Вы находитесь на странице: 1из 12

Regular Paper

Application of Axiomatic
Design and Design
Structure Matrix to the
Decomposition of
Engineering Systems
M. D. Guenov* and S. G. Barker

Department of Power, Propulsion and Aerospace Engineering, School of Engineering, Cranfield University, Wharley End,
Bedfordshire, MK43 0AL, United Kingdom

Received 25 February 2004; Accepted 1 August 2004, after one or more revisions
Published online in Wiley InterScience (www.interscience.wiley.com).
DOI 10.1002/sys.20015

A design decomposition-integration model, named COPE, is proposed in which Axiomatic
Design Matrices (DM) map Functional Requirements to Design Parameters while Design
Structure Matrices (DSM) provide structured representation of the system development
context. In COPE, the DM and the DSM co-evolve. Traversing between the two types of
matrices allows for some control in the application of the system knowledge which surrounds
the decision making process and the definition of the system architecture. It is argued that
this approach describes better the design process of complex products which is constrained
by the need to utilise existing manufacturing processes, to apply discrete technological
innovations and to accommodate work-share and supply chain agreements. Presented is an
industrial case study which demonstrated the feasibility of the model. © 2004 Wiley Peri-
odicals, Inc. Syst Eng 8: 29–40, 2005

Key words: axiomatic design, design structure matrix, systems decomposition, systems

*Author to whom all correspondence should be addressed (e-mail:
Standards for the engineering of systems such as EIA
Contract grant sponsor: UK EPSRC Research Grant GR/R37067 632 and ISO/IEC 15288 support a seamless process of
under the Systems Integration Initiative. converting customer needs into systems/technical re-
Systems Engineering, Vol. 8, No. 1, 2005
quirements, which are subsequently transformed into
© 2004 Wiley Periodicals, Inc. logical representations (architectural design) and fi-


nally into physical solution representations. The proc- allows the designer to attach constraints and text frag-
ess is iterative (see Fig. 1) due to the incomplete infor- ments to each item within each layer of the system
mation available at the outset of the project and also the decomposition, and also provides a good level of trace-
large number of constraints involved—technical, eco- ability of nonfunctional requirements throughout the
nomic and even political. The systems approach of design decomposition. SLATE, however, would only
tackling the problem includes the deployment of Inte- produce results which reflect the experience of those
grated Product Teams (IPT). These are composed of a using it. Observations made during the industrial case
variety of specialists from different functional disci- study suggest that, in practice, even experienced sys-
plines, ideally representing different phases of the prod- tems designers are too quick to follow a particular
uct lifecycle to ensure that the design process will solution path, relying heavily on existing knowledge
consider, as early as possible, all relevant requirements and past concepts.
and constraints. IPTs are now a common practice in There are also other methods of structuring the de-
industries such as the defense and aerospace. What sign process, such as the N2 chart [Lano, 1977, 1979],
seems to be lacking, however, are high-level support and the Design Structure Matrix (DSM) approach
tools which could help the systems teams and architects [Steward, 1967, 1981]. The latter approach provides the
draw together consistently a vast amount of information ability to group related design elements. There are
needed for the requirements and the design decompo- software tools available which are based on the DSM.
sition of the system. Currently there exist a number of
These include DeMAID [Rogers, 1999], which was
requirements management tools, which fulfill only par-
developed by NASA to facilitate the decomposition and
tially this need. For instance, “Acclaro” [Suh, 2001;
sequencing of design activities, and more recently,
Axiomatic Design Software Inc. website] is designed
PlanWeaver (ADePT [Adept Management website])
to evolve the systems design in accordance with the
which has been applied mainly in the construction
rules of Axiomatic Design Theory (ADT). The software
allows the systems designer to enter the top-level Func-
This paper presents work which has concentrated
tional Requirements (FRs) and Design Parameters
(DPs), and to decompose and map those in two tree particularly on the integration of ADT and DSM. The
hierarchies and associated design matrices. However, conjecture is that ADT and DSM are complementary
Acclaro is concerned primarily with functional decom- parts of the decomposition–integration problem, where
position, and not with explicit constraint mitigation and the former is more concerned with mapping of func-
control. SLATE, on the other hand [Talbott et al., 1994a, tional requirements to design parameters, while the
1994b], now part of the EDS Team Center suite of latter is better suited to modeling the interactions and
software, is a good example of a powerful requirements the integration of the design parameters. The main
management system. It provides constructs not only to objective of the work is to investigate to what extent
build requirements and product hierarchies, but also those two methods can be integrated and to evaluate the
approach in a realistic industrial setting. The potential
value of this work is that it provides a means of relating
component integration to system functionality, which
otherwise is not available but is essential during the
early stages of the design process. The following two
sections present a brief summary of ADT and DSM,
respectively. Section 4 outlines the Methodology, while
Section 5 describes the industrial case study undertaken
to test and evaluate the approach. Finally conclusions
are drawn and future work outlined.


The underlying hypothesis of the Axiomatic Design

[Suh, 1990, 2001] is that there exist fundamental prin-
ciples that govern good design practice. The main dis-
tinguishable components of the ADT are domains,
Figure 1. High-level view of the systems engineering process hierarchies, and design axioms. The foundation axioms
(adapted from ANSI/EIA-632-1998). are:

Figure 2. Decomposition by zigzagging.

Axiom 1. Maintain the independence of the func- be determined for the same hierarchical level in the
tional requirements.1 physical domain. This iterative process is called zigzag-
Axiom 2. Minimize the information content of the ging (see also Tate [1999] for a more thorough treatment
design. of the decomposition problem).
Zigzagging also involves the other domains since
Axiom 1 requires that Functional Requirements manufacturing considerations may constrain design de-
(FRs) be independent of one another, which enables cisions, while “over-specified” requirements could vir-
each FR to be satisfied without affecting any of the other tually prohibit the discovery of feasible design
FRs. Thus there is no coupling of function where it can solutions. At each level of the design hierarchy, the
be avoided, and the design remains as uncomplicated as relations (the dependencies) between the FRs and the
possible. Axiom 2 provides a quantitative measure of DPs can be represented in an equation of the form:
the merits of a given design, and thus it is useful in
selecting the best among the designs which satisfy FR = [A]DP, (1)
axiom one [Suh, 2001]. Generally, the design that uses
where each element of the design matrix [A] can be
the least information is superior. This is explained in
expressed as Aij = ∂FRi/∂DPj (i = 1,…, m and j = 1,…,
more detail later in this section. According to the ADT, n). Equation (1) is called the design equation and can
the design process takes place in four domains (Fig. 2, be interpreted as “choosing the right set of DPs to
adapted from Tate [1999] and Suh [2001]): Customer, satisfy given FRs.” This is required by Axiom 1 (Inde-
Functional, Physical and Process. Through a series of pendence). Each element Aij is represented as a partial
iterations, the design process converts customer’s needs derivative to indicate dependency of a FRi on a DPj. For
(CNs) into Functional Requirements (FRs) and con- simplicity the value of an element Aij can be expressed
straints (Cs), which in turn are embodied into Design as 0 (i.e., the functional requirement does not depend
Parameters (DPs). Functional Requirements can be de- on the particular design parameter), or otherwise X.
fined as “a minimum set of independent requirements Depending on the type of resulting design matrix [A],
that completely characterises the functional needs of the three types of designs exist: uncoupled, decoupled and
product in the functional domain”; Design Parameters coupled (Fig. 3).
are “the key physical variables in the physical domain Uncoupled design occurs when each FR is satisfied
that characterise the design that specifies the FRs” [Suh, by exactly one DP. The resulting matrix is diagonal, and
2001, p 14]. DPs determine (but also can be affected by) the design equation has an exact solution; i.e., Axiom 1
the manufacturing or Process Variables (PVs). The de- is satisfied. When the design matrix is lower triangular,
composition process starts with the decomposition of the resulting design is decoupled, which means that a
the overall functional requirement—in practice this sequence exists, where by adjusting DPs in a certain
should correspond to the top system requirement. Be- order, the FRs can be satisfied. The design matrix of a
fore decomposing a FR at a particular hierarchical level coupled design contains mostly nonzero elements, and
in the functional domain, the corresponding DP must thus the FRs cannot be satisfied independently. A cou-
It appears that software engineers realized this earlier, while pled design can be decoupled, for example, by adding
learning from some bad designs in the automotive industry [see components to carry out specific functions—however,
Stevens, Myers, and Constantine, 1974: 139]. this comes at a price.

Figure 4. Probability distribution of a design parameter. The

solid line represents uniform distribution, the dotted line a
nonuniform distribution (adapted from Suh [1990]).

Figure 3. Examples of design types. From top to bottom:

Uncoupled, Decoupled, and Coupled Design.
(manufacturing) system; common range is the overlap
between design and system ranges, and specifies the
range within which the FR(s) can be met [Suh, 2001].
One additional factor that affects coupling is the The information content of a system can be computed
number of FRs, m, relative to the number of DPs, n. If by summation of the individual information contents of
m > n, then the design is either coupled or the FRs all DPs only if the latter are probabilistically inde-
cannot be satisfied. If m < n, then the design is redun- pendent. Frey, Jahangir, and Engelhard [2000] proved
dant. (Note that in both cases the design matrix [A] is that simple summation cannot be performed for decou-
not square.) pled designs and offered a method for computing infor-
In addition to the design axioms, ADT is governed mation content. There is no exact method for computing
by a set of rules, which are formalized into theorems the information content of coupled designs.
and corollaries. Of particularly relevance to this re-
search are Corollary 3 and Theorem 5 which originate
from the first axiom (Independence):

• Corollary 3 states: “Integrate design features in a The DSM can be seen as a successor to the N2 chart
single physical part if the functional requirements [Lano, 1977, 1979, Becker, Ben-Asher, and Ackerman,
(FRs) can be independently satisfied in the pro- 2000], which was used for some years by the Systems
posed solution.” Engineering community, before the introduction of the
• Theorem 5 states: “When a given set of FRs is DSM. The DSM has become increasingly popular as a
changed by the addition of a new FR, by substi- means of planning product development, project plan-
tuting one of the FRs with a new one, or by ning and management, systems engineering, and organ-
selection of a completely different set of FRs, the izational development [Browning, 2001, 2002]. The
design solution given by the original DPs cannot concept dates back to the 1960s and was not actually
satisfy the new set of FRs. Consequently, a new published until 1981 [Steward, 1981]. The Design
design solution must be sought.” Structure Matrix (DSM) is a compact, matrix repre-
sentation of a system/project. The matrix contains a list
The Second Design Axiom states that minimizing the of all constituent subsystems/activities and the corre-
information content of a DP increases the probability sponding information exchange and dependency pat-
of success of satisfying a function. The information terns. That is, what information pieces (parameters) are
content is defined by the equation required to start a certain activity and where does the
information generated by the activity feed into—i.e.,
I = log 
system range 
. (2) which other tasks within the matrix utilize the output
 common range 
information [Design Structure Matrix website]. There
In Figure 4 design range is the tolerance within which are three different configurations of the matrix (Fig. 5).
the DP can vary; system range is the capability of the The simple, parallel configuration, in which the design

Figure 5. DSM configurations.

elements (e.g., design parameters or activities) are fully larization, and planning of organizational topolo-
independent of each other, the sequential, “decoupled” gies
configuration, in which the second parameter is de- • The temporal DSM, which is primarily used for
pendent upon the output of the first, and the fully mapping of design processes and for scheduling
coupled configuration, in which parameters are interde- and management of activities, over time
pendent upon each other. Figure 6 shows how a DSM
This research is concerned with the use of DSM to
may be used. The crosses below the diagonal in figure
model the integration and connectivity (logical and
6a represent a forward flow of information, while those
physical) between design embodiments of system ar-
above the diagonal depict a feedback loop. The DSM
chitectures, and to trace the effects of this integration
can be used to reorder, or “partition” design elements,
on the functionality of the system.
to minimize feedback loops (Fig. 6b). This is achieved
While DSM provides a powerful technique for the
by a process of “triangularization” [Browning, 2002],
analysis of design interactions within a complex devel-
to render the matrix as “lower-triangular” as possible.
opment program, it appears to be less effective in inno-
Sets of feedback loops can also be removed, by “tear-
vative design. As Dong and Whitney [2001, p 11] point
ing” them from the matrix—in practice this means
out, it is not possible to “obtain a DSM for a new
producing a set of assumptions for the (as yet) unknown product that has never been designed before.” Addition-
information. The DSM must then be repartitioned. ally, Dong and Whitney maintain that, although the
DSMs can also be used to create and sequence modules DSM captures the interactions between elements within
or clusters [Sharman and Yassine, 2004], which are a system, it fails to record explicitly the reasons for
either mutually exclusive, or minimally interacting. these interactions.
This absorbs any (remaining) iteration within the sys-
tem. This is shown in Figure 6c.
There are two types of DSM in terms of application: 4. DECOMPOSITION-INTEGRATION
• The static DSM, essentially a square matrix, used
for the representation of systems design architec- ADT postulates that a zigzagging process for FR-DP
ture interface, design decomposition and modu- mapping that takes place in a “solution neutral” envi-

Figure 6. Examples of DSM forms: (a) Basic, (b) Partitioned, (c) Clustered.

ronment ensures better design. This, however, can term which encompasses chunks of new technology
rarely happen in practice, particularly in complex prod- which is mature enough to be applied in the new product
uct environments, where economic considerations dic- with potential competitive advantage.
tate maximum possible utilization of mature designs As the decomposition progresses, essential design
and existing manufacturing processes [Guenov, 2002]. studies must be performed, including spatial layout,
In addition, the design process must capture the inter- performance and capacity constraints checks, and/or
actions among the design parameters, including geome- more detailed design of particular, riskier aspects of the
try, spatial layout, interfaces (e.g., logical and physical product, resulting in more interactions between the
connectivity), and so forth. As discussed in the previous design parameters. As a consequence, Corollary 3 of
section, DSM is a mature method capable of capturing ADT may be violated or requirements may need to be
the interaction between design parameters, but by defi- added or changed at a higher level, which would require
nition, it is fully known only for products that have a repetition of the decomposition phase (see Theorem
already been designed. 5 of ADT).
These considerations led us to the idea that the The first step to linking ADT and DSM was made
decomposition integration problem would be better by Dong and Whitney [2001], who showed that if the
modeled as a co-evolution of ADT matrices and DSMs. Axiomatic design matrix (DM) can be expressed ana-
The generic representation of the process which we lytically and one design parameter (DP) is dominant in
named COPE (decomposition-integration of COmplex satisfying a particular functional requirement (FR),
Product Environments) is depicted in Figure 7. In then the triangulated design matrix is equivalent to the
COPE, ADT is used to map the decomposition of DSM of the design parameters. The algorithm consists
functional requirements to design parameters (FR-DP). of three major steps:
At each decomposition level various knowledge 1. Construct the Design Matrix (DM).
sources are consulted in order to take into consideration
2. In each row of the DM choose the dominant (the
constraints originating from all stakeholders. The
largest) entry.
knowledge sources include unstructured ones (e.g., em-
3. Permute the matrix by exchanging rows and col-
ployees’ tacit knowledge and various unpublished
umns so that all dominant entries appear on the
documents) as well as structured/coded sources. Exam-
main diagonal.
ples of the latter include government regulations, DSMs
of past designs (also processes), company’s own design The authors show that choosing the dominant entries is
standards, synthetic and analytical models, knowledge important as it ensures the convergence of the design
bases, and so forth. Technology “bricks” is a generic process.

Figure 7. Schematic representation of the COPE Decomposition-Integration process.


Unfortunately, when designing complex (physical is the “Architectural” DSM, that is, a DSM derived only
or software) systems the FRs are satisfied by modules from the functional view of the product. At this stage a
or other (sub)systems or components and therefore the certain amount of layout/spatial design may need to be
design matrix cannot be represented analytically; and it done, including possible integration of components.
may not necessarily be square either. In order to over- This will be subjected to considerations regarding in-
come this restriction, we have devised a decomposition novation, cost, weight, capacity and other constraints,
procedure (Fig. 8) in which the design and structure which are taken into account by applying the knowl-
matrices co-evolve. edge sources, as discussed above (see also Fig. 4). The
The procedure starts with the construction of the DM resulting design may affect the functionality of the
of the FR-DP hierarchy. At this level the dominant DPs systems, for example, grouping components together
are identified. The DM is manipulated [see Dong and may couple functions. If that is the case, requirements
Whitney, 2001] so that it becomes lower triangular with may need to change or more design parameters may
dominant entries, “X,” on the main diagonal. The result need to be added. This means that the design decompo-

Figure 8. COPE decomposition–integration procedure.


sition has to be reconsidered from the highest level (see the comparison between “As Is” and “As Could Be”
Theorem 5). The decomposition ends when a leaf node would indicate any areas where the existing design
is encountered, that is, further decomposition will could have been improved, thus confirming (or not) our
change the subject of the requirement. In terms of initial hypothesis. Additionally, we expected that the
systems engineering standards (see, e.g., ANSI/EIA- comparison would identify a set of potentially useful
632), a leaf node can be approximately described as a features that have not yet been addressed by the existing
specified/assigned requirement. systems engineering tools.
Perhaps it is worth emphasising that the COPE pro- The findings of the study noted that possibly the key
cedure aims to retrace the AD-DSM connection at each constraint was that this was an update rather than a new
level of the decomposition rather than at reaching a leaf system, and that any design was therefore constrained
node of the DP hierarchy. This is justified by the fact by what already existed. However, the analysis of re-
that in practice one can be very restricted in performing quirements by the BAE SYSTEMS Project appears to
the entire decomposition in a “solution neutral” envi- have been relatively unstructured—and, indeed, an on-
ronment (as advocated by ADT)—there are, for exam- going process. This, when coupled to a predominantly
ple, cost considerations at the outset of the project “bottom-up” rather than “top-down” analysis tech-
which result in targets on reusability of components and nique, meant that the requirements were not decom-
manufacturing processes from previous products and posed fully, and therefore the relation of requirements
/or targets on commonality of components with other to design was not fully analyzed. This has the potential
existing product lines. Furthermore, one sometimes for delay in the form of extended or unnecessary rework
needs to decompose much deeper one branch of the or iteration—the need for which the Project admitted
FR-DP hierarchy relative to other branches in order to was not accounted for. The work breakdown was de-
de-risk the solution. For example, one may need to scribed as being intuitive, and based largely on previous
know if a particular material can be used before con- experience. Thus work appeared to be assigned to teams
tinuing further with the decomposition since this par- without consideration of how all of the modules could
ticular material may affect performance constraints or be integrated together successfully.
interfaces with other components. Thus we see our The design of the system was forced down specific
approach as compliant with the deeper localized studies avenues. This was mainly because it was time con-
that need to be preformed in practice at the conceptual strained and there was a history of rival bids, which,
and preliminary design stages before proceeding fur- through the prime contractor, led to decisions regarding
ther with the development process. the design being taken that were beyond the control of
the BAE SYSTEMS team. Furthermore, the fact that
5. CASE STUDY certain data was provided much later than scheduled,
has also placed certain restraints upon the system design
In order to test our approach regarding ADT (and, process. The organizational need requiring the design,
subsequently, DSM), a case study was undertaken in initially at least, to be planned in conjunction with two
conjunction with COPE Project sponsor BAE SYS- specific suppliers, also applied certain constraints.
TEMS. The chosen study concerns the upgrading of an The need to trace and analyze design decisions and
air defense system. The primary form taken by our changes has been a central requirement in the design of
research was a series of interviews with key members complex products for quite some time [Guenov, 1996].
of the project. These ascertained the nature of the re- In the context of this case study, the DSM offered a
quirements, and the structure of the design. This infor- potential solution to the problem of tracing the effect of
mation was then decomposed using axiomatic design contextual issues such as project scheduling, and event
techniques to identify connectivity between require- sequencing, upon the system design (Pimmler and Ep-
ments and design, and how each impacted the other. pinger, 1994; Ulrich and Eppinger, 1999). Through the
This was known as the “As Is” solution. Impacting application of ADT and DSM, our case study research
factors, or constraints, both within the system of design, discovered elements of the system design and its con-
and of an organizational nature, were studied to model text which required integration to provide a successful
their effect on the process of system design. This was design solution. For example, given that the chosen
validated with the BAE SYSTEMS Project. A parallel design utilized two sensors, one to track the position of
analysis was conducted to determine how the system incoming targets, the other to track an outgoing missile,
could have been designed, had ADT [Suh, 1990, 2001] with the outgoing missile potentially being able to be
and engineering design standards been applied. This seen by both sensors simultaneously, could be seen as
became the “As Could Be” solution. It was intended that an integration issue which requires the application of

Figure 9. 2nd-level Design Matrix (DM).

Figure 10. 2nd-level “Architectural” DSM.

both ADT and DSM. An example of such an application

is described below. second between 2.2.4 and 2.2.3. However, these feed-
The approach for deriving the DSM from the DM is back loops had not been anticipated in the design.
“leveled” so that each hierarchical level of the FR-DP Analysis of the model revealed that the likely reason for
decomposition hierarchy is evaluated in turn. The first their appearance in the structural DSM was a leveling
step is to construct the Design Matrix (DM). For the
issue: DP 2.2.4 verifies that the output of previous DPs
purposes of demonstrating the approach, a 2nd-level
is correct. But the corresponding FR for this DP occurs
abstraction of the ADT decomposition hierarchy will be
concentrated upon. A sample of the 2nd-level BAE at a lower level of decomposition in the DM. This raised
SYSTEMS DM is shown in Figure 9. an issue with the “As Is” model (this being the existing
The design appears to be a redundant one as the solution, modeled in Figs. 9 and 10). A review of the
number of design parameters is greater than the number requirements then led to the creation of an “As Could
of functional requirements. At this stage it is important Be” solution, which is shown in Figure 11a. This at-
to identify the primary relationships between require- tempted to use axiomatic design method [Suh, 1990,
ments and design—i.e., if more than one design pa- 2001] to create a design independently of the existing
rameter is used to satisfy a requirement, identify the “As Is” model, to see if improvements were theoreti-
dominant one, and where appropriate, secondary rela- cally possible to the design. The main changes (the
tionships if they exist (they do not in this example). This original FR/DP nomenclature has been retained for ease
is step two of the approach. The primary relationships of comparison) are an additional FR, stating the need to
are marked with large X’s, while secondary relation-
verify the output of processing tasks, and an additional
ships could be depicted with small x’s. The DM now
DP which enables a further verification task, that of
describes the functional linkages between the Func-
tional Requirements (FRs) and the Design Parameters comparing successive outputs to enhance reliability of
(DPs). This serves to define the manner in which the
DPs will satisfy the FRs. However, the effects of deci-
sions relating to the system such as cost, capacity, and
physical integration are not dealt with particularly well.
As a result, a DSM structure can be generated to accom-
modate these issues. The next step of the approach
concerns the derivation of an “Architectural” DSM.
This is illustrated in Figure 10. A “dummy” FR, denoted
by “?,” is added to make the matrix square. Asterisks
are used to denote blank cells on the diagonal.
The “Architectural” DSM is created through a com-
bination of a method proposed by Dong and Whitney
[2001], and triangulation. Thus the rows and columns
are permuted to produce a (predominantly) lower trian-
gular matrix. The vertical axis is renumbered according
to the DP order, and the “Architectural” DSM is the
result. This is the equivalent of the DM, in DSM form,
and provides the interface between the DM and DSM.
This stage of the case study analysis revealed a
problem with the design solution. Figure 10 shows a Figure 11. (a) 2nd-level “As Could Be” Design Matrix; (b)
feedback loop between DPs 2.2.3 and 2.2.2, and a 2nd-level “As Could Be” “Architectural” DSM.

missile/target detection, up from a lower level of the

This simplifies the FR-DP relationship—whereas,
previously, FR 2.2.2 had been satisfied by three DPs
which embodied the verification algorithms at a lower
level of decomposition, the requirement can now be met
independently of processing operations. This may incur
a cost overhead, depending on the technology used to
implement it, but is representative of the ADT design Figure 13. 2nd-level interaction between processor cards.
philosophy, being in accordance with corollary three of
axiomatic design [Suh, 1990, 2001]. Finally, FRs 2.2.1
processors. The first number denotes the card from
and 2.3.1 of the “As Is” solution (Fig. 9) perform the
which the communication is initiated, and the second
same operation. These have therefore been combined
number indicates which card receives the communica-
into one. The design solution remains uncompromised
as the DPs can be scheduled to cater for the newly
An example is that DP 2.2.1, stationed on card 3,
“compressed” FR. It can be seen in Figure 11b that this
communicates with DP 2.2.5, which appears on card 2.
has resolved the feedback issue.
The DSM can be separated out into layers (shown by
The effects of decision-making on the matrices now
Figs. 12 and 13), to give the system engineer a filtered
need to be modeled. The decisions can regard a range
view of how each of the design elements interacts with
of criteria, such as constraints, physical and logical each other in regard to a different design perspective.
connectivity, and software or hardware integration. In This is illustrated in Figure 14.
the BAE SYSTEMS case, the need exists to combine So far, we have shown that the model can highlight
software functionality onto processor cards. The physi- the effect of integration issues of a system design. This
cal connectivity of the design and the interaction of the is shown by the unbroken arrow in Figure 14. However,
functions across the processors are therefore relevant. it is also necessary to be able to understand how the
Using the “As Is” solution as a basis, the creation of nature of the design is affected by these issues. This
these DSM can be demonstrated by Figures 12 and 13. necessitates backtracking from the layered DSM to the
For the physical connectivity (denoted by 0’s), it can DM, to understand changes which may have been
be seen that of the “processing” DPs (2.2.1, 2.2.2, and caused to the original FR-DP relationship. This is indi-
2.2.3), 2.2.3 operates independently, while 2.2.2 pro- cated by the dotted arrow to the right hand side of Figure
vides data to 2.2.1. DPs 2.2.1 and 2.2.3 then feed 2.2.5. 14, illustrating how the effect of a constraint or decision
The verification DPs (2.2.4 and 2.2.5) also operate regarding integration may alter the balance between
independently, with both passing information to the FRs and DPs on the original DM.
output channel (DP 2.2.6) on completion of the verifi- To provide an example, currently, the processing is
cation task. This completes an operational cycle. verified prior to output. This involves two DPs (2.2.5
The integration between software cards, shown in and 2.2.4), which currently appear on processor cards
Figure 13, is described by numbers. The numbers on the two and three, as shown in Figure 13. If, however, the
diagonal indicate which of three processors the DP is capacity and speed constraints of the processor cards
assigned to. Thereafter, numbers separated by a comma dictated that they be positioned on the same card, then
indicates that two DPs communicate over separate new physical connectivity would appear on the appro-
priate DSM, as both DPs would now be linked via the
same processor card. This is denoted by the shaded area
upon the layered DSM in Figure 14. This in turn would
dictate a new sequence in which the two DPs could be
executed, bringing about a coupling, and altering the
structural DSM/DM, as indicated by the dashed arrow
and letter A.
A further example of this is that an FR specified the
need to detect objects a, b, and c at range x in climate
condition y. To meet this requirement, a DP provides
the required capability. However, cost and capability
Figure 12. 2nd-level physical connectivity between design constraints applied to the DSM mean that it is not
parameters. possible to use the DP. The alternative is a DP which,

Figure 14. Layered DSM and its effect upon the DM.

unfortunately, can only detect objects a and c at range neering life cycle and therefore COPE must be evalu-
x, and not in condition y. It can be seen that the process ated in a wider context. So far we have consulted and
of deriving the DSM to study constraints etc. and their tried to take into account the established standards for
effects has now revealed that the FR on the original DM engineering of systems, however, standards implemen-
cannot be met. Thus the FR must be changed to accom- tation is company dependent. In this view, further de-
modate the DSM constraints, and this changes the de- velopment, evaluation and validation of the method, as
sign, as dictated by theorem five of ADT [Suh, 1990, well as a specification of the requirements for a decom-
2001]. Consequently, a new design solution must be position tool form the scope of our future work. Cur-
sought. rently we are experimenting with the integration of
ADT, DSM, and Requirements Management tools
[Guenov and Barker, 2004].
A novel design decomposition model for complex ACKNOWLEDGMENTS
product environments (COPE) is presented. It com-
bines Axiomatic Design with Design Structure Matrix This work is funded by UK EPSRC Research Grant
to accommodate the iterative nature of the decomposi- GR/R37067 under the Systems Integration Initiative.
tion-integration process. The research to date has dem- We are indebted to BAE SYSTEMS for their help with
onstrated that this approach can bring significant the Case Study. We thank the anonymous referees for
benefits. An industrial Case Study, conducted as part of their helpful comments.
the research and reported here, has shown that the
DM-DSM arrangement can be used to identify the
existence of potential conflicts in the design solution,
and allows groups of design parameters to be explored Adept Management website: http://www.adeptmanagement.
in greater detail. This approach also appears to provide com/
a level of control and transparency to the systems design ANSI/EIA-632-1998, Processes for engineering a system,
process, and gives the systems architect the opportunity Electronic Industries Alliance, Government Electronics
to study proposed changes before they are made, and to and Information Technology Association, Engineering
understand how any such change(s) may produce a Department, Arlington, VA (approved January 7, 1999).
knock-on effect throughout the design solution. Axiomatic Design Software Incorporated website:
Despite the potential which COPE has demonstrated
O. Becker, J. Ben-Asher, and I. Ackerman, A method for
so far, we have not yet solved completely the problem
system interface reduction using N2 charts, Syst Eng 3(1)
of mixing various levels of details during the decompo- (2000), 27–37.
sition process. This is important since deeper localized T.R. Browning, Applying the design structure matrix to sys-
design studies cannot be avoided in practical situations, tem decomposition and integration problems: A review
it is a part of the de-risking process. Secondly, the and new directions, IEEE Trans Eng Management 48
decomposition process forms only a subset of the engi- (2001), 3, 292–306.

T.R. Browning, Process integration using the design structure J.L. Rogers, Tools and techniques for decomposing and man-
matrix, Syst Eng 5(3) (2002), 180–193. aging complex design projects, J Aircraft 36(1) (1999),
Qi Dong and D. Whitney, Designing a requirement driven 266–274.
product development process, Proc 13th Int Conf Des D.M. Sharman and A. Yassine, Characterizing complex prod-
Theory Methodol (DTM 2001), September 9–12, 2001, uct architectures, Syst Eng 7(1) (2004), 35–60.
11–20, Pittsburgh, PA. W.P. Stevens, G.J. Myers, and L.L. Constantine, Structured
Design Structure Matrix website: http://www.dsmweb.org/ design, IBM Syst J 13(2) (1974), 115–139.
D.D. Frey, E. Jahangir, and F. Engelhard, Computing the D.V. Steward, The design structure system, Document
information content of decoupled designs, Res Eng Des 67APE6, General Electric, Schenectady, NY, September
12 (2000), 90–102. 1967.
M.D. Guenov and S. Barker, Requirements-driven design D.V. Steward, The design structure system: A method for
decomposition: A method for exploring complex system managing the design of complex systems, IEEE Trans Eng
architecture, Proc ASME 2004 Int Des Eng Tech Conf Management EM-28, 3 (August 1981), 71–74.
Comput Inform Eng Conf (DETC’04), Salt Lake City, UT, N.P. Suh, The principles of design, Oxford University Press,
September 28–October 2, 2004, to appear. New York, 1990.
M.D. Guenov, Complexity and cost effectiveness measures N.P. Suh, Axiomatic design: Advances and applications, Ox-
for systems design, Manufacturing Complexity Network ford University Press, New York, 2001.
Conf, Downing College, Cambridge, UK, April 9–10, M.T. Talbott, H.L. Burks, R.W. Shaw, M. Amundsen, K.K.
2002, 455–466. Hutchison, and D.D. Strasburg, Method and Apparatus for
M.D. Guenov, Modelling design change propagation in an System Design, U.S. Pat. No. 5,355,317, October 11,
integrated design environment, Computer Model Simula- 1994(a).
tion Eng 1 (1996), 353–367. M.T. Talbott, H.L. Burks, R.W. Shaw, M. Amundsen, K.K.
ISO/IEC 15288 (System Life Cycle Processes), DPC: Hutchison, and D.D. Strasburg, Method and Apparatus for
01/647707, Version 6.0, Bsi, London, 2001. Aiding System Design, U.S. Pat. No. 5,357,440, October
R.J. Lano, The N2 Chart, Informational Report, TRW, 1977, 18, 1994(b).
California. D. Tate, A roadmap for decomposition: Activities, theories,
R.J. Lano, A technique for software and systems design, and tools for system design, Ph.D. Thesis, Department of
North-Holland, Amsterdam, 1979. Mechanical Engineering, Massachusetts Institute of Tech-
K.R. McCord and S.D. Eppinger, Managing the integration nology, Cambridge, MA, February 1999.
problem in concurrent engineering, Working Paper 3594- K.T. Ulrich and S.D. Eppinger, Product design and develop-
93-MSA, MIT Sloan School of Management, Cambridge, ment, 2nd edition, McGraw-Hill Education (ISE Edi-
MA, 1993. tions), New York, 1999.
T.U. Pimmler and S.D. Eppinger, Integration analysis of
product decompositions, ASME Des Theory Methodol
Conf, Minneapolis, MN, September 1994, 343–351.

Marin D. Guenov is a Senior Lecturer (Advanced Engineering Methods) in the School of Engineering,
Department of Power Propulsion and Aerospace Engineering at Cranfield University, UK. He holds an
MEng in Mechanical Engineering and a Ph.D. in Materials Handling Systems and Operational Research.
Currently Dr. Guenov leads national and international multidisciplinary research projects supported by
the European aerospace industry in the areas of design of complex systems, modeling and simulation for
synthetic environments, multidisciplinary design analysis and optimization, and infrastructures for
collaborative design. Dr. Guenov is a member of the Royal Aeronautical Society and The Association of
Cost Engineers and is a Charted Engineer.

Stephen G. Barker is a Research Officer in the School of Engineering, Department of Power Propulsion
and Aerospace Engineering at Cranfield University, UK. Dr. Barker holds a BSc. (Hons.) in Computer
Studies, and researched Applied Engineering Process Management for his Ph.D. Currently, Dr. Barker is
part of the COPE (COmplex Product Environment) research team, which investigates “Decomposition-
Integration” issues within Complex Product Development programs. Dr. Barker has previously worked
as a lecturer and tutor in the fields of Network Communications and Operations Management.