Vous êtes sur la page 1sur 9

IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 27, NO.

2, MARCH 1997 137

The New Taxonomy of Systems: Toward an


Adaptive Systems Engineering Framework
Aaron J. Shenhar, Senior Member, IEEE, and Zeev Bonen, Life Member, IEEE

Abstract— Systems engineering is developing rapidly, while example, many influential project management and systems
new standards are created and new tools are being developed. engineering texts and handbooks focus on the universal set
However, the theoretical understanding and the conceptual foun- of project management activities such as organizing, planning,
dation of systems engineering are still in their early stages. For
example, although real-world systems exhibit considerable differ- controlling, and monitoring [8], [16], and systems engineering
ences, there is very little distinction in the literature between the tasks such as requirements analysis, functional allocation, trade
system type and the description of its actual system engineering studies, and system synthesis [4], [7], [14], [23]. Several
pursuit. We suggest here a new approach to systems engineering. exceptions should be mentioned, however [2], [3], [6], [28],
It is based on the premise that the actual process of systems
[31]. In practice, as most project managers assert [22], there
engineering must be adaptive to the real system type. Using this
concept, we present a two-dimensional (2-D) taxonomy in which are great differences among systems and among the processes
systems are classified according to four levels of technological of their creation. Consider, for example, the case of building
uncertainty, and three levels of system scope. We then describe a new cross-city highway, as compared to the development
the differences found in systems engineering styles in various of a new space vehicle, or the case of making a new model
areas, such as system requirements, functional allocation, systems
design, project organization, and management style. We also
of a small appliance such as a coffee maker, as compared
claim that adapting the wrong system and management style may to the effort of building the English Channel tunnel. Natu-
cause major difficulties during the process of system creation. rally, these efforts are called projects, and they may employ
Two examples will be analyzed to illustrate this point: the famous some system engineering procedures; however, it is clear that
Space Shuttle case and one of the system development projects their differences outweigh their resemblance. Although some
we studied.
distinctions among systems have been mentioned previously,
from a conceptual perspective, the following question still
I. INTRODUCTION remains open: do all systems look alike, and, for that matter,
are all systems engineering processes really the same?
T HE CREATION of complex man-made systems probably
has its historical roots in early civilization. Today, with
the development of management theory, such creations are
This paper is based on our ongoing field study of recent
years, that has so far collected data on more than 250 projects
often linked to modern concepts of systems engineering. As in the United States and Israel. Data was collected in the
a problem-driven field, the practice of systems engineering defense and the commercial sector, and in a variety of in-
is rapidly growing, while more standards are being written dustries such as aerospace, electronics, computer, mechanical,
and additional tools are being developed. However, as a chemical, and construction. The purpose of this study was
theoretical discipline, systems engineering is quite new and to identify the differences among systems, to investigate
probably not well understood. Most research literature on contingencies in the processes of system creation, and to
the systems engineering process is still young and suffers develop a conceptual foundation for the distinction among
from a scanty theoretical basis. Originated in the defense systems and projects. Our basic proposition is that both project
and space industries, attempts are being made to introduce management style and systems engineering practice differ with
the ideas of systems engineering into the commercial and each specific kind of system and that management attitudes
consumer-oriented industries. However, these attempts have must be adapted to the proper system type. Our methodology
had only limited success, since managers and engineers in such was based on a combination of qualitative and quantitative
industries do not see the relevance of defense-based systems methods encompassing the process of building theory from
engineering to their fields. The reason, again, is perhaps the case study research and traditional statistical analysis based
tenuous conceptual foundation of systems engineering. on data accumulated through structured questionnaires.
One of the difficulties in developing a better understanding We suggest a conceptual 2-D taxonomy for the classification
of systems engineering is that little distinction has been made of systems. In our framework, systems are classified according
in the literature between the system type and its strategic to their technological uncertainty at the moment of project
type, or its systems engineering and managerial problems. For initiation and system scope which is their location on a hier-
Manuscript received April 26, 1995; revised March 20, 1996. archical ladder of systems and subsystems. After presenting
A. J. Shenhar is with Stevens Institute of Technology, Hoboken, NJ 07030 the basic model, we will describe some highlights from our
USA (e-mail: ashenhar@stevens-tech.edu). study as they relate to systems engineering. In particular,
Z. Bonen is with Bar-Ilan University, Besa Center for Strategic Studies,
Israel. we will outline the differences that were found in systems
Publisher Item Identifier S 1083-4427(97)00134-3. engineering practices between different system types. We will
1083–4427/97$10.00  1997 IEEE
138 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 27, NO. 2, MARCH 1997

A. The Technological Uncertainty Dimension


1) Type A—Low-Tech Projects: Type A projects are those
projects that rely on existing and well-established technologies
to which all industry players have equal access. Although
such projects may be very large in scale, no new technology
is employed by the firm at any stage. Typical projects in
this category are construction, road building, and other utility
works that are common in the construction industry.
The main characteristic of this kind of project is that no
development is required and usually no system testing is
conducted. The system requirements are usually set by the
customer prior to signing the contract and before the formal
initiation of the project execution phase. System functional
requirements are usually simple and straightforward, and are
Fig. 1. A 2-D taxonomy of engineering projects and systems. often static. The design cycle of such systems is usually
characterized by a single cycle and no design iterations are
then use several famous projects to illustrate the virtue of performed, while the final design of the system is usually
our framework. We will also discuss two cases of system frozen prior to the project initiation. The management and sys-
misclassification. They are the famous Space Shuttle program tems engineering style of low-tech projects can be described
and one of the system development projects in our study. as a firm and formal style, in which one should stick to the
initial design, and no changes are allowed or required.
II. THE 2-D TAXONOMY 2) Type B—Medium-Tech Projects: Type B projects rest
mainly on existing technologies; however, such systems
Innovation literature has often used a traditional distinction incorporate a new technology or a new feature of limited
between radical and incremental innovation [1], [3], [9], [10], scale. Typical projects of this kind include engineering efforts
[12], [32]. Radical innovation is frequently based on new of incremental innovation, as well as improvements and
technology, and it often results in products new to the firm modifications of existing products and systems. They may
and new ventures unrelated to existing businesses. Incremental involve new models in well-established product lines, such as
innovation, in contrast, introduces relatively minor changes automobiles or consumer electronics; they can also include
in existing products, and generally takes the form of product defense projects that involve upgrades of existing systems.
modifications, upgrades, derivatives, and line extensions. As Such projects are found in almost all industries including
our studies revealed, this low–high dichotomy is insufficient mechanical, electronics, aerospace, and chemical.
to capture the full-spectrum of engineering tasks [26]. The Medium-tech projects require some development work and
first dimension is based on levels of technological uncertainty some testing. Their requirements are mainly set in advance;
as perceived by the organization at the time of the project’s however, some changes may be introduced during the prod-
initiation [13]. It includes four different levels, designated as: uct development phase. This process often involves a joint
A—low-tech, B—medium-tech, C—high-tech, and D—super- effort of the contractor and customer. It may also require
high-tech. the involvement of potential customers in the process. The
The second dimension is based on the complexity of the system’s required functions, although still simple, are more
system as expressed by the different hierarchies inside a dynamic than in low-tech systems, and often include more
product. Since products are composed of components and than one mode of operation. The design cycle of such systems
systems of subsystems, hierarchies in systems may include would frequently consist of more than one iteration, and the
many levels [5], [25], [29]. However, our findings exhibit a final system design is usually fixed early, normally during the
predominant split between three clusters of engineering project first quarter of the project. The style of these projects can be
styles. The dimension is called system scope and it includes described as moderately firm; however, it is more flexible than
the following three designated levels: assembly, system, and in type A projects. More communication—both formal and
array (see Fig. 1). informal—is also required, since more tradeoffs and changes
are made.
III. SYSTEM CLASSIFICATION, SYSTEM 3) Type C—High-Tech Projects: Type C projects are de-
CHARACTERISTICS, AND SYSTEMS ENGINEERING STYLES fined as projects in which most of the technologies employed
How could systems be classified according to the 2-D are new, but existent—having been developed prior to the
model, and what are the distinct system characteristics, sys- project’s initiation. Many projects in the high-tech industry
tems engineering, and management styles used for different that involve a new family of products or a new technology
kinds of systems and projects? As found, different concerns generation produce products of this kind. Similarly, most new
characterize the two dimensions. In this section, we define defense development efforts would normally be characterized
the different types of systems and describe the associated as high-tech system projects, incorporating new technologies
managerial and systems engineering characteristics. These that have been recently developed [11]. Typical industries are
characteristics are summarized in Tables I and II. electronics, computers, and aerospace.
SHENHAR AND BONEN: TOWARD AN ADAPTIVE SYSTEMS ENGINEERING FRAMEWORK 139

TABLE I
SYSTEM CHARACTERISTICS BY LEVEL OF TECHNOLOGICAL UNCERTAINTY

Integrating several new technologies for the first time ob- involves intensive interaction and the use of multiple formal
viously leads to a high level of uncertainty and risk. Type C and informal communication channels.
projects therefore require extensive development and testing 4) Type D—Super-High-Tech Projects: Type D projects are
efforts, and building of several prototypes along the way. based primarily on new, not entirely existent, technologies.
System requirements are derived interactively with a strong Some of these technologies are emerging, others are even
involvement of customers or potential users, and many changes unknown at the time of the project’s initiation. The project’s
are introduced. Functional allocation is complex, and functions execution period is, therefore, devoted in part to the devel-
are dynamic and multimodal. The design cycle of these opment of new technologies, testing, and selection among
systems is also iterative, usually entails two or more cycles, alternatives. This type of development project obviously en-
and the system design freeze takes place at a later stage than tails extreme levels of uncertainty and risk; it is relatively rare,
in type B systems. It often occurs as late as the second quarter and is usually carried out by only a few and probably large
or the midpoint of the project. The systems engineering and organizations or government agencies. Typical products are
management style of high-tech systems can be described as the development of a new nonproven concept, or a completely
moderately flexible, since many changes are expected and new family of systems, and they are usually confined to the
are a natural part of this type of system development. It few leading electronics and aerospace industries. The most
140 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 27, NO. 2, MARCH 1997

TABLE II
SYSTEM CHARACTERISTICS BY SCOPE LEVEL

known example of such a project was the Apollo moon-landing is never made before the second or even the third quarter
mission [21]; however, other smaller scale projects can also of the project. To summarize, the management and system
be found in this category. engineering style of these projects can be described as highly
Type D projects require extensive development—both of flexible to accommodate the long periods of uncertainty and
the new technologies and the actual system. Their develop- frequent changes. Managers and engineers must live with
ment frequently requires building an intermediate, small-scale continuous change for a long time; they must extensively
prototype, on which new technologies are tested and approved increase interaction, be concerned with many risk mitigation
before they are installed on the prototype. System requirements activities, and adapt a “look for trouble” mentality.
are hard to determine; they undergo enormous changes and
involve extensive interaction with the customer. Obviously, B. The System Scope Dimension
the system functions are of similar nature—dynamic, complex, 1) Scope 1—An Assembly: An assembly is a collection of
and often ambiguous. A super-high-tech system is never components and modules, combined into a single unit. A
completed before at least two, but very often even four, typical assembly may perform a well-defined function within
design cycles are performed, and the final system design freeze a larger system, hence constituting one of its subsystems;
SHENHAR AND BONEN: TOWARD AN ADAPTIVE SYSTEMS ENGINEERING FRAMEWORK 141

it can also be an independent, self-contained product that creation must be greatly formal and bureaucratic; however,
performs a single function of a limited scale. A radar receiver, some informal relationships always develop at this level be-
a missile’s guidance and control unit, or a computer’s hard- tween the management team and some subcontractors and
disk are common examples of assemblies (subsystems) within customers.
larger systems; compact disc players, coffee makers, washing 3) Scope 3—An Array: An array is a large collection of
machines, and other household appliances can be considered systems functioning together to achieve a common purpose.
independent assemblies of the second kind. Such products Different terminology for an array may be a “supersystem,”
are either functioning as autonomous, or they require only a “system of systems,” or the Greek word “systema,” all
a limited amount of human interaction. being an expression of the array’s nature as a conjunction
The creation of assemblies is usually performed within one or conglomeration of systems. Arrays are usually dispersed
organization, and often under the responsibility of a single over wide geographical areas and normally include systems
functional group. It involves mostly technical people with of many kinds. A nation’s air defense system, consisting of
limited staff support or no staff at all. Systems engineering early warning radar, command and control centers, combat
in assemblies is mostly concerned with design related issues aircraft and ground to air missiles, is a good example of such
such as design for manufacturability, reliability, etc., and much a super system. Similarly, the public transportation network
less with total system organization. The core of the systems of a large city, or the city’s highway system, may also be
engineering activity is the process of concurrent engineering considered typical arrays. Arrays achieve wide-range missions,
and the involvement of crossfunctional teams. Resources plan- based on the simultaneous functioning of many systems, and
ning is relatively simple, with no more than 100 activities in their operation involves the association and interaction of
the network or PERT, and it is often performed manually. many people.
Control is simple as well; it often involves reporting only To manage the effort of building or even improving an
to top management or main contractor, with the majority of array, a program management office is usually established
documents being technical rather than managerial. In essence, as an umbrella organization which coordinates the efforts of
therefore, managing the creation of an assembly is usually various system project organizations. This office is usually
made in an informal way, using a team, or family-type style. staffed with legal, administrative, and finance experts who
2) Scope 2—A System: A system is a complex collection provide the proper control in each discipline. This office
of interactive units and subsystems within a single product, is usually less occupied however, with technical matters;
jointly performing a wide range of independent functions to those are left to the separate projects. It is also much less
meet a specific operational mission or need. A system consists concerned with system integration. Sometimes the leading
of many subsystems (and assemblies), each performing its organization is hiring another system house organization to be
own function and serving the system’s major mission. Radar, the super-system integrator. Many system engineering efforts
computers, missiles, and, for that matter, entire aircraft are are directed toward ensuring coordination among various
typical examples of systems performing independent tasks. system projects. Another major concern of array projects
Systems are capable of performing a complex mission, and involve environmental and government regulations, which are
their use involves considerable man-machine interaction. becoming critical due to the dispersed nature of the final
Organizing a system project usually involves a main con- product. Planning such projects is done through a master
tractor who establishes a project management team and who central plan, which is followed by detailed planning at the
is, in turn, responsible to managing and coordinating many systems level. The total number of activities may reach tens
internal and external subcontractors. The management team of thousands, and project planning tools are often inadequate
includes technical personnel, administrative staff, and sys- for planning purposes of array projects. In fact, array projects
tems engineering people. The systems engineering at this normally develop their own tools for planning, coordinating,
level encompasses the activities at the lower level, but it and control. Most central control in these projects involves
primarily involves most of the traditional aspects, such as managerial and financial issues, and they require extensive
system requirements, functional allocation, system synthesis, amounts of documentation. Finally, the management style of
and tradeoffs. One of the major difficulties in the system level such a program is primarily formal, bureaucratic, and tight,
is the problem of system integration, and that of achieving since the central office is usually remote from the separate
an overall optimized and effective system. The systems engi- projects, and most of their connections are formal and in
neering process at this level also requires a detailed systems written form.
engineering management plan. Planning resources in a system
project is a complex activity. It requires advanced and com-
puterized tools, since the system development network may IV. INTEGRATING THE
include hundreds or even thousands activities. Control—both TWO-DIMENSIONS—INTERACTION EFFECTS
managerial and technical—must be tight and formal, and As discussed above, when moving from lower to higher
extensive technical and administrative documents are used, technology systems while keeping the system unchanged, the
usually according to a formal standard. For example, one of concerns of managers and engineers are shifting toward addi-
the major concerns is configuration control and configuration tional technical issues: more design cycles, more testing, and
management to keep track of all the changes and detailed longer periods of ambiguity in requirements and specifications.
baseline of design. In summary, management style of system Similarly, increased scope with constant uncertainty results
142 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 27, NO. 2, MARCH 1997

in additional concerns for formal administrative issues: more technologies. Other examples of high-tech projects are the
planning, tighter control, more subcontracting, and more bu- development efforts of the first generation of videocassette
reaucracy. When, in addition, one moves simultaneously along recorders introduced by JVC and Sony in the mid-1970’s
both dimensions, from low- and medium-tech assemblies [24]. However, according to our framework, it is an assem-
toward high-tech and super-high-tech systems and arrays, new bly. Lockheed’s SR-71 Blackbird reconnaissance aircraft was
challenges must be addressed, and, as found, there is a strong initiated in the late 1950’s with no clear technology available,
interaction effect between the two dimensions. and almost no known configuration [15]. Retrospectively, we
First, it is at this end of the model that systems engineering may classify it as a super-high-tech system. New York City’s
is practiced at its best. The development of large multidis- transit authority capital program of modernizing the city’s
ciplinary systems, which consist of many subsystems and transit infrastructure would be classified as a low-tech array
components, involve the incorporation of most modern sys- [20]; the English Channel tunnel is a medium-tech array, since
tems engineering tools and processes to optimally harmonize it involves some new challenges and risks of digging such a
an ensemble of subsystems and components. Such tools are long undersea tunnel [18]. Finally, the U.S. strategic defense
almost not present in low-tech or medium-tech system projects. initiative, or as it was often called “Star Wars” [17], is the only
System engineers in high-end projects are a natural part of example we could site of a super-high-tech array development
the management team, and they are responsible for creating program.
the system’s architecture and configuration and incorporating
a multitude of technologies that must cohere in the final VI. CONSEQUENCES OF MISCLASSIFICATION
design. They are also accountable for the system’s reliability, Systems engineering and program management must be
maintainability, supportability, human factors, and economic conducted according to the proper style and be adapted to
feasibility of the entire engineering effort. the system type. Such style should be chosen according to the
In lower-tech systems, assembling, integrating, and testing classification presented above. In that sense, one may see the
all subsystems into one unit is usually not an issue, since above discussion as a description of project and system “ideal
all pieces readily fit together. Integration difficulties become types.” When the proper style is employed, we claim, that the
prominent, however, in the higher uncertainty type system chances for project success are much higher. However, when a
projects. In these projects, the successful production of the wrong style is utilized, or the when the system is misclassified,
separate subunits is one thing, while integrating them into one this may result in substantial difficulties and delays in the
working piece is another. Problems of interfaces, energy dis- process of the system creation. For example, working on a
sipation, electromagnetic interference, and even lack of space, project perceived as type C, when it is really a type D,
require long and tedious processes of assembly, numerous may lead to problems which are beyond the competence
tests, and necessary trade-offs in many of these projects. within the framework of the organization. It may require
Finally, there are problems of configuration and risk man- higher technical skills, additional resources, longer periods of
agement. Special software must be used in high-uncertainty development and testing, and much later design freeze. On
high-scope projects to keep track of all the decisions and the other hand, when a project is actually a type C, and it is
changes, and to identify the potential interactions that would wrongly perceived as type D, many overruns and difficulties
occur with each change. Such projects develop a cumulative may be attributed to technology, while they are simply the
database for their design baseline, which, in some cases, must result of bad administration and weak leadership. To illustrate
be updated on a day-to-day basis. As for risks, although all some of these difficulties, the framework of this work will
projects involve a certain level of risk, higher-scope higher- be used to analyze the style and management of two system
tech projects are more sensitive to the need for sytematic development projects.
risk analysis and risk mitigation. In such projects, a risk
management procedure is usually integrated into the program. A. The Space Shuttle Program
It often involves issues of design control, personnel qualifica- On Jan. 28, 1986, the space shuttle Challenger took off
tions, procurement management, quality control, and budget on its last mission, tragically exploding just 73 seconds into
and schedule control. launch. The tragedy shocked the American nation and se-
riously damaged NASA’s image, which had been built on
V. A RETROSPECTIVE CLASSIFICATION OF FAMOUS PROJECTS
a history of 25 years of continuous success. The shock
Several famous projects can be easily coded according to of the tragedy was followed by months of investigation,
this paper’s classification. For example, IBM’s project of and it has been widely analyzed in many articles and other
building the personal computer in the early 1980’s would writings. Using the model presented in this paper, we suggest
be characterized in this framework as a high-tech system a new way of looking into the Shuttle program. It is based
development project, since it was based on new but existent on the proposition that from its inception, this ambitious
technologies. Similarly, development projects of defense sys- system was designed and managed while employing the wrong
tems such as the Patriot missile are usually based on recent management and systems engineering style. We also claim that
technologies that were developed prior to the main project a proper classification at project initiation would have led to a
effort. Such projects are normally contracted following a well- different systems engineering style and culture, and might have
structured procedure of defense acquisition to contractors that helped eliminate many of the circumstances which eventually
demonstrate a viable capability and sufficient maturity of led to the Challenger tragedy.
SHENHAR AND BONEN: TOWARD AN ADAPTIVE SYSTEMS ENGINEERING FRAMEWORK 143

The space shuttle was designed to enable man to function 5) NASA had no prior experience in reusable thermal
in an extraterrestrial environment by using the same vehicle protection materials.
on repeated flights. The program was first proposed in 1969 6) Never before had a manned space vehicle undergone its
following the successful Apollo moon landing as an element first orbital flight with a live crew.
of an interplanetary transportation system that would enable 7) Finally, the shuttle was to become the first American
the United States to land a crew on Mars in the 1980’s. At space vehicle without a crew escape system.
that time, however, the enthusiasm and space competition that Developing the space shuttle turned out to be one of the
characterized the Apollo era had gone, and the shuttle’s initial most difficult and exasperating engineering challenges of the
high-cost proposals encountered low priorities. space age, if not the most difficult of all. Engineering and
NASA officials discovered that their main hope of winning design problems encountered during the development phase
congressional support for the shuttle in the atmosphere of were enormous, as illustrated by the following two examples.
indifference and skepticism of that time was to emphasize First, there were problems involving the main engines.
its cost effectiveness, namely, its reusability and its low These engines were expected to be fully developed by 1977,
development price-tag. The notion became that this new but between Mar. 1977 and Nov. 1979 there were 14 engine
innovation could be built almost entirely within the state-of- test failures caused by faulty seals, uneven bearing loads,
the-art, largely from off-the-shelf parts which either existed or cracked turbine blades, and many other defects. Eight of
could be readily fabricated. The space agency’s projection of the failures resulted in fires which damaged the engines
the shuttle was simply that of a “space-adapted airplane,” and or the test-stand. NASA’s early commitment to the initial
this was expressed by saying: “We have the technology, we design, based as it was on a success-oriented strategy, made
have the skills, we know how to build it” [19]. Eventually, it imperative to keep going, even when “Murphy’s law”
a moderate, low-cost reusable space shuttle program was took over and anything that could have failed actually did.
approved and its final configuration was selected in 1972. The What was missing in this part of the program was a prior,
shuttle’s first orbital flight was scheduled for 1978 while the much longer experimental phase of separate testing of the
shuttle fleet was to become operational in 1979. new technology of high-pressure hydrogen–oxygen engine
The program involved three major components: the orbiter, components, before bringing the fully assembled engine to
which is powered by three main liquid hydrogen–oxygen the test-stand [19].
rocket engines, the external tank which carries the liquid fuels The second example relates to the thermal protection diffi-
for the orbiter’s main engines and is not reusable, and two culties. The main thermal protection of the orbiter was made of
solid rocket boosters which supply the main thrust at liftoff. special tiles using a new material which had never been tested
It is clear now that NASA’s initial attitude toward the on any previous space vehicle. Its main problem appeared to be
space shuttle project was one which in our notation would the stability of the tiles installed. Each one of the 30 000 tiles
be classified as a type C. The belief, however, according to was different, and it had to be manually installed by skilled
which this invention could be built almost entirely within the craftsmen. After the first vehicle was flown from California to
state-of-the-art, based on existing technologies, turned out to Florida, thousands of tiles were missing and many others failed
be unrealistic. The problem of building such a huge vehicle the pull tests conducted to examine their stability. The thermal
that can function effectively in two different mediums, both in protection turned out to be a nonstable add-on to the orbiter
space and in the atmosphere, was too difficult to accomplish, structure, causing the agency many disruptions and continual
using the “success-oriented” attitude [30]. Eventually, the delays. In retrospect, this shield should have been built into
program suffered overruns amounting to almost three years in the vehicle’s structure, rather than glued in as an add-on. Had
schedule delays and 60% in unexpected costs, all accumulated the agency run earlier flight tests, perhaps on a smaller scale
even before the first flight. reentry vehicle, it might have chosen a different solution to
the thermal protection problem than the one that was initially
The following list demonstrates some of the main risks that
selected [19].
existed at the beginning of the project:
These and many additional problems encountered during
1) It was the first vehicle which was designed to fire hy- the shuttle’s development phase did not change NASA’s initial
drogen–oxygen engines at ground level in the proximity “go ahead” attitude, and in 1982 under the pressure of demands
of an external fuel tank and to be reused again at least for an intensive flight rate, the shuttle’s fleet was declared
50 times. “operational,” a state for which, in many respects, the system
2) Although NASA had already been using solid-fuel ve- was not prepared [30].
hicles for launching small unmanned spacecraft, solid In retrospect, although at the time of the Shuttle’s approval
boosters for manned flights was a technology new to NASA was under severe pressure to cut costs, it seems that
the agency. the program should have been managed using a different
3) Never before had such a large winged “space aircraft” philosophy. In the context of this paper, a type D system
reentered the atmosphere. management scheme was probably more appropriate than the
4) Never before had a 75-ton glider descended through the one actually used. If such a philosophy had been used, then the
entire regime of hypersonic and supersonic speeds to final configuration selection, the freezing point of the design,
make a pinpoint landing 5000 miles from the point of and the operational declaration would have been scheduled for
reentry. a much later point, leaving open the possibilities of using other,
144 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 27, NO. 2, MARCH 1997

more thoroughly tested, and perhaps improved technologies. In goal was to develop a new aiming and control system to
addition, a type D philosophy would have vested the program be installed on an existing military platform. Previous ex-
managers and engineers with a different systems engineering perience of the main contractor with earlier generations of
attitude toward risks, possible development problems, and this kind of system was not sufficiently indicative of the
probability of failure, and would have created the need for level of complexity and difficulty associated with this project.
a much higher level of communication among the various Consequently, the project management team chose a traditional
parties involved. management style. It turned out, however, to be a completely
Returning to the Challenger tragedy, we may note here new experience to the firm, particularly in terms of integrating
that the presidential commission that investigated the accident new technologies for the first time. As it turned out, the project
confirmed that it occurred because of a leak of a vaporized went into a serious crisis and was saved only after significant
material out of the right solid rocket booster. This leak caused a reconstruction and a change of managerial style and attitude,
flame that resulted in an enormous explosion. The commission and only after substantial overruns occurred. We maintain that
concluded that the failure was the result of faulty design of a careful early analysis of the project characteristics could have
an aft field joint, which was unacceptably sensitive to factors led management into a different style right from inception and
such as temperature, physical dimensions, the properties of would have helped to save considerable resources.
the materials, the effects of reusability, etc. The commission Using our framework in retrospect, our analysis shows that
also concluded that the decision to launch the Challenger the project should have been classified as a high-tech type
was wrong. Those who made that decision were unaware of C system level project. The initial management attitude that
the recent history of problems concerning the o-rings on the was employed on this program could be seen, however, as
joint and were unaware of written recommendations of the that of a medium-tech type B assembly project. Most of
contractor against the launch at temperatures below 53 F. these projects are usually improvement or modification efforts
Among its recommendations, the commission suggested entailing only minor advancements in-terms of technology,
performing a new design of the faulty solid rocket motor and they involve building separate units without much regard
joint and seal and forming an independent solid rocket motor to their interaction. The project proved to be much more
design oversight committee. The commission also recom- complicated and it required, in fact, a more careful design
mended reviewing the shuttle program management structure, and an increased level of flexibility in management style as
redefining the program manager’s responsibilities, reviewing required by type C system projects. The difference can clearly
all critical items, performing hazard analyses, establishing be seen when looking at the initially scheduled time for design
an office of safety, reliability and quality assurance, and freeze which was set at a very early stage of the program, and
improving communication among various NASA’s units and was in sharp contrast to the actual, much later, time when the
subcontractors. design was finally frozen.
Looking back on the shuttle’s program after the accident, The second problem emerged during the system integration
one cannot help but ask whether if NASA had adopted a phase. The integration stage was not sufficiently predicted,
different attitude toward the program, this tragic accident and even not planned for in detail until after it had started
would not have occurred? Of course, we may never know. and initially failed. The initial assumption was that this phase
What we may say, however, is that if the agency had initially would be short and simple. The first subsystems and units
employed a managerial form that is closer to a type D style, systems were indeed supplied on time, after they had been
it would have scheduled many of its decisions related to subjected to partial, nondynamic testing. However, when as-
technology and configuration to a later data, and, as mentioned, sembled together onto the final platform, they did not meet
it would have adopted a different and a much more intense many performance requirements. The consequences required
communication pattern inside the program. It would certainly renewed system engineering, additional feasibility studies, and
not have conducted this program by using the success-oriented redevelopment of two of the subsystems. The result was
approach, and would have been more keenly aware of the extensive budget overrun and schedule delay.
possibility of trouble. In conclusion, the main problem lay in misclassification of
One must note however that when the space shuttle program the project’s initial uncertainty and misevaluation of its scope.
was approved and its budget set, NASA was under extreme In the developer’s initial estimation, the system was nothing
pressure from the administration to prove the system’s cost- more than an add-on improvement to a system of a previ-
effectiveness and to considerably cut development expenses. ous generation. Similarly, the complex link between various
Under such circumstances, NASA abandoned its initial plan system components, and between the system and platform,
to build a small-scale test vehicle to make certain it had was almost completely ignored. The failure to predict the
all the technology in hand before embarking on a full-scale technical and integration difficulties resulted in a managerial
vehicle. But these circumstances probably also nurtured in the and technical crisis to which only extensive top management
agency the success-oriented type C attitude that prevailed in support helped the project to become a success after all.
the program since its inception.
VII. CONCLUSION
B. A New Military System Development Project Systems engineering is a wide-range activity, and it should
The second case is a research and development project not be handled in the same form for all kinds of systems.
we investigated during the course of our studies [27]. The Above all, system engineering requires a proper attitude,
SHENHAR AND BONEN: TOWARD AN ADAPTIVE SYSTEMS ENGINEERING FRAMEWORK 145

concept, and philosophy. To design and manage systems [20] S. J. Manne, and L. Collins, “Reconstructuring an aging infrastructure,”
effectively, one must choose the appropriate concept and adopt Project Manage. J., Apr. 9–24, 1990.
[21] C. R. Pellegrino and J. Stoff, Chariots for Apollo. New York:
the right attitude. Only then can one choose the specific tools Athenaeum, 1985.
and practices for executing the project. In this paper, we have [22] J. K. Pinto and J. G. Covin, “Critical factors in project implementation:
A comparison of construction and R&D projects,” Technovation, vol.
suggested classifying systems into four classes according to 9, pp. 49–62, 1989.
their level of technological uncertainty and into three classes [23] E. Rechtin, Systems Architecting. Englewood Cliffs, NJ: Prentice-Hall,
of system scope. Each class requires a different concept and 1991.
[24] R. S. Rosenbloom and M. A. Cusumano, “Technological pioneering and
a distinctive philosophy for effective systems engineering and competitive advantage: The birth of the VCR industry,” Calif. Manage.
management. Rev., vol. 24, no. 4, pp. 51–73, 1987.
The framework given here is not conclusive; it could [25] A. J. Shenhar, “On system properties and systemhood,” Int. J. General
Syst., vol. 18, no. 2, pp. 167–174, 1991.
be further refined and investigated. For example, one may [26] “From low to high-tech project management,” R&D Manage.,
consider additional uncertainties for the analysis of system vol. 23, no. 3, pp. 199–214, 1993.
[27] A. J. Shenhar and S. Alkaher, “Meeting the high technology system inte-
management—market uncertainties, for example. In this case gration challenge: The BAT project case history,” working paper, Center
the uncertainty dimension may represent a different kind of for the Development of Technological Leadership, Univ. Minnesota,
uncertainty and would require different skills. There is also Minneapolis, 1994.
[28] L. W. Steele, Innovation in Big Business. New York: Elsevier, 1975.
a need for more work in order to develop better and more [29] J. P. Van Gigch, Applied General Systems Theory, 2nd ed. New York:
advanced tools for system engineering in specific classes of Harper and Row, 1978.
systems. Researchers in the field of engineering and manage- [30] F. Webster, “The space shuttle Challenger incident: Showcase project,”
Project Manage. J., vol. 18, pp. 41–68, June 1987.
ment science, as well as practitioners and consultants, may [31] S. C. Wheelwright and K. B. Clark, Revolutionizing Product Develop-
find the framework of this work useful for seeking additional ment. New York: Free Press, 1992
[32] G. Zaltman, R. L. Duncan, and J. Holbek, Innovations and Organiza-
insights into the managerial differences among systems, and tions. New York: Wiley, 1973.
for finding better and more effective ways to manage the
creation of different kinds of systems.

REFERENCES Aaron J. Shenhar (S’74–M’76–SM’86) received


five academic degrees in electrical engineering, sta-
[1] W. J. Abernathy and J. M. Utterback, “Patterns of industrial innovation,”
tistics, and engineering economic systems from the
Technol. Rev., pp. 40–47, June/July 1978.
Technion–Israel Institute of Technology, Haifa, and
[2] N. Ahituv and S. Neumann, “A flexible approach to information system
Stanford University, Stanford, CA.
development,” MIS Quart., pp. 69–78, June 1984.
[3] S. B. Blake, Managing for Responsive Research and Development. San He is now Institute Professor of Management
Francisco, CA: Freeman, 1978. at Stevens Institute of Technology, Hoboken, NJ.
[4] B. S. Blanchard and W. J. Fabricky, System Engineering and Analysis, He has 20 years of business and management ex-
2nd ed. Englewood Cliffs, NJ: Prentice-Hall, 1990. perience in a high-technology environment. In the
[5] K. Boulding, “General systems theory: The skeleton of science,” Man- business world, he has managed the development
age. Sci., vol. 2, pp. 197–208, Apr. 1956. of several high-technology projects involving radar
[6] J. I. Cash, Jr., W. F. McFarlan, and J. L. McKenney, Corporate and computer technology. As executive at Rafael, the Armament Development
Information Systems Management. Homewood, IL: Irwin, 1988. Authority of Israel, he was Corporate Vice President and President, Electronic
[7] W. P. Chase, Management of Systems Engineering. Malabar, FL: Systems Division. Since beginning a second career in academia, he has been
Krieger, 1985. involved in teaching and research in the areas of technology and innovation
[8] D. I. Cleland and W. R. King, Systems Analysis and Project Manage- management, project and product development, systems engineering, and
ment. New York: McGraw-Hill, 1983. the management of professionals in high-technology organizations. He has
[9] R. D. Dewar and J. E. Dutton, “The adoption of radical and incre- published three books and more than 40 refereed articles. He is a recognized
mental innovations: An empirical analysis,” Manage. Sci., vol. 32, pp. consultant and well known speaker in technology-based organizations such as
1422–1433, 1986. 3M, Honeywell, Loral, the Trane Company, and Israel Aircraft Industry.
[10] J. E. Ettlie, W. P. Bridges, and R. D. O’Keffe, “Organizational strat-
egy and structural differences for radical vs. incremental innovation,”
Manage. Sci., vol. 30, pp. 682–695, 1984.
[11] R. J. Fox, The Defense Management Challenge: Weapons Acquisition.
Boston, MA: Harvard Business School Press, 1988.
Zeev Bonen (A’51–M’56–LM’92) was born in
[12] C. Freeman, The Economics of Industrial Innovation, 2nd ed. Cam-
Jerusalem, Israel, in 1926. He received the Electrical
bridge, MA: MIT Press, 1982.
[13] J. R. Galbraith, Organization Design. Reading, MA: Addison-Wesley, Engineer degree from the Technion–Israel Institute
1977. of Technology, Haifa, in 1951, and the Ph.D. degree
[14] J. W. Hunger, Engineering the System Solution. Englewood Cliffs, NJ: in nonlinear control from Cambridge University,
Prentice-Hall, 1995. Cambridge, U.K., in 1961.
[15] C. L. Johnson and M. Smith, Kelly: More Than My Share of It All. From 1952 to 1979 and 1981 to 1989, he
Washington, DC: Smithsonian, 1985. was with Rafael, the Armament Development
[16] H. Kerzner, Project Management: A Systems Approach to Planning, Authority, Israeli Ministry of Defense. He served
Scheduling, and Controlling, 5th ed. New York: Van Nostrand Rein- as President of Rafael from 1970 to 1978 and 1982
hold, 1994. to 1987. He has published many papers on weapons
[17] R. M. Lawrence, Strategic Defense Initiative. Boulder, CO: Westview, development and on the evolution of battlefield systems. Since his retirement
1987. in 1989 he has engaged in various consulting activities. Since 1994, he has
[18] J. K. Lemley, “The channel tunnel: Creating a modern wonder of the been Senior Research Fellow, Bar Ilan University Besa Center for Strategic
world,” PmNetwork, July 14–22, 1992. Studies, Israel.
[19] R. S. Lewis, The Voyage of Columbia. The First True Spaceship. New In 1970, Dr. Bonen received the Israel Defense Prize for the development
York: Columbia Univ. Press, 1984. of the Shafrir air-to-air missile.

Vous aimerez peut-être aussi