46th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics & Materials Conference AIAA 2005-2205

18 - 21 April 2005, Austin, Texas

Aircraft Design Support using Knowledge Engineering

and Optimisation Techniques

M.J.L. van Tooren∗, M. Nawijn∗, J.P.T.J. Berends∗, E.J. Schut∗

Nomenclature
A Area, m2
b Stiffener pitch, mm
Dij ij-th element of D matrix, Nmm
E Young’s modulus, N/mm2
h Stiffener height, mm
I Second moment of inertia, mm4
L Rib distance, mm
LB Lower design variable boundary
n0 Number of 0o layers in the skin
n45 Number of ±45o layers in the skin
ns0 Number of 0o layers in the stiffener
ns45 Number of ±45o layers in the stiffener
t Ply thickness, mm
UB Upper design variable boundary
W Weight/mass, kg
ρ Density, kg/dm3
σ Normal stress, N/mm2
X, Y, S Material failure stress in the material principal axis system, N/mm2
τ Shear stress, N/mm2

Multi-disciplinary optimisation of aircraft is normally restricted to a solution domain

defined by a selection of design variables. Optimisation theory however makes a distinc-
tion between design variables and design parameters. For aircraft design problems, vari-
ables specify limited differences within an aircraft configuration while parameters relate
to complex variations within a configuration and inter-type differences, i.e. differences in
configuration. During an optimisation, parameters are normally fixed and the optimisation
is limited to finding a combination of values for the design variables that will minimise or
maximise an objective function like weight or range. The mathematics required to optimise
at a higher level and support the choice between different concepts are not available nor
are product models that allow variation between configurations during the optimisation
process. In this paper the latter problem is addressed and the use of Knowledge Engi-
neering for parametric modelling of aircraft is discussed. It will be shown that a proper
combination of object oriented programming, rule based instantiation of objects and a ge-
ometry engine allows parametric modelling in the optimisation sense. The principle and
∗ Delft University of Technology, Systems Integration of Aircraft, Kluyverweg 1, 2629 HS Delft (the Netherlands),

www.sia.lr.tudelft.nl

1 of 16

American Institute of Aeronautics and Astronautics

Copyright © 2005 by M. van Tooren. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
 implementation of High Level Primitives (HLPs), i.e. functional building blocks, in a so-
called Multi-Model Generator (MMG) is shown to be a proper approach to the problem
of parametric modelling of complex products. It will also be shown how these parametric
models can be used and initialised in so-called Design and Engineering Engines (DEEs).
A DEE facilitates initiation of design parameters and variables, instantiates HLPs and
creates Multiple Models to support the required multi-disciplinary view on the system as
required in multi-disciplinary analysis and optimisation. The DEE offers a framework for
design decisions in the conceptual design phase. An example of a DEE implementation is
shown. The Initiator component of the DEE estimates the starting values for parameters
and variables and is an optimiser by itself. An example is given for the initiator component
in a sample DEE for composite aircraft tails.

I. Introduction
In the last decade several new aircraft configurations have been suggested for civil transport aircraft. The
blended wing body, box wing and joint wing aircraft1 are some illustrative examples. However, none of these
new concepts have made it to an application stage so far. The financial risks involved with deviations from
the so-called Kansas City type of aircraft5 are high and prohibitive for exciting experiments. In addition some
of the ’oldies’ in the field of conceptual aircraft design claim that the dominant configuration is dominant for
many good technical reasons and none of the so-called innovative configurations offers sufficient advantages
or even advantages at all that would justify its application to civil transport aircraft. Their right or wrong
will never be proven based on real experience so we will have to obtain virtual experience to verify their
statements.
This virtual experience should be obtained through multi-disciplinary design studies of the innovative
configurations, including parameter studies, what-if studies and trade-offs, to find the required proof. Some
experience with this approach was gained in the European MOB project2,4 on Multi-Disciplinary Optimisa-
tion of Blended Wing Body Aircraft. The project, discussed in more detail in the next section, showed that
for automated support of aircraft conceptual and preliminary design a number of issues needs to be resolved.
A selection of these open issues and potential solutions will be discussed in the other sections of this paper:
• the paradigm of Design and Engineering Engines (DEEs),
• parametric modelling based on KBE principles that allow efficient generation of design configurations,
• software frameworks that support the development of DEEs,

• the use of optimisation techniques and rule based reasoning for the initiation of the parametric models
for use in a DEE.

II. Experience obtained from the European MOB project

The objective of the European MOB project was to develop an automated, multi-site and multi-disciplinary
design process, that could be used for aircraft design and optimisation. A conceptual design of a Blended
Wing Body type of aircraft was chosen to validate the design process.
Optimisation theory makes a distinction between design variables and design parameters. For aircraft
design problems, variables specify differences within an aircraft configuration while parameters relate
to inter-type differences normally addressed as differences in configuration. During an optimisation,
parameters are normally fixed and the optimisation is limited to finding a combination of values for the
design variables that will minimise or maximise a certain objective function like weight or range for the
specific configuration. The MOB project showed that parametric modelling of aircraft is not possible with
traditional computer aided design (CAD) tools. A solution to this problem was to use Knowledge Based

2 of 16

American Institute of Aeronautics and Astronautics

Engineering (KBE) tools that allow this parametric modelling in the optimisation sense. A typical example
of a parametric change (change in configuration) was the addition of fins and canards.
The MOB study lead to the definition and implementation of a so-called Computational Design Engine
(CDE) Fig.1. This CDE was a collection of modelling and analysis tools connected through the internet
that allowed the multi-disciplinary optimisation of a freighter version of a Blended Wing Body aircraft2, .4

Figure 1. The MOB Computational Design Engine concept

One of the most important components of the CDE was the geometry module. The name geometry
module is a bit misleading since this module was actually a Multi-Model Generator (MMG). The main
characteristic of the MMG is that it can extract multiple views on the product model that resides inside the
MMG. The geometry is just one of the views, and non-geometric data can be extracted from the product
model as well. The functionality of the MMG is illustrated in Fig. 2.
A specific methodology was developed based on High-Level Primitives (HLPs)4,7 , which are fully para-
metric models of aircraft functional building blocks. These HLPs are object-oriented modules that include
rule bases for context sensitive instantiation. With these blocks a wide variety of aircraft can be modeled.
The HLP approach also made it possible to arrive at a true multiple-view on the design.
The project showed some important gaps in the current engineering tools and methodology. Especially
multiple view product modelling, process control of multi-site, multi-disciplinary design projects, equivalent
modelling6 and design information exchange need major improvements. Some of these topics will be discussed
in the following sections.

3 of 16

American Institute of Aeronautics and Astronautics

Figure 2. Role of the ICAD Multi-Model Generator inside the MOB Computational Design Engine. The
elements of the MMG are shown together with the models produced by this MMG

III. The paradigm of the Design and Engineering Engine

A Design and Engineering Engine (DEE) is defined as an advanced design environment, where the design
process of complex products can be supported and accelerated through the automation of non-creative and
repetitive design activities. A DEE is typically multi-site (the components of the DEE are distributed over
different physical locations) and multi-disciplinary.
In practice, a DEE consists of a collection of commercial of-the-shelf (COTS) components connected by
a framework. This concept is illustrated in Fig.3. Important elements of the DEE are process control and
initiators. These elements will be discussed now in more depth. The other components of the DEE like the
evaluator and the optimiser will not be considered here.
Process control is required to start and control each process in the loop, supply it with the proper input
and direct its output to the proper location. Flexibility requires that this process is platform independent
and can function in a constellation of computers, operating systems and tools.
In the DEE the initiator generates the initial values for the parameters and the variables inside the HLPs.
An example of an initiator is the process that calculates initial values for the parameters that define the
instantiation of structural elements in a HLP, e.g. a wing trunk containing spars, ribs and skins. If the HLPs
are used to feed a Finite Element (FE) analysis, it is required that the FE-tool is fed with proper input
data. During the conceptual design it is not desirable to have too much detail in the FE-model so equivalent
models6 should be used. For a structural design tool, the initiator must be able to initiate the structural
parts based on a structural concept selected by the designer. For example, if the designer specifies a blade
stiffened composite shell solution for the skin, the initiator must be able to estimate a first set of feasible
values for the design variables describing this structural concept. The initiator that will be further discussed
in section VI estimates a first skin lay-up, stiffener pitch, stiffener lay-up and dimensions based on a number

4 of 16

American Institute of Aeronautics and Astronautics

of a priori design decisions.

IV. Parametric modelling based on KBE principles

KBE based parametric modelling is required to capture commonality in a variety of configurations, Fig.4,
or to capture variation within a configuration, Fig.5. The examples shown in these figures have been created
in the KBE package ICAD. Both cases show the type of variation that can be controlled with parameters.
Within each configuration, determined by a parameter set, variation of design variables is possible. These
variables can address sweep angles, span, dihedral etc.
The HLPs used for the configurations shown are the wing trunk, the fuselage, the connection elements
and the engines4,3 . These HLPs and their use are illustrated in Fig.6. The wing trunk HLP can be used as
building block for wings, tails, control surfaces, fins, winglets and any other wing-like aircraft component.
Some possible configurations within a BWB topology are shown also in Fig.6.
The HLPs are reusable pieces of object-oriented code (modules) that are initiated based on rules and
controls. The parameters and variables inside the HLPs are addressable by basically any program and
therefore these product models can be used in analysis and optimisation loops. The concept of the HLPs
is implemented using Knowledge Engineering (KE, a term preferred by the author over Knowledge Based
Engineering) instead of traditional CAD. KE allows the definition of product models based on product
structure and rule bases for the selection and instantiation of components within that structure. When
properly used the resulting KE-model allows extraction of different geometrical and non-geometrical views
on the product after instantiation of the model with a specific set of parameter and variable values. The
objects that are responsible for this are called capability modules and are mixed-in in the HLP.
The parametric models become useful only when they can be integrated in a framework that takes care
of the initiation of the parameters and variables of the HLPs and that takes care of the coupling of the model

Figure 3. The Design and Engineering Engine (DEE) concept

5 of 16

American Institute of Aeronautics and Astronautics

Figure 4. Commonality between different aircraft configurations captured with KBE based Parametric Mod-
elling

Figure 5. Variation within an aircraft type captured with KBE based Parametric Modelling

Figure 6. The basic HLPs to define different aircraft configurations. Illustrated is how the wing trunk HLP
can be used to define a family of aircraft. The wing trunk HLPs are interconnected with the connection
elements. Note that all aircraft shown are made with the same number of wing trunks and with wing trunks
and connection elements only, so with a true parametric model

to analysis tools like Finite Element Modelling (FEM) and Computational Fluid Dynamics (CFD) to derive
design properties. This is explained further in the next section.

6 of 16

American Institute of Aeronautics and Astronautics

 V. Software frameworks for Design and Engineering Engines
Traditional, MDO frameworks focus on top-down execution of a string of tools. This can be seen as a
data-push scenario in which output data of a process is pushed into a consecutive process. An engineer
defines the problem in a Graphical User Interface (GUI) and actively chooses the string of tools to be executed
to solve the problem. Data communication between the tools and the execution of the consecutive tools is
managed from a central process. The main problems with this type of top-down execution are flexibility and
re-execution when faults in a subprocess emerge. As learnt from MOB, re-execution after the error solving
can be time-consuming and should not be necessary as tools in the preceding part of the execution string
already produced acceptable results.
These traditional frameworks suffer from inflexibility because they rely on strict execution schemes often
based on file names and locations. They are not flexible in execution nor when inputs or outputs change
in the course of time. For a DEE in which parametric modelling and knowledge engineering are primary
technologies this approach is insufficient.
In order to solve these problems, a data-pull scenario is suggested. Implemented with autonomous helper
programs, called agents, which communicate with each other in order to exchange data to their discipline
tools. At least four functions of such a framework can be identified:

• Resource management:
Dynamic management of entry, registration and exit of tools in the DEE
• Resource interfacing:
supply of a standard for resource wrapping and resource communication
• Process executions support:
starting and stopping of processes in a multi-thread fashion based on DEE-components service requests
• Information flow control:
supplying the DEE with a controlled data space.

The first two functions have a direct link with resource planning. The last two functions allow the
framework of agents to fulfil their basic transformation process.
It has to be noted that terminology comes from human group co-operation and project management. A
collection of agents and their discipline tools can be seen as a multi-disciplinary group or design team in
which several functions and specialisations are distributed over the group members. These group members
will combine their individual capacities to perform the task. In literature on group processes and social
behaviour this is called indirect co-operation.

A. Levels of aggregation in the DEE framework (DEEF)

At least three levels of aggregation within the framework can be identified:

• Disciplinary tools: Responsible for the basic transformation process of a discipline.

• Helper Agents: Responsible for the earlier mentioned functions of the DEE.
• Design and Engineering Engines: A collection of agents that are needed for a single Design and
Engineering Engine in order to solve a technical problem.

These levels of aggregation are presented in Fig.7. For each aggregation level an actor can be identified
as well. A Specialist is responsible for the development and enhancement of the discipline tool. He has
in depth knowledge of his or her field of expertise. This actor should make his or her tool able to handle
different instantiations of a parametric product view without any user intervention. Development of this

7 of 16

American Institute of Aeronautics and Astronautics

Figure 7. Three levels of aggregation (disciplinary tool, helper agent, DEE) and four actors (Specialist,
Integrator, Operator, Maintainer) within a collection of agents can be identified.

discipline tool can be done outside the DEEF. When convinced of a correct functioning tool, the tool can
be committed to the DEEFa .
An Integrator is responsible for development of the agent framework. The role of this actor is to
integrate and communicate with Specialists and come to interface agreements. Simple transformation of
dataformats can be placed in an input or output filter helping the specialist tool. Specialists are treated as
users of the framework without in-depth knowledge of the implementation.
An Operator is responsible for solving technical problems by operation of the DEE. This actor is using
the discipline tools arranged in a DEE as a user.
A Maintainer is responsible for operation of the DEEF. This role is facility oriented. This actor is
responsible for correct operation of the specialist tools, network hardware and operating systems on which
the DEEF is implemented.

B. Basic discipline tool transformation

The basic transformation process of the disciplinary tools and their agents can be seen as a steady state
model8 as presented in Fig.8.
The steady state model consist of several groups of functions, called zones. The input, output and
control zones in the model are implemented in the helper agent. The functions in the transformation
zone is implemented by the Specialist in the discipline tool. External and shared libraries are provided to
a Source Configuration Management (SCM) can help significantly in the ongoing development of a discipline tool, being

able to separate and later merge development code and stable production code. SCM can be implemented using a centralised
network accessible repository system like Subversion (SVN) or Concurrent Versions System (CVS).

8 of 16

American Institute of Aeronautics and Astronautics

 Figure 8. Steady State model by In ’t Veld.8

the discipline tool and are controlled (kept up to date) by the agent.
The incoming datastream is coded and labelled as belonging to a specific DEE, model instantiation or
optimisation run number. Each data object is mapped to an input requirement. Input filtering of the data
is done to prevent insane data entering the discipline tool. An input buffer is used to store temporarely
incoming data. Sane data is always accepted also when the discipline tool is occupied with another dataset.
The main transformation function is provided by the Discipline tool. It can rely on possible shared
libraries which are kept up to date across the DEEF by the agent. Output data is scanned for sanity in
the output filter and possible emerged errors are captured. The processed data is stored in the output
buffer and this buffer is the key to a drastic reduction in re-execution time when errors occur. Errors are
redistributed through the framework and when intervention is required, fed to the user via a web-application.
Outgoing data requested by another agent is finally made available by packing and labelling it for retrieval
(In the model it is known as the decoding function).

C. Inter-agent communication
The inter-agent communication protocol is based on conventional IP. All agents are equipped with a built-
in webserver to present status information to a user interface and to other agents. The object oriented
source code is identical for all agents. As discipline tools each have their own preferable architecture, the
agents encompassing the tools will have to run on that same architecture as well. By using cross-platform
methodologies in the code this is achieved.
Each agent is capable of performing any of the four functions defined in the start of this section. Because
not all functions need to be fulfilled in every agent, seniority in the framework determines which agent
performs these so called master functions. When this master agent becomes unavailable other agents will
rediscuss seniority to take over its position. This and all communication between the agents is direct in a
truly peer-to-peer fashion.
Agents themselves have no knowledge of the whole DEE structure. They only have knowledge of which

9 of 16

American Institute of Aeronautics and Astronautics

types of data the encompassing discipline tool depends on in order function properly. For example: A
structures discipline tool, being able to perform a FE calculation, depends on loads from an initiator tool
and a structural model from a Multi Model Generator.
The basic workflow between agents is as follows. An agent is queried for its output, either by another
agent or a by user interaction via a webinterface. First, the agent determines if the requested data is available
in its output buffer or can be generated from an already available input. If this is not the case, the agent
knows its direct generic dependency and is querying the framework who can provide this specific input data.
Agents that can deliver will contact the requesting agent directly and discussion start who is able to provide
the input data in timely fashion.
A prototype of the framework with agents in under development.

VI. The composite panel initiator

The wing trunk HLP allows the selection of a structural concept for its instantiation. The skins of the
wing trunk, for example, can be selected to be made of composites. This section will discuss the initiation
of parameters and variables of such composite skins in the wing trunk using optimisation techniques.

A. Initiator Optimisation Problem Definition

In general composite panels can be discretely stiffened or be of the sandwich type. Here we will discuss only
discrete stiffened panels and more specific, only blade stiffened panels. The method however, is generic and
can be easily applied to other stiffener types and sandwich panels.
The parameters (fixed during the optimisation which is, in the current context, the initial sizing) for the
blade-stiffened panels are the material types for stiffeners and skin. More precise, it is assumed that both
stiffeners and skin are built up from fibre reinforced materials with layers of fibres in the ±45o directions, and
layers of fibres in the 0o direction (parallel to the stiffener direction). Stacking sequence is also predefined:
the 0o fibres are assumed to be in the centre of the laminates. The lay-up of stiffeners and skin is not assumed
to be identical.
The optimisation problem is stated as follows:

minimize f (x)
subject to h(x) = 0 (1)
g(x) ≤ 0

where x is the vector of design variables, f is the objective function and h, g are the equality and
inequality constraints respectively. In the case of the blade stiffened panel the vector of design variables is
h o o o o
i
x = n±45
sk , n0sk , n±45
st , n0st , b, h, L (2)

The objective function is the total weight per unit area of the panel. It is a function of the design variables
and is defined as follows:
f = (Wskin + Wstiffeners + Wrib ) /bL (3)
where,
Wskin = (ρ0sk t0sk + ρ±45sk t±45sk ) bL
Wstiffener = (ρ0st t0st + ρ±45st t±45st ) bL (4)
Wrib = cb (c is a constant)
The constraints on the design variables are determined by the failure modes of the panel. Four different
failure modes are identified, local skin buckling between the stiffeners, flexural buckling of the skin- stringer
combination, stiffener buckling and ply failure. These constraints are formulated as follows:

10 of 16

American Institute of Aeronautics and Astronautics

 g1 : σskin − σskin buckling ≤ 0 Local skin buckling
g2 : σflexural − σflexural buckling ≤ 0 Flexural buckling of skin − stringer combination
g3 : σstiffener − σstiffener buckling ≤ 0 Stiffener buckling (5)
g4 − g7 : TH − 1 ≤ 0 Tsai − Hill failure criterion for 0o and ± 45o in skin
and stiffener respectively

Two different methods are used to evaluate the buckling constraints. Local skin buckling (g1 ) and flexural
buckling(g2 ) constraints are evaluated by the following analytical formulas, respectively,

2π 2 p 
σskin buckling = 2
(t0sk + t±45sk ) D11 D22 + D33 (6)
b

π 2 EI
σflexural = (7)
L2 A
The stiffener buckling constraint (g3 ) is evaluated by a finite element program, since no simple analytical
design formula is available.
The Tsai-Hill failure criterion used in constraints g4 − g7 is defined as,

σx2 σx σy σy2 2
τxy
TH = − + + (8)
X2 X2 Y2 S2
Finally, to obtain a properly posed problem, table 1 defines upper and lower boundaries on the design
variables.
The optimisation problem is solved using the MATLAB optimisation toolbox. Two materials have been
defined for the optimisation problem. The first is a unidirectional pre-preg, the second is a angle-ply, see
Table 2. The Sequential Quadratic Programming (SQP) algorithm is chosen as the solution algorithm. For
the evaluation of the stiffener buckling constraint g3 an external call is made by MATLAB to an FE program.

(a) local skin buckling (b) flexural buckling (c) stiffener buckling

Figure 9. Buckling modes considered in the composite panel initiator

11 of 16

American Institute of Aeronautics and Astronautics

Design variable Units lower bound upper bound
panel width (b) mm 60 100
stiffener height (h) mm 10 50
no. 0o plies in skin - 1 80
no. 0o plies in stiffener - 1 80
no. ±45o plies in skin - 1 80
no. ±45o plies in stiffener - 1 80
panel length (L) mm fixed at 500

Table 1. Upper and lower bounds on design variables

Property SI Units Value

T700 Uni-directional
E11 N/mm2 98000
E22 N/mm2 8000
2
G12 N/mm 3100
ν12 - 0.3
tply mm 0.125
T700 Angle-ply
E11 N/mm2 9980.3
E22 N/mm2 9980.3
2
G12 N/mm 11679
ν12 - 0.60973
tply mm 0.125

Table 2. Laminate properties used in the optimisation problem

12 of 16

American Institute of Aeronautics and Astronautics

B. Results

(a) Design variables (b) Buckling constraints

(c) Reserve factor based on Tsai-Hill

Figure 10. Composite panel initiator results with varying compressive load intensities.

Figures 10 show the results of the optimisation process for a range of compressive load intensities.
Fig.10(a) shows the design variables as a function of load intensity. Fig.10(b) and 10(c) show the objective
function, the buckling constraints and the material failure constraints for a range of load intensities.

13 of 16

American Institute of Aeronautics and Astronautics

 (a) (b)

(c) (d)

Figure 11. Spider plots of design variables relative to their lower and upper bounds at specific load intensities.

C. Discussion
Initially, the results of the composite panel initiator showed several ’spikes’ in some regions of the compressive
load intensity range (Fig.12). An investigation of the optimisation problem showed that the reason for this
behaviour was the sensitivity of the solution to the initial values of the design variables.
The spikes disappear when the solution for a previous lower load intensity is used as the start vector for
the next higher load intensity. Efforts should be made to order the optimisation of numerous panels from low
to high load intensities. When large changes in load intensities occur, the startvector for the optimisation
should be carefully chosen. Two other approaches could be investigated to improve the robustness. One
could use a more resilient optimisation algorithm like Simulated Annealing (SA) instead of the currently
used gradient based SQP. Or alternatively make use of Design of Experiments (DoE) to generate multiple
start vectors in the vicinity of the expected solution and select the best solution.
The results for the design variable values as a function of load intensity seem to be logic from an engi-
neering standpoint. The optimiser steers the number of layers such that they make the best contribution to

14 of 16

American Institute of Aeronautics and Astronautics

 Figure 12. Composite panel initiator without the use of the previous solution as start vector.

either stiffness or strength, depending on the load intensity and the related active constraints. If a former
optimal design vector is used as input for a consecutive higher load intensity the algorithm will stick to a
similar structural solution, while there may be a marginally more optimal solution nearby. This is the case
in two areas of load intensities in Fig.10, namely around 18000 N/mm and 24000 N/mm, resulting in a
small bump in the weight results. Based on regression these bumps are respectively 2.5% and 1.0% of the
expected optimal weight solution.
No sudden changes occur from one solution to the other. This smooth behaviour of the output allows
the process to be used as an initiator for structural components like wings and tails. The initiator can do
the automated preliminary sizing based on the designers choice for outer mould line and structural concept.

VII. Conclusions
The use of Knowledge Engineering tools allows the definition of true parametric models in the optimisation
sense. Implementation of these models in a Design and Engineering Engine (DEE), operating within a robust
framework, could enable selection of the optimal aircraft configuration for a specific use case. Optimisation
plays a major role in the DEE both at the Initiator stage, where initial parameter and variable values have
to be assigned, and at the top-level optimisation supporting selection of the optimal configuration.
So far the DEE principle has been applied to design problems limited in complexity. The real challenge
will be to use the DEE to support the selection of the optimal aircraft configuration for a specific mission
profile. To achieve this, High-Level-Primitives and Initiators must be developed for all aircraft elements of
importance in the conceptual design phase. In addition the Multi-Model Generator must include models for
all relevant system behaviour to allow for the required Multi-Disciplinary Optimisation.

15 of 16

American Institute of Aeronautics and Astronautics

 References
1 Beukers A., van Tooren M.J.L., and Vermeeren C.A.J.R. The development of aircraft structures in the next century. The

Aeronautical Journal, 107(1072), June 2003.

2 Morris A.J. MOB a european distributed multi-disciplinary design and optimisation project. 9th AIAA/ISSMO Symposium

on MDO, Atlanta, Georgia, 2002.

3 Cerulli C., Meijer P.B., and Tooren M.J.L. van. Parametric modelling of aircraft families for load calculation support.

45th AIAA/ASME/ASCE/ AHS/ASC Structures, Structural Dynamics and Materials Conference, Palm Springs, April 2004.
4 La Rocca G., Krakers L., and van Tooren M.J.L. Development of an icad generative model for blended wing body aircraft

design. 9th AIAA/ISSMO Symposium on MDO, Atlanta, Georgia, 2002.

5 Green J.E. Civil aviation and the environmental challenge. The Aeronautical Journal, 107(1072), June 2003.
6 Lisandrin P. and Tooren M.J.L. van. High-order finite elements reduced models for use in a flutter design tool. 45th

AIAA/ASME/ASCE/ AHS/ASC Structures, Structural Dynamics and Materials Conference, Palm Springs, April 2004.
7 van Tooren M.J.L. Sustainable knowledge growth, inaugural speech. Delft University of Technology, the Netherlands,

March 5th 2003.

8 Ten Haaf W., Bikker H, and Adriaanse D.J. Fundamentals of Business Engineering and Management. Delft University

Press, first edition, 2002.

16 of 16

American Institute of Aeronautics and Astronautics