Price 2006

ARTICLE IN PRESS
Progress in Aerospace Sciences 42 (2006) 331–376

www.elsevier.com/locate/paerosci
An integrated systems engineering approach to aircraft design

M. Price, S. Raghunathan, R. Curran
Centre of Excellence for Integrated Aircraft Technologies, School of Mechanical and Aerospace Engineering,
Queen’s University Belfast, Belfast BT9 5AG, UK
Abstract
The challenge in Aerospace Engineering, in the next two decades as set by Vision 2020, is to meet the targets of reduction
of nitric oxide emission by 80%, carbon monoxide and carbon dioxide both by 50%, reduce noise by 50% and of course
with reduced cost and improved safety. All this must be achieved with expected increase in capacity and demand. Such a
challenge has to be in a background where the understanding of physics of flight has changed very little over the years and
where industrial growth is driven primarily by cost rather than new technology.
The way forward to meet the challenges is to introduce innovative technologies and develop an integrated, effective and
efficient process for the life cycle design of aircraft, known as systems engineering (SE). SE is a holistic approach to a
product that comprises several components. Customer specifications, conceptual design, risk analysis, functional analysis
and architecture, physical architecture, design analysis and synthesis, and trade studies and optimisation, manufacturing,
testing validation and verification, delivery, life cycle cost and management. Further, it involves interaction between
traditional disciplines such as Aerodynamics, Structures and Flight Mechanics with people- and process-oriented
disciplines such as Management, Manufacturing, and Technology Transfer.
SE has become the state-of-the-art methodology for organising and managing aerospace production. However, like
many well founded methodologies, it is more difficult to embody the core principles into formalised models and tools. The
key contribution of the paper will be to review this formalisation and to present the very latest knowledge and technology
that facilitates SE theory. Typically, research into SE provides a deeper understanding of the core principles and
interactions, and helps one to appreciate the required technical architecture for fully exploiting it as a process, rather than a
series of events.
There are major issues as regards to systems approach to aircraft design and these include lack of basic scientific/
practical models and tools for interfacing and integrating the components of SE and within a given component, for
example, life cycle cost, basic models for linking the key drivers. The paper will review the current state of art in SE
approach to aircraft design and identify some of the major challenges, the current state of the art and visions for the future.
The review moves from an initial basis in traditional engineering design processes to consideration of costs and
manufacturing in this integrated environment. Issues related to the implementation of integration in design at the detailed
physics level are discussed in the case studies.
r 2007 Elsevier Ltd. All rights reserved.
Keywords: Systems engineering; Aircraft design; Interactions; Economics; Digital manufacturing; Nacelle–wing–pylon integration;
Automated design
Corresponding author. Tel.: +44 28 9097 4178; fax: +44 28 9097 5576.
E-mail address: m.price@qub.ac.uk (M. Price).
0376-0421/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.paerosci.2006.11.002
ARTICLE IN PRESS
332 M. Price et al. / Progress in Aerospace Sciences 42 (2006) 331–376
Contents
1. Introduction and background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333

1.1. General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333
1.2. Stasis in design—an over-constrained problem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
1.3. Issues in multi-disciplinary optimisation and integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
1.4. SE—what is it and what can it provide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
1.5. The problem statement and challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336
1.6. Approaches for these problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336
2. Integration of design and analysis into SE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336
2.1. The current typical design process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336
2.1.1. Conceptual design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
2.1.2. Preliminary design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
2.1.3. Detail design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338
2.1.4. General engineering design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338
2.1.5. CAE tools and their role in design support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338
2.2. SE process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
2.3. Gaps in the process. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
2.4. Rethinking the analysis and design process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
2.5. Application of DADI in aircraft design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
3. Identification and measurement of interfaces and emergent behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346
3.1. Systems modelling to assess the system versus behavioural modeling . . . . . . . . . . . . . . . . . . . . . . . . . 346
3.2. System characteristics and measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347
3.3. Simulation driven design environments and analysis fidelity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348
3.4. The technical architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349
3.5. Identifying interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349
3.6. Parameters versus constraints. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
3.7. Bounding the design space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
3.8. Reducing emergent properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
3.9. Risk. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
3.10. Unspecified architectures in dynamic systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
3.11. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
4. Economics and digital manufacturing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
4.1. Cost and SE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
4.2. A life cycle view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
4.3. Cost as a customer requirement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354
4.4. Relationship back to design: multi-disciplinary design synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354
4.5. Cost integrated design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
4.6. Approaches to cost modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 356
4.7. Cost modelling at various stages of the life cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
4.7.1. Development cost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
4.7.2. Production cost: in-house . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
4.7.3. Production cost: procurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
4.7.4. Operations cost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
4.8. Virtual product design and manufacture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360
4.9. Cost optimisation and design trade-off . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
4.10. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362
5. Collaborative design and the virtual enterprise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363
6. Applications and case studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363
6.1. Integration of aerodynamics and manufacturing for cost reduction . . . . . . . . . . . . . . . . . . . . . . . . . . 363
6.1.1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363
6.1.2. Development of an optimised design to reduce nacelle manufacturing cost . . . . . . . . . . . . . . . 363
6.1.3. Prediction of aerodynamic performance and cost arising from surface tolerance . . . . . . . . . . . . 365
6.1.4. Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365
6.2. Trade-offs and integration at conceptual design level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365
6.3. Aerodynamic integration of wing–pylon and nacelle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367
ARTICLE IN PRESS
M. Price et al. / Progress in Aerospace Sciences 42 (2006) 331–376 333
6.3.1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367

6.3.2. Lift, drag and flow field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 368
6.3.3. Factors influencing aerodynamic integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369
6.3.4. Numerical simulations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371
6.3.5. Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371
7. Concluding remarks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372
1. Introduction and background Perhaps the biggest challenge in this suite of

problems is the consideration of the wider systems
1.1. General in which aircraft operate, including the ATM
system, general transport and political systems. In
The aviation industry, including both manufac- the current view many of these aspects appear as
tures and airlines, generates more than $1000 bn constraints on the aircraft designer, making solu-
annually for the global economy [1]. This does not tions harder to find. However, if such issues can be
account for its secondary impact on the efficiency or incorporated into the design process then gains can
viability of other sectors which depend on air be made. For example, relaxation of design and
transportation. Growth is predicted to continue performance constraints for short range aircraft in
over the next 15 years at approximately 5% per favour of economics and environmental considera-
annum with corresponding impacts on revenue, the tion may favour turboprops over jets. But this is still
air transportation systems and the environment in consideration of conventional configurations and
[2,3]. The impact on society has been enormous, and it is necessary to consider other than the civil
the thirst for rapid transport has meant that growth transport industry for more radical ideas.
of the industry has not relied on continuing large Many successful unmanned air vehicles (UAVs)
scale innovations in design and the fundamental have been developed for various civil and military
shape and form of aircraft have not changed roles which do not follow conventional configura-
significantly in the last 50 years [4,5]. Aircraft tions [7,8]. More adventurous technology pro-
configurations which are prevalent today have been grammes such as Helios [9] and Spaceship One
highly successful, reliable and profitable and there [10] are showing new technologies and radical
has therefore been no incentive to change these. As designs but these are driven by the skill and
a result the design process has not changed either, imagination of unique individuals. They cannot be
and although advances in technology has provided a product of existing design methods since these are
advanced materials, systems, analysis and design largely based on empirical equations, which were
tools, these have been developed in the context of derived from existing aircraft [11–14]. This ap-
existing design procedures and business processes. proach is naturally conservative since similar results
The consequence has been improvements in effi- are always produced and this is an excellent and
ciency and reliability of product and process but no reliable approach for relatively static or slowly
shifts in the design paradigms. evolving capability, as has been the case for aviation
However, the emergence of serious issues such as since the 1960s. But the new social, legislative and
global warming, noxious emissions and noise environmental challenges now sit alongside as very
pollution coupled with a requirement to reduce the strict constraints and are difficult to overcome.
overall cost of aircraft is driving the need for radical There is a need for new design paradigms which can
new paradigms. It is relatively straightforward to account for both traditional and non-technical
show that changes in individual technologies will disciplines within the same framework so that such
struggle to produce the gains needed to meet the issues become design parameters to be worked with,
new stringent targets, and advances over a broad rather than forming constraints on the problem.
range of issues within the entire air transport This paper identifies some of the key technical
systems will need to be considered, such as aircraft challenges in developing conceptual design tools
flying long ranges in hops, and perhaps flying at and approaches in an integrated environment. The
lower altitudes [6]. field is so broad that it will be difficult to address
ARTICLE IN PRESS
everything, for example there are many problems in ceptual design is highly influenced by a number of
the design, analysis and costing of flight control different technical and non-technical disciplines and
systems, and the size and complexity of real time the capability to integrate these is needed to design
software used in these is a field in its own right. As an aircraft. This issue has many different facets.
will become clear the leap to fully integrated design Firstly, there is the obvious problem of getting data
solutions inclusive of all disciplines is an enormous from one analysis to another, that is, basic data
one. Therefore, this paper focuses on moving from transfer. Although this is difficult it is not impos-
the current well understood traditional design sible at a given level of detail. But this is the second
foundations by considering the process used to issue. At different stages in design, and as under-
generate manufacturing, design, analysis and cost standing evolves, the model detail, or fidelity,
models. This creates a strong synergy with the differs. At early stages models tend to be low
systems engineering (SE) approach. The principles fidelity and broad, but later models are high fidelity
embodied in this approach are applicable to a and focused. In some disciplines, particularly social
broader range of disciplines. This general approach and environmental disciplines, understanding is still
then allows design synthesis to become an integral so sparse that reliable models at any level do not
part of the SE process rather than a supporting exist.
function. By including a greater range of disciplines Analysis fidelity relates to the degree of detail and
the design space shifts and opens up to provide accuracy contained in a given analysis model,
alternative and more optimised solutions. namely, how faithful the model behaviour is to the
Before detailing developments in new approaches component it represents. Low fidelity models tend
it is worth focussing briefly on the key problems in to use simple equations and look up tables and
the above-mentioned areas and highlighting the often will not have any geometric models asso-
resulting challenges in a more technical way. ciated. Medium fidelity models will mostly have
some form of linear analysis and high fidelity
1.2. Stasis in design—an over-constrained problem models contain a lot of detail and often model
non-linear behaviour. Thus, to make trade-offs at
The basic issue with current approaches to early design stages low fidelity models are used,
conceptual design is that they result in highly since they are fast and many different aspects and
constrained design spaces. Of course, this is disciplines may be considered. High fidelity models
necessary since the initial problem is extremely tend to be used in very local areas later in the design
difficult and must be simplified to enable a solution phase since they are expensive in time and resources
to be found. The problem is that existing methods [15]. It becomes obvious very quickly that the
close up the design space very quickly. In starting appropriate mix of these is needed to optimise the
with a blank sheet, requirements are provided, development of the model to get good reliable
normally a range and payload, maybe with a speed. answers for as many configurations as possible
Even this limited selection of constraints locks in an [16,17].
L/D ratio for the aircraft, and variations now Nickol [18] presented a table of key analysis
mostly occur via the choice of engines and the best disciplines and associated fidelities providing an
shape of fuselage to carry the payload. Initial initial range of typical tools used in design. His
estimates are made based on previous knowledge table shows that even at this level of detail there
and these high level parameters will then tend to are gaps in capability (Fig. 1(a)). But the picture is
converge to similar values. By the time aerody- even more taxing than this. Adding cost and
namics, structures and manufacturing are looked at manufacturing to the list of disciplines shows
in more detail the constraints are set. even more gaps. But the real problem from a design
point of view is in integration. At any point in the
1.3. Issues in multi-disciplinary optimisation and design, trade-offs may be required focused on one
integration given aspect. The data needed for that may come
from several other disciplines each at different
The challenges posed in constraining the problem levels of fidelity. For example, a structures FEA
lead naturally into thoughts on the multi-disciplin- model may take data from low fidelity aerodynamic
ary nature of the problem and how so many models, control system models, high fidelity CAD
disparate disciplines can come together. The con- models, or even low fidelity structures models
ARTICLE IN PRESS
Analysis Discipline
Aerodynamics Structures & Noise Emissions Flight Geometry Cost Manufacturing Maintenance
Weights Control
Systems
Low Empirical Empirical Empirical

Empirical Vehicle Piano,
Methods Methods Methods
Methods (e.g. GAP Sketch Pad SEER, GAP GAP
(e.g. (e.g. (e.g.
EDET) (VSP) PRICE
FLOPS) FLOPS) FLOPS)
Medium Basic
Vort ex Lattice Basic Noise Vortex
Fidelity
Structures
(e.g. (e.g. GAP Lattice (e.g. GAP GAP GAP GAP
(e.g.
WINGDES) ANOPP) VORSTAB)
ELAPS)
High High End
CFD Advanced CAD Virtual Factory
FEM (e.g.
(e.g. FUN2D, Noise (e.g. GAP GAP systems GAP Simulations GAP
NASTRAN)
Fluent) AV ATAR) (e.g. (e.g. DELMIA)
CATIA)
b
Analysis Discipline
Aerodynamics Structures & Noise Emissions Flight Geometry Cost Manufacturing Maintenance
Weights Control
Systems
Low Empirical Empirical Empirical Empirical

Vehicle Piano,
Methods (e.g. Methods Methods Methods
GAP Sketch Pad SEER, GAP GAP
EDET) (e.g. (e.g. (e.g.
(VSP) PRICE
FLOPS) FLOPS) FLOPS)
Medium Basic
Vort ex Lattice Basic Noise Vortex
Fidelity
Structures
(e.g. (e.g. GAP Lattice (e.g. GAP GAP GAP GAP
(e.g.
WINGDES) ANOPP) VORSTAB)
ELAPS)
High High End
CFD Advanced CAD Virtual Factory
FEM (e.g.
(e.g. FUN2D, Noise (e.g. GAP GAP systems GAP Simulations GAP
NASTRAN)
Fluent) AV ATAR) (e.g. (e.g. DELMIA)
CATIA)
Fig. 1. Table of disciplines versus fidelity (modified from Nickol [18]): (a) extension of Nickol’s table to include cost, manufacturing
and maintenance; (b) interaction of disciplines and fidelities results in a web of interactions and linkages, in this scheme a potential 93
bi-directional linkages.
(i.e. taking loads from a more idealised model). The 1.4. SE—what is it and what can it provide
table shown in Fig. 1(b) is a modified version of
Nickol’s table now with nine disciplines and three SE has several formal definitions one of which is
levels of fidelity. This gives 729 bi-directional ‘‘an interdisciplinary, collaborative approach that
linkages. There are a multitude of issues here derives, evolves, and verifies a life-cycle balanced
including simple considerations of what data is system solution’’ [19]. In practice, this process drives
needed as input or what is provided for as output the design development in many aerospace compa-
from a given analysis. For example, 1D linear FE or nies and in wider engineering industry. It brings in
fluids models do not provide the richness of data consideration of many issues beyond the technical
available or needed in 3D simulations, and con- design challenge [20]. Issues such as risk, derivation
versely the 3D simulations provide too much data of the functional and physical architecture of the
for 1D models or simple parameter studies. Filtering product come to the fore and design synthesis
of data is therefore needed. This is a web of data becomes a supporting technology helping to make
and methods, and for full integration an approach is the decisions support the development of the system
needed which will more naturally permit control architecture of the product.
and development of models with clear paths and In doing this, the focus moves away from
filters for the data. technology itself to consideration of the requirements
ARTICLE IN PRESS
of the customer and what function the product 3. Digital manufacturing and economics.
should have. This seems a natural business view of 4. Collaborative design and the virtual enterprise.
design and development but the design methods
commonly used in practice have their roots in the In this review the first three of these are addressed
days when technology was the driver and hence this in some detail in Sections 2, 3 and 4. Collaborative
approach brings a different viewpoint. It makes the design and the virtual enterprise is a field now
designer think about the wider issues including expanding and applying new ideas which deserves a
manufacturing and cost, and the system in which review in its own right. Introductory references only
the vehicle will operate, therefore bringing many non- are given for this in Section 5.
technical disciplines into view.
Unfortunately many such disciplines do not have 1.6. Approaches for these problems
analytical models or any kind of model which
naturally blends with traditional engineering dis- To address these challenges fully will of course
ciplines. Moreover the resulting systems are very require very significant effort over a long period,
complex and it is difficult to develop sufficient however, focussing on the process can make
understanding of behaviour to cover all eventua- substantial progress and this is the approach being
lities. Different system elements interact in many taken more often in this field [16,17,22]. Specifically
ways and eventual performance or behaviour may the model building process is redefined to fit better
not always be as predicted. Such emergent beha- with SE with a strong emphasis on integrating
viour is a real challenge for engineers. through fidelities and across disciplines. Since model
fidelity and unknowns enter the picture, considera-
1.5. The problem statement and challenges tion is given to the introduction of a technical
architecture [23] to capture errors and emergent
It is clear that the current industrial environment behaviours. This is eventually linked to digital
has significantly expanded the remit and expecta- manufacturing to provide a path from concept
tions placed on designers. It is no longer sufficient to through to product.
provide a high performance product that satisfies The application focus of these developments has
strict technical requirements. The inclusion of been on airframe design problems in providing a
requirements definitions and life-cycle issues, capability to carry out trade-offs on configurations
through to eventual decommissioning, has resulted at as high a level as possible.
in a much broader definition of multi-disciplinarity
and now encompasses non-technical disciplines 2. Integration of design and analysis into SE
which often do not have analytical models. New
models for these disciplines are needed but equally It is useful to first consider the more traditional
there is a need for a broader framework within approach and assess how it can be used more
which these models may be embedded. The design effectively before describing how design synthesis
process then requires a new paradigm in which can be integrated with SE.
analysis and design syntheses are tightly integrated
with SE to facilitate the wider life-cycle issues. These 2.1. The current typical design process
are difficult enough, but the global nature of
aerospace and the vast monetary sums required to Perhaps the simplest description of the design
develop products requires that design teams are process is the linear diagram which follows require-
multi-national and the products are designed in ments through conceptual design, preliminary and
collaboration across what is known as the ‘Virtual finally detailed design before manufacture begins
Enterprise’ [21]. (Fig. 2).
The four key challenges emerging from this are At this point requirements will be taken as a given
then seen as: which provide parameters, constraints and targets
for the design. As integration progresses, it will be
1. Integration of design and analysis methods into a seen that requirements should be brought more into
SE framework. the design process [24], but traditionally require-
2. Identification and measurement of interfaces and ments act more as constraints on the design.
emergent behaviour. Manufacture is also more traditionally seen as the
ARTICLE IN PRESS
Requirements Conceptual Preliminary Detail Manufacture
Cl,
Range, Weight, Cd,
Fig. 2. Simple schematic of the design process.
Payload, Lift, S,
Cruise Drag, A,
recipient of design details, therefore the three central Speed Fuel Lt,
phases have dominated the technical design arena. Coefficient C
etc
2.1.1. Conceptual design

The conceptual design phase addresses the high-
Fig. 3. Increasing numbers of design parameters.
est level questions about the proposed aircraft. In
particular the main requirements and desired func-
as the design proceeds more values are being fixed
tions are considered, and normally a number of
and the design space is being gradually reduced.
potential configurations are outlined which will
However, at the same time the number of design
undergo a trade-off. The solution which best
parameters is increasing (Fig. 3).
matches the requirements will be chosen. Like any
Completion of the conceptual phase occurs when
design process this phase is highly iterative, but it
a satisfactory level of functionality and meeting of
tends to be the most open and unconstrained phase
the requirements is achieved. At this point sufficient
of aircraft design so the largest number of design
data exist for basic sizing of the internal structure
solutions can be tried here.
and a better definition of aerodynamic surfaces, etc.
The result of the conceptual phase should be a
The problem has now been sufficiently constrained
configuration with the basic size and arrangement of
for these to have meaningful solutions, and so the
its main aspects, such as the wing, empennage,
preliminary design phase may begin.
engines, fuselage, control surfaces, etc. These data
Of course in practice, especially in the large
are built from a few initial equations and empirical
commercial aerospace companies, the procedure is
data that provide a starting point for the design to
more complex and follows a more rigorous SE
proceed. In this traditional approach for powered
approach, however the basic steps and data are
aircraft two key equations dominate the design and
essentially the same.
these relate to the initial estimate of weight, and the
range equation which provides for the lift to drag
ratio and fuel efficiency coefficient 2.1.2. Preliminary design
W crew þ W payload Once the concept has been accepted and the
W0 ¼ . (1) design space is sufficiently well defined, or con-
1 W f =W 0 W e =W 0
strained, preliminary design may proceed. Raymer
Eq. (1) provides an estimate of W0 based on the [11] describes this as the point when all the big
weight of crew (Wcrew), payload (Wpayload), fuel questions have been answered. Each major system
weight (Wf), and the empty weight (We) of the aircraft is now considered again and more
detailed estimates of size, thickness, material, etc.
VL W i1
R¼ ln . (2) are made. In terms of the airframe this may be each
CD Wi
major section. For example, the fuselage skin
Eq. (2) provides an estimate of the range, R, thickness, number of stringers, frames will be given
based on the lift to drag ratio (L/D), the velocity (V) initial values. Estimates of internal structural loads
and the specific fuel consumption (C), and weight will then be obtained using finite element models.
over the cruise segments. Much complexity and difficulty here can come from
This basic data can then be used for initial sizing the many load cases needed in practice, which can
of the fuselage, wing and tail, and further empirical number in hundreds. Obtaining an acceptable
equations can guide key dimensions thus providing solution which covers the worst cases without being
estimates of the areas, lengths and positions. The over designed for normal in-service conditions is a
concept of volume coefficients in combination with significant challenge, even with today’s technology.
moment arms provides this, and is useful then in Typical airframe loads models have approximately
iterating to better estimates. It is notable now that 200,000 degrees of freedom. At this point in the
ARTICLE IN PRESS
tskin,
Cl,
Cd,
tweb,
S,
A, dframe,
Lt,
Etc
etc
Fig. 5. Structural details include holes, rivet positions, etc.
follow a SE approach where there is continual

Fig. 4. Further proliferation of the design parameters.
iteration between the phases and several design
reviews to ensure the work proceeds as smoothly as
possible. There are differences when the role of
design connections, local cut-outs and other such Concurrent Engineering and Integrated Product
details have not been established, but even so the Teams is considered, since these bring some issues
number of parameters has escalated again. For from production and operation to the early design
example, the fuselage now has a length, radius, skin phases [25,26]. But, in general this approach is still
thickness, frame spacing, stringer spacing and each valid, and moreover, is common across the en-
of these has areas and thicknesses as well (Fig. 4). gineering and design sector [27–29].
Once again the design space has been further The current approach to aircraft (or any) design
constrained to enable the details to be added. can be summarised as being based on conventional
configurations using empirical methods at the high-
2.1.3. Detail design est level supplemented by sophisticated multi-
Successful review and acceptance of the prelimin- disciplinary simulations at more detailed levels.
ary design allows the detail design phase to proceed. Because this core approach has not changed in
The design team now expands by an order of many years and it cannot be easily adapted to non-
magnitude and work begins on developing the standard aircraft, or easily account for the impact of
detailed drawings to enable the aircraft to be new designs, processes or materials which are
manufactured. However, design and analysis con- needed to address the grand challenges in aircraft
tinues through detailed analyses, such as stressing of design.
the final shape and form of the components.
Manufacturing details are added, such as holes, 2.1.5. CAE tools and their role in design support
blends and assemblies defined specifying exactly Although the basic design process has not
how the aircraft will be built. Much of this work changed much there have been major advances in
proceeds independently of the preliminary design the supporting technology, particularly in the
but refers back to this, and periodic design reviews computer aided design arena. CAD and analysis
again keep the detailed definitions within acceptable tools now provide the capability to develop designs
bounds. The number of design variables now at a fraction of the time and cost as before (Fig. 6).
explodes, including parameters on rivet pitch, joint This speed up has permitted more scenarios to be
details, etc. (Fig. 5). It is highly unusual for any considered and has resulted in a greater certainty in
detail change to affect the overall concept. Design of predicted behaviour. They have created opportu-
such local features is highly constrained and there is nities in sharing data between disciplines [30–32]
usually little room for manoeuvre. and in a more collaborative enterprise between sites
or companies [33–35]. In fact, these advances have
2.1.4. General engineering design focused attention on the limitations of design
The basic design procedure described here has models and the differences between the model world
many variations depending on the accepted practice and the real world [36]. The practicalities of
in a given company and their capabilities. Many generating multi-disciplinary models means they
ARTICLE IN PRESS
are often built using different parameters than the has been used as a process management tool in
designers have used. That is, the parameters used at which the functional and physical architectures are
the high level design are not necessarily used in linked to enable closer co-ordination and manage-
detailed design models. The control and management of such complex systems as aircraft [16,20].
ment of design parameters is needed in this new Since SE provides the linkage between the customer
environment [37] (Fig. 7). requirements and the eventual product shape and
performance it is well placed to provide the
2.2. SE process integrated environment needed. However existing
analysis tools used in design synthesis are built from
Having considered the design/analysis issues it is a different ethos, and do not easily blend into the SE
useful to turn to the SE process which is used to approach.
control the design and development of products. SE Compare the SE view of the design process in
Fig. 8 to that in Fig. 2. Here the system is evolved in
depth and complexity from initial requirements
through to eventual manufacture of a complete
product. Analysis is used in a design synthesis loop
supporting the system development and there is
continual feedback to function and requirements.
Analysis models are therefore being used to ensure
requirements are being satisfied as the design
iterates towards a stable solution. However, the
same basic flow of the process from requirements to
product manufacture is still evident.
Requirements
Requirements
Management Loop
Functional Analysis and

Allocation Design Synthesis
loop
System Synthesis
Fig. 6. Complex finite element analysis models can be generated

quickly and automatically from CAD systems. Fig. 8. Systems engineering process model.
Fig. 7. Global design parameters are naturally different from those used local analysis model.
ARTICLE IN PRESS
1.0
Aircraft
1.3 1.4 1.5

1.1 1.2
Operation Propulsion Avionics/Control
Airframe Other Systems Interfaces System Systems
1.1.1 1.1.2 1.1.3 1.1.4

Fuselage Wing Nacelle Empennage
1.4.1 1.4.2
Engine Fuel
1.1.2.1 1.1.2.2 1.1.2.3 Data used for

Pylon Flaps Allerons analysis cuts across
systems
Fig. 9. Simplified system schematic.
2.3. Gaps in the process Closer inspection of even this simple example
shows that detail is added to the system architecture
Although in taking such a high level view of both in a different way than the analysis models which
design processes it is easy to see analogies and fit, always focus on the physical (i.e. geometry)
the implementation of such complex processes in instantiation of the design. It is the system
practice clouds the issue, and meaningful linkages architectures which drive the eventual manufacture
are more difficult to see. For example, analysis of of the aircraft, with analysis used to support the
some detailed areas may be necessary to help decisions made. Thus, to measure and define system
support the system definition at a higher level, and function, many analysis types are needed which
in general it is analysis which provides the metrics have different information and levels of detail.
used to measure the system performance. Thus, Traditionally, systems integration is seen as the
design models may cut across or include many domain of the individual company since it is so
systems or sub-systems, and conversely the mea- tightly associated with the final product, but
surement of a given system may need several analysis methods are more generic and cut across
different analyses depending on context. systems views. This leads to the core of the problem.
The simple schematic in Fig. 9 illustrates this for The two approaches are not quite in synch. As
simplified airframe and propulsion systems. The detail is added to the systems architecture it is not
airframe is broken down into its major sub-systems, necessarily added to the design models. The natural
such as the wing, and similarly the propulsion consequence of this has been to focus on the final
system is broken down into its major sub-systems. geometry of the ‘‘as manufactured’’ product as
Although these are separate systems they are linked being the closest representation of the real product.
in design and analysis. For example, some loads for Everyone then draws on this baseline for the core
the wing are provided by the engine thrust and fuel information. But for integration this raises some
weight; and the geometry of the wing and nacelle immediate questions. Firstly, the final geometry is
are affected by the engine choice. From the airframe not known until the design is complete, so it is
design perspective the design model is usually a unclear as to how early and intermediate designs
finite element loads master model which will should be handled. Secondly, each discipline re-
incorporate elements of all these systems involved. quires different views or idealisations, so sophisti-
It therefore cuts across many systems and analysis cated tools are needed to abstract idealised models
uses parameters from all of these. This raises the from detailed geometry. This is compounded by
issue that optimisations may be either over concurrent efforts to broaden the disciplines used in
strained, or may result in local optima which cannot early design to include others, such as manufactur-
be achieved in practice because remote parameters ing and economics, where models tend to be
are being separately optimised as part of their study. different (for example, a statistical model) and their
It is difficult to control such issues without fidelity may not easily match the other disciplines.
constraining the problem further. This is still one of the major bottlenecks in analysis
ARTICLE IN PRESS
and design. Thirdly, the final geometry as an directly connected to the original conceptual mod-
absolute does not necessarily show the key design els, and so feedback is not always straightforward.
parameters. In fact, they are often hidden or masked Secondly, teams from different disciplines take the
in the model building process with individual relevant data from the CAD model for their
designers adding local parameters as appropriate problems and work using this. However, as the
for their design problem. Finally, the geometry does design is evolving they will only update as their own
not represent a unique map of the systems problem is affected [15]. Thus, a divergence in the
architecture and therefore systems interfaces and data being used by different discipline teams occurs.
interactions are not transparent. Interfaces are The final consequence is that integration and multi-
specified by the designer but interactions may disciplinary simulations become difficult to manage
emerge as a consequence of this. since the divergence of models and data means that
the interfaces between the models are more difficult
2.4. Rethinking the analysis and design process to manage.
To address these problems the approach taken by
The preceding discussion provides one view of a Armstrong et al. [15], Mawhinney et al. [49], and
commonly recognised problem which has spawned Price et al. [50] sets the key objective as to ensure
several attempts at providing integrated design that models evolve from and always refer back to
environments [38–42] which use parametric models the identical basic conceptual model. In doing this it
and multi-disciplinary optimisation. The commer- is therefore the model building process that is
cially available tools provide capability to link tackled rather than any of the plethora of CAE
analysis models and the design can define the tools. One logical consequence is that the concep-
analysis sequence and process [43–45]. That is, they tual model is used to drive the design through all
provide functionality and naturally this is used phases conceptual, preliminary and detail. The
within existing design processes. But this does not CAD system is no longer the pivotal element and
solve the problem of integration. takes its place alongside the analysis disciplines. The
By considering the process by which models are geometry is then simply one view (the geometric
built and the consequences of this it has been view) of the product data, just as a finite element
proposed that in bringing together SE and tradi- model is an idealised view of the structural
tional analysis tools opportunities for progress in configuration of the product.
design synthesis using multi-disciplinary analysis Automation is entirely possible and very detailed
models appear [46–49]. airframe models can be produced in conjunction
The design of aircraft begins with concept with the relevant analysis models. This approach is
sketches of basic configuration and layout. These entirely consistent with the SE approaches des-
basic configurations are normally of key lines or cribed in the previous section, and claims a step
surfaces, representing a skeleton of the structure or change in the ability to trade-off designs and
simply its key features. Initial analyses are often produce more reliable concept evaluations at an
carried out for order of magnitude assessments of early stage before requirements and product archi-
behaviour, e.g. stiffness, strength. The design then tectures are fixed.
evolves or develops from these key lines and The previous sections highlighted major issues
features, until it reaches a point where it is which need to be addressed in an integrated, SE
sufficiently well advanced for a 3D CAD model to environment. The complexity of many of the issues
be developed. At this point the design moves onto arising when detail is considered makes it difficult to
the preliminary design phase where the concept is tackle these individually. A single approach addres-
taken to a greater level of detail and understanding. sing all these simultaneously has been developed
Once the basic CAD model is in place details, such [51–53] recognising that much of the problem is
as joint configurations, are added as knowledge of process oriented rather than technology oriented.
the design evolves in the detail design phase. Starting with the premise that the SE framework
Between each of these phases there is a break or is a sound basis for aircraft design and manufacture,
disjoint in the process, and there are several then the task can be refocused on bringing existing
consequences to this. Firstly the CAD model is the technology can be brought within this framework. If
core, from which development teams extract and the model building process is based on the SE
share their data, but it has not necessarily been approach and develops in synch with the systems
ARTICLE IN PRESS
Point Line Surface

sweeps a sweeps a sweeps a
line surface solid
Fig. 12. Fuselage cross-section, plus central axis generates the
barrel.
Fig. 10. Basic principles of sweeping, for lines, surfaces and
solids.
There is now sufficient data to develop, for
example, beam models for loads, and therefore to
D begin to get better estimates of structural size and
L L
weight. Fig. 12 illustrates this for a fuselage barrel.
As each detail is added it is always referenced to its
T immediate parent parameters, thus maintaining
D T
linkages between them in the model hierarchy. For
T
example, rivet holes in a stringer flange can be
W
W placed relative to the neutral axis of the stringer
(Fig. 13).
Fig. 11. Initial point in space representing the aircraft can be Note that it is the key parameters, such as radius,
swept to generate the initial key lines of the wings and fuselage. or number of stringers, that now drive the geometry.
The systems model can now be kept in synch as
shown in Fig. 14. The fuselage model could be
architectures, then many of the problems can be created as a shell or solid as needed.
overcome, at least to the point where new technol- The same principle of course applies to analysis
ogy is needed to move forward. The resulting models. Fig. 15 shows the simple schematic aircraft
process is capable of starting with a few base fuselage and wing, with corresponding FEA and
requirements or parameters, and then adding detail CAD models. The key design parameters drive
as the system architecture develops. both, but the FE model keeps stiffeners as lines with
The process proposed to do this has been called associated cross-sections, while the CAD model uses
Dimensional Addition and Detail Insertion those cross-sections to generate a full solid repre-
(DADI). The idea is very simple. Sweeping a point sentation.
along a path creates a line. Sweeping a line along a These examples show how models can be kept
path creates a surface, and sweeping a surface along consistent for multi-disciplinary analysis but since
a path creates a solid (Fig. 10). This notion allows the models are process driven it is also straightfor-
the definition of key points and lines to be used in ward to include local models and mix beam, shell
design. Sweeping then generates the appropriate and solid representations. For example, at a joint
geometry. This concept was first applied in such a between fuselage frame and stringer the sweeping
systematic way by Armstrong et al. [46]. More operations can use the cross-section of a shell up to
recently it has been extensively developed by the joint locality, then switch to a solid at the joint
Mawhinney, Price et al. for airframe applications (Fig. 16). Thus, the interfaces are clearly marked
[47–52]. For an aircraft, beginning with the top level and analysis models can be modified accordingly
it is represented as a point, which is adequate for with multi-point constraints (MPCs) or other
initial estimates of weight, fuel load, etc. as used in connection methods [54,55].
the equations in Section 2.1.1. As the design Since it is essential to embed any process in the SE
develops, i.e. moving on to basic sizing, the point framework it is interesting to observe how this
can be swept to form a line or lines representing the logical approach sits alongside a systems architec-
fuselage, wing and tail (Fig. 11). ture as shown in Fig. 17. They reflect each other
ARTICLE IN PRESS
Parameters indicating station

along the fuselage – all
referencing the neutral axis
Fuse lage
Forward Fuselage Centre Fuselage Aft Fuselage
Upper Half Lower Half
Starboard Side Panel Parameters

relative to
Parameters
parent panel
referencing the Port Side Panel
neutral axis
Stringer 1
Rivet Holes
Parameters
Frame 1 relative to
frame neutral
Cut Out 1 axis
Fig. 13. Systems breakdown can be used in guiding parameter flow through the models.
very naturally and moving up and down the tree 3. The common approach promotes generic defini-
becomes very straightforward. Common naming tion of models and a better understanding of the
conventions can be used which correspond to the requirements from other disciplines or systems
work breakdown structure (WBS) from the SE design teams.
process. As the system is developed and more detail 4. Using this simple approach requires existing
is added to the design the models remain in synch. functionality so no extensive development of
For example, as detail including say cut-outs is new tools is necessary.
added to a fuselage panel this is built from the source
higher level model, maintaining the hierarchy. Of course, this is a somewhat simplistic view and
In essence DADI approach uses the same process considerable detail is required in how models are
to build all the geometric models used in the aircraft coupled together and how parameters are passed up
design process. By using existing functionality it is and down the tree. There is a significant amount of
possible to generate CAD, FEA, CFD and other work required to fill in the details here particularly
models using the same ordered process, and this in reference to matching with systems functionality.
process can be overlaid with the systems architec- However, much of this work is needed anyway to
ture so that the design models and systems break- develop integrated analysis and design environ-
down remain in synch. This requires some change in ments, and the main argument in this work is that it
view of model building but leans more towards should be carried out within the context of SE
defining a consistent approach in concert with the processes. Further work on some of these detailed
business process than being a new approach. issues is described in Mawhinney et al. [51],
The benefits of using this are: Castagne et al. [56], Curran et al. [57] and Early
et al. [58].
1. Use of a common process helps reduce the
number of tools needed in the creation of
geometry and analysis models in general. 2.5. Application of DADI in aircraft design
2. By driving the design and analysis models from
the higher level parameters design intent is easier Mawhinney et al. [51] describes how DADI
to capture and maintain as the product evolves. is used in aircraft design for conventional
ARTICLE IN PRESS
Aircraft
Fwd Fuselage Centre Fuselage Other sub systems
Upper Half Other sub systems
LHS Panel
Location relative to
fuselage central axis
Frame 1
Location relative to parent
panel
Cut Outs
Relative to parent
Fig. 14. Overlay of CAD model and systems breakdown, showing parametric linkages as part of the architecture.
Conceptual model
idealized as beams FEA
Shell Sweep
Profile
Solid Sweep
Shell Profile
CAD Solid
Fig. 16. Mixed dimensional model showing local solids at the

FEA frame–stringer junction.
CAD
Fig. 15. The same process generates CAD and FEA models.
ing beings from the junctions or interfaces as they

appear. If the structure at the interface is similar
configurations and demonstrates how unconven- then the conventional long fuselage or wing
tional configurations such as blended wing body appears, but it is equally viable to change direction
and distributed wings like Helios, can emerge quite to generate the multiple pylons as in Helios, or to do
naturally from this process. In this, as each sub- no sweeps for the fuselage and leave only the
system is considered, geometry creation and sweep- blended wing (Fig. 18).
ARTICLE IN PRESS
Aircraft
Fwd Fuselage Centre Fuselage Other sub systems
Upper Half Other sub systems
LHS Panel
fuselage central axis
Frame 1
parent panel
Cut Outs/stiffeners
Relative to parent
Fig. 17. The analysis models are overlaid with the systems model to maintain consistency across the architecture.
Fig. 18. DADI used to generate CAD and FEA models of different aircraft configurations.
ARTICLE IN PRESS
3. Identification and measurement of interfaces and bines information from analysis models in the
emergent behaviour systems architectures. This latest research will be
described here, with future directions.
The previous section considered some of the The aim of that then is to develop a methodology
issues around integration frameworks and SE and that identifies interactions between elements of a
presented a summary of some of the latest develop- system by measuring of the performance of a
ments in linkages between MDO design models and system, and what happens to both inputs and
the SE process. Although this process will provide a outputs of that performance analysis. Four key
background framework it does not directly solve issues were considered as follows: (1) reductionist
any of the issues surrounding the identification of versus holistic system design, (2) analysis fidelity, (3)
interfaces and those unexpected interactions which real versus perceived system characteristics, (4)
result as the system elements come together. The simulation driven design environments and analysis
identification of the interfaces between systems and fidelity.
the result of any interactions is a key aspect of SE It can be seen that these relate to the generic
theory. Aircraft are complex products combining challenges in integrated frameworks and thus the
many systems and components into a single two are closely linked. The tool produced can be
functioning entity, with interactions occurring used for interface management, and enables in-
through both physical and functional connections. formed decisions about design pathways to be
Identifying and measuring such interactions is made. Each of the four key issues addressed
difficult, and new tools are being developed which highlight interesting problems and these are re-
can use engineering information to identify them, viewed in the following sections.
and their potential outputs.
In SE different views of the system are used to 3.1. Systems modelling to assess the system versus
describe it. These are known as the requirements, behavioural modeling
physical and functional architectures (R–F–P) and
they may be supplemented by technical and In attempting to understand and analyse systems
dynamic architectures which address system beha- two common philosophies exist, the reductionist
viour. Initial high level system descriptions are and the holistic. The reductionist approach to SE
gradually broken down as the design develops and involves decomposing the system into its sub-
these architectures of the system become hierarch- systems, components and down to the individual
ical and complex. Local systems are designed in parts. The result is an evolving a physical archi-
detail and then brought together to form the tecture, and similarly functional and requirements
working product. The main problem is that as more architectures. The main benefit of this approach is
layers are added to this hierarchy it becomes more that it breaks the problem down into bite sized
difficult to identify which parameters are affected at pieces which can be solved but it lacks the essential
local level, and which are affected at a global level. information which is provided by the holistic system
Interactions can occur in systems which are view, which provides details of the overall system
physically remote and sometimes these interactions behaviour.
are indirect and hidden from the initial view. It is the overall system architecture, both its
Unfortunately analysis capability may not be entities and their associated relationships, that
sufficient to capture behaviour accurately enough strongly influences the overall behaviour of the
and consequently interactions are often overlooked, system. In general, complex systems also exhibit
leading to unexpected behaviour in the final system. behaviour, which no subset of their elements has,
This is called emergent behaviour, and is often as a result of the interaction of individual compo-
unwelcome. nents of systems [60–63]. Such behaviour is called
Engineering analysis, as described in the previous emergent behaviour. Flight itself is an emer-
sections, is normally focused on predicting overall gent behaviour which arises when propul-
behaviour and is not oriented to explicitly determin- sion systems (engines) are connected with lift
ing interactions. However, by looking at system generating systems (wings). Emergent behaviour,
architecture it is possible to map and identify where wanted or unwanted, is difficult to predict and will
interactions may occur. Early et al. [58,59] achieves itself further increase the overall perceived complex-
this by using a database application, which com- ity of the system. Complex systems also contain
ARTICLE IN PRESS
non-linear interactions between components, result- order to perform a system analysis, it is usually
ing in evolving states which are dependent on the necessary for the disciplinary analyses to be
intra-system connections and the environment in executed in some sequence, and as the number of
which they sit [64]. The result of this is that disciplines and the complexity of the system
individual system components may be analysed increases, it becomes increasingly difficult to gen-
and their behaviour understood, within the limits of erate this sequencing correctly in order to obtain the
analysis capability, the interactions between com- required information [66]. From the perspective of
ponents of a system add such complexity that the simplifying the system design, fully modular archi-
resulting systems are inherently difficult to analyse tectures would be developed, in which each module
and predict. Early et al.’s work [58,59] indicates that of the architecture has a distinct function, with each
directly studying the system interfaces reveals useful module connected to another via a few, well defined
knowledge about the system. An interface can be interfaces [67]. In the limiting case, all interactions
defined as ‘‘y.a common boundary or meeting between modules should occur over these predefined
point between two different systems or proces- interfaces, and all the system behaviour should be
sesy’’ indicating that any connection between encompassed by the individual module behaviour
systems can be termed an interface, and is not only and interactions across the defined interfaces. In
limited to the more commonly perceived physical practice this is not normally the case, and many
interfaces. The technical design process generates complex system designs are somewhere between
many such interfaces, or relationships, which fully modular and fully integral, with multiple
compliment the physical and functional interactions interactions occurring across often unknown or
most readily identified within SE. unidentified interfaces.
Moreover, in the reductionist approach, systems In most cases the physical architecture is easily
are continually broken down until a single compo- decomposed in the physical domain, in which each
nent or manageable subsystem is arrived at. As entity is designed to perform a function. Although
more levels are added to this hierarchy long chains the physical system entities may be obtained
of dependence arise, and links can become difficult through a logical decomposition, the attributes
to trace through the overall hierarchy. For a (e.g. the span, chord, sweep angle of a wing) are
complex system, the arrangement of the system obtained through the results of the associated
entities and multiple relationships between those analyses, and these can in turn promote interactions
entities become more difficult to manage as the between remote systems which have no physical
design evolves and more detail is added. The main interface. Although there is a sequence associated
obstacle to reducing the level of unpredicted with analyses, with the output from analysis A being
behaviours within system design is a clear under- required before analysis B may be performed, and
standing of and the ability to predict, all the so on, there is also a certain amount of flexibility in
interactions, both physical and functional, which the design process which can allow alternative
exist within the system design, and not only pathways to be taken. Further, the fidelity of the
considering those system interfaces which arise as analysis methodology used can also influence the
a result of inheritance. Thus, by complimenting the design pathway which may be taken.
reductionist models with systems models evolving As the system design evolves, there are many
interactions can be identified early and then choices which may be taken with regards to the
controlled and modified. manner in which the design process will continue, as
for example with the fidelity of the analysis. For
3.2. System characteristics and measurements example, is a full FEA analysis required, or will a
simple empirical calculation suffice? Although these
A complex system is composed of multiple sub- choices may be self-evident for an experienced
systems, each sub-system having required analyses engineer, the overall architectural effects of such
performed on them in order to derive their choices may not always be obvious. For example,
attributes [19,20,65]. These analyses usually trans- non-inheritance, analysis-driven interfaces which
late into a corresponding disciplinary analysis, and may be overlooked. The fidelity of the design
neglecting experimental validation studies, these analysis will also influence the resulting output
disciplinary analyses can have corresponding ana- parameter set obtained. If too low a fidelity model is
lysis software (commercial, in-house or legacy). In used, it is possible that subsequent analyses will be
ARTICLE IN PRESS
affected, corrupting the design chain. Equally, in given product, a series of tools need to be made
other cases, a high-fidelity model may be unneces- available to the user [58]. Open frameworks
sary. enabling collaborative multi-disciplinary design
There are several other noteworthy examples of have been recommended and several implementa-
emergent properties which require prior knowledge tions of such environments exist. The framework
of the interactions which exist within the system described in Mawhinney et al. [52,53] and Mon-
design. Cost is often viewed as a method of ensuring aghan et al. [54] presented methods by which the
reliable capital and operating cost assessment, but design process could be approached in order to
alternative methodologies such as the genetic causal adopt a design-driven rather than a model-driven
model costing model [68,69] require knowledge of design procedure. This open framework is based
the system architecture, and is a prime example of within a spreadsheet, operating in a heterogeneous
currently evolving methodologies which are reliant working environment (four scripting languages and
upon an understanding of the entire system design. six different applications), with modelling ranging
Again the ability to identify measurable system from 3D solids through to 1D beams [53]. It is
interactions enables the appropriate architectural capable of performing sequential analysis from the
decisions to be made during the systems design simple empirical calculations performed within the
process. This can be though of in general terms of spreadsheet environment, through to more involved
the system characteristics. calculations performed in CatiaTM and FluentTM.
The system characteristics (cs) are a function of Since that approach is most hierarchical and allows
these system parameters (ps), such that cs ¼ F(ps) as evolution from low fidelity to high fidelity detail
shown by Krus [70]. Both the characteristics of the models it is highly suited in the SE approach.
system and the parameters of the system may be Previously it was described how such a framework
subject to constraints (i.e. the constraints on the can be used within the SE design process but it has
characteristics of the system will effect the system further implications for holistic system measure-
parameters which may be defined, and similarly, ment as will now be discussed.
constraints on the system parameters will affect the Whilst the R–F–P systems breakdowns provide
characteristics of the system). The functional views of the system and links they are predicated on
relationship F is dependent upon the defined the systems engineer’s view of how the product
relationships between the parameters of the set ps, should be structured, informed of course by
and as such in order to define the function in its analyses. Analysis capability or accuracy can mask
entirety, it is necessary to discover what the or reveal behaviour. For example, consider an
parametric relationships are. If these relationships aircraft fuselage section with sub-models of a barrel
are not fully defined, the system characteristics cs segment and a single skin-stiffener element [54]. At
will also be ill-defined. Furthermore, the system the most idealised level this may be represented by a
objective function (fobjf) (such as minimisation of beam model, and hence changes at more local levels
the operational cost) is in general a function of the will be masked since beam theory does not naturally
coupled system characteristics and the system account for such detail. If the functionality of the
parameters fobjf ¼ fobjf (cs, ps) [71]. In order then system is examined, the fuselage has a property
to fulfil the objective function, it is necessary to be relating to the number of passengers it can carry,
able to define the functional relationship, which although its constituent sub-systems (the barrel
may only be done if all of the appropriate system segments, and at a lower level, the skin-stiffener
relationships have been identified correctly. element) do not exhibit this capability on their own.
The level of abstraction will then determine the
3.3. Simulation driven design environments and behaviours of the system which are visible. The issue
analysis fidelity of fidelity is an important one for the determination
of appropriate models/methods to be utilised at
The design framework within which the systems each stage of the process, as the examination of high
and analyses evolve is itself not a single entity, but a fidelity models using high fidelity methods is often
combination of software and hardware linked and computationally intensive, and in some cases,
controlled in such a way to promote an efficient inappropriate. The consequence is that the assumed
working environment. In order to be able to easily system behaviour is based on the underlying
and effectively develop such a framework for a analysis models and as the real product evolves it
ARTICLE IN PRESS
may exhibit behaviours which are unexpected, and level’ system entities in this case are the physical
possibly unwelcome. constraints placed upon the system (the independent
The general form of these observations is useful in parameters), which then flow-down through the
the context of systems modelling and measurement. architecture to their sub-ordinate dependent para-
An entity within a system has a series of attributes, meters, related through their analysis procedures.
an input parameter set (dependent and independent These affect both the system characteristics and
parameters which constrain the type of analysis performance.
which may be performed), an analysis methodology The system ‘technical’ architecture is an extension
(which imposes limits upon the fidelity of the overall to the description of the system design. Much in the
design), and an output parameter set (which will same way the functional architecture describes the
possibly form constraints for subsequent analyses). ‘function’ of each system component and the
The entity may be governed by one or several of physical architecture includes a description of the
these relationships, many of which are recursive in components which perform these functions, the
nature. The definition of these input-analysis- technical architecture contains a complete set of
output relations (Fig. 19) enables the design process rules which govern the interconnection and inter-
to be logically mapped to the physical system dependence of the elements of the system so that
architecture. The sequencing of this process is also they will achieve the requirements. The development
independent of the manner in which the physical of this technical architecture is a hierarchical
architecture is decomposed, ensuring that the process, much the way the other three views are
potential interfaces are not overlooked. The rela- developed. This technical architecture describes the
tionship between the analysis input set (D(z)) and hierarchical design process, linking the parameters
the output parameter set (F(x)) can be described by which are used within the design process to the
a function, where F(x) ¼ f(D(z)). The function f is systems that they describe (Fig. 20). The role of the
given by A(z) (i.e. the function f is described by the technical architecture is really to characterise the
analysis routine which operates on the input data set behaviour of the system.
to produce the dependent data set F(x)). In practice
this means that analysis fidelity is taken into 3.5. Identifying interactions
account and therefore all measurable interfaces
become visible. Supplementing standard architectures with a
technical one that includes analysis methods opens
3.4. The technical architecture the way for assessing interactions since measuring
capability is now accounted for. By accepting that
Having considered the impact of system break- that the level of modelling abstraction affects the
down, measurement and the design environment it outcome of any system analysis it becomes clear
is now possible to look at the system characteristics that interactions will be either included or omitted
more generally and consider what can be deter- according to analysis fidelity. If the system is
mined about them. analysed at a high level of abstraction, then many
The development of the technical architecture is of the interactions between systems may not be fully
very different from the flow-down R–F–P architec- explored, and as such the behaviours of the system
tures commonly encountered in SE. The initial ‘high may be described incorrectly. Conversely, if a low
level of abstraction is used, it is possible that while
the system will be fully described, that the level of
detail may be too much, increasing design turn
around time and cost to unacceptable levels. Within
the technical architecture, parameters may ‘skip’
generations (i.e. a dependent/independent variable
need not always appear at every stage within the
architecture, but may flow down without having
any impact on the intervening stages), and not all of
the lower level entities are dependent parameters;
new independent parameters may be introduced at
Fig. 19. Relationships between system entities. appropriate points within the architecting process.
ARTICLE IN PRESS
Fulfilling a need; by doing; physically described by
R (0) F (0) P (0)
R (1.1) R (1.2) F (1.1) F (1.2) P (1.1) P (1.2)
Function Performed to Physical System to

Fulfil a Requirement Fulfil a Function
T (0)
Provides Initial High Level Describes characteristics

Parameter Conditions T (1.1) T (1.2) of Physical Systems
characterised by behaviour…
Fig. 20. Relationship of RFP architectures to the ‘Technical’ architecture.
The relationships between the ‘entities’ (the

design parameter sets—input and output) is deter-
mined by the analysis methodology applied to the
input parameter set, and as such the relationships
may be altered by changing the type and fidelity of
the analysis methodology (simple analytical/multi- D1 D2
body/FEA/CFD models). The change in the rela-
tionship type will also have an effect on the overall
technical architecture, modifying the pathway A1 A2
through the design process which is taken, as S1 S2
different design methods may introduce different
dependencies, and as such change the dependency S1/S2=1:n where n=number of systems in the architecture
A1/A2=1:m where m=number of available analysis methods
chain. This also opens up the possibility of D1/D2=0:p where p=number of degrees of freedom.
interactions between systems ‘disappearing’ as de-
Fig. 21. Description of technical inter-system relationships.
sign analyses are changed. The simplest description
of the technical relationship is that each system
within the physical system architecture will poten-
tially interact with another system in the same or there is no interaction (0). If there is an
hierarchy (Fig. 21). If the relationship between S1 interaction, then this interaction will be dependent
and S2 is equal to 1 (a relationship between the two upon the analyses that the parameters are passing
systems exists), then the nature of this relationship is between (A1 and A2), and the degrees of freedom of
then determined by the nature of the analyses (the each of the analyses (D1 and D2).
analysis type (A1) that output the parameters from Thus, consideration of the effect of the analysis
S1 and passed them to S2 to undergo a further fidelity on the systems architecture and the identi-
analysis (A2)), and further described by the fidelity fication of measurable links between system ele-
of that analysis model used (i.e. beam, shell, solid ments will provide a capability to identify the
model, etc.). S1 and S2 indicate the interactions impact of design changes and the eventual observed
between the systems arising from the technical behaviour of the system. Early et al. [58,59]
parameters (if there are m systems within the implemented these concepts in a relational database
architecture, then two systems either interact (1) which contains the systems elements and their
ARTICLE IN PRESS
associated sets of input and output parameters.

They give examples of how this can be used in
design to trace the influence of design parameters
and their impact on other system elements. An
interesting aspect of this work is that secondary and
tertiary effects often arise which can result in
apparently local design changes having major
system impacts.
3.6. Parameters versus constraints
The issue of the impact of design parameter

changes raises other problems. In particular, the Fig. 22. System flexibility, change and time [59].
effect of constraints as they flow down through the
system. Having too many or tight constraints
restricts the solution domain, but having too few
where every parameter is a variable can leave the cycle has an increased lifespan, then it must be
design space too open. This leads to the concept of capable of coping with uncertainty and change in
system flexibility which is important particularly for order to be accepted as ‘flexible’. In order to design
systems which have a long life and will be subject to for flexibility, it is necessary to consider flexibility as
change over time. For example, current complex a property of the system which allows it to respond
engineering systems are being designed for increas- to changes in its initial objectives and requirements,
ingly longer design lifetimes. In most instances, occurring after the system design has been initi-
increasing a system design lifetime is driven by (and alised. In order to distinguish between robustness
often justified by) traditional economic analysis, but and flexibility, the relationship between the envir-
this logic ignores the effects of rapid technology onment in which the system exists, and the
advancement, and systems becoming obsolete long objectives of the system needs to be considered.
before they are due for retirement. In many cases, The optimised design is one in which the environ-
the initial circumstances from which the original ment is known and the objectives are unchanging—
system requirements were derived have changed or the most common approach for short lifespan, low
been modified, and as such flexibility is a key value products. The distinction between robust and
property which should be embedded in so-called flexible design comes within the system objectives—
‘high-value’ assets. a system design capable of meeting its objectives
Flexibility is understood as the ability to respond within a non-static environment is termed ‘robust’,
to change, but although this is essential, it is not a whereas in order to cope not only with changing
property which is distinguishable from robustness environmental issues, but also changing objectives,
when described in this manner. Flexibility is a the system must be flexible. This relationship
subjective description, which is particular to the between flexibility and robustness of a design as a
system under consideration, but there are some key function of the system objectives and environment
features which can be used to describe the system can be depicted graphically (Fig. 23).
flexibility [70]. There should be an understanding of Given this, the development of a flexible frame-
the time reference associated with the change (when work must incorporate both the capability to adapt
it is happening during the life-cycle of the system); a to both changing requirements and a changing
characterisation of the manner in which the system environment, leading to a requirement to develop
is changing (the environment, the system itself or methodologies which enable engineers to predict
the customer needs); and finally, and most impor- whether a system is capable of fulfilling these
tantly in the context of this study, the ability to rank requirements. Quantitative methods for predicting
different designs according to their flexibility the ability of a system to cope with change has
through metrics. previously been considered in detail in the realm of
Fig. 22 indicates the relationship which exists software engineering, with suites of applicable
between the three major system concerns of time, metrics identified in order to measure identified
uncertainty and flexibility [20]. If the system life characteristics of flexible systems.
ARTICLE IN PRESS
not appear to be a formal approach to addressing this

problem within the aircraft design arena, other than
what is available through robust design methods. This
Changing
System Objectives
Poor Design Flexible Design

is a difficult challenge since by their nature they are
unpredictable. However, it is possible to at least
identify where limitations of knowledge are, by having
the capability to trace the bounds of parameters and
Fixed
Optimized Robust Design functions. This is a possible avenue for progress. This
Design
better definition of the design and operating space
reduces uncertainty and will facilitate testing system
Environment response to unexpected inputs and system changes.
Fixed /Known Changing/Unknown
3.9. Risk
Fig. 23. System flexibility [59].
The SE process has a strong element of risk
The general area of change management has management embedded in the design process.
become more topical in recent years and a body of However, much of the estimating task remains in
work is beginning to emerge on how to incorporate the domain of specialist engineers and their
system changes and deal with the consequences technical knowledge of the modes of failure of a
[72,73]. This capability will be needed in particular given element. There are software tools which can
for future applications on very long lived systems as estimate risk but these are oriented towards soft-
aircraft are now becoming. ware and are not integrated tightly with the design
process. The technical architecture and tools men-
3.7. Bounding the design space tioned here can provide some capability to have
elements of risk assessment linked with the design
This simple view of using relational databases is synthesis process but again this topic has yet to be
basically an alternative view on set-based ap- an integral part of the design process.
proaches. In the case of systems the parameters
and interfaces themselves become the focus of 3.10. Unspecified architectures in dynamic systems
attention rather than the absolute values of the
behavioural or performance models. This set-based Much existing work on system measurement and
approach was successfully used by Sobek and Ward control relates to having a given system architecture
[74] from Toyota when they used set intersections in place. However it is now widely recognised that
and overlaps to obtain alternative solutions in over a life-cycle that architectures behave dynami-
subsystems which converge to eventually satisfy cally, changing in their nature and behaviour. This
initial requirements. Early et al. [59] presents this in is true even for quite fixed products as it is for more
a more generalised form where the parameter sets obvious dynamic systems, such as defence forces. It
are intersected to identify a generic design space. is interesting that much work focusing on char-
Both variants advocate beginning with a very open acterising dynamic systems to obtain better under-
design space and gradually closing it down. How- standing of their behaviour over time, with concepts
ever, it is clear that to achieve this in general relating to how to add or remove system elements
requires a capability to measure the system and and what affect this has on the system. Interaction
identify relevant interactions. tools and set-based approaches have a role to play
here in that it is necessary to fix an architecture.
3.8. Reducing emergent properties Interaction tools can identify physical and func-
tional interfaces, and therefore specify the architec-
As stated previously complex systems exhibit ture at any given instant.
behaviours, which none of their subsystems have,
due to the interaction of those subsystems [60–63]. 3.11. Summary
These additional behaviours are sometimes necessary
for the functioning of a system, but often emergent This section has highlighted some of the interest-
behaviours are unplanned and undesired. There does ing work ongoing in SE with particular reference to
ARTICLE IN PRESS
how this is being applied to design. It appears that 100 95%

85% Lifecycle Cost
although there is a body of work in software Determination
80 70%
engineering the application to aircraft design is still
sparse. More work is needed to provide the 60
encompassing integration framework needed for 35%
40
aircraft design. 22% Cost Reduction
20 Opportunities
4. Economics and digital manufacturing
Conceptual Detailed Design Production Operations
Up to this point it is the design environment Design Support
which has been the key focus of attention. However,
Fig. 24. Loss of cost control over the life cycle.
the physical construction of the product and its
subsequent life are more important in the end. The
current tight economic circumstances are driving the synthesis, the approach needs to incorporate tech-
need for more effective and efficient design ap- nical cost drivers that flow-down to the required
proaches and awareness of customer requirements design level and physical configuration but must
and their impact on design. The integrated tools also cope with life cycle issues relevant to acquisi-
described previously can now be combined with the tion cost and customer requirements. This includes
latest digital manufacturing and costing tools to a wide range of elements such as non-recurring
produce information on the production, cost and development costs, recurring production costs (in-
life-cycle impact of the product. Virtual factories cluding amortised non-recurring elements), and the
can be viewed alongside virtual product models to recurring operations costs.
reveal assembly and maintenance issues before the The customer requirements detailed in Fig. 8
configuration is finalised. provide the main input to the SE process, which
The main benefits of this are the introduction of then feeds into the requirements loop that iterates
manufacturability and cost considerations into the between Requirements Analysis, and Functional
design arena early in the product development cycle. Analysis and Allocation. The result of the require-
The potential then exists for the definition, valida- ments loop is a functional architecture, with
tion, management and delivery of fully optimised, identified interfaces between all of the identified
cost effective manufacturing solutions to produc- performance and other limiting requirements. A
tion environment quickly and efficiently. In addi- supporting systems engineering costing (SEC)
tion, it will allow the tracking of manufacturing and methodology must first be able to operate at this
component costs as production networks develop level. The functional architecture then feeds into the
from the earliest conceptual design stage. The design loop which iterates between Functional
corresponding impact on risk for the manufacturers Analysis and Allocation, and Design Synthesis.
is major, with many issues being resolved before The design loop sees the functional architecture
production even begins. The paper addresses the transformed into a physical architecture with
need for an approach to cost modelling that will concepts being defined along with configuration
facilitate the development of analysis tools that can items, system elements and physical interfaces; and
operate in complex [75–77] interdisciplinary envir- the preferred product and process solutions being
onments, which are increasingly being developed selected. Therefore, the SEC methodology must also
with a SE philosophy. be able to operate at this level and preferably, to be
recursive in nature for data continuity and to
4.1. Cost and SE maintain the integrity of the cost breakdown
structure (CBS) between levels.
The importance of integrating cost into the
development process is highlighted in Fig. 24, which 4.2. A life cycle view
illustrates the increasing difficulty in influencing
both production and lifecycle costs. Already at the Fig. 25 illustrates the conceptual model of how
end of the conceptual design phase, typically 70% of cost analysis and control can be integrated into a
lifecycle cost is determined while the opportunity to life cycle view of the SE methodology. The product
further reduce costs is only 35%. To provide design life cycle is broken down into four main phases, with
ARTICLE IN PRESS
major milestones associated with the passing from one with instances of redesign due to limited unantici-
stage to the next. It can be seen that the core activities pated emergent behaviour.
of SE, as shown in Fig. 25, are applied sequentially In order to address this issue, Price et al. [23]
during each phase of the life cycle. In addition, the introduced the concept of a system technical
costing that occurs during the design phase is architecture, an additional view of the system in
illustrated as a top down process that matures in which the technical and cost linkages induced by
complexity and fidelity as the design definition system analysis through the design phases could be
becomes fully complete at the Critical Design Review identified, and the effect of modifications on design
(CDR). The design phase is also associated with a then reflected into the more traditionally held
flow-down of the requirements that are then allocated requirements, functional and physical architectures.
within the functional architecture. Subsequent to the Initial consideration of system requirements gives
CDR, the verification phase is illustrated as a bottom- rise to a series of high level design characteristics,
up process where parts and components are fabricated which form the initial high level parameter defini-
and integrated into assemblies and sub-systems that tions for system design (so-called independent
ultimately, are tested, verified and validated at an design parameters, as they have no ancestry within
operational systems level. the system architecture). Through a series of
operations on these parametric sets, a series of
dependent parameter sets are introduced (for which
4.3. Cost as a customer requirement a clear set of parent elements may be determined).
These independent and dependent parameter sets
The architectural make-up or analysis framework provide the technical definition of the system
of the system has a direct influence on the emergent components contained within the system architec-
behaviours which it exhibits, and that is true also of ture. By constructing the architecture in this
cost. In the case of all architectural modelling for manner, a clear and traceable pathway through
system design, the interactions between components the design process is developed, which supports the
give rise to the ‘sum is greater than the whole’ additional architectural models, with clear linkages
property observed in the majority of systems, and in between each, as shown previously in Fig. 20.
most cases, it is this emergence which is designed
for. On the flipside, however, this emergence may 4.4. Relationship back to design: multi-disciplinary
also manifest itself in undesirable behaviours (so- design synthesis
called side-effects), leading to a need to establish
methodology which will effectively identify, and Fig. 26 illustrates the multi-level, multi-fidelity
eventually control and drive system development to framework being developed to achieve design and
produce more cost effective, stable system designs, cost synthesis. On the left-hand side of the figure the
Fig. 25. Systems engineering model for acquisition costing.

ARTICLE IN PRESS
aircraft structural design is first assessed through a affordability, product quality and market timeliness
simple frame analysis, which then is used to develop are the three key elements of competitiveness.
solid models to represent a more realistic geometric Sheldon also points out that there are two funda-
definition. The solid models can then be used to carry mental engineering approaches to controlling cost:
out more accurate analysis of structural integrity and namely, (1) designing for cost and (2) costing for
cost, etc., to be fed iteratively into the sizing design. Within aerospace, Dean and Unal [80] are
procedure. Further levels of abstraction can be well known for promoting such considerations
imposed down to localised detailed modelling, within NASA. Although Sheldon defines the Design
depending on the maturity of the design process in For Cost (DFC) methodology as being driven by
the development stages following the process illu- management imposed cost targets, this is usually
strated in Fig. 25. On the right-hand side of Fig. 26 referred to specifically as Design To Cost (DTC)
the integration of SEC is illustrated with fidelity, [81]; implying that a cost target has to be met and
discipline and time dimensions. This relates directly adhered to. DFC is generally taken to mean that the
to the need for the additional technical architecture design process is mindful of cost. Many authors
presented in Fig. 20 as the integration of quantitative now believe that imposing strict target costing on
modelling, e.g. for cost in this case, requires that engineering design, as for DTC, is not effective as it
multi-fidelity cost models are required to facilitate tends to result in inferior design that still overshoots
the SE approach in order to enable the coupling to the poor cost estimates used as the initial guidance
the other multi-fidelity models that characterise the [82]. It is more important to give designers
multi-disciplinary system design space. supportive costing tools that facilitate the product
definition process by linking design decisions to
4.5. Cost integrated design estimated cost impact.
DFC/DTC tools tend to be application specific
Cost is an important attribute of any product and and highly customised within the aerospace industry
highly relevant to the engineering design process [82–85]. For example, Gieger and Dilts [86] have
[77,78]. Sheldon [79] has stated that customer presented an automated design-to-cost tool which
Fig. 26. Integrating disciplines that drive design at various levels of definition.
ARTICLE IN PRESS
can be linked to a CAD package in order to provide high level or a lower level that links discrete
the estimated cost of machined parts from a computational models [104]. Marx et al. [105] have
particular material. Within aerospace, this would linked MDO to Life Cycle Analysis by defining high
be most relevant to the detailed design process for a level objective functions that encompass the life
range of parts from smaller complex machinings but cycle needs of aircraft, supported by necessary
could be extended to larger fuselage frames of disciplinary models which facilitate the optimisation
machined-finish aluminium forging for example. process through a linkage that is defined by an
The cost tool interprets the machined part using objective function. There are various Direct Oper-
Feature Based Modelling [87] and classes it accord- ating Cost (DOC) models available, which tend to
ingly using Group Technology [88]. Various costing be of a parametric nature [106,107], which allow the
modules then plan and cost the machining process trade-off of design parameters and which can be
using a mixture of Activity Based Costing [89,90] linked to manufacturing models to couple the
and analogous costing in a comparative manner. impact on production [81,108].
Taylor [91] has also advocated a Feature Based The impact of the work of Boothroyd and
approach to aerospace cost estimating and this is Dewhurst [108] in highlighting the need for a
often used in traditional aircraft cost estimating, methodology that links the impact of design
although in a less formalised manner. decisions on manufacture is well referenced. Stoll
In analogous costing deviations in the design or [109] has also addressed many of the organisational
manufacture of a new product are used to account and implementation aspects of DFMA while other
for alterations in the initial cost estimate [91]. Apart authors were also reporting the important linkage
from the analogous, ABC and Feature Based between DFMA and LCC [110].
techniques, there are a range of other methods for The relevance to LCC is still prevalent as stated
generating the actual cost estimates from input data by Murman et al. [111] who defines better–fas-
and constraints [92] including: regression-oriented ter–cheaper life cycle needs in terms of value-
parametrics [93], bottom-up costing, fuzzy logic [94] oriented cost, performance and time functions.
and neural networks [95]. The parametric estimating Marx et al. [105] have presented a parametric
technique [96,97] is widespread within aerospace solution for linking life cycle needs back to design.
and varies greatly from being based on purely A more detailed analysis platform for manufactur-
statistical significance, to being more causal in ing cost drivers has been developed by Rais-Rohani
nature; being either linear, exponential (logarithmic [112], where he incorporates many of the relevant
linearity) or polynomial in form. manufacturing issues in terms of parametrically
Pugh [98] has advised that a top-down cost defined complexity factors; including: compatibility,
estimation should be carried out even before the complexity, quality, efficiency and coupling. Rais-
aircraft development process begins. Thurston and Rohani’s work is integrated into the aircraft design
Essington [99] advocates a holistic approach to the process using a three-tier MDO methodology [113].
design process that is appropriate at the concept Sandoz [114], a chief engineer on the Boeing 747,
stage where a product is defined in terms of a has presented a value-oriented approach to the
measure of its utility value to the customer. This integration of requirements for aircraft structures
includes cost in a multi-attribute analysis [100] of already in the early seventies. Other authors have
the design that can then be mathematically opti- continued to address the impact on manufacturing
mised [101]. Another form of this design methodol- by characterising the various manufacturing pro-
ogy has also been applied by Collopy and Eames cesses for fuselage panel parts [115], along with the
[102] to satisfy the more holistic design requirements associated assembly processes [116], with respect to
of an Unmanned Arial Vehicle (UAV). key design drivers and cost. Much of the work has
In general, this approach can be traced back to again been industrial-oriented and focuses on
much of the classic research within the aerospace assessing the trade-off between technologies or
industry into parametric optimisation [103]: the materials [117].
identification of key design parameters that drive
performance and which can be optimised when 4.6. Approaches to cost modelling
combined in mathematical formula. Much of the
current mainstream research is focused on Multi- Various alternative approaches to cost modelling
Disciplinary Optimisation (MDO), whether at a exist [118,119]. Bottom-up or Detailed Estimating
ARTICLE IN PRESS
(DE) is the least formulated costing technique and drivers is seen in the context of product families.
entails the gathering of all cost information that can The model adopts the scientific principle of categor-
be directly attributed to the final article, to be isation but also incorporates the scientific require-
cumulated into a highly detailed model. Analogous ment of utilising causal relations. Essentially, the
Costing and Case Based Reasoning (AC/CBR) are generic methodology has the following principles:
techniques that principally rely on the similarity and
differentiation of like-products to ensure that the 1. Genetic principle: Cost is classified into families
cost estimate is comparative to a previous instance. according to product and process, to identify
An automated form of analogous-type costing uses likely commonality through shared cost drivers.
Neural Networks and Fuzzy Logic (NN/FL) to 2. Causality principle: Cost is formulated into a
formulate relations that link independent product- relation as a function of the design attributes which
related attributes to the dependant costs under gives rise to a cost potential, using: weights, part
consideration. Parametric Estimating (PE) typically counts, sizing, and material selection, etc.
entails the linking of cost to high-level product
parameters through probabilistic analysis in order
to establish estimating relations that will combine The concept of using the Causal principle to
into a cost estimating model [120,121]. Commercial establish relations with a strong scientific basis is
versions, such as PRICE-H and SEER-DFM, offer clear, however, not well practiced in cost estimating
a facilitating environment and functionality that where statistical significance is often used as the
allows an organisation to calibrate the model with primary requirement. This entails a high degree of
their own historical data in order to tailor it to their risk as there is rarely an attempt to define the range
specific financial needs and business environment. of application of the model or its sensitivity to
However, as well as historical data, the calibration design and manufacturing change, relative to the
process requires expert judgment, assumptions and historical data used to establish the statistical
opinion [122–125]. Finally, there are financial relation. In terms of the Genetic principle: within
accounting techniques, such as Activity Based aerospace, design tends to be derivative and there-
Costing (ABC), which record the use of resources fore, the costing blueprint can be seen in previous
in determining a more accurate estimate of the total
cost, including capital cost rates and overheads.
It has been established that ideally, any facilitat-
ing costing methodology should be able to operate
and interface at various levels and during all stages
of the life cycle. This has been a fundamental driver
in the development of a generic approach termed
Genetic Causal Costing (GCC) [126–130], which is
shown in Fig. 27 in application throughout the life
cycle. This is conceptualised in Fig. 28 where the
causal definition of the relation of cost to design Fig. 28. The genetic-causal costing model.
Techniques:
Life cycle phase: DE AC/CBR NN/FL PE ABC GCC
Concept and technical development
System development and demo
Production and acquisition
Operations and disposal
Fig. 27. Matrix of cost estimating techniques.

ARTICLE IN PRESS
aircraft. Engineering manufacturing costs can be fied as a cost family or category while the
classified according to materials, fabrication processes independent variables should be of a causal nature.
and assembly, while additional costs can be grouped It is well known that weight is a good indicator of
according to support, quality and inspection, and cost and is often used in parametric cost analysis
general factory overheads. There is also a large [131]. However, it is recommended that part count
distinction between recurring and non-recurring also be considered as a more causal driver. For
costs. A broader life cycle analysis should consider example, weight is more loosely coupled to cost in
the non-recurring cost of the development process the fact that ‘more stuff is likely to cost more’, i.e.
plus additional company recurring overheads, all of more material requiring more design effort in order
which should be reflected in the acquisition cost. to define it. However, in this instance part count,
However, the key aspect to this understanding of cost even with a slightly lower statistical significance
is that it is genetically inherited and that it originates (R2 ¼ 0.92 as apposed to R2 ¼ 0.93 with weight as
through the engineering design definition, which the independent variable), is assessed to be more
dictates the materials to be processed and assembled causal as it relates directly to the number of
into some product configuration. Environmental drawings to be generated to define the assembly
factors, such as machining rates and labour rates, configuration.
can be assumed to be fixed at the product definition
stage, although logistics and supply chain manage- 4.7.2. Production cost: in-house
ment introduce a realistic variance. Following the Genetic-Causal principle, cost can
be disaggregated into families according to material,
4.7. Cost modelling at various stages of the life cycle treatments, fabrication, and assembly; each of
which can be further broken down as required.
4.7.1. Development cost The plots to the left on Fig. 30 show the make cost
Development cost is largely treated as a non- (fabrication) being related simply to stringer length,
recurring element, i.e. a one-off effort that is not for a particular family of stringers, whether T, Z, L
repeated, although in practice design modification Section, etc. The regression analysis of the data
does occur during the production stage as the requires both a constant and a slope in order to
manufacturability of the design solution becomes characterise the relation to some linear approxima-
evident. Fig. 29 presents some data relating to the tion, shown in the upper and lower graphs,
non-recurring design effort recorded for fuselage respectively. Subsequently, the distributions (uncer-
sections. The design hours, as the dependant tainty characterisation) for these statistical compo-
variable, are being modelled as a function of both nents can be utilised through Monte Carlo analysis
part count and weight. It is highlighted that there is in order to model the likely variation in cost due to
a minimal 15% delta in the cost being predicted the variance observed (or imposed relative to some
depending on the independent variable. In terms of criteria), thereby producing a more robust cost
the Genetic-Causal approach, design cost is identi- estimate. This is illustrated to the right where both
distributions have been used by the Monte Carlo
analysis to provide the rigorous assessment of the
3000 1.2 likely distribution of cost.
Part count R2 = 0.9176
2500 Weight 1 4.7.3. Production cost: procurement
Normalised weight
2000
R2 = 0.931
0.8 Currently, 70–80% of aerospace parts are pro-
Part count
cured from the supply chain. Procurement cost

1500 0.6 modelling is characterised by the analysis of large
data banks of part prices and often only very basic
1000 0.4
information regarding the manufacture, that being
500 0.2 too sensitive or proprietary in the opinion of the
± 15% supplier. Fig. 31 shows some results obtained from
0 0
0 0.2 0.4 0.6 0.8 1 1.2
various techniques of linking the procured cost for
Normalised design hours
the causal drivers identified, using both regression
analysis and neural networks to develop the
Fig. 29. Modelling design cost using high level cost drivers. formulation. The RMSE value shown can be
ARTICLE IN PRESS
100
90
Make cost [£]
80 Frequency
y = (Ax + (B + sB) Cumulative %
70 Histogram AUTR
250 120%
60
y = (Ax + (B - s ) 100%
B 200
50
80%
Frequency
40 150
60%
30 100
40%
0 50 100 150 200 250 50 20%
Stringer length [in]
100 0 0%
0.00260
0.00290
0.00321
0.00351
0.00382
0.00412
0.00443
0.00473
0.00504
0.00534
0.00565
0.00595
0.00625
0.00656
0.00686
0.00717
0.00747
0.00778
90
Make cost [£]
80 Bin
y = (A + s ) x + B
A
70 Monte Carlo analysis
60
y = (A - sA) x + B
50
40
30 Uncertainty characterisation
0 50 100 150 200 250
Stringer length [in]
Fig. 30. Modelling production cost at a cost component level.
3.0
2.5
2.0
RMSE Value
1.5
1.0
0.5
0.0
1 2 3 4 5 6 7
Group Number
Neural Network Test Data Regression Neural Network Training Data
Fig. 31. Predictions for procured cost based on neural net and regression techniques.
interpreted as the non-dimensional error, where 4.7.4. Operations cost

Group Numbers or levels 6 and 7 represent the Fig. 32 presents a pie chart detailing the propor-
much small data bins for a particular part family, tional contribution, with ownership being the
machined bulkheads in this case. Group 5 analysis is financing of the acquisition cost at 54%. The
returning an accuracy level of approximately 5–10% engineering effort expending in the early develop-
for all the parts considered in that larger bin, while ment stages influences ownership, maintenance
Groups 6 and 7 show that machined bulkheads are (13%) and fuel costs (15%), while crew (13%),
not a particularly well modelled sub-category. This landing fees (2%) and insurance (3%) are largely
reduced accuracy is due to the lack of design independent. It is striking that the chart supports
instances and to the wide variation in design the view that more effort is required in reducing
definition, relative to the bulk of parts considered acquisition cost as it has over 3 times the influence
in Group 5. of performance (fuel burn).
ARTICLE IN PRESS
Insurance manufacturing cost, whose inclusion into the

Landing 3% operating cost is what is giving the increased saving
2%
over the minimal weight criteria.
Maintenance
13%
4.8. Virtual product design and manufacture
Digital manufacturing tools can be used to link

Fuel
15%
Aircraft the design Bill of Material (e-BOM) to the
54% manufacturing BOM (m-BOM), with each being
regarded as different views of the one BOM, thereby
reducing the cost of consolidation currently re-
Crew quired with separate design and manufacturing data
13% systems. The manufacturing BOM can then be
indented and configured into a hierarchical Work
Fig. 32. Typical breakdown of operating cost for a regional jet.
Breakdown Structure (WBS) that can then be used
to formulate network planning in order to optimise
• Wider stringer spacing the work flow. Key products of this process are the
• Heavier stringers
Importance of manufacturing
• Thicker skin
Integrated Master Plan (IMP) and Integrated
Master Schedule (IMS), which are used to manage
saving in direct operating
4000
3500 this effort.
3000 A Product–Process–Resource (PPR) hub is used
cost US $/m2
2500 at the heart of the system to integrate the design

2000 BOM with the manufacturing BOM, the parts being
1500
attributed with additional manufacturing and cost
1000
500
information at that stage. The assembly process is
0 then defined as consisting of some combination of a
min min min min
W doc
number of operations with associated precedence,
mat mfc
W = total weight, mat = bare material cost,
time and cost. The network analysis can then be
c = total manufacturing cost, doc = direct operating cost) carried out to determine the optimal detailed work
flow, thereby fully defining the most cost effective
Fig. 33. Modelling operating costs according to various optimi- solution to each task identified in the Integrated
sation criteria. Master Plan.
Verification and validation is crucial to any
Fig. 33 presents the results of an optimisation company improving its business and can be used
addressing the trade-off of manufacturing cost and to review and modify the design under considera-
operational cost for a fuselage panel pictured on the tion and its manufacture. However, design mod-
left (also considered in Section 4.9). The impact of ification is a very expensive procedure, especially
manufacturing cost on DOC as an optimisation due to the aircraft certification procedure. Also,
criterion is highlighted, as well as the significant such information could be used at a much earlier
change in the design solution; with wider stringer stage to find a more optimal solution in a less
pitch, heavier stringer design and thicker skin constrained design space. This is made possible
thickness. The cost is being assessed in terms of through manufacturing simulation, which allows
the reduction in operating cost (US$/m2 of panel) detailed manufacturing processes to be accurately
from some baseline design position, where the modelled and integrated into the early development
optimisation criteria is either minimising weight processes and thereby providing a virtual verifica-
(W), material cost (mat), manufacturing cost (mfc) tion and validation capability. Although true
or operating cost (doc). It is not surprising that verification and validation process will not be
minimal material cost is not a good criterion for complete until production has commenced, the
reducing operating cost while operating cost is some manufacturing simulation is linking the value of
20% better than both weight and manufacturing that process back into the design currently being
cost. The parity between weight and manufacturing developed, thereby reducing its cost through design
cost is further evidence of the importance of and manufacturing optimisation.
ARTICLE IN PRESS
Fig. 34. Simulation as a tool for verification of optimal manufacturing solutions.
Fig. 34 shows plots of the process and operator Stringer area

efficiencies (left and right graphs, respectively) b =2hts
simulated for the fuselage section shown in
Fig. 34; fitting tending to be higher than riveting in
ts
the plots. It is highlighted that the utilisation h
improvement is resulting in more steady and efficient
labour deployment. It can be seen from the process t
plots to the left in Fig. 34 that the two processes
represented, fitting and riveting, are being used at Fig. 35. Modelling of the panel for structural.
much more stable and constant utilisation rates. This
concurrent planning thereby reduces cycle time and
maximises the effective utilisation of labour. The borne by the operator. Ultimately, there was a
plots to the right-hand side highlight the latter point trade-off between driving design according to
where it can be seen that the fitting and riveting minimal weight and driving it according to reduced
processes required a more constant number of manufacturing cost. The analysis of cost was
operators after the optimisation. The stepping in facilitated with a Genetic-Causal cost modelling
the upper plot signifies varying numbers of operators methodology and the structural analysis was driven
throughout the assembly process, who can not be by numerical expressions of appropriate failure
redeployed in the short terms and therefore, are modes that utilise Engineering Sciences Data Unit
frequently idle, reducing the process efficiency. The (ESDU) reference data. However, this preliminary
improvement illustrated resulted in a 20% improve- study did not consider surplus value and manufac-
ment in the assembly cost, within the designated turer’s profit, which are focused on in this paper.
cycle time constraints for required throughput. The panel was modelled as shown in Fig. 35 and
the active design variables were chosen to be:
4.9. Cost optimisation and design trade-off stringer pitch b, stringer height h, skin thickness t,
stringer thickness ts and rivet pitch rp.
Rather than only looking at manufacturing cost, The panel was first optimised for maximum
a preliminary study showed that DOC, which is the theoretical efficiency, Z. In fact this is equivalent to
cost of transporting a given weight of aircraft minimising the cross-sectional area of the skin and
structure during the aircraft’s life span, could also stringers, but excluding cleats, rivets and other
be considered in terms of the impact of weight on secondary items. The optimisation was then carried
fuel burn, in addition to the acquisition cost to be out for minimum total weight, minimum material
ARTICLE IN PRESS
W = total weight, mat = bare material cost, mfc = total manufacturing cost, doc = direct operating cost.
Panel
Optimised for η b h t ts r0
efficiency η 0.693 42.8 27.6 0.85 1.61 31.1

Minimum W 0.632 71.5 31.0 1.6 1.6 61.6
Minimum mat 0.628 65.5 27.0 1.09 2.53 41.9
Minimum mfc 0.383 192.3 38.64 2.52 6.07 124.7
Minimum doc 0.517 125.1 28.2 1.97 3.73 83.7
All dimensions in mm (values in italics indicate that the limits of validity in local buckling have been reached)
Fig. 36. Panel dimensions after optimisation.
cost, minimum total manufacturing cost and mini- (W = total weight, mat = bare material cost,
mum DOC. It was found that the various criteria for mfc = total manufacturing cost , doc = direct operating cost)
optimisation lead to widely differing panel dimen-
Panel Saving in:
sions (see Fig. 36). Minimisation of direct operating
optimised for:
cost leads to a stringer pitch almost triple that of the W mat mfc doc
theoretical optimum, at the same time more than
minimum W 1.60 -11 807 2898
doubling the skin and stringer thickness and the rivet
pitch. Increased stringer pitch implies a reduced minimum mat 0.99 36 680 2335
number of connecting cleats, and this together with minimum mfc -2.29 -108 1186 2872
increased rivet pitch leads to substantial cost savings
in assembly. The key observation is that the minimum doc 0.58 -26 1122 3539
optimisation is tending to drive the design towards
All cost savings are in US $ per m2 of panel, weight in kg per m2
fewer stringers that are larger and with a greater rivet
(values in italics indicate an increase)
pitch, which is consistent with the recorded impact of
riveting on manufacturing cost. Fig. 37. Savings according to the choice of objective.
The results in terms of DOC are detailed in
Fig. 37 where the first column shows the quantity
minimised in the optimisation, and the others present that the operational cost of fuel burn will go up
the relative changes. Positive values denote reductions making the aircraft more susceptible to increased fuel
relative to the reference panel, while negative values costs as well as being less environmentally efficient.
indicate an increase. There was a substantial reduction
in both weight and DOC when the panel was 4.10. Conclusions
optimised for minimum total weight, rather than for
maximum theoretical efficiency. The minimisation of This section has presented a methodology for the
material cost and total manufacturing cost, also show life cycle integration of cost into the SE approach,
improvements with regard to DOC although the termed systems engineering costing (SEC). This has
minimisation of manufacturing cost induces a weight included an assessment of cost modelling technol-
penalty. Optimisation for minimum DOC rather than ogy, and the application of cost modelling to a
for minimal weight shows a further improvement of number of key tenants of SE, including: Require-
10% of the total DOC. This might be considered a ments and Functional Analysis; Life Cycle Analysis;
favourable result, since much structural optimisation Multi-disciplinary Design Synthesis; Risk Manage-
is performed for minimum weight; it is being implicitly ment; Cost Management; Planning and Scheduling;
assumed that this also reduces cost. The actual saving and Verification and Validation. Cost management
of 660$/m2 would relate to a rough order of components were disaggregated according to the
magnitude DOC saving of $350,000 for the complete development stage, production stage (including in-
barrel section of the fuselage. house and procured), and operations stage. The
The price of the aircraft is reduced to make the integration of cost into SEC has been facilitated by
producer more competitive and the customer gains the Genetic-Causal costing model and ultimately, it
early through reduced acquisition cost and its has been shown that this can be achieved to good
through-life ramifications in financing. However, result. The key principles of cost categorisation and
there is a shift in responsibility to the customer in causality were highlighted as the fundamental
ARTICLE IN PRESS
tenants of the Genetic-Causal modelling approach. performance and safety measures [110,136]. A typical
Therefore, it is concluded that the SE approach offers DOC breakdown of the Airbus 320 class of aircraft
a facilitating framework for the effective integration of (150 passengers and 2800 nautical mile range, Fig. 38
cost into the life cycle balanced design of aircraft. shows that the aircraft cost contribution to DOC is
more than four times that of the fuel contribution.
5. Collaborative design and the virtual enterprise This is even more extreme for smaller aircraft with a
lower payload range. It is understood in the aerospace
The preceding sections have addressed issues industry that any reduction in DOC requires not only
relating to integration of technical and economic a better understanding of aerospace disciplines but
tools to facilitate design. However, in today’s also an understanding of the sensitivity of each
industrial context aerospace development takes discipline to the other. Work in this are began with
place across company, national and international a novel concept of trade-off study between costs
boundaries. There is a minefield of problems in associated with two design drivers, aerodynamics ,
relation to collaborative working, ranging from and hence fuel, and manufacturing were conducted by
technical data transfer to intellectual property rights Kundu et al. [136,137], Curran et al. [138], Kundu et
and even to national security. This is, in many ways, al. [139] and Woods et al. [140]. The concept was
the real challenge: to develop a virtual enterprise extended to trade-off studies between wider ranges of
with virtual collaborative design environments design drivers; including aerodynamics, structural
[132,133]. The technologies summarised in this configuration, manufacturing and assembly as pre-
paper are pre-requisites to this but there are also sented in Sanchez et al. [141].
major research initiatives in both Europe [134] and A particular component of an aircraft was chosen
the USA [135] which are addressing this. This topic for a detailed programme of research. This was the
is too broad in itself to cover within this paper. nose cowl of two generic nacelles. The road map for
DOC estimation consisted of the development of a
6. Applications and case studies rapid cost model based on the baseline design and
validated against the later design. Cost reduction
The foregoing sections have shown how the design measures for the structural part fabrication and
process can be made more inclusive and that their assembly was investigated. The savings were
integration across disciplines and through depth can extrapolated for the aircraft and corresponding
be achieved and can provide more flexibility in the DOC savings for the mission profile were estimated.
design space. However, this is not the whole story.
The understanding of the physics of problems where
complex systems interface is, in itself, a challenge. The 6.1.2. Development of an optimised design to reduce
practical issue of getting data both for physical design nacelle manufacturing cost
and also commercially sensitive cost and intellectual The major objectives involved: (a) development
property data inhibits the development of new generic of an accurate cost model and (b) development of a
rules for design. In this section a more detailed view
of the practical implementation of this integrated Depreciation Nav & Land.
21% 16%
philosophy are presented through two case studies. A
third case study is included which briefly reviews the
progress and challenges of integration in under-
Crew
standing the physics of the wing–pylon interface to 20%
reinforce the need for linkage between design/analysis
disciplines and the SE process. Maintenance
13%
6.1. Integration of aerodynamics and manufacturing
for cost reduction Insurance
1%
Fuel
6.1.1. Introduction
12% Interest
In recent years, customer needs in the commercial 17%
Aerospace sector have been governed by reduced lead-
time and cost, while also looking for enhanced aircraft Fig. 38. Typical DOC distribution [141].
ARTICLE IN PRESS
methodology for evaluation of the cost benefits of the in-house capability issues they were not con-
Design for Manufacturing and Assembly (DFMA) sidered in this project.
in addition to tradeoffs between aerodynamics and The methodology applied was based on the
manufacturing. factors/indices of nacelle B, in relation to the
A rapid cost-modelling methodology was devel- baseline cost of nacelle A, and the associated cost
oped that was specifically aimed at catering for the drivers mentioned earlier. The total manufacturing
industrial needs of DFMA during the conceptual cost of the nacelle was the sum of the individual
design phase. The Rapid Cost Model was based on costs of each of the four nacelle components, as
parametric methods taking into consideration the given below for the nose cowl cost CNC, being the
effects of engineering decisions on cost and the link sum of the following six items:
between design statistics and manufacturing cost. In X5
the process of cost modelling, methodologies were C NC ¼ C 0 ¼ C 0Mat þ C 0Fab þ C 0Asm
i i
also established for cost data collection, definition þ C 0Sup þ C 0Amr þ C 0Misc ,
and accounting. From a baseline cost, the method
demonstrated a fast and relatively accurate predic- where subscripts ‘Mat’ stands for material, ‘Fab’ for
tion for the later design. fabrication ‘Asm’ for assembly, ‘Sup’ for support
The task concentrated on cost modelling of the ‘Amr’ for amortisation and ‘Misc’ for miscellaneous
structural elements of generic nacelles. The task was cost.
conducted in two steps. First, a detailed cost model Methodologies were developed for each of the six
was developed for the ‘Nose Cowl’ and then the cost components. For example, nose cowl manu-
model was extended to the complete nacelle. Two facturing cost C 0Man consisted of two items: parts
nacelles, A and B, were considered in the modelling. fabrication and parts assembly to finish. Man-hours
Nacelle A was an existing product and that was required for the fabrication of each part and
taken as the baseline design, whereas nacelle B was a assembly were a combination of operations: ma-
newer design with higher specification standards chining, forming, fitting and mounting into jigs.
and 50% more thrust than the baseline turbofan of Manufacturing cost was expressed as
that family of nacelles. The aerodynamic mould-
lines of both the nacelles were similar, but their Manufacturingcost ¼ rates manhours
structural design philosophy, hence the sub-assem- learning curve factor
bly (tooling concept) differed. The manufacturing size factor manuf: philosophy:
cost of the finished product consisted of: cost of
material (raw and finished product); cost of parts or
manufacture; cost of parts assembly to finish the Xm Xn
product; cost of support (to ensure quality); C 0Man ¼ C 0Fab þ C Asm ¼ ½ðK size Þ0:5 i
½ i F 1 F 2 . . . F n m
amortisation of non-recurring costs; and additional ðman hour rates learning factorÞFab
miscellaneous cost for contingencies, etc. In that Xm Xn
work the actual cost data are classified ‘commercial þ ½ðK size Þ0:25 i ½ i F 0i ðman hour rates
in confidence’, so relative results were presented. learning factorÞAsm .
Structural components and Engine Built Unit
(EBU), e.g. anti-icing units and valves, were This analysis led a simple tool for the calculation
considered in the cost modelling, but the EBU costs of nacelle B nose cowl cost, as given by
were separated and are not included in the presented
research. Eleven cost drivers in two groups were C NC_B ¼ 0:8306Mat_NacA þ 1:0878C 0Fab_NacA
identified for the analysis. Group 1 related to in- þ 0:759C 0Asm_NacA þ 0:05 ð0:828C 0Mat_NacA
house data within the organisation and consisted of
eight cost drivers, namely: Size, Material, Geome- þ 1:0878C 0Fab_NacA þ 0:759C 0Asm_NacA Þ.
try, Technical Specification, Structural Design
Analysis of industrial cost data showed that the
Concept, Manufacturing Philosophy, Functionality,
nose cowl A cost fractions were as follows:
Man-hour rates (overheads, etc.). Group 2 cost
drivers consisted of Role (e.g. military or commer- C 0Mat_NacA =C NoseCowl_A ¼ 0:408; C 0Fab_NacA =C NoseCowl_A
cial), Scope and Condition of Supply, and Pro-
gramme Schedule. As they were not concerned with ¼ 0:349C 0Asm_NacA =C NoseCowl_A ¼ 0:149.
ARTICLE IN PRESS
Hence the relative cost of nacelle B was expressed ing cost, as mentioned previously. A trade-off study
as between fuel cost and manufacturing cost was
performed on an isolated nacelle and the analysis
C NoseCowl_B =C NoseCowl_A ¼ 0:92. extended to a complete aircraft. Manufacturing
tolerance relaxation at eleven key manufacturing
Analysis of components such as the fan cowl,
features on the surface assembly of an isolated nacelle
thrust reverser and tail cone followed the same
was studied without unduly penalising parasite drag,
procedure as the nose cowl. In developing this
using the cost model developed. The drag estimation
methodology several aspects of manufacturing and
was based on ESDU data corrected for pressure
assembly were taken into consideration. These
gradients determined by CFD and experimental
include, part count for the assembly, man-hours
studies. The drag data and aircraft operation were
involved in fabrication and assembly, rework/
used in evaluating the fuel cost. A large amount of
concessions and quality of the surface finish. The
manufacturing cost data was obtained and analysed in
analysis of certain components was for re-design
detail for the effect of manufacturing tolerances on
change only, where the material type and manu-
cost. Fig. 39 shows DOC and drag variation with
facturing process were not affected. In that case, the
surface tolerance relaxation. The Direct Operating
frames were manufactured in three sections and
Cost of Aircraft was estimated using Association of
joined with splices and rib clip attachments. These
European Airlines (1989) ground rules. The payload-
were re-designed into one-piece frames that did not
range for the mission profile was kept constant
require the use of the clips. The results of the
throughout the aircraft performance analysis.
approach adopted on Nacelle B (higher specifica-
With conservative estimation, given below is the
tions and 50% increase in thrust) led to a part count
typical aircraft cost reduction through such a
reduction of 24% compared to Nacelle A. The
DFMA approach to tolerance allocation, with fuel
increase in structural requirements from the much
price taken at US$0.75 per gallon. The study
higher thrust is highlighted by the fact that fastener
resulted in approximately 1.28% DOC saving for
count actually increased, e.g. by 8% on the nose
a 2% saving in aircraft cost involving no drag rise,
cowl in order to provide the required stiffness.
and additionally, approximately 0.42% DOC saving
However, there was a saving of 12% in assembly
for a further 1% saving in aircraft cost for tolerance
costs due to the reduction in part count, and the
relaxation that did involve drag rise. The total of
increased number of complex components, which
1.7% DOC savings translates into savings of $530
were not manufactured in-house, influenced the
per sortie for the Airbus 320 class of aircraft. With a
material cost increase of 16%. The total reduction
typical annual utilisation of 500 sorties, that totals
in cost was found to be 2% per kilo of structure and
to $265,000 per aircraft. For smaller aircraft the
therefore the evaluation methodology has been
percentage savings could be higher.
applied to show that DFMA has been able to
improve the manufacturability of the design to
6.1.4. Concluding remarks
maintain, and even reduce cost slightly, regardless
In this integrated cost performance study the
of a 50% increase in technical design specifications.
practical issues bringing detailed results through to
early design emerge. The work undoubtedly fol-
6.1.3. Prediction of aerodynamic performance and lowed the SE approach, but in practice the fidelity
cost arising from surface tolerance of the analysis models and availability of data is a
The prediction of aerodynamic performance challenge every project encounters. Indeed the
arising out of surface tolerances involved two tasks: commercial sensitivity of data, particularly costs
(a) trade-off studies between aerodynamics and data, but also design methods looms large as a
manufacturing to optimise allocation of the toler- difficulty in the development of integrated design
ances and (b) prediction of performance and DOC. methodologies.
Aircraft surface smoothness requirements were
aerodynamically driven with tighter manufacturing 6.2. Trade-offs and integration at conceptual design
tolerance to minimise drag, where the tighter the level
tolerance, the higher is the assembly cost in the
process of manufacture. A typical DOC distribution The previous sections illustrated some detailed
shows that fuel cost is only a fraction of manufactur- issues in relation to integration, indicating that there
ARTICLE IN PRESS
Fig. 39. Trade off between aerodynamic and manufacturing tolerances for cost [139].
Fig. 40. Under wing and fuselage mounted engines are compared for their effect on an upper wing panel.
are many technical details which also needed to be models of the aircraft. The final results were
sorted out before full integration is possible, and compared to make the design choice. The optimisa-
even with that there will be elements of under- tion and cost estimation followed the basic proce-
standing needed for the problem physics as more dure developed by Curran et al. [128] which carried
elements are brought together in an integrated out a number of assessments on the variance
framework. This section documents one project between minimum weight optimisation and mini-
[142] which looked towards a tighter integration of mum cost in general. The completed system model
design, manufacturing and cost within the industry was also assessed using the systems analysis tools
standard SE process attempting to embody the developed by Early et al. [58] to obtain estimates for
principles more faithfully from the outset. which parameters have effects through the system
In that project the design process and systems and which are damped out, either due to low
measurement technologies were applied to a rela- sensitivity or due to low analysis fidelity.
tively simple configuration trade-off. The trade-off The results of the study confirmed that for the
was to assess the effect of the engine location on a upper wing panel chosen that the minimum weight
conventional aircraft, the choices being, under wing design costs 60% more than the minimum DOC
or on the rear fuselage (Fig. 40). To provide a focus design (Fig. 41). The minimum weight design has
for the study an upper wing panel near the wing many more stiffeners and a thinner skin. However,
root was chosen. The aim of this local study was to the minimum DOC design cost is only 41% heavier
consider the effect of manufacturing process on the than the minimum weight design. These differences
cost and performance of the chosen configurations. are likely to be exaggerated when durability and
Optimisations were performed to obtain the panel damage tolerance over the life is accounted for.
design for minimum weight and for minimum DOC With respect to the two panels in this study it is
based on the loads coming from the high level clear that the wing mounted engines had better cost
ARTICLE IN PRESS
Costs Table Configuration 1 Configuration 2
Wing Fuselage Wing Fuselage

Minimise for:
Weight $13,975 $11,521 $15,097 $11,846
DOC $8,665 $7,176 $9,803 $6,955
b
Weights Table Configuration 1 Configuration 2
Minimise for: Wing Fuselage Wing Fuselage
Weight 71.8 Kg 32.4 Kg 89.6 Kg 34.1 Kg
DOC 105.3 Kg 42.2 Kg 128.9 Kg 48.6 Kg
Genetic Weight
56.7 Kg 28.6 Kg 85.7Kg 32.1 Kg
Optimisation
Fig. 41. Cost and weight comparisons based on minimum weight and cost: (a) comparison table for cost of the panels in $; (b) comparison
table for weight of the panels in kg.
and weight results but at least the integrated The depth and complexity of many of the studies
approach also allows savings to be identified over illustrate the difficulty of providing integrated
the conventional approaches. The minimum DOC models in industrial practice. The geometry for
for configuration 2 cost 50% less than the minimum aerodynamic integration for a typical by pass
weight design for configuration 1. This approach is Engine turbo fan engine consists of Wing Body,
very positive in providing a rationale for design Nacelle, Pylon, Core Cowl and Core Cowl Bifurca-
decisions and a basis for more rigorous, quantifiable tion. In general, the installation and integration of
design trade-offs. The flow of manufacturing the nacelle could be with symmetrical plane of the
information back to design relaxes the constraints nacelle vertical to either the horizontal plane or
and allows other structural configurations to be wing plane.
considered. The weakness in that study is that the It is understood that the Installation of pylon and
large differences, particularly in the minimum nacelle on aircraft wing at transonic flows has
weight design, are likely to be reduced by simple significant effect on wing circulation, chord wise
practical manufacturing constraints. and span wise load distribution, shock position, the
surface boundary layer on and its interaction with
6.3. Aerodynamic integration of wing– pylon and shock wave and therefore lift and drag. An effective
nacelle integration requires an understanding of the links
and interfaces between several disciplines; Wing and
6.3.1. Introduction Nacelle Aerodynamics, Propulsion systems and its
The previous sections espoused the use of SE to location, Structure, and Noise and is a compromise
facilitate integrated models which include non- in aircraft design between requirements for take off,
technical disciplines. In practice of course this is cruise, decent and engine conditions. Further, one
often easier said than done. This section contains a has also to consider the effect of integration on
brief review of the aerodynamic research on noise, vibration and flutter, manufacturing, main-
wing–pylon–nacelle systems [143–158]. There are tenance and repairs. As highlighted by Berry [147],
many research issues even in this detailed area, and in relation to a typical transonic aircraft such as
complex physical phenomena which need to be Boeing 777, integration of power plant with the
understood before their use in integrated models. airframe is complex and has no unique solution.
ARTICLE IN PRESS
The current state of art on aerodynamic integra- 0.8

tion of wing–pylon–nacelle is described in this
section. 0.7
0.6
6.3.2. Lift, drag and flow field
Typical effect on span wise loading on a wing 0.5
body without (WB) and with a pylon–nacelle
CL
(WBPN) has been studied for a generic twin engine 0.4
wide body aircraft of Airbus type and derived DLR
F4 configuration. The experiments were carried out 0.3
in the ONERA S2MA wind tunnel calculations
0.2
based on the DLR Euler code CEVCATS [144].
Both the experiments and calculations show com- 0.1
pared to a wing body, a significant reduction in lift
with the installation of the pylon–nacelle. The 0
reduction in lift is most pronounced, as high as -4 -3 -2 -1 0 1 2 3
30%, at the location of the engine. It was also Alpha
noticed that the engine installation produces a WBPN Exp WBPN CFD WB CFD WB Exp
discontinuity in the slope of the wing lift (circula-

Fig. 42. Lift coefficient versus angle of attack [158].
tion), distribution due to flow separation at that
location, observed in experiments. Lift and drag
results at M ¼ 0.75 for DLR F6 Wing body
0.8
configuration with one of the nacelle installation
were presented in Broderson [143]. The results 0.7
shown there are from calculations with DLR TAU 0.6
software based on RANS and experimental data
from tests conducted in the ONERA S2MA wind 0.5
tunnel. 0.4
The installation drag was calculated from
CL
0.3
C DðInstÞ ¼ C Dðwith engineÞ C DðCleanÞ C DðinternalÞ . 0.2
Accepting the difference between CFD analysis 0.1

and experiments of a constant shift of about 16 drag 0
counts, both the experiments and CFD show the 0 0.01 0.02 0.03 0.04 0.05
-0.1
pylon–nacelle installation drag arising due to shift
in shock position and boundary layer separation at -0.2
Cd
the engine location can be about 10–11 drag counts. WBPN Exp WBPN CFD WB CFD WB Exp
Results of computation compared with the
experimental results for the baseline configuration Fig. 43. Lift coefficient versus drag coefficient [158].
with the wing body nacelle and pylon (WBNP) and
wing body alone (WB) from the work of Devine
et al. [158] are shown in Figs. 42 and 43,
respectively. tion in mesh generation. To meet the mesh
These investigations were based on an adopted requirements for an accurate simulation using
commercial computational code, FLUENT 6TM, unstructured meshes is difficult, especially in the
with an unstructured mesh generated using near wall regions. A solution to this problem was
ICEMCFD can be used to predict the drag to create a hybrid mesh, which allows for suffi-
associated with changes in geometry can be cient grid resolution in the near wall region. To
predicted with accuracy comparable to those of create a hybrid mesh, using ICEMCFD pris-
other predictions. The unstructured meshing tech- matic layers were grown from a surface mesh. The
niques have the advantage of the relative automa- solver was run in coupled implicit mode using a
ARTICLE IN PRESS
Spalart-Allmaras [150] turbulence model due to it

performance in the 2nd AIAA drag prediction
workshop.
Results presented here are for M ¼ 0.75 and
R ¼ 3 106. It can observed from these figures that
when compared with the experimental data, the
computations over predict the lift (Fig. 42), at every
design point by 12% and drag by 8% (Fig. 43). The
results for lift and drag from a commercial code
compare favourably with corresponding results
from DLR obtained with the TAU code [143].
In comparison with experiments, while the total
lift and drag values are not predicted accurately (i.e.
within 1–2 drag counts from experimental), the
change in lift and drag due to the nacelle and pylon
interactions is predicted accurately enough. For
Fig. 45. Computed streamlines inboard side of pylon [158].
example at CL ¼ 0.5 the experimental installation
drag was 34 drag counts while the predicted was 36
drag counts. Within a range of lift coefficients
between 0.45 and 0.65 the predicted installation aerodynamic factors. The Aerodynamic factors
drag was within 2 drag counts of the experimental include the overall nacelle shape and its orientation,
result. This means validated CFD tools such as nacelle internal and external contours, wing geo-
Fluent6 can be used for predicting changes in lift metry, the type of boundary layer on the wing,
and drag for different design cases. Figs. 44 and 45 nacelle and pylon; laminar or turbulent, the nacelle
are the computed streamlines at the wing root position relative to the wing; vertically and hor-
(above wing) and the inboard side of the pylon izontally, pylon geometry, and exhaust system, and
(below the wing). the non-aerodynamic factors include the engine
bypass ratio, ground clearances during take off and
landing, flutter boundaries, noise, weight and cost.
6.3.3. Factors influencing aerodynamic integration However, these effects are not fully understood. The
There are several factors influencing an integrated major source of increase in drag can be either due to
design of wing–pylon–nacelle as noted by Berry the pylon geometry [144] or due to channel effect
[147]. Some are aerodynamic and others non- arising out of the nacelle, pylon and wing [145] or
most likely to a combination of both.
Laminar flow technology (nacelle/wing) offers
potential benefits in meeting ACARE targets in
reducing fuel consumption and noise. There are
validated design tools [156] available for predicting
the aerodynamic features of laminar flow nacelle. The
positioning of transition does have an influence on the
wing–pylon–nacelle aerodynamic interactions [157].
Rumsey et al. [157] showed that the differences in
pressure distributions on a wing with pylon and
nacelle between transition, specified at 5.7% chord,
and fully turbulent transition near the leading edge.
It appears that the transition does not have much
effect on the boundary layer separation inboard.
There is only small shift in the shock position.
Transition specified gave higher or lower drag by 10
drag counts depending on the type of solver used.
But the effect of having an extensive laminar flow
Fig. 44. Computed streamlines at wing root [158]. over a nacelle or a wing on the aerodynamics of
ARTICLE IN PRESS
nacelle–wing–pylon interaction is not fully under- 0.6

stood. 0.55
Results based on a study of a 150 seat advanced 0.5
turbo prop aircraft [146], without the presence of 0.45
Cl
0.4
the propellers illustrated the significant effect of
0.35 Compression Plyon
nacelle geometry on pressure distribution on the DLRF6
0.3
wing. The results show that a suitably contoured
0.25
nacelle can reduce the shock strength leading 0.2
to a reduction in both the wave drag and viscous 0.025 0.027 0.029 0.031 0.033 0.035 0.037 0.039 0.041
drag. This could be attributed to reduced ac- Cd
celerations and velocities on wing with gradual
Fig. 46. Lift versus drag of baseline case to compression pylon
changes in the combined cross sectional area wing
case [158].
and nacelle.
Investigations by Wilhelm [153] to understand the
complexity of interactions between wing geometry Results for lift and drag of compression pylon
and pylon nacelle are based on an inverse interactive with conventional one for wing pylon with the
design method with an Euler solver. The wing presence of nacelle 9170 are compared in Fig. 46.
configuration was DLR F6, and the nacelle config- The results shown here are for a fixed Mach
uration was CFM56. The nacelle was closely coupled number of 0.76 and three angles of incidence, 11,
to the wing with a narrow channel flow inboard and 01 and 11. For every design point the compression
strong shock waves on wing, pylon and nacelle. pylon produces small increases in lift and drag. At
Wilhelm showed that a small variation in nacelle zero incidence, the lift and drag coefficients for the
geometry inboard can produce a significant reduction compression pylon were 0.432 and 0.03456, respec-
in shock strength and therefore a reduction in nacelle tively, compared to the corresponding values for the
drag. An alternative way of reducing interference conventional pylon of 0.426 and 0.03498, respec-
drag is by the changes in the wing geometry. A tively. For a same lift coefficient, the compression
significant reduction in shock strength and therefore pylon produced a reduction of, approximately 5
interference drag is achieved by a small in geometry drags. This reduction was mainly from a reduction
of the lower surface of the wing inboard. in the pressure drag. The breakdown was a 0.5%
The influence of pylon geometry can be gleaned reduction in the skin friction drag and a 1.8%
from the experiments performed at NASA Langley reduction in the pressure drag.
Research Centre [145,149,154] on propulsion inte- Systematic investigations performed by Godard
gration. The configuration was 1/24 scale model a et al. [146], Rudnik and Rossow [152] looked at the
high wing transport aircraft and experiments were effect of nacelle–pylon location in relation to the
performed several Pylon geometry and in Mach wing. The configuration was a generic transport
number range 0.7–0.8. Typical results for pressure type DLR F6. Analyses were performed by using
distributions on the wing, inboard and outboard at both ONERA-CANARI and DLR-CERVATS
a Mach number of 0.8 showed the sensitivity of the codes and results compared with the tests at
pressure distribution and the shock position on the ONERA S2MA wind tunnel. Five positions of
wing to the pylon geometry, and demonstrated the pylon location were investigated; three horizontal
need for a careful contouring of pylon/wing for locations and two sets of vertical positions were
reducing the adverse effects of loss lift due to investigated. The results showed that the flow field
acceleration of air flow with the presence of pylon. and therefore installation drag, was sensitive to the
Variations in pylon designs [149] rather than, say horizontal position but not that sensitive to the
nacelle or wing designs, offer fewer design con- vertical position of the nacelle.
straints. In these investigations a compression pylon The results show that a considerable reduction in
design produced the lowest installation drag. This drag can be achieved by suitably locating the pylon
design was subjected to several wind tunnel tests at in relation to the wing. The best result for minimum
NASA Langley in the late 1980s and compared to drag was found to be furthest upstream from the
several different pylon designs. However, these wing. The reason for this was when the nacelle
experiments were conducted without the presence closest to the wing in the horizontal position had a
of a nacelle. large effect on the effect area change, between the
ARTICLE IN PRESS
wing, pylon and nacelle in the direction of flow. This step was the use of Euler codes with viscous
led to an acceleration of the local flow and a corrections [151]. In recent years Reynolds Aver-
stronger shock wave resulting in increase in drag aged Navier–Stokes (RANS) calculations were
and reduction in lift. However, mounting a nacelle performed [152]. The ability and computational
the further the nacelle further away from the wing resource required of each in predicting the inter-
has the greater the structural weight. It should be ference effect is different. Euler calculations can
emphasised that in a practical design, there are predict the change in lift went the nacelle and pylon
constraints in engine position and these include is installed and is computationally inexpensive.
structure, flutter, ground clearance, nacelle internal However, the accuracy of the lift values is inade-
aerodynamics, and noise. Therefore, further inte- quate and drag implications cannot be predicted.
gration with other disciplines is necessary for full Euler calculations with viscous effects can predict
understanding of the implications of these results. the lift more accurately and also the changes in lift.
Integration of large BPR nacelles has received However, the drag implications cannot be predicted
attention by investigators in recent years [148,149]. and the computational cost is three times the basic
The largest BPR engine today is on the Boeing 777, Euler calculations [155]. RANS calculations can not
which has an engine BPR of 9. It is well understood only predict the lift accurately but also drag.
that high BPR engines have higher aerodynamic However, the computational cost can be as much
efficiency, lower noise but greater or same level of as hundred times greater than the basic Euler
interference effects and reduced ground clearance calculations [155].
for take off and landing [147]. Currently very high To acquire accurate solutions, CFD analyses
bypass ratio (BPR 10) and ultra high by pass ratio must be conducted by using a proper computational
(BPR 15) engines are being considered for aircraft domain and grid quality in consideration of the flow
propulsion systems. Rudnik and Rossow [152] features to be expected. The drag prediction work-
showed the dependency of fuel consumption and shop [157] emphasises the importance of the size
installation drag on by pass ratio. The installation and the quality of the mesh and turbulence
drag increase due to increase in by pass ratio is modelling, transition prediction and viscous model
primarily due to increase in wetted area of the in prediction of aerodynamic lift and drag of wings
nacelle. body and wing body and nacelle–pylon interactions.
The configuration was ALVAST twin jet narrow With a medium size grid the accuracies in predicting
body wing fuselage model. Three types of engines drag with wing body interactions due to grid, code
were investigated: a conventional turbofan simula- used, turbulence model used, transition fix and the
tor (TF), BPR ¼ 5, a very high Bypass ratio viscous model are 5, 5, 7, 10, 4 drag counts,
simulator (VHBR), BPR ¼ 10, and a Ultra high respectively. The corresponding figures for wing–
bypass ratio (UHBR), BPR ¼ 15. The effect of body pylon and nacelle are 11, 10, 15, 13 and 5 drag
increasing in the by pass ratio (engine diameter), on counts, respectively. Although, in comparison with
the upper surface of the wing is to reduce the shock experiments, while the total lift and drag values are
strength and move the shock forward, which is not predicted accurately, the change in lift and drag
equivalent to reducing the wing incidence and due to the nacelle and pylon interactions is
should produce reduced lift. This should also predicted within 2 drag counts of the experimental
produce a decrease in drag. On the lower surface results. Therefore, validated CFD tools such as
of the wing increase in the by pass ratio increases Fluent6 can be used for predicting changes in lift
the negative pressures, due to increase in airflow and drag for different pylon nacelle configurations.
velocities on that surface. This is more so pro-
nounced inboard rather than the outboard wing 6.3.5. Concluding remarks
section with UHBR. This should increase drag on The aerodynamic integration for drag reduction
the wing lower surface and the nacelle. for a transonic aircraft is sensitive to the detailed
geometry of the wing, pylon–nacelle, the location of
6.3.4. Numerical simulations pylon–nacelle in relation to the wing, the type of
The CFD methods used to simulate the nacelle, boundary layer on the surfaces, and the engine
pylon and wing interference effects have evolved bypass ratio. Although the prediction methods
over the past 15 years. The earlier calculations were available for drag are not accurate enough when
performed using inviscid Euler codes [150]. The next compared with experiments, the change associated
ARTICLE IN PRESS
with different wing–pylon–nacelle design configura- interpretations in industry and academia relating to
tions can be predicted with reasonable accuracy both breadth and depth, for example combining
of 2 drag counts. A more effective approach in models even with a single discipline, such as
aerodynamic design of aircraft for drag reduction, aerodynamics is a form of integration for several
apart from novel configurations that may bring a systems or sub-systems. These studies raised many
step change in drag, may require area ruling with questions on detailed understanding and it is clear
the presence of nacelle and pylon and the tail plane that introduction of other disciplines will further
for both pressure and induced drag, and detail complicate the picture. The potential of a fully
considerations to both the wetted area and bound- integrated system from design through to manufac-
ary layer development for skin friction drag. This turing and life-cycle costs is of major benefit to
focused review of an integrated approach to design industry and widely applicable beyond the aero-
of wing body–pylon–nacelle shows that there is space sector. There is no question that in design
much detail embedded under the SE design philo- synthesis and analysis within SE the integration of
sophy. These details provide both an understanding aerodynamic-structures-manufacturing and CAD
of the physics of problems as well as new design for reduced cost/reduced lead-time can be per-
rules or guidelines. These integrated models form formed at least to first-order accuracy. However, the
one of the key technologies required to meet the whole life cycle modelling of an aircraft within the
ACARE targets of significant reductions in aero- context of SE, for performance/cost/environment/
dynamic drag and therefore fuel consumption. safety, is at an infant stage as the integration is
highly complex involving additionally people, pro-
7. Concluding remarks duct and processes which do not fit easily with
engineering analyses. The major challenge is to
The many new challenges emerging in the aero- provide this environment such that technical details
space industry today are resulting in new initiatives can be accounted for appropriately, without over-
and insights to the role of engineering design and whelming the high level design challenges and where
analysis and in particular the need for collaborative the influence of data and parameters throughout
multi-disciplinary approaches to solving the pro- and across systems is clearly understood.
blems ahead. It is evident that major gains will be
difficult to achieve even with step changes in a single
technology. It is the integrated approach that will References
permit the exploration of new design spaces. Within
this paper the state of the art in SE and the [1] Aerospace & Defense Industries Association of Europe.
integration of traditional analysis disciplines in this Facts and Figs. 2004, 2004.
[2] International Civil Aviation Organization (ICAO). Annual
framework were reviewed and some of many
Report of the Council, 2004.
exciting projects were summarised. Some key [3] Advisory Council for Aeronautics Research in Europe
lessons are emerging from this varied and volumi- (ACARE). Strategic Research Agenda 2 Executive Sum-
nous work. Firstly there are technical challenges in mary, October 2004.
modelling and analysis capability and major work is [4] Paul D, Pratt D. History of flight vehicle structures
required in refining the understanding of the 1903–1990. AIAA J Aircraft 2004;41(5).
[5] Mohagheh M. Evolution of structures design philosophy
relationships between system entities and the and criteria. AIAA J Aircraft 2005;42(4).
analysis models that are used to represent them. [6] AEIGT Technology Subgroup. Greener by Design, 2004.
Several key issues in this area have been reviewed. [7] Nietz TC, Baber S. An innovative UAV design, In: AIAA
In order to help bridge the gap between traditional 3rd ‘‘Unmanned Unlimited’’ technical conference, work-
engineering analysis and the analysis of more shop and exhibit, Paper no. AIAA 2004-6380, Chicago, IL,
20–23 September 2004.
complex systems an understanding of SE and how [8] Crouse G, Sash T. Conceptual design of a long-endurance,
physical models can be fully embedded in this tilt-boom unmanned aerial vehicle. In: AIAA 3rd ‘‘Un-
environment is still needed. There has been progress manned Unlimited’’ technical conference, workshop and
in generic behavioural models, and in characterising exhibit, Chicago, IL, 20–23 September 2004.
[9] Velev O. Regenerative fuel cell system for an unmanned
systems but there is not yet strong evidence of how
solar powered aircraft. In: AIAA-2000-2873 ASME, IEEE,
this can be capitalised on to shift the design process et al., intersociety energy conversion engineering conference
forward. The latter section illustrates some of the and exhibit (IECEC), 35th, Las Vegas, NV, 24–28 July
complexities faced in practice. Integration has many 2000.
ARTICLE IN PRESS
[10] McKee M. Pioneering private space flight. New Scientist [33] Stark J. Engineering information management systems:
2004;2479. beyond Cad/Cam to concurrent engineering support
[11] Raymer D. Aircraft design: a conceptual approach. AIAA (Automation in Manufacturing). Van Nostrand Reinhold,
Education Series, 1999. 1992.
[12] Roskam J. Airplane design part I: preliminary sizing of [34] Crnkovic I, Asklund U, Dahlqvist A. Implementing and
airplanes. Design Analysis & Research Corporation, 1989 integrating product data management and software config-
[13] Jenkinson LR, Simpkin P, Rhodes D. Civil jet aircraft uration management. Boston, MA: Artech House Inc.;
design. London: Butterworth-Heinemann; 1999. 2003.
[14] Fielding JP. Introduction to aircraft DESIGN. Cambridge: [35] Burden R. PDM: product data management. Resource
Cambridge University Press; 1999. Publishing, 2003.
[15] Armstrong CG, Monaghan DJ, Price MA, Ou H, Lamont [36] Morris AJ, Vignjevic R. Consistent finite element structural
J. Integrating CAE concepts with CAD geometry. In: analysis and error control. Comput Methods Appl Mech
Topping BHV, Bittnar Z, editors. Engineering computa- Eng 1997;140:87–108.
tional technology. Stirling: Saxe-Coburg Publications; [37] Frey EK. Parametric modelling techniques for aircraft
2002. p. 75–104. structures, AIAA Paper 99-1363, 1999. p. 1411–5.
[16] Blair JC, Ryan RS, Schutzenhofer LA, Humphries WR. [38] Vaughan C. Product data management and the engineering
Launch vehicle design process: characterization, technical analysis environment. NAFEMS, 2000
integration, and lessons learned. NASA, TP-2001-210992, [39] Peak RS, Fulton RE, Nishigaki I, Okamoto N. Integrating
2001. engineering design and analysis using a multi-representa-
[17] Knight NF, Stone TJ. Rapid modeling and analysis tools: tion approach. Eng Comput 1998;14(2):93–114.
evolution, status, needs and directions. NASA, CR-2002- [40] La Rocca G, Krakers L, van Tooren M. Development of
211751, NASA Langley Research Center, 2002. an ICAD generative model for blended wing-body aircraft
[18] Nickol C. Conceptual design shop. In: Presentation to design, In: AIAA/ISSMO symposium on multidisciplinary
conceptual aircraft design working group (CADWG21), analysis and optimization, 9th, AIAA, AIAA-2002-5447,
September 2004. 2002. p. 453–62.
[19] INCOSE-TP-2003-016-02. Systems engineering handbook. [41] Wallace DR, Abrahamson S, Senin N, Sferro P. Integrated
INCOSE Technical Product, 2004. design in a service marketplace. Comput-Aided Des
[20] Blair JC, Ryan RS, Schutzenhofer LA, Humphries WR. 2000;32(2):97–107.
NASA systems engineering handbook. NASA Report SP- [42] Senin N, Wallace DR, Borland N. Distributed object-based
6105, 1995. modelling in design simulation marketplace. ASME J Mech
[21] Camarinha-Matos LM, Afsarmanesh H. The virtual Des 2003;125:2–13.
enterprise concept. In: IFIP conference proceedings, vol. [43] ISight, Engineous Software Inc. Cary, NC, Version 9, 2005.
153 archive, proceedings of the IFIP TC5 WG5.3/ [44] Fiper, Engineous Software Inc. Cary, NC, Version 1.6,
PRODNET working conference on infrastructures for 2005.
virtual enterprises: networking industrial enterprises, 1999. [45] Model Center, Phoenix Integration Inc. Blacksburg, VA,
[22] Jameson A. Re-engineering the design process through Version 6.1, 2005.
computation. AIAA J Aircraft 1999;36(1):36–50 [0021- [46] Armstrong CG, McCune RW, Ou H, Price MA, McMillan
8669]. AJ. Modelling requirements for dimensional addition. In:
[23] Price M, Early JM, Curran R, Benard E, Raghunathan S. 5th world congress on computational mechanics, vol. 1,
Identifying interfaces in engineering systems. AIAA J 2002. p. 442 ISBN:3950155414.
2006;44(3):529–40. [47] Mawhinney P, Price M, Armstrong CG, Ou H, Murphy A,
[24] Frezza ST, Levitan S, Chrysanthis P. Requirements-based Gibson A, et al. Using idealised models to enable
design evaluation. In: 32nd ACM/IEEE design automation integration and analysis driven design. In: AIAA’s 3rd
conference, 1995, p. 76–81. annual aviation technology, integration, & operations
[25] Ranky PG. Concurrent/simultaneous engineering: meth- (ATIO) Forum, AIAA Paper 2003-6747, Denver, Novem-
ods, tools & case studies, Series World class CIM series. ber 2003.
Guildford: CIMware; 1994. [48] Mawhinney P, Price M, Armstrong C, Curran R, Murphy
[26] Peters JF. The transition of functional organizations to A, Early J, et al. Design and analysis integration using
integrated product teams on the space station program. systems engineering for aircraft structural design. In: 4th
NCOSE 1995. p. 767–73. AIAA aviation technology and operations forum, Chicago,
[27] Ullman DG. The mechanical design process, 3rd ed. New 20–22 September 2004.
York (Maidenhead, UK): McGraw-Hill; 2003. [49] Mawhinney P, Price M, Curran R, Benard E, Murphy A,
[28] Guindon R. Designing the design process: exploiting Raghunathan S. Geometry-based approach to analysis
opportunistic thoughts. Hum–Comput Interact integration for aircraft conceptual design. In: 5th annual
1990;5(2&3):305–44. aviation technology, integration, & operations (ATIO)
[29] Whitney DE. Designing the design process. Res Eng Des forum, AIAA-2005-7481, Washington, DC, September
1990;2(1):3. 2005.
[30] CATIA. Dassault Systemes, Paris, France, Version 5 [50] Price M, Mawhinney P, Curran R, Armstrong C, Murphy
Release 15, 2005. A. Geometry centred airframe design, In: AIAA’s 5th
[31] StarCD, CD-adapco /www.cd-adapco.comS. annual aviation technology, integration, & operations
[32] Patran. MSC Software Corp, Santa Ana, CA, Version 2004 (ATIO) forum, Paper No. AIAA-2005-7479, Washington,
Release 2a, 2004. September 2005.
ARTICLE IN PRESS
[51] Mawhinney P, Price M, Armstrong CG, Ou H, Murphy A, symposium of the international council on systems
Gibson A, et al. Using idealised models to enable engineering, San Jose, CA, 10–12 August 1994.
integration and analysis driven design. In: AIAA’s 3rd [68] Curran R, Raghunathan S, Price M. A review of aircraft
annual aviation technology, integration, & operations cost modelling and genetic causal approach. Prog Aero-
(ATIO) forum, AIAA Paper 2003-6747, Denver, November space Sci J 2004;40(8):487–534.
2003. [69] Curran R, Rothwell A, Castagne S. A numerical method
[52] Mawhinney P, Price M, Armstrong C, Curran R, Murphy for cost-weight optimisation of stringer-skin panels. J
A, Early J, et al. Design and analysis integration using Aircraft 2006;43(1):264–74.
systems engineering for aircraft structural design. In: 4th [70] Krus P. Systems engineering in aircraft system design. In:
AIAA aviation technology and operations forum, Chicago, 11th annual international council on systems engineering,
20–22 September 2004. Melbourne, Australia, 2001.
[53] Mawhinney PN. CAE integration for aircraft structural [71] Kroo I, Altus S, Braun R, Gage P, Sobieski I. Multi-
design. PhD thesis, Queen’s University Belfast, 2005. disciplinary optimisation methods for aircraft preliminary
[54] Monaghan DJ, Lee KY, Armstrong CG, Ou H. Mixed design. In: Fifth AIAA/USAF/NASA/ISSMO symposium
dimensional finite element analysis of frame models. In: on multidisciplinary analysis and optimization, Panama
Proceedings of the 10th ISOPE conference, Seattle, USA, City, FL, AIAA 94-4325, 7–9 September 1994.
2000. [72] Clarkson PJ, Simons C, Eckert C. Predicting change
[55] Shim KW, Monaghan DJ, Armstrong CG. Mixed dimen- propagation in complex design. In: ASME design
sional coupling in finite element stress analysis. Eng engineering technical conference, DETC1002-DTM-
Comput 2002;18(Part 3):241–52. 21698, 2001.
[56] Castagne S, Curran R, Rothwell A, Price M, Murphy A. [73] Flanagan TL, Eckert CM, Simth J, Eger T, Clarkson PJ. A
Development of a precise manufacturing cost model for the functional analysis of change propagation. In: Interna-
optimization of aircraft structures. In: AIAA’s 5th annual tional conference on engineering design, Stockholm, 2003.
aviation technology, integration, & operations (ATIO) [74] Sobek DK, Ward AC. Principles From Toyota’s set-based
forum, Paper No. AIAA-2005-7438, 2005. concurrent engineering process. 1996 ASME design en-
[57] Curran R, Price M, Raghunathan S, Benard E, Crosby S, gineering technical conferences and computers in engineer-
Castagne S, et al. Integrating aircraft cost modelling into ing conference, 96-DETC/DTM-1510, Irvine, CA, 18–22
conceptual design. Concurrent Eng: Res Appl (CERA) August 1996.
2005;13(4):321–30. [75] Ostwald PF. Engineering cost estimating. Englewood
[58] Early J, Price M, Mawhinney P, Curran R, Benard E, Cliffs, NJ: Prentice-Hall; 1992. p. 576 ISBN:0-13-276627-2.
Raghunathan S. Framework for modelling and simulation [76] Department of Defense. Systems engineering fundamentals.
within a flexible integration environment. AIAA’s 4th Defense Acquisition University Press.
annual aviation technology, integration, & operations [77] Wierda L. Design-oriented cost information: the need and
(ATIO) forum, Paper No. AIAA-2004-6457, Chicago, IL, the possibilities. J Eng Des 1990;1(2):147–67.
September 2004. [78] Hoult DP, Meador CL, Deyst J, Dennis M. Cost awareness
[59] Early J, Price M, Curran R, Raghunathan S, Benard E, in design: the role of data commonality. SAE technical
Castagne S. Constraint management for engineering paper, Number 960008, 1996.
systems, In: 5th annual aviation technology, integration, [79] Sheldon DF, Huang GQ, Perks K. Design for cost: past
& operations (ATIO) forum, AIAA-2005-7374, Washing- experience and recent development. J Eng Des
ton DC, September 2005. 1991;2(2):127–39.
[60] Poon J, Maher ML. Emergent behaviour in co-evolution- [80] Dean EB, Unal R. Elements of designing for cost. In:
ary design. Artificial Intelligence in Design, 1996. Proceedings of AIAA 1992 Aerospace Design Conference,
[61] Jankovich L, Jankovich S, Chan AHC, Little GH. Irvine, CA, February 1992.
Emergent modelling for structural design. In: Conference [81] Wood MJ. Design to cost. New York: Wiley Interscience;
on complexity and complex systems in industry, 20–22 1989.
September 2000. [82] Curran R, Rush C, Roy R, Raghunathan S. Current cost
[62] Crick F. The astonishing hypothesis. London: Simon & estimating practice in aerospace. In: Proceedings of
Schuster; 1994. concurrent engineering: research and applications
[63] Damper RI. Emergence and levels of abstraction. Int J Syst (CE2002), 2002. p. 894–902.
Sci 2000;31(7). [83] Curran R, Kundu A, Ray A, Woods K, Crosby S,
[64] Ottino JM. Engineering complex systems. Nature Raghunathan S, et al. Aerospace cost estimating for
2004;427(399). competitive design for manufacture. In: Proceedings of
[65] Jackson S. Systems engineering for commercial aircraft. concurrent engineering, 2002. p. 885–93.
Aldershot: Ashgate Publishing Limited; 2000. [84] Roy R, Jones P. Developing an integrated approach to
[66] Isaacs A, Sudhakark PM, Mujundar A. Design and design and manufacturing cost modelling. In: Proceedings
development of MDO framework. In: International of CE2000 conference, Lyon (France), 17–21 July 2000.
conference on modeling, simulation, optimization for p. 31–9.
design of multi-disciplinary engineering system, Goa, India, [85] Heinmuller B, Dilts DM. Automated design-to-cost:
2003. application in the aerospace industry. In: Annual meeting
[67] Clymer JR. The role of reductionism and expansionism in of the Decision-Science-Institute, vols. 1–3, San Diego, CA,
system design and evaluation. In: 4th international 22–25 November 1997. p. 1227–9 [chapter 569].
ARTICLE IN PRESS
[86] Gieger TS, Dilts DM. Automated design-to-cost: integrat- [106] Roskam J. Airplane design, vol. 8. Roskam Aviation/
ing costing into the design decision. Computer Aided Des Engineering Company, 1985.
1996;28(6/7):423–38. [107] Westphal R, Scholz D. A method for predicting direct
[87] Brimson JA, Downey PJ. Feature technology: a key to operating costs during aircraft system design. Cost Eng
manufacturing integration. CIM Rev 1986;2(3):21–7. 1997;39(6):35–9.
[88] Bashir HA, Thompson V. Estimating design complexity. [108] Boothroyd G, Dewhurst P. Product design for assembly.
J Eng Des 1999;10(3):248–57. Boothroyd Dewhurst. Wakefield, RI, USA, 1990 [first
[89] Cooper R, Kaplan RS. Measure cost right: make the right edition published in 1983].
decisions. Harvard Bus Rev 1988;66(5):96–103. [109] Stoll HW. Design for manufacture: an overview. Appl
[90] Cokins G. ABC can spell a simpler, coherent view of costs. Mech Rev 1998;39(9):1356–64.
Computing Canada, 1 September 1998. [110] Bloom HM. Design for manufacturing and the life cycle.
[91] Taylor IM. Cost engineering-a feature based approach. In: In: Proceeding of the NSF design theory 88 conference,
85th meeting of the AGARD structures and material panel, 1998. p. 302–12.
Aalborg, Denmark, October 13–14, 14: 1–9, 1997. [111] Murman EM, Walton M, Rebentisch E. Challenges in the
[92] Curran R, Kundu A, Raghunathan S, Eakin D. Costing better, faster, cheaper, era of aeronautical design, engineer-
tools for decision making within integrated aerospace ing and manufacturing. Aeronautical J 2000:481–8.
design. J Concurrent Eng Res 2002;9(4):327–38. [112] Rais-Rohani M. A framework for preliminary design of
[93] Smith AE, Mason AK. Cost estimation predictive model- aircraft structures based on process information. NASA
ling: regression versus neural network. Eng Economist Grant NAG-1-1716, 1998.
1997;42(2):137–62. [113] Rais-Rohani M, Greenwood AG. Product and process
[94] Villarreal JA, Lea RN, Savely RT. Fuzzy logic and neural coupling in multidisciplinary design of flight vehicle
network technologies. In: 30th aerospace sciences meeting structures. In: 7th AIAA/NASA/USAF/ISSMO sympo-
and exhibit, Houston, TX, 6–9 January 1992. sium on multidisciplinary analysis and optimisation, Paper
[95] Bode J. Neural networks for cost estimation. Am Assoc No AIAA 98-4820, St. Louis, MO, USA, 2–4 September
Cost Eng 1998;40(1):25–30. 1998.
[96] Roy R, Bendall D, Taylor JP, Jones P, Madariaga AP, [114] Sandoz PL. Structural design of future commercial trans-
Crossland J, et al. Development of airframe engineering ports, Paper No AIAA 73-20, 1973.
CERs for military aerostructures. Second world manufac- [115] Noton BR. Cost drivers in design and manufacture of
turing congress (WMC’99), Durham, UK, 27–30 Septem- composite structures. Compos Struct Anal Des
ber 1999. 1987:419–28.
[97] Department Of Defence (DoD). Parametric estimating [116] Noton BR. Cost drivers and design methodology for
handbook, 2nd ed. DoD; 1999. automated airframe assembly. In: Proceedings of 31st
[98] Pugh P. Working top-down: cost estimating before devel- international SAMPE symposium, 7–10 April 1986
opment begins. J Aerospace Eng Part G 1992;206:143–51. p. 1441–55.
[99] Thurston DL, Essington SK. A tool for optimal manu- [117] Ermanni P, Ziegmann G. Cost-efficiency of highly inte-
facturing design decisions. Manuf Rev 1993;6(1):48–59. grated fuselage structures-comparison between metals and
[100] Keeney RL, Raiffa H. Decisions with multiple objectives: composites. In: Proceedings of SAMPE advanced materi-
preferences and value trade-offs. New York: Wiley; 1976. als: cost effectiveness, quality control, health and environ-
[101] Vanderplaats GN. Numerical optimization techniques for ment, 1991. p. 347–59.
engineering design: with applications. New York (Maiden- [118] Curran R, Kundu A, Raghunathan S, Eakin D. Costing
head, UK): McGraw-Hill; 1984. tools for decision making within integrated aerospace
[102] Collopy PD, Eames DJH. Aerospace manufacturing cost design. J Concurrent Eng Res 2002;9(4):327–38.
prediction from a measure of part definition information. [119] Mileham AR, Currie GC, Miles AW, Bradford DT. A
Warrendale, PA: SAE Publications; 2001. parametric approach to cost estimating at the conceptual
[103] Kundu AK, Raghunathan S, Curran R, Cather G. Cost stage of design. J Eng Des 1993;4(2):117.
modelling as a holistic tool in the multidisciplinary systems [120] Pugh P. Working top-down: cost estimating before devel-
architecture of aircraft design—the next step design for opment begins. J Aerospace Eng 1992;206:143–51.
customer. In: 41st AIAA aerospace sciences meeting and [121] Herrera A. Design for manufacture and assembly, a cost
exhibit, USA, 6–9 January 2003, Reno, NV. effective tool supporting lean design initiatives. In: Pro-
[104] Butterfield J, Yao H, Curran R, Price M, Armstrong CG, ceedings of international forum on DFMA; 1998.
Raghunathan S. Integration of aerodynamic, structural, [122] Rush C, Roy R. Analysis of cost estimating processes used
cost and manufacturing considerations during the con- within a concurrent engineering environment throughout a
ceptual design of a thrust reverser cascade. In: 42nd AIAA product life cycle. In: Proceedings of the 7th international
aerospace sciences meeting and exhibit, AIAA Paper 2003, conference on concurrent engineering, Lyon, France,
5–8 January 2004. Technomic Publishing Co. Inc.; 2000. p. 58–67.
[105] Marx WJ, Mavris DN, Schrage DP. Effects of alternative [123] Arrow KJ, Arrow SS. Methodology problems in airframe
wing structural concepts on high speed civil transport life cost-performance studies. Rand Corporation document
cycle costs. In: Conference proceedings of 37th AIAA/ no.: RM-456-PR, 33; 1950.
ASME/AHS/ASC structures, structural dynamics, and [124] Levenson GS, Barro SM. Cost-estimating relationships for
materials conference, Salt Lake City, Paper No. AIAA- aircraft airframes. Rand Corporation document no.: RM-
96-1381, UT, 15–17 April 1996. 4845-PR, 99; 1966.
ARTICLE IN PRESS
[125] Roy R, Bendall D, Taylor JP, Jones P, Madariaga PA, cost for trade-offs on aircraft configurations. In: AIAA’s
Crossland J, et al. Identifying and capturing the qualitative 6th annual aviation technology, integration, & operations
cost drivers within a concurrent engineering environment. (ATIO) forum, Paper No. AIAA-2006-7739, Witchita, KS,
Advances in concurrent engineering. Pennsylvania, USA: September 2006.
Technomic Publishing Co. Inc.; 1999. p. 39–50. [143] Broderson O. Drag prediction of engine-airframe inter-
[126] Curran R, Kundu A, Raghunathan S, McFadden R. ference effects using unstructured Navier–Stokes calcula-
Influence of manufacturing tolerance on aircraft direct tions. AIAA J Aircraft 2002;39(6):927–35.
operating cost (DOC). J Mater Process Technol 2003; [144] Rossow CC, Godard J, Hoheisel H, Schmitt V. Investiga-
239–47, ISSN:0924-0136. tion of propulsion integration effects on a transport aircraft
[127] Curran R, Raghunathan S, Price M. A review of aircraft configuration. AIAA J Aircraft 1994;31(5):1022–30.
cost modelling and genetic causal approach. Prog Aero- [145] Henderson WP. Airframe/propulsion integration at transo-
space Scie J 2004;40(8):487–534. nic speeds. J Eng Gas Turbines Power Trans ASME
[128] Curran R, Rothwell A, Castagne S. A numerical method 1991;113(1):51–9.
for cost-weight optimisation of stringer-skin panels. [146] Godard JL, Hoheisel H, Rosscow CC, Sscmitt V.
J Aircraft 2005;43(1):264–74 [January–February 2006]. Investigation of interference effects for different engine
[129] Curran R, Watson P, Cowan S, Mahwinney J, Raghu- positions on a transport aircraft configuration. In:
nathan S. Development of an aircraft cost estimating model Hoheisel H, editor, Proceedings of DLR workshop on
for program cost rationalization. In: Proceedings of the aspects of airframe integration for transport aircraft,
Canadian Aeronautics and Space institute (CASI), April, German Research Centre, DLR, TR96-01, Brunswick,
Montreal, 2003. Germany, 1996.
[130] Curran R, Castagne S, Rothwell A, Price M, Murphy A. [147] Berry D. The Boeing 777 engine/airframe integration
Integrating manufacturing cost and structural requirements aerodynamics design process. International Council of the
in a systems engineering approach to aircraft design. In: Aeronautical Sciences (ICAS) 94-6.4.4, 1994.
Proceedings of the 46th AIAA/ASME/ASCE/AHS/ASC [148] Rudnik R, Rossow CC. Numerical simulations of engine/
structures, structural dynamics & materials conference, airframe integration for high bypass engines. ECCOMAS 2000.
AIAA, Reston, VA, 2005. [149] Ingraldi AM, Kariya TT, Re R, Pendegraft OC. Inter-
[131] Department of Defense. Parametric estimating handbook, ference effects of very high bypass ratio nacelle installations
2nd ed. DoD; 1999. on a low-wing transport. J Eng Gas Turbines Power Trans
[132] Goranson HT. The agile virtual enterprise : cases, metrics, ASME 1992;114(4):809–15.
tools. Quorum Books; 1999. [150] Rossow CC, Hoheisel H. Numerical study of interference
[133] Camarinha-Matos LM, Afsarmanesh H, Garita C, Lima C. effects of wing mounted advanced engine concepts.
Towards an architecture for virtual enterprises. J Intell International Council of the Aeronautical Sciences (ICAS)
Manuf 1998;9(2):189–99. 94-6.4.1, 1994.
[134] VIVACE—/www.vivaceproject.comS. [151] Jie L, Fengwei L, Qin E. Numerical simulation of transonic
[135] Fiper. Cary, NC: Engineous Software Inc.; 2005. Version flow over wing-mounted twin-engine transport aircraft.
1.6. J Aircraft 2000;37(3):469–78.
[136] Kundu A, Raghunathan S, Morris W. Effect of manufac- [152] Rudnik R, Rossow CC. Numerical simulations of engine/
turing tolerances on aircraft aerodynamics and cost. In: airframe integration for high bypass engines. Aerospace Sci
World manufacturing congress, Paper 711-003, New Technol 2002;6:31–42.
Zealand, November 1997. [153] Wilhelm R. An inverse design method for engine nacelle
[137] Kundu A, Raghunathan S, Cooper RK. Effect of aircraft and wings. International Council of the Aeronautical
smoothness requirements on cost. Aeronautical J 2000; Sciences (ICAS) 2004-2.7.2, 2004.
104(1039): 415–20 [Paper 2389]. [154] Carlson JR, Lamb M. Integration effects of pylon
[138] Curran R, Kundu A, Raghunathan S, McSpadden R. geometry on a high-wing transport airplane. NASA
Impact of the aerodynamic surface tolerance on aircraft technical paper 2877, 1989.
cost driver. J Aerospace Eng Proc. IMechE Part C [155] Mavriplis D. Aerodynamic drag prediction using unstruc-
2001;215:29–39. tured mesh solvers. VKI lecture series, CFD-based aircraft
[139] Kundu A, Watterson JK, Raghunathan S, McFadden R. drag prediction and reduction, 2003.
Parametric optimisation of manufacturing tolerances [156] Riedel H, Horstmann KH, Ronsheimer A, Stizmann M.
at the aircraft surface. J Aircraft AIAA 2002;39(2): Aerodynamic design of a natural laminar flow nacelle and
271–9. design validation by flight testing. Aerospace Sci Technol
[140] Woods K, Curran R, Raghunathan S. University— 1998;2(1):1–12.
industry technology transfer in the aerospace industry— [157] Rumsey CL, Rivers SM, Morrison JH. Study of CFD
the way forward. In: Proceedings of the 19th CASI validation on transport configurations from the second
conference, Canadian Aeronautical and Space Institute, drag prediction workshop. In: AIAA 43rd aerospace
Montreal, 2003. sciences meeting and exhibit, 5–8 January, Reno, NV,
[141] Sanchez M, Kundu AK, Hinds BK, Raghunathan S. A USA, paper 2004–0394.
methodology for assessing manufacturing cost due to [158] Devine R, Crawford B, Benard E, Cooper R, Raghunathan
tolerance on aerodynamic surface features on turbo fan S. A computational investigation of propulsion integration.
nacelles. Int J Adv Manuf Technol 1999;14:894–900. In: AIAA 4th aviation technology, integration and opera-
[142] Price M, Curran R, Murphy A, Raghunathan S, Early J, tions (ATIO) forum, Paper No. AIAA 2004-6340, Chicago,
Castagne S, et al. Integrating design, manufacturing and IL, 20–22 September 2004.

Price 2006

Uploaded by

Copyright:

Available Formats

Price 2006

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Price 2006

Uploaded by

Copyright:

Available Formats

ARTICLE IN PRESS

Progress in Aerospace Sciences 42 (2006) 331–376

An integrated systems engineering approach to aircraft design

1. Introduction and background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333

6.3.1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367

1. Introduction and background Perhaps the biggest challenge in this suite of

Low Empirical Empirical Empirical

Low Empirical Empirical Empirical Empirical

Requirements Conceptual Preliminary Detail Manufacture

2.1.1. Conceptual design

follow a SE approach where there is continual

Functional Analysis and

Fig. 6. Complex ﬁnite element analysis models can be generated

1.3 1.4 1.5

1.1.1 1.1.2 1.1.3 1.1.4

1.1.2.1 1.1.2.2 1.1.2.3 Data used for

Fig. 9. Simpliﬁed system schematic.

Point Line Surface

Parameters indicating station

Forward Fuselage Centre Fuselage Aft Fuselage

Upper Half Lower Half

Starboard Side Panel Parameters

Fwd Fuselage Centre Fuselage Other sub systems

Upper Half Other sub systems

Fig. 16. Mixed dimensional model showing local solids at the

ing beings from the junctions or interfaces as they

Fwd Fuselage Centre Fuselage Other sub systems

Upper Half Other sub systems

Fulfilling a need; by doing; physically described by

R (0) F (0) P (0)

R (1.1) R (1.2) F (1.1) F (1.2) P (1.1) P (1.2)

Function Performed to Physical System to

Provides Initial High Level Describes characteristics

Fig. 20. Relationship of RFP architectures to the ‘Technical’ architecture.

The relationships between the ‘entities’ (the

associated sets of input and output parameters.

3.6. Parameters versus constraints

The issue of the impact of design parameter

not appear to be a formal approach to addressing this

Poor Design Flexible Design

how this is being applied to design. It appears that 100 95%

Fig. 25. Systems engineering model for acquisition costing.

Life cycle phase: DE AC/CBR NN/FL PE ABC GCC

Concept and technical development

System development and demo

Production and acquisition

Operations and disposal

Fig. 27. Matrix of cost estimating techniques.

cured from the supply chain. Procurement cost

Fig. 30. Modelling production cost at a cost component level.

interpreted as the non-dimensional error, where 4.7.4. Operations cost

Insurance manufacturing cost, whose inclusion into the

Digital manufacturing tools can be used to link

2500 at the heart of the system to integrate the design

Fig. 34. Simulation as a tool for veriﬁcation of optimal manufacturing solutions.

Fig. 34 shows plots of the process and operator Stringer area

efficiency η 0.693 42.8 27.6 0.85 1.61 31.1

Fig. 36. Panel dimensions after optimisation.

Costs Table Configuration 1 Configuration 2