, von Spakovsky and Moorhouse, 2008)……………………………………..34
Figure 2.11 Optimum fuel weight with and without AFS-A DOF for objectives 1, 2, and 3
(Periannan, von Spakovsky, and Moorhouse, 2008)…………………………………………..34
Figure 2.12 Hypersonic vehicle configuration (Markell, 2005). ................................................ 35
Figure 2.13 A physical representation of the forebody and inlet component of the hypersonic
vehicle along with design and operational decision variables that govern the flow
characteristics throughout the inlet (Markell, 2005). ................................................................. 36
Figure 2.14 Propulsion subsystem components and airframe subsystem (Brewer, 2006). ........ 42
X
Figure 2.15 Total scramjet vehicle mission (Brewer, 2006). ....................................................... 42
Figure 3.1 Supersonic fighter aircraft mission from the RFP found in Mattingly, Heiser, and
Pratt (2002). ............................................................................................................................... 48
Figure 3.2 Free-body diagram of the aircraft (Rancruel, 2002). .................................................. 49
Figure 3.3 Engine system layout (Rancruel, 2002). ..................................................................... 60
Figure 3.4 Engine Station Definitions (Periannan, 2005). ........................................................... 60
Figure 3.5 ECS layout and components (Muñoz and von Spakovsky, 2001). ............................ 69
Figure 3.6 ECS station definitions (Rancruel, 2002). .................................................................. 70
Figure 3.7 Geometric parameters of the offset-strip fin (Muñoz and von Spakovsky, 1999). .... 72
Figure 3.8 VC/PAOS layout and station definitions (Rancruel, 2002). ....................................... 76
Figure 3.9 Schematic of the fuel loop subsystem (Rancruel, 2002). ........................................... 79
Figure 3.10 Oil loop subsystem schematic. ................................................................................. 82
Figure 3.11 Notional central hydraulics subsystem layout (simplified). ..................................... 86
Figure 3.12 Notional electrical subsystem schematic (simplified). ............................................. 90
Figure 4.1 Physical decomposition of a 2-unit system. ............................................................. 100
Figure 4.2 Multi-level optimization resulting in a set of nested optimizations. ........................ 103
Figure 4.3 An example of three subsystems and their associated coupling functions. .............. 104
Figure 4.4 Notional flow diagram of the application of the ILGO decomposition strategy to the
two-unit system of Figure 4.2. ................................................................................................. 106
Figure 4.5 Aircraft subsystem interactions and coupling functions. ......................................... 108
Figure 4.6 Diagram of optimization problem solution approach using ILGO. ......................... 140
Figure 5.1 Sensitivity analysis of morphing-wing effectiveness for different wing- and fuel-
weight penalties (Smith et al., 2007). ...................................................................................... 144
Figure 5.2 Total exergy destruction plus fuel exergy loss for each of the nine subsystems after
the first ILGO iteration. ........................................................................................................... 148
Figure 5.3 Variation of the mission segment fuel burned with variations in aspect ratio and the
sweep angle for mission segment 4 (subsonic climb at Mach 0.536 from a 20,000 ft to 41,700
ft altitude). ................................................................................................................................ 153
Figure 5.4 Variation of the mission segment fuel burned with variations in aspect ratio and the
sweep angle for mission segment 5 (subsonic cruise at Mach 0.656 at a 41,700 ft altitude). . 154
XI
Figure 5.5 Variation of the mission segment fuel burned with variations in aspect ratio and the
sweep angle for mission segment 17 (supersonic cruise at Mach 1.5 and 30,000 ft altitude). 155
Figure 5.6 Variation of the mission segment fuel burned with variations in thickness-to-chord
ratio for mission segment 4 (subsonic climb at Mach 0.536 from a 20,000 ft to 41,700 ft
altitude). ................................................................................................................................... 155
Figure 5.7 Variation of the mission segment exergy destruction with variations in aspect ratio
and the sweep angle for mission segment 5, subsonic cruise at Mach 0.656 at 41,700 ft altitude.
.................................................................................................................................................. 157
Figure 5.8 Variation of the mission segment exergy destruction with variations in aspect ratio
and the sweep angle for mission segment 17, supersonic cruise at Mach 1.5 and 30,000 ft
altitude...................................................................................................................................... 157
XII
List of Tables
Table 2.1 Mission Segment Definition and Description. .............................................................. 18
Table 2.2 AFS-A aerodynamics and model equations. ................................................................ 19
Table 2.3 Mission specifications (Rancruel, 2003). .................................................................... 26
Table 2.4 Comparison between the optimum ATA and the aircraft proposed by Mattingly,
Heiser, and Daley (1987), (Rancruel, 2003). ............................................................................. 28
Table 2.5 Comparison of the optimal combustor models (Markell, 2005). ................................ 38
Table 2.6 Optimal decision variable values for the energy and exergy based optimizations of a
scramjet engine with fixed thrust (Markell, 2005). .................................................................... 39
Table 2.7 Optimal design decision variable values for the single segment optimizations (Markell,
2005). ......................................................................................................................................... 40
Table 2.8 Optimal operational decision variable values for the partial mission (Markell, 2005).
.................................................................................................................................................... 40
Table 2.9 Optimal vehicle fuel mass flow rate comparison (Markell, 2005) .............................. 40
Table 2.10 Mission segment details (Brewer, 2006). .................................................................. 42
Table 2.11 Mission design and operational decision variables for the inlet, nozzle, combustor,
and airframe (Brewer, 2006). ..................................................................................................... 43
Table 2.12 Samples of results populations for sparse and dense optimal solution spaces. Note
that the very large numbers (i.e. E+15) represent infeasible solutions (Brewer, 2006). ........... 45
Table 2.13 Optimal objective function results (Brewer, 2006) ................................................... 46
Table 3.1 Air-to-air fighter (AAF) mission segments and details. .............................................. 48
Table 3.2 Master flight equation and governing flight equations. ............................................... 50
Table 3.3 Lift and drag equations for the AFS-A. ....................................................................... 50
Table 3.4 Mission segment model equations. .............................................................................. 52
Table 3.5 Weight fraction model equations. ................................................................................ 54
Table 3.6 Main subsystem weight calculations and 0TW . ............................................................. 55
Table 3.7 AFS-A exergy destruction rate equations. ................................................................... 59
Table 3.8 Low-bypass turbofan engine station definitions. ......................................................... 60
Table 3.9 Diffuser and nozzle equations...................................................................................... 61
Table 3.10 Fan and high pressure compressor equations. ........................................................... 62
Table 3.11 Burner and afterburner calculations. .......................................................................... 63
XIII
Table 3.12 High and low pressure turbine equations. .................................................................. 64
Table 3.13 Turbine cooling mixer and exhaust mixer equations. ................................................ 65
Table 3.14 Thrust and engine performance calculations. ............................................................ 65
Table 3.15 Inlet and nozzle drag and installed thrust equations. ................................................. 66
Table 3.16 PS exergy model equations. ....................................................................................... 67
Table 3.17 Thermodynamic model of the ECS (Periannan, 2005). ............................................. 70
Table 3.18 Geometric and heat transfer models of the compact heat exchangers. ...................... 72
Table 3.19 ECS exergy destruction rate equations. ..................................................................... 75
Table 3.20 VCPAOS model equations. ....................................................................................... 76
Table 3.21 Fuel loop subsystem thermodynamic model equations. ............................................ 80
Table 3.22 OLS pump work equations. ....................................................................................... 82
Table 3.23 OLS heating load equations. ...................................................................................... 83
Table 3.24 OLS exergy destruction equations. ............................................................................ 84
Table 3.25 Actuator flow estimation calculations. ...................................................................... 86
Table 3.26 Central hydraulic subsystem heating load equations (Majumar, 2003). .................... 88
Table 3.27 CHS subsystem exergy destruction equations (Bejan, 1996). ................................... 88
Table 3.28 Fighter aircraft power generation/empty weight estimate. ........................................ 91
Table 3.29 Electrical subsystem generator work. ........................................................................ 91
Table 3.30 ES heating load model equations. .............................................................................. 91
Table 3.31 ES exergy destruction model equations. .................................................................... 92
Table 3.32 Flight controls subsystem weight equations (Raymer, 1999). ................................... 94
Table 3.33 FCS actuator electrical power and fluid power requirements. ................................... 94
Table 3.34 FCS exergy destruction equations. ............................................................................ 95
Table 4.1 Number of coupling functions associated with each subsystem. ............................... 107
Table 4.2 AFS-A fixed-wing design and operational decision variables and inequality
constraints. ............................................................................................................................... 134
Table 4.3 AFS-A morphing-wing design and operational decision variables and inequality
constraints. ............................................................................................................................... 135
Table 4.4 AFS-A mission decision variables and inequality constraints................................... 135
Table 4.5 PS design and operational decision variables and inequality constraints. ................. 136
XIV
Table 4.6 ECS optimization synthesis / design and operational decision variables and inequality
constraints. ............................................................................................................................... 136
Table 4.6 FLS optimization decision variables and inequality constraints. .............................. 138
Table 4.8 VC/PAOS optimization synthesis / design and operational decision variables and
inequality constraints. .............................................................................................................. 139
Table 4.8 OLS optimization operational decision variables and inequality constraints. ........... 139
Table 5.1 Comparison of the optimum morphing-wing gross takeoff weights with a 6x wing-
weight penalty and the optimum fixed-wing gross takeoff weight (Smith et al. 2007). .......... 145
Table 5.2 Optimal fixed- versus morphing-wing AAF configuration and performance parameters
for the subsonic cruise and the supersonic penetration mission segments (Smith et al. 2007).
.................................................................................................................................................. 145
Table 5.3 AAF subsystem weights and the percentage of AAF empty weight after the first ILGO
iteration. ................................................................................................................................... 148
Table 5.4 AAF Subsystem percent weight reduction versus ILGO iteration (Rancruel, 2002). 149
Table 5.5 Projected AAF subsystem weights versus ILGO iteration based on the ILGO
progression from Rancruel (2002). .......................................................................................... 149
Table 5.6 Extrapolated nine-subsystem AAF gross takeoff weight and empty weight versus
ILGO iteration based on the ILGO progression from Rancruel (2002). .................................. 150
Table 5.7 Extrapolated AAF subsystem system-level optimum weights after seven ILGO
iterations along with the percentage of AAF empty weight. ................................................... 151
Table 5.8 Extrapolated subsystem optimum weights versus the optimum subsystem weights
from Rancruel (2002). .............................................................................................................. 151
Table 5.9 Baseline AAF configuration and performance for mission segment 4, 5, and 17. .... 152
XV
Nomenclature
A Area, aspect ratio FCS flight controls subsystem
AR inlet area FLS fuel loop subsystem
AAF air-to-air fighter g gravitational constant
AAFS airframe subsystem - aerodynamics
GA genetic algorithm
b wing span ILGO iterative local-global optimization
hB horizontal tail span cbK landing gear cross beam factor
BCA best cruise altitude dwK delta wing factor
BCM best cruise Mach mcK 1.45 for mission completion required, 1 otherwise
BCLM best climb Mach tpgK landing gear tripod factor
DC drag coefficient vsK sweep wing factor
0DC parasite drag coefficient vshK Constant = 1.425 if variable sweep wing, 1.00 otherwise
LC lift coefficient L lift, wing length
DwaveC wave drag coefficient aL Electrical routing distance (ft) from generators to avionics to cockpit
feC skin friction coefficient mL length of main gear
CHS central hydraulic subsystem nL length of nose gear
D drag mass flow rate
DOF degree of freedom M Mach number
ECS environmental controls subsystem cN number of crew
e Weissinger span efficiency enN number of engines
ES electrical subsystem genN number of generators
desEx exergy destroyed lN length of nose gear
wF fuselage width at horizontal interception sN number of flight control systems
XVI
uN Number of hydraulic utility functions (typ. 5-15)
V velocity
zN ultimate load factor stallV stall velocity
q dynamic pressure PAOSVC / vapor compression/PAO loop subsystem
Q flow rate, heating load W weight, work
OLS oil loop subsystem lW length of main gear
ORS optimum response surface dgW design gross weight
P pressure EmptyW aircraft empty weight
TOP power takeoff from PS TOW aircraft takeoff weight
PAO polyalphaolefin
PS propulsion subsystem
R specific gas constant, additional drag
Greek
S planform area of the wing angle of attack, fan bypass ratio
csS total area of flight control surfaces
design bleed air ratio
htS horizontal tail area taper ratio, shadow prices
refS reference area of wing LE leading edge sweep angle
wetS wetted area of wing pressure ratio, pi
sfc specific fuel consumption temperature ratio
t time
T thrust, temperature
instT installed thrust
tT total temperature
reqT thrust required
ct / thickness to chord ratio
v valve ratio
1
Chapter 1
Introduction
The energy-based economy is driving a new technology to the aircraft industry:
morphing-wing technology, or the ability of an aircraft to change the shape of its wings during
flight, is being researched heavily by both the military/government and the private aircraft
industry. The goal of the application of morphing technology is to develop an aircraft that can
adapt or change its aerodynamic performance to fly dissimilar missions or dissimilar mission
segments within a mission more efficiently than a fixed-wing aircraft.
A prime platform for investigating morphing technology is the fighter aircraft. Currently
designed fighter aircraft are being used for multiple roles, depending on the branch of the
military or even country in which the aircraft is being used. The designer of a fixed-wing aircraft
would find the dissimilar roles and requirements of the aircraft to be a design challenge, to say
the least. Even more of a challenge would be to design an aircraft that can perform all the roles
required in the most efficient manner, which has been proven to be impossible (e.g. a subsonic
high-endurance reconnaissance aircraft can not perform the role of a supersonic fighter aircraft
more efficiently than the fighter aircraft, nor vice versa). Morphing technology allows a system
designer to design an aircraft that can adapt to its flight conditions in order to meet the
performance requirements in the most efficient manner.
1.1 Morphing-Wing Aircraft
“Morphing wings” is the new catch phrase in the aerospace research industry today.
Morphing technology employed in aircraft wings has been proven, at least on a conceptual level,
to allow aircraft to outperform their fixed-wing counterparts over the entire mission in fuel
savings due to drag reduction and improved lift-to-drag ratios. “Morphing wings” can imply the
ability of an aircraft to change its aerodynamic performance from a very simple morphing, such
as flaps or slats, to a more extensive morphing such as variable wing length, sweep, and chord
lengths. An example of morphing wing aircraft is the Lockheed Martin Unmanned Air Vehicle
(UAV) concept which came out of the DARPA Morphing Aircraft Structures program is shown
in Figure 1.1. This concept has the ability fold its wings, effectively changing the flight
characteristics (by varying the wetted area and aspect ratio) of the aircraft extensively to allow a
2
mission that could include reconnaissance, loiter, and attack/low observeability configurations in
the same vehicle, encompassing the requirements of a hunter-killer mission. The hunter-killer
mission would involve first searching for and identifying the target, then destroying it with the
shortest delay in time between identification and destruction as possible. This type of mission
would normally be performed by a “package” or group of aircraft specializing in certain portions
of the hunter-killer mission. The Lockheed Martin UAV concept allows the package to be
eliminated, as the concept can perform all tasks in the hunter-killer mission due to the ability to
morph its wings.
Figure 1.1 Lockheed hunter-killer morphing aircraft concept (Bowman, Sanders, Weisshar, 2002).1
This variable geometry wing can rotate on a pivot, which significantly reduces the drag
of the wing in high-speed flight. The wing leading edge is near perpendicular to the flight
direction for takeoff and high endurance mission segments for maximum lift, but is rotated (or
swept back) to effectively reduce the induced drag and parasitic drag (due to wave drag) for
supersonic flight. Other examples of morphing wings have been investigated for the purpose of
building a single aircraft that can perform the duties of a group of aircraft. One such example is
1 Reprinted with permission from author.
3
the oblique-wing concept. This concept was investigated and tested under the DARPA Oblique
Flying-Wing Project with the NASA AD-1 (Ames Dryden AD-1).(Curry and Sim, 1982; Dryden
X-Press, 1979) which is shown in Figure 1.2.
Figure 1.2 NASA Oblique-Wing Demonstrator, the NASA AD-1 (Dryden X-Press, 1979)
1.2 Aircraft System/Subsystem Synthesis/Design
The synthesis/design process of the AAF is driven by the mission flown which defines
the customer requirements for the aircraft to be developed. The mission flown in this thesis
work is very similar to Rancruel (2003) and Butt (2005) but with some changes which are
discussed at length in Chapter 3. The flow chart of the synthesis/design process is shown in
Figure 1.3.
The first synthesis/design stage is the conceptual synthesis of the design. Viable
solutions to the requirements are developed and analyzed for cost, feasibility, manufacturability,
etc. in this stage. Synthesis, design, and operational variables are also investigated at this level
to verify the feasibility of the design. The conceptual synthesis stage has the most variation in
model design and performance as often the requirements are adjusted at this step as well,
depending on the level of technology required / available. The best solution for the requirements
is eventually chosen at this step based on multiple performance measures (e.g. cost, weight, ease
of manufacturing, performance, etc.)
The next stage is the preliminary synthesis/design. The configuration of the best solution
from the conceptual synthesis is frozen at this step, and more detailed testing and major
subsystem design commences on the solution. Databases of analytical data and performance
testing are compiled at the preliminary synthesis/design step as well. For example, the engine
4
sizing, weight, compressor and turbine stages, bypass ratios, etc. would be determined at this
stage of the synthesis/design process.
Figure 1.3 Aircraft synthesis/design stages (Raymer, 20062; Rancruel, 2002).
Following the preliminary synthesis/design stage is the detailed design stage. Actual
subsystems would be prototyped and set up for manufacturing, and subsystem performance in
the detailed design stage verified. The performance of the total system would be verified /
estimated again at this point as the subsystems are prototyped and tested. Tooling for
manufacturing would also be developed at the detailed design stage.
The final synthesis/design stage is manufacturing. Hopefully, after the first production
model is made, the design still meets all the requirements that were originally posed by the
customer. Often this is not the case and the synthesis/design needs to be revised slightly (or
perhaps drastically) to meet the original performance requirements. The prevalence of the total
system not meeting original performance requirements is evidence that a better synthesis/design
process/method may be needed.
2 Copyright 2006 by Daniel P. Raymer. Reprinted with permission from author.
5
Finally, each synthesis/design step often has a different design group
synthesizing/designing and optimizing a subsystem and its components somewhat independently
from all the other design groups. A more integrated synthesis/design/optimization process might
reduce the number of occurrences of the final synthesis/design not meeting the original
requirements.
1.3 Modeling and Simulation
An aircraft can be viewed as a system of subsystems. Each subsystem is a component or
group of components that can be logically separated from the rest of the components that make
up the system. Subsystem boundaries may be determined using a variety of criteria such as
logical physical boundaries, thermodynamic boundaries, time boundaries, etc.
After the subsystems are clearly defined, a model of each of the subsystems is developed.
A subsystem model can take on many different forms: a thermodynamic model, a geometric
model, an aerodynamic model, a kinetic model, etc. These models can be analytical, empirical,
or semi-empirical and can be zero-D or lumped-parameter models or 1-D, 2-D, or 3-D
computational fluid dynamics (CFD) model. Typical of the types of subsystem are geometric,
aerodynamic, and thermodynamic used in subsystem modeling, simulation, and large-scale
optimization are models with lumped parameter distributions. A lumped parameter distribution
indicates that properties that would normally have an infinite distribution of values are
represented by an averaged value. For example, the temperature of the air in a turbofan engine
would have an infinite distribution of temperatures throughout the engine; however, the
temperatures are “lumped” or averaged to one temperature at a given station in the engine and
calculated as such in the model equations.
Typically, each subsystem of an aircraft is defined as a group of components that perform
a given function. For example, the engine or propulsion subsystem (PS) provides the thrust to
the aircraft and power to other subsystems and is easily separable, thermodynamically from the
rest of the aircraft. Thus, it is considered a subsystem. The airframe subsystem (AFS) is
responsible for providing the aerodynamics required to fly the mission as well as house the other
subsystems. The fuel loop subsystem (FLS) consists of all the components associated with the
fuel tank and a set of associated heat exchangers which help condition the fuel and deliver it to
6
and from the engine. Other subsystems are defined in a similar fashion and each has a range of
operating conditions and parameters for the function it must perform.
Once the subsystem components and configurations have been defined, every subsystem
model (lumped parameter or otherwise) must be written in an engineering software language to
represent the model equations which simulate the subsystem behavior.
The coded model must then be validated to verify that it produces operating parameter
values close to previously published ones or which from engineering experience seem
reasonable. After validation, each subsystem is integrated with the other subsystems. Integrating
a subsystem involves defining each inter-subsystem interaction by assigning appropriate
variables between subsystems. For example, the thrust required by the airframe subsystem-
aerodynamics (AFS-A) to fly a specific part of the mission segment is assigned as a required
thrust from the engine. Thus, the thrust required by the AFS-A is an interaction between the
engine and AFS-A.
When all the subsystems are integrated, the modeling of aircraft is complete and the code
ready to use to simulate the aircraft’s behavior for a given design point. The simulation gives
feedback as whether or not the aircraft has been designed in a way that enables it to successfully
fly at that design point (usually the most difficult part of the mission to fly, i.e. supersonic flight),
and if so, how well the design flies at that design point compared to other syntheses
(configurations) / designs. When multiple simulations are ran and compared to each other to find
the best configuration / design, it can be said that the system is being optimized. Conventionally,
this is done by trial and error and not by using large-scale optimization. Furthermore,
conventionally, this process of synthesis / design of fixed-wing aircraft is performed only at the
design point and off-design operation is simply verified. However, large-scale optimization
would greatly enhance this process as would inclusion of the off-design mission segments
directly in the synthesis/design process along with the optimization. This requires a mission
integrated approach which is in fact a necessary approach when applying morphing-wing
technology to an aircraft. This is discussed along with large-scale optimization in the following
section.
7
1.4 Large-Scale Optimization and Mission Integration
When mathematical optimization is applied to the process of finding the “best” system
synthesis / design of a complex system, it is called large-scale optimization. To find this “best”
or optimum system requires a performance metric or optimization objective function by which
numerous syntheses / designs can be compared. For example, an optimization objective for a
power plant synthesis / design may be to maximize the power output or to minimize the
operating cost of the power plant. The optimization takes place by varying the system synthesis /
design and operating parameters and then evaluating system performance by running a system
simulation with the given synthesis / design and operating parameters. The independent system
synthesis / design variables are called the synthesis / design decision variables, while the
independent system operating variables are called operational decision variables. The variables
are called “decision” variables because they are being varied, or decided upon, during the
optimization to find the best system configuration and design.
A complete system optimization often requires that both the synthesis / design and
operational decision variables be varied to find the optimum system synthesis / design as
knowledge of system behavior during operation as well as the system’s physical characteristics
are needed to find the optimum system. The highly integrated and tightly coupled (i.e.
everything influences everything else) nature of synthesizing / designing an aircraft typically
requires that both synthesis / design variables and operational decision variables participate in
the optimization as is done in this thesis research. Unfortunately, a complete system
optimization for even a single design point can prove to have a very large computational burden
depending on the complexity and size of the system being optimized. This creates a need for
efficient and effective large-scale optimization algorithms and additional tools and methods to
manage the computational burden of optimizing such a large system.
Adding morphing-wing technology to an aircraft further increases the complexity of the
optimization problem. To fully investigate the benefits of morphing-wing technology, a single
design point optimization is not sufficient. Aircraft performance must be evaluated over the
entire mission since morphing wing technology is intended to improve the off-design
performance of the aircraft. If the aircraft is being synthesized / designed with only the design
flight conditions in mind, morphing-wing technology cannot show any benefit over a fixed-wing
aircraft. Of course, mission integrated optimization compounds the computational burden
8
significantly as additional operational decision variables are required for each operational
condition. In reality, this would require a dynamic, real-time optimization. Computationally,
this adds an additional computational burden to an already computationally challenged problem.
However, for purposes of synthesis / design, this dynamic behavior can be approximated by a
quasi-stationary approach. This is done by splitting the entire aircraft mission as defined, for
example, in Mattingly et al. (2002), into mission segments (see Figure 1.4). Separating the
mission into segments is a form of time decomposition which is common to aircraft synthesis /
design and makes the aircraft modeling quasi-stationary instead of dynamic. Operational
decision variables are then applied to each of the mission segments for each of the subsystems,
and the entire aircraft is optimized over the entire mission. In a sense, the aircraft no longer has
a “design point” per se but is instead optimized for its overall performance in the entire flight
envelope, from takeoff to landing.
Mission integrated synthesis/design is absolutely necessary for studying the effects of
morphing-wing technology in fighter aircraft; however, mission integration synthesis / design
and optimization creates a huge computational burden. The computational burden increase is
primarily due to an increase of degrees of freedom (DOF) or the number of variables the
optimization algorithm has “free” to decide upon for system simulation. An additional problem
is the non-linear response of aircraft performance with respect to the decision variables. This
and the large number of DOF as well as the fact that these may include a mix of discrete and
continuous variables limits the optimization algorithms that can be used, as many are unable to
find the overall aircraft optimum due to the variable search method used.
A solution to the dilemma of too many DOF which does not require a reduction in their
number is the use of decomposition strategies. A solution to the difficulties caused by the non-
linearities and the mix of discrete and continuous variables is the use of heuristic optimization
algorithms (.e.g., genetic algorithms, simulated annealing, etc.) or the use of hybrid heuristic-
nonheuristic optimization algorithms (e.g., surrogate, model-based optimization, etc.). A non-
heuristic algorithm is typically a gradient-based algorithm such as, for example, sequential
quadratic programming (SQP), generalized reduced-gradient (GRG), steepest descent, etc. A
discussion of the decomposition strategies and optimization algorithms used in this thesis
research appears in Chapter 4. However, a brief discussion of decomposition for large-scale
optimization is given in the following section.
9
1.5 Decomposition for Large-Scale Optimization
The size of a complete aircraft system optimization problem may require a decomposition
strategy to handle the huge (and perhaps prohibitively so) complexity of the optimization. There
are several different types of decomposition strategies (Frangopoulos, von Spakovsky, and
Sciubba, 2002) among which physical decomposition is a major player. Most of the physical
decomposition strategies in the literature can be characterized as Local-Global Optimizations
(LGOs). With LGO, the system is physically decomposed into a set of subsystems or units each
of which is characterized by its own set of decision variables. Each set of decision variables is
strictly local, i.e. affects in the main system only its corresponding unit. In a complex system,
however, there is always another set of decision variables which is not strictly local and which
acts at the so-called system level to affect some or all of the units. It is this division which leads
to a multi-level optimization problem in which at the system level an optimization occurs with
respect to its set of decision variables, while at the unit level individual optimizations of each
unit are carried out with respect to each individual set of unit-level decision variables. To
maintain the integrity of the overall system optimization, i.e. one which takes into account all of
the decision variables simultaneously, the unit-level optimizations must occur many times within
the system-level optimization, resulting in a set of so-called nested optimizations. Although such
a set can only approximate the so-called single-level optimization which occurs without
decomposition, it nonetheless may do so quite closely. However, the computational burden is
much greater than with the single-level optimization and may in fact be prohibitive.
To address this last problem as well as a number of others discussed in Chapter 4, a
physical decomposition strategy called Iterative Local-Global Optimization (ILGO) has been
developed and applied by Muñoz and von Spakovsky (2000a,b,c,d, 2001a,b) and Rancruel and
von Spakovsky (2006, 2004a,b, 2003a,b) to eliminate this problem of nested optimizations.
ILGO does this by eliminating the system-level optimization and, thus, the need for nested
optimizations and does so by bringing the system-level information down to the unit-level. This
is done by maintaining subsystem interactions via a set of coupling function and shadow price
pairs, which measure the effect that changes in the coupling functions have on the system
objective function. Elimination of the nested optimizations reduces much of the computational
burden of LGO and has the further advantage of permitting the subsystem optimizations to take
10
place simultaneously. ILGO, coupling functions, and shadow prices are discussed at more length
in Chapter 4.
1.6 The Use of Exergy Analysis
An emerging analysis tool for aircraft systems is exergy or available energy analysis.
The available energy of a system is defined as the largest amount of energy that can be
transferred from a system to a weight in a weight process while bringing the system to a mutual
stable equilibrium with a notional reservoir (Gyftopoulos and Beretta, 1991, 2004). The
usefulness of this analysis technique is realized in multi-component, highly-coupled energy
systems where identifying and correcting performance losses is made difficult using traditional
energy balance methods. Energy balances treat all forms of energy as equivalent, without
differentiating between the quality (ability to produce useful work) of energy crossing the system
boundary. Hence the energy from a high temperature source is treated in the same way as the
energy rejected to a low temperature sink. Energy balances do not provide information about
internal losses. For example, an energy balance for an isolated system in a not stable equilibrium
state shows that the process the system undergoes incurs no losses. This, however, is not true! In
fact, most of the causes of thermodynamic losses in thermal, chemical, and mechanical processes
such as heat transfer across finite temperature differences, mixing, combustion, and viscous flow
cannot be detected with energy balances since these losses are not associated with a loss of
energy (which can neither be created or destroyed) but instead with a decrease in the quality of
the energy.
An exergy or available energy approach overcomes these deficiencies since exergy
accounts for this loss in quality. The rate of loss of exergy internal to the system (i.e. the rate of
irreversibilities or entropy generated) provides information about the true inefficiencies of the
system. Hence an exergy analysis of a multi-component system such as the propulsion subsystem
(PS) of an aircraft indicates the extent to which its components contribute to the inefficiency of
the overall aircraft system. Unlike the energy method which is based on the First Law of
Thermodynamics alone, the concept of exergy and irreversibility is based on both the First and
Second Laws of thermodynamics. The irreversibility of a system, sub-system, or component can
be found by doing an exergy balance which combines unsteady or steady state balances of mass,
energy, and entropy into a single balance of exergy. The overall exergy destruction rate which is
11
directly proportional to the overall entropy generation rate and which appears in this single
balance can be determined directly from this balance. To determine the individual exergy
destruction rates which comprise this overall rate, a set of independent phenomenological
equations relating the exergy destruction rates to the various irreversible phenomena present in a
given process must be used.
Since the intention of this thesis research is to investigate morphing wing feasibility using
modeling, simulation, mission integration and optimization, exergy analysis is an additional tool
which can be employed for finding locations in the aircraft system needing the most
improvement and for categorizing the internal and external losses which occur in and out of the
aircraft. In other words, an exergy analysis can provide information about which components or
set of variables may be good candidates for further optimization or re-optimization and is useful
for setting up guidelines for process improvements since no matter what aerodynamic,
thermodynamic, kinetic or geometric phenomena are being modeled, a common quantity, i.e. the
exergy, can be used as the measure of process improvement or performance.
1.7 Thesis Objectives
The overall goal of this thesis work is to investigate the feasibility of applying morphing-
wing technology to an air-to-air fighter (AAF). The feasibility is investigated by modeling,
simulating, analyzing, and optimizing a morphing-wing fighter aircraft and comparing it to a
fixed-wing fighter aircraft. The aircraft model is an expansion by four subsystems on the five-
subsystem model developed and implemented by Rancruel (2003) and is intended to bring the
number of subsystems up to that of a complete aircraft system. It, thus, consists of nine
subsystems including the following: airframe subsystem-aerodynamics (AFS-A), propulsion
subsystem (PS), environmental controls subsystem (ECS), fuel loop subsystem (FLS), vapor
compression/polyalfaolephin subsystem (VC/PAOS), oil loop subsystem (OLS), central
hydraulic subsystem (CHS), electrical subsystem (ES), and flight controls subsystem (FCS). The
first five of these exist in the aircraft model of Rancruel (2003) while the latter four are
researched, developed, implemented, and validated here. Two additional subsystems exist in
both aircraft models, namely, the fixed and expendable payload subsystems. However, neither
involves synthesis/design or operational degrees of freedom (DOF) even though each plays a
role in the synthesis/design process. All the subsystems used are the same between the
12
morphing- and fixed-wing models with the exception of additional DOF for the morphing AFS-
A which are included to account for the wing geometry changing during flight.
To achieve the overall goal outlined above, the first objective of this thesis research
involves the research and development of four conceptual-level, lumped-parameter subsystem
models (thermodynamic and geometric) and extensive modification of the other five high-
fidelity, lumped-parameter subsystem models (aerodynamic, thermodynamic, and geometric)
which have already been fairly well documented in a previous Ph.D. dissertation (Muñoz, 2000)
and M.S. thesis (Rancruel, 2003) and in publications by Muñoz and von Spakovsky (2000a,b,c,d,
2001a,b) and Rancruel and von Spakovsky (2006, 2004a,b, 2003a,b). The subsystems are as
listed above. Note, however, that the AFS-A involves separate fixed-wing and geometry
morphing aerodynamics models. These along with the PS, ECS, FLS, and VC/PAOS are
modified from previous work and implemented in iSCRIPT™, which is a software language and
tool set being developed specifically for the purpose of aircraft mission integrated
synthesis/design modeling, analysis, and large-scale optimization by TTC Technologies, Inc.,
(TTC) as part of a U.S. Air Force AFRL Phase II SBIR in which Virginia Tech is a participant.
The remaining four subsystems, the OLS, ES, CHS, and FCS, are developed and implemented
specifically for this thesis work.
The second objective of this thesis research is to integrate all of the subsystems into an
overall aircraft system capable of flying an entire fighter aircraft mission. This requires defining
a complete mission able to thoroughly exercise the capabilities of both the fixed- and the
morphing-wing aircraft. It also requires assigning the proper wing- and fuel-weight penalties due
to morphing-wing technology (both explained in Chapter 3).
A third objective is to apply large-scale optimization using physical decomposition (i.e.
ILGO) to the mission integrated synthesis/design of both the fixed- and morphing-wing fighter
aircraft and then analyze and compare the results. The main comparison for the fixed- versus the
morphing-wing aircraft is a sensitivity study of the effects of the wing weight and fuel weight
penalties associated with the morphing technology as was done in Butt (2005) but with a much
more detailed and complete aircraft system.
The final or fourth objective is a study of the morphing-wing parameters that provide the
highest payoff in terms of fuel savings for three mission segments: subsonic climb, subsonic
cruise, and supersonic cruise. The effect of varying the wing parameters on the fuel burned is
13
compared to the exergy destruction plus exergy loss due to unburned fuel for the AFS-A to see a
correlation in trends. The results and conclusions drawn are then compared to those given in
Periannan, von Spakovsky, and Moorhouse (2008).
14
Chapter 2
Literature Review
Much research has been done in the areas of morphing-wing technology, exergy analysis,
and optimization, and how they relate to aircraft synthesis / design. This chapter reviews some
of the previous work done in this area in each of these areas.
2.1 Benefits and Design Challenges for Morphing Aircraft
Moorhouse, Sanders, von Spakovsky, and Butt (2005, 2006) illustrate the potential
benefits of applying morphing wings to an aircraft as well as the design challenges that are yet to
be overcome in applying morphing technology to aircraft. The aerospace industry has shown a
history of taking a current technology and expanding it to yield the next generation of aircraft.
Often the new designs end up looking like the old designs albeit with new materials or more
sophisticated electronics. The current aerospace customer needs a more affordable aircraft with
expanded mission capabilities. Technologies are already being applied to wings to allow wing
shape and, thus, aerodynamic performance to change depending on the requirements of the flight
conditions. An early example of wing morphing, i.e. wing twisting, was used by the Wright
brothers for roll control and is being ‘re-invented’ by the active aeroelastic wing (AAW)
program. Currently, variable wing sweep or “swing wings” are employed on many fighter
aircraft to allow better cruise endurance but without sacrificing high velocity flight performance.
Low order shape control is being used on aircraft in the form of Fowler flaps and
ailerons. These technologies have been hugely successful, enabling aircraft to reduce stall
speeds, increase lift, etc. to perform mission segments more successfully than otherwise may
have been possible. However, more extensive shape control or ‘morphing aircraft structure’ is
desired to allow drastic wing planform area and aerodynamic performance changes during flight.
The benefits of applying morphing wings to a fighter aircraft were investigated with a
model and a set of optimizations by Butt (2005; see Section 2.2 for details on the investigation).
The optimization results show that drastic morphing of an aircraft wing, including wing sweep,
wing length, and root and tip chord lengths could allow a fighter aircraft to use significantly less
fuel than a fixed-wing counterpart even for a much heavier morphing-wing aircraft. A point
perhaps not obvious to a designer is the fact that design constraints, which are determined by the
15
mission requirements, are relaxed due to the ability to morph the aircraft wings. Figure 2.1
shows an example of an aircraft synthesis/design space with the dark shaded region. However,
when morphing technology is used on the aircraft, the design space is expanded to include the
cross-hatched region.
Figure 2.1 Effect of morphing on the synthesis/design space of thrust to weight (T/W) vs wing loading (W/S)
(Periannan, von Spakovsky, and Moorhouse, 2008).
The increase in the design space indicates that morphing wings allow the aircraft to
perform the same mission as a fixed-wing counterpart, but with more flexibility in the design of
the wing loading and thrust to weight ratio.
Morphing-wing technology certainly shows very promising results for implementation on
paper; however, building a morphing-wing presents new challenges in design, implementation,
and control. The difficulty is how can the morphing technology be mechanized? The model in
Butt (2005) uses mission segment-by-segment morphing or a single configuration for the entire
mission segment. Would real-time shape control provide more benefit? One concept for real-
time shape control is the mission adaptive wing (MAW). An example of a trailing edge MAW is
shown in Figure 2.2.
The MAW allows the wing lift and drag to be dynamically adjusted (or morphed in ‘real
time’) during flight to outperform a fixed-wing counterpart. An example of such behavior could
be observed in a sustained turn. Figure 2.3 shows the benefit of employing MAW technology to
maximize the lift/drag ratio to allow a much smaller turning radius over that of a fixed-wing
aircraft. The MAW shows great promise as does the morphing-wing aircraft in Butt (2005).
Additional design risk is added when real-time morphing is implemented; furthermore, the level
16
Figure 2.2 Mechanization of the mission adaptive wing (MAW) trailing edge (Periannan, von Spakovsky,
and Moorhouse, 2008).
of aircraft integration required to successfully implement morphing technology is substantially
higher than what traditional design methods dictate today. The authors believe a new design
methodology must be used: the vehicle is considered a device in which all components are
Figure 2.3 Example of the benefit of mission adaptive wing (MAW) technology (Periannan, von Spakovsky,
and Moorhouse, 2008).
optimized to minimize exergy consumption at the system level. The authors assert three things
must be changed in traditional design methods. The first is in the philosophy of how structures
are designed. Wing structures must be designed for deformation characteristics from the start of
17
the design. The second change is that a better understanding of how to optimize the distribution
of sensors and actuators in addition to structural properties, such as mass and stiffness, needs to
be developed. The final change is to investigate the scalability of actuators and the degree of
deformation from a systems level perspective that is best suited to the design problem. The
exergy-based concept design method proposed by the authors shows good promise in solving
this third item as exergy is a metric common to all systems. The exergy-based objective function
for optimization requires an integration of the entire vehicle during the design process rather than
the traditional design assumption that minimum weight is the best design.
There are other challenges in employing morphing wings that must be overcome. The
example in Figure 2.3 should show that even for a simple model, non-linear design and analysis
is required in order to avoid the significant error which comes from assuming linearity. Flight
control issues of a wing capable of shape changing to modify roll characteristics, drag, and even
perhaps wing loading profiles would pose a significant design challenge. Previously, control
gains were set based on wind-tunnel testing with a re-design if flight testing showed
improvement was needed. The current design method is to use a neural network to actively
adjust gains based on actual in-flight aerodynamic performance. Perhaps this technology could
be employed to control morphing wings as well or would it be trying to catch up to the
constantly changing flight characteristics (forming too great of a computational burden)?
These authors as well as others have clearly shown that employing adaptive structures in
the next generation of aircraft will yield significant benefits in operating cost and performance
for aircraft that perform significantly dissimilar mission segments. They have also shown that
current design methodologies will be unable to accomplish the level of integration required to
successfully design (and eventually manufacture) a morphing-wing aircraft. Non-linear design
methods and an exergy-based optimization metric may be required to reach this level of
integration and some questions need answering as well. How fast do the wings need to change
shape? If they change slowly, will traditional flight control subsystems be sufficient for flight
control? If they change quickly, what will be the flight controls needed to compensate for
aerodynamic instability during shape change? These questions and design challenges must be
answered and overcome to continue to ‘re-invent and extend’ the adaptive structures that started
with the Wright brothers.
18
2.2 Study of Morphing-wing Effectiveness in Fighter Aircraft
Previous work has been done in the area of morphing-wing technology at Virginia Tech
(Butt, 2005). An afterburning turbojet propulsion subsystem (PS) and the aerodynamic aspects
of a morphing airframe subsystem (AFS-A) are used to perform a feasibility study of employing
morphing-wing technology in a supersonic fighter aircraft flying a mission as depicted in Figure
2.4 taken from Mattingly, Heiser, and Daley (1987). Butt separates the entire mission (originally
14 segments) into 22 segments to allow a more accurate estimation of aircraft behavior over the
entire mission (e.g., separates a subsonic-supersonic acceleration into 3 segments: subsonic,
transonic, and supersonic). The mission flown is a subset of the 22 mission segments in that
segments 3-21 are what are actually flown. The mission segments are detailed in Table 2.1.
Figure 2.4 Mission Profile by segment or leg (Mattingly, Heiser, and Daley, 1987).
Table 2.1 Mission Segment Definition and Description.
Mission Segments No. Name Description 1 Warm up 1 minute, military power 2 Take off 5 min at idle, take off + roll rotation 3 Subsonic Acceleration 1 Accelerate to climb speed 4 Subsonic Acceleration/Climb Climb at BCM/BCA 5 Subsonic Cruise 1 Cruise until cruise + climb range = 280 km 6 Combat Air Patrol/Loiter Patrol for 20 min and 9150 m and Mach for best endurance 7 Subsonic Acceleration 2 Accelerate to Mach 0.99 8 Transonic Acceleration 1 Accelerate to Mach 1.2 9 Supersonic Acceleration 1 Accelerate to Mach 1.5
10 Supersonic Penetration Mach 1.5 for 185 km 11 Combat Turn 1 Mach 1.6, one 260 degree 5g turn at max power
19
No. Name Description 12 Combat Turn 2 Mach 0.9, two 360 degree 5g turns at max power 13 Subsonic Acceleration 3 Accelerate from M = 0.9 to 0.99 14 Transonic Acceleration 2 Accelerate from M = 1 to M = 1.2 15 Supersonic Acceleration 2 Accelerate from M = 1.2 to M = 1.6 16 Deliver Expendables Deliver 2 AMRAAM, 2 AIM-9L, and ½ ammunition 17 Escape Dash M = 1.5 for 46 km 18 Supersonic Climb Climb to BCM/BCA from M = 1.5 to M = 1.2 19 Transonic Climb Climb to BCM/BCA from M = 1.2 to 1 20 Subsonic Cruise 2 Cruise at BCM/BCA for range of 278 km 21 Loiter Loiter at 9150m for 20 min at M for best endurance 22 Landing Land
2.2.1 Airframe Subsystem – Aerodynamics (AFS-A)
The aerodynamics model is developed from Raymer (1999), Mattingly, Heiser, and Pratt
(2002), Andersen (1998), and Nicolai (1975). Table 2.2 displays the airframe subsystem –
aerodynamics (AFS-A) calculations and model equations.
Table 2.2 AFS-A aerodynamics and model equations.
Component Variable Description Model Equation Master Flight Equation
T Thrust
g
Vh
dt
dWVDT
2
2
V Velocity
Drag
D Drag SqCD D
DC Drag coefficient
02
1 DLD CCKC
where 2min_
"0 min dragLDD CKCC
Induced Drag 1K
Induced drag factor (Nicolai, 1975)
Subsonic: eAR
K
11
Supersonic: LC
K1
1
Parasitic Drag 0DC
Parasitic Drag Coefficient
Subsonic: ref
wetfeDD S
SCCC min0
Supersonic: Dwaveref
wetfeD C
S
SCC
0
Wave Drag DwaveC
Coefficient of drag due to shock waves
S
qDC wave
Dwave/
20
Component Variable Description Model Equation
waveq
D
Wave drag efficiency factor
where for M = 1 to 1.2:
2max
2
9*2*2
l
A
q
D
q
D
HSwave
and for M > 1.2
HS
LEWD
waveq
DME
q
D
100
12.1386.0177.0
57.0
where HaackSearsHS
q
D
q
D
Exergy
PardesxE
Exergy destruction rate due to parasitic drag
T
VDTxE
Pardes00
where: SqCD D00
T
VDTxE i
desInd0
where: SqCD Dii InddesxE
Exergy Destruction Rate due to Induced Drag
The master flight equation is derived from a force balance on the aircraft. The drag on the
aircraft, D, must be overcome by the thrust, T. The drag coefficient equation for an uncambered
wing is also used in the model and the Wessinger span efficiency from Nicolai (1975) is used to
calculate the induced drag factor, K1. The lift curve slope, LC , is determined from supersonic
normal force curve slope charts presented in Raymer (1998), Nicolai (1975), and Anderson
(1999) where, for a high-performance uncambered aircraft dragLC min_ is approximately zero.
The parasitic drag coefficient, 0DC , takes on two different forms depending on whether
or not the aircraft is in the subsonic or supersonic flight regime. Cfe is the equivalent skin
friction coefficient and is chosen to be 0.0035 as recommended for an air force fighter (Raymer,
1998). Swet and Sref are the wetted areas of the aircraft and the exposed wing area, respectively;
and CDwave is the coefficient of drag due to the formation of shock waves during supersonic
flight. CDwave is calculated using a correlation for the Sears-Haack body wave drag where EWD is
an empirical wave drag efficiency factor which ranges from 1.8 to 2.2 for supersonic fighter
aircraft, M is the flight Mach number of the aircraft, and LE is the leading-edge sweep angle.
21
The exergy destruction rate in the AFS-A is due to the parasitic and induced drag losses,
and the equations are also shown in Table 2.2. The morphing AFS-A has a total of 4
synthesis/design and 72 operational decision variables because for each of the 18 mission
segments flown, 4 wing parameters are varied including wing sweep, wing length, root chord
length, and tip chord length. The fixed-wing AFS-A had a total of 4 synthesis/design decision
variables.
2.2.2 Propulsion Subsystem (PS)
The propulsion subsystem (PS) used in Butt (2005) is from Saravanamutto, Rogers, and
Cohen (2001). A schematic of an afterburning turbojet engine is shown in Figure 2.5. The
turbojet PS has 3 synthesis/design decision variables which are the compressor design pressure
ratio and the design corrected mass flow rates for the compressor and turbine. There are a total
of 44 operational decision variables for the PS: 18 compressor pressure ratios, 18 burner fuel/air
ratios, and an additional 8 afterburner fuel/air ratios.
Figure 2.5 Schematic of a single-spool turbojet engine with afterburner.
2.2.3 Most Important Results from Butt (2005)
The results of the large scale optimization show that the benefits of wing morphing are
very promising. The morphing aircraft take-off weight was established by multiplying the fixed-
wing weights and take-off fuel by penalty factors due to the morphing actuator weights and
power requirements. Figure 2.6 shows a sensitivity study of the morphing-wing fuel
consumption with respect to weight and fuel penalty factors. The optimum fixed-wing fuel
consumption is also displayed.
Fuel
Compressor Shaft
Burner
Burner
Turbine
Air
Diffuser
Afterburner
Products
Nozzle
22
Figure 2.6 Morphing-wing and fixed-wing fuel consumption comparison (Butt, 2005).
Figure 2.6 shows the wing weight penalty factors as well as the fuel consumption penalties. For
example, a morphing-wing aircraft that has wings 4 times the weight of the fixed-wing and
carries 25% more fuel at take off for actuator power uses approximately 2900 lb of fuel, which
shows an improvement of 10% over the fixed-wing aircraft at 3200 lb of fuel. Figure 2.6
indicates that even for very unreasonable morphing-wing weight and actuator power
consumption penalties (up to a factor of 7 for wing weight and a factor of 2 for the fuel weight) a
better-performing fighter aircraft can be attained by employing morphing technology.
2.3 Decomposition Strategies for Large-scale Aircraft Synthesis/Design
Optimization
The work of Muñoz (2000) illustrates multiple optimization algorithms as well as
different autonomous decomposition strategies for optimizing highly coupled, highly dynamic
energy systems. The optimization algorithms discussed and used include the following gradient-
based algorithms: the Method of Feasible Directions (MFD) and Sequential Quadratic
Programming (SQP) and the following nongradient-based optimization algorithms: simulated
annealing (SA) and Genetic Algorithms (GAs). The decomposition discussed and utilized
include time decomposition and physical decomposition. The latter in the literature can
generally be described as Local-Global Optimizations (LGO). An exception to this, is the
strategy developed by Muñoz (2000) and Muñoz and von Spakovsky (2000, a, b, c, d; 2001 a, b)
called Iterative Local-Global Optimization (ILGO) which is unique and addresses the major
23
drawback to LGO which is its large computational burden. Decomposition methods allow the
designer to create more complicated system models than would otherwise be possible for use
with large-scale optimization.
LGO works by physically decomposing a system into a set of units for which a set of
unit-level optimization sub-problems are established within the context of a reduced (i.e. smaller
than the original) system-level problem. This results in a set of a nested optimization problems
(i.e. the unit-level within the overall system-level) as well as a system-level optimum response
surface (ORS) which can be constructed implicitly or explicitly on the basis of the many unit-
level interactions (or coupling functions) between units. A search of this surface results in the
global optimum solution for the system.
Figure 2.7 illustrates the system-level ORS for a two-unit system as well as the unit-level
ORS generated by the sub-problem optimizations which in turn result in the system-level ORS
and the large computational burden with which LGO is strapped. In this figure, u12, and u21
represent the coupling functions between units 1 and 2; 2z is the set of local or unit-level decision
variables used in the sub-problem optimization of unit 2; and 2z , 2z
, and 2z are the objective
functions for the unit- and system-level optimization problems. The * superscript indicates
either a unit- or system-level optimum.
Now, in contrast to LGO, ILGO significantly reduces the computational burden by
eliminating the need for nested optimizations. It does so by embedding the system-level
optimization at the unit-level and in the process transforms the unit-level optimization sub-
problems into so-called system-level, unit-based optimizations. This eliminates the need to
implicitly or explicitly construct the system-level ORS which ILGO accomplishes by using ORS
slope information in the form of shadow prices as well as changes in the coupling functions to
move towards the system-level optimum. It is these shadow prices and coupling function
changes which embed the system-level information at the unit-level.
24
Figure 2.7 Unit- and system-level optimum response surfaces (ORSs; Rancruel, 2003).
Muñoz (2000) and Muñoz and von Spakovsky (2000 a, b, c, d, 2001 a, b) apply ILGO to
the optimization of a supersonic aircraft consisting of the following integrated subsystems: the
propulsion subsystem (PS), the environmental controls subsystem (ECS), the airframe
subsystem-aerodynamics (AFS-A), the expendable and permanent payload subsystems (EPAY
and PPAY), and the equipment group (EG). These authors show that ILGO can be used as an
effective physical decomposition strategy for large-scale optimization by comparing the results
from a mission-integrated synthesis/design optimization of the supersonic aircraft with and
without the use of ILGO. The results of this comparison show that the final system optimums
are within 0.5% of each other. Note, however, that in order to make this comparison, the actual
aircraft optimization problem with 153 degrees of freedom (109 for the ECS, 44 for the PS, and
0 for the AFS-A, EPAY, PPAY, and EG) had to be reduced down to one with only 52 degrees of
freedom (38 for the ECS, 14 for the PS, and 0 for the AFS-A, EPAY, PPAY, and EG) in order to
be able to solve the problem in a reasonable time frame without decomposition. Thus, for the
comparison, the optimizations with and without decomposition are based on this latter number.
A more involved application of ILGO and further validation of the effectiveness of this
approach is found in Rancruel (2003) and Rancruel and von Spakovsky (2006, 2004 a, b, 2003 a,
b). In their work, a more complete fighter-aircraft model with five subsystems with optimization
25
degrees of freedom (DOF) and three subsystems without DOF is optimized. A list of the
subsystems included is as follows:
PS – modeled with a modern performance code for on- and off- design performance
(FAST™ by Honeywell);
ECS – bootstrap type subsystem;
FLS – modeled as a transient subsystem;
Vapor Compression / Polyalphaolefin Loops Subsystem (VC/PAOS) – a vapor
compression refrigeration cycle with high and low temperature heat exchanger networks;
AFS-A - the only non-energy based subsystem; and,
EPAY, PPAY, and EG – involving no optimization DOF.
ILGO is used to physically decompose the system in order to be able to do the large-scale
synthesis / design of the tactical fighter aircraft, and optimization involving 493 DOF. ILGO
permits what had never been done before for highly dynamic, non-linear systems, namely a close
approach to the theoretical condition of “thermoeconomic isolation (TI)” (Frangopoulos and
Evans, 1984; von Spakovsky and Evans, 1993). TI is defined as the condition under which each
subsystem (or unit resulting from physical decomposition) of a system can be optimized
independently of the other subsystems and yet still result in a system optimum identical to what
would be attained without decomposition. For a more complete discussion, the reader is referred
to Rancruel (2003). Furthermore, as shown in Muñoz (2000), ILGO can be effectively used for
the large-scale optimization of energy systems consisting of several subsystems and many DOF,
while Rancruel (2003) shows that ILGO can also be extended to non-energy based subsystems
such as the AFS-A. By including the AFS-A with DOF in the aircraft synthesis / design
optimization process, overall aircraft performance can be improved.
The mission profile used in the synthesis / design optimization is the same as shown in
Figure 2.4 and used in Butt (2005). However, the mission is split up into segments slightly
differently than in Butt. The mission segment details are shown in Table 2.3.
The decomposition of the model is shown in Figure 2.5 along with the coupling functions
for the aircraft model used by Rancruel. The AFS-A uses the component buildup method for
parasitic drag estimation detailed in Raymer (2000). The supersonic parasitic drag is estimated
26
Table 2.3 Mission specifications (Rancruel, 2003).
Phase Description
1 Warm-up and take off, field is at 600 m pressure altitude with T=310 K. Fuel allowance is 5 min at idle power for taxi and 1 min at military power for warm-up. Take-off roll plus rotation must be ≤ 450 m on the surface with a friction coefficient = 0.05. STALLTO VV *2.1
2 Accelerate to climb speed and perform a minimum time climb in military power to best cruise mach number and best cruise altitude conditions (BCM/BCA)
3 Subsonic cruise climb at BCM/BCA until total range for climb and cruise climb is 280 km 4 Descend to 9150 m 5 Perform combat air patrol loiter for 20 min at 9150 m and a Mach number for best endurance 6 Supersonic penetration at 9150 m and M = 1.5. Range = 185 km
7
Combat is modeled by the following: Fire 2 AMRAAM missiles Perform one 360 deg., 5 g sustained turn at 9150 m, M=0.9 Accelerate from M = 0.8 to M = 1.6 at 9150 m at max. power Fire 2 AIM-9Ls and ½ ammo
Conditions at end of combat are M = 1.5 at 9150 m 8 Escape dash, at M = 1.5 and 9150 m for 46 km 9 Using military power, do a minimum time climb to BCM/BCA
10 Subsonic cruise climb to BCM/BCA 11 Subsonic cruise climb at BCA/BCM until total range from the end of combat equals 278 km 12 Descend to 3000 m 13 Loiter 20 min at 3000 m and a Mach number for best endurance
14 Descend and land, field is at 600 m pressure altitude with T = 310 K. A 2 s free roll plus braking distance must be ≤ 450 m. Runway has a friction coefficient = 0.18. STALLTD VV *15.1
using a correlation to the Sears-Haack body wave drag. The drag due to lift or induced drag is
estimated using the leading edge suction method also detailed in Raymer (2000). The thrust
requirements from the PS for each mission segment is determined using the master flight
equation (see Table 2.2) modified to account for individual subsystem drag due to the ECS air
requirements and the thermal management subsystem (TMS) which includes the VC/PAOS and
FLS. The empty weight of the aircraft is estimated using the component group weight method
from Raymer (1987) which was developed by studying previously built fighter aircraft and
applying regression analysis to the data.
The ECS is a bootstrap – type ECS similar to what is used in the F-16. The ECS provides
conditioned air to the pilot in the cockpit and the air-cooled avionics. This subsystem requires
bleed air from the PS and has two bleed ports: high and low pressure. Both bleed ports cannot be
active at the same time. The bootstrap-type ECS is defined as such because the compressor is
driven off its own process airflow rather than some other means of compressor power (such as
electrical power or shaft power from the PS).
27
Figure 2.5 Subsystems and subsystem coupling functions (Rancruel, 2003).
The aircraft also uses compact heat exchangers with off-set strip fins for the
thermodynamic model of all the heat exchangers; the F-16 uses the same type of heat
exchangers. Kays and London (1998), Shah and Webb (1982) and Shah (1981) provide the
thermodynamic models for the heat exchanger configurations considered here. The VC/PAOS
cools the portions of the avionics that require liquid cooling and uses the FLS as a heat sink as
well as a ram air heat exchanger as a cooling mechanism. The FLS is used as a heat sink for the
rest of the aircraft subsystems including the Central Hydraulic Subsystem (CHS) and the Oil
Loop Subsystem (OLS). Most of the fuel needed to cool the hydraulic subsystem and oil loop
subsystem is burned by the engine, but if any excess fuel is needed for cooling duties, it is cooled
in a ram air/fuel heat exchanger before being returned to the fuel tank.
The evolution of optimal weight with respect to the ILGO iterations for the takeoff
weight, AFS-A weight, fuel weight, and PS weight are shown in Figure 2.6. Notice that the
slope of the weight vs. ILGO iteration curve becomes nearly zero as the ILGO iterations increase
28
for all the subsystems displayed. This indicates an optimum has been reached, even possibly the
global optimum.
Figure 2.6 Evolution of the gross take-off weight, fuel weight, AFS-A weight, and PS weight at different
points of the iterative local-global optimization (ILGO) approach (Rancruel, 2003).
Finally, the optimum aircraft configuration is compared to the aircraft design proposed by
Mattingly, Heiser, and Daley (1987) to validate the results. The comparison is shown in Table
2.4. It is clear that the ILGO approach yields a superior aircraft to the one proposed by
Table 2.4 Comparison between the optimum ATA and the aircraft proposed by Mattingly, Heiser, and Daley (1987), (Rancruel, 2003).
Optimum Mattingly, Heiser, and Dailey (1987)
TOW , (lb) 22,396 23,800
FuelW , (lb) 7,194 7,707
refSTOW / , (lb/ft2) 61.49 64
TOSL WT / 1.13 1.27
refS , (ft2) 364.2 371.87
SLT , (lb) 25,306 30,226
AFSW , (kg) 3,100 4,200
29
Mattingly, Heiser, and Daley (1987). Furthermore, the results of Rancruel (2003) are reasonable
and the models behave as they should and are, thus, validated.
The objective functions used by Rancruel include minimize gross takeoff weight and
fuel consumption. Total cost was also defined as a system-level objective function; however, the
work of Muñoz (2000) shows that cost is linearly related to gross takeoff weight and is, thus, not
explicitly used as an objective function.
More work has been done with optimizing different objective functions in supersonic
fighter aircraft applications. The thesis work of Periannan (2005) investigates multiple objective
functions including exergy destruction minimization and the effects of the overall optimum
found with respect to allowing non-energy based subsystems to participate in the optimizations.
2.4 Effects on Aircraft Synthesis / Design of Different Objective Functions
The work of Periannan (2005) and Periannan, von Spakovsky, and Moorhouse (2008)
investigated five different objective functions, or figures of merit, for the analysis of a fighter
aircraft. The comparison is constructed to study the differences, if any, in the optimum vehicle
found for each objective function. The aircraft subsystem is assumed to consist of three
subsystems: a PS, an AFS-A, and an ECS. Initially, only the energy-based subsystems are
included with DOF in the optimization, namely, the PS and ECS. The AFS-A participates in the
optimization but without DOF. These optimizations are carried out for four of the five following
objectives (i.e. 1, 2, 4, and 5):
1. Minimize the gross take-off weight:
FUELECSPSETO WWWWW (2.1)
2. Minimize the exergy destruction in the PS and ECS plus exergy lost due to unburned fuel
loss:
FuelLossECSDEStotPSDEStotobj ExExExEx __2 (2.2)
3. Minimize the exergy destruction in the PS, ECS, and AFS-A plus exergy loss due to
unburned fuel loss:
FuelLossAAFSDEStotECSDEStotPSDEStotobj ExExExExEx ___2 (2.3)
4. Maximize thrust efficiency
fuelfuel
thrustthrust LHVm
W
(2.4)
30
5. Maximize thermodynamic efficiency
Max
FuelLossECSDEStotPSDEStotthrustthermo W
ExExEx
W
W __
max
1 (2.5)
where maxW is the maximum work rate the PS could provide if no other sources of losses existed
in the aircraft (an ideal system) or
FuelLossECSDESTotPSDESTotThrustMax ExExExWW __ (2.6)
Note that there is no AFS-A term included for the exergy destruction; however, it is included in a
second set of optimizations which include not only these AFS-A losses but AFS-A DOF as well.
These optimizations are discussed later in this section.
The PS used in all of these optimizations is a low-bypass afterburning turbofan. The
model equations are based on Mattingly, Heiser, and Daley (1987). Both the AFS-A and the ECS
are based on Rancruel (2003) and Muñoz (2000). All modeling for the AFS-A and PS is done
using the gPROMS™ dynamic modeling environment. The ECS is modeled using a C code
coupled to the gPROMS™ models for the other two subsystems. gPROMS™ is also used for all
optimizations. It uses a gradient-based approach, i.e. a sequential quadratic programming (SQP)
algorithm developed for mixed integer nonlinear programming (MINLP) problems.
The PS and ECS used by Periannan (2005) and Periannan, von Spakovsky, and
Moorhouse (2008) are described at length in Chapter 3 of the present thesis so their descriptions
will not be repeated here. The AFS-A model equations are similar to those described in Chapter
3. The only source of irreversibility modeled in the AFS-A is the parasitic drag or zero-lift drag
given by:
T
uDTEx Parasitic
DES AAFStot
0_
(2.7)
where 0T is the dead state temperature (set to the sea-level ambient temperature), ParasiticD is the
parasitic drag, u is the aircraft velocity, and T is the ambient temperature of the aircraft.
Prior to the optimizations, a parametric exergy analysis is performed on three
components in the PS to determine the effect of specific decision variables on the exergy-based
objective functions. The compressor pressure ratio, fan bypass ratio, and turbine inlet
temperatures are varied to see their effect on the exergy destruction rate, specific thrust, and
specific fuel consumption (SFC). Four different mission segments are examined including
31
warm-up/takeoff acceleration (mission segments 1 and 2), climb (mission segment 5), and
supersonic penetration (mission segment 8). Only the parametric study for the supersonic
penetration mission segment is presented here. Figure 2.7 shows the effects on the PS exergy
destruction rate and specific thrust with the variation in the three PS design decision variables.
Figure 2.7 Variation of vehicle specific thrust and exergy destruction rate with fan bypass ratio and turbine inlet temperature for a fixed compressor pressure ratio of 8 for the supersonic penetration mission segment (Periannan, 2005).
Figure 2.7 proves to be very informative as to the behavior of the PS with respect to the
exergy-based objectives. Notice that the exergy destruction rate for a given specific thrust (and
compressor ration) generally decreases with decreasing turbine inlet temperature and bypass
ratio. The trade-off that this implies is that the lower bypass ratios result in better, more efficient
PS designs provided the turbine inlet temperature is lowered as well. Furthermore, mission
segment 8, the turbine inlet temperature has less of an effect on exergy destruction than the
bypass ratio. This, however, may not be the case for other mission segments as shown in
Periannan (2005).
Figure 2.8 shows the response of the exergy destruction rate and specific thrust to
changes in compressor pressure ratio and fan bypass ratio. The fixed parameter is the turbine
inlet temperature, which is set to 1700° K. It can clearly be seen that a higher pressure ratio is
beneficial for the overall performance of the PS since the highest pressure ratio results in the
highest specific thrust and lowest exergy destruction rate. Furthermore, the lower the losses (i.e.
the rate of exergy destruction), the smaller the vehicle with higher specific thrust, which has the
consequence of reducing the cost of the aircraft and the total fuel consumption.
32
Figure 2.8 Variation of vehicle specific thrust and exergy destruction rate with fan bypass ratio and compressor pressure ratio for a fixed turbine inlet temperature of 1700° K for the supersonic penetration mission segment (Periannan, 2005).
Finally, Figure 2.9 shows the effects on the PS exergy destruction rate and specific fuel
consumption variations with bypass ratio and compressor pressure ratio. The obvious conclusion
from this figure is that the highest pressure ratio and lowest bypass ratio produce the best
performing PS. However, the requirements for thrust may demand that more airflow be moved
by increasing the bypass ratio above the minimum constraint. In addition, note that at higher
compressor pressure ratios, the effects of bypass ratio on the exergy destruction rate and specific
fuel consumption is less than at lower pressure ratios.
Upon completion of the exergy analysis, Periannan (2005) and Periannan, von
Spakovsky, and Moorhouse (2008) return to the comparison of the optimization results for the
various objectives. As mentioned above, objectives 1, 2, 4, and 5 are compared to show which,
if any, of the objective functions yield a better overall vehicle. Recall that for this set of
optimizations only the PS and ECS have DOF. Figure 2.10 shows the optimum gross takeoff
weight yielded for each of the objective functions. All three runs for each objective function are
shown to give some confidence that a global optimum was found. The first optimum for each
objective function being higher than the subsequent two runs indicates that for the first
optimization of each objective that a local optima may have been found rather than the global
33
Figure 2.9 Variation of vehicle specific fuel consumption and exergy destruction rate with fan bypass ratio
and compressor pressure ratio for a fixed turbine inlet temperature of 1700 K for the supersonic penetration mission segment (Periannan, 2005).
optimum. However, the second two runs being nearly identical indicate that the global optimum
was likely found for all four objective functions. The result that none of the objective functions
yield a better aircraft is due to the fact that although the AFS-A participates in the optimization,
it has no DOF. In Butt (2005) and Brewer (2006), the thrust efficiency objective function
consistently yields the poorest design in terms of performance and fuel usage, the one for a
supersonic fighter and the other for a hypersonic vehicle. The difference in these optimizations
is that AFS-A DOF are included.
The next set of optimizations conducted in Periannan include AFS-A DOF. Only
objective functions 1 and 3 are used. Using objective function 5 instead of 2 allows the inclusion
of exergy losses (equation (2.7)) due to the AFS-A. Comparisons of the optimum results include
those with and without AFS-A DOF. Figures 2.10 and 2.11 summarize these results. For
example, the former shows that optimizing using objective 3 with AFS-A DOF reduces the gross
takeoff weight by about 4.5% over the optimum aircraft found without AFS-A DOF and with
objective 1. This difference is even more pronounced, as seen in Figure 2.11, where the fuel
weight is reduced by 9.8%. Even in a comparison with the optimum vehicle found using
objective 1 with AFS-A DOF, objective 3 with AFS-A DOF produces an optimum vehicle with a
5.8% reduction in fuel weight.
34
Figure 2.10 Optimum gross takeoff weight with and without AFS-A DOF for objectives 1, 2, and 3
(Periannan, von Spakovsky and Moorhouse, 2008).
Figure 2.11 Optimum fuel weight with and without AFS-A DOF for objectives 1, 2, and 3 (Periannan, von
Spakovsky, and Moorhouse, 2008).
Finally, although the work of Periannan (2005) and Periannan, von Spakovsky, and
Moorhouse (2008) also show the strengths of the gradient-based optimization in terms of speed,
a weakness of this type of algorithm is the need of generating several feasible but very different
initial points with which to start the optimizations. In a complex large-scale optimization
5000
6000
7000
8000
9000
10000
11000
objective 1 objective 3 objective 2
Gro
ss T
akeo
ff w
eig
ht
in k
g
With AFS-A DOF
Without AFS-A DOF
2000
2200
2400
2600
2800
3000
3200
3400
objective 1 objective 3 objective 2
Fu
el W
eig
ht
in k
g
With AFS-A DOF
Without AFS-A DOF
35
problem with many DOF, this can be a rather daunting task, to say the least. Even a single
starting point is problematic. The need for several very different ones is due to the susceptibility
of such algorithms to getting stuck at local optima. Genetic or hybrid genetic-gradient based
optimization algorithms are examples of how these difficulties can be overcome, and the
following sections provide illustrations of the application of the former to a hypersonic vehicle
using both exergy and non-exergy based objective functions.
2.5 Exergy Methods for the Development of High Performance Vehicle Concepts
The renewed interest in hypersonic vehicles has fueled new design technologies to be
employed in hypersonic vehicles. The work of Markell (2005) illustrates applying exergy
analysis and exergy destruction minimization to the hypersonic vehicle. Markell developed a 1-
D hypersonic vehicle model and a partial 3-segment hypersonic mission and a detailed exergy
model for his thesis work. Two objective functions are compared in the work: an exergy based
objective function and a more traditional objective function, namely, the thrust efficiency. The
hypersonic vehicle consists of two subsystems, a propulsion subsystem and an airframe
subsystem. The propulsion subsystem consists of a inlet, combustor, and nozzle component.
Markell performed multiple optimizations on just the propulsion subsystem for a single mission
segment and also optimized the entire vehicle for the partial three-segment mission. The 1-D
hypersonic vehicle is shown in Figure 2.13.
Figure 2.12 Hypersonic vehicle configuration (Markell, 2005).
The forebody serves as the means of compressing the incoming airflow with oblique
shock waves forming at the leading edge of the vehicle. The inlet further compresses the
incoming air with more oblique shocks forming off subsequent turning angles before entering the
combustor / cowl area of the propulsion subsystem. The forebody and inlet design is extremely
important in hypersonic vehicle flight due to no other means available for compressing the air
36
entering the combustor. The forebody design is also instrumental in maximizing the mass
capture into the combustor as well.
Figure 2.14 shows a diagram of the design variables associated with the forebody and
inlet of the hypersonic vehicle. The operational decision variable shown in the figure is the
angle of attack, or . Energy exchange is also included in the forebody and inlet model to
modify the flow to maintain shock-on-lip conditions which maximizes the mass capture area into
the combustor when the forebody and inlet must operate at off-design conditions.
Figure 2.13 A physical representation of the forebody and inlet component of the hypersonic vehicle along with design and operational decision variables that govern the flow characteristics throughout the inlet (Markell, 2005).
The exergy model for the hypersonic vehicle is somewhat involved, as losses are tracked
for friction, shock waves, combustion, and mixing throughout the vehicle. The inlet exergy
destruction has two contributors to total exergy destruction: friction and shocks. The entropy
rate due to friction in the inlet is given by:
1
2lnt
tirr
P
PR
m
Sfric
(2.8)
where m is the mass flow rate, R is the specific gas constant, and 12 tt PP is the total pressure
ratio in the inlet. The entropy generation rate due to the three oblique shocks in the inlet is
expressed by:
1
3
11 lnln31i
ti
ti
i
ipi
irr
P
PR
T
Tc
m
Sshocks
(2.9)
where the three oblique shock temperature and pressure properties are required to find each
oblique shock entropy rate. The total irreversibility rate for the inlet then becomes:
37
31
shocksfricinlet irrirrirr SSS (2.10)
The combustor is a constant area, hydrogen-fueled combustor in which the flow is
constrained to stay supersonic, defining a scramjet-type propulsion subsystem. The flow can
become subsonic if the combustor length is too long, inlet area is too small, or the initial
combustor flow Mach number is too low (designated 4M in Figure 2.14). The mixing equations
account for incomplete combustion but do not include species dissociation due to the
computational burden required to track dissociation. Heat loss from the combustor through the
combustor walls is also tracked using the Eckert Reference Enthalpy Method (Heiser and Pratt,
1994). The total irreversibility rate for the combustor is then
mixirrinccombirrhtirrfricirrCOMBirr SSSSS (2.11)
where the terms to the right of the equals are the entropy generation rates due to irreversible
losses resulting from friction, heat transfer, incomplete combustion, and mixing, respectively.
The nozzle heat losses are modeled using the Eckert Reference Enthalpy Method (Heiser
and Pratt, 1994), with an adjustment to the overall heat transfer rate, as it is known to over
predict the nozzle heat transfer rate. Half the heat transfer rate calculated by the Eckert method
is used to account for the nozzle plume separating from the vehicle in flight as suggested by
Riggins (2003). The frictional losses are calculated with a skin friction coefficient suggested by
Riggins (2004). Including the mixing and heat transfer in the nozzle, a control volume of the
nozzle an entropy generation of:
w
wallj
jj
j
jjmixnozzirr T
Q
P
PRsysymS
6
7
16
17
log (2.12)
where jy and js indicate the mole fraction and entropy of constituent j and the subscripts 6 and
7 indicate the nozzle entrance and exit conditions, respectively.
The airframe subsystem-aerodynamics (AFS-A) is modeled to account for pressure wave
forces, center of gravity, as well as frictional forces from skin friction. A simple diamond-airfoil
wing and elevons are also modeled with the AFS-A to provide additional lifting surface and to
balance vehicle moments during flight. Shock expansion theory is employed to calculate the
drag and lift of the diamond airfoils (Anderson, 2001). The total drag on the vehicle is converted
38
to a frictional force, fF , on the AFS-A which yields the total entropy generation rate due to
friction, i.e.
t
firr T
uFS
fric (2.13)
where u is the speed of the aircraft and tT the local skin temperature. The total AFS-A entropy
generation rate including shock losses is:
31shocksfricaero irrirrirr SSS (2.14)
The combustor is initially optimized for a single mission segment to compare to a
published hypersonic combustor model from Riggins (1997). The model in Riggins (1997) was
presented for a single mission segment optimization, so it was necessary to only optimize a
single mission segment for validation. The results show good agreement (see Table 2.5), so the
combustor model is considered validated. The model is then used to develop a number of
parametric studies which are not repeated here. The x(m) in Table 2.5 represents the station in
the vehicle. Zero meters is the inlet entrance, five meters is the combustor entrance, six meters is
the nozzle entrance, and eleven meter is the engine exit. M, T, P, and u are the Mach number,
temperature, pressure, and velocity of the flow in the vehicle, respectively.
Table 2.5 Comparison of the optimal combustor models (Markell, 2005).
Riggins Model (1996) Markell (2005)
x (m) M T (K) P (N/m2) u (m/s) M T (K) P (N/m2) u (m/s)
0 12.0 200 1000 3400 12 200 1000 3402.3
5.0 6.22 679 70900 3255.4 6.26 674.4 70417 3259.2
6.0 1.72 4350 647000 2283 1.73 1.73 645804 2288.3
11.0 5.12 1110 5500 3423 5.11 5.11 5541 3423.9
Next the optimum scramjet engine is determined for a fixed thrust based on two different
objectives. The first is an energy based figure of merit called overall the overall efficiency,
which is defined as
pthprf hm
Tu
0
0 (2.15)
where 0 , T, u0, fm , prh , th , and p are the overall efficiency, engine thrust, vehicle velocity,
fuel mass flow rate, fuel enthalpy, and thermodynamic and propulsive efficiencies, respectively.
39
The second objective function is the exergy based optimization objective function that uses
concepts from both the 1st and 2nd laws of thermodynamics. This objective function is defined as
trelease
irr
T
TQ
ST
0
0
1
1
(2.16)
where is the thermodynamic efficiency, 0T is the dead state temperature (see Gyftopolous and
Beretta, 1991), irrS is the (irreversible) entropy generation within the system, releaseQ is the
heat loss to the environment, and tT is the temperature of the system at the point of heat loss.
The results for these two optimizations are shown in Table 2.6. The two different objective
functions produce nearly the same optimum scramjet engine in that the fuel mass consumed fm ,
is nearly identical between the two optima. The variables in Table 2.6 are shown on the
hypersonic vehicle in Figure 2.4.
Table 2.6 Optimal decision variable values for the energy and exergy based optimizations of a scramjet engine with fixed thrust (Markell, 2005).
Obj. Function )(mX fb )(mX cowl )(mX ramp )(fb )(nozz
Obj. 1 8.6114 13.652 2.9049 4.6026 17.066 Obj. 2 8.4019 13.896 3.1784 4.5669 17.549
)(mLcomb cowl% )/( skgm f )( 2mveh
Obj. 1 1.1148 0.0016 2.6363 0.5 26.039 Obj. 2 1.1523 0.0113 2.6754 0.5 26.025
The partial mission flown consists of three segments: a Mach 8 cruise for 1000 nm, acceleration
and climb from Mach 8 to Mach 10 in less than 90 seconds, and finally a Mach 10 cruise for
1000 nm. Initially, each mission segment was optimized individually to find the best hypersonic
vehicle for a given mission segment based for each of three separate objectives: thrust efficiency,
exergy destruction rate, and the exergy destruction rate plus the rate of the exergy loss. The
results of these optimizations are shown in Table 2.7. Notice that the exergy destruction
objective function produces combustor lengths at or near the minimum of 0.5 m. This is due to a
deficiency in the second objective function which ignores the rate of fuel exergy lost out the back
end of the engine. Nonetheless, the results for all three objectives show that the combustor is the
largest source of exergy destruction.
40
Table 2.7 Optimal design decision variable values for the single segment optimizations (Markell, 2005).
Mach #
Obj. Funct.
)(mX fb )(mX cowl )(mX ramp )(fb )( )(nozz )(mLcomb cowl%
8 0 8.5303 12.619 2.4418 2.5501 1.7227 17.512 0.8933 0.0207
desxE 8.4284 14.313 3.1839 1.0000 0.9547 15.524 0.5000 0.0736
8-10
0 8.6716 13.616 3.2772 1.4914 1.2096 17.999 0.9451 0.0009
desxE 8.4101 13.441 2.7657 1.0365 2.2349 17.335 0.5007 0.0860
lossfueldes xExE 8.5617 12.603 2.6580 1.0000 1.1521 12.8238 0.8802 0.2432
10
0 8.4094 11.969 2.4620 1.7887 1.2350 15.842 0.7017 0.0136
desxE 8.5113 14.267 3.0703 1.2493 0.9818 16.057 0.5003 0.2272
lossfueldes xExE 8.4127 13.265 3.2994 1.0577 1.7537 15.537 0.7775 0.2019
The final comparison made in Markell (2005) is between optimizations to determine the
optimal hypersonic vehicle that can fly all three mission segments of the partial mission
optimally. The optimal operational decision variable values for the three mission segments are
shown in Table 2.8 and each vehicle optimal fuel mass flow rate is compared in Table 2.9. The
interesting result is that the third objective function produces a vehicle that flies at shallower
angles of attack, , and has a lower fuel mass flow rate, fm , than either of the other two
Table 2.8 Optimal operational decision variable values for the partial mission (Markell, 2005).
Objective Function
)( )( 2mS wing
1 2 3 1 2 3
0 1.7724 1.2 0.9 7.2033 5.76515 9.3670
desxE 1.4453 0.8 0.8 7.6983 9.2350 0.4869
lossfueldes xExE 1.1933 0.4 0.3 0.7908 3.2372 0.5065
Table 2.9 Optimal vehicle fuel mass flow rate comparison (Markell, 2005).
Objective Function 0
lossfueldes xExE
1 0.5434 0.5779
2 1.9856 2.3892
3 0.6944 0.7070
optimizations. In Table 2.8 the optimal vehicle for the third objective function requires much
less effective wing area, wingS , to fly the mission than that for the optimal vehicle based on the
41
first objective. In Table 2.9, the fuel mass flow rates required are compared and show that the
third objective (one of the exergy-based objectives) yields a better performing hypersonic vehicle
with an overall savings of 6.5% fuel over entire mission. The work of Markell lays the
framework for Brewer (2006), who uses the 1-D hypersonic vehicle to fly an entire hypersonic
mission.
2.6 Integrated Mission-Level Analysis and Optimization of High Performance
Vehicle Concepts
The work of Brewer (2006) involves the construction (in collaboration with Markell,
2005) of a 1-D hypersonic vehicle model and using a genetic algorithm (GA) to find the
hypersonic vehicle configuration that flies a Mach 6 to Mach 10 flight envelope in the most
efficient manner. All the hypersonic vehicle designs found to date in the literature are based on a
single mission segment or flight condition. The vehicle design in Brewer (2006) is based on a
mission which includes cruise, acceleration / climb, deceleration / descend, and turn mission
segments. This thesis work furthermore includes a comparison between three objectives: thrust
efficiency maximization (the traditional propulsion subsystem design optimization objective),
minimization of fuel mass consumption (the traditional weight-based design optimization
objective), and the minimization of exergy destruction plus fuel exergy loss (the non-traditional
design optimization objective).
As in Markell (2005), the 1-D hypersonic vehicle model consists of two subsystems: the
propulsion and airframe subsystems. Irreversible loss mechanisms modeled to account for
exergy destruction include losses due to shocks, friction, heat transfer, mixing, and incomplete
combustion. The airframe modeling includes trim and force accounting, while the inlet includes
the modeling of energy addition of subtraction to or from the flow in order to maintain shock-on-
lip operating conditions at all operating points. The hypersonic vehicle is shown in Figure 2.5.
The mission flown during the vehicle design optimizations is shown in Figure 2.16 and the
corresponding flight details are given in Table 2.10. The hypersonic vehicle has a total of seven
design decision variables including forebody position, fbX , cowl position, cowlX , ramp 1
position, 1rampX , forebody angle, fb , combustor length, combL , nozzle expansion angle, nozz ,
42
Figure 2.14 Propulsion subsystem components and airframe subsystem (Brewer, 2006).
Table 2.10 Mission segment details (Brewer, 2006).
Segment Description
1 Accelerate and climb from Mach 6 (at 23.2 km) to Mach 8 (at 26.9 km), t = 90 sec
2 Mach 8 cruise for 600 sec
3 Accelerate and climb from Mach 8 (at 26.9 km) to Mach 10 (at 30.0 km), t = 90 sec
4 Mach 10 cruise for 600 sec 5 Perform a 180°, 2g sustained turn at Mach 10 6 Descend and decelerate to Mach 6
Figure 2.15 Total scramjet vehicle mission (Brewer, 2006).
and percent nozzle length, nozz% . The single operational decision variable is the angle of attack
of the vehicle, 0 . The design and operational variables for the hypersonic vehicle are given in
Table 2.11.
43
The first optimization objective used is maximizing thrust efficiency. The overall
mission thrust efficiency is found from a weighted average of each mission segments thrust
efficiency, or
5
0
5
iif
iiif
tm
tm
i
i
(2.15)
Table 2.11 Mission design and operational decision variables for the inlet, nozzle, combustor, and airframe (Brewer, 2006).
where fim is the fuel mass flow rate, it is the segment time, and i is the mission segment thrust
efficiency. The second optimization objective function used is minimize fuel mass burned which
is defined as
5
oiififuel tmgW . (2.16)
where g is the gravitational constant. The final optimization objective function is that of
minimizing the exergy destruction plus exergy lost from unburned fuel in the combustor and is
defined as
5
0,,,,0
iiilossitotalirrilossDes txESTExEx (2.17)
where itotalirrS ,, is the total rate of entropy generated by the vehicle for the mission segment, i,
while ilossxE , is the rate of exergy loss due to unburned fuel in segment i. The irreversibilities
44
included in the total rate of entropy generated are those for the inlet, combustor, nozzle, and
airframe. The vehicle was optimized using a genetic algorithm (GA) Queuing Multi-Objective
Optimizer (QMOO) (Leyland, 2002). This type of algorithm is more suited to a hypersonic
vehicle optimization problem than a gradient-based optimizer due to the nature of the problem:
highly constrained, mixed-integer variables, and non-linear spaces in the solution space would
cause a gradient based method to often get stuck in local optima. There is also the difficulty of
finding one or more initial points in such a highly constrained problem with a gradient based
approach. QMOO was developed by Leyland and Molyneaux and Laboratoire d’Energetique
Industrielle (Laboratory of Industrial Energy Systems, LENI) at the Ecole Polytechnique
Federale de Lausanne (EPFL). Both the hypersonic vehicle and QMOO are developed in
MATLAB enabling a straightforward coupling of the GA and the model.
Due to the random nature of the GAs, multiple optimizations are needed to establish that
the algorithm has indeed found the global optimum instead of a local optimum. This problem,
however, is not exclusive to GA optimizations since many gradient-based methods require
multiple runs as well to verify that a true optimum (whether it be the global or “best” local
optimum) has been found.
The model validation is performed by comparing against previously published results
(Riggins, 1996; Bowcutt, 1992; and Starkey, 2004). The model results of Brewer (2006) are all
well within reasonable tolerances of those reported in these references. Note also, that at the
time of publication, there was no known documentation of mission-level optimization results for
a hypersonic aircraft in the literature and very little for single-segment hypersonic optimizations.
After validation, the optimizations for the three different objective functions are run. The
hypersonic mission proves to be a very difficult problem to solve, as the optimizations took up to
two months to complete, running constantly from start to finish on personal computers with
processors ranging in speeds from Pentium III equivalent clock speeds of 1.5 to 3.05 Ghz with
512 to 1024 MB of internal memory. The computational cost is due to the sparse optimal
solution space for the hypersonic mission. This is in contrast to an “easier mission to fly” or a
system that has a dense optimal solution space. A random sample of the solutions for a
hypersonic aircraft mission versus a morphing-wing supersonic aircraft mission (Butt, 2005) is
shown in Table 2.12. The sparse optimal solution space only shows two feasible solutions in the
45
random sample, while the dense solution set shows feasible solutions in every sample space
shown. The average number of feasible solutions in the 125-member population for the
hypersonic optimization is just 10%, while the average number of feasible solutions in the 70-
member population for the morphing-wing optimization is nearly 100%.
The sparse optimal solution space requires the constraints of the variables shown in Table
2.11 be restricted further to allow the optimum solution to be found in a more timely manner (1 –
2 weeks). The initial forebody angle is found by the GA (QMOO), while the rest of the angles
Table 2.12 Samples of results populations for sparse and dense optimal solution spaces. Note that the very large numbers (i.e. E+15) represent infeasible solutions (Brewer, 2006).
Sparse Optimal Solution Space Dense Optimal Solution Space Min. Exergy Destruction Plus Fuel Exergy Loss
Population Sample [GJ] (Brewer, 2006) Min Fuel Usage Population Sample [kg] (Butt, 2005)
5.0000E+15 5.0000E+15 5.0000E+15
5.0000E+15 5.0000E+15 5.0000E+15
5.0000E+15 5.0000E+15 5.0000E+15
5.0000E+15 5.0000E+15 5.0000E+15
5.0000E+15 1.4005E+02 5.0000E+15
1.4008E+02 5.0000E+15 5.0000E+15
5.0000E+15 5.0000E+15 5.0000E+15
5.9966E+02 5.9854E+02 6.1114E+02
5.4164E+02 5.3917E+02 5.6801E+02
5.9471E+02 5.5250E+02 5.7282E+02
5.7131E+02 5.8799E+02 5.7668E+02
5.7147E+02 5.5866E+02 5.5966E+02
5.9706E+02 5.7667E+02 5.4164E+02
5.8924E+02 5.6730E+02 6.0444E+02
on the hypersonic body are found by iteratively trying different ramp angles until a feasible
solution allows progression to the next angle on the forebody. This method, although very
computationally expensive, is still faster than simply allowing the GA to find a solution by
setting all the ramp angles to optimization decision variables as was attempted by Muñoz.
Brewer discovered that one of the weaknesses exhibited by QMOO was its inability to
suppress significant digits assigned to decision variables. A large amount of computational time
is wasted due to the fact that for each decision variable, QMOO would only vary the last 12 to 15
significant digits of the decision variable from one generation to the next (unless the decision
variable is flagged as an integer, of course). Any decision variable that is not defined as an
integer automatically has 15 significant digits and Brewer could not find a way to modify this
easily within the MATLAB QMOO code.
Despite the large amount of initial work and required tailoring of the design/operational
decision variable constraints, Brewer is able to attain results for all three objective functions.
These results are given in Table 2.13. As can be seen from the table, the second and third
46
Table 2.13 Optimal objective function results (Brewer, 2006).
Objective Function
Max. Thrust Efficiency Min. Fuel Mass Min. Exergy Destroyed
+ Fuel Exergy Lost Run 1 2 3 1 2 3 1 2 3
Thrust Efficiency [%]
34.52 34.48 32.97 29.38 26.40 26.41 28.76 29.89 26.57
Fuel Mass [kg]
1911 1911 2227 1717 1720 1744 1732 1797 1801
Exergy Destruction
[GJ] 153.7 151.1 166.2 145.2 139.3 140.2 140.0 142.3 143.1
objective functions, minimize fuel mass and minimize exergy destroyed plus fuel exergy lost,
respectively, yielded very similar optima. The explanation for the similarities can be found in
one simple fact: the main source of exergy, or available energy, on the vehicle is found in the
fuel. If the fuel mass required to fly the mission is minimized this is, in reality, minimizing the
largest source of exergy destruction: fuel combustion. The thrust efficiency objective function,
which is purely a “first law of thermodynamics” performance metric, yields a significantly worse
performing vehicle in terms of fuel mass burned and exergy destruction than the other “second
law of thermodynamics” based objective functions, despite having a higher thrust efficiency than
the other two optimal vehicles.
47
Chapter 3
Model Description and Synthesis/Design Problem Description
This chapter discusses the synthesis/design problem as well as the subsystem models
used in this thesis work.
3.1 Problem Definition
Solving the air-to-air fighter (AAF) synthesis/design optimization problem starts with
developing appropriate subsystem models for the aircraft. Next, the mission is examined in
detail to enable mission segments (or logical pieces of the mission) to be defined. Each
subsystem is then prepared to “fly the mission” by defining the interactions with the other
subsystems. Individual subsystem model convergence is checked by running optimizations on
the decoupled or non-interacting subsystems before they are integrated. Finally, the subsystems
are integrated and the coupling functions and optimization synthesis / design and operational
decision variables are defined. Multiple optimizations for each different objective function and
AFS-A configuration (i.e. morphing / fixed-wing) are then run to verify that a global optimum
has been reached. The subsystems modeled in this thesis include the following:
Airframe subsystem – Aerodynamics (AFS-A);
Propulsion Subsystem (PS);
Environmental Controls Subsystem (ECS);
Fuel Loop Subsystem (FLS);
Vapor Compression / Polyalphaolefin Subsystem (VC/PAOS);
Oil Loop Subsystem (OLS);
Electrical Subsystem (ES);
Central Hydraulic Subsystem (CHS);
Flight Controls Subsystem (FCS).
The fighter aircraft mission flown is shown conceptually in Figure 3.1. The details of the
mission are derived from a request for proposal (RFP) found in Mattingly, Heiser, and Pratt
(2002) and the mission segments are given in Table 3.1.
48
Figure 3.1 Supersonic fighter aircraft mission from the RFP found in Mattingly, Heiser, and Pratt (2002).3
Table 3.1 Air-to-air fighter (AAF) mission segments and details.
Mission Segment
Description
1 Warm-up and take off, 2000 ft altitude, 1 min for military power warm up, takeoff ground roll + 3 s rotation distance < 1500 ft, STALLTO VV *2.1
2 Accelerate to best subsonic climb Mach (BCLM) 3 Minimum time to climb to 20,000 ft at military power 4 Continue minimum time to climb to best cruise Mach (BCM1) and best cruise altitude (BCA1) 5 Subsonic cruise until total distance for climb/cruise is 150 nmi 6 Perform combat air patrol loiter for 20 min at 30,000 ft and best mach for endurance (BCM2) 7 Accelerate to Mach 0.8 8 Accelerate to Mach 1.2 9 Accelerate to Mach 1.5, total time for acceleration t < 50 s
10 Supersonic Penetration at Mach 1.5 until total distance for accel + supersonic penetration is 100 nmi, supercruise if possible (no afterburning)
11 Combat segment: perform 360 degree, 5 g sustained turn at 30,000 ft, M = 1.6 12 Combat segment: perform two 360 degree, 5 g sustained turns at 30,000 ft, M = 0.9 13 Combat Segment: Accelerate from Mach 0.8 to Mach 1.0 in max power 14 Combat Segment: Accelerate from Mach 1.0 to Mach 1.2 in max power 15 Combat Segment: Accelerate from Mach 1.2 to Mach 1.6 in max power 16 Combat segment: drop payload of 2 AIM-9L’s and 250 rds of 25mm ammunition (1309 lb) 17 Escape dash at M = 1.5 and 30,000 ft for 25 nmi, supercruise if possible 18 Climb/decelerate to BCM and BCA at military power, no distance credit
19 Subsonic cruise at best cruise mach (BCM3) and best cruise altitude (BCA2) until total distance is 150 nmi from escape dash
20 Loiter for 20 minutes at 10,000 ft and best mach for endurance (BCM4) 21 Descend and land
3 ©AIAA, reprinted with permission.
49
Note that each best cruise Mach or best cruise altitude for a given mission segment is unique to
the other best cruise Mach or best cruise altitudes for the aircraft in different mission segments
(e.g. 21 BCMBCM and 21 BCABCA ). The subsequent sections in Chapter 3 detail the
previously bulleted subsystems that comprise the AAF, starting with the AFS-A.
3.2 Airframe Subsystem
The airframe subsystem-aerodynamics (AFS-A) is the largest subsystem in the AAF and
serves not only as the structure required to house the other subsystems but also as the subsystem
that produces lift and the aerodynamics required to fly the mission. The airframe houses all the
subsystems mentioned in Chapter 3 as well as the rest of the items not detailed in the following
sections. The items not detailed are accounted for simply by a fixed weight value.
The AFS-A is developed from Raymer (2000) as well as Rancruel (2003) and is based on
drag-polar relationships. The free-body diagram of the aircraft is shown in Figure 3.2. Analysis
of Figure 3.2 and an energy balance will yield the master flight equation which is given in Table
Figure 3.2 Free-body diagram of the aircraft (Rancruel, 2003).
3.2. The “clean” drag term of the master flight equation, D , is the drag due to AFS-A
aerodynamics. Additional drag from other subsystems (such as the FLS ram/air heat exchanger)
is accounted for in the term, R . The thrust, T , is the installed thrust of the propulsion subsystem
(PS) which is detailed in Section 3.3.
The lift, L , and drag, D , terms are determined using the lift-drag polar relationship from
Mattingly, Heiser and Daly (1987) and Raymer (2000) which is detailed in section 3.2.1. The
analysis of the mission follows the lift-drag discussion. The AFS-A weight equations are
50
presented in Section 3.2.3 and, finally, the calculation of the takeoff weight completes the
discussion of the AFS-A.
Table 3.2 Master flight equation and governing flight equations.
Component Variable Description Model Equations
Master Flight Equation
T Thrust
g
V
dt
dW
dt
dhWVRDT
2
2
D Clean Drag R Additional Drag V Aircraft Velocity W Aircraft instantaneous weight
Governing Flight Equation
T Installed Thrust
g
Vh
dt
d
VW
RD
W
T ii
i 2
1 2
iV Velocity of aircraft at segment, i
ih Altitude of aircraft at segment, i g Gravitational acceleration
3.2.1 Lift and Drag
The total drag on the aircraft is a combination of parasitic drag or “zero-lift drag” as well
as drag due to lift or “lift-induced drag”. This section details the lift and drag relationships and
equations that were used to develop the aerodynamics model of the AFS-A. In short, the drag
model used is the component buildup method called the “Parasite-Drag Buildup Method”
detailed in Raymer (2000). The supersonic wave drag is calculated using the Sears-Haack
supersonic wave drag correlation. A summary of the general equations for the lift and drag
relationships is given in Table 3.3.
Notice that the total drag on the aircraft is a combination of the drag due directly to the
performance characteristics of the aircraft as well as the additional drag from other subsystems
shown in the master flight equation in Table 3.2. The wave drag is calculated using the Sears-
Table 3.3 Lift and drag equations for the AFS-A.
Component Variable Description Model Equations
Clean Drag
D Clean Drag
DqSCD
where
2LDoD KCCC
q Dynamic Pressure
S Wing Reference Area
DC Total Drag coefficient
K Lift factor
Parasitic Drag DoC
Coefficient of parasitic drag
tailDofuselageDowingDoDo CCCC ___
canopyDowaveDomiscDo CCC ___
51
Component Variable Description Model Equations
Wave Drag waveDoC _ Coefficient of shock wave drag
HaackSears
LEwd
waveq
DME
q
D
100
12.1386.0177.0
57.0
S
q
D
C waveDowave
Lift-induced Drag
LC Coefficient of lift
qS
WC TO
L
Weight fraction of aircraft
Haack wave drag correlation shown in Table 3.3. The wave drag constant, wdE , is set to 2 in
this thesis work in order to correspond to currently built fighter aircraft. The clean drag is the
only drag accounted to the AFS-A in the exergy destruction calculations (see Section 3.2.5);
however, other subsystem drag must be included in the master flight equation to accurately
predict AFS-A performance.
3.2.2 Mission Analysis
The mission consists of many dissimilar mission segments as well as a payload drop.
Some important equations for the mission segments are given in Table 3.4 with a subsequent
discussion of the individual mission segments.
The first mission segment includes a 60 second warm-up at military power (full throttle
in the main burner, no afterburning), then takeoff/acceleration, and finally, a ground roll for
takeoff. The total distance for acceleration and ground roll is constrained to not exceed 1500 ft.
The final aircraft speed at the end of the ground roll is set to 1.2 times the estimated stall speed,
stallV , which is initially estimated in order to calculate the drag and lift characteristics. The actual
stall speed is recalculated, and the actual drag and lift parameters are iterated until convergence
to increase accuracy. The aircraft velocity at the end of segment 1 is 1.2* stallV .
The second mission segment is a horizontal acceleration segment with a specified time of
30 seconds. The aircraft accelerates from 1.2* stallV to the best climb Mach (BCLM). The best
52
climb Mach is a degree of freedom (DOF) for the second mission segment in the AFS-A
optimization problem.
Table 3.4 Mission segment model equations.
Component Variable Description Model Equations
Takeoff model
A Warm-up weight fraction 60*1 SFC
W
W
W
W
TO
A
i
fA
max2 LTO
Astall CS
WgV
stallB VV 1.1
2
/exp BB V
TRDg
SFC
stallc VV 2.1
g
VV
TRD
SFC BCC 2
50/
exp22
CBATOW
W 1
1
stallV Stall speed
maxLC Takeoff lift coefficient
BV Velocity at end of acceleration
B Takeoff acceleration weight fraction
CV Velocity at end of ground roll
1 Final weight fraction for takeoff
1W Weight of aircraft after mission segment 1
Climb model
G Glide ratio (vertical velocity/horizontal velocity) Ae
CG
W
T Do
2
where for 30LE
64.0*045.0178.1 68.0 ARe
for 30LE
1.3cos*045.0161.4 15.068.0 LEARe
SbAR
2
AR Aspect Ratio
e Weissinger span efficiency
DoC Coefficient of parasitic drag
b Wing span
S Wing reference area
Sustained turn model
d Time to turn 1
2
2
ng
Vxd
Aeq
SWn
SWn
Cq
D
L
Do
1
DLn
W
T
x Number of turns
n Load factor for turn
53
The third and fourth mission segments are sequential, constant Mach, minimum time-to-
climb segments in military power from 2000 ft to the best cruise altitude (BCA1). The climb
angle is determined by the excess thrust available from the PS. The horizontal distance traveled
for the third and fourth mission segments is tracked because the total distance for segments three,
four, and five is 150 nmi.
Mission segment five is a cruise segment at best cruise Mach (BCM1) and altitude
(BCA1). The BCM1 and BCA1 are AFS-A DOF. Mission segment six is a 20 minute loiter
segment at 30,000 ft at best Mach for endurance (BCM2), which is also an AFS-A DOF.
Mission segments seven, eight, and nine split up the acceleration from BCM2 to Mach
1.5: BCM2 to Mach 1.0, Mach 1.0 to Mach 1.2, and, finally, Mach 1.2 to Mach 1.5. Each of the
three segments is time constrained to 15 seconds for a total acceleration time of 45 seconds.
Mission segment ten is a supersonic penetration mission segment at Mach 1.5 that
continues until the total distance traveled for segments seven, eight, nine, and ten is 100 nautical
miles. Segment ten specifies to supercruise (supersonic cruise without afterburning) if possible.
The combat simulation consists of a single 5-g sustained turn at Mach 1.6 and two 5-g
sustained turns at Mach 0.9, acceleration from Mach 0.8 to Mach 1.6 at maximum power (full
throttle for the main burner and afterburner), and a payload drop of 1309 lbs which is calculated
from firing two AIM-9L’s and 250 rounds of 25 mm ammunition. The combat simulation ends
with the aircraft at Mach 1.6, and a supercruise (if possible) escape dash at 30,000 ft from 25
nautical miles follows the combat simulation.
Mission segment eighteen is a deceleration / climb at military power to best cruise Mach,
BCM3, and best cruise altitude, BCA2. This mission segment usually does not include a weight
drop as the energy equation for this segment involves a trade of kinetic energy for potential
energy. Mission segment nineteen is simply a 150 nautical mile cruise at BCM3 and BCA2.
The final mission segment before landing is a 20 minute loiter / observation mission
segment at best cruise mach and best cruise altitude until the total distance since the escape dash
is 150 nautical miles.
The descent and landing mission segment has a 6 minute time allowance with no weight
drop for the descent. The final weight fraction to account for the landing, ground roll, braking
and taxi is set to 997.0if WW . The weight fraction calculations are presented in the
following section.
54
3.2.3 Weight Fraction Model
The gross weight of the AAF decreases after each mission segment due to the burning of
fuel in the PS. The gross weight decreases are tracked in each mission segment using weight
fractions, or initial to final mission segment weight ratios that are calculated based on the aircraft
energy change over the mission segment, specific fuel consumption (SFC), and flight
conditions/aerodynamics.
The mission segments are first categorized into mission segment type: for example, loiter,
cruise, horizontal acceleration, constant speed climb, climb / acceleration, etc. The master energy
equation for the weight fraction calculation is shown in Table 3.5 as well as the main derivatives
of the master equation with a discussion following.
Table 3.5 Weight fraction model equations.
Component Variable Description Model Equations
Master weight fraction equation
if WW Final/initial weight ratio
g
Vh
uV
sfc
W
W
W
W
i
i
i
fi 2)1(
exp2
1
where
TRDu /
sfc Specific fuel consumption V Aircraft velocity g Gravitational acceleration
i Weight ratio for segment i
Climb h Altitude
huV
sfc
W
W
i
f
)1(exp D Clean drag
R Additional drag
Acceleration T Installed thrust
g
V
uV
sfc
W
W
i
f
2)1(exp
2
Cruise S Range for cruise
SCC
V
sfc
W
W
LD
i
fexp
Loiter t Time for loiter
tKCV
sfc
W
WD
i
f14exp
The weight fraction of the aircraft starts out at 1 and decreases throughout the mission
until the aircraft lands. The aircraft weight is the product of each mission segment weight
fraction, i , multiplied by the gross aircraft takeoff weight, TOW . Wing loading, SW , is a
strong function of weight fraction, and the average weight of the aircraft is used to calculate the
wing loading for each mission segment. The average weight fraction is iterated to increase
accuracy when a large weight drop occurs in a mission segment. The takeoff weight calculations
are detailed in the following section.
55
3.2.4 Calculation of WTO
The gross takeoff weight, TOW , of this model is primarily developed from component
weight estimates in Raymer (2000). The method of estimating fuel weight and empty weight
from Mattingly, Heiser, and Pratt (2002) was initially used. However, this method proved to be
somewhat slow and sometimes non-convergent for this application. The method used for this
thesis is to allow gross takeoff weight to be a design decision variable and participate in the
optimization. Using this method requires an initial fuel weight to be estimated to conceptually
size some of the components. For this work, the initial fuel weight was estimated to be 40% of
0TW . Each subsystem and component weight is then estimated to find the empty weight of the
aircraft, EMPTYW . The actual fuel weight is found by subtracting 0TW from EMPTYW . If the
aircraft runs out of fuel during the mission, the chosen TOW is simply thrown out and a new TOW
is chosen along with a new set of decision variable values.
The conceptual sizing method in Raymer (2000) details an iterative process to decrease
the amount of fuel at take-off to the amount required to fly the mission to avoid taking off with
excess fuel. This iteration is not used with the following justification: a lower TOW can be found
by simply picking a better AFS-A design. This avoids spending computational time trying to
make a likely “worse” solution improve by finding the actual amount of fuel the aircraft needs to
fly the mission. Allowing the 0TW to be a DOF forces the optimizer to find the takeoff weight
and corresponding aircraft design and operational decision variable values that minimize the
difference between fuel used for the mission and fuel at takeoff. Table 3.6 outlines the main
subsystem weight calculations and the final 0TW equation. Note that the Nomenclature section
defines the variables that are used but not defined in Table 3.6.
Table 3.6 Main subsystem weight calculations and 0TW .
Component Variable Description Model Equations
AAFSW
wingW Wing Weight (lb)
806.0260.00.2
_
4.0785.0622.05.0
10001316.3
)/()(0103.0
htzdg
h
wtailh
rootwzdgvsdwwing
SNW
B
FW
ctASNWKKW
tailhW _ Horizontal Tail
Weight (lb)
gearmainW _ Main Landing Gear Weight (lb)
gearnoseW _ Nose Landing Gear Weight (lb)
56
Component Variable Description Model Equations
fuselageW Fuselage weight (lb)
525.05.0290.0_
973.025.0_
)(
)(
nwnllgearnoze
mlltpgcbgearmain
NLNWW
LNWKKW
685.0849.05.025.035.0499.0 WDLNWKW zdgdwffuselage
eds
enddvginductionair
DLL
NKLKW
373.0
498.1182.0643.0_
/*
29.13
zSLenmountsengine NTNW 579.0795.0_ 013.0
inductionairW _ Inlet duct weight (lb)
mountsengineW _ Engine mount weight (lb)
PSW PSW Engine weight (lb)
81.0exp)(063.0 25.01.1 MTW SLPS
FLSW FLSW Fuel Loop subsystem weight (lb)
DuctingMechanicalTankFuelHXFLS WWWWW _
ECSW ECSW
Environmental controls subsystem weight (lb)
DuctingMechanicalHXECS WWWW
PAOVCW / PAOVCW /
Vapor compresson/PAO loop subsystem weight
MechanicalHXVCPAOS WWW
CHSW CHSW Central hydraulic subsystem weight
664.0887.16 uvshCHS NKW (kg)
OLSW OLSW Oil loop subsystem weight
023.182.37 enOLS NW
FCSW FCSW Flight controls subsystem weight
127.0484.0489.0003.0*28.36 cscsFCS NNSMW
ESW Rated
System generating capability (kW, or kVA)
091.010.010.0152.02.172 genacmcES NLNRatedKW
TOW Gross takeoff weight fuelFCS
ESCHSECSFLSVCPAOSECSPSAAFSTO
WW
WWWWWWWWW
3.2.5 Morphing-wing Considerations
The AFS-A must account for morphing-wing technology, not only with respect to
aerodynamic performance but also with respect to takeoff weight and in-flight weight fractions.
In order to account for aerodynamics performance, five wing geometry parameters are allowed to
vary over the mission:
Aspect ratio, A ;
57
Wing length, L ;
Wing sweep angle, LE ;
Thickness to chord ratio, ct ; and
Taper ratio, .
The AFS-A morphing is treated in a quasi-stationary sense in that the wing geometry may
change for each of the 21 mission segments, but the change is considered an instantaneous
change (no time allowance for wing geometry change is considered). In other words, each of the
mission segments has a unique wing configuration for the morphing AFS-A. Note that the
payload drop (mission segment sixteen) has no time allowance, thus, the wing configuration for
that segment has no effect on the aircraft aerodynamics or performance. The morphing-wing
AFS-A has a total of 20 unique wing configurations for the 21 mission segments flown. The
fixed-wing AFS-A has a single configuration for the entire mission and does not change wing
geometry during the mission.
The actuation cost and additional weight of adding morphing wings to the aircraft must
also be taken into account in the morphing-wing aircraft. This is done by using wing weight and
fuel penalties to account for the additional actuators required for morphing wings and the energy
required to morph the wings, respectively.
The wing weight penalties are factors that are multiplied by the equivalent fixed-wing
weight and are added to the aircraft takeoff weight. For example, if an aircraft wing with a
specified geometry weighs 2,000 lb and the morphing wing weight penalty is 2, the actual wing
weight will be 4,000 lb. A difficulty in calculating the wing weight of a morphing-wing is
picking the geometry at which to calculate the equivalent wing weight before multiplying by the
wing weight penalty. Analysis of the wing weight equation in Table 3.6 indicates that the largest
wing weight is yielded by using the following values of the morphing-wing geometry:
Lowest aspect ratio, A ;
Largest wing length, L ;
Largest sweep angle, LE ;
Smallest thickness to chord ratio, ct ; and
58
Largest taper ratio, .
Using this method to establish the wing weight puts the morphing-wing aircraft at the largest
disadvantage with respect to gross takeoff weight. Note that the work of Butt (2005) used a
sweep angle of zero to establish the equivalent fixed-wing weight. Thus, the sensitivity study
presented at the end of Butt (2005) should have larger wing penalties showing a greater
advantage over the fixed-wing weight than in this thesis work. Consequently, the performance
metric to compare the fixed-wing performance to the morphing-wing performance in this work
will be the gross takeoff weight, TOW .
The fuel weight penalties are used to account for the power required to morph the wings.
After the aircraft takeoff weight and total fuel weight is established, the fuel weight penalties are
multiplied by the total fuel weight to find the morphing fuel weight. The morphing fuel is
reserved for morphing the wings and is not available to the PS to fly the mission. For example,
if the aircraft has 10,000 lb of total fuel, and the fuel penalty is 25%, then there will be 2,500 lb
of “morphing fuel”, leaving 7,500 lb of fuel available to fly the mission. The morphing fuel is
expended over the mission at the same rate as the fuel penalty. For example, if the fuel penalty
is 25% and the aircraft burns 1000 lb of fuel on a mission segment, 250 lb of morphing fuel is
used as well (in addition to the 1000 lb used to fly the aircraft) to account for the energy required
to morph the wings. Note that if excess fuel is sized at takeoff, then excess morphing fuel will
also be sized at takeoff which implies that excess fuel will be carried throughout the mission.
The aircraft weight fractions are updated at the end of each mission segment to reflect the
morphing fuel utilized. The usage of the morphing fuel is not taken into account in the exergy
destruction calculations since how exactly it is being used, i.e. the details of the actuators, is
unknown at this design stage. The AFS-A exergy model is presented in the following
subsection.
3.2.6 AFS-A Exergy Model
The exergy destruction in the AFS-A is composed of two unique parts: the exergy
destruction due to parasitic drag and that due to lift-induced drag (or simply induced drag). The
parasitic drag and induced drag equations are shown in Section 3.2.1 and the exergy destruction
rate equations are given in Table 3.7.
59
The exergy destruction in the AFS-A was tracked for every mission segment. The only
mission segment that has zero exergy destruction is the instantaneous payload drop mission
segment 16 (see Table 3.1).
Table 3.7 AFS-A exergy destruction rate equations.
Component Variable Description Model Equations
Exergy destruction rate due to parasitic drag
PardesxE Exergy destruction rate due to parasitic drag
ambdes T
VDTxE
Par00
where Doparasite qSCD
0D Zero lift or parasitic drag q Dynamic pressure S Wing reference area
DoC Coefficient of parasitic drag
Exergy destruction rate due to lift-induced drag
InddesxE Exergy destruction rate due to lift induced drag
amb
ides T
VDTxE
Ind0
where iDi qSCD
2LD KCC
i
iD Drag due to lift or lift induced drag
LC Coefficient of lift
Lift factor
Tightly coupled with the AFS-A is the propulsion subsystem which is detailed in the
following section.
3.3 Propulsion Subsystem
The propulsion subsystem (PS) used in this thesis work is a low bypass afterburning
turbofan engine with the on- and off-design models based primarily on Mattingly, Heiser, and
Pratt (2002). The following subsections detail the PS model used in this thesis work.
3.3.1 PS Layout and Station Definitions
PS is made up of the following components: fan, high pressure compressor, burner /
combustor, high pressure turbine, low pressure turbine, exhaust mixer, afterburner, and nozzle.
The PS system layout and station definitions are shown in Figure 3.3. The station definitions
must be clear to track properties and operating conditions through the PS. Thus, Figure 3.4 and
Table 3.8 give further explanation and formally define the PS nomenclature. Notice in Figure
3.4 that the bleed air consumption is not shown.
60
Figure 3.3 Engine system layout (Rancruel, 2002).
Figure 3.4 Engine Station Definitions (Periannan, 2005).
The bleed air is used by the environmental control subsystem (ECS), which is detailed in Section
3.4 and Section 4.5.
Table 3.8 Low-bypass turbofan engine station definitions.
Station Description 0-1 Free stream to diffuser inlet 1-2 Diffuser 2-3 Fan entry to high pressure compressor entry 3 High pressure compressor exit 3’ Fan exit to bypass duct 3a Burner entry 4 Burner exit 4a Coolant mixer 1 exit, high pressure turbine entry 4b High pressure turbine exit 4c Coolant mixer 2 exit 5 Low pressure turbine exit, mixer entry 5’ Fan bypass duct exit to mixer entry 6 Exhaust mixer entry, afterburner entry 7 Afterburner exit, nozzle entry
7 8 9 1
Bleed air Cooling air #2
Cooling air #1
High pressure spool
Low pressure spool
Fan High
pressure Compressor
5 4c 4a
4 3a
3 3’ 2
Low pressure turbine
Coolant mixer 2
Last rotor
First rotor
Coolant mixer 1
Mixer After
Burner
5’
6
Burner Inlet Nozzle
61
Station Description 8 Exhaust nozzle throat 9 Exhaust nozzle exit
The turbofan engine is a much more complicated engine than the turbojet used in Butt
(2005) and requires flow and energy balancing through the engine and somewhat involved
iterations to find the off-design performance. In short, the PS is modeled with a reference or
“design” engine and each mission segment is considered as being “off design” at conditions at
which the engine was not specifically designed to operate most efficiently. The thermodynamic
model of the PS follows.
3.3.2 PS Thermodynamic Model
The thermodynamic models and design equations of the PS components are detailed in
this section. The off-design operation of the PS is used for every mission segment, while the
design calculations are used to “build” a reference engine. The off-design simulation requires
either an equation solver or iterative process. The latter is used for this thesis model. A
comprehensive list and order of calculation of the off-design equations are not repeated here, but
may be found in Mattingly, Heiser, and Pratt (2002).4
3.3.2.1 Free Stream and Diffuser
The engine analysis must start at the free stream conditions. First, the known properties
of the air are converted to total or stagnation properties. Temperature and pressure are first
converted from standard atmospheric conditions to the actual conditions of the aircraft, which are
corrected using tables and interpolation from Heiser and Pratt (1994). The equations and
constants for freestream and diffuser entry and exit properties are shown in Table 3.9.
Table 3.9 Diffuser and nozzle equations.
Component Variable Description Model Equations
Freestream
r Isentropic freestream temperature recovery ratio
1
202
11
rr
r M
rt TT 00
rt PP 00
r Isentropic freestream temperature recovery ratio
0tT Total/stagnation temperature
0tP Total/stagnation pressure
4 ©AIAA. The low-bypass turbofan equations are reprinted with permission. See the Mattingly, Heiser, and Pratt
(2002) for the comprehensive list of turbofan equations.
62
Component Variable Description Model Equations
Diffuser
d Total pressure ratio
0
2
t
td P
P = Rspecd max
97.0max d
0
2
t
td T
T =1
Rspec =1 for M0 ≤ 1
35.10 1075.01 MRspec for M0 > 1
maxd Total pressure due to wall friction
d Total temperature ratio
Rspec Ram recovery coefficient
Nozzle
n Total pressure ratio
7
9
t
tn P
P
17
9 t
tn T
T n Total temperature ratio
3.3.2.2 Fan and High Pressure Compressor
The fan and high pressure compressor components are driven by the low and high
pressure turbines, respectively, via the low pressure and high pressure spools, respectively. The
fan and high pressure compressor are used to compress the incoming air before combustion. The
fan runs at a lower pressure ratio than the high pressure compressor, and typically spins slower
than the high pressure spool as well. The fan and high pressure compressor calculations are
given in Table 3.10.
Table 3.10 Fan and high pressure compressor equations.
Component Variable Description Model Equations
Fan
'c Total pressure ratio 2
3'
'
t
tc P
P ,
2
3'
'
t
tc T
T
1'
1/1
'
''
c
c
cc
=
1
1
2
3
23
2
t
t
tt
t
P
P
TT
T
TcmW pccc ''' . =
1
1
2
3
'
2''
' c
c
t
t
c
tpcc P
PTcm
fandr
fandrfanfanC mm
'c Total temperature ratio
'c Efficiency
'cW Power
Ccm ' Corrected mass flow rate
Compressor
cH Total pressure ratio
1
1/1
cH
cHcH
cHcH
=
1
1
3
'3
3'3
'3
t
t
tt
t
P
P
TT
T
'3
3'
t
tcH P
P ,
'3
3
t
tcH T
T
cH Total temperature ratio
cH Efficiency
63
Component Variable Description Model Equations
cHW Power
1.
1
3
33
'
' cH
cH
t
t
cH
tpcHcHpcHcHcH P
PTcmTcmW
cHcdr
cHcdrcHcHC mm
'
' cHCm Corrected mass flow rate
c Total pressure ratio for the compressor section cHcc '
3.3.2.3 Main Burner and Afterburner
The fuel is added to the airstream in the main burner and afterburner (if being used) and
ignited. Note that the main burner only adds fuel to the core air flow, while the afterburner adds
fuel to both the core and bypass air as both streams are combined in the mixer before entering the
afterburner section. Fixed efficiencies and pressure ratios are used for both the main burner and
afterburner. The equations are given in Table 3.11.
Table 3.11 Burner and afterburner calculations.
Component Variable Description Model Equations
Burner
burn Efficiency 98.03344
PRfAB
apcatptburn hm
TCmTCm
97.03
4 at
tburn P
P ,
at
tburn T
T
3
4
0TC/hf
pcbPR
cHcr '
0
4
TC
TC
pc
tpt
burn Total pressure ratio
burn Total temperature ratio
f Fuel/air ratio
Enthalpy ratio
Afterburner
AB Efficiency 97.06677
PRfAB
tpmtpABAB hm
TCmTCm
97.06
7 t
tAB P
P ,
6
7
t
tAB T
T
ABpcABPR
ptpMMtLmtHmABAB TCh
CCff
0
2121
/
/
1
11
0
7
TC
TC
pc
tpABAB
AB Total pressure ratio
AB Total temperature ratio
ABf Fuel/air ratio
AB Enthalpy ratio
64
3.3.2.4 High and Low Pressure Turbines
The high and low pressure turbines allow expansion and work extraction in the core of
the engine. The high pressure turbine vanes are cooled by coolant mixers with bleed air extracted
from the fan. The equations describing the high and low pressure turbine behavior are given in
Table 3.12.
Table 3.12 High and low pressure turbine equations.
Component Variable Description Model Equations
High Pressure Turbine
tH Total pressure ratio
t
t
tH
tHtH
1
1
1
,
4
4
t
bttH P
P
/11
11
'
'
121 cHcrmH
cHcrtH f
tHmbcHcdr
tHmbcHcdrtHtHC mm
1'
1'
tH Total temperature ratio
tH Efficiency
tHCm Corrected mass flow rate
Low pressure Turbine
tL Total pressure ratio t
t
tL
tLtL
1
1
1
,
ct
ttL P
P
4
5
//11
/111
'
'
2121 cHcrtHtHmL
mPTOcrtL f
Ca
( Where, 00 TCm/PC pcTOTO )
tLmtHmbcHcdr
tLmtHmbcHcdrtLtLC mm
21'
21'
tL Total temperature ratio
tL Efficiency
tLCm Corrected mass flow rate
3.3.2.5 Coolant Mixers and Exhaust Mixer
Turbine cooling is required to avoid exceeding material design temperature limits in the
turbines in the PS (the limit imposed is 3200 R). The turbine cooling is performed in the coolant
mixers with air bled off of the high pressure compressor. The amount of turbine cooling
required depends on the temperature of the burner.
After the turbine section, the bypass and core airstreams are combined in the mixer
section of the turbofan engine. The mixed streams then enter the afterburner section. The
equations for the coolant mixers and exhaust mixer are given in Table 3.13.
65
Table 3.13 Turbine cooling mixer and exhaust mixer equations.
Component Variable Description Model Equations
Coolant Mixer 1
21, Coolant mixer ratios 000,16/2400421 tT
where RTt 24004
otherwise 021
1cmix Total temperature ratio 121
'1211 11
/11
f
f cHcrcmix
11 cmix 1cmix Total pressure ratio
Coolant Mixer 2
2cmix Total temperature ratio 2121
11'2121
2 11
}{11
f
f tHcmixcHcrcmix
12 cmix 2cmix Total pressure ratio
Exhaust Mixer
M Total temperature ratio
),,(
),,()'1(
1
/1
11'
666
555
6
5
5
6
'
5555'
6
5
5
6
21215
5
''
'
RMMFP
RMMFP
A
A
P
P
TCTC
C
C
T
T
fm
m
Mt
tM
tptp
p
p
t
tM
M Total pressure ratio
' Mixer bypass ratio
3.3.3 Thrust and Performance Calculations
The overall performance equations of the PS are detailed in Table 3.14. The uninstalled
thrust is the engine thrust produced without any losses attributed to the engine cowl or nozzle
drag.
Table 3.14 Thrust and engine performance calculations.
Component Variable Description Model Equations
Overall Engine Performance
0f Total fuel consumption
1/11 210 ABfff
nABMtLtHbcHcdrt
P
P
P
P '9
0
9
9
2
11
9
99 1
1
2
AB
AB
P
PM t
AB
88 /)1(990
9
PP
C
C
T
T
t
ABpAB
pc
2
11
9
9
0
9 8
8
11
P
P
V
V t
r
AB
000 TgRMV ccc where 174.32cg lbm*ft/(lbf*s2)
9
9
P
Pt Total/static pressure ratio in nozzle
9M Nozzle exit Mach
0
9
T
T Overall
temperature ratio
0
9
V
V Overall velocity
ratio
0V Aircraft velocity (ft/s)
66
Component Variable Description Model Equations
Thrust
0m
F Uninstalled specific thrust (lbf/lbm/s)
0
0
20
90
0
9
9
00
0
90
0
0
3600
/1
11
11
1
mF
fS
M
PP
T
T
V
V
R
Rf
V
Vf
g
V
m
F
cc
ABc
S Uninstalled specific fuel consumption (1/h)
F Uninstalled thrust (lbf)
Subsonic inlet drag is estimated by assuming a worst case scenario of massive flow
separation at the lip of the inlet and no recovery of additive drag. Additive drag is defined as the
positive drag acting on the streamtube which encloses the air that enters the engine inlet
(Mattingly, Heiser, and Daley, 1987).
Supersonic inlet drag is estimated somewhat conservatively using the idea of the inlet
“swallowing” its projected image. This, in effect, means that the inlet area, 1A , is larger or equal
to the flow capture area, 0A , and that the inlet must have the ability to vary its geometry. Also,
the excess air captured by the inlet must be vented via boundary layer bleed ports. The excess
air vented is at a lower velocity than the aircraft, thus, creating drag from the momentum loss.
The supersonic inlet loss coefficient reflects this momentum loss drag. The inlet and nozzle drag
and installed thrust equations are given in Table 3.15.
Table 3.15 Inlet and nozzle drag and installed thrust equations.
Component Variable Description Model Equations
Inlet Drag
inlet Inlet loss coefficient
For M0 < 1.0:
000
20
0
120
0
1
1
0 1
aMmFg
MA
AM
T
T
M
M
cinlet
For M0 > 1.0:
00
2
1
200
0
1
1
1
1
21
amFg
MMA
A
cinlet
0M Freestream Mach
1M Inlet Mach
0A Area of engine capture streamtube
1A Inlet area
0M Freestream Mach
1M Inlet Mach
0T Freestream temperature
1T Inlet temperature
67
Component Variable Description Model Equations
Nozzle Drag
nozzle Nozzle loss coefficient
For M0 < 0.8:
00
0
9100 2
am
Fg
A
AACM
c
D
nozzle
For 0.8 < M0 < 1.2:
00
0
9100 2
am
Fg
A
AACM
c
DP
nozzle
For M0 > 1.2:
1
exp4.11
2.1 20
200
M
M
C
MC
D
D
10A Nozzle inlet area
9A Nozzle throat
0a Freestream speed of sound
DC Convergent nozzle pressure drag coefficient
DPC Experimental pressure drag coefficient
Installed Thrust T Installed thrust (lbf)
inletnozzleFT 1*
The DC term and DPC terms in Table 3.15 are nozzle pressure drag coefficients derived
using the “integral mean slope” (IMS) method discussed in Mattingly, Heiser, and Pratt (2002).
3.3.4 PS Exergy Model
The exergy model for the PS was developed on a component by component basis. This
accounting structure provides a much more detailed picture of losses and inefficiencies which
allows a designer to very quickly see areas needing improvement within the PS. The exergy
destruction calculations are shown in Table 3.16. Notice also that the unburned fuel is taken into
account in the exergy destruction equations.
Table 3.16 PS exergy model equations.
Component Variable Description Model Equations
Fan fandesxE _
Exergy destruction rate in the fan
2
3
2
320_
''lnln
P
PR
T
TCmTxE cpcfandes
High Pressure Compressor
cHdesxE _
Exergy destruction rate in the high pressure compressor
'3
3
3
330_ lnln
' P
PR
T
TCmTxE cpccHdes
68
Component Variable Description Model Equations
Burner burnerdesxE _
Exergy destruction rate in the burner
fuelcorefuelburnerdes hmxE **35.0 __
High Pressure Turbine
tHdesxE _
Exergy destruction rate in the high pressure turbine
4
4
4
440_ lnln
P
PR
T
TCmTxE b
ta
bptatHdes
Low Pressure Turbine
tLdesxE _
Exergy destruction rate in the low pressure turbine
c
ttL
c
tptctLdes P
PR
T
TCmTxE
4
5
4
540_ lnln
Mixer mixerdesxE _
Exergy destruction rate in the mixer
r
iiMpMimixerdes yR
p
pR
T
TCmxE
1
''
_ lnlnln
Afterburner ABdesxE _
Exergy destruction rate in the afterburner
fuelABfuelABdes hmxE **35.0 __
Unburned unburnlossxE _
Exergy loss rate due to the fuel lost out the back of the PS
ABdes
ABburnerdesbunburnloss
xE
xExE
_
__
*
1*1
The PS is not only closely coupled to the AFS-A but also to other aircraft subsystems.
One of these is the environmental controls subsystem (ECS), which uses bleed air from the PS as
the working fluid to perform its cabin and avionics cooling duties (see Figure 3.5). The ECS is
detailed in the following section.
3.4 Environmental Controls Subsystem
The environmental controls subsystem (ECS) is responsible for cooling the low-heat
generation avionics boxes and keeping the cabin temperature and humidity at comfortable levels
for the pilot. The system is a bootstrap system similar to that used in the F-16. The ECS consists
of four compact heat exchangers, a water separator, and an air-cycle machine.
3.4.1 ECS Layout and Definitions
The ECS flow rate and input air conditions are determined by two bleed ports located in
the PS. The low pressure bleed port is located midway on the high pressure compressor in the
PS, while the high pressure bleed port is located immediately after the high pressure compressor.
Figure 3.5 shows the bootstrap type ECS along with the PS. The bleed air input pressure to the
ECS is controlled via a pressure regulating valve (PRV).
69
Figure 3.5 ECS layout and components (Muñoz and von Spakovsky, 2001).
Following the path of the bleed air flow from the PS shows that the bleed air is first
cooled in the primary heat exchanger via ram air. A portion of the bleed air is then compressed
and cooled first through the secondary heat exchanger, then through the bleed air/PAO heat
exchanger, and finally through the regenerative heat exchanger. The air is then expanded over
the turbine that drives the compressor. The water separator removes the water from the air
which finally makes it to the cabin and low-heat generating avionics (i.e. air-cooled avionics).
The following subsections detail the ECS thermodynamic model, heat exchangers, and ECS
exergy model.
3.4.2 ECS Thermodynamic Model
The thermodynamic model of the ECS consists of multiple components. The ECS model
is based primarily on the work of Muñoz and von Spakovsky (2001a,b), Periannan (2005), and
Rancruel (2003). The station definitions may be seen in Figure 3.6, and the model equations for
the ECS components are given in Table 3.17.
70
Figure 3.6 ECS station definitions (Rancruel, 2002).
Table 3.17 Thermodynamic model of the ECS (Periannan, 2005).
Variable Description Model Equation
9T Load thermodynamic temperature loadTT 9
9P Load thermodynamic pressure loadPP 9
8T Water separator inlet temperature 98 TT
8P Water separator inlet pressure wsPPP 98
6
7
T
T Turbine temperature ratio
1
6
7
6
7 11P
P
T
Ttb
t Turbine efficiency
Fv Turbine velocity factor
14028
Y
YT
NFv
in
1
1
PRY
6T Regenerative hot-side exit temperature. Cmin is the smallest of the heat capacities C5 and C8’. Also, rhx is the heat exchanger effectiveness.
5
8556 C
C'TTTT min
rhx
6P
Regenerative heat exchanger hot-side exit
pressure. A correlation is used for the pressure
drop. 5
56 1 P
P
PP
Preconditioning
Sub-system
Bootstrap
Sub-system
Preconditioning
Sub-system
Bootstrap
Sub-system
3‘
3‘
4‘
8‘
71
Variable Description Model Equation
5T
Bleed air / hot PAOS heat exchanger hot-side
exit temperature. Cmin is the smallest of the heat
capacities C4’ and C14.
'
min'rhx_pao/bleed' C
CTTTT
414445
5P Bleed air / hot PAOS heat exchanger hot-side
exit pressure. A correlation is used for the
pressure drop.
''
PP
PP 4
45 1
'T4 Secondary regenerative heat exchanger hot-side
exit temperature. Cmin is the smallest of the heat
capacities C4 and C10.
4
444 C
CTTTT min
ohx_ondsec'
'P4 Secondary regenerative heat exchanger hot-side
exit pressure. A correlation is used for the
pressure drop.
44
4 1 PP
PP '
3
4
T
T Compressor temperature ratio
11
1
1
3
4
3
4
P
P
T
T
cp
c Compressor efficiency
shaftw Shaft work of the compressor and turbine 7634 hhhhwshaft
3T Primary heat exchanger hot-side exit
temperature. Cmin is the smallest of the heat
capacities C2 and C12.
2
223 C
CTTTT min
ohx_pri
3P Primary heat exchanger hot-side exit pressure. A
correlation is used for the pressure drop. 2
23 1 P
P
PP
k Ram inlet scoop mass flow ratio. ukuii
fr Ram inlet scoop pressure recovery factor
ofo PrPi
iT
Ram scoop inlet temperature (“i” equals 10 or
12 and Ti* is the temperature at sonic
conditions; see Anderson, 1984). 212
1
i*
i
i
MT
T
iP
Ram scoop inlet pressure (“i” equals 10 or 12
and Pi* is the pressure at sonic conditions; see
Bejan (1996).
2
1
212
11
ii*
i
i
M)(MP
P
oiP
Ram scoop inlet stagnation pressure (“i” equals
10 or 12 and Poi* is the stagnation pressure at
sonic conditions; see Davenport (1983).
)(i
i*o
oi M)(
MP
P
i
12
12
1
121
72
Variable Description Model Equation
iD
fL
4
Ram scoop inlet augmented friction factor; f
assumed equal to 0.01 and L and D are the
length and the diameter of the ram air duct,
respectively, see Davenport (1983).
2
2
2
2
12
1
2
114
i
i
i
i
i M)(
M)(ln
M
M
D
fL
dragD Drag due to the presence of the ram air inlet and
outlet (e). Pressure drag has been ignored. edrag uumD
1311orP Ram air pressure just ahead of the ram scoop
exit (states 11 or 13). 12101210
1311 1 oror
or PP
PP
3.4.2.1 Heat Exchangers
The heat exchanger thermodynamic models used in this thesis work are primarily from
Shah (1981) and Kays and London (1984). The models are compact, offset-strip fin type heat
exchangers and focus on the liquid/air heat exchange, although the thermodynamic model may
be applied to any general heat exchanger with the proper derivation from the general case. The
geometric parameters of the offset-strip fin are shown in Figure 3.7.
Figure 3.7 Geometric parameters of the offset-strip fin (Muñoz and von Spakovsky, 1999).
The weights of the heat exchangers are found using the fin geometry and density. Table
3.18 shows the compact heat exchanger model equations.
Table 3.18 Geometric and heat transfer models of the compact heat exchangers.
Variable Description Model Equation
Lh Hot-side length Assigned value
Lc Cold-side length Assigned value
t s h=b
a
73
Variable Description Model Equation
hD Hydraulic diameter thhlsl
shlDh
2
frA Frontal area nfr LLA
Ratio of minimum free flow area to frontal area fr
O
A
A
Heat transfer area / volume between plates hD
4
t
f
A
A Fin area / total area
hs
hs
A
A
t
f
2
pN Number of plates ahh
ahLnN
rb
bnpp 2
2
bpV
rpV Volume between plates, bleed and ram air side
bprbp hNLLVb
)1(
bprbp hNLLVr
A Heat transfer area VpA
OA Minimum free flow area L
ADA h
O 4
G Mass velocities OA
mG
eR Reynolds number
he
DGR
f
j
Friction and Colburn coefficients (Muñoz and von Spakovsky, 1999)
For *ee RR (laminar flow)
2659030530185607422062439 ....eR.f
0678014990154105403065220 ....eR.j
f
j
Friction and Colburn coefficients (Muñoz and von Spakovsky, 1999)
For 1000 *ee RR (turbulent flow):
2423068200093602993086991 ....eR.f
1733019550103704063024350 ....eR.j
wA Wall conduction area )N(LLA prbw 12
wR Wall thermal resistance ww
w Ak
aR
l Fin length tb
l 2
H Height abbnabH chplatesh 22
74
Variable Description Model Equation
fA Finned area
2
12
kt
hm , HLAfr
ml
mlf
)tanh( ,
A
AffO 11
frA Frontal area
f Fin efficiency
o
Outside overall surface efficiency
n Mixture molar flow rate
OA
mG
,
k
cP p
r
32
r
p
P
jGch
j Colburn factor
G Maximum mass velocity
Pr Prandtl number
h Heat transfer coefficient
U Overall heat transfer coefficient h
phcpcmin cncnC ,min
hph
cpcmax cncnC ,max
rOw
bO hAR
hAUA 111
, maxminr C/CC
minC
UANTU
minC Minimum heat capacity
maxC Maximum heat capacity
rC Heat capacity ratio
NTU Number of transfer units
Effectiveness
1
11 780220 .
r.
r
NTUCexpNTUC
exp
P The core pressure drop
e
ie
m
i
o
e
ic
iK
A
Af
K
P
G
P
P
2
22
1
1211
2
The ECS exergy model is discussed in the following subsection.
3.4.3 ECS Exergy Model
The ECS exergy model losses are primarily due to the heat exchangers and air cycle
machine (consisting of a turbine/compressor) losses. No exergy destruction is attributed to the
“conditioned” air exiting the cockpit and avionics after it is used. Any excess bleed air from the
PS is considered to be negligible. The exergy destruction rate equations for the ECS heat
exchangers are shown in Table 3.19.
75
Table 3.19 ECS exergy destruction rate equations.
Component Variable Description Model Equations
Cold Side coldP
DESxE
Exergy destruction rate on the cold side of the heat exchanger due to the pressure drop
outcoldincoldcoldcoldcoldP
DES PPTRmxE /ln0
Hot Side hotP
DESxE
Exergy destruction rate on the hot side of the heat exchanger due to the pressure drop
outhotinhothothothotP
DES PPTRmxE /ln0
Temperature gradient
TDESxE
Exergy destruction rate due to a temperature drop
outhot
inhot
outcold
incold
T
T
T
TPcoldcoldPhothot
TDES T
dTcm
T
dTcmTxE
1
0
Total Exergy Destruction Rate
HXDESxE Total exergy destruction rate in the heat exchangers
hotP
DEScoldP
DEST
DESDES xExExExEHX
The exergy destruction calculations for the compressor and turbine ECS components are
found in Table 3.16 as one turbine or compressor is treated the same as another in terms of the
exergy destruction. The ECS interacts with the PS via bleed air and the vapor compression /
polyalphaolefin subsystem (VC/PAOS), which is detailed in the following section.
3.5 Vapor Compression / PAO Subsystem
The vapor compression / PAO subsystem (VC/PAOS) consists of three loops: the cold
PAO loop, the hot PAO loop, and the vapor cycle. The VC/PAOS serves as a heat sink for the
ECS via a PAO / bleed air heat exchanger and for the high-heat generation avionics via the cold
PAO loop. The hot PAO loop rejects heat to the fuel loop subsystem (FLS) and to a ram air /
PAO heat exchanger. The motor driving the compressor on the vapor cycle and the pumps on
the hot and cold PAO loops are all driven via electrical power from the electrical subsystem
(ES). The VC/PAOS layout and station definitions are shown in Figure 3.8.
76
Figure 3.8 VC/PAOS layout and station definitions (Rancruel, 2002).
3.5.1 VC/PAOS Thermodynamic Model
The thermodynamic models of the vapor cycle compressor and the PAO / ram air heat
exchanger in the VC/PAOS are based on a perfect gas model. The load temperature (or the
temperature of the ‘Avionics Box’ in Figure 3.8) is given by Figliola, Tipton, and Ochterbeck
(1997), and the physical model is from Greene (1992). The model equations for the VC/PAOS
thermodynamic model are presented in Table 3.20.
Table 3.20 VCPAOS model equations.
Component Variable Description Model Equations
Initial Condition loadT
Load temperature of the high-heat generation avionics
loadTT 10
77
Component Variable Description Model Equations
Cold PAO Loop Pump
AvioQ
Heat transfer rate required by the liquid cooled avionics
pPAOPAO
Avio
Cm
QTT
1011
PAOm Mass flow rate in the cold PAO loop 01109
01109
01109
01109
TT
TT
hh
hh ssp
pW Pump work rate 110901109 PPvTTcWP
0110901109
01109
PPvTTc
PPvp
pipeboxAvioEvapp PPPPPP _01109
pPAOC Heat capacity of the PAO
p Pump efficiency
pP Pressure rise across the pump
Evaporator
loadQ
Heating load due to the avionics box and the pump temperature rise
)( 910 hhmQ vapload
)()( 91078 hhmhhmQ paovapEvap
Vap
EvapEvap
VapEvap m
Qh
m
CTThh
7
min7978
9
min109910 C
CTTTT evap
99
10910 1 P
P
PP
77
878 1 P
P
PP
EvapQ Heating load on the evaporator
Evap Evaporator effectiveness
vapm Vapor mass flow rate in the evaporator
Compressor
cW Compressor work rate
11
1
1
8
5
8
5
P
P
T
T
cp
85 hhmW vapc
net
load
W
QCOP
COP Vapor compression cycle coefficient of performance
Condenser
EvapQ Heat transfer rate in the evaporator
CompEvapCond QQQ
)()( 42165 hhmhhmQ paovapCond
Vap
CondCond
VapEvap m
Qh
m
CTThh
5
min4556
4
min45412 C
CTTTT evap
55
566 1 P
P
PP
44
41212 1 P
P
PP
CompQ Heat transfer rate in the compressor
Cond Condenser effectiveness
CondQ Heat transfer rate in the condenser
Expansion Valve valPr Pressure ratio of the
expansion valve
0706 hh assumes isenthalpic 0)()( 6767 PPvTTcv
CondEvapcomval PP PrPr
78
Component Variable Description Model Equations
Hot PAO Loop Pump
p Pump efficiency
01201
01201
01201
01201
TT
TT
hh
hh ssp
0120101201 PPvTTcW p
0120101201
01201
PPvTTc
PPvp
pW Pump work rate
pPAOC Heat capacity of the PAO
Bleed Air / Hot PAO Heat Exchanger
hxpaobleed _/
Bleed air / PAO heat exchanger effectiveness
1
min_1_/12 C
CTTTT inbleedhxpaobleed
11
2 1 PP
PP
Fuel / Hot PAO Heat Exchanger
hxpaofuel _/ Fuel / PAO heat exchanger effectiveness
2
min_2_/23 C
CTTTT infuelhxpaofuel
22
3 1 PP
PP
Ram Air / PAO Heat Exchanger
hxpaoram _/ Ram air / PAO heat exchanger effectiveness
3
min_3_/34 C
CTTTT inramrhxpaoram
33
4 1 PP
PP
Note that the VC/PAOS ram air inlet has the same thermodynamic model as the inlet
ducts in the ECS. The heat transfer model for the condenser and evaporator is from Liu and
Kakac (2000). The compact heat exchanger model for effectiveness is from Incropera and
DeWitt (1990) and applies to single-pass, cross-flow heat exchangers with unmixed fluids. The
thermodynamic properties of the PAO are based on data from the CRC Handbook (1976);
Zabransky et al. (1996); and the JANAF Thermochemical Tables (1998).
3.5.2 VC/PAOS Exergy Model
The exergy component models of the VC/PAOS are effectively the same as those of the
ECS. No additional unique components are introduced in the VC/PAOS; and, thus, the exergy
destruction equations are not repeated here.
The VC/PAOS interfaces with a number of subsystems, including the ECS (cooling loads
for the bleed air and liquid-cooled avionics), the ES (to power the pumps and compressor), and
the FLS which serves as a heat sink for the hot PAO loop. This latter subsystem is discussed
next in the following section.
79
3.6 Fuel Loop Subsystem
The fuel loop subsystem (FLS) consists of a fuel tank, pumps, fuel lines, and controls
required to supply fuel to the PS as well as the hardware necessary to use the fuel as a heat sink
by the other subsystems. The FLS thermodynamic model is based on the work of Rancruel
(2003) and Periannan (2005). The FLS system schematic is shown in Figure 3.9.
Figure 3.9 Schematic of the fuel loop subsystem (Rancruel, 2002).
The fuel from the fuel tank is first pressurized by the pump and then is heated by the hot
PAO loop via a compact heat exchanger. The fuel is then passed through the fuel / oil heat
exchanger where it is further heated by the engine cooling/lubricating oil in the oil loop
subsystem (OLS). The fuel is interfaced with the central hydraulic subsystem (CHS) as a heat
sink again before finally being burned in the PS. Notice that the fuel loop does not end at the PS
but allows for excess fuel to be pumped through the loop if additional cooling capacity is needed
80
to meet the heating load requirements of the other subsystems. The additional fuel is cooled in
the fuel / ram-air heat exchanger before it is returned to the fuel tank to avoid heating the fuel
beyond acceptable limits.
3.6.1 FLS Thermodynamic Model
The thermodynamic model of the FLS is similar to the ECS and VC/PAOS in that
compact heat exchangers are also used. The model equations for the heat exchangers can be
found in section 3.4.2.1. The fuel tank temperature is not monitored as was done in Rancruel
(2003), but rather a constraint is imposed on the allowable temperature of the excess fuel
returning to the fuel tank. The FLS thermodynamic model equations are presented in Table 3.21.
Table 3.21 Fuel loop subsystem thermodynamic model equations.
Component Variable Description Model Equations
Fuel tank
outfuelm _ Mass flow rate of fuel leaving the tank
T
TTTC
TΤCmm
addpao
paofuel
outpaoaddpaooutfuelfuelp
outpaoinpaopaopoutfuel
_
/
___
___
)(*
)(*)*(
SFCmm outfueladdfuel __
)(** ___/ TTCmQ addfueloutfuelpfuelHXPAOfuel
addfuelfuelp
inramPStofuelminpramfuelPStofueladdfuel mC
TTCmTT
_
___/___ *
)(*)*(*
addfuelm _
Fuel added to the fuel required by the PS
addfuelT _
Temperature of the fuel returning to the tank
SFC Specific fuel consumption (by the PS)
PStofuelT __
Temperature of the fuel to the PS
Fuel / PAO heat exchanger
paofuel / Fuel / PAO heat exchanger effectiveness
)T(T*Cp)*m(
)T(T*Cp)*m(
Q
Q
fuel_inpao_inmin
pao_outpao_inpao
max
actual
fuel/pao
Fuel / Oil heat exchanger
oilfuel / Fuel / oil heat exchanger effectiveness )T(T*Cp)*m(
)T(T*Cp)*m(
fuel_inoil_inmin
oil_outoil_inoil
fuel/oil
Fuel / Hydraulic Oil heat exchanger
hydfuel /
Fuel / hydraulic oil heat exchanger effectiveness
)T(T*Cp)*m(
)T(T*Cp)*m(
fuel_inhyd_inmin
hyd_outhyd_inhyd
fuel/hyd
Fuel / Ram-Air heat exchanger
ramfuel / Fuel / ram air heat exchanger effectiveness )T(T*Cp)*m(
)T(T*Cp)*m(
Q
Q
ram_infuel_inmin
ram_inram_outram
max
actual
fuel/ram
81
The FLS serves as the main thermal management subsystem (TMS) in the aircraft. The
FLS exergy model is described by the same equations as for both the VCPAOS and ECS, and
thus, the equations are not presented again in this section. The FLS interfaces with a number of
subsystems as a heat sink, one of which is the OLS which, as previously mentioned, is the
subsystem responsible for cooling and lubricating the engine bearing surfaces. The OLS is
discussed in the following section.
3.7 Oil Loop Subsystem
The oil loop subsystem (OLS) lubricates and cools the PS bearing surfaces. The OLS
may perform secondary functions as well such as cooling the auxiliary power unit (APU) or
operating thrust reversers; however, secondary OLS functions are not modeled in this thesis.
The oil loop subsystem used in this thesis is similar to the type used in the Pratt and
Whitney F100-PW-100 (early F-15 engine) with a few exceptions: a fuel/oil cooler is used here
instead of the air / oil cooler as was used in the F100-PW-100. A simplified diagram of the OLS
is shown in Figure 3.10.
Aircraft generally have two different configurations for OLSs: either a “hot tank” or a
“cold tank” configuration. The former configuration (see Figure 3.10) pumps the oil directly
from the scavenger pumps to the oil tank. The supply (or pressure) pump moves the hot oil first
through the oil cooler and finally to the oil nozzles at the various bearings. The “cold tank”
configuration has the scavenger pumps moving the oil through the oil cooler before heading to
the tank, thus, the name “cold tank.” It is advantageous to have a “hot tank” design in fighter
aircraft because the oil and air separation is more efficient.
Multiple scavenger pumps are located throughout the bearing sumps in a jet engine for
redundancy as well as for location (a sump is usually needed at each main bearing location). The
scavenger pumps have much greater total capacity than the supply pump so the scavenger pumps
will inherently pump a quantity of air as well as oil. This requires oil and air separation and an
OLS typically has an internal deaerator on the tank. All the pumps on the OLS are driven by a
gearbox on the low pressure spool of the PS.
82
Figure 3.10 Oil loop subsystem schematic.
The supply pump in an OLS is usually an internal/external gear, rotary, positive-
displacement type pump which can effectively pressurize any system over its design limits if
enough shaft power is available, and the pump does not encounter mechanical failure. This
necessitates placing relief valves throughout the system to avoid over-pressurization.
3.7.1 OLS Thermodynamic Model
The OLS interacts with the PS, FLS, and AFS-A in the aircraft. The interaction with the
PS is not only via the heating load, as previously mentioned, but also via the shaft power
required to power the supply, booster, and scavenger pumps. The work required by the OLS
pumps is calculated using the equations in Table 3.22.
Table 3.22 OLS pump work equations.
Component Variable Description Model Equation
Pump Work
pumpW Shaft power required
mpump
PQW
*
where
fluid
mQ
m Pump efficiency
Q Volumetric flow rate
83
The OLS heating load on the FLS must be established to have a fully integrated model.
The development of the OLS model proved to be difficult because of the lack of information on
this subsystem. Two statements found in Hudson (1986) were used along with some PS
performance/setting correlations to develop the heating load model for the OLS. Hudson (1986)
states that the OLS cooling load is 31% of the total aircraft heating load and that the ECS is 21%
of the total aircraft heating load during cruise conditions. These two statements are used as the
basis of heating load model for the OLS. The ECS heating load is well-defined; thus, a
relationship for the OLS was developed relative to the uninstalled cruise thrust and ECS heating
load during cruise conditions. Cruise was defined as the flight conditions for mission segment
19.
Once the ECS heating load is established for segment 19, the OLS heating load can be
estimated using the aforementioned correlation provided by Hudson (1986). This requires the
maximum and minimum thrust at the given cruise conditions which are found using the thrust
correction for altitude (versus sea level take-off thrust) and a maximum/minimum estimated
engine operating speed (which is found using 11,500 RPM for maximum rpm, and 4,200 RPM
for minimum-sustainable RPM). Next, a percentage of the total available thrust can be
developed for the engine/aircraft pair for cruise conditions from the maximum and minimum
thrust for the engine and the required thrust to fly at cruise conditions. For example, the thrust
required for cruise is 25% of the maximum thrust of the engine at those operating conditions.
This percentage is carried over to the heating load as a correlation between uninstalled thrust and
heating load on the OLS. Thus, for any uninstalled thrust setting in the PS, a corresponding
heating load on the OLS can be found. The heating load equations and correlations used in the
OLS model are given in Table 3.23.
Table 3.23 OLS heating load equations.
Component Variable Description Model Equation
OLS Heat Exchanger Load
HXccOLSfuelQ _/
Cooling load on the fuel from the OLS at cruise conditions
HXccECSfuelHXccOLSfuel QQ _/_/21
31
HXccECSfuelQ _/
Cooling load on the fuel from the ECS at cruise conditions
HXOLSfuelQ _/
Cooling load on the fuel from the OLS at a given sea level, equivalent thrust SLTT max
maxT Maximum thrust at altitude
cruiseT Uninstalled cruise thrust from
84
Component Variable Description Model Equation
the PS
std
ambient
P
P
maxmax
* RPMT
TRPM cruise
cc
cruise
cruiseHXccOLSfuel T
TTQQ
TRPM
RPMT
min_/min
maxmax
minmin
1
min_/
min
min_/
* QTslopeQ
TT
QQslope
QTHXOLSfuel
cruise
HXccOLSfuelQT
minT
Estimated minimum thrust at altitude from the RPM relationship
ccRPM Estimated engine RPM at cruise
QTslope
Slope of the heating load versus the installed the thrust curve (assumes a linear relationship)
T Uninstalled thrust
The heat exchanger in the OLS is the same as the compact heat exchangers detailed in
section 3.4.2.1. Note that the weight of the heat exchanger is accounted to the FLS since the heat
exchanger is owned by the FLS and not the OLS in this thesis work.
A schedule of stainless steel tubing and predicted mass flow rates of the OLS were
initially used to model the OLS. The results showed that the pressure drops due to frictional
losses were negligible in terms of the exergy destruction and pumping losses. Thus, the final
OLS model used in this thesis work does not estimate tube sizing or calculate friction losses.
3.7.2 OLS Exergy model
Exergy destruction is caused in the OLS by pressure losses due to fluid friction or
restriction, temperature losses due to heat transfer across finite temperature differences,
unrestrained expansions, or mechanical inefficiencies. The exergy equations for the OLS are
given in Table 3.25.
Table 3.24 OLS exergy destruction equations.
Component Variable Description Model Equation
Exergy Destroyed
0T Ambient temperature
1
1
midealactualPump PQWWxE
outhot
inhot
outcold
incold
T
T
T
TPcoldcoldPhothot
TDES T
dTcm
T
dTcmTxE
1
0
Q Volumetric flow rate
m Mechanical efficiency of the pump
PumpxE Exergy destruction rate due to pumping irreversibilities
TDESxE
Exergy destroyed due to the heat exchanger temperature gradient
85
It is important to note that for this subsystem, there are actually a total of 6 pumps that
generate entropy: the main pump, booster pump, and the four scavenger pumps. The total exergy
destruction in the OLS is approximated by finding the pumping losses and exergy destruction in
the fuel/oil heat exchanger. We now turn to the central hydraulic subsystem (CHS) which is the
next subsystem that interfaces with the FLS by means of a heat interaction. The CHS is detailed
in the following section.
3.8 Central Hydraulic Subsystem
Hydraulics are used to actuate flight control surfaces as well as various other systems in
the aircraft (e.g. the landing gear, nose-wheel steering, etc.). Hydraulics are attractive to use in
aircraft because they are able to transfer large amounts of power from a central location to where
it is needed by means of small diameter hoses. However, despite the advantages of hydraulics,
traditional central hydraulic subsystems (CHS) may have a limited future in fighter aircraft. The
More Electric Aircraft (MEA) initiative started the fighter aircraft industry moving in the
direction of eliminating the CHS by using electric actuation rather than hydraulics. State-of-the-
art (SOTA) aircraft have replaced the hydraulics on primary flight control surfaces with electric
actuators and motors. Ultimately, the CHS will likely be eliminated from future fighter aircraft
for environmental reasons, reliability, maintainability, and operations and support (O & S) costs.
A traditional CHS is modeled in this thesis work, although MEA considerations are taken into
account in the electrical subsystem.
The CHS has the following characteristics: accessory gearbox driven pumps, triple
redundant hydraulic lines for flight critical loads, and a fuel/hydraulic oil heat exchanger. The
two hydraulic pumps are typically interconnected by power take-offs from the accessory drive
gearbox. The following hydraulic power consumers on the fighter aircraft have been modeled in
the CHS model: ailerons, tail (assume fully moveable), rudder, and landing gear. An example of
a CHS layout is shown in Figure 3.11.
3.8.1 CHS Thermodynamic Model
The CHS thermodynamic model includes sizing the actuators, pumps, and estimating the
flow required by the non-flight critical subsystems that are hydraulically actuated. Sizing the
86
hydraulic actuators requires an estimate of the flight control surface and rudder areas as well as
the landing gear weight and drag. A high-resolution sizing of the flight control surfaces would
require a dynamic analysis of the aircraft as well as roll characteristics and even possibly wind
tunnel testing. Obviously, that level of detail is not feasible for a conceptual design study such
as the present one. Thus, currently built aircraft were studied for aileron and rudder sizing
estimations and corresponding control surface sizing. Main and nose landing gear specifications
were also estimated to determine actuator sizing.
The flight control actuator sizing must be sized by the most constrained mission segment
to ensure that the flight control surfaces can operate properly throughout the mission. In this
case, the maximum opposing force to the flight control surfaces is when the aircraft is flying at
the highest dynamic pressure. The actuator flow estimation and sizing equations are shown in
Table 3.25.
Table 3.25 Actuator flow estimation calculations.
Component Variable Description Model Equation
Actuator Flow Estimation
maxF Theoretical maximum available force from an actuator (losses included)
PDF 2max 7854.0
PDFavail27.0
D Piston diameter (m)
P Pressure at the piston (N/m2)
availF Force available (approximate)
Flight Controls
High Lift Devices
Landing Gear
Braking
Cargo Bay
Hydraulic Reservoirs
Gearbox Driven Pumps
Electric Pump
Valve
Figure 3.11 Notional central hydraulics subsystem layout (simplified).
87
Component Variable Description Model Equation
Q Flow rate into the piston (m3/s)
act
actcyl
VD
Q
VAQ
*4
*
2
where
reqFF max
actV Velocity of the actuator (m/s)
cylA Area of the cylinder (m2)
Actuator Design
reqF Max force on the actuator
CSLareq AqCF ** max
2max 2
1VMAXq
aLaC *2
maxq Maximum dynamic pressure encountered by the aircraft
CSA Surface area of the control surface
LaC Coefficient of lift for a flat plate (simple)
a Angle of the control surface to the freestream (radians)
Although, as stated above, the actuators must be sized for the highest load they
encounter, the pump sizing method is somewhat different. The CHS pumps are sized to the
highest flow required at the lowest RPM. This design point usually occurs during landing, when
the engine is at a relatively low operating speed and many hydraulic functions are taking place
such as extending the landing gear, extending the flaps, deploying the slats, etc. Thus, the main
gear and nose gear are sized for the aircraft, and the total flow rates required estimated.
The actuator sizing for the flight control surfaces determine the flow rates and little
deviation from the design flow rate is observed even at a lower opposing force (lower dynamic
pressure). When the flow rate is known for the highest flow rate mission segment, the pump can
be sized to that flow rate and pressure requirement. A side effect of traditional hydraulic pump
sizing is that the pump is generally oversized in high engine RPM situations which sacrifices
pump operating efficiency. However, if an electrically powered pump were used instead of the
shaft-driven pump, this disadvantage would be removed.
Now, as to the working fluid, the hydraulic oil must be cooled due to heat generated by
friction in the actuators and hydraulic lines. The hydraulic oil is cooled via a fuel/hydraulic oil
heat exchanger. Similarly to the OLS, the heat exchanger weight is accounted to the FLS and is
not included in the CHS. The CHS heating load equations are given in Table 3.26. Note that the
equation units are listed in Table 3.26 because the constant in the heating load equation is not for
a general case.
88
Table 3.26 Central hydraulic subsystem heating load equations (Majumar, 2003).
Component Variable Description Model Equation
Heating Load Estimation
CHSE Heating load on CHS heat exchanger (kW)
CHSCHS PQE 1*10*966.38 5 Q Flow rate of oil (gpm)
P System pressure (psi)
CHS System efficiency
in
outCHS W
W
where
actactout VFW
PQWin
outW Actuator work required
actF Force required in the actuator
actV Velocity required from actuator
inW Work supplied to actuator
For the CHS model, the system pressure is set to 4000 psi, while the system efficiency,
, varies from 20% to nearly 90% depending on the required actuator power and the power
supplied to the actuator. The actuator equations are presented in more detail in Section 3.10.
The exergy model for the CHS is detailed in the following section.
3.8.2 CHS Exergy Model
The CHS destroys exergy due to irreversibilities in the heat exchanger, actuator
inefficiencies, frictional losses in the hydraulic lines, and inefficiencies in the hydraulic pumps
and motors. The exergy destruction equations are given in Table 3.27. Note that the frictional
Table 3.27 CHS subsystem exergy destruction equations (Bejan, 1996).
Component Variable Description Model Equation
Exergy Destroyed
genS Entropy generation rate
inpgen T*
Pm
T
Tln*cmS
1
2 for:
inp T*c*P
0TSxE genCHS
1
1
midealactualPump PQWWxE
CHSxE Exergy destruction rate
Fluid density
0T Dead state temperature
Q Volumetric flow rate
m Mechanical efficiency of the pump
PumpxE Exergy destruction rate due to pump irreversibilities
losses in the hydraulic lines are the cause of exergy destruction as well. Thus, the hydraulic lines
are sized to permit no more than a 25% pressure drop from the pump to the actuator in full-flow
conditions.
89
Finally, CHS power takeoff requirement is a small contributor to the total engine power
takeoff term, (see Table 3.12). More important in this sense is the electrical subsystem (ES),
which is responsible for the majority of the shaft power required from the PS. The ES is, thus,
detailed next.
3.9 Electrical Subsystem
Aircraft ESs have been a subject of much research recently. The main reason is the More
Electric Aircraft (MEA) program started by the Air Force in 1991 (Pearson, 1998; Weimer,
2003; Cloyd, 1997; Moir, 1999). The goal of the program was to transition currently built and
future aircraft away from traditional shaft-powered subsystems and towards electrically powered
ones. The MEA program required a more reliable, higher output, and more survivable electrical
power generation and distribution subsystem.
The ES model is based on MEA considerations as well as on a second program focused
on the power distribution subsystem called the Power Management and Distribution System for a
More Electric Aircraft (MADMEL). Northrup Grumman built a demonstrator of the
technologies developed under this program, and the ES components sized in this thesis are based
on the MADMEL demonstrator.
The model that is used here is patterned after the F-35 in that it employs two integrated
270 VDC starter/generators (IS/Gs) and an integrated power unit (IPU). The main generators are
switched reluctance machines (SRMs) with multiple channels, each of which are supported by a
channel (non-electrically isolated) SRM (Elbuluk and Kankam, 1996). Switched reluctance
machines are chosen due to high power densities and advantages in reliability and fault
tolerances compared to synchronous and induction machines. The power requirements are set up
to model an aircraft that has electro-hydrostatic or electro-mechanical actuators (EHA or EMA,
respectively) since industry seems to be moving in that direction. The 270 VDC generation is
claimed to have better efficiency than previously built aircraft generators as well as being
required for high powered/high voltage flight control actuators. 5 The power distribution
subsystem includes 28 VDC and 115 AC converters. The battery weight and chargers is also
included in the ES weight. A simplified schematic of the ES is shown in Figure 3.12.
5 Raymer (1999) states that an F-16 sized plane with electrically powered flight control actuators requires about 80
kW of additional power generation capacity.
90
minimum of three phases. The IPU is also an SRM; however, it is a three-phase, two-
This thesis work involved performing a small survey of currently built aircraft and their
power generation capabilities to establish an MEA-based power generation capacity for the
aircraft. Notably the F/A-18, F-22, F-16, F-35 and Eurofighter Typhoon were studied to
establish generating capacity guidelines. The power generation capacity was correlated to the
gross takeoff weight of the aircraft. The fighter aircraft previously built have lower electrical
loads than the newer models because traditional aircraft have more mechanically actuated
subsystems. For example, the Eurofighter Typhoon has an empty weight of approximately
11,000 kg and has two 30 kVA generators supplying 115/200 VAC, 400 HZ, three-phase power.
The weight-to-power-generation ratio for the Typhoon is 5.5 VA/kg6, which is significantly
lower than the MEA based F-22 and F-35 which are approximately 9.7 W/kg and 8.3 W/kg,
respectively. The guidelines developed for ES generating capacity and component weights are
shown in Table 3.28. The weight equation for the electrical subsystem may be found in Section
3.2.4.
6 VA (volt-amperes) is a measure of alternating current (AC) power, while W is a measure of direct current (DC)
power. For the purposes of this work, the AC power is considered to have a power factor of near unity which makes the units of VA nearly equivalent to W. This allows a direct comparison between the weight-to-power-generation ratios with a VA or W rating.
270 VDC BUS #1 270 VDC BUS #2
270 VDC Integrated
Starter/Generators
DC/AC Converter
DC/DC Converter
115 VAC BUS 270 VDC BUS 28 VDC BUS
DC/DC Converter
DC/AC Converter
115 VAC BUS270 VDC BUS 28 VDC BUS
Power Distribution Centers (PDCs)
IS/G #1
IS/G#2
IPU
Batteries
Figure 3.12 Notional electrical subsystem schematic (simplified).
91
Table 3.28 Fighter aircraft power generation/empty weight estimate.
Component Variable Description Model Equation
Non-MEA Aircraft Power Generation
tradkWR _ Estimated system electrical power generation capacity for a non-MEA aircraft (kW) EtradkW WR *2.6_
EW Empty aircraft weight (kg)
MEA Aircraft Power Generation mea_kWR
Estimated system electrical power generation capacity for an MEA aircraft (kW)
Emea_kW W*.R 38
3.9.1 ES Thermodynamic Model
The ES requires shaft power from the PS and cooling from the VC/PAOS. The power
takeoff (or power extraction) from the low pressure spool of the PS is simply the amount of shaft
power required by the generators for a given mission segment. The generating efficiency is set
to 85%, and the power takeoff equations are shown in Table 3.29.
Table 3.29 Electrical subsystem generator work.
Component Variable Description Model Equation
Electrical Subsystem Work
ES_TOP Power takeoff required for electrical subsystem
g
genoutESTO
WP
_
_
genoutW _
Electrical work rate required for a given mission segment
g Generator efficiency
Some of the ES components must be cooled to avoid overheating. The main generators
and components are designed to be cooled by the VC/PAO subsystem. Note that the subsystem
would be designed to route PAO through various heatsinks and hot areas in the electrical
components rather than through a heat exchanger. Table 3.30 gives the heating load equations
from the generators and the ES components. The transmission lines generally do not require
active cooling because they are oversized to handle current spikes.
Table 3.30 ES heating load model equations.
Component Variable Description Model Equations
Generator Heating Load
genESPAOQ _/
Cooling load on the PAO loop from the ES generators
sThm
STWT
TQ
dt
STUdirrs
kk
o
0
001
For steady state, no heat interactions and no mass interactions:
irrESTOgenout STPW 0__0
thus
genoutW _ Electrical work rate required for
a given mission segment
ES_TOP Power Take-off required for the ES
g Generator efficiency
0T Ambient Temperature
92
Component Variable Description Model Equations
genirrS _
Entropy generated in the generators
genoutESTOirr WPST __0
From an entropy balance, the entropy that must be removed from the component is as follows:
genirrgenESPAO
ST
Q_
0
_/
Component Heating Load
cinW _
Electrical work rate entering a component coutcincirr WWST ___0
Similarly to the generators, the entropy that must be removed in a heat interaction from the component is as follows:
cirrcESPAO
ST
Q_
0
_/
coutW _ Electrical work rate leaving a
component
cirrS _
Entropy generated in a component
cESPAOQ _/ Cooling load on the PAO loop
from the ES components
Note that the heating load in both the generators and components is the exergy
destruction rate, which is the ambient temperature multiplied by the entropy generation rate in
that component. This makes sense because the components are converting some form of
electrical energy into another with no planned heat interactions. If the components were 100%
efficient, the energy balance would indicate no active component cooling is required; however,
real components are obviously less than 100% efficient.
3.9.2 ES Exergy Model
The exergy destroyed in the ES is primarily due to power generation and component
losses. The three main components are as follows:
Generators;
Other electrical components (inverters, converters, ELMCs, etc.); and
Transmission lines (ohmic heat loss).
The equations for the ES exergy destruction are shown in Table 3.31. We now conclude with the
final subsystem, i.e. that for flight control.
Table 3.31 ES exergy destruction model equations.
Component Variable Description Model Equation
Exergy Destroyed in the ES Components
irrS Entropy generation rate
irre_in
e_outkk
ST)W
W(QT
T
dt
)STU(d
0
00 1
For steady state and no heat interaction:
irreoutein STWW 0__0
thus
0T Ambient temperature
cESxE _ Exergy destruction rate for a
component
e_inW Electrical work rate into the component
93
e_outW Electrical work rate out of the component
eouteinirrcES WWSTxE __0_
e_inclostc_ES W)(WxE 1
c Component efficiency
Exergy Destroyed in the ES Generators
genESxE _ Exergy destroyed in the generators
ES_TOggen_out PW
gen_outES_TOgengen_ES WPSTxE 0
ES_TOP Power takeoff required for the ES
gen_outW Work rate leaving the generators
g Generator efficiency
Exergy Destroyed in the Lines
transESxE _
Exergy destruction rate due to transmission losses (Elgerd, 1998)
RixE transES2
_
A
lR
i Current in the line
R Total resistance of the line Resistivity of the transmission line
A Cross sectional area of the line
l Length of the conductor
Total Exergy Destruction in the ES
ESxE Rate of exergy destruction in the ES transESgenEScESES xExExExE ___
3.10 Flight Controls Subsystem
The flight controls subsystem (FCS) consists of the actuators required to operate the
flight control surfaces as well as the control hardware associated with the actuators. The two
main types of flight control actuators that are discussed here are the traditional hydraulic
actuators and electro-mechanical/hydraulic actuators. Having one type of actuator or the other
changes many items on an aircraft. For example, traditional hydraulic actuators require a central
hydraulic subsystem (CHS) including accessory gearbox driven hydraulic pumps, triple
redundant hydraulic lines, emergency hydraulic power subsystems, a hydraulic oil reservoir, etc.
Electro-mechanical/hydraulic style actuators are localized, meaning they are installed as a single,
easily-replaceable unit that locally houses its own oil (if electro-hydraulic) or gear system
(electro-mechanical) and has no other dependence on the rest of the aircraft except for a power
connection and control connections. Electrically powered actuation devices (EPAD) have been
tested in currently built fighter aircraft and are being used on production aircraft as well.
94
3.10.1 FCS Thermodynamic Model
The FCS weight must be estimated for this model. The equation in Table 3.32 is based
on previously built fighter aircraft and requires the number of hydraulic functions, the number of
mechanical functions, the total area of the control surfaces, and the yawing moment of inertia of
the aircraft. After the weight of the FCS has been estimated, it must be included in the empty
weight of the aircraft.
Table 3.32 Flight controls subsystem weight equations (Raymer, 2006).
Component Variable Description Model Equation
Flight Controls Subsystem Weight
FCSW Weight of flight controls subsystem (kg)
127.0484.0489.0003.028.36 cscsFCS NNSMW
M Mach number
csS Total area of control surfaces (ft2)
sN Number of flight control surfaces
cN Number of crew
The actuator sizing was detailed in the CHS model so it will not be repeated here.
However, the power required by the actuators must be estimated to determine the CHS system
efficiency (see Table 3.26). The actuator power model equations are shown in Table 3.33. Note
that Majumdar (2003) states that a well designed actuator should have a range of losses from 2%
to no more than 8%. Also note that the larger diameter piston cylinders generally have lower
loss percentages because the break-away force and frictional forces are small compared to the
force being applied to the piston.
Table 3.33 FCS actuator electrical power and fluid power requirements.
Component Variable Description Model Equation
Actuator Electrical and Fluid Power
elecP Electrical power Required
t
dFP
PP
actact
act
actelec
*
actP Actual power required to
move the load (no losses)
act Actuator Efficiency act 92-98%
fluidP Fluid power supplied to the actuator15 QPPfluid *
Q Flow rate of hydraulic fluid
3.10.2 FCS Exergy Model
The FCS destroys exergy due to inefficiencies in the actuators alone, since the other
losses will be attributed to the AFS-A (drag), ES (if electrically powered actuators), or CHS (if
95
fluid powered actuators). The exergy destroyed is, thus, simply the power supplied to the
actuator, minus the actual power translated to the flight control surface by the actuator. The
exergy destruction equations are given in Table 3.34.
Table 3.34 FCS exergy destruction equations.
Component Variable Description Model Equation
Exergy Destruction rate in the FCS
FCSxE Exergy Destroyed
1
1
actreqFCS WxE
reqW Work rate required by actuator
The subsystem models included in the AAF have now been discussed. The following
chapter details the system synthesis/design, problem definition, and solution approach taken for
this thesis work.
96
Chapter 4
Large-scale System Synthesis/Design Optimization Problem
Definition and Solution Approach
This chapter discusses the optimization problem definition, optimization decision
variables and limits, subsystem integration, a decomposition approach called iterative local
global optimization (ILGO), the solution approach, and the iSCRIPT™ engineering
modeling/optimization software as applied to the fighter aircraft system.
4.1 AAF Aircraft System Synthesis/Design Optimization Problem
The fighter aircraft system consists of subsystems that are very tightly coupled in that one
subsystem can affect the other subsystem operation and/or design significantly by changing a
single design or operational decision variable. The nine subsystems modeled all play a role in
the optimization. However, some such as the ES, FCS, and CHS do so via a set of system-level
degrees of freedom (DOF) called coupling functions, while the remaining subsystems do so visa
vie both a set of coupling functions and a set of local (subsystem specific) decision variables.
The former subsystems, thus, in effect participate only passively in the optimization since they
have no local decision variables but nonetheless still alter their configurations, sizes, and energy
consumption since these are dependent on the aircraft geometry (e.g., the CHS and FCS) and
aircraft size (e.g., the ES). The interdependence of the subsystems is further illustrated by
considering the PS and ECS. The ECS is dependent on the PS bleed port air properties in that
any operational or design change in the PS requires a subsequent change in the ECS operational
decision variables. Also, any change in the requirement of bleed air in the ECS changes the
specific thrust of the PS and, thus, the overall aircraft performance.
The optimization problem presented by the fighter aircraft requires large-scale
optimization methods to handle the size of the problem. The objectives listed in Chapter 1 are
accomplished by performing the following optimizations:
Perform an optimization of the entire aircraft consisting of 9 subsystems with a
morphing-wing AFS-A and with exergy destruction and exergy fuel loss
minimization as the objective;
97
Perform an optimization of the PS and fixed-wing AFS-A with fuel burned
minimization as the objective;
Perform an optimization of the PS and morphing-wing AFS-A with fuel burned
minimization as the objective, make a comparison to the fixed-wing results and
distinguish this work from that of Butt (2005);
Perform a parametric analysis on the morphing AFS-A based on minimum fuel
burned; and
Perform a parametric analysis on the morphing AFS-A based on minimum exergy
destruction and perform analysis between minimum fuel burned results and minimum
exergy destruction results.
For the purposes of the large-scale optimizations affected in this research, a decomposition
strategy is used. It is described in the following sections. However, before giving this
description, the system-level optimization problem is defined in the next section.
4.1.1 System-Level Optimization Problem Definition
Three different objective functions are utilized in the definition of the system-level
optimization problem in this thesis: i) total exergy destruction plus fuel exergy lost, ii) gross
takeoff weight, and iii) fuel burned. Each is minimized with respect to a set of decision variables
and equality and inequality constraints.
Nine-Subsystem AAF Aircraft System-Level Optimization Problem
The first of the objective minimizations is expressed as a minimization of the total exergy
destruction, totaldesEx _ , plus the fuel exergy loss, lossfuelEx _ , due to unburned fuel lost out the
back end of the PS, i.e.,
Minimize
lossfuelFCSdesESdes
CHSdesOLSdesVCPAOSdes
FLSdesECSdesPSdesAFSdeslossfueltotaldes
ExExEx
ExExEx
ExExExExExEx
___
___
______
(4.1)
w.r.t YX
,
98
subject to 0/
FCS
ES
CHS
OLS
PAOSVC
FLS
ECS
PS
AAFS
h
h
h
h
h
h
h
h
h
H , 0/
FCS
ES
CHS
OLS
PAOSVC
FLS
ECS
PS
AAFS
g
g
g
g
g
g
g
g
g
G (4.2)
where the vectors X
and Y
represent the synthesis / design and operational decision variables,
respectively. The vectors of equality constraints, H
, represent the geometric and
thermodynamic models for each subsystem. The vectors of inequality constraints, G
, represent
the physical limits on the independent and dependent variables of the system.
Two-Subsystem AAF Aircraft System-Level Optimization Problem
The second objective function minimization is that of minimizing the gross takeoff
weight and is expressed as follows
Minimize PSAFSTO WWW (4.3)
w.r.t. PSPSAAFSAAFS YXYX
,,,
subject to
,0
AAFSh 0
AAFSg (4.4)
,0
PSh 0
PSg (4.5)
Where the vectors of equality constraints, AAFSh
and PSh
, represent the geometric and
thermodynamic models, and the vectors of inequality constraints, AAFSg
and PSg
, represent the
physical limits imposed on the independent and dependent variables. The third objective
function is minimized in a similar fashion and is given by
99
Minimize
n
iFueliburnedFuelburnedFuel iloss
WWW1
__ (4.6)
where n is the total number of mission segments, and i is the mission segment. Equation (4.6)
is minimized with respect to the same set of decision variables and subject to the same set of
constraints as for the previous objective. Furthermore, it can be shown that this last objective is,
in fact, equivalent to the previous one since the amount of fuel burned is directly proportional to
the gross takeoff weight of the aircraft. For this reason, only objectives one and three are used to
generate the results presented in Chapter 5.
Now having defined the system-level objective of the optimization, a discussion of the
need for decomposition in large-scale optimization is presented next.
4.1.2 Need for Decomposition
Large-scale engineering optimization problems may require decomposition techniques or
strategies in order to make the manageable or even solvable. Such techniques reduce the
problem into sub-problems. The aircraft optimization problem in this thesis work is much too
large to handle as a single problem; and thus, both physical and time decomposition strategies
are used. Physical decomposition results in a set of aircraft subsystems and boundaries, which
for the aircraft of this thesis are detailed in Chapter 3. The particular physical decomposition
technique used here is called iterative local-global optimization (ILGO) developed by Muñoz
and von Spakovsky (2001a,b, 2003) and applied to a 5-subsystem fighter by Rancruel and von
Spakovsky (2003, 2004).
Time decomposition is used to split the mission, represented by a timeframe from
mission start to mission completion, into the stationary mission segments of a quasi-stationary
description. The mission segments are defined based on the aircraft flight characteristics (e.g.
climb, cruise, loiter, accelerate, etc.) or requirements (e.g. drop a payload, fly a distance, loiter an
amount of time). Individual mission segments are modeled as steady state and transient behavior
is approximated via the quasi-stationary description. The resulting mission segments are given
in Table 3.1 of Chapter 3. Initially, every mission segment was split into 5-7 smaller segments to
increase fuel burn/gross weight calculation accuracy. However, it was found that in mission
segments with little-changing flight characteristics (e.g., loiter, cruise), a single time step gave
sufficient accuracy. Note that decreasing the number of time steps in each segment significantly
100
decreases the computational burden as well, since the aerodynamics, weight fractions, etc. are
updated for each time step within the mission segment. Now, before applying these
decomposition strategies to the system-level optimization problems, a description of ILGO is
given in the following sections.
4.2 Iterative Local-Global Optimization (ILGO) Approach
The purpose of this section is not to give a detailed discussion of ILGO since it was not
developed in this thesis work, but rather to give a top-level overview of the ILGO decomposition
strategy for large-scale optimization. The reader is referred to Muñoz and von Spakovsky
(2000a,b,c,d; 2001a,b) for details on this strategy. A basic discussion of local-global
optimization (LGO), however, must precede the discussion of ILGO since the former is the basis
for most if not all of the physical decomposition strategies found in the literature. In the process,
the differences between LGO and ILGO are highlighted and the uniqueness of the ILGO
approach revealed.
4.2.1 Local-Global Optimization (LGO)
Section 2.3 of Chapter 2 discusses the existence / requirement of the ORS (see Figure 2.7)
in the LGO approach and also the computational burden required to explicitly or implicitly
generate the ORS for a problem of any magnitude. For further explanation, consider the
following system problem that has been decomposed into two subsystems or units as shown in
Figure 4.1. The vectors 1Z
and 2Z
are the so-called local-level decision (i.e., independent)
variables for each unit and the arrows between the two units are represented by vectors of the
functions, 12u
and 21u
, that couple the two units.
Figure 4.1 Physical decomposition of a 2-unit system.
1Z 2Z
21
12
u
u
2R 1R
101
The system-level objective function, C , is expressed as a sum of the unit-contributions,
i.e., 21 CC . Each of the unit contributions to the system-level objective function has a set of
terms that define the unit contribution. For example, for the first unit, one might have that
1111 capitalCRkC (4.7)
In equation (4.7), 1R is an external resource used by unit 1 (e.g., fuel or power) and 1capitalC is a
function related to the size of the unit (e.g., weight/volume or cost) while 1k is a conversion
factor.
For each of the units, the coupling functions bring resources from the other unit
depending on the needs of the receiving unit. The amount of these resources and the external
ones depend on the values of the decision variables of the coupled units. So, looking only at the
contribution of the first unit to the system-level objective function, C , the following expression
is written:
1212121121111 1,,,,,, ZCZZZuZZZuZRkC capitalsyssys
(4.8)
where 1capitalC is, for example, the capital cost in a thermoeconomic problem and sysZ
the set of
system-level decision (i.e., independent) variables which cannot be assigned strictly locally, i.e.,
to one unit only. In thermodynamic problems, this term is either ignored as is done in a
stationary system problem or is converted into a physical term such as weight for a non-
stationary system problem. The reader is referred to the Evans-El-Sayed formalism for more
information about thermoeconomics in the context of decomposition (Evans and El-Sayed,
1970).
Considering both units with the system-level optimization problem is as follows:
Minimize
221212112222
121212112111
2
1
,,,,,,
,,,,,,
ZCZZZuZZZuZRk
ZCZZZuZZZuZRkC
capitalsyssys
capitalsyssys
(4.9)
w.r.t. sysZZZ
,, 21
subject to the primary constraints
02
1
h
hH (4.10)
102
02
1
g
gG (4.11)
and to the secondary constraints
0,, 2112
sysZZZu (4.12)
0,, 2121
sysZZZu (4.13)
where the vectors of equality constraints, H
, represent the physical and thermodynamic models
for each subsystem. The vectors of inequality constraints, G
, represent the physical limits on
the independent and dependent variables associated with each unit in the system. Equations 4.12
and 4.13 indicate that the coupling functions take on the values
and
constrained by the
following two expressions:
maxmin 1212 uu
(4.14)
maxmin 2121 uu
(4.15)
Problem (4.9) can be physically and mathematically decomposed into a set of two subproblems
which must be repeatedly solved for different values of
and
constrained within the limits
set by expressions (4.14) and (4.15). The two subproblems are, thus, expressed as
Subproblem 1:
Minimize 11111 1,, ZCZRkC capital
(4.16)
w.r.t. 1Z
subject to 01
h (4.17)
01
g (4.18)
Subproblem 2:
Minimize 22222 2,, ZCZRkC capital
(4.19)
w.r.t. 2Z
subject to 02
h (4.20)
02
g (4.21)
Note that for a subproblem, either
is fixed for a given optimization and
is calculated (i.e., a
result of the subproblem optimization) or vise versa. For example, in the case of subproblem 2,
if
is fixed for subproblem 1, then it must be calculated in subproblem 2; conversely, if
is
103
fixed for subproblem 1, then it must be calculated in subproblem 2. Also, note that the repeated
optimizations required of subproblems 1 and 2 in effect result in a set of nested optimizations for
which the inner part of the nest is comprised of the set of subproblem optimizations and the outer
part of the nest of a single system-level optimization with the
and
as decision variables.
Thus, the LGO decomposition results in a multi-level optimization which reduces an
overall system-level problem into a set of smaller subsystem problems which in theory should be
easier to solve. In fact, for large-scale system optimization problems involving many degrees of
freedom and non-linearities, such a multi-level approach may be the only way to arrive at a
solution. However, it introduces an additional computational burden due to the nesting. This
may become so large that it renders the problem unsolvable at least from a practical standpoint.
The multi-level optimization resulting from the application of LGO to the two-unit system of
Figure 4.1 is illustrated in Figure 4.2.
Figure 4.2 Multi-level optimization resulting in a set of nested optimizations.
Note that in this figure, there are a total of three optimization loops shown: one each for units 1
and 2 as well as a third at the system-level. Multi-level optimization requires a new unit-level
optimization for each iteration (i.e., completion) of the system-level optimization loop. As
already mentioned, such nesting can become very computationally burdensome. ILGO was
Unit- system
interactions
System-level optimization
Unit-level optimizations
System - optimizer
Simulations
Unit 1 Unit 2
104
specifically developed to address this difficulty. The following section gives an overview of this
approach.
4.2.2 ILGO Approach
Iterative local-global optimization (ILGO) is a decomposition strategy for large-scale
optimization developed by Muñoz and von Spakovsky (2000a,b,c,d; 2001a,b) which eliminates
the need for nested optimizations found with LGO. It not only reduces the computational burden
of multi-level optimization but also allows the optimal decentralized development of aircraft
subsystems in the context of an overall system-level optimization.
ILGO eliminates the nesting of LGO by embedding at the unit-level the system-level
information found in the outer loop of the LGO optimization. In doing so, the outer optimization
loop is no longer needed since it is implicitly present at the unit-level. The embeddiong is
accomplished via a set of coupling functions and a set of associated “shadow prices”. The
coupling functions are simply the unit-level or subsystem-level interactions required as inputs or
outputs by each physical unit. These functions, thus, “couple” or integrate the unit-level
problems with each other. Figure 4.3 shows an example of a simple three unit system with
2112 ,uu
unit 2 and unit 1 interactions
3113,uu
unit 1 and unit 3 interactions
3223,uu
unit 2 and unit 3 interactions
Figure 4.3 An example of three subsystems and their associated coupling functions.
coupling functions. The functions’ associated shadow prices measure changes in the optimal
values of the local (unit-level) functions with respect to changes in the coupling functions. The
shadow prices allow the decomposed optimizations to progress along the system-level ORS in
the direction of the system-level optimal solution. Furthermore, the unit-level objective
functions used by LGO are morphed into system-level unit-base objective functions in ILGO.
For example, the unit-level objective functions (equations (4.16) and (4.19)) of the two-unit
3113,uu
3223,uu
2112 ,uu
Unit 2
Unit 1
Unit 3
105
system depicted in Figure 4.1 are transformed into system-level unit-base objective functions as
follows:
)1(210
221
)1(120
2120
*211' uuCCC (4.22)
)2(210
121
)2(120
1120
*122 ' uuCCC (4.23)
where the last here terms in each of these equations are 1st order Taylor Series expansions about
some reference point “o” on the ORS relative to changes of the coupling functions 12u and 21u
with respect to unit 1 (equation (4.22)) or unit 2 (equation (4.24)), respectively. The s' in these
equations are the partial derivatives associated with these expansions. They are, in fact, the
shadow prices. Note that for purposes of simplification, the vectors 12u
and 21u
in equations
(4.16) and (4.19) have been assumed to contain only a single coupling function each (i.e. 12u or
21u ). As should be evident, optimizing equations (4.22) or (4.23) in effect optimizes not only
the local objective (equation (4.16) or (4.19)) of each unit but each unit’s system-level effects via
the additional terms appearing in equations (4.22) and (4.23). Thus, each local optimization not
only optimizes the unit but also optimizes the system as a whole. For more details of the ILGO
formalism, the reader is referred to Muñoz and von Spakovsky (2000a,b,c,d; 2001a,b), Rancruel
(2002, 2005), and Rancruel and von Spakovsky (2005, 2006).
To understand how the process of optimization with ILGO proceeds, a notional flow
diagram is given in Figure 4.5. As with Figure 4.2, there are two units participating in the
optimization. The key differences with ILGO are no system-level simulations required and the
outer loop optimization has been eliminated. Each system-level, unit-based optimization results
in a new set of shadow price and coupling function values which are then used in a new ILGO
iteration unless convergence has been reached. Based on past experience, the number of
iterations required for convergence for even complex systems is less than six or seven.
In the following sections, the ILGO decomposition is applied to the AAF aircraft
optimization problem which is the subject of this thesis.
106
4.3 System-Level, Unit-Based Synthesis/Design Optimization Problem Definitions
Section 4.1 defines the system-level optimization problem. The following sections define
the system-level, unit-based synthesis/design optimization problems for each subsystem starting
with the AFS-A. However, the subsystem integration and coupling functions for the AAF fighter
are defined first in the next section.
4.3.1 Subsystem Integration and Coupling Functions
Subsystem integration is necessary for the overall system optimum synthesis/design to be
found. Subsystem integration, in a programming sense, requires subsystem interactions to be
defined explicitly between the subsystem models. In the context of the fighter aircraft system,
subsystem integration is a very involved process due to the number of subsystem interactions
present. An additional difficulty is added to the integration process when a subsystem has a two
– way interaction (e.g., a heating interaction between two subsystems) since this may involve a
pinch temperature difference constraint that affects both subsystems.
Unit-level optimizations
New ILGO Iteration
- optimizer Decomposition
(define subsystem boundaries, coupling)
System-level information is sent to the system-level unit-base subproblems
(shadow prices, coupling function changes)
System-level, unit-based optimization results
ILGO Start
Unit 1 Unit 2
Convergence
Figure 4.4 Notional flow diagram of the application of the ILGO decomposition strategy to the two-unit system of Figure 4.2.
107
While a subsystem interaction is simply an operating parameter (e.g., bleed air
temperature) or a design parameter (e.g., subsystem weight) required by one subsystem from
another, applying the ILGO decomposition to a system requires that the subsystem interaction
not be treated as a real-time interaction as it would in an overall system simulation, but rather as
an independent variable or fixed parameter during a local or unit-level optimization. The
interactions are updated after each ILGO iteration to effectively maintain the subsystem
interactions despite the de-coupled local optimizations. The subsystem interactions are, thus,
called “coupling functions” within the context of ILGO as they are not to be confused with the
subsystem interactions as found in the LGO decomposition which are maintained in real time
and result in the multi-level optimizations of the LGO approach. The reader is referred to the
work of Muñoz and von Spakovsky (2000a,b,c,d; 2001a,b), Rancruel (2002, 2005), and Rancruel
and von Spakovsky (2005, 2006) for details on coupling functions and the updating methods.
The AAF aircraft system of this thesis has numerous subsystem interactions and each
subsystem is highly dependent on the other subsystems for both design and operational
considerations. The fighter aircraft subsystem interactions are shown in Figure 4.5 with an
explanation of the main interactions following the figure. Figure 4.5 does not give a
comprehensive list of all of the subsystem interactions; however, Figure 4.5 does clearly
illustrate the interdependence and tight coupling of the aircraft subsystems. A comprehensive
list of all coupling functions is not given here because of the very large number involved, i.e.
488. However, the number associated with each subsystem is given in Table 4.1 where for the
Table 4.1 Number of coupling functions associated with each subsystem.
Subsystem Input (fixed within
subsystem) coupling functions
Output (variable within subsystem) coupling
functions
AFS-A 88 88
PS 115 162
ECS 93 72
VC/PAOS 50 41
FLS 94 41
ES 6 21
FCS 6 1
CHS 6 41
OLS 30 21
Totals 488 488
108
application of ILGO used here the input coupling function and associated shadow price values
are assumed fixed in any given ILGO iteration while the output coupling function and shadow
price values are determined by each subsystem optimization in a given ILGO iteration.
Interaction
2112 ,uu
Uninstalled thrust, OLS power takeoff
3223,uu
Bleed air properties, bleed air mass flow required
4224 ,uu
Fuel flow rate required, FLS power takeoff
62u
ES power takeoff
5115 ,uu
OLS weight, mission segment time
5445,uu
FLS weight, drag, ram air properties, mission segment time
5225,uu
PS weight, drag, AFS-A thrust required, specific fuel consumption, mission segment time
6556 ,uu
ES weight, TOW , mission segment time
5335,uu
ECS weight, drag, ram air properties, mission segment time
9339 ,uu
Bleed air/PAO heat exchanger properties
7557 ,uu
CHS weight, mission segment time
8778,uu
Control surface sizing, hydraulic oil flow rate required
8558 ,uu
FCS weight, aircraft wing and tail geometry, mission segment time, flight conditions
9559 ,uu
VC/PAOS weight, drag, ram air properties, mission segment time
Figure 4.5 Aircraft subsystem interactions and coupling functions.
Note that each of the subsystem connections in Figure 4.1 has a two way arrow which
indicates that there is a coupling function vector for each direction of the interaction. The one
exception is the interaction between the ES and PS, 62u
, which represents the shaft power
takeoff (i.e. power extraction from the low pressure turbine), ,TOP required from the PS to
generate power for the aircraft. There is no interaction required by the PS from the ES.
2. PS
PSPS
PSPS
GH
YX
,
,
5. AFS
AFSAFS
AFSAFS
GH
YX
,
,
3. ECS
ECSECS
ECSECS
GH
YX
,
, 4. FLS
FLSFLS
FLSFLS
GH
YX
,
,
9. VC/PAOS
VCPAOSVCPAOS
VCPAOSVCPAOS
GH
YX
,
,
6. ES
ESES GH
,
1. OLS
OLSOLS
OLSOLS
GH
YX
,
,
8. FCS
FCSFCS GH
,
7. CHS
CHSCHS GH
,
5115 ,uu
8778,uu
8558 ,uu
12u
2112 ,uu
3223,uu
9339 ,uu
5225,uu
5335 ,uu
9559 ,uu
62u
5445,uu
4224,uu
6556 ,uu
7557 ,uu
109
The subsystem interactions are now briefly discussed starting with the AFS-A. Every
aircraft subsystem has a weight associated with it, and each subsystem weight must be defined as
a coupling function to the AFS-A. In addition, the AFS-A requires a certain amount of thrust to
fly the aircraft as well as specific fuel consumption, ,sfc rates for the corresponding thrust.
Thus, both the thrust required and the corresponding sfc are interactions with the PS. Also, the
subsystems that are associated with inlet ducts (i.e., the ECS, FLS, and VC/PAOS) have
momentum drag associated with the inlet / exit ducts and these drag values are an interaction
with the AFS-A as well. An additional AFS-A subsystem interaction results from the wing
geometry of the AFS-A and the FCS. The wing geometry is used as a basis for estimating
control surface sizing in the FCS. Note that the morphing-wing AFS-A wing geometry that is
used to size the FCS is the same geometry as used for establishing the wing weight (see
subsection 3.2.5).
The PS is the next most highly integrated and interdependent subsystem. The PS requires
power takeoff (or power extraction from the low-pressure spool), ,TOP values from the OLS, ES,
FLS, CHS, and VCPAOS (the CHS and VCPAOS power takeoff interactions with the PS are not
shown in Figure 4.1). The ECS requires bleed air temperature and pressure from both the high
and low pressure bleed ports of the PS. In addition, the ECS specifies the bleed air mass flow
rate required from the PS, which affects the total performance of the PS depending on the mass
flow rate required and which bleed port is selected by the ECS. The OLS requires uninstalled
thrust values from the PS to estimate the heating load on the FLS/OLS heat exchanger. The PS
also outputs the sfc for each mission segment to the FLS (and to the previously mentioned AFS-
A).
The VC/PAOS interfaces with the ECS via the bleed air / polyalphaolefin (PAO) heat
exchanger. The PAO temperature and heat exchanger physical geometry are all coupling
functions between the ECS and VC/PAOS. The model and ILGO decomposition was initially
set up to only have a heat interaction as a coupling function, but the optimization then created a
thermodynamically impossible heat exchanger (i.e., the exit temperature of the bleed air side or
“hot side” of the heat exchanger was significantly lower than the PAO side or “cold side” exit
temperature). The thermodynamic characteristics of the heat exchanger, thus, had to be “owned”
by one of the unit-level optimizations to avoid this scenario. Consequently, the bleed air/PAO
110
heat exchanger geometry is sized in the ECS code, while the heat exchanger weight and exergy
destruction is accounted to the VC/PAOS.
The interdependence of the subsystems can be seen very easily when one observes the
effects on the rest of the aircraft of changing one subsystem parameter. For example, if the
weight of a subsystem increases 100 kg, the aircraft requires more lift during cruise, more thrust
during acceleration, and subsequently more fuel. Also, if the increased weight is due to a
subsystem requiring additional cooling (e.g., additional power generating capacity in the ES
increases the subsystem weight and cooling requirements), then the VC/PAOS is affected and
perhaps even the ECS ram air inlet duct sizing, etc. Needless to say, if one item on an aircraft is
changed significantly, it generally requires a synthesis / design or operational change in many
other subsystems on the aircraft due to the high level of integration in such systems. This effect
is even more pronounced if the aircraft is being optimized to find the best synthesis/design and if
“over designing” a subsystem is not an acceptable solution.
Having outlined and discussed the coupling functions for the AAF aircraft system and its
subsystems, the optimization decision variables and their associated variable constraints are
discussed next.
4.3.2 AFS-A System-Level, Unit-Based Synthesis/Design Optimization Problem
Definition
The AFS-A synthesis/design and operational decision variables are presented in Section
4.4 as are the variable inequality constraints, which are based on typical fighter aircraft
configurations (Raymer, 2002) and previous work done in fighter aircraft optimization
(Periannan, 2004; Rancruel, 2005; Butt, 2005; Smith et al. 2007). The system-level objectives
that used for the AFS-A in this thesis work are minimizing the exergy destruction plus exergy
loss due to unburned fuel exiting the rear of the PS and minimizing the fuel burned. Both are
determined with respect to the entire mission.
Now, it is only for the former system level objective, which is used for the nine-
subsystem AAF aircraft system optimization, that the ILGO decomposition strategy is used.
Thus, a set of system-level unit-base optimization problems for the nine-subsystems must be
determined as described in Section 4.2. That for the AFS-A is as follows:
111
AFS-A System-Level, Unit-Based Exergy Destruction Optimization Problem
Minimize
n
iireqireqTiisfciousij
iousiju
j iAltM
j iiit
j
n
iijij
Dj
jjWFCSdes
CHSdesESdesOLSdesPAOSVCdes
FLSdesECSdesPSdesAAFSdesAAFSdes
Tsfcu
AltMt
DWEx
ExExExEx
ExExExExEx
1varvar
3
1
5
1,
6
1
5
1
4
1 1
8
1
*
****/
***
,
'
(4.24)
w.r.t. AAFSAAFS YX
,
subject to
0
AAFSh (4.25)
0
AAFSg (4.26)
where the vector of equality constraints, AAFSh
, represents the aerodynamic and geometric
model of the AFS-A and the vector of inequality constraints, AAFSg
, the physical limits
imposed on the subsystem. The superscript “*” indicates the optimum value from the previous
ILGO iteration of a given subsystem objective function.
Problem (4.24) represents the system-level minimization resulting from varying the AFS-
A (or local) decision variables only. The local or unit-based objective function is denoted by
AAFSdesEx
and the expression for the system-level, unit-based objective function, AAFSdesEx
' ,
includes products of paired shadow prices and associated coupling functions. As to the shadow
prices given in problem (4.24), the first summation represents the subsystem weights, the first
double summation term represents the drag of the inlet ducts (one term for each mission segment
for the ECS, FLS, and VC/PAOS), the double summation term represents the time for the
mission segments that are functions of the mission decision variables (i.e., the mission segment
112
time is determined by ,1BCM ,2BCM ,BCLM etc.), the next double summation is for the three
subsystems that require the mission decision variable values from the AFS-A (i.e., the PS, ECS,
and FLS), the next coupling function represents the miscellaneous AFS-A coupling functions not
listed here, and the last summation couples the AFS-A and the PS with the thrust required and
fuel consumption rate. All told, the AFS-A has 88 input and 88 output shadow prices (see
Section 4.3.1). The shadow prices for the coupling functions given in Problem (4.24) are defined
as
8
1
*
j j
jdes
jW W
Ex (4.27)
3
1 1
*
j
n
i ij
jdes
ijD D
Ex (4.28)
8
1
5
1
*
j i i
jdes
it t
Ex (4.29)
3
1
5
1
*
, ,j i i
jdes
iAltM AltM
Ex (4.30)
iousij
jdes
iousiju u
Ex
var
*
var
(4.31)
and the final two shadow prices represent a total of 40 unique shadow prices, one for each of the
mission segment legs excluding the payload drop mission segment
i
dessfc sfc
ExAAFSPS
i
*
(4.31)
i
AAFSPS
ireqreq
desT T
Ex
*
(4.32)
Note that there are both input and output shadow prices for every subsystem. For example, an
input shadow price for the AFS-A is equation (4.24) since the specific fuel consumption, ,sfc is
an interaction with the PS and is defined in the PS. Thus, the value of the shadow price is
constant within the AFS-A system-level, unit-based optimization. In contrast, an output shadow
113
price example is equation (4.32) as the thrust required to fly the aircraft is determined in the
AFS-A and is output to the PS. Were all the shadow prices for the AFS-A explicitly stated here,
88 input and 88 output shadow prices would be defined for the AFS-A for a total of 176 shadow
prices.
Equations (4.27) through (4.32) represent the effect of the marginal changes in the
optimum value of the system-level, unit-based objective function for the AFS-A due to changes
in the coupling functions. Problem (4.24) has sixteen additional terms, ,jW ,ij
D ,it
,, AltM ,variousiju
,
isfc
and
ireqT that represent the variations in the coupling functions
and are defined as
8
1
*
jjjj WWW
(4.33)
3
1 1
*
j
n
iijijij
DDD
(4.34)
6
1
5
1
*
j iijijij
ttt
(4.35)
6
1
5
1
*,,,j i
ijijijAltMAltMAltM
(4.36)
*varvarvar iousijiousijiousij uuu (4.37)
n
iiii
sfcsfcsfc1
* (4.38)
n
iireqireqireq TTT
1
* (4.39)
where the superscript "" indicates the optimum coupling function value from the previous ILGO
iteration. Similarly to the shadow prices above, equations (4.33) to (4.39) are not
comprehensive; the AFS-A has 176 total equations, one for each shadow price, but the examples
provided here are adequate for generating the other equations. The PS system-level, unit-based
synthesis/design optimization problem definition is next.
114
4.3.3 PS System-Level, Unit-Based Synthesis/Design Optimization Problem
Definition
The PS system-level, unit-based optimization problem is defined as follows:
PS System-Level, Unit-Based Exergy Destruction Optimization Problem
Minimize
n
i
iinstiinstT
ihlbleedihlbleedPihlbleedihlbleedT
ibleedibleedmiTOiTOPiisfc
iousiiousiui
iitPSPSWCHSdes
ESdesOLSdesESdesPAOSVCdes
ECSdesAAFSdeslossfueldesPSdeslossfueldesPSdes
T
PT
mPsfc
utWEx
ExExExEx
ExExExExExEx
1,_,_,_,_
varvar
5
1
*
****/
**__
''
(4.40)
w.r.t. PSPS YX
,
subject to
0
PSh (4.41)
0
PSg (4.42)
where the vector of equality constraints, PSh
, represents the thermodynamic and geometric
model of the PS and the vector of inequality constraints, PSg
, the physical limits imposed on the
subsystem. The superscript “*” indicates the optimum value from the previous ILGO iteration of
a given subsystem objective function.
Problem (4.40) represents the system-level minimization resulting from varying the PS
decision variables only. The local or unit-based minimization objective function is denoted by
PSdesEx plus an additional term, ,_ lossfueldesEx to account for the unburned fuel lost out the back
of the PS expressed. The expression for the unit-based, system-level objective function,
lossfuelPS desdes ExEx_
'' , includes products for paired shadow prices and associated coupling
functions. The first represents the weight of the PS, the second represents the time coupling with
115
the AFS-A, the third represents the various other coupling functions, and the six within the
summation are the PS interactions. The shadow prices are defined as
PS
AAFSdes
PSW W
Ex
*
(4.43)
5
1
*
i i
PSdes
it t
Ex
(4.44)
iousij
jdes
iousiju u
Ex
var
*
var
(4.45)
And the next six shadow prices which represent a total of 240 independent shadow prices are
defined as
i
PAOSVCFLSAAFSdes
isfc sfc
Ex
*/,, (4.46)
iTO
CHSPAOSVCESFLSdes
iTOP P
Ex
*,/,, (4.47)
ibleed
ECSdes
ibleedm m
Ex
*
(4.48)
ihlbleed
ECSdes
ihlbleedT T
Ex
,_
*
,_
(4.49)
ihlbleed
ECSdes
ihlbleedP P
Ex
,_
*
,_
(4.50)
iinst
OLSAAFSdes
iinstT T
Ex
*, (4.51)
where equations (4.44), (4.45), (4.46), (4.47), (4.48), (4.49), (4.50), and (4.51) represent a total
of 5, 11, 60, 80, 20, 40, 40, and 20 unique shadow prices, respectively. All told, the PS has 115
input, or fixed, shadow prices and 162 output (or variable i.e. internally calculated) shadow
prices for a total of 277 shadow prices for the PS.
Equations (4.43) through (4.51) represent the effect of marginal changes in the optimum
value of the system-level, unit-based objective function for the PS due to changes in the coupling
116
functions. Problem (4.40) has nine additional terms, ,PSW ,it ,variousiu
,i
sfc ,iTOP
,ibleedm ,,_ ihlbleedT ,,_ ihlbleedP and
iinstT that represent variations in the coupling functions
listed above and are defined as
*PSPSPS WWW (4.52)
5
1
0*
iiii ttt (4.53)
*varvarvar iousijiousijiousij uuu (4.54)
*
ii sfcsfcsfci
(4.55)
iTOTOTO PPPii
* (4.56)
ibleedbleedbleed mmmii
* (4.57)
ihlbleedhlbleedhlbleed TTT
ii*
,_,_,_ (4.58)
ihlbleedhlbleedhlbleed PPP
ii*
,_,_,_ (4.59)
iinstinstinst TTTii
* (4.60)
where the superscript "" indicates the optimum coupling function value from the previous ILGO
iteration. Similarly to the AFS-A, equations (4.52) to (4.60) are not comprehensive; the PS has
277 total equations, one for each shadow price, but the examples provided here are adequate for
generating the other equations. The ECS system-level, unit-based synthesis/design optimization
problem definition is next.
4.3.4 ECS System-Level, Unit-Based Synthesis/Design Optimization Problem
Definition
The ECS system-level, unit-based optimization problem is defined as follows:
ECS System-Level, Unit-Based Exergy Destruction Optimization Problem
117
Minimize
n
i ihlbleedihlbleedPihlbleedihlbleedT
ibleedibleedmiECSiECSD
iousiiousiui
iit
inletinletAHXPAObleedHXPAObleedW
ECSECSWPAOSVCdes
PSdesAAFSdesECSdesECSdes
PT
mD
ut
AW
WEx
ExExExEx
1 ,_,_,_,_
varvar
5
1
_/_/
*/
**'
(4.61)
w.r.t. ECSECS YX
,
subject to
0
ECSh (4.62)
0
ECSg (4.63)
where the vector of equality constraints, ECSh
, represents the thermodynamic and geometric
model of the ECS and the vector of inequality constraints, ECSg
, the physical limits imposed on
the subsystem. The superscript “*” indicates the optimum value from the previous ILGO
iteration of a given subsystem objective function.
Problem (4.61) represents the system-level minimization resulting from varying the ECS
decision variables only. The local or unit-based minimization objective function is denoted by
.ECSdesEx The expression for the unit-based, system-level objective function,
ECSdesEx' , includes
products for paired shadow prices and associated coupling functions. The first shadow price
represents the weight of the ECS, the second represents the weight of the bleed air / PAO heat
exchanger, the third represents the area of the inlet for the ram air / PAO heat exchanger, the
fourth represents the time coupling with the AFS-A, the fifth represents the various other
coupling functions, and the four within the summation are the ECS interactions. The shadow
prices are defined as
118
ECS
desW W
ExECSAAFS
ECS
*
(4.64)
HXPAObleed
PAOSVCdes
HXPAObleedW W
Ex
_/
*/
_/
(4.65)
inlet
desA A
ExECSPAOSVC
inlet
*/ (4.66)
5
1
*
i i
ECSdes
it t
Ex
(4.67)
iousij
jdes
iousiju u
Ex
var
*
var
(4.68)
and the next four shadow prices which represent 144 independent shadow prices are defined as
iECS
AAFSdes
iECSD D
Ex
*
(4.69)
ibleed
PAOSVCPSdes
ibleedm m
Ex
*/, (4.70)
ihlbleed
PAOSVCPSdes
ihlbleedT T
Ex
,_
*/,
,_
(4.71)
ihlbleed
PAOSVCPSdes
ihlbleedP P
Ex
,_
*/,
,_
(4.72)
where equations (4.67), (4.68), (4.69), (4.70), (4.71), and (4.72) represent a total of 5, 13, 28, 28,
48, and 48 unique shadow prices, respectively. The ECS has a total of 93 input and 72 output
shadow prices for a total of 165 shadow prices.
Equations (4.64) through (4.72) represent the effect of marginal changes in the optimum
value of the system-level, unit-based objective function for the ECS due to changes in the
coupling functions. Problem (4.61) has nine additional terms, ,ECSW ,_/ HXPAObleedW ,inletA
,it ,variousiu
,iECSD ,
ibleedm ,,_ ihlbleedT andihlbleedP ,_ that represent variations in the
coupling functions listed above and are defined as
*ECSECSECS WWW (4.73)
119
*_/_/_/ HXPAObleedHXPAObleedHXPAObleed WWW (4.74)
*inletinletinlet AAA (4.75)
5
1
0*
iiii ttt (4.76)
*varvarvar iousijiousijiousij uuu (4.77)
*iii ECSECSECS DDD (4.78)
ibleedbleedbleed mmm
ii*
(4.79)
ihlbleedhlbleedhlbleed TTTii
*,_,_,_
(4.80)
ihlbleedhlbleedhlbleed PPPii
*,_,_,_
(4.81)
where the superscript "" indicates the optimum coupling function value from the previous ILGO
iteration. Equations (4.73) to (4.81) are not comprehensive but rather represent 165 total
equations for the ECS, one for each shadow price, but the examples provided here are adequate
for generating the other equations. The VC/PAOS system-level, unit-based synthesis/design
optimization problem definition is next.
4.3.5 VC/PAOS System-Level, Unit-Based Synthesis/Design Optimization Problem
Definition
The VC/PAOS system-level, unit-based optimization problem is defined as follows:
VC/PAOS System-Level, Unit-Based Exergy Destruction Optimization Problem
Minimize
120
4
1 ,_,_
,_,_
1
_/_/'//
varvar
5
1
_/_/
//*
/
*
/
*
/
*
///'
i ihlbleedihlbleedP
ihlbleedihlbleedTibleedibleedm
n
i FLSfuelFLSfuelm
iHXPAOfueliHXPAOfuelQiPAOSVCiPAOSVCD
iousiiousiui
iit
inletinletAHXPAObleedHXPAObleedW
PAOSVCPAOSVCWPAOSVCFLSdesPAOSVCECSdes
PAOSVCPSdesPAOSVCAAFSdesPAOSVCdesPAOSVCdes
P
Tm
m
QD
ut
AW
WExEx
ExExExEx
(4.82)
w.r.t. PAOSVCPAOSVC YX // ,
subject to
0/
PAOSVCh (4.83)
0/
PAOSVCg (4.84)
where the vector of equality constraints, PAOSVCh /
, represents the thermodynamic and geometric
model of the VC/PAOS and the vector of inequality constraints, PAOSVCg /
, the physical limits
imposed on the subsystem. The superscript “*” indicates the optimum value from the previous
ILGO iteration of a given subsystem objective function.
Problem (4.82) represents the system-level minimization resulting from varying the
VC/PAOS decision variables only. The local or unit-based minimization objective function is
denoted by PAOSVCdesEx
/while the expression for the system-level, unit-based objective function,
PAOSVCdesEx/
' , includes products for paired shadow prices and associated coupling functions.
The first pair accounts for the effects of changes in the weight of the VC/PAOS, while the
second pair represents the weight of the bleed air / PAO heat exchanger, the third represents the
area of the inlet for the ram air / PAO heat exchanger which is sized in the ECS. The fourth pair
represents the time coupling with the AFS-A, the fifth represents the various other coupling
functions. The pairs within the summation from i to n account for the effects of changes in the
121
VC/PAOS inlet drag, the heat interaction with the FLS via the fuel / PAO heat exchanger, and
the mass flow rate of the fuel in the FLS while the three shadow price / coupling function pairs
within the final summation account for the effects of changes in the bleed air / PAO heat
exchanger working fluid properties.
The formulation of the bleed air / PAO heat exchanger pair is unique to the VC/PAOS
and ECS. The heat exchanger is “owned” by the ECS and the influence of this heat exchanger
on both the ECS and VC/PAOS is represented by more than just a heat interaction coupling
function. The properties of both working fluids (bleed air and PAO) are defined as coupling
functions so the optimization does not find a thermodynamically impossible heat exchanger such
as one having the hot-side exit temperature lower than the cold-side inlet temperature. Thus,
additional coupling functions are defined for the PAO working fluid properties at the bleed
air/PAO heat exchanger so that the optimization can size the heat exchanger properly.
Furthermore, the heat exchanger weight is passed from the ECS to the VC/PAOS as a coupling
function as the bleed air / PAO heat exchanger weight is accounted in the VC/PAOS. The
shadow prices for the VC/PAOS are defined as
PAOSVC
AAFSdes
PAOSVCW W
Ex
/
*
/
(4.85)
HXPAObleed
AAFSdes
HXPAObleedW W
Ex
_/
*
_/
(4.86)
inlet
ECSdes
inletA A
Ex
*
(4.87)
5
1
*/
i i
PAOSVCdes
it t
Ex
(4.88)
iousij
jdes
iousiju u
Ex
var
*
var
(4.89)
and the next six shadow prices, which represent 84 independent shadow prices, are defined as
iPAOSVC
AAFSdes
iPAOSVCD D
Ex
/
*
/
(4.90)
122
iHXPAOfuel
FLSdes
iHXPAOfuelQ Q
Ex
_/
*
_/
(4.91)
iFLSfuel
FLSdes
iFLSfuelm m
Ex
*
(4.92)
ibleed
ECSdes
ibleedm m
Ex
*
(4.93)
ihlbleed
ECSdes
ihlbleedT T
Ex
,_
*
,_
(4.94)
ihlbleed
ECSdes
ihlbleedP P
Ex
,_
*
,_
(4.95)
where equations (4.88), (4.89), (4.90), (4.91), (4.92), (4.93), (4.94), and (4.95) represent a total
of 5, 4, 20, 20, 20, 8, 8, and 8 unique shadow prices, respectively. The VC/PAOS has a total of
50 input shadow prices and 48 output shadow prices for a total of 96 shadow prices.
Equations (4.85) through (4.95) represent the effect of the marginal change in the
optimum value of the coupling function on the system-level, unit-based objective function for the
VC/PAOS due to changes in the coupling functions. Problem (4.82) has seven additional terms,
,/ PAOSVCW ,_/ HXPAObleedW ,inletA ,it ,variousiu
,/ iPAOSVCD ,_/ iHXPAOfuelQ ,FLSfuelm
,ibleedm ,,_ ihlbleedT and ihlbleedP ,_ that represent the effect that the variation in the
VC/PAOS decision variables has on the coupling functions listed above, equations (4.85)
through (4.95), and are defined as
*/// PAOSVCPAOSVCPAOSVC WWW (4.96)
*_/_/_/ HXPAObleedHXPAObleedHXPAObleed WWW
(4.97)
*inletinletinlet AAA (4.98)
5
1
0*
iiii ttt (4.99)
*varvarvar iousijiousijiousij uuu (4.100)
123
*/// iii PAOSVCPAOSVCPAOSVC DDD (4.101)
*
_/_/_/ iHXPAOfueliHXPAOfueliHXPAOfuel QQQ (4.102)
ibleedbleedbleed mmm
ii*
(4.103)
ihlbleedhlbleedhlbleed TTTii
*,_,_,_
(4.104)
ihlbleedhlbleedhlbleed PPPii
*,_,_,_
(4.105)
where the superscript "" indicates the optimum coupling function value from the previous ILGO
iteration. Similarly to the ECS, equations (4.96) to (4.105) are not comprehensive but rather
represent 96 total equations for the VC/PAOS, one for each unique shadow price. The FLS
system-level, unit-based synthesis/design optimization problem definition is next.
4.3.6 FLS System-Level, Unit-Based Synthesis/Design Optimization Problem
Definition
The FLS system-level, unit-based optimization problem is defined as follows:
FLS System-Level, Unit-Based Exergy Destruction Optimization Problem
Minimize
n
i iPAOSVCOLSCHSFLSiPAOSVCOLSCHSFLSQiisfc
iFLSTOiFLSTOPiFLSiFLSD
iousiiousiui
iitFLSFLSWOLSdes
CHSdesPAOSVCdesECSdes
PSdesAAFSdesFLSdesFLSdes
Qsfc
PD
utWEx
ExExEx
ExExExEx
1 /,,/,,
varvar
5
1
*
**/
*
**'
(4.106)
w.r.t. FLSFLS YX
,
subject to
0
FLSh (4.107)
124
0
FLSg (4.108)
where the vector of equality constraints, FLSh
, represents the thermodynamic and geometric
model of the FLS and the vector of inequality constraints, FLSg
, the physical limits imposed on
the subsystem. The superscript “*” indicates the optimum value from the previous ILGO
iteration of a given subsystem objective function.
Problem (4.106) represents the system-level minimization resulting from varying the FLS
decision variables only. The local or unit-based minimization objective function is denoted by
.FLSdesEx The expression for the unit-based, system-level objective function,
FLSdesEx' , includes
products for paired shadow prices and associated coupling functions. The first pair accounts for
the effects of variations in the FLS weight, the second pair for the mission segments with
variable time, the third pair for various other interactions, while the pairs in the summation
reflect the effect of variations in the FLS coupling functions. The FLS shadow prices are defined
as
FLS
AAFSdes
FLSW W
Ex
*
(4.109)
5
1
*
i i
FLSdes
it t
Ex
(4.110)
iousij
jdes
iousiju u
Ex
var
*
var
(4.111)
And the next four shadow prices, which represent 120 independent shadow prices, are defined as
iFLS
AAFSdes
iFLSD D
Ex
*
(4.112)
iFLSTO
PSdes
iFLSTOP P
Ex
*
(4.113)
i
PSdes
isfc sfc
Ex
*
(4.114)
125
iPAOSVCOLSCHSFLS
PAOSVCOLSCHSdes
iPAOSVCOLSCHSFLSQ Q
Ex
/,,
*/,,
/,,
(4.115)
where equations (4.110), (4.111), (4.112), (4.113), (4.114), and (4.115) represent a total of 5, 9,
20, 20, 20, and 60 unique shadow prices, respectively. The FLS has a total of 94 input shadow
prices and 41 output shadow prices for a total of 135 shadow prices.
Equations (4.109) through (4.115) represent the effect of the marginal change in the
optimum value of the coupling function on the system-level, unit-based objective function for the
FLS due to changes in the coupling functions. Problem (4.106) has seven additional terms,
,FLSW ,it ,variousiu
,
iFLSD ,iFLSTOP ,isfc and iPAOSVCOLSCHSFLSQ /,, that represent
variations in the coupling functions listed above and are defined as
*FLSFLSFLS WWW (4.116)
5
1
0*
iiii ttt (4.117)
*varvarvar iousijiousijiousij uuu (4.118)
*
FLSFLSFLS DDDii (4.119)
*
iFLSiFLSiFLS TOTOTO PPP (4.120)
*iii sfcsfcsfc (4.121)
iPAOSVCOLSCHSFLSiPAOSVCOLSCHSFLSiPAOSVCOLSCHSFLS QQQ *
/,,/,,/,, (4.122)
where the superscript "" indicates the optimum coupling function value from the previous ILGO
iteration. Similarly to the ECS, equations (4.116) to (4.122) are not comprehensive but rather
represent 135 total equations for the FLS. The OLS system-level, unit-based synthesis/design
optimization problem definition is next.
4.3.7 OLS System-Level, Unit-Based Synthesis/Design Optimization Problem
Definition
The OLS system-level, unit-based optimization problem is defined as follows:
OLS System-Level, Unit-Based Exergy Destruction Optimization Problem
126
Minimize
n
iiOLSFLS
iOLSFLSQiuninstiuninstT
iiousiiousiuiitOLSOLSWFLSdes
PSdesAAFSdesOLSdesOLSdes
QT
utWEx
ExExExEx
1
5
1varvar
*
**'
(4.123)
w.r.t. OLSOLS YX
,
subject to
0
OLSh (4.124)
0
OLSg (4.125)
where the vector of equality constraints, OLSh
, represents the thermodynamic and geometric
model of the OLS and the vector of inequality constraints, OLSg
, the physical limits imposed on
the subsystem. The superscript “*” indicates the optimum value from the previous ILGO
iteration of a given subsystem objective function.
Problem (4.123) represents the system-level minimization resulting from varying the
OLS decision variables only. The local or unit-based minimization objective function is denoted
by OLSdesEx while the expression for the system-level, unit-based objective function,
OLSdesEx' ,
includes products for paired shadow prices and associated coupling functions. The first pair
accounts for the effects of changes in the weight of the OLS, the second pair represents the time
coupling with the AFS-A. The pairs within the summation from i to n account for the effects of
changes in the uninstalled thrust from the PS and the heat interaction with the FLS via the fuel /
oil heat exchanger. The OLS shadow prices are defined as
OLS
AAFSdes
OLSW W
Ex
*
(4.126)
5
1
*/
i i
PAOSVCdes
it t
Ex
(4.127)
127
iousij
jdes
iousiju u
Ex
var
*
var
(4.128)
and the next two shadow prices, which represent 40 independent shadow prices, are defined as
iuninst
AAFSdes
iuninstT T
Ex
*
(4.129)
iOLSFLS
FLSdes
iOLSFLSQ Q
Ex
*
(4.130)
where equations (4.128), (4.129) and (4.130) represent 5, 20, and 20 unique shadow prices,
respectively. The OLS has a total of 30 input shadow prices and 21 output shadow prices for a
total of 51 shadow prices.
Equations (4.126) through (4.130) represent the effect of the marginal change in the
optimum value of the coupling function on the system-level, unit-based objective function for the
FLS due to changes in the coupling functions. Problem (4.123) has five additional terms,
,OLSW ,it ,variousiu
,iuninstT and iOLSFLSQ that represent the effect that the variation in
the OLS decision variables has on the coupling functions listed above and are defined as
*OLSOLSOLS WWW (4.131)
5
1
0*
iiii ttt (4.132)
*varvarvar iousijiousijiousij uuu (4.133)
*
uninstuninstuninst TTTii (4.134)
iOLSFLSOLSFLSOLSFLS QQQ
ii*
(4.135)
where the superscript "" indicates the optimum coupling function value from the previous ILGO
iteration. Again, equations (4.131) to (4.135) are not comprehensive but rather represent 51 total
equations for the OLS. The CHS system-level, unit-based synthesis/design optimization
problem definition is next.
128
4.3.8 CHS System-Level, Unit-Based Synthesis/Design Optimization Problem
Definition
The CHS system-level, unit-based optimization problem is defined as follows:
CHS System-Level, Unit-Based Exergy Destruction Optimization Problem
Minimize
n
iiCHSFLS
iCHSFLSQiCHSTOiCHSTOP
iousiiousiuCHSCHSW
FLSdesPSdesAAFSdesCHSdesCHSdes
QP
uW
ExExExExEx
1
varvar
***'
(4.136)
subject to
0
CHSh (4.137)
0
CHSg (4.138)
where the vector of equality constraints, CHSh
, represents the thermodynamic and geometric
model of the CHS and the vector of inequality constraints, CHSg
, the physical limits imposed on
the subsystem. The superscript “*” indicates the optimum value from the previous ILGO
iteration of a given subsystem objective function. Note that the CHS has no design decision
variables and/or operational decision variables, thus there are no vectors of design decision or
operational decision variables associated with problem (4.136) and only the shadow prices and
coupling function pairs vary during the local optimization of the CHS.
Problem (4.136) represents the system-level minimization resulting from varying the
CHS interactions only as the CHS has no decision variables. The local or unit-based
minimization objective function is denoted by CHSdesEx while the expression for the system-
level, unit-based objective function, CHSdesEx' , includes products for paired shadow prices and
associated coupling functions. The first pair accounts for the effects of changes in the weight of
the CHS, the second pair accounts for the effects of changes in various coupling functions in the
CHS, and the pairs within the summation from i to n account for the effects of changes in the
129
uninstalled thrust from the PS and the heat interaction with the FLS via the fuel / hydraulic oil
heat exchanger. Notice that problem (4.136) has no shadow price and coupling function pair
representing the AFS-A mission segment times since the CHS is calculated within the OLS unit-
level optimization and the time variables are directly assigned. The CHS shadow prices are
defined as
CHS
AAFSdes
CHSW W
Ex
*
(4.139)
iousij
jdes
iousiju u
Ex
var
*
var
(4.140)
iCHSTO
PSdes
iCHSTOP P
Ex
*
(4.141)
iCHSFLS
FLSdes
iCHSFLSQ Q
Ex
*
(4.142)
where equations (4.140), (4.141), and (4.142) each represent 6, 20, and 20 unique shadow prices,
respectively. The CHS has a total of 6 input shadow prices and 41 output shadow prices for a
total of 47 shadow prices.
Equations (4.139) through (4.142) represent the effect of the marginal change in the
optimum value of the coupling function on the system-level, unit-based objective function for the
CHS due to changes in the coupling functions. Problem (4.136) has four additional terms,
,CHSW ,var iousiu
,iCHSTOP and iCHSFLSQ that represent the effect that the variation in the
CHS degrees of freedom has on the coupling functions listed above and are defined as
*CHSCHSCHS WWW (4.143)
*varvarvar iousijiousijiousij uuu (4.144)
*
iCHSiCHSiCHS TOTOTO PPP (4.145)
iCHSFLSCHSFLSCHSFLS QQQ
ii*
(4.146)
130
where the superscript "" indicates the optimum coupling function value from the previous ILGO
iteration. Again, equations (4.143) to (4.146) are not comprehensive but rather represent 47 total
equations for the CHS. The ES system-level, unit-based synthesis/design optimization problem
definition is next.
4.3.9 ES System-Level, Unit-Based Synthesis/Design Optimization Problem
Definition
The ES system-level, unit-based optimization problem is defined as follows:
ES System-Level, Unit-Based Exergy Destruction Optimization Problem
Minimize
n
iiESTO
iESTOP
EmptyEmptyWi
iitESESW
PSdesAAFSdesESdesESdes
P
WtW
ExExExEx
1
5
1
**'
(4.147)
subject to
0
ESh (4.148)
0
ESg (4.149)
where the vector of equality constraints, ESh
, represents the thermodynamic and geometric
model of the ES and the vector of inequality constraints, ESg
, the physical limits imposed on the
subsystem. The superscript “*” indicates the optimum value from the previous ILGO iteration of
a given subsystem objective function.
Problem (4.147) represents the system-level minimization resulting from varying the ES
decision variables only. The local or unit-based minimization objective function is denoted by
ESdesEx while the expression for the system-level, unit-based objective function, ESdesEx' ,
includes products for paired shadow prices and associated coupling functions. The first pair
accounts for the effects of changes in the weight of the ES, the second pair is the AFS-A mission
131
segment time, the third pair represents the AAF empty weight, and the pair within the summation
accounts for the effects of changes in the shaft power extracted from the PS. Note that the ES
has no design decision variables and/or operational decision variables, thus there are no vectors
of design decision or operational decision variables associated with problem (4.147) and only the
shadow prices and coupling function pairs vary during the local optimization of the ES. The ES
shadow prices are defined as
ES
AAFSdes
ESW W
Ex
*
(4.150)
5
1
*
i i
ESdes
it t
Ex
(4.151)
Empty
AAFSdes
EmptyW W
Ex
*
(4.152)
iESTO
PSdes
iESTOP P
Ex
*
(4.153)
where equations (4.153) represents 20 unique shadow prices. The ES has a total of 6 input
shadow prices and 21 output shadow prices for a total of 27 shadow prices.
Equations (4.150) to (4.153) represent the effect of the marginal change in the optimum
value of the coupling function on the unit-based, system-level objective function for the ES.
Problem (4.147) has four additional terms, ,ESW ,it ,EmptyW and iESTOP that represent the
effect that the variation in the ES degrees of freedom has on the coupling functions listed above
and are defined as
*ESESES WWW (4.154)
5
1
0*
iiii ttt (4.155)
*EmptyEmptyEmpty WWW (4.156)
n
iiESTOiESTOiESTO PPP
1
*
(4.157)
132
where the superscript "" indicates the optimum coupling function value from the previous ILGO
iteration. Equations (4.154) through (4.157) are comprehensive for the ES in this thesis work
and represent 27 total equations for the ES. The FCS system-level, unit-based synthesis/design
optimization problem definition is next.
4.3.10 FCS System-Level, Unit-Based Synthesis/Design Optimization Problem
Definition
The FCS system-level, unit-based optimization problem is defined as follows:
FCS System-Level, Unit-Based Exergy Destruction Optimization Problem
Minimize
sssWsA
WALbrrCttC
FCSFCSWAAFSdesFCSdesFCSdes
WA
WALbCC
WExExEx
_
_
'
_
_
*
(4.158)
subject to
0
ESh (4.159)
0
ESg (4.160)
where the vector of equality constraints, FCSh
, represents the thermodynamic and geometric
model of the ES and the vector of inequality constraints, FCSg
, the physical limits imposed on
the subsystem. The superscript “*” indicates the optimum value from the previous ILGO
iteration of a given subsystem objective function. Note that the FCS has no design decision
variables and/or operational decision variables, thus there are no vectors of design decision or
operational decision variables associated with problem (4.155) and only the shadow prices,
representing the subsystem interactions, vary during the local optimization of the FCS.
Thus, problem (4.158) represents the system-level minimization resulting from variation
in the FCS interactions only. The local or unit-based minimization objective function is denoted
by FCSdesEx while the expression for the system-level, unit-based objective function,
FCSdesEx' ,
133
includes products for paired shadow prices and associated coupling functions. The first pair
accounts for the effects of changes in the weight of the FCS. The subsequent paired shadow
prices and coupling functions account for the effects of changes in the wing tip and root chord
lengths, wing span, wing length, aileron chord / wing chord, and aileron span / wing span. The
FCS shadow prices are defined as
FCS
AAFSdes
FCSW W
Ex
*
(4.161)
t
AAFSdes
tC C
Ex
*
(4.162)
r
AAFSdes
rC C
Ex
*
(4.163)
b
ExAAFSdes
b
*
(4.164)
L
ExAAFSdes
L
*
(4.165)
WA
ExAAFSdes
WA _
*
_
(4.166)
ss
AAFSdes
sWsA WA
Ex
_
*
_
(4.167)
The ES has a total of 6 input shadow prices and 1 output shadow prices for a total of 7 shadow
prices. The ES is calculated within the OLS, thus no coupling functions related to mission
decision variables are required.
Equations (4.161) to (4.167) represent the effect of the marginal change in the optimum
value of the coupling function on the system-level, unit-based objective function for the FCS.
Problem (4.158) has seven additional terms, ,FCSW ,tC ,rC ,b ,L ,_WA and ss WA _
that represent the effect that the variation in the FCS degrees of freedom has on the coupling
functions listed above and are defined as
*FCSFCSFCS WWW (4.168)
134
*ttt CCC (4.169)
*rrr CCC (4.170)
*bbb (4.171)
*LLL (4.172)
*___ WAWAWA (4.173)
*___ ssssss WAWAWA (4.174)
where the superscript "" indicates the optimum coupling function value from the previous ILGO
iteration. Again, the FCS has no design decision or operational decision variables that
participate in the optimization. This concludes the discussion of the system-level, unit-based
synthesis/design optimization problem definitions for the AAF model developed in this thesis
work.
4.4 Optimization Decision Variables and Variable Constraints
This section details the optimization synthesis / design, operational, and mission decision
variables and constraints for each of the subsystems included in this thesis work. The first
subsystem to be detailed is the AFS-A. The fixed-wing AFS-A decision variables and variable
constraints are given in Table 4.2
Table 4.2 AFS-A fixed-wing design and operational decision variables and inequality constraints.
Component Design Decision Variables Constraints
Fixed-wing AFS-A
TOW Gross takeoff weight (lb) 000,60000,10 TOW
wingL Wing length (ft) 6530 wingL
LE Leading edge sweep angle 6015 LE
AR Aspect ratio 102 AR Taper ratio 10
ct
Thickness-to-chord ratio 15.006.0 ct
tailAR Tail aspect ratio 5.65.3 tailAR
tailct
Tail thickness-to-chord ratio 25.006.0 tailc
t
135
The fixed-wing AFS-A has only a single configuration for the mission and, therefore, has
a total of 8 design decision variables. Note that the gross takeoff weight, TOW , participates in the
optimization as a means to establish the aircraft takeoff weight and fuel weight.
The morphing-wing AFS-A design and operational decision variables and variable
constraints are given in Table 4.3. Note that the design decision variables are listed first and
only take one value for the entire mission. The operational decision variables each have 20 (one
for each mission segment) unique values, thus, the morphing-wing AFS-A has 20 unique wing
configurations which brings the total number of optimization decision variables for the AFS-A to
103.
Table 4.3 AFS-A morphing-wing design and operational decision variables and inequality constraints.
Component Design Decision Variables Constraints
Morphing-wing AFS-A
TOW Gross takeoff weight (lb) 000,60000,10 TOW
tailAR Tail aspect ratio 5.65.3 tailAR
tailct Tail thickness-to-chord ratio 25.006.0
tailct
Operational Decision Variables Constraints
L Wing length (ft) 6530 L
LE Leading edge sweep angle 6015
AR Aspect ratio 102 AR Taper ratio 10
ct
Thickness-to-chord ratio 15.006.0 ct
The AFS-A also has 6 mission decision variables that participate in the optimization. This set of
variables is given in Table 4.4. The variable names correspond to the mission segments detailed
Table 4.4 AFS-A mission decision variables and inequality constraints.
Component Mission Decision Variables Constraints
AFS-A Mission
1BCM Best cruise Mach #1 0.116.0 BCM 2BCM Best cruise Mach #2 0.126.0 BCM 3BCM Best cruise Mach #3 0.136.0 BCM 1BCA Best cruise altitude 1 (ft) 000,551000,30 BCA
2BCA Best cruise altitude 2 (ft) 000,552000,30 BCA
BCLM Best climb Mach 0.15.0 BCLM
in Table 3.1. The mission variables could have been included in any of the other subsystems;
but since the mission decision variables affect the AFS-A performance and weight fractions
significantly, it makes sense to have them participate in the AFS-A optimization. Note that each
136
of these mission decision variables affects the time required to fly the mission which is an
integration concern for all the subsystems. Also note that the takeoff weight, TOW , is a decision
variable for both the fixed-wing and morphing-wing AFS-A optimizations.
The next set of subsystem decision variables and inequality constraints to be discussed
are those for the PS. The model of the PS has a reference or design engine so there are decision
variables associated with the reference conditions of the engine as well. The decision variables
and corresponding variable constraints for the PS are given in Table 4.5. Note that the mission
defined in Table 3.1 requires military or maximum power for some of the mission segments
which removes the decision variables for the corresponding mission segment. Thus, the PS has a
total of 11 design decision variables and 27 operational decision variables.
Table 4.5 PS design and operational decision variables and inequality constraints.
Component Design Decision Variables Constraints
Propulsion Subsystem
0M Design Mach number 7.18.0 0 M
Alt Design altitude 000,50000,30 Alt
0m Design total mass flow rate (lb/s) 4006.0 0 m
c Design total compressor pressure ratio 0.300.6 c
'c Design fan pressure ratio 0.60.2 ' c
Design bypass ratio 85.025.0
4tT Design burner temperature (R) 32001000 4 tT
7tT Design afterburner temperature (R) 36001000 7 tT
5M Design mixer Mach number for core 6.04.0 5 M
TOP Design power takeoff (kW) 0.4000.0 TOP
Design bleed air ratio 03.000.0 Operational Decision Variables Constraints
4tT Burner temperature (R) 32001000 4 tT
7tT Afterburner temperature (R) 36001000 7 tT
Next comes the ECS. The decision variables and the corresponding constraints are
shown in Table 4.6. The ECS has a total of 23 synthesis / design decision variables. Note,
however, that there are two possible configurations for the regenerative heat exchanger. This
requires that another independent set of regenerative heat exchanger variables must be used as
well bringing the total to 28 synthesis / design decision variables.
Table 4.6 ECS optimization synthesis / design and operational decision variables and inequality constraints.
Component Design Decision Variables Constraints
137
Primary Heat Exchanger
bFin Bleed-air fin type 2011 bFin
rFin Ram-air fin type 2011 rFin
bL Bleed-air side length, m 9.05.0 bL
rL Ram-air side length, m 9.006.0 rL
nL Non-flow length, m 9.05.0 nL
Secondary Heat Exchanger
bFin Bleed-air fin type 2011 bFin
rFin Ram-air fin type 2011 rFin
bL Bleed-air side length, m 9.05.0 bL
rL Ram-air side length, m 9.006.0 rL
nL Non-flow length, m 9.05.0 nL
Regenerative Heat Exchanger
hFin Hot-side (bleed air) fin type 2011 hFin
cFin Cold-side (from the water separator) fin type
2011 cFin
hL Hot-side length, m 5.03.0 bL
cL Cold-side length, m 3.015.0 cL
nL Non-flow length, m 5.03.0 nL
Bleed Air / PAO Heat Exchanger
bFin Bleed-air fin type 2011 bFin
rFin PAO fin type 2011 rFin
bL Bleed-air-side length, m 9.05.0 bL
rL PAO-side length, m 9.006.0 rL
nL Non-flow length, m 9.05.0 nL
Compressor desPR Design compressor pressure ratio
0.38.1 desPR
Inlet Duct 1 1A Inlet duct 1 area, cm2 2208 1 A
Inlet Duct 2 2A Inlet duct 2 area, m2 2208 2 A
Component Operational Decision Variables Constraints
Pressure Regulating Valve vvPR Pressure setting 0.60.1 vvPR
Bleed Port Bleed Bleed port selection (high or low pressure bleed port)
)1,0(Bleed
Air Cycle Machine hotv Hot-air bypass 0.10.0 hotv
Splitter bypassv Compressor bypass 0.10.0 bypassv
Regenerative Heat Exchanger regm Regenerative heat exchanger mass flow rate, kg/s
2.00.0 regm
138
Because of certain difficulties encountered during the optimization process in this thesis
work, the ECS flies only 4 of the 21 mission segments to establish the ECS subsystem synthesis /
design and subsystem interactions. The mission segments flown are warmup and takeoff
(segment 1), subsonic climb (segment 4), supersonic turn (segment 11), and subsonic cruise
(segment 19). This brings the total number of operational variables to 20, though if the entire
mission were used, 100 operational decision variables would be required for the ECS. (In the
work of Rancruel (2003), the subsonic cruise mission segment was used as the design segment.)
The next subsystem to be discussed is the FLS. The synthesis / design decision variables
and constraint limits for the FLS are given in Table 4.7. The FLS optimization only has 8
Table 4.7 FLS optimization decision variables and inequality constraints.
Component Design Decision Variables Constraints
Ram Air / Fuel Heat Exchanger
rFin Ram air fin type 2011 rFin
fFin Fuel side fin type 2011 fFin
rL Ram air side length, m 9.005.0 rL
fL Fuel side length, m 9.005.0 fL
nL Non-flow length, m 9.005.0 nL
Fuel/Oil Heat Exchanger
cL Cold-side length (m) 90050 .L. c
hL Hot-side length (m) 9010 .L. n
nL Non-flow length (m) 9010 .L. n
synthesis / design decision variables since the heat exchangers are sized in other subsystems.
Note that the fuel added to the PS-required fuel mass flow rate, addfuelm _ , is an operational
variable. However, the fuel added to the flow is iterated within simulation and does not
participate as a decision variable in the optimization (see section 3.6.1). Thus, the FLS has no
operational decision variables.
The synthesis / design and operational decision variables and inequality constraints for
the VC/PAOS are shown in Table 4.8. The VC/PAOS flies only a single mission segment to
establish its synthesis / design conditions and performance. The mission segment chosen is the
most constrained mission segment of the four flown by the ECS which is the segment with the
highest cooling load required from the cold PAO loop. If the entire mission were being
optimized, a total of 60 operational decision variables would participate in the optimization;
139
however, for this thesis work, the VC/PAOS has 20 synthesis / design decision variables and
only 3 operational decision variables.
Table 4.8 VC/PAOS optimization synthesis / design and operational decision variables and inequality constraints.
Component Synthesis / Design Decision Variables Constraints
Condenser
vFin Vapor-side fin type 2011 vFin
pFin Liquid-side fin type 2011 pFin
vL Vapor-side length, m 9.05.0 vL
pL Liquid-side length, m 9.006.0 pL
nL Non-flow length, m 9.05.0 nL
Evaporator
vFin Vapor-side fin type 2011 vFin
pFin Liquid-side fin type 2011 pFin
vL Vapor-side length, m 9.05.0 vL
pL Liquid-side length, m 9.006.0 pL
nL Non-flow length, m 9.05.0 nL
Ram Air / PAO Heat Exchanger
hFin Hot-side (bleed air) fin type 2011 hFin
rFin Ram-air fin type 2011 rFin
hL Hot-side length, m 5.03.0 hL
rL Ram-air side length, m 9.005.0 rL
nL Non-flow length, m 5.03.0 nL
Fuel / PAO Heat Exchanger
pFin Bleed-air fin type 2011 pFin
fFin Fuel-side fin type 2011 fFin
pL Bleed-air side length, m 9.05.0 pL
fL Fuel-side length, m 9.005.0 fL
nL Non-flow length, m 9.05.0 nL
Component Operational Decision Variables Constraints
Mass Flow Rates
vaporm Vapor mass flow rate, kg/s 2.22.0 vaporm
hotpaom _ Hot PAO loop mass flow rate, kg/s 5.32.0 _ hotpaom
coldpaom _ Cold PAO loop mass flow rate, kg/s 5.32.0 _ coldpaom
The next subsystem is the OLS. Its operational decision variables and variable
constraints are given in Table 4.9. The OLS heat exchangers are sized in the FLS. The OLS has
a total of 20 operational decision variables.
Table 4.9 OLS optimization operational decision variables and inequality constraints.
Component Operational Decision Variables Constraints
OLS operational variables oilm Mass flow rate of oil, kg/s 0.40 oilm
140
Finally, the remaining subsystems, the ES, CHS, and FCS are passive subsystems in
terms of the optimization as they have no local design or operational decision variables
associated with them in this thesis work. They do, however, participate via system-level degrees
of freedom, i.e. via coupling functions. Thus, they must be considered in the subsystem
integration as their operation has a direct effect on the overall system-level objective function
and on system performance.
4.5 Solution Approach
Both physical and time decomposition are applied in this thesis work to the AAF aircraft
which is modeled and optimized based on 9 subsystems separated by physical or thermodynamic
boundaries and flown over a mission separated into 21 time segments. Even though the AFS-A,
PS, ECS, FLS, VC/PAOS, CHS, OLS, ES, and FCS are modeled and optimized separately,
coupling functions update the subsystem interactions between ILGO iterations. Each of the
subsystem optimizations, thus, represent a unit (or local) component of the overall system
optimization problem.
Figure 4.6 shows the notional flow of the ILGO approach for the AAF aircraft system.
Starting from the individual subsystem optimizations in Figure 4.6, the ILGO process is as
follows:
Figure 4.6 Diagram of optimization problem solution approach using ILGO.
Simulate Dependent Subsystems
Assign shadow prices to the subsystems and pass these and their updated coupling functions to the subsystems
Individual Subsystem Optimizations
AFS-A
PS
ECS
FLS
VCPAOS
OLS
ILGO Iteration
Calculate the system-
level objective function
Calculate the new shadow
prices
CHS ES FCS
Update the coupling functions
141
1. The initial subsystem optimizations start with arbitrary values for the coupling functions
and shadow prices, but within specified limits.
2. After the AFS-A, PS, ECS, FLS, OLS, and VC/PAOS are optimized, the CHS, ES, and
FCS are simulated based on outputs from the optimized subsystems.
3. The overall system-level objective value is calculated.
4. New shadow prices are calculated based on the initial result.
5. The ILGO iteration count is incremented.
6. The coupling function values are updated based on the individual subsystem optimization
results.
7. Finally, new shadow prices are assigned to the appropriate unit-level subproblems and
these and their associated updated coupling function values are passed to the subsystems
so that the next set of individual subsystem optimizations can proceed.
The explanation of ILGO provided here is merely a top-level overview. The reader is once again
directed to Muñoz and von Spakovsky (2000a,b,c,d; 2001a,b), Rancruel (2002, 2005), and
Rancruel and von Spakovsky (2005, 2006) for a detailed explanation of ILGO and the other
facets of the ILGO approach mentioned here. The software package used in this thesis work is
briefly described in the following subsection.
4.6 iScript™ Scripting Language and Optimization
The software used for the optimization of this thesis work is called iSCRIPT™ which is
in development by TTC Technologies, Inc. iSCRIPT™ was developed under an Air Force Small
Business Innovative Research (SBIR) Phase II project. The reasons for using iSCRIPT™
include the following attributes that make it attractive for this thesis work:
Quick learning curve for the programming syntax;
Built-in optimization tools;
Component-based programming structure;
Automated ILGO; and
Automated parallelization for the optimization.
142
Another large contributing factor to the decision to use iSCRIPT™ is the fact that the
subsystems were already mostly written (or being written) in iSCRIPT™ as a demonstration to
the Air Force and industry. The optimization in iSCRIPT™ is performed on two levels. The
unit or subsystem level is optimized using a genetic algorithm (GA) developed by researchers at
the Laboratoir d’energetique industrielle at the Ecole Polytechnique Federale de Lansanne in
Lausanne, Switzerland (Leyland, 2002). The parameter set in the original implementation was
condensed to five parameters by TTC including the population size, initial evaluations,
convergence criteria, mutation frequency, and maximum number of generations to simplify the
user interaction with the method. To arrive at a system-level optimum, the ILGO decomposition
strategy is implemented in iSCRIPT™.
GAs are based on Darwin’s theory of natural selection or “survival of the fittest.” The
initial population or set of optimization decision variable values is generated by stepping through
the range of values possible for each optimization variable. After the initial population is
generated, the algorithm reorders the optimization variables based on the values of the objective
functions they yield, i.e., the solutions are ordered from the best to the worst solution based on
the value of the objective. The GA then selects the better sets of optimization decision variable
values, called “parents”, mutates them slightly, then repopulates the bottom half of the
population with the mutated optimization decision variable values or “offspring” of the parents.
The new population members or sets of decision variable values are then evaluated in the model
and the entire population is again reordered from the best to the worst solution and the selection-
mutation-repopulation process continues until the convergence criteria is met or until the
maximum number of generations specified is reached.
The implementation of ILGO in iSCRIPT™ uses a gradient-based method to converget
the decomposed system to a system-level optimum. In this approach, gradients are computed
using the shadow prices of the coupling functions between the subsystems. The gradients are
used along with the coupling functions to search the system-level ORS for the overall system
optimum. The shadow prices represent the effect of the unit-level coupling functions on the
system-level objective function. Such an approach allows multiple subsystem optimizations to
take place at the same time since each subsystem is effectively isolated from the others while at
the same time assuring a system-level optimum by periodically updating the subsystem
interactions between ILGO steps via coupling functions.
143
Chapter 5
Results and Discussion
This chapter discusses the results obtained during this thesis work. The first section starts
with a set of results from the first phase of this thesis work that is an extension of the work of
Butt (2005) from a paper presented at the 2007 AIAA Thermophysics Conference held in Miami,
FL (Smith, et. al, 2007). The next section presents results and analysis of a partial optimization
of the 9 subsystem AAF that is written in iSCRIPT™ along with predicted optimal results and
analysis. The final section is a parametric study of the morphing AFS-A behavior.
5.1 Two-Subsystem Optimization Results
The work of Smith, et al. (2007) studies the benefits of using morphing wing technology
in an AAF. Two subsystems are modeled: the AFS-A, both morphing- and fixed-wing models,
and a turbofan PS as described in Chapter 3. The AFS-A models were developed and used by
Butt (2005). The morphing-wing AFS-A is physically similar to that described in Chapter 3 with
a few differences which are as follows:
root- and tip-chord lengths are operational decision variables while the AFS-A
described in Chapter 3 uses the taper ratio as the operational decision variable to
establish the root- and tip-chord lengths;
morphing AFS-A wing-weights are established based on setting the wing sweep to
zero. This makes the wing-weight estimate lower than if the wing had some amount
of sweep, thus, the metric to compare fixed-wing and morphing-wing results in
Smith et al. (2007) is gross takeoff weight; and
wing-weight penalties and fuel-weight penalties are established in the manner
described in Section 3.2.5.; however, the excess fuel that is carried to power the
actuators that morph the wings is not expelled in Smith et al. (2007) in order to
match what was done in Butt (2005).
The mission flown in Smith et al. (2007) is similar to that described in Chapter 3 but with
some differences. The AAF model does not fly the entire mission but rather flies a subset of the
144
mission and no DOF are associated with the mission (i.e. the mission segments are fixed and do
not participate in the optimization) as was done in the work of Butt (2005).
The optimization is handled in MATLAB™ using a genetic algorithm developed by
Leyland (2002). The fixed-wing AFS-A has 4 synthesis/design decision variables (wing length,
wing sweep, and root- and tip- chord lengths) while the morphing-wing AFS-A has the same
number of synthesis / design decision variables and a total of 72 operational decision variables
(18 sets of wing length, wing sweep, and root- and tip- chord lengths). The turbofan PS has 9
synthesis/design decision variables (Mach number, mass flow rate, altitude, compressor pressure
ratio, fan pressure ratio, bypass ratio, main burner temperature, afterburner temperature, and
mixer Mach number) and 28 operational decision variables (nineteen main burner temperatures
and nine afterburner temperature settings). The optimization objective function used is that of
minimizing the fuel burned over the mission.
Figure 5.1 shows the sensitivity of the total fuel consumed over the mission with respect
to the wing- weight and fuel- weight penalties for the morphing-wing AAF. Note that the fixed-
wing result is shown in Figure 5.1 with the red horizontal line. The shaded area below the red
line indicates the fuel savings region (i.e. any combination of wing-weight and fuel-weight
Figure 5.1 Sensitivity analysis of morphing-wing effectiveness for different wing- and fuel-weight penalties (Smith et al., 2007).
145
penalties within the shaded area outperform the fixed-wing result in terms of fuel burned over
the mission). As mentioned previously, the metric to compare the fixed-wing and the morphing-
wing result is the gross takeoff weights. A sample of the optimum gross takeoff weights for the
6x wing-weight penalty and the fixed-wing gross takeoff weight optimum for the AAF is given
in Table 5.1.
Table 5.1 Comparison of the optimum morphing-wing gross takeoff weights with a 6x wing-weight penalty and the optimum fixed-wing gross takeoff weight (Smith et al. 2007).
Morphing-Wing Fuel Penalty and Gross Takeoff Weights Fixed- Wing
Fuel Penalty 3% 10% 25% 50% 75% 100% N/A
(lbm) 19,473 19,917 20,844 22,330 23,757 25,138 14,959
The morphing-wing gross takeoff weights shown in Table 5.1 are only for the 6x wing-
weight penalty while the final column in Table 5.1 gives the optimum gross takeoff weight for
the fixed-wing AAF. The reader should notice that the two highest fuel penalties, 75% and
100%, for the 6x wing-weight penalty burn more fuel than the fixed-wing result shown in Figure
5.1. However, these two aircraft have gross takeoff weights 59% and 68% heavier, respectively,
than the fixed-wing gross takeoff weight of 14,959 lbm. The interesting result is that a morphing-
wing aircraft that has a gross takeoff weight 49.3% higher (for the 6x wing-weight penalty and
50% fuel-weight penalty) than the fixed-wing aircraft burns approximately 100 lb less fuel than
the fixed-wing AAF over the entire mission.
Table 5.2 shows the comparison between the fixed-wing AAF and the morphing-wing
AAF configurations for two mission segments, namely, the subsonic cruise segment and the
supersonic penetration segment. The flight conditions of these two mission segments are at
Mach 0.9 and a 42,000 ft altitude and at Mach 1.5 and a 30,000 ft altitude, respectively. The first
four rows in Table 5.2 are AFS-A design/operational parameters while the final four rows in
Table 5.2 are engine operating conditions and design parameters.
Table 5.2 Optimal fixed- versus morphing-wing AAF configuration and performance parameters for the subsonic cruise and the supersonic penetration mission segments (Smith et al. 2007).
Morphing-Wing Fixed-Wing
Mission Cruise Supersonic Penetration
Cruise Supersonic Penetration
Wing length (ft) 35.50 29.09 41.4301
146
Morphing-Wing Fixed-Wing
Mission Cruise Supersonic Penetration
Cruise Supersonic Penetration
Leading edge wing sweep (deg) 13.16 43.63 41.7168
Root-chord length (ft) 4.04 4.00 5.0138
Tip-chord length (ft) 1.53 1.68 2.6809
Burner temperature (°R) 2365.7 3198.7 2956.5 2938.8
Fuel consumption (lbm) 76.8 712.2 210.1 662.2
Design air mass flow rate (lbm/s) 154.3 158.9
Design Mach 1.35 1.55
Note that the fixed-wing leading edge sweep angle is nearly the same as the morphing-
wing for the supersonic penetration mission segment. This indicates that the fixed-wing design
tended toward the more stressing aerodynamic flight condition: supersonic flight. The
morphing-wing AAF, however, with the ability to change its wing geometry, increased the
leading edge sweep angle and decreased the wing length for the supersonic mission segment.
Also notice that the morphing-wing AAF moved to a smaller sweep angle to create a high aspect
ratio wing during the cruise segment with the end result of burning only about 1/3 of the fuel of
the fixed-wing AAF in cruise.
The explanation for the morphing-wing AAF burning slightly more fuel than the fixed-
wing AAF for the supersonic mission segment is simple: the two AAFs have nearly the same
aerodynamic properties during supersonic flight from their wing geometry; however, the
morphing-wing AAF is heavier and, thus, requires more thrust for sustained supersonic flight.
The increase in thrust requirements for the morphing-wing AAF can be seen by the higher burner
temperature (i.e., higher throttle setting).
Note also that the design air mass flow rates are nearly the same for the morphing- and
fixed-wing AAF with the morphing-wing design air mass flow rate being slightly lower than that
for the fixed-wing engine. The telling result here is the design Mach number. Notice in Table
5.2 that the morphing-wing AAF has a much lower design Mach number than the fixed-wing
AAF (1.35 versus 1.55). This indicates that the optimum engine for the morphing vehicle tends
towards subsonic flight as that is where the morphing-wing aircraft shows the greatest benefit
over the fixed-wing aircraft. The physical characteristics and operating conditions shown in
147
Table 5.2 clearly highlight the potential benefits of morphing-wing technology in fighter aircraft
for the AAF consisting of a PS and AFS-A.
Results for the nine-subsystem AAF described in Chapter 3 are presented next.
5.2 Nine-Subsystem Results
The second phase of this thesis work started out by developing the models for the nine
subsystems (i.e. AFS-A, PS, ECS, VC/PAOS, FLS, OLS, ES, CHS, and FCS) described in
Chapter 3, coding and testing the models for a single flight condition, and then updating the code
so that each of the subsystems could fly the 21 segments of the mission described in Chapter 3.
The next step was to integrate the subsystems so that they could “talk” to each other via the
coupling functions within ILGO. The third step was to validate each individual subsystem
within the total AAF system, and the final step was to optimize the AAF using the ILGO
decomposition strategy. All but the final step were carried to completion within this thesis work
due to time and resource limitations. Thus, the following sections present a partial ILGO of the
nine-subsystem AAF system starting with a preliminary synthesis/design exergy analysis.
5.2.1 Preliminary Synthesis/Design Analysis
The nine-subsystem AAF that is partially optimized has the morphing-wing AFS-A as
described in Chapter 3. The set of results presented here represents the subsystem configurations
after the first ILGO iteration. The total exergy destruction plus fuel exergy loss is shown in
Figure 5.2 and the subsystem weights are given in Table 5.3.
Note that the AFS-A and PS dominate the exergy destruction (plus exergy fuel loss) of
the other subsystems in the AAF. This is primarily due to the fact that they consume the largest
amounts of exergy in the AAF: the AFS-A consumes exergy in terms of drag, the PS consumes
exergy during fuel combustion. The other subsystems are not insignificant in terms of aircraft
operation; but in terms of the overall exergy destruction plus fuel exergy loss objective function,
they could be considered negligible. The fact that the AFS-A and PS destroy and lose the most
exergy over the mission would indicate that they are also the subsystems with the most potential
for improvement.
An initial examination of the exergy destruction results shown in Figure 5.2 for the OLS
may lead one to think the result is in error; however, the OLS exergy destruction is only being
tracked for the pump losses. Also, the exergy destruction in the fuel / oil heat exchanger is
148
Figure 5.2 Total exergy destruction plus fuel exergy loss for each of the nine subsystems after the first ILGO
iteration.
accounted to the FLS, and no other energy processes are included in the OLS model. Finally,
taking into account that the weight of the OLS is 0.1% of the gross takeoff weight, the total
exergy destruction for the OLS appears to be reasonable.
When considering the subsystem weights given in Table 5.3, the PS and AFS-A again
have the largest contribution to the gross takeoff weight of the AAF but in a lesser sense. The PS
and AFS-A are easily the heaviest of the subsystems. The resulting weights of the partially
optimized subsystems are given in Table 5.3.
Table 5.3 AAF subsystem weights and the percentage of AAF empty weight after the first ILGO iteration.
AFS-A PS ECS VC/PAOS FLS ES CHS OLS FCS Weight (lbm) 14980.8 5418.3 1297.1 1608.1 491.5 929.5 325.3 37.8 429.3
% empty weight 58.7% 21.2% 5.1% 6.3% 1.9% 3.6% 1.3% 0.1% 1.7%
Note that the subsystem total exergy destruction plus fuel exergy loss (in Figure 5.2) and
weight results (in Table 5.3) are for a partially optimized solution as only the first ILGO iteration
had completed after a few months of execution. The perceived cause of the lengthy execution is
discussed at the end of this chapter. Previous results of the ILGO decomposition applied to a
completed five-subsystem AAF synthesis / design optimization (Rancruel, 2002) are used to
extrapolate the AAF objective function values and subsystem weights found in the present thesis
work towards a system-level optimum. These results are discussed next.
2.6E+07
2.5E+08
6.1E+041.7E+05
4.7E+052.5E+05 3.5E+05
2.2E+01
2.6E+04
1
10
100
1,000
10,000
100,000
1,000,000
10,000,000
100,000,000
1,000,000,000
AFS‐A PS ECS VCPAOS FLS ES CHS OLS FCS
Total Exergy Destruction Plus Fuel Exergy Loss (kJ)
Subsystem Total Exergy Destruction Plus Fuel Exergy Loss
149
5.2.2 Projected Optimum and Comparison
Although only a partial optimization was completed for the nine-subsystem AAF in this
thesis work, previous results of an ILGO decomposition applied to a fixed-wing AAF are used to
predict the optimum for the nine-subsystem AAF system. For example, the work of Rancruel
(2002), previously discussed in Chapter 2, applies ILGO to a fixed-wing AAF system flying a
very similar mission. The weight of each of the subsystems modeled in Rancruel (2002) is
tracked versus each ILGO iteration and the percentage improvement of the subsystem weights is
shown in Table 5.4. If the weight reduction percentages are applied to the result of the first
Table 5.4 AAF Subsystem percent weight reduction versus ILGO iteration (Rancruel, 2002).
ILGO Iteration WTO WPS WECS WVC/PAOS WFLS
Average Reduction
1 -- -- -- -- -- --
2 13.0% 17.1% 20.0% 22.0% 17.9% 18.0%
3 5.3% 9.5% 15.3% 15.6% 15.2% 12.2%
4 3.2% 2.6% 8.2% 7.4% 12.8% 6.9%
5 1.0% 4.5% 3.6% 3.2% 2.9% 3.0%
6 1.4% 1.9% 1.9% 2.1% 2.4% 1.9%
7 0.2% 0.8% 1.9% 2.1% 0.8% 1.1%
ILGO iteration given in the previous section, the resulting optimum AAF system weights are
extrapolated to those given in Table 5.5. Note that no percentage improvement is assumed for
the OLS due to the fact that the OLS weight equation (see Table 3.6) is only a function of the
number of engines. The ES weight is calculated based on AAF empty weight, which is a
Table 5.5 Projected AAF subsystem weights versus ILGO iteration based on the ILGO progression from Rancruel (2002).
ILGO #
AFS-A (lb)
PS (lb)
ECS (lb)
VCPAOS (lb)
FLS (lb)
ES (lb)
CHS (lb)
OLS (lb)
FCS (lb)
1 14980.8 5418.3 1297.1 1608.1 491.5 929.5 325.3 37.8 429.3
2 11611.4 4491.5 1037.7 1255.1 403.7 926.2 325.3 37.8 419.9
3 10705.5 4063.7 879.2 1059.0 342.3 913.6 325.3 37.8 414.4
4 9988.4 3956.8 807.1 980.5 298.4 905.8 325.3 37.8 409.3
5 9944.7 3778.6 778.3 949.2 289.6 903.5 325.3 37.8 408.0
6 9622.1 3707.3 763.9 929.6 282.6 899.9 325.3 37.8 404.4
7 9628.4 3678.8 749.5 909.9 280.4 899.5 325.3 37.8 404.4
150
function of the other subsystem weights. Thus, an iteration for the ES weight (which is a
function of the AFS-A empty weight) is required. The FCS weight is scaled based on the AAF
empty weight, EmptyW , (which consists of the sum of the subsystem weights) but only decreases
weight at a reduced rate compared to EmptyW due to the nature of the FCS equation (see Table
3.6). The FCS weight prediction equation derived is as follows for the ith ILGO iteration:
489.0)1(
1 1
i
FCSiAAFS
iiAAFSi
FCS WW
WW (5.1)
Finally, the CHS weight is not a function of aircraft weight but rather of the number of
hydraulic functions and the level of redundancy on the AAF aircraft. Thus, no improvement was
assumed when projecting the optimum CHS result.
The AAF gross takeoff weight and fuel weight are also extrapolated from the first ILGO
iteration result to completion based on the results from Rancruel (2002) and these extrapolated
results are given in Table 5.6. Note that the empty weight percent improvement is result of the
Table 5.6 Extrapolated nine-subsystem AAF gross takeoff weight and empty weight versus ILGO iteration based on the ILGO progression from Rancruel (2002).
ILGO # TOW (lbm)
% Improvement EmptyW (lbm) %
Improvement
0 60000.0 -- 39665.5 --
1 38599.5 -35.7% 25517.8 -35.7%
Extrapolated Results:
2 33590.4 -13.0% 20508.7 -19.6%
3 31822.5 -5.3% 18740.8 -8.6%
4 30791.2 -3.2% 17709.5 -5.5%
5 30496.5 -1.0% 17414.9 -1.7%
6 30054.6 -1.4% 16972.9 -2.5%
7 29995.6 -0.2% 16913.9 -0.3%
gross takeoff weight and subsystem weight improvement predictions. The predicted system-
level optimum subsystem weights after seven ILGO iterations for the nine-subsystem AAF are
shown in Table 5.7. The optimum results from Rancruel (2002) after seven ILGO iterations are
for a fixed-wing AAF system and greater improvements are expected from an AAF with
morphing-wing technology so the extrapolated optimums shown here may indeed be
conservative.
151
Table 5.7 Extrapolated AAF subsystem system-level optimum weights after seven ILGO iterations along with the percentage of AAF empty weight.
AFS-A PS ECS VC/PAOS FLS ES CHS OLS FCS Weight (lbm) 9628.4 3678.8 749.5 909.9 280.4 899.5 325.3 37.8 404.4 % empty weight 56.9% 21.7% 4.4% 5.4% 1.7% 5.3% 1.9% 0.2% 2.4%
The AAF aircraft system sized in Rancruel (2002) was a much smaller aircraft overall
than the one optimized here. However, some conclusions can be made about the integrity of this
model from a comparison between the two. The extrapolated subsystem weights from this thesis
work are compared to the subsystem weight results from Rancruel in Table 5.8. The
Table 5.8 Extrapolated subsystem optimum weights versus the optimum subsystem weights from Rancruel (2002).
TOW PSW ECSW AFSW FLSW PAOSVCW /
Rancruel (2002) (lbm) 22443.1 2275.2 573.2 6834.3 704.4 511.5 Predicted Optimum (lbm) 29995.6 3678.8 749.5 9628.4 280.4 909.9 % Difference 34% 62% 31% 41% -60% 78%
extrapolated optimums compare relatively well to the results from Rancruel (2002) with one
exception: the weight of the FLS. The explanation for this difference may be because only one
ILGO iteration had completed for the nine-subsystem AAF aircraft system. The initial coupling
function values are arbitrarily set (within limits) for this first ILGO iteration and the subsystem is
optimized without any knowledge of the actual coupling function values from the other
subsystems until this first iteration has completed. Obviously, after this first iteration, the
coupling function values are no longer arbitrary and are updated with actual values for the
subsystem interactions. In other words, the weight of the FLS could have actually increased in
the second ILGO iteration due to having real interactions with the other subsystems. However,
further investigation of the FLS behavior must be left to future work as must verification of the
extrapolated system-level optimum found here.
The following section briefly addresses the highest-payoff operational decision variables
and their effect on the morphing AFS-A model.
5.3 Parametric Study of the Morphing-Wing AFS-A
The benefits of adding morphing-wing technology to an aircraft that flies dissimilar
mission segments has been addressed in the literature review (Butt, 2005) and in the first set of
152
results presented in this thesis (Smith et al. 2007). The morphing-wing AFS-A used in both
studies has been allowed to morph significantly. The benefits of a limited amount of morphing
such as that due to variable wing sweep, high-lift devices, etc. is known. However, the ability to
significantly morph the wing in ways that are not widely used in production (e.g., via wing
length and taper ratio) are addressed in this section via multiple parametric studies.
The first parametric study shows the variation of mission segment fuel burned for three
flight conditions: subsonic climb (Mach 0.536 from a 20,000 ft to 41,700 ft altitude), subsonic
cruise (Mach 0.656 at a 41,700 ft altitude) and supersonic cruise (Mach 1.5 at a 30,000 ft
altitude) which correspond to mission segments 4, 5, and 17, respectively, as found in the
partially optimized subsystem of the nine-subsystem AAF aircraft. A second parametric study of
the subsonic cruise and supersonic cruise mission segments is conducted to compare the exergy
destruction versus the fuel consumption results from the first parametric study to see the trends
of the former with respect to the fuel burned performance metric for the mission segment.
The baseline AFS-A configuration and performance for the 4th, 5th, and 17th mission
segments is given in Table 5.9. Note that the sweep angles, LE , wing lengths, L , and
thickness-to-chord ratios, ct / , are not significantly different between the subsonic and
supersonic mission segments. Furthermore, the performance metric for the mission segment is
the fuel burned. Note also that the PS is not re-optimized for these parametric studies and any
Table 5.9 Baseline AAF configuration and performance for mission segment 4, 5, and 17.
Subsonic Climb
Segment 4
Subsonic Cruise
Segment 5
Supersonic Cruise
Segment 17
LE (deg) 30.3 27.4 27.9
L (ft) 53.4 53.8 52.5
AR 6.09 3.12 3.11
0.30 0.23 0.23
ct / 0.072 0.071 0.073
refS (ft2) 691 731 349
AAF Performance
DL CC / 11.1 11.3 3.4
Fuel Burned (lbm) 437.7 1179.8 225.3
Mission Fuel Burned (lbm)
11299.5
153
decrease in AFS-A fuel burned is due to the energy equation for the specific flight condition (see
Table 3.5) with fixed inputs for sfc from the PS.
The first parametric study looks at the effect of varying the aspect ratio and sweep angle
with respect to the fuel burned for the mission segment. The parametric study performed on the
subsonic climb segment, mission segment 4, is shown in Figure 5.3. The resulting effect of
Figure 5.3 Variation of the mission segment fuel burned with variations in aspect ratio and the sweep angle
for mission segment 4 (subsonic climb at Mach 0.536 from a 20,000 ft to 41,700 ft altitude).
changing the aspect ratio and sweep angle is very small since changing the aspect ratio by 10%
and the sweep angle by 20% from the baseline only yields a change in fuel consumption of 5.7
lbm. The sweep-angle variation shows very little effect on performance as the three different
curves representing the different sweep angles are very close to each other, while the aspect ratio
must be increased beyond the “knee” in the curve to see any appreciable change in the mission
segment. In this case, it appears the partially optimized solution for this mission segment is near
the unit-level optimum solution as little improvement is shown over the baseline fuel
consumption of 437.7 lbm and little sensitivity is observed to fuel burned for changes in sweep
angle or aspect ratio for the subsonic climb mission segment.
The second flight condition (subsonic cruise at Mach 0.656) results show some different
trends and appear in Figure 5.4. Notice that similar to the subsonic climb segment, there is a
436
437
438
439
440
441
442
443
2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 3.5
Fuel Burned in M
ission Segm
ent (lbm)
Aspect Ratio
‐20% Sweep
+20% Sweep
Baseline Sweep
154
definitive “knee” in the curve for the mission segment fuel burned; however, a continued
increase in aspect ratio has a negative effect on the fuel burned. Also, the sweep angle has a
greater effect in this case as compared to the subsonic climb mission segment. Note also that
increasing both wing sweep angle and aspect ratio causes fuel consumption to spike, indicative
of the loss of wing surface area (because the wing length is held constant). Finally, note that
the fuel burned for mission segment 5 for the partially optimized AAF is 1179.8 lbm and little
Figure 5.4 Variation of the mission segment fuel burned with variations in aspect ratio and the sweep angle
for mission segment 5 (subsonic cruise at Mach 0.656 at a 41,700 ft altitude).
improvement potential is shown in Figure 5.4 (only a 0.8% improvement in fuel consumption for
a 20% sweep angle decrease and a 10% increase in aspect ratio) so that the design of the AFS-A
may have already been near the unit-level optimum solution for the subsonic cruise mission
segment as well.
However, subsonic and supersonic flight may yield different results. The results for the
supersonic parametric study are shown in Figure 5.5. The baseline configuration burns 225 lbm
of fuel, but a 20% increase in sweep angle to 36.4 deg for the supersonic mission segment burns
only 190 lbm of fuel for a 15.8% decrease in fuel consumption. A 10% increase in aspect ratio
yields an 11.7% decrease in fuel burned or 198 lbm. The most improved point in Figure 5.5 is for
a 20% increase in sweep angle and 20% increase in aspect ratio for a 31% decrease in fuel
consumption over the baseline configuration. This parametric study indicates that the partially
1160
1170
1180
1190
1200
1210
1220
1230
1240
1250
1260
1270
2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9
Fuel Burned oin M
ission Segm
ent (lbm)
Aspect Ratio
‐20% Sweep
BaselineAR
+10% AR
‐10%
AR
+20% AR
155
optimized solution, the baseline, is not fully optimized as the parametric study yields a
significant decrease in fuel consumption.
Figure 5.5 Variation of the mission segment fuel burned with variations in aspect ratio and the sweep angle
for mission segment 17 (supersonic cruise at Mach 1.5 and 30,000 ft altitude).
A second study is conducted to see the variation of fuel consumption with variations in
the thickness-to-chord ratio, ct / , and taper ratio, , for both the subsonic and supersonic
mission segments. The subsonic climb mission segment results are shown in Figure 5.6.
Figure 5.6 Variation of the mission segment fuel burned with variations in thickness-to-chord ratio for
mission segment 4 (subsonic climb at Mach 0.536 from a 20,000 ft to 41,700 ft altitude).
156
Changing the taper ratio showed no measureable effect on the subsonic climb mission, but the
thickness to chord ratio did have some effect and the result also indicates the baseline aircraft
configuration was at a near-optimum value.
The supersonic cruise mission segment study showed less than a 1% change for fuel
consumption for an increase or decrease of the thickness-to-chord ratio by 10% and/or a 20%
increase or decrease in the taper ratio. This highlights the fact that in supersonic flight, the wing
thickness and taper ratio make a smaller contribution to the total drag via the skin friction drag
and that the total drag is dominated by the wave drag in supersonic flight. The subsonic cruise
segment was similarly insensitive to changes in the thickness-to-chord and taper ratio with
respect to fuel consumption.
The conclusion from the taper ratio and thickness to chord ratio study is that either the
taper ratio and thickness to chord ratio have a negligible effect on the aircraft performance in
cruise flight, or more likely that the drag buildup method and lift calculations do not have the
resolution required to effectively measure their effect on aircraft performance. The only
measurable effect found on fuel consumption was the thickness-to-chord ratio during the
subsonic climb segment. Furthermore, it should be emphasized that these changes are made
about a point (i.e., synthesis/design), which is already optimal or nearly so. Thus, the conclusions
drawn cannot be generalized to syntheses/designs, which may be far from optimal. Because the
change in fuel consumption for significant change in thickness-to-chord ratios and taper ratios
was negligible for the subsonic and supersonic cruise segments, the results will not be given
here.
The results of the first two parametric studies for varying aspect ratio, sweep angle, taper
ratio, and thickness-to-chord ratio are based on using fuel burned as the performance metric. To
show what the results would be if the performance metric were changed to exergy destruction, a
final parametric study is conducted. This study examines the variation in mission segment
exergy destruction with variations in aspect ratio and sweep angle for the subsonic cruise and
supersonic cruise mission segments (segments 5 and 17, respectively). The baseline sweep and
aspect ratio are shown in Table 5.9; and the taper ratio is varied up to 20% from the baseline,
while the sweep angle is increased and decreased 20% from the baseline. The subsonic cruise
results are shown in Figure 5.7, and the supersonic cruise results are shown in Figure 5.8.
157
Figure 5.7 Variation of the mission segment exergy destruction with variations in aspect ratio and the sweep
angle for mission segment 5, subsonic cruise at Mach 0.656 at 41,700 ft altitude.
Figure 5.8 Variation of the mission segment exergy destruction with variations in aspect ratio and the sweep
angle for mission segment 17, supersonic cruise at Mach 1.5 and 30,000 ft altitude.
The subsonic cruise results in Figure 5.7 follow the fuel burned results in Figure 5.4
relatively closely. Note that the “bump” for the ‘+20% Sweep’ curve in Figure 5.4 barely
appears in Figure 5.7 (i.e., is significantly smoothed out). There may be multiple reasons for the
158
fuel consumption result being more volatile than the exergy destruction result, such as the fact
that the fuel burned is technically dependent on the PS, which in this case is considered to be
constant. The bump may disappear with additional PS optimizations. That being said, the trends
of the exergy destruction results follow the fuel burned results quite well for the subsonic cruise
segment.
The supersonic cruise segment total exergy destruction trends shown in Figure 5.8 follow
the mission segment fuel burned trends shown in Figure 5.5 also rather well. The conclusion at
least for the case when the AFS-A is optimized by itself for a fixed sfc is that the exergy
destruction is directly proportional to the fuel consumption of the AAF in supersonic cruise and
subsonic cruise. However, as shown in Periannan, von Spakovsky, and Moorhouse (2008)
where a three-subsystem AAF aircraft with AFS-A, PS, and ECS is optimized, this may not
always be the case. Furthermore, the exergy destruction result presented here appears to have a
smoother response than the fuel burned result for changes in the AFS-A wing geometry.
159
Chapter 6
Conclusions/Recommendations
This thesis work produced a number of conclusions and recommendations related to the
morphing-wing AAF synthesis/design optimization problem. The first set of results from Smith
et al. (2007) was meant to be an extension of the work of Butt (2005) and a stepping stone for the
new PS model being used in the nine-subsystem AAF. The work and research required to have
all 9 subsystems fly the mission and integrating the subsystems for the nine-subsystem AAF
proved to be the bulk of the work for this thesis. However, a few conclusions and
recommendations are drawn, which include the following:
1. The morphing-wing AFS-A and turbofan PS results compare well with Butt (2005). The
high wing-weight and fuel-weight penalties for morphing-wing technology still show an
improvement over a fixed-wing AAF flying the same mission. However, the morphing-
wing-fuel and wing-weight penalties are arbitrarily set due to a lack of information about
the methods used to morph the wings. Further study is recommended to establish the
actual structural weight penalties and power requirements for the morphing-wing
technology.
2. The results given in Smith et al. (2007) allude to the possibility of a new paradigm or way
of thinking to be adopted for supersonic fighter aircraft design. A fixed-wing supersonic
AAF design can typically be viewed as “design a supersonic AAF that can fly at subsonic
speeds” while the morphing-wing supersonic AAF design could be viewed as “design a
subsonic AAF that can fly at supersonic speeds.” In other words, morphing-wing
technology is an enabling technology that allows aircraft designers to move away from
designing aircraft for supersonic flight, enabling them instead to concentrate on an
efficient aircraft design that has the capability of flying at supersonic speeds.
3. The partially optimized nine-subsystem results show a single one inconsistency with
previous results from Rancruel (2002), namely, in the FLS weight. However, this is
somewhat to be expected due to the fact that the set of results produced here is not for a
fully optimized solution and does not have the interactions with the other subsystems
established at actual values since only the first ILGO iteration completed. That being
160
said, the FLS weight equations and subsystem interactions may need to be studied in
more detail to verify model accuracy.
4. The extrapolated optimum solutions for the nine-subsystem AAF are likely conservative
compared to the results shown in Rancruel (2002) because of the additional benefit of
morphing-wing technology. The PS in particular should have a smaller design mass flow
rate and design Mach number, which would likely, along with the benefits of morphing-
wing technology, drive the empty weight lower than what is found here.
5. The difficulty of optimizing the nine-subsystem AAF proved to be in calculation time
and resources available. It is believed that the calculation time of the iSCRIPT™
software package is significantly slower than what was expected, in particular, when
calculation-heavy code is running. Initial partial optimizations of the nine-subsystem
AAF showed the estimated time-to-optimize at nearly 7 months using parallel processing
on an Intel Q6600-based (quadruple core, 2.4 GHz) PC.
6. Further validation of every subsystem model and additional integration within the AAF
was not feasible during this thesis work considering the calculation / optimization time
and human resources available to do the work. Thus, additional work on improving
calculation times and on subsystem validation and integration not only at the unit-level
but also at the system-level is recommended. Full validation and verification of the nine-
subsystem AAF requires a significant amount of computing power using the iSCRIPT™
software platform.
7. A recommendation for further study would be to use advanced fixed-wing fighter aircraft
designs such as a blended-wing-body (BWB) to represent the most advanced fixed-wing
technology available. The aerodynamics model of a BWB would likely require a higher-
fidelity drag model to model the BWB’s flight performance compared to that of the
morphing-wing AFS-A on a conceptual level.
8. The negligible effect of varying the thickness-to-chord ratio and taper ratio in the
morphing-wing parametric studies for subsonic and supersonic cruise flight highlights the
possibility that the component drag buildup method used does not have the resolution
required to show a measurable effect of varying these wing geometry parameters. If
further study is to be done with the drag model from Raymer (2003), it is recommended
161
that the thickness-to-chord ratio be set to a constant parameter for cruise segments and
not participate as a DOF in the optimization except perhaps during climb segments.
9. The high-payoff wing parameters to morph should be investigated further. The benefits
of the application of variable wing sweep is well known via the F-14 Tomcat. However,
drastic wing morphing may not be possible in the near future and the best-value-
geometry morphing with the highest payoff and least risk should be identified via cost /
benefit studies and pursued.
10. The structural weight and power required for discrete types of morphing need to be
established (e.g., the penalties for wing length morphing, the penalties for variable wing
sweep, etc.), and conceptual-level optimizations should be used to find the best set of
morphing parameters to study.
11. The exergy destruction and fuel consumption were compared to find trends in their
predictions of the AAF for both supersonic and subsonic cruise mission segments. The
exergy destruction objective trend directly correlates with the fuel burned objective trend
(at least for the AFS-A optimized in isolation) but has a smoother response than the fuel
burned objective for changes in the AAF wing configuration. The supersonic cruise
flight condition matches better than the subsonic cruise flight condition. Recommended
further study in this area would be to perform a parametric study of the AFS-A, but
allowing the PS and possibly another subsystem such as the ECS to optimize (or at least
vary the throttle setting) for each set of AFS-A wing geometries to see if the direct
correlation between fuel consumption and exergy destruction is maintained, or instead
follows the trends shown in Periannan, von Spakovsky, and Moorhouse (2008).
162
References
1. Anderson Jr., J. D., 2001, Fundamentals of Aerodynamics, 3rd Edition, New York: McGraw-Hill.
2. Anderson, J.D. , 1984, Fundamentals of Aerodynamics, New York: McGraw-Hill. 3. Anderson, J.D. Jr., 1999, Aircraft Performance and Design, WCB/McGraw-Hill 4. Bejan, 1996, Entropy Generation Minimization, pg 37 5. Bejan, A. 1996. Entropy Generation Minimization: The Method of Thermodynamic
Optimization of Finite-Sized Systems and Finite-Time Processes. Boca Raton, FL: CRC Press, Inc.
6. Beretta, G. P., Gyftopoulos, E. P., Park, J. L., and Hatsopoulos, G. N., 1984, Quantum Thermodynamics: A New Equation of Motion for a Single Constituent of Matter, Il Nuovo Cimento, 82 B, 2, pp. 169-191.
7. Beretta, G. P., 2006, Steepest-Entropy-Ascent Irreversible Relaxation Towards Thermodynamic Equilibrium: The Dynamical Ansatz that Completes the Gyftopoulos-Hatsopoulos Unified Theory with a General Quantal Law of Causal Evolution, International Journal of Thermodynamics, ICAT, Istanbul, Turkey, Sept., vol. 9, no. 3.
8. Bolz, R. E., & Tuve, G. L. (1973). CRC handbook of tables for applied engineering science. 2d ed. Editors: Ray E. Bolz [and] George L. Tuve. CRC handbook series. Cleveland, Ohio: Chemical Rubber.
9. Bowcutt, K., Hypersonic Aircraft Optimization Including Aerodynamic, Propulsion, and Trim Effects, AIAA Paper No. 19925055, 1992.
10. Bowman, J., Sanders, B., Weisshaar, T., 2002, Evaluating the Impact of Morphing Technologies on Aircraft Performance, AIAA Paper 2002-1631.
11. Brewer, K.M., 2006. Exergy Methods for Mission-Level Analysis and Optimization of Generic Hypersonic Vehicle Concepts. Department of Mechanical Engineering, Virginia Polytechnic Institute and State University, M.S. Thesis.
12. Butt, J.R., 2005. A Study of Morphing Wing Effectiveness in Fighter Aircraft using Exergy Analysis and Global Optimization Techniques. Department of Mechanical Engineering, Virginia Polytechnic Institute and State University, M.S. Thesis.
13. Chase, M. W. 1998. NIST-JANAF thermochemical tables. Journal of physical and chemical reference data, no. 9. [Washington, D.C.]: American Chemical Society
14. Cloyd, J.S.; A status of the United States Air Force’s More Electric Aircraft Initiative. Energy Conversion Engineering Conference, 1997. IECEC-97.
15. CRC, 1976. Handbook of Tables for Applied Engineering Science. CRC Press. 16. Crowe, 2001. Engineering Fluid Dynamics. pg 16, 418-419. 17. Curry, R.E. and Sim, A.G. 1982, “The Unique Aerodynamic Characteristics of the AD-1
Oblique-Wing Research Airplane,” AIAA paper 82-1329, AIAA 9th Atmospheric Flight Mechanics Conference, Aug 9-11 1982, San Diego, CA.
18. Davenport, C.J., 1983, “Correlations for Heat Transfer and Flow Friction Characteristics of Louvered Fin,” AIChE Symposium Series, No.225, 79, 19-27.
19. Dryden X-Press, 1979, AD-1 Construction Completed, Feb 23, p. 2. 20. El-Sayed, Y.M. and Evans, R.B., 1970, Thermoeconomics and the Design of Heat Systems,
Journal of Engineering for Power, ASME Transactions, vol. 118, October, N.Y. 21. Elgerd, O.I., van der Puije, P.D. 1998. Electric Power Engineering. pg 74, 75
163
22. Figliola R. S., Tipton, R., Ochterbeck, J.M.,1997. “Thermal Optimization of the ECS on an Aircraft with an Emphasis on System Efficiency and Design Methodology,” Society of Automotive Engineers, Inc. 971241 pp 137-143.
23. Frangopoulos, C.A., Evans R.B., 1984, Thermoeconomic Isolation and Optimization of Thermal System Components, Second Law Aspects of Thermal Design, HTD Vol. 33, ASME, N.Y., N.Y., August.
24. Frangopoulos, C.A., von Spakovsky, M. R., Sciubba, E., 2002, A Brief Review of Methods for the Design and Synthesis Optimization of Energy Systems, International Journal of Applied Thermodynamics, ICAT, Istanbul, Turkey, December, vol. 5, no. 4.
25. Greene R., 1992, Compressors, Selection, Usage, and Maintenance, McGraw Hill Inc, USA. 26. Gyftopoulos, E. P., and Beretta, G. P., 1991, Thermodynamics: Foundations and
Applications, Macmillan Publishing Co., New York. 27. Gyftopoulos, E. P. and Beretta, G. P., 2005, Thermodynamics – Foundations and
Applications, Dover Pub., N. Y. 28. Hatsopoulos, G. N. and Gyftopoulos, E. P., 1976a,b,c,d, A Unified Quantum Theory of
Mechanics and Thermodynamics – Part I: Postulates, Part IIa: Available Energy, Part IIb: Stable Equilibrium States, Part III: Irreducible Quantal Dispersions, Foundations of Physics, 6, 1, pp. 15-31, 2, pp. 127-141, 4, pp .439-455, 5, pp. 561-570.
29. Heiser, W. H., and Pratt, D. T., 1994, Hypersonic Airbreathing Propulsion, AIAA, Inc., Washington, D.C.
30. Hudson W. A. and Levin M. L., 1986. Integrate Aircraft Fuel Thermal Management. Rockwell International Corp. El Segundo, CA. Intersociety Conference on Environmental Systems (16th: 1986: San Diego, Calif.). Aerospace environmental systems: proceedings of the Sixteenth ICES conference. Pg 11-25.
31. Incropera F. and DeWitt D., 1990, Fundamental of Heat and Mass Transfer. John Wiley and Sons, Inc.
32. Kakac S. and Liu H., 2002, Heat Exchangers Selection, Rating, and Thermal Design. CRS press.
33. Kays, W. M. and London, A. L., 1984, Compact Heat Exchangers, McGraw-Hill, Inc. pp 14, 16, 17, 18, 166, 287.
34. Kotas, T.J., 1985, The Exergy Method of Thermal Plant Analysis. 35. Leyland, G.B., 2002, Multi-Objective Optimization Applied to Industrial Energy Problems,
Department of Mechanical Engineering, University of Auckland, M.S. Thesis. 36. Majumdar, 2003, Oil Hydraulic Systems: Principles and Maintenance, McGraw-Hill, Inc.
pp 100-101, 133, 256, 320-323. 37. Maldonado, M.A.; Shah, N.M.; Cleek, K.J.; Walia, P.S.; Korba, G.J.; 1997; Power
Management and Distribution System for a More-Electric Aircraft (MADMEL) – program status, Energy Conversion Engineering Conference. Proceedings of the 32nd Intersociety, Volume 1, 27-July—August. Page(s): 274-279
38. Markell, K. C., 2005, Exergy Methods for the Generic Analysis and Optimization of Hypersonic Vehicle Concepts, Department of Mechanical Engineering, Virginia Polytechnic Institute and State University, M.S. Thesis.
39. Mattingly, J.D., Heiser, W.H. and Daley, D.H., 1987, Aircraft Engine Design, AIAA Education Series, New York, New York.
40. Mattingly, J. D., Heiser, W.H., and Pratt, D.T., 2002, Aircraft Engine Design, AIAA Education Series, Washington, DC.
164
41. Moir, I.; “More-electric aircraft-system considerations, Electrical Machines and Systems for the More Electric Aircraft (Ref. No. 1999/180)”, IEE Colloquium on 9 Nov. 1999, Page(s):10/1 - 10/9
42. Moir, I., Seabridge, A., 2001, Aircraft Systems: mechanical, electrical, and avionics subsystems integration, pg 164
43. Moorhouse D. J., Sanders B., von Spakovsky, M.R., and Butt, J., 2005, Benefits and Design Challenges of Adaptive Structures for Morphing Aircraft, International Forum on Aeroelasticity and Structural Dynamics 2005 (IFASD’05), AIAA, DLR, Bogenhausen, Germany, June 28th to July 1st.
44. Moorhouse D. J., Sanders B., von Spakovsky, M.R., and Butt, J., 2006, Benefits and Design Challenges of Adaptive Structures for Morphing Aircraft, The Aeronautical Journal, Royal Aeronautical Society, London, England, vol. 110, no. 1105, March, 2006, p 157-162.
45. Moran, M. J., 1989, Availability Analysis: a Guide to Efficient Energy Use, ASME Press, New York, New York.
46. Muñoz J. R., von Spakovsky M.R., 1999, A Second Law Based Integrated Thermoeconomic Modeling and Optimization Strategy for Aircraft / Aerospace Energy System Synthesis and Design (Phase I - Final Report), final report, Air Force Office of Scientific Research, New Vista Program, December.
47. Muñoz, J.R., von Spakovsky, M.R., 2000, The Use of Decomposition for the Large-Scale Synthesis / Design Optimization of Highly Coupled, Highly Dynamic Energy Systems - Theory, International Mechanical Engineering Congress and Exposition – IMECE’2000, ASME, Orlando, Nov. 5-10, AES-Vol. 40.
48. Muñoz, J.R., von Spakovsky, M.R., 2000, The Use of Decomposition for the Large-Scale Synthesis / Design Optimization of Highly Coupled, Highly Dynamic Energy Systems - Application, International Mechanical Engineering Congress and Exposition –IMECE’2000, ASME, Orlando, Nov. 5-10, AES-Vol. 40.
49. Muñoz, J.R., von Spakovsky, M.R., 2000, Decomposition in Energy System Synthesis / Design Optimization for Stationary and Aerospace Applications, 8th AIAA/NASA/USAF/ISSMO Symposium on Multidisciplinary Analysis and Optimization, Long Beach, CA, September 6-8.
50. Muñoz, J.R., von Spakovsky, M.R., 2000, An Integrated Thermoeconomic Modeling and Optimization Strategy for Aircraft / Aerospace Energy System Design, Efficiency, Costs, Optimization, Simulation and Environmental Aspects of Energy Systems (ECOS’00), Twente University, ASME, Netherlands, July 5-7.
51. Muñoz, J.R., von Spakovsky, M.R., 2001, A Decomposition Approach for the Large Scale Synthesis/Design Optimization of Highly Coupled, Highly Dynamic Energy Systems, International Journal of Applied Thermodynamics, March, vol. 4, no. 1.
52. Muñoz, J.R., von Spakovsky, M.R., 2001, The Application of Decomposition to the Large Scale Synthesis/Design Optimization of Aircraft Energy Systems, International Journal of Applied Thermodynamics, June, vol. 4, no.2.
53. Muñoz, J.R., von Spakovsky, M.R., 2003, Decomposition in Energy System Synthesis / Design Optimization for Stationary and Aerospace Applications, AIAA Journal of Aircraft, special issue, Vol. 39, No. 6, Jan-Feb.
54. Nicolai, L., 1975, Fundamentals of Aircraft Design, published by the author. 55. Nist-janaf thermochemical tables, fourth edition. 1998. [S.l.]: American Institute Of Phy.
165
56. Pearson, W.; The More Electric/All Electric Aircraft – A Military Fast Jet Perspective; All Electric Aircraft (Digest No. 1998/260) June 1998; Page(s) 5/1 – 5/7
57. Periannan, V., 2005, Investigation of the Effects of Different Objective Functions/Figures of Merit on the Analysis and Optimization of High Performance Aircraft System Synthesis/Design, Department of Mechanical Engineering, Virginia Polytechnic Institute and State University, M.S. Thesis.
58. Periannan, V., von Spakovsky, M. R., and Moorhouse, D. J., 2008, “A study of various energy- and exergy-based optimisation metrics for the design of high performance aircraft systems”, The Aeronautical Journal, vol. 112, no. 34
59. Periannan, V., von Spakovsky, M.R., Moorhouse, D., 2008, Energy versus Exergy-Based Figures of Merit for the Optimal Synthesis/Design of High Performance Aircraft Systems, The Aeronautical Journal, Royal Aeronautical Society, London, England, vol. 112, no. 1134, pp 449-458.
60. Rancruel, D. F., 2003, A Decomposition Strategy Based on Thermoeconomic Isolation Applied to the Optimal Synthesis/Design and Operation of an Advanced Fighter Aircraft System, Department of Mechanical Engineering, Virginia Polytechnic Institute and State University, M.S. Thesis.
61. Rancruel, D. F., von Spakovsky, M. R., 2003a, “Decomposition with Thermoeconomic Isolation Applied to the Optimal Synthesis/Design of an Advanced Fighter Aircraft System”, International Journal of Thermodynamics, ICAT, Istanbul, Turkey, September, vol. 6, no. 3.
62. Rancruel, D. F., von Spakovsky, M. R., 2003b, “A Decomposition Strategy Applied to the Optimal Synthesis/Design and Operation of an Advanced Fighter Aircraft System: A Comparison with and without Airframe Degrees of Freedom”, International Mechanical Engineering Congress and Exposition – IMECE’2003, ASME Paper No. 44402, N.Y., N.Y., November.
63. Rancruel, D. F., von Spakovsky, M. R., 2004a, “Advanced fighter aircraft sub-systems optimal synthesis/design and operation: Airframe integration using a thermoeconomic approach”, 10th AIAA/ISSMO Multi-disciplinary Analysis and Optimization Conference, Aug. 30 - Sept. 1, Albany, New York, vol. 6, p 3403-3415.
64. Rancruel, D. F., von Spakovsky, M. R., 2004b, “A Decomposition Strategy based on Thermoeconomic Isolation Applied to the Optimal Synthesis/Design and Operation of an Advanced Tactical Aircraft System, Efficiency, Costs, Optimization, Simulation and Environmental Aspects of Energy Systems” (ECOS’04), Guanajuato, Mexico, ASME, July.
65. Rancruel, D. F., von Spakovsky, M. R., 2004, “Use of a Unique Decomposition Strategy for the Optimal Synthesis/Design and Operation of an Advanced Fighter Aircraft System”, 10th AIAA/ISSMO Multi- disciplinary Analysis and Optimization Conference, Aug. 30 - Sept. 1, Albany, New York.
66. Rancruel, D., 2005, Dynamic Synthesis/Design and Operation/Control Optimization Approach Applied to a Solid Oxide Fuel Cell based Auxiliary Power Unit under Transient Conditions, Department of Mechanical Engineering, Virginia Polytechnic Institute and State University, Ph.D. Dissertation.
67. Rancruel, D. F., von Spakovsky, M. R., 2006, “A Decomposition Strategy based on Thermoeconomic Isolation Applied to the Optimal Synthesis/Design and Operation of an Advanced Tactical Aircraft System”, Energy: The International Journal, Elsevier, available
166
on-line at http://dx.doi.org/10.1016/j.energy.2006.03.004, vol. 31, no. 15, December, p 3327-3341.
68. Raymer, Daniel P., 1999, Aircraft Design: A Conceptual Approach, AIAA Education Series, Reston, VA.
69. Raymer, Daniel P., 2006, Aircraft Design: A Conceptual Approach, Fourth Edition, AIAA Education Series, Reston, VA.
70. Riggins, D.W., “Evaluation of Performance Loss Methods for High-Speed Engine and Engine Components”, Journal of Propulsion and Power, Vol. 13, No. 2, March-April 1997.
71. Riggins, D. W., “High-Speed Engine/Component Performance Assessment Using Exergy and Thrust Based Methods,” NASA Contractor Report 198271, 1996.
72. Riggins, D. W., Private Communication, July 2003 - July 2004 73. Roskam, J., Airplane Design Part: II, 1985, pp 110, 185-186. 74. Saravanamuttoo, H.I.H., Rogers, G.F.C, Cohen, H., 2001, Gas Turbine Theory, Pearson
Education Limited, Harlow, England. 75. Shah, R. K., 1981, Compact Heat Exchanger Design Procedure in Heat Exchangers:
Thermal Hydraulic Fundamentals and Design, S. Kakac, AE Bergles, and F. Matinger editors, Hemisphere Publishing Corporation, pg 495-536
76. Shah, R.K. and Webb, R.L., 1982, Compact and Enhanced Heat Exchangers, in Heat Exchangers: Theory and Practice, J. Taborek, G.F. Hewitt, and N. Agfan editors, Hemisphere Publishing Corporation, pp 435-468
77. Smith, C. E., von Spakovsky, M.R., 2007, Time Evolution of Entropy in a System Comprised of a Boltzmann Type Gas: An Application of the Beretta Equation of Motion, International Mechanical Engineering Congress and Exposition – IMECE’2007, ASME Paper No. IMECE2007-42933, N.Y., N.Y., November.
78. Smith, K., Butt, J., von Spakovsky, Moorhouse, D., 2007, A Study of the Benefits of Using Morphing Wing Technology in Fighter Aircraft Systems, 39th AIAA Thermophysics Conference, Miami, FL, June 25-28.
79. Starkey, R. P., “Scramjet Optimization for Maximum Off Design Performance,” AIAA Paper No. 20043343, 2004.
80. Szargut J., Morris D.R., Steward F.R. (1988). Exergy Analysis of Thermal, Chemical and Metallurgical Processes. 332 pp. Hemisphere Publ. Corp., New York.
81. von Spakovsky, M.R., and Evans, R.B., 1993, Engineering Functional Analysis (Part I), Journal of Energy Resources Technology, ASME Transactions, Vol. 115, No. 2, N.Y., June.
82. von Spakovsky, M.R., 2008, The Second Law: A Unified Approach to Thermodynamics Applicable to All Systems and All States, Keenan Symposium: Meeting the Entropy Challenge, MIT, American Institute of Physics (AIP), AIP Proceedings, Cambridge, MA, October 2007.
83. von Spakovsky, M.R., Smith, C. E., Verda, V., 2008, Quantum Thermodynamics for the Modeling of Hydrogen Storage on a Carbon Nanotube, International Mechanical Engineering Congress and Exposition – IMECE’2008, ASME Paper No. IMECE2008-67424, N.Y., N.Y., Oct.-Nov.
84. Weimer, J.A.; “The Role of Electric Machines and Drives in the More Electric Aircraft”, Electric Machines and Drives Conference, 2003. IEMDC’03. IEEE International, Volume 1, 1-4 June 2003. Page(s): 11-15
167
85. Zábranský, M., Ruzicka, V., Majer, V., and Domalski, E.S., (1996). Heat capacity of liquids: Critical review and recommended values. Journal of physical and chemical reference data, no. 6. [Washington, D.C.]: American Chemical Society.
168
Appendix A
Fan Performance Map Code
function [eff] = fan(xfan, Pi_cprim, Pi_cprimR) Pi_cprimR_map = 1.821; PRf = Pi_cprim/Pi_cprimR*Pi_cprimR_map; %Curve fitted equations for fan map, mattingly page 159, the elseif loops %progress from left to right across the fan map and checks for above stall %conditions or too low of flow rates. Low flow rates will set the %efficiency of the fan to 70%. Design fan efficiency is 0.89 (mattingly %pg 125. xact = xfan; ystall = 0.00013089775410*xact^2 -0.00518069645313*xact + 1.26052318037349; ycal2 = 0.00000005785153*xact^4 -0.00001162747108*xact^3 + 0.00092003348112*xact^2 -0.02526078062955*xact + 1.35610991970601; %left 80% ycal3 = 0.00000000685532*xact^5 -0.00000278371525*xact^4 + 0.00044925428731*xact^3 - 0.03590113760590*xact^2 + 1.42444807870004*xact + -21.31564085462291; %right 80% ycal4 = 0.00000265672539*xact^3 -0.00041471513906*xact^2 + 0.02982941208547*xact + 0.48191930572447; %left 82% ycal5 = 0.00000740839366*xact^3 -0.00168038353132*xact^2 + 0.13433234460665*xact -2.40691063865553; %right 82% ycal6 = -0.00000154556212*xact^3 + 0.00021313171573*xact^2 + 0.01364862191487*xact -0.13343794806055; %left 84% ycal7 = 0.00000187730332*xact^4 -0.00063352000447*xact^3 + 0.08021250388788*xact^2 -4.50398173025479*xact + 95.86165686002784; %right 84% yop = 0.00017181510217*xact^2 -0.01465955277341*xact + 1.56921934760171; %operating line if ((xact > 110 | xact < 30)) %if actual percentage of design flow rate is out of operating range set eff to 75% eff = 0.78; else if ystall >= PRf & PRf > ycal2 & PRf > yop & xact <= 90 %if PRf is greater than 78% eff, and less than stall, calc eff - assumes next line is 78% eff eff = 0.8-0.02*(PRf-ycal2)/(ystall-ycal2); else if ycal4 < PRf & PRf <= ycal2 & ystall >=PRf & PRf > yop & xact<=90 %PRf is less than 82% eff, greater than 80% eff, less than 90%, calc eff eff = 0.82-0.02*(PRf-ycal4)/(ycal2-ycal4); else if ycal4 < PRf & ystall >=PRf & PRf > yop & xact>90 %PRf is less than 82% eff, greater than 80% eff, over 90%, calc eff eff = 0.82-0.02*(PRf-ycal4)/(ystall-ycal4); else if ycal4 > PRf & PRf >= ycal6 & ystall >=PRf & PRf > yop & ((xact <= 100 & xact >=75)) %greater than 82% eff, above the 84% curve only eff = 0.84-0.02*(PRf-ycal6)/(ycal4-ycal6);
169
else if ycal4 > PRf & PRf >= yop & xact >100 %greater than 82% eff, over 100% eff = 0.89-0.07*(PRf-yop)/(ycal4-yop); else if ycal4 >= PRf & PRf >=yop & xact < 75 %greater than 82% eff, left of the 84% curve, over operating line eff = 0.89-0.05*(PRf-yop)/(ycal4-yop); else if yop < PRf & PRf <= ycal6 & ((xact<100 & xact >75)) %greater than 84% eff, over operating line eff = 0.89-0.05*(PRf-yop)/(ycal6-yop); else if PRf >= ycal3 & PRf<= ycal5 & ((xact>=65 & xact<=110)) %between 80 and 82% efficient, under operating line eff = 0.82-0.02*(ycal5-PRf)/(ycal5-ycal3); else if PRf>=ycal5 & PRf<=ycal7 & ((xact>=73 & xact<=100)) %between 82 and 84% efficient, only under 84% curve, under operating line eff = 0.84-0.02*(ycal7-PRf)/(ycal7-ycal5); else if PRf>=ycal7 & PRf<=yop & ((xact>=73 & xact<=100)) %between 84 and 89% efficient, under operating line eff = 0.89-0.05*(yop-PRf)/(yop-ycal7); else if PRf>=ycal5 & PRf<=yop & xact<73 %between 82 and 89% efficient, left of 84% curve, under op line eff = 0.89-0.07*(yop-PRf)/(yop-ycal5); else if PRf>=ycal5 & PRf<=yop & xact>=100 %between 82 and 89% efficient, right of 84% curve, under op line eff = 0.89-0.07*(yop-PRf)/(yop-ycal5); else eff = 0.78; %if points are off the fan map, give 78% default efficiency end end end end end end end end end end end end end