Multi-disciplinary Design of Aircraft Fuselage Structures

Michel van Tooren 1 and Lars Krakers.2

Delft University of Technology, The Netherlands

The fuselage structure of civil passenger transport aircraft has mechanical duties
combined with the task to protect passengers against the cold and excessive noise in the
different flight phases. Current design practice for fuselage structures is a sequential
approach in which mechanical requirements are satisfied first after which the acoustical and
thermal requirements are satisfied in a corrective fashion. It can be questioned if this
approach will lead to an optimal weight design. Therefore a design study was started to
compare sequential versus concurrent multi-disciplinary design of a simplified fuselage
structure. A Design and Engineering Engine (DEE) has been built that is able to generate
(input for) mechanical, acoustical and thermal models of stiffened and unstiffened simplified
fuselage sections with optional floors subjected to mechanical, acoustical and thermal loads.
So far the DEE was used to perform simple what-if studies. The DEE is now extended to
optimize a stiffened cylinder for minimum weight, subjected to mechanical, acoustical and
thermal constraints. The acoustical models have been validated with tests in which the
predicted transmission loss is compared with the actual behavior of stiffened and unstiffened
cylinders. Optimization is done using a SQP algorithm applied to response surfaces obtained
with a DOE approach. The study does not show yet that using MDO in this specific design
problem can be beneficial to reduce weight.

Nomenclature
HFl = Floor height p = fuselage internal pressure
L = fuselage section length tc = Sandwich skin core thickness
M = fuselage bending moment tf = Sandwich skin facing thickness
Q = fuselage shear force tflc = Floor core thickness
R = fuselage radius, constraint tfllf = Floor facing thickness
dfr = Frame width tfr = Frame thickness
dstr = Stringer width tinc = Interior core thickness
hfr = Frame height tinf = Interior facing thickness
hstr = Stringer height tskin = Skin thickness
m = Nr of frames tstr = Stringer thickness
nbottom = Nr of stringers bottom
ntop = Nr of stringers top

I.Introduction

A IRCRAFT fuselages are subject to a wide envelope of mechanical loads. Flight and ground loads result in
shear, bending and torsion while the internal pressurization leads to an internal stress system that can be
described, in a highly simplified way, with the vessel equations. At the same time the fuselage is there to protect the
passengers against a hostile environment that can be characterized by noise levels, temperatures and radiation levels.
It is unclear if an MDO approach, in which multi-disciplinary design and sizing of the structures is done in an

1
Professor, Faculty of Aerospace Engineering, Kluyverweg 1, 2629 HS Delft, The Netherlands; MDO TC-Member.
2
PhD Student, Faculty of Aerospace Engineering, Kluyverweg 1, 2629 HS Delft, The Netherlands.

1
American Institute of Aeronautics and Astronautics
integral way, should be preferred (with respect to resulting fuselage weight) over a sequential approach in which
mechanical sizing is done first, after which the protection functions are incorporated with features like thermal
blankets and noise suppression techniques that do not contribute to the load carrying function of the fuselage.

A simplified fuselage is taken as design study object. The basic element is a stiffened cylinder, including a floor like
division of the cross section, subjected to mechanical loads, external noise fields and a low external temperature.
The impact of structural concept (sandwich shells versus different discretely stiffened shells) and of thermal and
acoustical insulation design solutions on the overall weight is studied using an MDO computational approach.

A Design and Engineering Engine (DEE)1 has been built and used as computational framework. The Multi-Model
Generator in this DEE is a Knowledge Based Engineering application that is able to generate (input for) mechanical,
acoustical and thermal models for subsequent analysis with a commercial finite element solver. The analyses are
used to build response surfaces with a DOE approach. A gradient based (SQP) or a genetic algorithm subsequently
explores the design space looking for minimum weight solutions. The most complex and uncertain part of the
computational effort is the acoustical analysis. Therefore it was decided to build and test two metal cylinders, one
stiffened and one unstiffened, for their response on a wide range of frequencies. The cylinders have been subjected
to external noise and the transmission loss has been measured using a scanning microphone on the inside of the
cylinder. The measurements have been used to verify and calibrate the calculations.

The acoustical properties of shell structures can be influenced in different ways. The type and placement of
stiffening elements has a large influence on vibration behavior. In addition, exchanging weight between skin and
stiffening elements can be used to influence the transmission losses. Besides optimizing current design concepts,
modern materials and active techniques can be used to influence the acoustical behavior of the fuselage. In this study
the application of visco-elastic materials and active materials (PZT actuators) is taken into account and the DEE is
able to generate models reflecting these alternative solutions.

It is shown that supporting multi-disciplinary design of complex products using MDO requires very powerful Multi-
Model Generators (MMG)1. A working MMG has been built, using Knowledge Based Engineering, that generates
mechanical, acoustical and thermal models for subsequent analysis in an FE-solver. Especially the creation of
acoustical models including active elements like PZT-actuators is a demanding task. It is shown that with the use of
KBE this problem can be effectively addressed. The principle of High Level Primitives and Capability Models,
ref.1, has been implemented to create parametric models that can automatically discretize the enclosed air in the
cylinder in the presence of stiffening elements and active elements like PZT actuators. The acoustic analysis is a
computationally intensive task. Having an acoustic analysis in an MDO problem requires the use of surrogate
models. Response surfaces have been successfully used in this case combined with a DOE approach. The study does
not show yet that using MDO in this specific design problem can be beneficial to reduce weight. However, the tools
developed and verified are such that a series of proper formulated optimization problems can now address this
question. This paper will show some first results.

II. Multi-disciplinary fuselage requirements

Aircraft fuselage design is guided by a large set of End section
multi-disciplinary requirements. In this paper we will Start section Q
use a reduced set of mechanical, acoustic and thermal M
requirements to size a sample, simplified, Airbus
p
A320 like fuselage. M

Mechanical Loads Q
The structural load case consists out of a bending
moment M, an internal pressure p and a shear load Q. p 55000 N/m2
Figure 2 illustrates the load case and defines the load M 4400000 Nm
levels. Q 600000 N
Noise requirements Figure 1. Simplified fuselage load case.
The two noise sources that are most important for
the design of acoustical insulation of a fuselage wall

2
American Institute of Aeronautics and Astronautics
 are the turbulent boundary layer noise and noise
from the engines. Both generate an overall noise
level that varies around 130 dB with the
presence of several peaks at the blade passage
frequencies (BPF) of the engines. The total
turbulent boundary layer noise level is relatively
constant along the fuselage. However, at the
front fuselage high frequencies are more present
and at the rear fuselage the lower frequencies
Normal speech are more present. The engines are responsible
for local effects. Especially in the engine
compressor, turbine or propeller plane the noise
levels are high.
The maximum acceptable noise level inside
a passenger cabin is currently around 75 dB and
decreasing due to the increasing attention on
passenger comfort. Since the human ear is less
sensitive for lower frequencies it is better to
express the allowable noise level in phons.
Figure 2. Human hearing sensibility. The figure shows how (Phons are defined as noise with equal loudness
the noise level in decibels can be expressed in phons as a as the noise in decibel at 1000 Hz,.e.g. noise of
functions of the frequency. 75 phons is perceived just as loud as noise of 75
dB(A) at 1000 Hz). Since it is difficult to
insulate low frequency noise, low frequency noise is an issue in sound insulation problems.
In general it can be stated that the sound transmission loss of the fuselage has to be such that the outside noise
(approximately 130 dB with BPF peaks) is reduced to the required weighted inside noise level of approximately 75
phons, i.e. the difference between the 130 dB line and the 75 phons line in Fig.2. The fuselage wall also has to be
able to damp noise peaks at certain frequencies.

Thermal requirements
For passenger comfort the temperature inside the fuselage should be around 20oC. The outside temperature can
range from -55 degrees (cruising altitude of 13 km) to +50oC. This gives a temperature difference (inside minus
outside temperature) of 75 or -30oC. This requirement implies that the heat flow caused by this temperature
difference may not exceed the capacity of the air conditioning system. In practice this means that the heat flow
through the fuselage wall should be minimized.

For this paper the mechanical loads, the outside noise levels and the outside temperature will be considered as
given facts. The improvements that can be achieved by silencing techniques for the engines and the boundary noise
will not be taken into account in the multidisciplinary analysis of the fuselage wall. Only the acoustical and thermal
insulation properties of the fuselage wall itself will be considered. Radiation issues are left out of the discussion.

III. The considered design options

In this study we consider an Airbus A-320 like fuselage section with a length of 8 meters, a radius of 1.98m and
one floor positioned at a height of 1.56m measured from the bottom of the fuselage. The fuselage is assumed to
consist of a stiffened or sandwich, frames, insulation blankets, interior panels and floor panels. The structural,
acoustic and thermal design options evaluated in this work are discussed below.

Structural design options

Two structural design options have been studied: a stiffened shell concept and a sandwich concept. The model of
the fuselage with the stiffened skin concept consists out of a skin stiffened with circumferential aluminium C-frames
and longitudinal aluminium z-stringers. The floor is constructed as a sandwich. Figure 3 illustrates the concept with
some detailed cross sections of the interior, floor panels and the C-frames and Z-stringers. A set of basic values for
the structural dimensions is chosen fictively and given in the table below. During the optimization the number of
design variables is kept limited by applying scaling factors to groups of dimensions, e.g. the stiffener area is varied

3
American Institute of Aeronautics and Astronautics
in the optimization instead of height, width and thickness separately. The model of the fuselage with sandwich skin
has frames but no stiffeners. Instead the skin consists of two facings and a core. The material properties used in the
this study are listed in Appendix A.

tflf
tinf tinc tinf tflc

tflf
Interior panel Floor panel
dfr
dstr
R

hstr hfr

Hfl
L dstr
tskin Z - Stringer C - Frame

Figure 3. The stiffened shell design option. The overall fuselage structure is shown on the left; details of
stiffeners, frames, interior and floor panels on the right.

Radius (R) 1.98 m Stringer height (hstr) 35 mm

Fuselage section length (L) 8.0 m Stringer width (dstr) 20 mm
Skin thickness (tskin) 1 mm Stringer thickness (tstr) 1 mm
Floor height (HFl) 1.56 m Interior facing thickness (tinf) 0.5 mm
Floor facing thickness (tfllf) 0.8 mm Interior core thickness (tinc) 4 mm
Floor core thickness (tflc) 8 mm Nr of stringers top (ntop) 8
Frame height (hfr) 120 mm Nr of stringers bottom (nbottom) 6
Frame width (dfr) 50 mm Nr of frames (m) 15
Frame thickness (tfr) 1.2 mm

Table 1. Dimensions used for the generalized fuselage model

tflf

tinf tinc tinf tflc

tflf
Interior panel Floor panel
dfr
R
tf
tc hfr

tf
Hfl

L Skin C - Frame

Figure 4. The sandwich shell design option. The overall fuselage structure is shown on the left; details of
sandwich skin, frame, interior and floor panels on the right.

4
American Institute of Aeronautics and Astronautics
 Radius (R) 1.98 m Frame height (hfr) 120 mm
Fuselage section length (L) 8.0 m Frame width (dfr) 50 mm
Floor height (HFl) 1.56 m Frame thickness (tfr) 1.2 mm
Skin facing thickness (tf) 0.6 mm Floor facing thickness (tfllf) 0.8 mm
Skin core thickness (tc) 10 mm Floor core thickness (tflc) 8 mm
Nr of frames (m) 3

Table 2. Basic dimensions used for the fuselage section based on the sandwich concept.

Acoustical insulation concepts

Acoustical insulation of the fuselage section is obtained with interior panels and insulation blankets. The
fuselage structure itself can either improve or deteriorate the insulation of the single, unstiffened reference skin,
depending on frequency and structural details (e.g. stiffening). In addition visco-elastic layers can be added to skin
and interior panels to increase the transmission loss. The interior panels are placed on the inside of the fuselage, on
top of the frames. The insulation blankets are assumed to be placed between the frames and have a thickness equal
to the frame height. Active noise reduction with the help of PZT sensors/actuators can be analyzed by the developed
tools but is not considered in the optimization discussed in this paper. The material properties related to the acoustic
insulation concepts under consideration are listed in appendix A.

Thermal insulation concepts

Thermal insulation of the fuselage section is achieved with insulation blankets. Also here the fuselage can
improve or deteriorate the thermal insulation of the single, unstiffened reference skin. The thickness of the insulation
blankets is identical to the height of the frames. The material properties related to the thermal insulation concepts
under consideration are listed in appendix A.

Material properties
The skin, frames and stringers are made of aluminium. The floor and interior panels are assumed to be sandwich
panels made of Fibrelam® produced by HEXCEL Composites. For this study Fibrelam® type 6100 Grade 1 is
chosen for the floor panels and Fibrelam® type 1100 Grade 2 for the interior panels. Fibrelam® type 6100 Grade 1
consists out of carbon phenolic facings and aramid phenolic honeycomb with a density of 139 kg/m3 and a cell size
of 1/8”. Fibrelam® type 1100 Grade 2 has an aramid phenolic honeycomb core with a density of 64 kg/m3 and a
cell size of 1/8” and glass phenolic facings. Both the floor and interior panels are fabricated with phenolic resin
because of its excellent low fire, smoke and toxic gas emission features as well as good corrosion and impact
resistance properties. Because it is difficult to find a complete set of material properties for insulation blankets used
in the Airbus A320, the material properties of the glass fiber blankets found in Ref.2 are used. All relevant material
properties of the different parts in the fuselage section model are given in Appendix A.

IV. The applied models

For each of the design options models have been selected or made to evaluate their properties in relation to the
requirements. Below the most important models are discussed briefly. For an extensive discussion of all models and
their fidelity see Ref.3.

Structural behavior
The stress distribution in the fuselage structure due to the external loading is obtained with an ABAQUS finite
element model. The resulting stress data is used to analytically check each component on strength and stability. The
structural FEM model only contains the structural parts of the fuselage section. The structural parts are the skin,
frames, stringers and the floor. The skin, floor and frame web are modeled in ABAQUS with shell elements ‘S4R’.
The floor is defined as a sandwich by their ABAQUS shell section definition.
The stringers are modeled with three dimensional beam elements ‘B31’. The beam section definition is used to
give the stringer beam elements the exact cross-sectional properties of the z-stringers with the ‘Arbitrary’ cross

5
American Institute of Aeronautics and Astronautics
section option. The frame flanges are modeled as beam elements with a circular cross section. An example of a
structural FEM model of the A320 like fuselage is given in Fig.6.A

The stress data obtained with the finite element analysis are processed to find the highest stresses per element
group. The element groups are linked to structural members (frames, stiffeners, skin) and subdivided with respect to
position in the cross section (crown, side, keel). The resulting stress levels are considered as the applied stresses.

The allowable stresses in the panel are linked to the following failure modes: skin buckling, stringer buckling,
allowable tensile and compressive stresses in the skin material, and allowable tensile and compressive stresses in the
stringer material. For the isotropic stiffened skin concept well-known local buckling and panel buckling formulas
are used. For the sandwich fuselage buckling formulas from Sullins3 are used.

Acoustic behavior
The sound transmission loss (TL) of the fuselage wall at low frequencies is determined by means of a FEM
analysis. The FEM model for acoustical analysis, Fig.6.B, consists of the complete fuselage section including the
skin, frames, stringers, floor, floor beams, insulation blankets and interior panels. Like for the thermal model the
skin insulation blankets and interior panels are modeled with solid brick elements. Also the air inside the fuselage is
included in the model. The air and insulation elements are of a different nature compared to the structural elements.
They are not capable in carrying any stresses. To enable ABAQUS to perform a coupled structural-acoustic analysis
these different elements have to be connected with each other with the ABAQUS ‘*TIE’ command. The ABAQUS
non-linear steady state dynamic analysis procedure is used to perform a frequency sweep.
The steady state dynamic analysis is a mode-based analysis. It means that the steady state dynamic analysis has
to be preceded with an eigenmode analysis. With the eigenmode analysis all eigenfrequencies of the frequency range
of interest are determined. Since within the turbulent boundary layer noise all frequencies are represented, all
eigenmodes of the fuselage structure should be excited. To ensure this an out of phase point load is used as
excitation force. During the steady state dynamic FEM analysis the acoustic pressure at chosen nodes (p) of the air is
recorded. TL is determined with:

TL=10log10(p/ppoint load)

Because of computational limitations the FE-analysis is not suitable for the high frequency range. For high
frequencies the model would require very small elements resulting in heavy models (in memory size) and extremely
long calculation times. Therefore analytical formulas are used from various literature sources2,5-24. An overview can
be found in Ref.3. The TL is calculated as the sum of the transmission losses caused by the different design options.
An overview of the model combinations can be seen in Table 3. The basic TL is the one of the skin (wall 1), which
can be either monolithic or sandwich. A visco-elastic layer can be added to the skin. An interior panel can be added
that acts as a second wall (wall 2). The interior panel can also have a visco-elastic layer. Wall 1 can have stiffeners

wall 1 visco 1 wall 2 visco 2 description

t nil nil nil TL of single skin (Fahy5)
t t nil nil TL of single skin with visco-elastic layer (Fahy5,
Derby&Ruzicka14 )
t nil t nil TL of double wall (Fahy5)
t nil t t TL of double wall with visco-elastic layer in the interior
panel (Fahy5, Derby&Ruzicka14)

Options Present nil

Wall 1 stiffeners add ∆TL for stiffeners (Fahy5, Von Venzke15) ∆TL = 0
Sandwich Replace single skin for sandwich skin (DIAB24) Single skin (Fahy5)
blanket add ∆TL for blanket (Beranek2) ∆TL = 0

Table 3. Parameters specifying which literature formulas are used to determine the TL.

in which case a correction to the TL is made. Finally an insulation blanket can be present which is also modeled
with a correction on the TL.

6
American Institute of Aeronautics and Astronautics
 The resulting TL is divided in three frequency ranges: 0-500, 500-5000, 5000-20000 Hz. The average TL in each
range is calculated and the three resulting values are used in the optimization. These values can also be translated in
phons and added to have a single constraint value. This is however not done in the current study.

Thermal behavior
The thermal behavior of the fuselage section is analyzed with an ABAQUS FE model, Fig.6C. This model
consists of half a fuselage including structure, insulation blankets and interior panels, all modeled with solid thermal
brick elements The thermal load case is defined as a constant incidence heat flux on the inside surface. The inside
surface is defined as the inside surface of the interior panel, the insulation blanket or the fuselage skin depending on
whether these components are included in the fuselage model. The constant incident thermal flux is specified as
incident surface energy per second.
The constant incident heat flux causes the inside surface of the fuselage to heat up. Through conduction the other
parts of the fuselage, which are in physical contact, will heat up. Because of the constant incident heat flux the
thermal model will reach a steady state when the temperature difference between the inside and outside fuselage
surface will remain constant. The type of analysis that is used is an uncoupled transient heat transfer analysis; no
interest is taken in the deformations caused
by the temperature distributions. Transient
frames
analysis is chosen because the temperature
distribution in the fuselage wall has to reach
Interior panel equilibrium between the incident energy and
the exiting energy over a period of time. The
difference between inside surface temperature
and the average outside surface temperature
at the steady state characterizes the thermal
insulation of the fuselage wall.

Practical generation of the models, the

stringers Multi-Model Generator (MMG)
Insulation
skin blankets During the optimization of the fuselage
section with respect to structural, acoustic and
thermal behavior, many different variants
Parameter true nil have to be evaluated. To support the
:top-floor? Top floor no top floor automatic generation of models a Knowledge
:bottom-floor? Bottom floor no bottom floor Based Engineering application has been built
:shell-skin? single skin sandwich skin in ICAD3. Based on parametric fuselage
:frames? frames no frames primitives with associated capability
:stringers? stringers no stringers modules1,3, a large range of different designs
:insulation? insulation blanket no insulation blanket can be generated by a so-called Multi-Model
:interior-panel? interior panel no interior panels Generator. An example of the main primitive
and its parameters and the associated
geometry is shown in Fig.5. The Multi-Model
Figure 5. The fuselage primitive defined in the Multi-Model Generator creates different models for the
Generator. structural, acoustical and thermal analysis,
Fig.6.
The MMG also supports the generation of non-cylindrical fuselage sections. Double bubble, triple bubble,
double floors etc can be easily modeled. Also double curved section can be specified. In that case also non-
continuous stiffeners can be handled. For this study, however, we stick to cylindrical fuselage sections.

The analytical models for the structural behavior and the high frequency acoustic behavior are programmed in
MatLab.

V. The analysis
The MDO process is structured using the Design Engineering Engine concept2. It can be considered a linkage of
numerical tools, each automating a part of the preliminary design process. The DEE enables the user to investigate

7
American Institute of Aeronautics and Astronautics
 Figure 6. The ABAQUS FE models created based on the input from the Multi-Model Generator.
A) Structural model; B) Acoustic model, C) Thermal Model
relations between chosen design variables and the resulting performance with respect to a set of user specified
requirements. The DEE used for the current study is shown in Fig.7. The process starts with specifying the
requirements for which a concept has to be initiated. The user can select the design options as discussed earlier.
With the Design of Experiments (DoE) method different sets of values are generated for the design
parameters/variables associated with the selected design options. The MMG generates for each of these sets the
input files for the corresponding models in ABAQUS or MatLab, by making use of the primitives, Fig.5 and 6.
The output generated by the MMG is used by the four different analysis modules of the DEE; the acoustic
module, which consists out of 3 sub-modules, the thermal module, the structural module and the weight module. The
last module determines the weight of the configuration, which will be used as one of the objective functions in the
optimization. Based on the evaluation of the design vectors specified by the DoE module, response surfaces can be
created for each design constraint and objective. The last step is to use these response surfaces for the optimization.
Both Genetic Algorithms (GA) and gradient based methods are used.

The Matlab program ‘L_GOPT’ developed by Lanzi25 is used for the DoE, the generation of response surfaces
and for the optimization. The Design of Experiments within the L_GOPT program is based on Latin hypercubes. In
this research the ‘maxmin’ criterion is used to select the sample points. The MaxMin criterion searches to maximize
the minimum distance between any pair of the sample points. The G_OPT program uses 4 types of Radial Basis
Function (RBF) approximations to determine the response surfaces (linear, cubical, thin plate spline and Gaussian).
The RBF approximation with the best fit is automatically selected by the program to define the response surface.
The G_OPT program uses the response surfaces to maximize or minimize a selected objective, taking into account
several constraints. Its Genetic Algorithm routine can also be used for the maximization or minimization of multi
objectives. This will result in a Pareto Set of solutions. Another option is the Sequential Quadratic Programming
routine, requiring a user specified a valid start vector.

Different analyses have been done on a verification case, a stiffened shell fuselage section and a sandwich shell
fuselage section. The verification case has been compared with experiments on different stiffened and unstiffened
cylinders3. Below the design variables, the constraints and the objectives are defined together with the design space
(Table 4) for an example analysis of an aluminum stiffened shell version of the fuselage section.

Five design variables that are chosen for the optimization of the stiffened skin fuselage concept:
1) The skin thickness
2) The number of frames
3) The number of stringers
4) The stringer factor
5) The interior panel core thickness.

The DEE is used to determine the performance w.r.t. the requirements. Five structural requirements are
considered:
1) The tensile stress criteria of the fuselage skin
2) The buckling criteria of the fuselage skin

8
American Institute of Aeronautics and Astronautics
 3) The tensile stress criteria of the stringers
4) The Euler buckling criteria of the stringers
5) The hoop stress criteria

Six different design cases have been analyzed:

1) Maximization of the temperature difference between the inside and outside surface of the fuselage wall
2) Maximization of the overall sound transmission loss in frequency range 1 [0-500Hz]
3) Maximization of the overall sound transmission loss in frequency range 2 [500-5000Hz]
4) Maximization of the overall sound transmission loss in frequency range 3 [5000-20000Hz]
5) Minimization of the structural weight of the fuselage section
6) Minimization of the total weight of the fuselage section including the structure, blanket and interior panels.

Customer Design of
requirements experiments

Concept generator ICAD input

& Initiator parameters

ICAD model
generator

ICAD FEM
model generator

Acoustic Thermal Structural Weight

ABAQUS acoustic Active noise ABAQUS thermal ABAQUS structural
LF FEM Analysis control tool: FEM Analysis FEM Analysis
Matlab
acoustic Python output ABAQUS
Python output Python output ICAD
HF Python
extractor MATLAB extractor extractor weights
analysis
TNO algorithm
Matlab Matlab Matlab
Postprocessor Postprocessor Postprocessor

Evaluator Response
Surfaces

Final Matlab
Configuration Optimiser

Figure 7. The Design Engineering Engine used for the Fuselage MDO problem

The analysis has been done in two steps. First response surfaces were built for the complete design space and a
GA based optimization is performed. Second, the response surfaces were refined by doing a second DoE loop
around the optimum found in the first step. These new surfaces were used for an SQP based optimization.

Skin thickness 1 – 2 mm
Number of frames 8, 9, 10, 11, 12, 14, 16, 18, 20, 22, 25
Number of stringers 20, 30, 40, 50, 60, 70, 80, 90, 100
Area of the stringer cross section 0.5 – 1.5
Thickness of the interior panel core 4 – 40 mm

Table 4. Definition of the design space.

9
American Institute of Aeronautics and Astronautics
 VI. Results and Discussion
The results for the sample design cases as defined in the previous section can be presented in the form of tables,
response surfaces and directly in the optimum result for a specific optimization. In the tables and figure below some
results are presented. Table 5 relates to Design Cases 5 and 6 and lists the results for the first optimization using the
GA algorithm; Table 6 relates to the GA results for Design Cases 1-4. The response surface, Fig.8, shows the
relation between skin thickness, stringer factor (cross sectional area relative to the basic solution described earlier)
and the TL at high frequencies.
Objective function--> Wtot Wtot Wstruc Wstruc
G_opt program settings
Number of members for
each generation 100 100 100 100
Number of generations 200 200 200 200
Cross-over probability 0.75 0.85 0.75 0.85
Mutation probability 0.15 0.15 0.15 0.15
Design variables
SkinThickness (mm) 1.5294 1.5078 1.5167 1.5255
NrFrames 13 11 12 13
NrStringers 100 100 100 100
StringerFactor 0.5071 0.5753 0.5388 0.5071
InteriorCoreThickness (mm) 4 4 4 4
Final performances:
Rhoop 0.8685 0.9103 0.8897 0.8697
Rskintensile 0.9466 0.9247 0.9370 0.9483
Rstringertensile 0.8042 0.7788 0.7928 0.8058
Rskinbuckling 0.9897 0.9929 1.0002 0.9963
Rstringerbuckling 0.9963 1.0019 0.9930 0.9952
DeltaT (oC) 108.90 113.59 111.45 108.79
TL1 (dB (A)) 38.99 39.77 39.37 38.98
TL2 (dB (A)) 79.03 77.62 78.33 79.07
TL3 (dB (A)) 108.23 107.40 107.84 108.21
Weighttot (kg) 1149.15 1151.95 1149.44 1148.01
Weightstruc (kg) 794.48 795.81 793.99 793.37

Table 5.Results of the GA Optimization procedure for Case 5, Minimum Total Weight, and Case 6,
Minimum Structural Weight. Two different settings of the GA routine have been used and are shown to give an
impression of the sensitivity of the results with respect to the GA settings.

The results so far show that all the

components of the DEE are working and
that the optimization can take place in two
steps, each fully automated, using the GA
for the first step and SQP for the final
optimization step. So far only optimizations
have been done using different, single,
objective functions and using the limits on
the design variables as bounds for the
design space. From Table 5 and 6 it
becomes clear that optimization for the
different disciplines within the limits on the
design variables gives substantially
different results. The next step therefore
will be to define clear constraints on the
acoustic and thermal properties and
Figure 8. Response surface showing relation between skin optimize for minimum total weight. This,
thickness, stringer factor and TL at high frequencies (5000-
20000 Hz) 10
American Institute of Aeronautics and Astronautics
however, has not been done yet and therefore a conclusion on the difference in final design configuration between
sequential design of fuselage sections and an MDO approach cannot be given.

Objectivefuntion --> Delta T TL 1 TL 2 TL 3

G_opt program settings
Number of members for
each generation 100 100 100 100
Number of generations 200 200 200 200
Cross-over probability 0.75 0.75 0.75 0.75
Mutation probability 0.15 0.15 0.15 0.15
Design variables
SkinThickness (mm) 2.0000 1.9922 2.0000 1.8471
NrFrames 8 8 12 20
NrStringers 80 100 90 80
StringerFactor 1.2765 1.0765 0.6137 0.8333
InteriorCoreThickness (mm) 4 7 5 6
Final performances:
Rhoop 0.7440 0.7671 0.7246 0.7199
Rskintensile 0.5891 0.5728 0.7387 0.7686
Rstringertensile 0.4835 0.4672 0.6269 0.6502
Rskinbuckling 0.3529 0.2956 0.6008 0.8459
Rstringerbuckling 0.4603 0.5130 0.9958 0.1609
DeltaT (oC) 127.46 122.75 115.40 97.49
TL 1 (dB (A)) 39.59 41.62 38.99 35.60
TL 2 (dB (A)) 82.20 81.35 87.58 76.21
TL 3 (dB (A)) 119.02 120.25 117.31 123.49
Weighttot (kg) 1387.23 1415.21 1283.52 1347.53
Weightstruc (kg) 1032.41 1044.35 918.22 972.90

Table 6 .Results of the GA Optimization procedure for Case 1 to 4, Maximization temperature diference
over the fuselage wall and maximization of TL over the three frequency bands.

VII. Conclusions and Recommendations

A DEE has been built for the Multi-disciplinary Design Optimization of fuselage sections, taking into account
structural, thermal and acoustical requirements. Different design options for the structural design, the acoustical
insulation and the thermal insulation have been studied. The DEE is equipped with a powerful Multi-Model
Generator, a KBE-based generative model that can deliver the required (input for) structural, acoustical and thermal
models. The two step optimization approach, using a GA to identify the area of interest and the SQP algorithm to
find the optimum within this area seems to be robust. Using different objective functions it was shown that the
design of a fuselage section is sensitive to requirements from different disciplines.
Next step will be to use the DEE to show the difference in resulting fuselage section weight between a sequential
and a MDO approach.

11
American Institute of Aeronautics and Astronautics
 Appendix A

Properties of the materials used in the study

Aluminium: (skin, frames and stringers) Glass phenolic: (Interior panel facings)
Modulus of elasticity 68.9 GPa Modulus of elasticity 20 GPa
3 3
Density 2710 kg/m Density 2550 kg/m
Poisson ratio 0.3 Poisson ratio 0.3
Thermal conductivity 140 W/m°C Thermal conductivity 0.24 W/m°C
Specific heat 900 J/kg°C Specific heat 1110 J/kg°C
Allowable material stress 210 MPa
Allowable hoop stress 85 MPa

Carbon phenolic: (Floor facings) Air:

Modulus of elasticity 18 GPa Bulk modulus 144120 MPa
3 3
Density 1800 kg/m Density 1.225 kg/m
Poisson ratio 0.3
Thermal conductivity 1 W/m°C
Specific heat 600 J/kg°C
Allowable material stress 150 MPa

Fibrelam® 6100 Aramid phenolic honeycomb Fibrelam® 1100 Aramid phenolic honeycomb
HRH-10-1/8-9.0 (Floor core) HRH-10-1/8-4.0 (Interior core)
Modulus of elasticity 620 MPa Modulus of elasticity 193 MPa
Shear modulus L-direction 120 MPa Shear modulus L-direction 59 MPa
Shear modulus W-direction 76 MPa Shear modulus W-direction 32 MPa
3 3
Density 139 kg/m Density 64 kg/m
Thermal conductivity 0.0675 W/m°C Thermal conductivity 0.0675 W/m°C
Specific heat 1300 J/kg°C Specific heat 1300 J/kg°C

Glass fibre blanket: (Insulation blankets)

2
Bulk modulus 118300 N/m
3
Density 9.6 kg/m
Thermal conductivity 0.036 W/m°C
Specific heat 1005 J/kg°C
Volumetric drag 1

Frequency dependent properties:

Frequency [Hz] 20 28 40 56 80 116 160 224 320 447 640
Sound attenuation [dB] 2 3 4 6 8 10 15 23 35 55 90
Blanket impedance 1190 1189 1188 1186 1184 1181 1176 1170 1130 1080 1000
Blanket wave length 2.25 2.00 1.75 1.5 1.25 1.03 0.81 0.60 0.45 0.33 0.024
Frequnecy [Hz] 891 1280 1778 2560 3548 5120 7079 10240 14125 20480
Sound attenuation [dB] 140 210 285 350 400 450 490 520 545 560
Blanket impedance 900 800 710 630 560 520 510 506 505 504
Blanket wave length 0.17 0.13 0.10 0.08 0.06 0.05 0.045 0.043 0.041 0.039

Acknowledgments
The authors thank TNO-TPD and NIMR for their financial support of this study.

References
1
Rocca, G La, & Tooren, MJL van, “Development of Design and Engineering Engines to support multidisciplinary design
and analysis of aircraft.” Proceedings Design Research in the Netherlands, 2005.
2
Beranek, L.L., Noise and Vibration control, Institute of noise control engineering, Washington DC, 1988.
3
Krakers, L.A., Multi-disciplinary design optimization of aircraft fuselage structures, PhD Thesis, to be published, 2007.
4Sullins, R.T., Smith, G.W., Spier, E.E., Manual for structural stability analysis of sandwich plates and shells, NASA report,
CR-1457, Langley, 1969
5
Fahy, F., Sound and structural vibration, radiation, transmission and response, Academic Press, London, 1989.
6
Smith, B.J., Peters, R.J., Owen, S., Acoustics and noise control, Edinburgh: Longman, 1996.

12
American Institute of Aeronautics and Astronautics
 7
Demmenie, E.A.F.A., Transmissieverlies van rompwandpanelen, NLR TP 94476 L (in Dutch), Amsterdam, 1994.
8
Leissa, A.W., Vibration of shells, NASA SP-288, Washington, 1973.
9
Heckl, M., “Vibrations of point-driven cylindrical shells”, Journal of the acoustical society of America, Vol 34, nr 10, 1962.
10
Willis, C.M., Daniels, E.F., Experimental study of noise reduction for an unstiffened cylindrical model of an airplane
fuselage, NASA TP-1964, Langley Virginia, 1981.
11
Morse, P.M., Vibration and Sound, McGraw-Hill, New York, 1948.
12
Cremer, L., Heckl, M., Ungar, E.E., Structure born sound, Springer Verlag, Berlin, 1973.
13
Kerwin, E.M., “Damping of flexural waves by a constrained viscoelastic layer”, Journal of the acoustical society of
America, Vol 31, nr 7, 1959.
14
Derby, T.F., Ruzicka, J.E., Loss factor and resonant frequency of viscoelastic shear-damped structural composites, NASA
CR-1269, Washington DC., 1969.
15
Venzke, G., von, Dammig, P., Fischer, H.W., “Der Einfluß von Versteifungen auf die schallabstrahlung und
Schalldammung von Metallwanden“, Acoustica Vol 29, nr 1, 1973.
16
Kurtze, G., Watters, B.G., “New wall design for high transmission loss or high damping”, The journal of the acoustical
society of America, vol 31, nr 6, 1959.
17
Ford, R.D., Lord, P., Walker, A.W., “Sound transmission through sandwich constructions”, Journal of sound & vibrations,
vol 5, nr 1, 1967.
18
Smolenski, C.P., Krokosky, E.M., “Dilatation-mode sound transmission in sandwich panels”, The journal of the acoustical
society of America, vol 54, 1973.
19
Dym, C.L., Lang, M.A., “Transmission of sound through sandwich panels”, The journal of the acoustical society of
America, vol 56, nr 5, 1974
20
Dym, C.L., Lang, M.A., “Optimal acoustic design of sandwich panels”, The journal of the acoustical society of America,
vol 57, nr 6, 1975.
21
Dym, C.L., Ventres, S.C., Lang, M.A., Transmission of sound through sandwich panels: A reconsideration, The journal of
the acoustical society of America, Vol 59, nr 2, 1976
22
Dym, C.L., Lang, D.C., Transmission los of damped asymmetric sandwich panels with orthotropic cores, Journal of sound
and vibration, vol 88, nr 3, 1983.
23
Oude Nijhuis, M., Analysis tools for the design of active structural acoustic control systems, PhD thesis, University of
Twente, 2003.
24
DIAB Technical Bulletin, “Acoustic behavior of sandwich panels”, DIAB technical services, DIABgroup
25
Lanzi, L., Optimisation of composite stiffened panels under post buckling constraints, PhD thesis, Politecnico di Milano,
2004.

13
American Institute of Aeronautics and Astronautics