See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/230688298

1.5.2 Systems Engineering in Aircraft System Design

Article  in  INCOSE International Symposium · July 2001

DOI: 10.1002/j.2334-5837.2001.tb02363.x

CITATIONS READS
5 3,059

1 author:

Petter Krus
Linköping University
311 PUBLICATIONS   1,770 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Swedish Aeronautical Research Center View project

System of Systems Engineering (SoSE) View project

All content following this page was uploaded by Petter Krus on 05 January 2018.

The user has requested enhancement of the downloaded file.

 Systems Engineering in Aircraft System Design
Petter Krus
Department of Mechanical Engineering
Linköping University
SE-58183 Linköping, Sweden

design. In particular traceability between upper

ABSTRACT level design requirements and desirables and detail
In order to be competitive, companies have to design should be ensured. In this way the proper
provide products that perform more and better than trade-off can be made even at a detail level. This is
earlier ones at a lower cost. This inevitably leads to in contrast to the conventional way, where certain
a higher density of functionality in products, and as properties are expected from subsystems and are
a consequence there is a trend to integrate different sent down as requirement. Using proper systems
technologies such as mechanical, electrical and engineering tools, it is possible to have engineers at
fluid subsystems and embedded software. More of subsystem and component level more involved and
the engineering effort is therefore placed on system more creative in pushing the top level properties in
architecture and system integration. It also becomes the most favourable direction.
more important to find and to express requirements It is recognised that very substantial parts of
more exactly in order to be able to emphasise the the design process consists of manual design work
right ones. It also means that the product has to involving the inspiration and creativity of the
more tailored the customer needs and that analysis designer, and that there is a lot of effort put into
of systems of systems where the product is to developing design methodologies and procedures
operate, becomes more important. that explicitly involves the role of the designer.
There are several levels of design from This paper, however, concentrates on the parts of
requirement analysis and system architecture down the design process that can be formalised as well as
to detail design There is a clear danger that systems how formal methods effects the design process as a
engineering activities are performed only at top whole. This includes optimisation methods,
level of a design. In order to have an impact on the sensitivity and robustness analysis, as well as
product development process it must, however, formal method for defining system requirements,
permeate all levels of the design in such a way that and methods to ensure traceability in design
a holistic view is maintained through all stages of GENERAL DESCRIPTION
the design.
One of the main contribution of systems
INTRODUCTION engineering is that it should make it possible to
Systems engineering is generally associated with allow each engineer at any level of the design to
the upper level of the design process. It is more relate to the top level requirements. Conversely, it
about system architecture than component design. should be possible to pinpointing critical areas in
At the chore of system engineering, however, lies the detail design from a top-level perspective. The
the idea that all the details in a design should be aircraft design process can loosely be described as
traceable to the requirement of the system. This is having three levels.
illustrated by two definitions of Systems
Engineering. Systems of systems

The interdisciplinary approach governing the

total technical effort required to transform a Aircraft
requirement into a system solution. (IEEE
P1220) configuration design

The branch of engineering dealing with the

design of a complex interrelation of many Subsystem design
elements (a system) to maximise an agreed-upon
measure of system performance.(McGraw Hill
Dictionary)
Figure 1.
In this paper systems engineering is understood System hierarchy
as formal methods to connect different aspects of
System of system analysis. At this, the top level, a advantage with this approach is that it allows the
whole scenario or mission is analysed. Here the designer to optimise complex non-linear systems in
primary system requirements are defined and a convenient way, directing the designers effort to
analysed. At this level there is also support for the requirements and system objectives rather than
breakdown into separate functions. It should also to the actual computation of system parameters.
be possible to bring models from the lower levels
of detail design up to this level for verification of Generating Concept Concept
the total system functions. It should also be requirement generation optimisation
specification and selection
possible to obtain the measurable on the impact of
detail design decisions on the over all system
performance. Meta-modelling
and post-optimal
Aircraft preliminary design. Requirements for the analysis
aircraft design have been defined at the top level.
At this level the conceptual design of the aircraft is
made. The relations between system requirements
and design parameters are quantified. This is done Concept optimisation and selection
both using detail computational methods for
structure and aerodynamics as well as simulation Parameter System
optimisation model
techniques for flight dynamics where system
models from the subsystem design are included.
Evaluation
The requirements and desirables of the subsystem (simulation)
are also defined at this level.

Subsystem design. At this level the basic aircraft Figure 2. System development
systems are designed. Relations between system process
requirements and system parameters are established
for the negotiation of the requirements. Computational models. In order to evaluate
Although systems engineering can have a different concepts, models at an adequate level of
profound effect on the product development details have to be established. This is a very
process, the most important effect is that using important task since the design becomes no more
systems engineering, all aspects of the detail design accurate than the underlying models. One of the
are possible to trace to top-level design most important shifts in paradigm occurring in
requirements and desirables, and vice-verse. This engineering system design may well be the
adds enormously to the quality of the work as well adoption of common system models as a
as it much easier at later stages to make foundation for system design. Efficient models for
modifications and redesigns. This is particularly complete aircraft systems simulation can be
true in the aeronautics industry where product established using newly developed tools, where
cycles are long and implicit requirements are complete systems can be simulated more or less in
forgotten over time. real time. This allows for a much more effective
product development process since a system can be
KEY TECHNOLOGIES tested in all stages of design.
The parts of the design process involved in this Meta-modelling techniques and post-optimal
research area are shown in Fig. 2. From this figure analysis for traceability in design. These
some key technologies needed for systems techniques are very closely tied to the requirement
development can be identified. specification and they involve the estimation of
sensitivities between design parameters and the
Requirement specification. Formal methods for functional characteristics. This can also be
establishing the requirements and the desirables of generalised to estimate the influence of aggregates
a design are needed in order to ensure the of parameters, such as whole subsystem on
traceability between requirements, desirables and requirements. This can be presented as an
the design. This also involves initial analysis of the aggregated design impact matrix. Sensitivities can
value of fulfilling certain customer requirements also be generalised to involve higher order
vs. the cost to do so. This also connects to the post- functions, such as quadratic, which means that
optimal analysis. approximate analytical expression for the relations
Concept optimisation and selection. Based on between design parameters and functional
computational models, optimisation of different characteristics. This is a valuable tool for studying
configurations can be carried out. Optimisation has parameter variations around a design point when
proven to be an extremely useful tool when the actual underlying models are computer
connected to evaluation models. These can for intensive, and for tying the top-level requirements
instance be dynamic simulation models. The and desirables to the low-level detail design.
 EXAMPLES OF TOOLS
R > Rref
The aggregated design impact matrix. The SLo< SLor
aggregated design impact matrix (ADIM) is a tool SL < SLr
to calculate and present the relative importance of W1 > W1r
whole components and subsystems (instead of
individual parameters) on different system
characteristics. The matrix formulation for PENALTY FUNCTIONS FOR
representing the mapping between functional CONSTRAINTS
requirements (FR:s) and design parameters (DP:s)
is a well established practice. (N P Suh 1990). This Since the allowed parameter space is only a small
is also the case in the house of quality (Hauser and fraction of the total parameter space it is necessary
Clausing). In this paper inspiration from these to handle the constrains as penalty functions, which
concepts are used to form a formal procedure means that solutions in the forbidden area can be
linking optimisation to a matrix representation of used to find the direction to the allowed parameter
design relations. space. A penalty function has a value that increases
When an optimal solution has been found in rapidly if constraints are violated. This ensures that
some way, it is of great interest to investigate the the solution will be pushed inside the constraints.
optimum from different points of view, i.e. to see Well inside, the penalty functions are set to zero.
the sensitivity of constraints, sensitivity to The total objective function is then set up as:
parameter variations and disturbances. In this way N

it is possible to revise the requirements and to f obj = f obj ,0 (∑ f p ,i + 1) (4)

detect areas in the design which requires high i =1
attention because the design is heavily dependent where
on them. For large systems, however, the overview f obj ,0 = W1 + W f (4)
is lost and some means to introduce hierarchy into
the analysis is essential, and therefore introduced in The total objective function for the example. For
this paper. the aircraft example the objective function and the
constraints are used to formulate the following total
APPLICATION: OPTIMISATION FOR
object function.
PRELIMINARY SIZING OF AN AIRCRAFT
In order to exemplify the use of optimisation and THE OPTIMISATION METHOD
the establishment of the aggregated design impact The COMPLEX-RF optimisation method used here
matrix, a very crude aircraft model is made. This is a modified version of the COMPLEX method by
model is of the kind used for first estimation of an Box (1965), which in turn is based on the
aircraft design that can be found in text books such SIMPLEX method. It is modified by introducing
as (E. Torenbeek 1981) or (Raymer 1989). some randomisation in the search. This avoids
Consider the following optimisation problem. The premature collapse of the method. It also include a
system characteristics cs are computed from the forgetting factor which ensures that the COMPLEX
system parameters ps is made up predominantly with recent parameter
sets, this has also been found to improve the
cs = F ( p s ) (1)
success rate. This method is a non-gradient method
and has been used very successfully over a wide
The object function is in general a function of range of problems and is characterised of simplicity
system characteristics and system parameters (It and robustness. This is, surprisingly, not reflected
can also be defined as a separate system in its rather limited popularity.
characteristic).
SENSITIVITY ANALYSIS
f objf = f objf (cs , ps ) (2)
In order to obtain sensitivities for the post optimal
Both the system characteristics and the system analysis a recursive least square (RLS) scheme is
parameters may be subject to constraints. used. Of course it would be possible to do variation
There are constraints in range R, and lift of and around the optimum to obtain the sensitivities, but
landing length, Slo, SL. there are also structural since the COMPLEX method goes trough a
constraints meaning that the weight of the unfuled sequence of evaluations of parameter sets that are
aircraft W1 must be greater than an estimated homing in on the optimum a more elegant solution
required weight W1r. The objective is to minimise is to use these evaluations instead. The recursive
the take off weight, which is W1+Wf. least square method continuously estimates the
sensitivities and is ready to present the estimate
fobjf = W1 + Wf (3) whenever the optimisation has converged.
If the system is complex and the sensitivity
subject to the constraints: matrix large, it may be difficult to get an overview
of the system since the different parameters may different system characteristics.
have values of different orders of magnitude. The
system characteristics are normally also of different
Propulsion Structure fuel
orders of magnitude. In order to make it easier to
Wing T W1 Wf
get an overview of the sensitivities some kind of
R 1.18 0 -0.05 0.45
normalised dimensionless sensitivities are needed. SLo 2.69 -0.29 0.37 0.03
One approach is to normalise the sensitivities with SL 1.78 0 0.33 0.03
respect to the nominal system parameters and W0r 0 0 0.79 0.21
system characteristics so that the normalised Vs 1 0 0.31 0.03
sensitivity shows how many percent the system VLo 1 0 0.31 0.03
characteristic is changed if a system parameter is VL 1 0 0.31 0.03
changed one percent.
The result from the optimisation and the
sensitivity identification is shown in the table Table 2. Aggregated design impact
below, showing the sensitivities in normalised matrix with seven original parameters
form.
This contains the optimum value of the object In this way, the matrix when shown at the top level
function and values for the design parameters B, will have the same appearance regardless of how
Cr, Ct, tc, T and W1 at the optimum. In addition detailed the underlying system is. It is also
the sensitivities between all the system parameters interesting to note that the values in the ADIM are
and system characteristics are shown. In this way it only changed slightly as a more detailed model is
is possible to estimate the tolerance for the used. From table 2 it can be concluded that for
parameters needed to fulfil the constraints. In range the wing and the amount of fuel are the most
addition to system characteristics associated with important components. If all parameters in the wing
requirements, stall speed at cruise, takeoff speed are changed one percent in the most favourable
and landing speed are also presented as system direction it will change the range 1.18%.
characteristics. The requirements on range R and Using the aggregated design impact matrix, it
lift-off distance are invoked since they have the is possible to allocate resources in a way that
same value as the requirement. The same thing reflect the real importance of different subsystems,
goes for W0r, which has the same value as the sum and to emphasise on the important areas. As the
of W1 and Wf. work progresses, the sensitivity matrix becomes
If the number of system parameters and system progressively larger and more complex. The
characteristics are the same it is possible to invert aggregated design matrix, however, remains largely
the sensitivity matrix and study the influence on the unchanged during the whole system development,
system parameters if constraints are moved. which means that a single format of information
can be maintained, at the top level, through the
B Cr Ct tc T W1 Wf
design.
R -0.53 -0.42 -0.18 -0.05 0 -0.11 1.03
SLo -2.44 0.2 -0.05 0 -2.52 3.2 0.27 COMPLETE AIRCRAFT SYSTEM
SL -0.02 -1.29 -0.47 0 0 0.92 0.08 SIMULATION
W0r 0 0 0 0 0 0.43 0.11
Vs -0.5 -0.35 -0.15 0 0 0.46 0.04 For more detailed level of system development,
VLo -0.5 -0.35 -0.15 0 0 0.46 0.04 modelling and simulation is of crucial importance
VL -0.5 -0.35 -0.15 0 0 0.46 0.04 for the understanding of system behaviour. In
aircraft, simulation has been strong in the area of
flight control. Modelling and simulation of the
Table 1. Normalised sensitivities. hydraulic systems has also a long tradition. The
rapid increase in computational power has now
AGGREGATED DESIGN IMPACT come to a point where complete modelling and
MATRIX simulation of all the sub-systems in an aircraft is
It is obvious that if the number of parameters is not far away. This means new challenges in dealing
increased substantially as it would in a real design with very complex multi domain systems.
as it becomes more detailed, the overview is The development in system modelling has
quickly lost. However, since normalised come to the point where complete modelling of
sensitivities are used, it is possible to aggregate all systems is possible, e.g. the complete hydraulic
parameters associated with one component by system and interfacing systems in an aircraft.
simply adding them together. This has, however, Complete modelling does not mean that all
the drawback that different sign in sensitivities may components are dealt with down to the very
cancel each other out. Therefore it is better to add smallest details of their behaviour. It does,
the absolute values of all sensitivities instead. This however, mean that all functionality is modelled, at
will at least give a value to the relative importance least qualitatively. Furthermore, in contrast with the
and hence the magnitude of impact from design usual problem oriented approach, the tests to be
changes in a component or subsystem to the simulated with the model are not explicitly known
when the model is established. The aim of this simulation. The distributed modelling technique is
thematic area is to develop a framework for a highly robust method suitable for simulation of
complete modelling of the functionality in an large-scale systems. It also makes it possible to
aircraft. Areas of interest are: partition the system for parallel processing. This
means that it is possible to simulate very large
• Flight dynamics model. systems within reasonable time. It is also a valid
• Actuator system model, including complete paradigm for multi domain system. This is in
system model of hydraulic system. contrast to conventional centralised solvers that
• Landing gear. flattens the system of equations for the simulation.
• Electrical power system. Furthermore, object oriented modelling is also a
base for object oriented design.
• Propulsion, including engine system and vector
Using the HOPSAN simulation tool it is
control system.
possible to model and simulate flight dynamics
• Fuel system together with actuation systems and propulsion
• Air system, cooling and ventilation. (and potentially any other system too), Fig 3. In
this way it is possible to analyse the influence of
The aim is to develop generic, query subsystems on the whole aircraft for full system
independent models that can be used as test beds optimisation, and to calculate sensitivities to
for analysis of a wide range of test applications. include in the aggregated design impact matrix.
This means that subtle couplings between different
subsystems can be detected and dealt with at an
early stage of system development. The main
objections against this approach has been the
computational resources needed to simulate such a
large system and the difficulty to model large
systems. Using a complete model it is possibility to
use one model for a wide range of analysis. From
mission performance to detailed behaviour on a
component level.
The rapid development in simulation methods
and the general increase in hardware performance
imply that simulations can be run thousands of
times faster than just a decade ago. As a Figure 3. Comple system simulation
consequence, there is no longer such need to use in the HOPSAN simulation package
very economic, more or less analytical methods in
system design. This means that design methods CONCLUSIONS
based on different kinds of numerical optimisation
Systems engineering is here understood as a means
for system design, are becoming much more
to connect all aspects of design in such a way that
important. Modelling and simulation has now come
each design decision can be checked against top
to a point where it is changing the way engineering
level requirement. In this paper some tools for
is done, the true potential of simulation being
system engineering in aircraft system design has
relished when it is connected with other tools to
been presented.
support the design process.
The aggregated design impact matrix ADIM is
OBJECT ORIENTED MODELLING a very powerful tool for system design and
TECHNIQUES analysis. The primary benefit from using ADIM is
that it is a hierarchical method to identify the
Object orientation is a powerful paradigm for importance of different areas in a design, and that it
modelling of complex systems since a close can be used at all levels in design. At the top level
mapping between the structure of the real system it can be used to quickly identify the areas of
and the model can be maintained at all levels of the importance in a design. At the bottom level it can
model. Using object-orientation it is possible to be used to represent the impact of an individual
model all aspects of a component in one object. design parameter on the top-level requirements. In
The objects can then be combined for system this way all engineers becomes involved in pushing
modelling. It is desirable to use a tool-independent the top-level system characteristics in the desirable
modelling language in order to have the models direction.
tool independent. One of the most important shifts in paradigm
The new modelling language Modelica has occurring in engineering system design may well
been developed for this purpose. A translator from be the adoption of common system models as a
Modelica to the HOPSAN simulation package has foundation for system design. This will allow for a
under development. The HOPSAN package utilises much more effective product development process
a distributed modelling technique that allows that since a system can be tested in all stages of design.
the object structure is maintained also through the Simulation models for complete vehicle simulation
 can be established, as well as performance models Mavris, D N and S Qiu, 'An Improved Precess for
for various aspects of the design. In this way it is the Generation of Drag Polars for use in
possible to carry out optimisation and sensitivity Conceptual/Prelininary Design, AIAA/SAE
analysis at a high level of design using detailed World Aviation Conference, San Fransisco,
system models. USA, 1999.
REFERENCES Raymer D P, 'Aircraft Design: A Conceptual
Approach', AIAA Education Series, ISBN 0-
Andersson J, Pohl J, Krus P: ‘‘Design of Objective 930403-51-7, 1989.
Functions for Optimization of Multi-domain
Systems’’, ASME Annual Winter meeting, Suh.N P ‘‘The Principles of Design’’, Oxford
November 15-20, Anaheim, USA. 1998 University Press 1990.
M. J. Box. A new method of constrained Torenbeek E, 'Synthesis of Subsonic Airplane
optimisation and a comparison with other Design', Kluwer Academic Publishers, ISBN
methods. Computer Journal, 8:42--52, 1965. 90-247-2724-3, 1982.
Hauser, J. R. and D. Clausing "The House of
Quality," The Harvard Business Review, May- BIOGRAPHY
June, No. 3, pp. 63-73. 1988.
The author is professor in Mechanical Engineering
Krus P, A Jansson, J-O Palmberg, 'Optimization Systems at the Department of Engineering,
Using Simulation for Aircraft Hydraulic Linköping University. He received a PhD at the
System Design', Proceedings of IMECH same University in 1988. In 1993 (- present) he
International Conference on Aircraft became head of the Master program in Mechanical
Hydraulics and Systems, London, UK, 1993 Engineering. Research areas include fluid power
control, modelling and simulation, aeronautics,
Krus P, S Gunnarsson, 'Numerical Optimization system optimisation and biomechanics. He is also
for Self Tuning Electrohydraulic Control active as a supervisor within the Swedish national
Systems'. Proceedings of JHPS International graduate school in engineering design, ENDREA.
Symposium on Fluid Power, Tokyo, Japan,
1993.

View publication stats

Previous Menu/Document