A Conceptual Design and Optimization Method for
Blended-Wing-Body Aircraft
Roelof Vos
∗
and Jorrit van Dommelen
†
Delft University of Technology, Delft, The Netherlands
This paper details a new software tool to aid in the conceptual design of blended-wing-
body aircraft. The tool consists of four main modules. In the preliminary sizing model a
class I estimate of the maximum take-off weight, wing loading, and thrust-to-weight ratio
is calculated. This information is used together with an initial guess of the 30 design
variables that to form a geometric model of the aircraft. From this geometric model four
disciplinary models are derived: an aerodynamic model, a model of the wing box structure,
a model for the cabin, and a model for the fuel tank. In the subsequent analysis module
refined weight estimation for the operating-empty weight is being calculated, as well as
the center-of-gravity shift during loading, the static margin, the main stability and control
derivatives, and the harmonic range. In the last module, these analysis results are compared
to 27 nonlinear constraints stemming from the top-level requirements and the aviation
regulations. A gradient-based optimization routine is employed to find a combination of
the design variables that satisfies all constraints while optimizing for harmonic range at
constant maximum take-off weight. Between 20 and 60 iterations are required to achieve
convergence. The tool has been set-up to allow for maximum configurational flexibility
such as forward-swept outer wings, under-the-wing engines, and twin vertical tails.
I. Introduction
For conventional aircraft configurations, such as the tube-and-wing (TAW) aircraft, commercially-off-
the-shelf (COTS) design tools are available for their conceptual design. Traditionally, these tools are based
on proven handbook methods and embedded in a software package (e.g. DARcorp’s AAA or PASS by
Desktop Aeronautics). Typically, these packages reduce top-level requirements along with constraints on
handling qualities and safety to a Class II airplane design. Subsequently, the design is further refined at the
preliminary design level where the fidelity of the analysis tools increases progressively.
At TU Delft continuous efforts have b een directed towards the implementation of a design and engineering
engine (DEE) that can facilitate multi-disciplinary design optimization (MDO) beyond the Class II. One
important benefit is the ability of the DEE to analyze and size both conventional and unconventional
configurations. As was demonstrated by Schut et al., the initial input vector for the DEE plays a pivotal
role in whether all the constraints can be met once the MDO process has converged.
1
It was argued that a
separate lower-fidelity DEE would be necessary to define an input vector that would yield feasible solutions
in the subsequent optimization process. This lower-fidelity DEE was termed the conceptual DEE because it
translated the top-level requirements to an initial input vector for the subsequent MDO process. Likewise,
the output from the subsequent DEE could be used as an input vector for a more advanced DEE, enabling
a multi-fidelity design and optimization process. This process is schematically depicted in Figure 1.
Considering the case of TAW aircraft, the very first initiator DEE is readily available in the form of the
aforementioned Class II airplane design tools. Because these tools are based on empirical relationships, the
likelihoo d of having a feasible design as an output of this first design step is rather high. For non-traditional
aircraft configurations, there are no recipes that ensure a feasible design at the end of the conceptual
design phase. On the other hand, it is desirable to assess different configurations on their feasibility at
∗
Assistant Professor, Faculty of Aerospace Engineering at Delft University of Technology, Kluyverweg 1, 2629 HS Delft, The
Netherlands, AIAA member
†
Engineer at Barge Master, Karel Doormanweg 9, 3115 JD Schiedam, The Netherlands
1 of 14
American Institute of Aeronautics and Astronautics
53rd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference<BR>20th AI
23 - 26 April 2012, Honolulu, Hawaii
AIAA 2012-1756
Copyright © 2012 by R. Vos and J. van Dommelen. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
Downloaded by TECHNISCHE UNIVERSITEIT DELFT on February 26, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2012-1756
Top level
requirements
New airplane
model
New airplane
model
New airplane
model
Figure 1. Initiator Component of a DEE Consists of Multiple DEEs of Simplified Problems (Modified from
Schut et al.
1
)
low computational cost. The internal optimization loop within the first DEE ensures that for a given
configuration, a limited number of design variables, and a reduced set of constraints, a design results that has
a high probability of being feasible. With carefully chosen sizing methods, analysis tools, and optimization
algorithms this can swiftly result in an adequate input vector for the subsequent DEE.
Of all the non-traditional airplane designs, the blended wing body is one of the few concepts that is
currently considered by academia and industry as a possible successor to the TAW aircraft. The idea of
having the body generate part of the lifting forces, thereby reducing the overall wetted area of the airplane
dates back to the early 20
th
century when airplane designers such as Dunne, Lippisch, and the Horton brothers
invented several all-lifting-vehicle (ALV) configurations.
2
Prior to World-War II Northrop developed the X-
35 flying wing (FW). After the war the X-35s were retrofitted with turbojet engines to become the B-49
bomb er. Even though speed and altitude performance of the B-49 were good, the airplane lacked sufficient
range and was canceled in favor of the Convair B-36. It was speculated that the range inferiority was inherent
to the jet-powered FW configuration.
3
Due to this controversy, the flying-wing concept was abandoned for
almost 40 years. It revived in the early 1990s with the introduction of the Northrop B-2, demonstrating
the feasibility of a high-speed FW aircraft. It also resulted in renewed discussion about the aerodynamic
performance of FW aircraft compared to TAW aircraft. An independent examination by Torvik on the range
performance of high-bypass-ratio FW aircraft demonstrated that prior assumptions on aspect ratio and wing
thickness substantiating the conclusion that FW aircraft had inferior range performance were not correct.
4
In 1991 a paper by Torenbeek compared aircraft of various wing volume to total volume ratios (X). He
recognized that the low wing loading of an FW aircraft results in a reduced aerodynamic efficiency at FL350.
A higher lift-to-drag ratio can only be achieved if the airplane is flown at higher altitude. Because engine
thrust is linearly dependent on the ambient pressure in the stratosphere, the size of the engine is likely to
grow with a higher cruise altitude. Torenbeek confirms this and comes to the following conclusions based on
a simplified analysis:
5
• The highest aerodynamic efficiency for a baseline TAW aircraft with X = 0.18 occurs for T/p = 9m
2
and amounts to L/D = 20.5.
• For engines with an installed thrust of T /p = 9m
2
the FW aircraft (X = 0.9) does not perform better
than the TAW aircraft, while L/D decreases rapidly with decreased installed thrust.
• To reach its full p otential, the FW aircraft requires 45% more thrust to reach the maximum L/D = 27.5.
• To achieve this L/D, the FW aircraft needs to cruise at FL450.
These conclusions were based on a hypothetical airliner with a gross weight of 450 tons, a cruise Mach
number of 0.85 and a total useful volume of 2000m
3
. Furthermore, it was assumed that the drag polar
2 of 14
American Institute of Aeronautics and Astronautics
Downloaded by TECHNISCHE UNIVERSITEIT DELFT on February 26, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2012-1756
could be represented by a simple two-term quadratic equation. These results indicated that the FW aircraft
certainly had the potential to improve the aero dynamic efficiency of a new generation of aircraft, provided
that they would fly at higher altitudes.
In 1951 it was Burnelli who conceived a design that bears close resemblance to what is commonly known
as the blended wing body (BWB) aircraft (see Figure 2). Since the 1980s Russian studies at TsAGI have
investigated the BWB concept and the technologies that are required to making them viable.
6
Developments
at McDonnell Douglas (later Boeing) throughout the 1990s confirmed the findings by Torvik and Torenbeek
by predicting a 20% higher lift-to-drag ratio for an 800 passenger BWB airplane compared to a conventional
baseline aircraft with equal capacity and range.
7
Subsequent research by Boeing and NASA on the BWB-450
demonstrated to have 32% better fuel burn properties compared to the Airbus A380, using similar technology
levels.
8
Test flights to investigate the performance and handling qualities on scale model (X48-B) have been
under way since 2007, with overall satisfactory results. Flight tests of a modified BWB configuration with
twin vertical tails (X-48C) to shield engine noise were scheduled to take place in late 2011.
9
Figure 2. Photograph of a Mo del Of Burnelli’s Advanced Transport Design from 1951
10
Multi-disciplinary optimization (MDO) has been demonstrated to be of great help in improving baseline
BWB designs. Due to the tight interconnection between wing planform shape and cabin volume allocation,
the optimization is significantly more complicated than for a traditional TAW aircraft. Given the short-
coupled nature of the aircraft in combination with the absence of an empennage complicates the problem
even more by making the planform shape responsible for the aerodynamic performance, balance, stability, and
control of the airplane. To refine the design of the BWB-450 a multi-disciplinary design tool (WingMOD)
was used to size the aircraft for performance, balance, stability, and control.
11, 12
Starting point for the
optimization was an aft-swept configuration with winglets doubling as vertical tails. In Europe a collaborative
project on the multi-disciplinary optimization of the BWB relied on distributed analysis tools that were
connected to a single MMG.
13
Both projects demonstrated how the total weight of the BWB airplane
could successfully be reduced while satisfying all the constraints stemming from regulations and top-level
requirements.
To facilitate MDO a parametric description of the airplane geometry based on high-level primitives
(HLPs) can swiftly generate geometric models of many diverse aircraft configurations.
14
Combining and
resizing the HLPs allows the designer to generate airplane designs beyond the TAW, such as the BWB
configuration
13
(see Figure 3). The design can be subjected to various analysis tools (aerodynamic,
15
struc-
tural,
16
aeroelatic,
17
acoustic
18
) each based on a single representation of the geometry. The results of these
modules can subsequently be used in flight mechanics tools to investigate performance and handling quali-
ties.
19, 20
The interface between the analysis tools and the geometry module is formed by so-called capability
modules (CMs) that translate the geometry to tool-specific input files. The output from each of the analysis
tools can be used in an optimization process by comparing it to the constraints and evaluating the objective
function. The assembly of the HLP-based geometry module and the various CMs has been termed the
multi-model generator (MMG). Together with the analysis tools, the optimization module, and the initial
input vector it forms the design and engineering engine (DEE).
Even though it has been demonstrated that MDO can successfully be used to improve existing BWB
designs in various ways, implementation in the conceptual design has been limited to aerodynamic and
structural analysis.
21
At the same time, no conceptual design tools are available to quickly size various BWB
configurations. Finally, if the multi-fidelity design optimization paradigm of Figure 1 is to be successfully
implemented altering the BWB configuration should be enabled whenever a configuration proves to be
infeasible in any of the lower level DEEs. This paper details a first step in the generation of the initial DEE
3 of 14
American Institute of Aeronautics and Astronautics
Downloaded by TECHNISCHE UNIVERSITEIT DELFT on February 26, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2012-1756
Figure 3. The High-Level Primitives Build Up Approach
14
that allows for various BWB configurations to be analyzed with relatively low-fidelity tools, resulting in a
Class II design of the BWB. Engine positioning, fin positioning, wing sweep, and landing gear disposition are
included in the configuration design. The following sections detail the methods that have been implemented
in the tool and discuss some of the obtained results.
II. Methodology
A. Conceptual DEE Structure
The conceptual DEE consists of four modules (see Fig. 4). The initiator module generates an input vector
based on traditional preliminary sizing methods. Subsequently, the MMG constructs a geometric model of
the aircraft and generates a simple aerodynamic and structural model. It also sizes fuel tank and cabin. In
the analysis module, the aerodynamic forces are evaluated, the weight and center-of-gravity (CoG) travel are
calculated, and the airplane is trimmed. Stability and control derivatives are calculated, the rotation speed,
the minimum control speed, and the cruise range are determined. In the evaluator module the analysis
results are compared to set constraints. The optimization routine tries to satisfy all the constraints and
subsequently find the airplane geometry with the longest range. Whenever the constraints are not met
(“acceptable?”) or the optimization has not converged (“converged?”), the optimizer alters the input vector
and runs the lo op again.
0
0
W/S
W/S
x
x
f(x)
preliminary
sizing
Top Level
Requirements
Initial
input vector
MMG
Analyzer
Converger/
optimizer
1. Converged?
2. Acceptable?
1
2
no
no
yes
yes
Performance
estimates
Optimized
BWB
DEE
Figure 4. Structure of DEE for Conceptual Design of the BWB
4 of 14
American Institute of Aeronautics and Astronautics
Downloaded by TECHNISCHE UNIVERSITEIT DELFT on February 26, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2012-1756
B. Preliminary Sizing
During preliminary sizing the top-level requirements along with the relevant certification specifications are
used to estimate the maximum take-off weight, thrust and wing surface area. The wing loading and thrust
loading are sized in the initiator based on requirements on range, take-off field length, cruise speed, and cruise
altitude. Assumptions on maximum lift coefficients in the relevant configurations as well as statistical data
on characteristic weights are made based on values from literature on BWB aircraft. In Fig. 5 the outline
of the preliminary sizing module is shown. The output of the preliminary sizing along with configurational
choices on the number of and position of engines, the fin position, and the wing sweep (forward or aft) are
input to the subsequent module (MMG).
Top Level Requirements
Drag polars
Weight fracs,
L/D, SFC,
OEW vs. MTOW
Performance estimates
N , R , E, W ,
N
p
s , s , γ
TO land FAR
PAX
W , W , W
f
TO OE
T, S, b
0
Class I weight
estimation
Wing loading
thrust loading
Preliminary Sizing
e
Figure 5. Structure of Preliminary Sizing Module
C. Multi-Model Generator
The MMG receives input from the preliminary sizing module and the vector containing the initial design
parameters. This input vector is automatically scaled such that the wing span and wing area match those
calculated in the preliminary sizing. The MMG translates this information into a geometric model of the
outer shell of the aircraft. The BWB planform is translated into five individual wing trunks. Each trunk has
a root and tip airfoil that needs to be selected by the user. In addition, initial values for sweep, taper, span,
twist, relative thickness, and mean chord are set by default in the input vector. During the optimization,
these parameters are changed. For the vertical tail a single input parameter can be selected (tail height).
All other geometric parameters of the vertical tail have either been fixed or related to the tail height via a
fixed relation. This has been done to keep the number of design variables as low as possible. In total there
are 30 design variables (excluding airfoil shapes for each of the sections). An example of an input vector for
the planform shape is presented in Table 1.
Parameter
Sect 1
Trunk 1
Sect 2
Trunk 2
Sect 3
Trunk 3
Sect 4
Trunk 4
Sect 5
Trunk 5
Sect 6
Unit
Chord 38.5 37.9 33.2 22.5 12.7 4.2 m
Span 1.0 2.7 5.7 8.9 21.4 m
Twist 1.0 1.0 deg
Sweep 31 63 55 44 38 deg
Dihedral 0 0 1.0 2.0 5.0 deg
Thickness-to-chord ratio 0.16 0.16 0.16 0.16 0.12 0.12 -
Table 1. Example Input Vector for BWB Planform.
5 of 14
American Institute of Aeronautics and Astronautics
Downloaded by TECHNISCHE UNIVERSITEIT DELFT on February 26, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2012-1756
