G Model

CIRP-1406; No. of Pages 23

CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

CIRP Annals - Manufacturing Technology

jou rnal homep age : ht t p: // ees .e lse vi er . com /ci r p/ def a ult . asp

Automating design with intelligent human–machine integration

Yue H. Yin (2)a,*, Andrew Y.C. Nee (1)b, S.K. Ong (1)b, Jian Y. Zhu (1)c, Pei H. Gu (1)d,
Lien J. Chen (2)e
a
Robotics Institute, Shanghai Jiao Tong University, Shanghai, China
b
Mechanical Engineering, National University of Singapore, Singapore, Singapore
c
Nanjing University of Aeronautics and Astronautics, Nanjing, China
d
Shantou University, Shantou, China
e
Pratt and Whitney, Hartford, CT, USA

A R T I C L E I N F O A B S T R A C T

Keywords: This paper reviews the state-of-the-art methodologies for automating design with intelligent human–
Design automation
machine integration from the perspectives of ontology and epistemology. The human–machine
Human–machine integration
integrated automating design paradigm is reviewed systematically based on a proposed prototype of
Intelligent design
Imaginal thinking human–machine integrated design, from the aspects of ontology-based knowledge management with
Ontology local-to-global ontology transitions, and epistemology-based upward-spiral cognitive process of coupled
design ideation. Particularly, imaginal thinking frame is proposed as the foundation of intelligent
human–machine interaction that puts human and machine on an equal platform. Further, this paper
presents implementations and applications of the automating design paradigm and concludes with the
identification of future trend.
ß 2015 CIRP.

1. Introduction comparisons in literature [103,117,177,205,206], they all share

1.1. Challenges
In the last century, the manufacturing industry has faced major
challenges posed by the shifting from mass production to mass
customization and personalization. Such challenges require the
companies to be able to react to unpredictable, rapid and
fluctuating market changes in a responsive and cost-effective
way [205]. These dynamical market changes are driven by fierce
global economic competition, better educated and more demand-
ing customers, and a rapid advancement in processing technolo-
gies, which reflect a new balance among economy, technology and
society [206]. Meanwhile, the past few decades have witnessed
the evolution of manufacturing paradigm from Dedicated
Manufacturing System (DMS) to Flexible Manufacturing System
(FMS) [35]. Due to the high initial investment of FMS, other
advanced manufacturing systems, such as Computer Integrated
Manufacturing System (CIMS), Agile Manufacturing System
(AMS), Reconfigurable Manufacturing System (RMS) [206] to-
gether with other Intelligent Manufacturing System (IMS) like
Fractal, Holonic and Multi-Agent Manufacturing Systems have
been developed. Although these manufacturing systems are
explicitly different from each other according to the detailed
* Corresponding author. Tel.: +86 2134206310.
E-mail address: yhyin@sjtu.edu.cn (Y.H. Yin).
three basic characteristics, namely intelligentization, modular-
ization and reconfigurability. Thus, nowadays companies not only
should be flexible to process a variety of existing products, but also
be scalable and reconfigurable to deal with new products in order
to survive in the global competition. Major progress of product
manufacturing has been achieved mainly through the develop-
ment and application of advanced manufacturing technologies
and tools [2,38,97,113,115,140,186,199].
While the advanced manufacturing systems and technologies
mentioned above have provided some cost-effective solutions to
mass customization and rapid responses to dynamic market
changes, they are still inadequate in handling all the challenges,
partly because of the fact that design science has come to a
bottleneck and researchers often neglect the time needed for
design of both new products and its manufacturing system. As
shown in Fig. 1, in order to cope with the short window of
opportunity for introducing new products, advanced manufactur-
ing paradigms, technologies and processes mentioned above have
dramatically helped to reduce the manufacturing system lead-
time of the products during the last few decades, including the
time for designing, building or reconfiguring the manufacturing
system, as well as that for ramping up to full-volume and high
quality production. However, the methodologies for designing
both products and manufacturing systems are far from maturity,
leading to the remaining long design time. While extensive
attention has been paid to advanced manufacturing technologies,
many other factors, including rather detrimental ones that could

http://dx.doi.org/10.1016/j.cirp.2015.05.008
0007-8506/ß 2015 CIRP.

Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
G Model
CIRP-1406; No. of Pages 23
2 Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
Fig. 1. The relationship between design and manufacturing.
affect the holistic product cycle are often neglected. It is
conceivable that advanced manufacturing technologies or
manufacturing systems alone can hardly address all the chal-
lenges. Design, which is usually regarded as the starting point or
fountainhead of product manufacturing, determines what the
product should be and where it is to be used and is the vital factor
one should emphasize on. Unfortunately, although design theory
and methodology have been quite well developed in the last few
decades, they are still lagging behind the development of advanced
manufacturing technologies and systems. In consequence, there
are few or no corresponding design theories or methodologies
(DTM) for product development and manufacturing processes.
Without appropriate DTMs guiding how and where to use these
modern technologies, the driving effect of advanced manufactur-
ing systems will be largely limited. Therefore, despite of the
manufacturing lead-time which has been shortened due to
technology development, the design time of both new products
and its manufacturing system has now become a dominant
limiting factor. The application of advanced design technologies
and tools in the early product design phase and the design of its
manufacturing system are the situations where the most impor-
tant decisions are made with respect to product functionality,
quality, manufacturability, cost, environmental performance and
life cycle for optimum design and development of products and
manufacturing systems.
Furthermore, as the core of a manufacturing system, handmade
and manually operated machines are being replaced by CNC, or
automated equipment such as autonomous machine tools and
robots. Therefore, the intelligentization of advanced manufactur-
ing paradigms and machines has become an important trend,
which necessitates artificial intelligence (AI)-based methodologies
to realize the automation of the design paradigm.
1.2. Intelligent human–machine integration
In some sense, automating design using a computer can be
regarded as a mimetic behaviour of human design activities, as
computer systems are expected to independently finish the entire
design process. Up to the present, despite numerous studies using
AI for design to enable the computers to be equally effective as
human decisions, few known systematic methods or scientific
basis for automating design have been developed. With the rapid
development of technologies like FEA for computer-aided design
(CAD), computers, endowed with fast and massive computational
and logical deduction abilities, are becoming increasingly more
powerful for simple design tasks requiring fast and cost-effective
performance. However, due to the intrinsic deficiency of compu-
ters in decision-making and evaluation, human participation
remains a necessity for many complex tasks involving qualitative
judgements and uncertainties. In contrast, humans tend to practise
intuitive thinking based on their rich experience and knowledge,
while compared with computers, humans are much weaker and
slower at logical deduction. Therefore, human–machine integra-
tion is a promising approach for automating design, combining the
quantitative computation and qualitative analytical advantages of
both while compensating the weaknesses of each other.
While it is well recognized that human–machine integration
should surpass the power of either human or machine intelligence
alone, advances in harmonizing the intelligent interaction between
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
human and machine have been slow, partly because of the gap
between human’s customary intuitive thinking and computer’s
logical thinking under the traditional AI framework. Recently, the
imaginal thinking method [207,209] has attempted to narrow this
gap. This method considers imaginal thinking, the synonym of
image-based thinking, to be the fundamental mode of human
thinking. It is defined as the forming of images and the holistic and
direct comparisons among them. Moreover, imaginal thinking is
supported by experimental psychologists [19] who have shown
that people actually use images, not descriptions as computers do,
to understand and respond to the outside world. Brooks’ [19,20]
achievements in action-based AI theory also show indirect
evidence of human imaginal thinking, which is different in
approach but equally satisfactory in result. Thus, imaginal thinking
is helpful to promoting automating design with intelligent
human–machine interaction, because design ideation can be
considered as the generation and transformation of generalized
images.
On the other hand, research [207,209] has proven the feasibility
of computer’s behavioural intelligence in mimicking human
imaginal thinking. The emergence and rapid development of 3D
modelling in CAD have seen the increasing ability of computers in
processing images. Moreover, virtual reality (VR) [3,112] and
augmented reality (AR) technologies [121] have been exploited as
powerful human–machine interaction tools by creating efficient
interactive environments. Although technology developments
have laid a firm foundation, the current AI level is still the most
influential limiting factor to the implementation of automating
design paradigm. Therefore, for situations where computer
systems alone are inadequate to complete the design tasks, the
guidelines based on imaginal thinking can help to build up an
automating design paradigm with intelligent human–machine
integration technologies, such as 3D images under a VR or AR
platform.
1.3. Automating design based on ontology and epistemology
It has been well recognized that design is essentially an iterative
process based on human experience, knowledge and creativity,
indicating that human’s intuitive creativity is vital for coming up
with a brand-new design solution, while AI and existing design
theories fall short to generate a general framework. Due to the fact
that the nature of human intelligence has not been fully discovered
and confusion still exists between the intelligence of human design
and the intelligent nature of design itself, a clear definition of DTM
has not yet been formulated. This further aggravates the relation-
ships among individual theories and methodologies such that
designers may feel difficult to choose a correct method to conduct
design. Moreover, the application of AI becomes essential as
required by the ultimate goal of automating design. This puts
forward higher requirements on the understanding of the inherent
relation between the intelligence of human design and the
intelligent nature of design, and the logical mapping to AI is even
more profound. Thus intractable problems arise when complex
connections among the intelligent nature of design, human
intelligence and AI need to be resolved.
Examining these problems from a higher perspective with
broader and deeper insights helps greatly in eliminating the
confusions. Philosophical views of design [86], which are concerned
with the explanation of the fundamental nature of existence and the
relationship between human and external knowledge, have been
dedicated to building the ultimate guideline for the understanding
of design. Actually all the design activities performed by human also
act as a significant means of obtaining knowledge. Two topics,
namely ontology and epistemology have been leading philosophical
research. Ontology is defined as the science of things that exist,
including the nature, the properties, the characters and the relations
of the objects in reality [69,70]. Epistemology, on the other hand, is
concerned with the methodology for achieving and correctly
understanding the knowledge. In the realm of design science, it
G Model
CIRP-1406; No. of Pages 23
Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
Fig. 2. The upward spiral characteristics of automating design.
specifically addresses the question of what is the relation between
the object of knowledge, and elaborates the influence of human on
the process of achieving knowledge. As shown in Fig. 2, the
development of automating design paradigm has actually been
undergoing an upward spiral process, which is different from the
well accepted iterative process viewed from traditional DTMs, but
quite similar to the development of philosophical research. The
ontological understanding of automating design also goes
through an evolution process. During the modification and
exchanges of each ontology, the understanding towards it keeps
evolving, and its connotation also keeps refreshing itself. With
the promotion of this quantitative-to-qualitative change, an
epistemological jump occurs eventually, resulting in a new
version of the ontology. Comparing to the epistemological
evolutions, the phase before the birth of a new ontology can be
considered a quasi-static or homeostatic process. With the help of
epistemology, the ontologies involved in automating design
becomes increasingly organized and structured: firstly formed
are the local ontologies, which will then be gradually integrated
into a global one by the collaboration and interaction of human
and machine. Similarly, the optimization in design must also
experience such local-to-global and uncertain-to-certain onto-
logical transitions. As can be seen from Fig. 2, ontology and
epistemology are actually inherently related to each other, i.e.,
ontology is the cognitive result of epistemology, while the
embodiment of epistemology must be based on ontology.
Moreover, the development of automating design has made
evident the importance of human–machine integration, as human
and machine are both the central intelligent agents that promote
the ontological evolution.
Therefore, based on the philosophical views of ontology and
epistemology, it is promising to solve the confusion challenging
the automating design paradigm by clarifying the relations among
these elements and providing deeper insights into them. The
superiority of philosophical views for the design process is
strongly supported by Horvath’s research when he gave his
classification of design theories and considered design philosophy
as an independent category [86]. He argued that design
philosophy is the highest level of speculative thinking of the
historical evolution of the design discipline and the fundamental
basis of design thinking. Numerous contributions have been made
since. Gero [59], for instance, takes an ontological view when he
built the function–behaviour–structure (FBS) design prototype
and thus clarified many design concepts and their relations. Their
work can be regarded as solid evidence for the effectiveness of
epistemology and ontology, from both the theoretical and
practical perspectives.
The motivation of this paper is to present a holistic automating
design paradigm with intelligent human–machine integration,
which aims at revealing the nature of the design process and
relieving human from design burden. Ontology and epistemology
are employed to clarify the confusions and complex mapping
relationships among the intelligent nature of design, human
intelligence and AI so that a firm theoretical foundation for
automating design paradigm could be established.
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
3
The remaining sections of the paper are organized as follows:
Section 2 provides an overview of the brief history of DTM including
several widely applied intelligent design theories which can be
regarded as the cornerstones in DTM development. Section 3
provides the basic definitions of ontology and epistemology, and
further discusses ontology-based design modelling from the
knowledge management perspective and ontology integration.
The intelligent human–machine interaction paradigm along with
the discussions on traditional AI frames and the imaginal thinking
frame, as well as the relevant support tools are also reviewed in
detail. Moreover, the epistemologies and the notion of coupled
design and coupled optimization in automating design are further
discussed. Section 4 demonstrates the implementation of the
automating design paradigm in the design of a large ultra-precision
grinding machine. Sections 5 and 6 contain the concluding remarks
and the prospection of future trends respectively.
2. Brief history of automating design
In the past 160 years, perhaps since Redtenbacher, Reuleaux,
Bach, Riedler and others who attempted to understand the design
process, a variety of DTMs have been developed and proposed.
Moreover, a recognized classic view is that a design theory is the
modelling and understanding of design, while design methodolo-
gies are about how to design or how a design should be [178]. This
period has also witnessed the development of automating design,
on which extensive studies have been conducted. The upward
spiral development of automating design shown in Fig. 2 can be
roughly divided into four stages, with each stage supported by
significant achievements that continually promoted the cognitive
level. Meanwhile, the progress of various aspects has contributed
to the spiral understanding of automating the design paradigm. In
particular, two dominating issues stand out: (1) Understanding of
the intelligent nature of design, which will continue to take time to
achieve, and the understanding needs to stay up-to-date. (2) The
technical level, which directly determines the implementation
methods and the thinking frame.
The evolvement of automating design is schematically shown in
Fig. 3, which emphasizes the changing inferential burden during
the design process of human and machine where the abscissa is the
time axis and the ordinate represents the design burden. The time
domain is roughly divided into three periods according to the
different developing stages of automating design: 1850–1980,
1980–1995, 1995–present. The inferential burden is composed of
two types, i.e., logical thinking and intuitive thinking, which are the
two distinct well-recognized thinking frames. The height of the
rectangle represents the entire design task, while the purple line
represents the design load performed by humans and the blue line
represents that performed by machines. Until the 1970s, humans
did nearly all the design tasks reflected by the purple line that stays
at the same height, indicating humans were performing all the
design tasks while the blue line stays almost on the ground
indicating machines had almost no role in the design process. The
unbalanced task allocation is due to the lack of computing and
computer-aided drafting facilities at that time. The situation had
been changed in the 1980s with the emerging and rapid
development of CAD and other analysis technologies, and the
changing trend is evident in Fig. 3. During the 1970s and 1980s, the
purple line starts to decline at the same rate as that of increase of
Fig. 3. Brief history of automating design.
G Model
CIRP-1406; No. of Pages 23
4 Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
the blue line, indicating that the design inferential burden was
gradually taken over by the computers. Revolutionary attempts
were made continuously and have given rise to the great forward
leap in the design industry, indicated by the rapidly rising blue line
in the 1970–1985 period because of the advent of computer
technology. The future of automating design seems to be
promising. However, people soon realized the innate deficiency
of machine intelligence, as mentioned in the previous sections.
When it went to the 1990s, industrial systems became more
complex and a large amount of information had to be handled in
the conceptual design phase. Either computer or human alone was
not able to handle such complex design tasks. As a result,
automating design was stagnant around 1995, as represented by
the flattening of the blue line in Fig. 3, whereas the design
community was lingering between hope and disappointment. It
was not long before researchers realize the possibility of human–
machine integration which would be able to achieve the
complementary advantages of both the intuitive thinking of
human and logical deduction of computer. The development of
automating design has made another stride in the 2000s. However,
there is obviously a limit to the machine’s inferential power due to
the intrinsic logical thinking frame, which is incongruous with
humans’ intuitive thinking frame, resulting in a wide gap between
human and machine. The aforementioned imaginal thinking
method has presented a hopeful sign of unifying the two
significant thinking frames, holding the promise to fully automate
the design process. The three stages of automating design are
discussed in detail below.
2.1. The foundation of automating the design process during the
period of 1850–1980
Initial efforts were mainly focused on the insights into the
intelligent nature of human’s design behaviour itself, leading to a
consensus on what design is and how to model the holistic design
process [8,114,139,181,202]. Pioneering works were done by some
German scholars. Redtenbacher [141] pointed out several char-
acteristics and principles, such as sufficient strength, sufficient
stiffness, low wear and friction, minimum use of materials, easy
handling, easy assembly, and maximum rationalization. Reuleaux
[145–147], the student of Redtenbacher, was one of the earliest to
think highly of human intelligence in design activities. He pointed
out that design evaluation should be done by the intelligence and
discretion of designers. In the 1920s, regarding the balancing of
conflicting demands, Erkens proposed a step-by-step approach
based on constant texting and evaluation [52,189–191]. He
suggested that this process must be continued until a network of
ideas, or the design emerges. After that, some seminal theories laid
the foundation for the development of automating design. Franke
created a comprehensive structure for transmission systems using
a logical-function analogy based on the elements with different
physical effects [55]. Hansen highlighted the significant influenc-
ing factors in the design process and classified them by defining a
system as a clearly delimited part of reality which is related to its
environment (U), structure (S) and function (F) [76]. Furthermore,
he realized that the core issue of engineering design is the analytic
and synthetic relations between the sets of functions and
structures [77].
Based on the above explorations, some basic concepts were
clarified and an initial consensus was reached about the intelligent
nature of automating design. Noteworthy is that Pahl and Beitz
[131] created a design framework including four stages, namely,
planning and clarification, conceptual design, embodiment design,
and detail design, and this frame has been widely adopted by now.
It not only provides preliminary theoretical guide for the
optimization processes in automating design, but also recognizes
the iterative nature of design. Furthermore, the frame elucidates
the fact that ‘‘conceptual design’’, starting from requirements to
configuration, is the core that mostly reflects the creative and
intelligent essence of design.
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
2.2. Attempts to fully automate the design process from 1980 to 1995
In earlier years, design theory related research was mostly
carried out in German-speaking countries. The situation has
changed since the 1980s when this field evoked a much broader
interest, especially in the United States. Meanwhile, with the
emerging and rapid development of CAD, AI and other analytic
technologies, the power of computers has been well-acknowl-
edged. As a result, many new methods have been introduced by AI
researchers interested in design. This gave rise to the scientific
formulation of knowledge, and the power of mathematics began to
take effect in design [177]. Nowadays, many simple tasks like
repetitive computing and drafting are assigned to computers,
resulting in the concept of automating design paradigm which put
machines at the centre of design activities with less human
intervention, and it is becoming a significant trend of design
research.
From then on, numerous studies have been dedicated to the
understanding of the contents of automating design. Some
scholars concentrated on the application of CAD in solving specific
problems with an algorithmic approach. Shirai et al. introduced a
compact and practical CAD/CAM system for progressive dies
[159]. Feng et al. described the basic principles and methods for
intelligent CAD of shafts, sleeves and cylindrical gears using
desktop computers [53]. Others focused on better solutions for
CAD/CAM integration and the building of intelligent CAD systems.
Melkanoff et al. discussed the capabilities and implementation of
an ‘‘intelligent’’ CAD system [116]. The system provides direct
feedback from manufacturing experience to the CAD process by
guiding the user to implement a design with the consideration of
manufacturing processes, economic manufacturing, and automat-
ed manufacturing. Gal-Tzur investigated an interactive hierarchi-
cal CAD/CAM system for the integrated design and manufacturing
of cams [56]. Kimura and Suzuki studied the representation and
manipulation of designers’ intent, which is highly effective to
support efficient product designs in CAD systems [93].
2.2.1. Thinking design machine based on axiomatic design
Aiming at solving the problems involved in the conceptual
design stage, Suh proposed axiomatic design, which is based on
two design axioms: (1) The Independence Axiom: maintaining the
independence between functions and requirements; (2) The
Information Axiom: minimizing the information contents of the
design. This theory attempts to create fundamental principles and
methodologies with the purpose of providing a firm scientific basis
for design [168–171].
In Suh’s opinion, although many CAD software and programs
have been developed, they only deal with the later stages of the
design process, such as geometric modelling and drawing [79,176],
whereas the actual design solution depends highly on the
designer’s individual knowledge and creativity. He is the first to
develop a ‘‘thinking design machine’’ (TDM) which can automati-
cally generate designs or design concepts. Based on Axiomatic
Design, the mapping result of TDM from the function domain to the
physical domain is not unique [45,58,105,117,172,173]. However,
although an infinite number of plausible design solutions may be
generated, not all of them are appropriate. Therefore in TDM, these
solutions are evaluated based on the design axioms.
2.2.2. Intelligent CAD based on general design theory (GDT)
GDT scientifically pursues a better understanding of design by
further aiming at two engineering goals. One is to serve as a
guiding principle for building future (intelligent) CAD systems. The
other is to provide the designer with a practical and efficient design
methodology. Technically, GDT is not purely a design theory
because it defines design knowledge mathematically and makes it
operable as mathematical sets. In this sense, GDT is a theory about
design knowledge and more so general knowledge [6,212,213].
Yoshikawa proposed that an intelligent design system must
perform the following tasks: process planning, tool selection,
G Model
CIRP-1406; No. of Pages 23
Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
fixtures and machine tools, determination of cutting conditions,
etc. Therefore, the systems must be capable of recognizing and
understanding the results of the design in order to translate the
geometrical data into those instructions. Design is a complicated
intellectual process consisting of various kinds of decisions such as
understanding the design specification, selection or creation of
components, determination of mapping methods between them,
and supplement of detailed shapes and dimensions, etc. All these
decisions should be made by analysing the functional require-
ments of a design object, which can be treated as a machine which
assembly of the components depends on the designer. Accordingly,
a CAD system assisting the design process is expected to be capable
of either making decisions for the designer or providing a good
environment that facilitates decision making. A CAD system with
such capabilities is referred to as an intelligent CAD system
[142,211].
TDM and intelligent CAD have formed the theoretical basis for
automating design fully. These two theories have been achieved
through different approaches but are equally effective in applying
machine intelligence to the stage of conceptual design, as TDM
provides a systematic frame of conceptual design for machine
processing, and intelligent CAD clarifies the human ability of
designing in a scientific way, meanwhile producing the practical
knowledge about design methodology. In spite of the deep
insights, these theories remain hard to put into practice, and this
leads to the notion of introducing ontology and epistemology into
design.
2.2.3. FBS prototype based on ontology
It is not until the last two decades that researchers started to
apply ontology [30,43,46,149,155] to automating design research,
and it was around the time when AI technology began to boom.
One representative theory is the FBS design prototype.
The basis for Gero’s FBS framework is formed by three classes of
variables describing different aspects of an object under design:
namely function (F), behaviour (B), and structure (S). A design
description (D) is figured out when a designer constructs
connections among the function, behaviour and structure of an
object through his knowledge and experience.
Moreover, Gero found that in most cases, designers begin with
incomplete information, and this has led to the question of how to
start a design before all the relevant information is available. To
formulate an answer, Gero proposed his design prototype, which is
a conceptual schema for representing a class of generalized
heterogeneous grouping of elements derived from resembling
design cases that provide the basis for the start and continuation of
the design, through integrating all the requisite knowledge
appropriate for the situation into one unified schema [59].
FBS prototype has made large influence on automating design
from two perspectives, i.e., it elucidates the iterative nature of
design behaviour, and proposes a feasible framework for the
integration of human knowledge, experience and creativity. As a
ground breaking effort, the FBS prototype has been inspiring more
systematic and global design frameworks that compensate its
relatively local descriptions of design. Based on the successful
modelling of FBS, further studies have been implemented. The
theory of RFBS [34] emphasizes on the determination of
requirement, which acts as the performance criteria of the product
or the expected behaviour of the system. Situated FBS [60]
proposes that the designer’s concepts, as a function of what he has
done, may change according to what he is ‘seeing’, i.e., the
interactions between the designer and the environment strongly
influence the course of design. To inspire new and creative design
ideas, cross-domain knowledge is required. By analysing the
function representations in different design domains, graphic
design suggests that the function representation can be general-
ized into a small number of categories [138]. This formalism is
employed in an explorative model of analogy-based design.
In addition, although not categorized into a specific knowledge
system, significant progresses based on intelligent computer
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
5
algorithms are abundant, such as modular design [73], adaptable
design [71], TRIZ [27,75,192], Pugh’s total design [136] and
concurrent engineering [117,163]. All these revolutionary
attempts have given rise to the great-leap development in the
design world. However, the innate disadvantages of either human
or machine have severely limited the progress in automating
design. In the 1990s, industrial systems have become more
complex and conceptual design has been challenged by massive
information. Meanwhile, modern technologies are increasingly
relying on computer’s strong processing power to perform
complex computation and simulation. Consequently, there is an
urgent need to find a way of complementing the intelligence of
human and machine.
2.3. Automating design with intelligent human–machine integration
from 1995 to present
The irreplaceable role of human creativity and inspiration has
been well recognized and numerous studies have been dedicated
to its rational explanation in a general framework. In Gero’s
perspective [59], during creative design which extends or moves
the state space of potential designs, the role of context and the
designer’s perception of it play an increasingly important part
[59]. Peirce pioneered in formalizing the abduction process
[21,78], which has been regarded as a general formalization of
intuitive creativity with logic [21,78]. Shah [157] proposed two
ways of generating creative designs, either intuitive or logical. The
intuitive approaches are capable of increasing the flow of ideas,
removing mental blocks, and increasing the chances of conditions
perceived to be promoters of creativity. Although it is manifested
that mimicking the designers’ intuitive thinking would be a great
help to automating design, the wide gap between intuitive and
logical thinking remains unsolved.
Currently, a practical way is to maximize human creativity
when conducting design activities, and this leads to the notion of
human–machine integration. Chen et al. [109,110,210] proposed
the idea of ‘‘human–machine integrated system’’ composed of
human intelligence and machine intelligence, which emphasizes
human’s role in design and the construction of human centric
design system. Computer system is no more the only components
of manufacturing system, but an important part connoting
machine and human. Despite the quite enlightening ideas, the
essential issue of the nature of thinking has not been solved in a
general framework [40]; moreover, it is still questionable whether
overemphasizing the role of human would facilitate the establish-
ment of automating design paradigm. Hence, after realizing the
advancement of human–machine integration, a natural question is
how to harmonize it. The bottleneck lies in the difficulty in
developing new theories and methods capable of perceiving the
data and reasoning in an intuitive way like humans in order to
maximize the power of human–machine integration. To solve this
problem, understanding human thinking framework holds the key
and several new theories have been proposed. Yin
[206,207,209,216] attempted to develop an AI framework based
on imaginal thinking. This exploration helps us to make one step
closer to the ultimate goal of enabling machines to conduct all the
design tasks with truly harmonized human–machine integration.
3. Human–machine integrated automating design paradigm
As shown in Fig. 4, the paradigm of human–machine integrated
automating design emphasizes the integration of machine’s
massive computation and logical deduction abilities and human’s
experience, intuition and creativity, whereas automating design is
essentially machine-centred with the ultimate goal of completely
emancipate human from the design workloads. It is mainly
composed of three areas: the prototype of human–machine
integrated automating design, ontological modelling of knowledge
management and epistemology-based automating rationale, and
intelligent human–machine interaction. Fig. 4 provides a global
G Model
CIRP-1406; No. of Pages 23
6 Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
Fig. 4. Human–machine integrated automating design paradigm. Note the loops of
kinematic optimization (1, 10 , 100 ), dynamical optimization (2, 20 , 200 ) and geometric
optimization (3, 30 , 300 ).
view of the whole automating design process that embodies the
human–machine integrated design notion with cross-domain,
collaborative and concurrent characteristics aiming at design
integration and collaboration.
The ontology-based design theories and methodologies have
been extensively studied and applied in the design realm
[30,69,70]. Ontology in design science can be considered as
common vocabulary or standardized terminology of specific
domain knowledge [70], and often provides the hierarchical
structure and the widely acknowledged semantic-network of that
knowledge [31]. For design of the general sense, ontology is
favourable for the grasping of a methodology or some domain
knowledge, while for automating design, the reasoning of the
entire design process or the knowledge is particularly crucial, as
the machine is expected to mimic the design ideation and
knowledge reuse process of humans [209]. This leads to the core
issue of automating design, i.e., the human cognitive principles and
models [207,209], which are however under broad debates and the
cognitive models remain coarse [201]. From the philosophical
view, reasoning or cognitive science is directly related to
epistemology, as introduced earlier. The philosophical concepts
of ontology and epistemology are tightly coupled with each other,
i.e., neither will make sense without the other. Therefore, ontology
in the design world has been inherently endowed with the
cognitive process or attributes, but is focused on a relatively static
structure, framework, or an exchange protocol, e.g., in the
representing manner of a semantic-network [185]. However,
aiming at the demands of reasoning in automating design, it
becomes a necessity to introduce the twinned concept of
ontology—epistemology into design science, and in effect the
relevant efforts have already unconsciously proposed concepts
that fall in this category.
Compared to ontology, epistemology depicts the dynamic
information flow among the entity concepts within the prescrip-
tive ontology. Although at the first glance epistemology in this
sense seems to be similar with semantic-network, the distinction
between them is evident: although a semantic network may also
formulate the information flow, usually the interoperable relations
are without teleology. On the contrary, epistemology is aimed at
reasoning with a clear purpose. It contains the reasoning
information flow that bridges the gap between the initial
knowledge state and the desired knowledge state. Examples
include the design prototypes illustrated in Figs. 4 and 5. Besides,
teleology is also the main difference between human and machine.
Design ideation of human is always purpose driven, also acting as
the prerequisite of creativity, while machines cannot generate a
‘‘purpose’’ autonomously, which is also a main challenge for
automating design. To this point, the preliminary definition of
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
Fig. 5. Ontology of automating design and its semantic-network [85].
epistemology can be formulated: epistemology specifies consen-
sual or formalized information flow of reasoning or cognitive
process towards an ontology. The following discussion on
automating design with human–machine integration would be
implemented from the perspectives of both ontology and
epistemology.
Viewing from the present technical limitations, a key challenge
in automating design is accomplishing the automated decision-
making under critical conditions of complex product design, when
faced with massive data, insufficient prior knowledge, missing and
contradicting data. Although a computer is highly effective in
relatively uncomplicated and structured design tasks, human–
machine integration is crucial in a way that designers can refer to
the embodied possible solutions rapidly generated by a computer,
meanwhile focus on creative works, decision-making and opti-
mization.
Effective human–machine integration depends largely on an
efficient human–machine interaction interface. The most widely
applied intelligentization method is the visualization interface,
which is an object-oriented system [3]. The integration of
computing visualization and application of VR/AR techniques
makes the system human-oriented [106,128,187]. On the other
hand, Internet of Things (IoT), as the backbone of the system, is the
basis for information sharing and resource allocation [7,197]. With
the help of these technologies, humans are left to define the
priority list of optimization, set specific design objectives and
constraints or select optional choices, whereas the automated
computer platform is responsible for tedious, specific, or repetitive
computation and logical tasks, releasing designers to concentrate
on more creative work.
The challenge of automating design with intelligent human–
machine integration can be concluded from three perspectives:
 ontological modelling and ontology-based knowledge manage-
ment of automating design;
 epistemologies of automating design; and
 thinking frames and support tools for intelligent human–
machine integration.
3.1. Ontology of automating design with intelligent human–machine
integration
Automating design is essentially a knowledge operation
process [180] based on human–machine collaboration that is
intrinsically continuous, dynamic, systematic, iterative and
coupled.
The epistemological framework of automating design has been
shown in Fig. 2. To facilitate the efficient communication between
human and machine, the knowledge for design is modelled and
represented using ontologies that formalize a well-structured
description for knowledge sharing and processing. Moreover,
ontology is especially powerful in formalizing a global theoretical
G Model
CIRP-1406; No. of Pages 23
Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
system. Therefore, accompanied by the evolutionary process of
ontologies promoted by epistemology, the earlier formed local
ontologies ought to be continually refined and integrated into a
global one. The fundamental elements of the local ones are of
course defined and formulated by humans, whereas for automat-
ing design, a foreseeable trend or demand is that the formation of
the global ontology based on local ones should be automated. This
requires a sophisticated framework of ontological modelling and
integration aiming at facilitating machine processing. This is also
one of the main tasks in human–machine integrated design and
collaborative design. The topics covering ontological modelling,
integration, ontology-based knowledge management, the episte-
mology of automating design, as well as the human–machine
integration interface are discussed in detail in the following
subsections, while one needs to start from the origin of the current
widely accepted ontologies in automating design, or the associated
prototypes.
3.1.1. Prototype of automating design
As the fundamental core of automating design, a prototype can
provide answers to questions such as ‘‘what should be automating
design?’’ and ‘‘how does it move forward?’’ The formation of the
automating design prototype can be dated back to the widely
accepted FBS design framework of Gero [59].
In spite of the further progresses, the improved FBS design
prototypes are still not fully suitable for automating design
because the highly brief and conceptualized description of design
does not quite fit into the logical processing architecture of
computers, i.e., the lower level deduction of the computer has to be
stipulated by the higher level human intelligence. Therefore, a
more detailed and specific framework is needed. Pahl and Beitz
[131,178] suggested that the knowledge concluded from the
developing of new products can be stored in the design knowledge
base for future use, while the staged results of the design phases
proposed by them can be further specified. Consequently, by
combining the advantages of their work and the FBS design
prototype, an automating design framework capable of multi-
objective optimization has been proposed considering both
ontological and epistemological views, meanwhile it integrates
human and machine intelligence by applying imaginal thinking
methods [206,207,209,216]. As shown briefly in the middle part of
Fig. 4, this prototype consists of five areas of domain knowledge,
namely requirement (R), function (F), configuration (C), structure
(S) and details (D), in which the last three terms can be collectively
referred to as ‘‘S’’ in the traditional FBS prototype. A more detailed
explanation of this prototype can be found in [81]. However, the
FBS prototype is aiming more at design of a general sense or
broader scope, while the subdivision of S into configuration,
structure and details is particularly focused on the essential
properties of automating design, since the inherent logical
structure of computers can only implement design step-by-step,
without the possibility of jumping directly from function to
structure or details like humans do. Thus, this most recent
prototype provides a framework more suitable for machine-
centred automating design process.
3.1.2. Ontology of automating design and its semantic-network
The prototype introduced above demonstrates the fundamental
notion of automating design well, and its structure can be
recognized easily from the existing design theories and methodol-
ogies [34,99,207], in which diverse terminological systems have
been established from different perspectives, especially when
computer science, information technology and artificial intelli-
gence [61] are heavily involved. Therefore, an effective tool that is
capable of unifying the expressions of the design rationale
becomes a necessity. Actually, the recognition of the prototypes
has gradually formed a solid consensual foundation for the
understanding of the design automation process, giving rise to
the introduction of ontology. Till today, ontology has been
extensively applied as it is recognized as a standardized
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
7
terminology, a formalized knowledge or an exchange protocol
[185] of some domain. Furthermore, ontology is essentially
equivalent to semantic-network, or ‘‘a graph with nodes repre-
senting concepts or individual objects and arcs representing
relationships or associations among the concepts’’ [29]. An
example is shown in Fig. 5, which illustrates the basic semantic-
network [85] of product ontology. It contains five main sub-classes,
namely requirement, function, component, product and behaviour,
which are linked by the semantic relations such as has_Attribute,
has_Port and has_Function etc.
On the other hand, the relationship between design and
knowledge is so tight that these two terms can be regarded as the
two facets of the same object. For example, the GDT theory was
commented by Tomiyama [177] as ‘‘not a design theory but an
abstract theory about knowledge and its operation’’. Thereby, the
significance of effective and unbiased representation and manage-
ment of knowledge is self-evident. For such reasons, the following
sections are devoted to the ontological modelling of automating
design from the perspective of knowledge management (KM).
3.2. Ontology-based knowledge management for automating design
Unified and unambiguous representation and management of
knowledge are critical to realize automated machine reasoning
and decision-making by human. Design knowledge can be treated
as a whole set of intuition, reasoning, insights, and experience,
which is interpreted by information and data [9]. It often
originates from a wide variety of sources—heuristic, qualitative,
quantitative, etc. Moreover, modern product development process
involves the collaboration of engineers from different organiza-
tions or different areas of expertise [149]. Thus, a consensus must
be reached on the understanding of one specification. Traditional
knowledge management systems often fail to provide adequate
support due to their inflexible data structures, hardwired
application procedures, as well as their limited ability to integrate
process and product knowledge. Additionally, the prevailing use of
natural language in knowledge representation has led to the
widely existing ambiguity due to personal terminologies [48]. It
also largely hinders the knowledge exchange between human and
machine. All these mismatches put the blame on the lack of a
common language or a common conceptual foundation.
With ontologies, the knowledge content of an area can be
standardized, and becomes a common basis for communication,
knowledge sharing and problem solving. Combining natural
language-based understanding and knowledge-based problem
solving [30], ontology has been used to realize uniform represen-
tation of design knowledge and the automatic mapping within it.
There are three key advantages of using ontology as a knowledge
management tool for automating design. First, the core contents of
an ontology are entities and relationships rather than structures
and forms, and this is consistent with the thinking style of human.
It greatly facilitates the integration of human’s knowledge and
experience. Second, an ontology is an incompact but organized
network with unambiguous semantics, which is advantageous in
knowledge updating and multi-disciplinary integration. Third, an
ontology can provide a uniform framework for describe the object
as well as the process of automating design.
3.2.1. Ontological modelling of knowledge
Ontological modelling of automating design knowledge is
essentially the development of a well-structured and explicit
knowledge base for automated rationale and decision making.
Hence, a coherent hierarchical structure of ontology should be
constructed, in order to facilitate the navigation, browsing and
searching of knowledge, which are the principal knowledge
operations a computer can perform. With the purpose of
formalizing all the knowledge concerned with the automating
design process, pioneering root concepts of ontologies identifying
the purpose and scope of taxonomies have been proposed.
Darlington et al. [43] developed three types of ontologies to
G Model
CIRP-1406; No. of Pages 23
8 Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
capture concepts and relationships associated with different
aspects of design, which are comprised of requirement ontology,
product finish ontology and machine motion ontology. Ahmed et
al. [1] categorized design ontologies into four root types: process,
function, physical product and issue. Although their works are
purely concentrated on the design process, it is worth noting that
they not only specify the generic concepts but also provide the first
platform for codifying the taxonomies for a specific design.
However, the differences of knowledge from different sources
like computer reasoning mechanism [201] or human experience
[32,91], etc., are not taken into account. Thus, it remains difficult to
distinguish the roles of computer and human in automating
design, making the development of an effective integration
interface even more difficult.
Based on the previous researches, for human–machine inte-
grated automating design, the upper level ontologies should be
classified into design ontology and resource ontology. The former
describes the product as well as the design process, which can be
further divided into product ontology and process ontology. The
latter is an integrated representation of human and computer
knowledge for automating design, as shown in Fig. 6. These
ontologies and their detailed semantic relations are discussed in
the following sections.
Fig. 6. Ontological taxonomies of automating design.
3.2.1.1. Ontology of product knowledge from the view of automating
design. Product ontology, which partly resembles a product
lifecycle management system, is developed to represent unambig-
uously and systematically the information, domain knowledge and
their relations with the artefact associated with the evolution of its
states throughout the cycle of automating design. With increasing
complexity and diversity of modern products, a meta-model has
been used to encourage the reuse of product models and the
flexible representation of knowledge [179]. A complete architec-
ture of product ontology consists of four layers: representation,
meta-ontology, model and instance. [203,209].
 Representation layer is aimed at choosing an ontological
representation language to formalize the knowledge models of
products. Typical representation languages encompass OWL, KIF
(Knowledge Interchange Format), UML, etc.
 Meta-ontology layer identifies general terminologies, vocabu-
laries of concepts and their relationships that are available to set
up a common foundation for different domains. It serves as a
common semantic framework on which product models for
specific design domains can be built through the inheritance or
reuse of concepts and relations in the meta-model.
 Model layer depicts the product models for specific domains
such as computer model and automobile model, which are
clustered using meta-ontologies.
 Instance layer is the top layer, which denotes a concrete product
design result that conforms to the models without violating any
constraint. The difficulties in modelling product knowledge can
be significantly reduced by constructing product models based
on existing cases.
Starting from the FBS model in the 1990s and its ontological
extension in the 2000s [60,61], three levels of ontologies have been
defined to categorize the different states of artefacts in the design
process. They are the function ontology, or ‘‘what the object is for’’,
the behaviour ontology, or ‘‘what the object does’’, and the
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
structure ontology, or ‘‘what the object consists of’’. RFBS model
[34], which is adapted from the FBS model, realized automated
mapping between functions and generic organs based on seman-
tics and ontologies, and further developed a mid-level ontology
which integrates lower level ones [83]. Yin’s imaginal thinking-
based human–machine integrated design methodology [207]
suggests that the whole automating design process can be viewed
as the transformation among the states of the product in terms of
its requirement, function, configuration, structure, detail and
behaviour. Based on the above discussion and the semantic
relationships among the meta-ontologies, the sub-ontologies
belonging to the product ontology of automating design are
introduced in detail below.
(a) Requirement ontology. This ontology is developed to unify the
specifications of the tasks, which are a set of constraints and
performance criteria posed by the system and customers.
Requirement ontology typically places the restrictions on
either functions or structures of products, which are described
in the form of behaviour. Ontological modelling of requirement
is extremely significant, due to the fact that an unambiguous
description of requirements can unify the initial imports,
which facilitate the automatic reasoning of automating design.
A knowledge-based tool for the specification of the require-
ments in automotive systems has been developed through the
application of ontology by Roy [149]. The ontology for
managing requirement has been developed in products,
information and process levels to achieve a consensual
understanding of a product between automotive OEM and
suppliers. Linet et al. [108] classified requirements into six
categories: physical, structural, functional, cost, performance
and requirements, depending on the properties of the artefact.
The relevant objects and their attributes were described based
on first-order logics. Christophe et al. [34] proposed an
effective requirement structure for a mobile robot, including
the expectations in performance as well as the constraints
according to the operating environment of the system.
In some studies, a similar ontological concept named
‘‘issue’’, is applied as a synonym of requirement. It refers to
the problems a designer considers during design ideation,
which are not always explicitly modelled and are limited to a
specific domain. Ahmed et al. [1] identified four classes of issue
ontologies, i.e., lifecycle, environment, interfaces, and func-
tional requirements of the product and their characteristics.
(b) Function ontology. Function represents the necessary knowl-
edge to explain the performance of a product or artefact aiming
at fulfilling customers’ needs, which can be transformed into
configuration in conceptual design phase, the early-stage of
automating design. Function ontology is developed in order to
automate this significant process to avoid the unexpected
uncertainties, as proposed in the intelligent CAD system
[6,179]. Plenty of studies have been dedicated to this issue.
Automated function-based synthesis methods rely on the
user’s ability to develop functional descriptions of the desired
product [28]. A formal function representation can provide a
standardized abstract set of function-related terminologies
that are independent of physical structure to strategically
guide the design [181]. Scientists have developed many
knowledge representation schemas for computers and expert
systems to enable the possibility of automatic conceptual
design. Hirtz and Stone [83] proposed an integrated ontological
functional basis with formal representation and a standardized
set of function-related terminologies to support functional
modelling. The taxonomies have been widely employed to
describe the function due to their advantages of completeness
and compatibility. The resulting taxonomy is referred to as the
functional basis with a set of functions (verbs) and flows
(nouns).
(c) Product ontology. It is consisted of sub-assemblies or parts and
serves the management of product design cycle. According to
G Model
CIRP-1406; No. of Pages 23
Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
the prototype of automating design, product ontology can be
divided into configuration, structure and details as the results
of conceptual, embodiment and detailed design phases,
respectively.
Configuration ontology specifically extracts classes encap-
sulating function and their intrinsic attributes. Thus, it is the
representation of all the legal combinations of components as
well as relevant technical and financial constraints. Researches
regarding the application of ontology to product configuration
have raised a broad interest in both industry and academia
recently. Soininen et al. [164] presented a general ontology of
configuration, which integrates the concepts of port, function,
constraint, attribute, etc. Configuration ontologies can be
formalized in Ontolingua [68], a knowledge representation
language combining first order logic and frame paradigms.
Altuna et al. [5] developed configuration ontologies in the
OBELIX project to support collaborative design of products and
services, and these ontologies are further categorized into
component ontology, constraint ontology and requirement
ontology, each of which has its own semantic structure. In their
work, general terminologies, vocabularies and relationships
are defined in the previously mentioned meta-model of the
four-layer architecture [203] to facilitate the reuse of
configuration models. An illustrative example of a ranger
drilling machine was employed to show the efficiency and
effectiveness of the method.
Structure ontology, on the other hand, is the embodiment of
the configurations, and is represented by the various design
models aiming at formalizing the components and their
relations. Chang [31] put forward the methods of ontology
development and utilization for DFM to study the most
important phases in the design process, including concept
categorization, class development, slot categorization and
development, and the identification and implementation of
the relations among the slots. The DFM ontology can be used to
capture the domain knowledge and define the structure and
relations of a data set. Last but not least, detailed ontology deals
with the geometric parameters, forms, dimensions and surface
properties of all the individual parts.
(d) Behaviour ontology. Behaviour ontology can be categorized into
two classes: expected and physical. Deduced from the
requirements and functions, the former indicates how the
system is supposed to operate. The latter represents the
characteristics of the product determined by the internal
factors like structure, material, etc., and is deduced from the
analysis of product ontology. The ontological modelling of
behaviour is essential to setting up a unified interface for
human decision making that relies on the evaluation of the
behaviours, which is one of the most important steps in
automating design.
3.2.1.2. Ontology of the automating design process. Process ontology
defines the basic terminologies for computer-centric automated
rationale and human-centric decision making in human–machine
integrated automating design. As introduced earlier, many studies
have been dedicated to the modelling of the design process [1], and
various taxonomies have been proposed from different dimen-
sions. These pioneering works laid the foundation for constructing
process ontology in automating design.
There are three theories originating from different perspective
of views. Proposed by Pahl and Beitz [131], design is defined as a
step-by-step decomposing and abstract-to-concrete process,
which can be divided into planning and clarification, conceptual
design, embodiment design and detail design. Their work revealed
the basic framework of automating design from a traditional view
of logical progression. Aiming at recognizing the states transfor-
mation of products during design ideation [59], design can be
described as a set of activities, including formulation, synthesis,
analysis and evaluation, etc., each of which is the task needing to be
automated. From the epistemological view, Sim and Duff [162]
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
9
pointed out that design is a cognitive process by defining the three
main classes of definition, evaluation and management, and
advocated that the design process can be elucidated using
knowledge level theories [124]. Accordingly, ontological model-
ling of knowledge plays the key roles in automating this cognitive
process.
3.2.1.3. Ontology of resource knowledge. Resource ontology can be
simply put as the modelling of available or accessible knowledge
during the automating design process. While the classification of
resource is crucial for its representation and use [9], it can be
implemented according to different dimensions or basis [29], such
as product vs. process (the content of description) [131], and
dynamic vs. compile (the state of existence) [165], etc. For
automating design, in order to distinguish the knowledge from
computer and human, the resource ontology can be classified into
explicit class and tacit class [129,182], whereas these different
classes of ontologies are closely connected with each other.
Explicit resource acts as the grounding basis for computer
reasoning, which aims at creating the intellectual platform for
automating design or automated inference, such as computer
algorithms, expert knowledge systems, problem solving routines,
optimization methods and mathematical tools stored in reposi-
tories. On the other hand, tacit resource is indispensable in
creating new values in a product. The resource ontologies for
human–machine integrated design clearly provide the hierarchi-
cal structure of related knowledge elements. On the whole, the
resource ontology can be specifically categorized into fundamen-
tal theory, optimization method and human experience.
Fundamental theory is the principle to guide the whole
automating design process, and can be further classified into
several sub-ontologies according to different disciplines, such as
design methodologies, mathematical and mechanical principles,
etc. These fundamental theories consist of axioms, theorems and
corollaries that have been proven to be correct during the
exploration of design activities.
Optimization, in particular, has established itself as a mainstay
in the field of engineering, and has become an essential part of
automating the design process. It helps the designer to address the
optimal result from numerous candidates that are in accordance
with the fundamental theories. Witherell et al. [201] proposed an
ontology-based approach called ONTOP for the management of
optimization knowledge. This approach reduces the time used to
be spent in recreating models from scratch, and further facilitates
the elimination of costly trial and error approaches. ONTOP has the
potential to remarkably improve the ability of the system to deploy
optimization knowledge.
Experience resource, which originate from human, can be
classified as declarative knowledge (know what), procedural
knowledge (know how) and causal knowledge (know why)
[70]. Chen [32] developed a method of ontology-based empirical
knowledge representation and reasoning. The method adopts OWL
to represent empirical knowledge in a structured way in order to
help knowledge requesters understand clearly the knowledge
without any bias. His method has been validated as being able to
facilitate the automation of tacit knowledge storage, management
and sharing, and provide the requesters with accurate and
comprehensive empirical knowledge for problem solving and
decision-making.
3.2.2. Automating integration: Local-to-global ontology transition
As discussed above, knowledge of special discourse domains
are modelled using local ontologies. However, each of the local
ontologies has its own independent semantic structure. Addi-
tionally, different ontologies often use different concepts or
terminologies, even when describing the same subject. An
ontology itself does not solve the interoperability problem, but
only raises the heterogeneity from the data level to the semantic
level [214]. Moreover, automating design can be represented as a
mapping and fusing process between the global ontology and
G Model
CIRP-1406; No. of Pages 23
10 Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
different local ontologies. Therefore, how to realize automatic
transition from the local ontologies to the global one becomes a
prominent issue of automating design [85].
Ontology integration is one of the critical techniques in setting
up the semantic relations and reconciling differences among
ontologies, as it enables semantic information exchange to be
implemented in a global manner [85]. Consequently, it amplifies
designers’ capability in the product development, and reduces the
development cycle time and cost in automating design. Research-
ers have proposed several frameworks for ontology integration
aiming at providing a guideline to realize a seamless automatic
integration. Chen [32] proposed a web service-based ontology
integration framework for the collaborative design of moulding
products. Ontology integration is regarded as the core part of the
design process, providing smooth and seamless flow of informa-
tion among different phases of the design. His work is supported by
Yang [204] and further extended to the manufacturing field.
Aiming at the human–machine integrated design of ultra-precision
grinding machine, Hong et al. [85] have divided the whole
integration task into four consecutive parts: local ontology,
ontology integration, global ontology and ontology searching, as
illustrated in Fig. 7. In his work, knowledge from different
domains, such as bearings, drivers, etc., are modelled by ontologies
separately and then integrated into a global ontology. After that,
multidisciplinary coordination and optimization of the design
implemented by different agents located diversely can be
harmoniously realized. Ontology integration can be divided into
two steps: ontology mapping and ontology merging, which are
accomplished through a public integration server that ensures the
automation and coordination of the process.
Fig. 7. Ontology-based framework of knowledge integration [85].
3.2.2.1. Ontology mapping. The complete process of automatic
ontology mapping and merging proposed is illustrated in Fig. 8
[85]. There are two ways for the mapping process: either by using
a shared ontology or by using a heuristic-based approach [214].
Concept mapping is proceeded with the local ontologies
established or provided by experts in some domain with the
purpose of identifying similar knowledge concepts in other
domains. Steps in ontology mapping are similarity matching,
similar concepts aggregation, and linking [33]. As the key point of
ontology mapping, the similarity matching between two concepts
should be firstly carried out. Yang et al. [204] proposed an
approach to calculate ontology similarity between a local ontology
and the target global ontology, in terms of the concept name,
essential information, and relationships. Their work is developed
by Zhan [214], who proposed that similarity comparison should be
composed of attribute similarity, composition similarity and
inheritance similarity. Hong et al. [85] proposed that a complete
matching process involves the comparison of concept names,
attributes, semantic relations and instances, etc., according to the
basic structure of ontology. In similarity matching for concept
names, concepts are firstly decomposed into terminology sets of
unit words, and the similarity grade between the two concepts can
be calculated with the Jaccard coefficient [74,96]. Ciocoiu et al. [37]
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
Fig. 8. Ontology mapping and merging process [85].
presented a method for building ontologies by providing semantics
for terminologies in a computer-interpretable format to achieve
interoperability among different engineering applications.
Ontology mapping plays an irreplaceable role in model-based
system engineering, and has become a promising solution for
automating design of complex mechatronic systems. A SysML-
based modelling approach was proposed to realize automated
mapping between design and simulation models in order to
facilitate system engineers to express the various behavioural
requirements [24]. Dori and Shpitalni [46] proposed a fundamen-
tal ontology construction approach for a generic product that
incorporates hardware and software components. The ontology
can serve as a basis for a knowledge model to cover the entire
product lifecycle, and can be applied in the VRL-KCiP Network of
Excellence in order to obtain the required knowledge and to
organize it in a consistent and meaningful form.
3.2.2.2. Ontology merging. After ontology mapping, similar con-
cepts in both the global ontology and local ontologies should be
merged. Ontology merging consists of two steps: (1) concept
merging; (2) relation reconstruction [204].
Concept merging means if a concept is the same as any of those
in the global ontology, then merge its concept name, attribute and
instance. Otherwise, create a new concept in the global ontology,
with the holistic information in local ontologies. After that,
semantic relations should be added into the global ontology,
including hierarchical and process relations.
3.2.3. Ontology-based knowledge sharing for collaborative design
The increasing complexity of modern products necessitates a
distributed and heterogeneous collaborative environment for
automating design. Concurrent and collaborative design of
products and processes have thus become essential, and successful
firms today are those that are able to manage concurrent activities
in an intra- or inter-organizational network and deliver value to
the marketplace. In a collaborative automating design environ-
ment, an ontological knowledge base is formed by assembling
design sharing, consistent interpretation of constraints and
requirements, and the evolution of a web of relations. Efficiently
obtaining the desired knowledge is the prerequisite to implement
a collaborative automating design, which is accomplished by
ontology searching and sharing.
3.2.3.1. Ontology searching and sharing. Knowledge can be classi-
fied into individual application (design process) and integrated
system (collaborative design) from the view of collaborative design
[29]. Gruber [70] proposed a method for the development of
ontologies with content-specific agreements for the sharing and
reuse of knowledge among software entities to facilitate the
G Model
CIRP-1406; No. of Pages 23
Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
automation of design process. Zhan et al. [212] presented a
semantic approach that uses ontologies to share the knowledge
related to product data in CAD/CAE applications and to integrate
design evaluation information provided by the individual applica-
tions. Their work greatly enhances the ability of the design teams
to automatic create, share, and exchange knowledge for solving
design evaluation challenges involving multiple applications and
views.
Human–machine interfaces among subsystems in collaborative
product development projects are mostly defined by interface
control documents [139]. Recently, a computer-aided methodolo-
gy for defining and controlling subsystem interfaces and ports
using ontologies has been presented. A software can now replace
interface control documents [85], and support collaboration based
on the proposed architecture and its functionality. The software
allows designers to load subsystem descriptions from a shared
repository and link them together by defining connectivity rules. It
can also report errors when port requirements or connectivity
rules are violated.
3.2.3.2. Semantic web. Semantic web is recognized as the third
generation web, which core uses ontology and descriptive logics.
With the explosion of data and information, which may further be
geographically dispersed, the web has become a primary resource
for automatic knowledge capture for automating design. To
overcome the difficulty caused by the heterogeneous formats
and representation with personal preference in knowledge capture
and sharing, the sematic web has been utilized to provide the
designers with seamless access to an effective collection of various
knowledge resources via the Internet.
Semantic web supports the ‘‘understanding’’ and automated
handling of knowledge by computer [12]. It provides a common
framework for the liberation of data based on the Resource
Description Framework (RDF), which integrates a variety of
applications using XML as the exchange syntax. Raw data in the
databases are collected, and connected to the models of the world
through ontology [195]. The framework of semantic web was first
proposed by Berners-Lee in 2000 [10] and revised along with the
recognition of the needs for effective user interfaces in 2006 [11],
as shown in Fig. 9.
Fig. 9. Pyramid of semantic web languages [11].
Existing knowledge models can be connected via the internet and
designers can access product models, reuse existing design
knowledge, and promote exchange of knowledge across multiple
projects and disciplines. OntoEdit has been introduced to integrate
numerous aspects of ontological engineering in the semantic web,
and terminologies are precisely defined for easy communication
among people and applications [174]. Li et al. [107] proposed a
semantics-based approach which combines heuristic reasoning
with the ontologies in semantic web to facilitate the knowledge
reuse in aircraft tooling design. Their work has solved the challenge
of semantic transmission between products, software and designers.
Zhang et al. [215] proposed an ontology-based modelling frame that
evolves along five consecutive layers: knowledge elicitation,
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
11
product modelling, ontology modelling, knowledge reuse, and
knowledge application layers, to facilitate knowledge searching and
sharing on the semantic web. Wang et al. [195] presented an RDF-
based knowledge model to realize the knowledge representation on
semantic web for design innovation. In their work, design
knowledge is treated as a content language dealing with the
messages in the multi-agent system.
After all, knowledge does not inherently exist but is obtained by
interpretation of information deduced from the analysis of data in
the form of observations, computational results and factual
quantities [29]. Therefore, a complete semantic web-based
knowledge capture and sharing system can be divided into four
consecutive layers: distributed knowledge elicitation, ontology-
based knowledge modelling, web service-based knowledge inte-
gration and knowledge reusing and sharing, as shown in Fig. 10. In
this scheme, diverse knowledge assets are wrapped as ontology-
based web services in order to facilitate knowledge consumption,
reuse and supply on the Semantic Web [215].
Fig. 10. Ontology-based design knowledge modelling for reusing and sharing on the
semantic web.
3.3. Intelligent human–machine interaction
The ontology-based knowledge management framework dis-
cussed previously also provides a well-structured foundation for
human–machine interaction, which is a necessity in the current
automating design paradigm. Human–machine interaction aims to
achieve seamless cooperative design between human and com-
puter so that they are able to perform mutual compensation
[207,208]. It is the intersectional area of philosophy, cognitive
science (design ideation), computer science and technology
(design rationale and representation) [26]. This indicates that
two basic issues have to be clarified: the thinking frame that unifies
human intuitive thinking and machine logic reasoning, and the
support tools for communication between human and machine.
Thinking frame is crucial because it serves as the fundamental
guideline for human–machine interaction in automating design. As
emphasized at the beginning of this review, it is of great value to
ravel out the confusions in the intelligent nature of design.
Nowadays, social and cognitive science in the computer industry
helps to clarify the real nature of design, instead of supporting a
received view [26]. Traditional AI frames, symbolism and
connectionism for instance, have developed monolithic body of
theories for logical thinking and intuitive thinking, respectively,
and both have been implemented successfully in solving practical
problems. However, in spite of their great achievements, the
harmony between them can hardly be reached and their kernel
theories are still incompatible with each other [207,208]. Thus, to
enable human–machine interaction to deal with the bewildering
relationships among the intelligent nature of design, human
intelligence and AI, an essential breakthrough is required. For this
issue, imaginal thinking [207,209] helps to achieve smooth
communication between human and computer, and the methods
based on this frame have been successfully applied to solve some
practical design problems [206,216].
On the other hand, even with a theoretically well-functioning
thinking frame, it still cannot be put into practice without technical
support of proper interface. Consequently, generalized images of
G Model
CIRP-1406; No. of Pages 23
12 Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
imaginal thinking need to be supported by various levels of
technical tools. Currently, the primary design embodiment media
are paper-pencil and digital CAD system, which are complemen-
tary to each other to a certain extent [166]. Paper-and-pencil
sketching affords fast and natural idea development in a creative
but ambiguous manner, and CAD systems allow for a more precise
and detailed design but lack of seamless and intuitive interaction
[130]. The combination of VR and AR as new visualization
mediums and promising interaction techniques provide unique
features which raise interest in both academic research and
industry. With the help of VR and AR techniques, a virtual product
prototype can provide excellent designer–computer interfaces,
allowing the designers to explore the fine details of a complex
product [17,121].
The intelligent human–machine integration can thus be
explored from two perspectives. One is the thinking frame that
tries to unify the human and machine intelligence, and the other is
the support tools for the representation of generalized images
which provide better interfaces and facilitate the mutual interac-
tion. Finally, the VR and AR based human–machine interface for
automating design will be presented.
3.3.1. Thinking frame for intelligent human–machine interaction
3.3.1.1. Traditional AI frame. Biologically speaking, neurons are the
fundamental elements of the human nervous system. Neurons
communicate with one another via synapses, where the axon
terminal of one cell impinges upon other neurons’ dendrites,
somas, or less commonly, axons. Human brain has a huge number
of synapses, the number of which goes to maximum when humans
are three-year-old and then declines with maturation. Although
much of human brain remains mysterious, the emerging role of
social and cognitive science is diverse, pervasive and critical [26].
Symbolism and connectionism are two of the theories that have a
long history and both own extensive supports [208]. However, it is
also unavoidable that there are disputations and divergence
between them.
The greatest contribution of symbolism to AI is the notion of a
physical symbol system hypothesis [123] first proposed by
Newell and Simon. The hypothesis considers all the intelligent
creatures as symbolic systems, and the intelligence comes from
the processing of symbol. However, there are still a lot of
uncertainties on whether artificial intelligence could be achieved
by symbolism. First, human does not rely merely on logical
thinking to solve problems, non-logical deduction also plays an
important role in human mental activity. Second, when the
knowledge database reaches a huge volume, how to manage it
and search the database in acceptable time becomes a main
technical problem (‘frame problem’). Finally, even the frame
problem was solved, the intelligent system based on symbolism
is still uncertain to produce human intelligence. After all,
searching the database is not the way human deals with
problems.
Connectionism, which imitates human neuron structure as the
main method to realize intelligence, uses a concept opposite to
symbolism [54]. It proposes that only if the machine has the same
structure as human brain, intelligence will have a great probability
to occur. The artificial neural network is the most well-known
approach. There are two central issues in realizing connectionism,
network structure and the learning method. As connectionism
aims at achieving artificial intelligence through the mimicking of
human brain structure, the intelligent feature largely depends on
the network structure. On the other hand, a prototype of artificial
neural network is rather ‘‘empty’’ without any expected intelli-
gence. Learning and training are essential in adjusting the network
structure and weights of connections among the neurons, so as to
form the problem-solving knowledge. In other words, in connec-
tionism, the learning problem is more crucial than the structural
problem. Therefore, the study on machine learning is pertinent
[208].
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
3.3.1.2. Human fundamental thinking frame. In a current view,
human fundamental thinking frame can be considered as being
based on images, as traditional AI frames have been challenged by
Brooks’ action-based AI theory which was introduced earlier
[19,20]. Experimental psychologists have also shown clear
evidence for image-based thinking processes [47], which is also
supported by the psychological researches in the field of design
where sketching has traditionally been thought as a core
conceptual tool during the design ideation process [14,42,57,67,
135,153,175,183]. The imaginal thinking frame taking image as
the fundamental characteristics agrees well with these findings,
and seems to be promising to serve as the fundamental frame that
unifies different thinking frames.
From the perspective of imaginal thinking, humans thinking
[3,19] is quite different from the logical step-by-step computa-
tional operations that computers perform in a framed world. The
images mentioned here are generalized, including not only the low
level information directly apperceived by the sensing apparatus of
human body, but also high-level information of experiences and
knowledge obtained by imitating and learning. Viewing from the
human nervous system, the information discussed here represents
signals in the huge neural network when stimulated by a certain
perception. If treating one input of a single neuron as one
dimensional, with an average 1011 neurons and 7000 synapses of
each neuron, high dimensional information is always generated by
the interactions of all the neurons. Such high dimensional
information is also a form of images, i.e., the information is not
limited to real images, but all that can be understood in human
mind as a general or even abstract form of images. Thus, any signal
form can be an image, which corresponds to certain excitation
pattern in the human neural network.
3.3.1.3. Imaginal thinking frame. It has been recognized that image
processing is the fundamental thinking mode [3,19]. As is shown
in Fig. 11, imaginal thinking can be divided into two categories of
learning procedure and non-learning procedure [208].
Fig. 11. The description of imaginal thinking.
The differences among imaginal thinking, symbolism and
connectionism can be well exemplified by the automating design
process. Any thinking activity, including design, is based on a
complete and systematic knowledge hierarchy, which is treated as
the combination of high dimensional images in imaginal thinking
and can be represented by ontologies, as discussed earlier. In
automating design, the aim of logical reasoning is to search the
solution ontologies that completely match the given requirements
(the matching images in imaginal thinking). In this sense, imaginal
thinking seems to agree with symbolism. However, the ‘‘If-Then’’
pattern in logic reasoning excludes the possibility of other values
between 0 (completely different) and 1 (exactly the same) when
image- or ontology-comparisons are made between the require-
ments and existent knowledge. As for the case where no matches
could be found when one’s former knowledge and experience fail
to solve the problem, human often reorganize the existent
knowledge obtained from mimicking and observation so that an
approximate solution can be figured out. This corresponds to the
learning process where intuitive thinking and imaginal thinking
G Model
CIRP-1406; No. of Pages 23
Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
Fig. 12. Roles of imaginal thinking.
reach a consensus. As specified, automating design is essentially an
upward spiral cognitive process. From the perspective of imaginal
thinking, this process can be explained using the continual
modification and creation of existent images and new images,
respectively, in accordance with the upgrading of ontologies
representing knowledge and experience.
Fig. 12 gives a clear illustration of the role of imaginal thinking
and its relation with other thinking styles. The two basic elements
of behaviourism, perception and action can be both denoted as
images, thus the whole design process can be seen as the
transformation of one image to another when seeking to generate
a satisfactory image.
So far, the relations among imaginal thinking and other
thinking frames have been well documented and can be concluded
as follows. Symbolism emphasizes step-by-step logical thinking.
The elementary logic can be considered as a type of low
dimensional image. Composed of these basic logical steps, the
design process can be illustrated as high dimensional images.
Therefore, logical thinking is a special type of imaginal thinking. On
the other hand, the connectionism emphasizes intuitive thinking
which has no clear representation of the reasoning process from
requirements to solution. Actually, such procedures only happen
when past knowledge and experience cannot meet the require-
ment, denoted as S1. . .Sn 6¼ S, and human brain goes for blind
search trying to find a similar neuron excitation pattern to solve
the problem. As long as a sudden stimuli generates a pattern which
meets the requirement, human takes this solution to generate
motion potential, which is denoted as S*  S. All the solutions,
either potential or practical, can be illustrated as high dimensional
images. Such type of intuitive thinking that cannot be represented
by mathematical logics or existing knowledge is another special
type of imaginal thinking [134].
Thus, imaginal thinking actually encompasses the character-
istics of both logical and intuitive thinking. It also has the chance to
solve the conflicts among artificial intelligence frames, symbolism,
connectionism, and behaviourism [208]. Besides, neural functions
and human thinking are tightly bound by imaginal thinking. It
further indicates the possibility for imaginal thinking to serve as
the fundamental frame of human thinking.
3.3.2. Support tools for intelligent human–machine interaction
The support tools in design aims at the smooth exchange of
knowledge and seamless human–machine integration to support
design decision making. There have been a number of studies on
how to achieve a smooth information exchange between human
and the external world, including paper drawing or machine
intelligence. While some other methods, verbal language for
instance, have been thought as overshadowing human insight
thinking [154], the utilization of image in a more generalized sense
as proposed by imaginal thinking, has become increasingly
popular. The key advantages are:
(1) Because humans tend to think in the form of image, the image-
based interface allows for less mental transformation in the
human brain [166]. Thus human designers would be able to
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
13
think in a more efficient way if machine intelligence deduces
and displays the solution in image.
(2) Sufficient levels of interactivity are the key aspects that make
current image-based representation different from traditional
graphic ones [111]. Modern techniques allow designers to see
the design solutions either from local or global perspectives.
Thus, it is much easier for designers to find the truth behind
appearance [111].
(3) Current computer science prefers to conduct design by
employing ‘‘virtual prototyping’’ techniques, which is also in
the form of image that presents, analyzes and tests the
simulation model of physical products. Besides the fact that
most computer-based analytical techniques such as finite
element analysis, weight and mass properties, kinematic
analysis, etc. are based on 3D image modelling, digital
modelling environments may employ other artificial stimuli
to promote the development process [166].
Some typical image-based representation tools and techniques
are discussed in the following sections.
3.3.2.1. Computer-aided design. CAD is the earliest and most
primitive interface for human–machine interaction. Initially, the
development of commercial CAD system was largely driven by fast
changing market requirements [90], most of which put emphasis
on efficiency and accuracy of product design rather than creativity.
Efforts were on reducing the product development costs by
automating the routine processes and increasing drawing produc-
tivity. Hence, although there have been a lot of CAD software, the
majority are dealing with the later stage of design, such as
geometric modelling and drawing [169,173]. Despite the
tremendous contribution made in automating the configuration
process, designers still have to make full decisions throughout the
design process, especially at the conceptual design stage where the
impact can be the greatest (as much as 70–80% of all the product
cost). Just as many researchers have argued and illustrated, the
later in the design process are changes made, the greater is the cost
of changes [71]. Thus, insufficient CAD development and
improvement in the downstream stage of design have become
the main obstacles. Current CAD systems overlook the possibility
of introducing new way of thinking and designing brought by the
advancement of machine intelligence [39]. Thus, CAD has to be
integrated into the earlier stage of the design process, i.e., the
conceptual design. Suh proposed the Thinking Design Machine
(TDM) supported by the axiomatic design theory. However, the
generation of creative ideas requires a broad range of knowledge
and considerable decision making ability, and this makes the TDM
remain at the theoretical level [45,58,105,117,172,173]. Yoshi-
kawa proposed the intelligent CAD system with similar consid-
eration as Suh [142,211].
Viewing from the development of CAD, it should be more of a
conceptual tool being able to develop new ways of perceiving and
conceiving design than just a technical drawing tool. CAD may
foster new patterns, relationships, or aesthetics expanding rather
than reducing designers’ creative options [90]. Moreover,
designers who focus on innovation through analogue tools alone
might effectively censor CAD during conceptual and schematic
design phases [63].
3.3.2.2. Sketching. Sketching has been recognized as the most
effective way and the most convenient interface to promote
human creativity. Combined with other graphic tools, it works as
the extension of human short term-memory, and is pivotal to the
conceptual design phase, which is largely seen as a cognitive
process with the emphasis on visualization and externalization
[87]. The limitation of human memory is one major factor that
restricts the cognitive efficacy in conceptual design [118,184]. One
well-accepted hypothesis, the Information Processing System (IPS)
provided by Newell and Simon, can be viewed as a ‘‘map’’ of the
locations in which information about the design may be stored
G Model
CIRP-1406; No. of Pages 23
14 Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
Fig. 13. The design environment [184].
(Fig. 13). Human memory is divided into internal and external
workspaces. The designer actually owns two types of memory:
short-term memory (STM) and long-term memory (LTM). STM is
very fast and powerful, directly accessed by the information stored
in LTM or accepted from external world. It is responsible for all the
design operations, such as visual perception and drawing creation.
Miller’s theory of chunking [118] states that human short-term
memory is limited to seven chunks of information. Since the
number of chunks that the STM can process is limited and
mechanical designs are getting increasingly more complicated, the
size of STM becomes a critical bottleneck for human designers.
Since textual information is so limited in characterizing forms,
graphic representation is the only reasonable memory extender for
mechanical designers [36]. When designers develop images
during the design process, they are communicating with images
and have to respond to them, through which they are able to ‘‘see’’
many ideas that did not exist before the drawing process [104].
It is strongly supported by the research in ergonomic and
psychological evaluation of design tools that sketching in the
design process is of great importance, especially in the field of
creative design [104]. The effectiveness in inspiring designers’
creative thinking has been proven in studies [104,152,167,184]. It
has been recognized that engineers are notorious for not being able
to think without making ‘‘back-of-the-envelope’’ sketches of rough
ideas. Sometimes these informal sketches serve to communicate a
concept to a colleague, but more often they just help the idea take
shape on paper. The role of drawings in engineering is more than
just to archive a concept or to communicate with others
[184]. Moreover, Schooler et al. [154] concluded from his
experiments that the relationship between language and thought
is not always symbiotic. Thus IX when words fail to work with
thought, it is the images which take the form of sketches and
drawings become very useful. In [167], the results even suggested
that paper-based sketching was more effective not only in
producing creative solutions than digital drawing but also in
supporting particular synthesis strategies.
Although good in stimulating and recording designers’ creative
thinking, traditional paper–pencil based sketching falls short when
analysis and optimization are needed after the conceptual design.
Thus it is necessary to make CAD compatible with sketching based
on machine intelligence, while the first step may be clarifying the
reason why paper-based sketching is more effective in creative
thinking.
In Ref. [104], Lawson pointed out the core problem is that the
communication between computer and human is conducted
through machine language instead of human language. It means
that the input of the CAD system probably requires a lot of highly
complicated orders. This prevents the designers from recoding
their inspirations quickly and effectively when the ideas are more
easily missed than seized. Thus, the designers have to translate
their ideas into the input of CAD systems. To overcome this
deficiency, CAD systems need to be more than tools to record well-
formed and structured concepts [184], particularly in several
ways:
 CAD systems must allow for sketching input.
 CAD systems must allow for a variety of interfaces for the
designer.
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
 CAD systems must recognize domain-dependent features and
treat them as entities.
The first tip indicates that it is unnecessary for all the graphical
representations to be as refined as that demanded by current CAD
systems, especially in conceptual design. Tip 2 suggests that future
CAD development needs to be driven by the ‘‘Design’’ rather than
‘‘Computer’’, and a good interface that is able to quickly record
designers’ ideas is preferred. Tip 3 means that the recognition of
functional features at diverse levels of refinement is needed.
3.3.2.3. Virtual reality and augmented reality. Currently, virtual
reality technology might be one of the most powerful means to
enhance intelligence and it provides the most streamline human–
machine interface by allowing the designer to explore refined
details. For example, in a CAVE-type simulation environment,
designers can walk inside a product and explore interference,
operations and other design details with a completely immersive
feeling. Augmented reality, on the other hand, allows the designers
to superimpose graphics, videos, or texts on a real product in a real
manufacturing environment and this can provide clear annota-
tions such as operation, inspection and assembly details.
3.3.2.3.1. Human–machine interface: VR view. Design using CAD
tools has been known to be too precise in the sense that exact
dimensions must be supplied and all lines and surfaces must join.
These requirements restrict the creativity of a designer as
conceptual design always begins with sketches. Both current
CAD and sketching are still the primary modelling media. Relevant
studies [89,132,166,200] have been conducted to analyse users’
requirements for immersive modelling and compare the sketching
capability of designers in 2D and 3D environment (Fig. 14). The
results suggest that satisfaction, fun of sketching as well as idea
generation and reflection reached the most positive range [22].
The main advantages expressed by participants have been
concluded as follows:
 Spatiality: The possibility ‘‘to work with the space’’, to create
spatial models and to perceive their spatial impact.
 One-to-one proportionality: The possibility to draw models ‘‘one-
to-one’’.
 Association: The possibility to ‘‘take existing objects into the
virtual space and work with them’’.
 Formability: The possibility to manually warp virtual models and
sketches which in turn allows for a gradual developing and
testing of ideas until they are ‘‘mature’’.
While many of the participants hold positive emotion towards
the virtual design environment, they still regard sketching and CAD
irreplaceable in their design activities. Perkunder et al. [135]
addressed the problem by considering automatic shape creation
using sketches as input to an immersive CAD environment in the
early phases of product design. The shape creation and deforma-
tion algorithms of FiberMesh were made available in a 3D
environment and the integrated tool was evaluated in order to
investigate whether this approach supports the conceptual design.
Fig. 14. Schematic of CAD and VR sketch tool [166].
G Model
CIRP-1406; No. of Pages 23
Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
An immersive modelling environment has been created by
merging sketching tools with CAD and VR technologies. A designer,
using such tools, can sketch out a concept in the form of a picture,
alter the perspectives and shapes, and finally arrive at a product
represented by engineering drawings [166]. The schematics of the
online integration of a CAD system with a VR-based sketching tool
are shown in Fig. 14. This process of design iteration is akin to the
imaginal thinking frame.
3.3.2.3.2. Human–machine interface: AR view. Augmented Reality is
an emerging technology that merges real and virtual objects in the
same environment. AR systems are highly interactive and an AR 3D
design space will be able to complement current 3D CAD systems.
However, there are only a few systems supporting product
modelling and modification either in the 2D or 3D environments.
Construct 3D [92], for instance, is good at constructing 2D and 3D
primitives, but lacks the capability of developing complex parts
containing these primitives. Thus, Park [133] proposed an AR-
based re-formable mock-up system, in which the shape, colour and
texture can be modified by the Portable Media Players. A gesture-
based AR design environment called GARDE [127] was developed
by Ng et al. In this system, designers are able to make the
modification, visualize the 3D model and evaluate the design
solution both in the model and even in conventional CAD software.
A developed system, named Augmented Reality Computer-Aided
Design Environment (ARCADE) [126], is an AR design space which
allows the users to create the functional models and 3D models,
and to evaluate the functional behaviour and ergonomics of the
design concept. An intuitive method for generating 3D models
using bare-hand interactions has been developed. The user can
create 3D models using a building block approach that is similar to
playing with virtual LEGOs, along with the extrusion approach
which is similar to the creation of 3D models with conventional
CAD systems.
By comparing VR/AR based interfaces with CAD and sketching
from the perspectives of supporting image-based representation, it
can be inferred that the virtual interface performs even better than
the others that have already been widely applied [166]. The
promising features are as follows:
(1) Three-dimensionality: Immersive VR/AR interfaces allow the
designers to design in the work space where product modelling
is in one-to-one scale [166]. Since the three-dimensional
visualization is how human represents objects mentally [64],
such interfaces provide great convenience for designers to
recall past knowledge and experience in the memory [103],
thus to reducing mental transformation burden.
(2) Interactivity: Comparing with CAD and paper-based sketching,
VR and AR are both highly interactive technologies which
provide high-bandwidth information exchange [166].
(3) Digitality: Different from traditional paper-and-pencil sketch-
ing, digital media allows for a direct integration of existing
models or prototypes [166].
Consequently, VR/AR based human–machine interface is so far
the best tool to achieve smooth communication among designers
and computers.
3.4. Epistemologies of automating design
As mentioned earlier, the introduction of epistemology into
design realm serves the purpose of explicitly formalizing the
information flow lying in the automating design process. It follows
from the previous sections that ontology equips automating design
with the foundation for human–machine mutual understanding
and collaboration between teams, by formulating a static
framework of taxonomy. However, in order to achieve the true
harmony between human and machine and maximize the machine
intelligence so that it can be move forward to the early stage of
design [196], it is particularly crucial to understand the cognitive
nature of automating design. Consequently, epistemology is very
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
15
helpful to share unambiguously and unbiasedly the understanding
of the design goals of various phases and levels, facilitating the
collaboration and ideation exchanges among teams.
The sections below are arranged as follows: First, the studies on
epistemology and the related rationale of the automating design
framework derived from the research of Pahl and Beitz [131] are
discussed, especially the four design stages. Next, an emerging
trend in conceptual design—the analogical design along with the
representative biologically inspired (bio-inspired) design is intro-
duced. In particular, as a key topic in automating design, the
capture (or search) of knowledge is discussed, taking bio-inspired
design as an example. Finally, the rationale and the corresponding
supporting tools of coupled design and coupled optimization from
the epistemological perspective are focused.
3.4.1. Rationale of cognition in automating design
The greatest merit of epistemology for promoting the develop-
ment of automating design is its descriptive power of the cognitive
mechanism [158] of humans during design ideation, which is
supposed to be mimicked by machine intelligence for the ideal
situation where the design is fully automated. As clarified by Goel
et al. [66], the future trend of CAD systems should be based on the
grounding in design cognition. Hence, it can be inferred that the
future of automating design would require the maturity of
‘‘machine cognition’’, which possibility is still controversial, the
paradigm of ‘‘machine learning’’ can be considered as a prototype.
Despite the difficulties, the first step is to make clear the human
cognitive process in design, and even for this topic, there remains
many unsolved issues [18]. Therefore, the information flow that
plays the dominating role in the epistemologies of the four
axiomatic design phases are elucidated first.
3.4.1.1. Planning and clarification. Planning and clarification refers
to the conversion from requirements to functions. In automating
design, this process depends substantially on human knowledge
and experience, because up till present, computer’s logical
architecture is still not able to perform the planning job
independently or identify correctly the requirements posed by
market and customers. Thereby this stage is currently the one that
fully requires human knowledge and experience. The knowledge
matrix has been proposed and it demonstrates well the
epistemological process involved, and can be interfered directly
by designers when the requirements change. A mapping can be
formulated and simplified as r = R  f, where r denotes the
requirement, f denotes the functions, and R denotes the knowledge
matrix [207].
At the beginning of the design, it is essential to set up a
requirement list acting as a guideline during design and a standard
for evaluation. Designers also have to compile the requirements
into technically and economically attainable ones, which is also the
prerequisite for conceptual design [131].
Since product development is becoming more complicated, it is
almost unrealistic to maintain the initial customer requirements
throughout the product development life cycle. To control the
manufacturing quality according to the required product design
quality, QFD was proposed [120]. Although it requires much effort
to perform comprehensively a whole QFD process, its effect has
been recognized in numerous industrial applications, particularly
automotive industry where there is little chance to modify the
product configuration during long range product development
[120] if the customer requirements are not captured satisfactorily
at the beginning. According to model-driven engineering, one
defect of Gero’s FBS model is that it does not include a requirement
phase and that the documentation phase becomes a secondary one
compared to the models [34]. This has been improved by the
previously introduced RFBS framework in which the requirement
models contain the performance criteria of the product and its
service functions [34].
The activity specification is a critical enabler for collaborative
product development through an extended enterprise [149]. Often
G Model
CIRP-1406; No. of Pages 23
16 Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
this activity involves numerous iterations before a shared
understanding of the product requirements since various omis-
sions, ambiguities and assumptions have to be resolved. Ontol-
ogies for the requirement management are developed in product,
information and process levels. The three levels are required to
represent the product requirements, the documents used in
communication and the business process for the changing
management [149].
3.4.1.2. Conceptual design. Conceptual design is recognized as the
most significant and creativity-requiring phase [81] that has the
dominating influence on the product quality. If any fault occurs in
this stage, the modifications in the later design processes would
cause much unnecessary cost [71]. Conceptual design mainly aims
at determining the principal solution, which can be achieved by
abstracting the essential problems, establishing function config-
urations, searching for suitable working principles and the
combination of these principles into a working configuration.
The results can take diverse forms, such as a function configura-
tion, a circuit diagram or a flow chart [131].
At present, conceptual design relies extensively on human
knowledge and experience, especially in creative design where
human’s intuitive thinking plays a central role [54]. Thus, for the
design to be automated, the machines or computers have to be
capable of mimicking human’s ideation or problem solving
process. This makes it inevitable to investigate the cognitive
mechanism or models [81] involved in the design ideation of
human, which is also exactly the focus of epistemological issues. As
a matter of fact, conceptual design can be regarded as a typical
Human Problem Solving (HPS) process, the theory of which was
proposed by Newell and Simon [125]. Nowadays, the Short Term
Memory (STM)/Long Term Memory (LTM) model for design
ideation has been widely accepted. Fig. 15 shows its explanation
towards human’s problem encoding and externalization process.
The biological line defines the physical perceptive boundary
(visual, acoustic, haptic, etc.), and the external knowledge in the
environment includes the knowledge stored in databases, books,
internet, etc. Crossing the biological line requires the externaliza-
tion of memories. The STM and LTM are separated by the
knowledge line, crossing which involves the information retrieval
and storage process. The core of the whole model lies in the
internal symbolic structure, or the memory representation, which
can be in the form of a semantic network. Based on this schema,
other cognitive models focused on memory storage, memory
retrieval, problem assimilation/solving [81], etc. have been
successively proposed. In particular, the example-exposure
cognitive model provides the preliminary theoretical basis for
analogical design and invention, and memory retrieval is closely
related to the knowledge capture/search process, which will be
discussed in detail in the next section.
Interestingly, to some extent, the current computer architec-
ture (the von Neumann architecture) is very analogous to the STM/
LTM model. For example, the perceptive receptor, central nervous
system, STM and LTM in human cognitive model correspond to the
input device/sensor, CPU, RAM and hard disk in a machine,
respectively. The largest difference between human and machine
comes from the fact that all these functional components are
Fig. 15. Cognitive model of encoding and externalization process, modified from
[81].
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
universally based on neurons for human but are separate for
machine. Moreover, the internal symbolic structure is apparently
absent in the machine’s RAM, which is also the biggest impedance
for machines to generate ‘‘intuition’’ as well as making proper
decisions during conceptual design.
Although machines are still far from replacing humans at the
present automating design level, the similarity between human
cognitive models and machine architecture has always been
inspiring the pioneering attempts. The first step to automate
conceptual design might be the decomposition of a problem, in
which the original complex problem is broken into smaller
‘‘chunks’’, thus the overall problem can be simplified and solved
concurrently. The use of Design Structure Matrix (DSM) [50] is
considered as ‘‘facilitating intelligent system decomposition and
integration analysis’’. Suh’s axiomatic design [169] is helpful to
problem decomposition, and it divides the design process into four
domains [169]. The axiomatic design theory has been used in a
wide range of industrial applications ranging from software design
to product and manufacturing systems design [170,171]. The
problem decomposition effort has led to the conception of
modularity, or modular design [72], in which the modules and
their hierarchical topologies can be expressed by the ontologies to
ease the information processing and configuration exchanges of
machine.
Due to the fact that the participation of human is indispensable
during conceptual design, the effective human–machine integra-
tion becomes pivotal. Actually, referring to the above cognitive
models, machine can act as the environment, external memory, or
even external knowledge during the integration, while the
significance of human is reflected by the STM, or the internal
symbolic structure that plays a key role in decision making. Hence,
intelligent design tools and knowledge management systems have
been developed aiming at maximizing human advantage. Yoshi-
kawa [211] proposed that an intelligent design system must cover
the tasks of process planning, tool selection and fixtures, and the
determination of cutting conditions, etc. Furthermore, the system
must be good at recognizing and understanding the results of the
design so as to translate the geometrical data into instructions.
Accordingly, intelligent CAD system [142,211] has been proposed
for providing an ideal environment that facilitates the designer’s
decision making. In addition, it is worth noting that sketching is
particularly preferred by the designers at conceptual design stage
[167], and it is recognized that freehand sketching is helpful to
creative ideation. Thus, sketching support has become a future
trend for CAD systems to make machines better serve as the
extended memory of humans. With the continuous efforts of the
researchers, one can anticipate that one day machines would be
equipped by their own internal symbolic structure and eventually
be able to make correct decisions independently, fully automating
the conceptual design phase.
3.4.1.2.1. Analogical design. In human design activities, analogical
reasoning is ubiquitous, i.e., when confronted with a new design
problem, the designers tend to search the experience or knowledge
repository for similar problems that have existing solutions, the
reformulation of which is expected to fit the new problem. One
may find it difficult to give a precise definition of analogical design,
but the related tasks can be clearly codified, as initially concluded
by Goel [188]: analogical design involves reminding and transfer of
knowledge about one design situation to another, and the
transferred content can be the knowledge of a design problem,
solution, domain or strategy, etc. According to Newell and Simon’s
theory [125], this process is referred to as ‘‘problem space search.’’
Even for creative design, the use of analogies is regarded as playing
a key role [188]. For automating design, especially at the
conceptual design phase, the machines are expected to be capable
of mimicking this analogical reasoning process of human, as
interpreted by the example-exposure cognitive model. Such type
of design ideation induced the branch of analogical design, which
has also been recognized as the emerging trend in future CAD
systems.
G Model
CIRP-1406; No. of Pages 23
Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
Presently, the epistemological principle or mechanism under-
neath the analogical design process is still unclear. Fortunately, AI
design researches have made extensive characterizations about
analogical design, for which Goel [65] identified four main issues:
(1) why, i.e., the task or goal for which analogical design is applied;
(2) what, i.e., the knowledge content to be transferred; (3) how, i.e.,
the methods involved in the reminding or transfer process; (4)
when, i.e., the strategic control of design process. Obviously, these
issues are tightly related to one another. To answer these
questions, case-based reasoning [25] and model-based reasoning
have been proposed. Furthermore, the analogical reasoning studies
in both AI and psychology or cognitive sciences suggest the generic
abstraction being the centric issue. Viewing from the teleological
perspective, several AI theories can be employed as the theoretical
framework for the machines to realize automating analogical
reasoning, such as SYN [16] (generic abstractions of analogical
design), DSSUA [137] (design prototypes as generic abstractions)
and IDEAL [13] (design patterns as generic abstractions).
As mentioned earlier, in automating design, the three core
problems in knowledge management is knowledge capture,
representation and rationale, while for analogical design, the
dominate process is the analogy or knowledge search. This can be
well demonstrated by the most typical analogical design–bio-
inspired design, which is particularly advantageous in highly
coupled and complex design problems, as discussed in detail
below.
3.4.1.2.2. Bio-inspired design. A satisfactory definition of bio-
inspired design was provided by Shu et al. [161] as the studies
of formation, structure, function, material or mechanism of
biological substance especially for the purpose of artificially
synthesizing similar products. Although bio-inspired design is a
typical example of analogical design, there exists a subtle
difference between them. According to Vattam et al. [188], bio-
inspired design is a rather ‘‘compound’’ type of analogical design
that generates compound solutions. Put in simple words, it is
implemented by the mapping of similarities between two distinct
knowledge domains, i.e., biology and engineering. Designers find
that bio-inspired design is especially powerful in designing highly
complex, versatile systems and solving coupled problems or
dilemmas, and its process can be divided mainly into two parts:
biological system identification and engineering system realization.
Similar with analogical design, analogy-based design theories
have been applied to bio-inspired design, analogical reasoning
methods such as analogy transfer and problem decomposition
have also been proposed [188]. As a developing subfield of design,
its cognitive principium is still under investigation, however, a
consensus has been reached that the key problem in bio-inspired
design is the correct selection of bio-analogy, as well as its proper
utilization, which directly determines the final quality of the
product, since it is basically an ideation method embodied in the
conceptual design phase. To achieve the unbiased understanding
of the bio-analogy, many methods have been proposed in recent
years. A database as well as a social network called Asknature [44]
was developed to provide the engineers with nature’s solutions
and the access to biologists, while the risks are that biologists may
favour the solutions out of their own expertise. A heuristic search
method was proposed by Helms et al. [80] to maximize the
efficiency and sufficiency of the search. A database-like tool named
BioTRIZ [41] is capable of transferring knowledge from biology to
engineering, and a conflict-pattern problem defining paradigm is
suggested. Most worth mentioning is the study conducted by Shu
[160], who proposed the natural-language based search method to
improve the quality of search results and the corresponding
strategies for the bio-analogy assessment. Actually, most of the
existing methods are based on database technology, search engine
technology, knowledge management platform for collaborative
design (e.g., consulting biologists online to avoid cognitive bias),
and semantic web technology.
The search of bio-analogy is a typical knowledge capture
process. Systematic studies have been conducted by Shu et al., and
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
17
they suggested that verbs should be the best choice as search
keywords in order to retain most potential analogy candidates. Of
course, adjectives are also acceptable as they directly describe the
attribute of the target analogy, while the use of nouns is
particularly discouraged as it would cause preconceived ideas
when specifying physical entities. They also propose that key-
words can be extended using the tool WordNet to expand the
search range. As a matter of fact, combining adjective-type and
verb-type lemmas would be beneficial because there are cases
when an adjective is hard to map onto verbs, while sometimes
adjectives against one function are not as abundant as verbs. In this
situation, the combination takes a complementary effect. More-
over, the requirements of being common and engineering-oriented
allow the lemmas to cover engineering terminologies, improving
the usability, efficiency and accuracy of the search.
As an emerging trend in the conceptual design of highly
complex and multi-objective products, bio-inspired design is
drawing increasingly more attention. However, as can be
imagined, the difficulty for automating bio-inspired design is
enormous, because understanding a bio-analogy correctly is even
harder for a designer, not to say a machine. Thus, during bio-
inspired design, an efficient human–machine integrated interface
and multi-disciplinary collaboration becomes more significant.
3.4.1.3. Embodiment design. The primary task in the embodiment
design phase is to further determine the physical structure, or
entrust the system with corresponding physical properties. For
example, mass, stiffness, damping, heat durability, conductivity,
etc. related to the material should be specified. Hence, the final
results are the specification of a layout [131], which involves the
transformation from concept into entity, or from abstract images
into representative ones. Embodiment design is roughly composed
of two aspects, the design and optimization of one part and that of
the interfaces among parts.
Automating the embodiment design phase is a highly
challenging task because numerous cross-domain factors need
to be considered simultaneously. To ease the autonomous
processing of computers, modular design, which turns the
independent sub-functions into disparate modules, has provided
a good operation platform for machine intelligence. Three
advantages exist when applying modularity to embodiment
design, first, various modules greatly narrow the gap between
human and machine, facilitating human–machine integration,
second, separate modules are much easier for machines to
understand, third, concurrent engineering can be achieved easily
since the clustering of different modules is decoupled and can be
designed simultaneously.
Modular design has been studied extensively, and various
methods have been proposed [82,131,151]. Pahl and Beitz [131]
pointed out the importance of modular design practice, and
proposed product modularity should be frequently stated as a goal
of good design practice.
Ulrich [98,99] developed the methods of using functional
structures for modularization by grouping the function-related
elements together, and stated that a modular product or
subassembly has a one-to-one mapping from the functional
structure to the physical components of the product. He further
argued that all the interfaces among the components are
decoupled. Yet, this definition accounts only for the functional
aspects and ignores all the other life-cycle features. It also
disallows multi-functional modules, while it does introduce an
important view that the manner in which the modules are
compiled together (interface) play a key role [62].
Erixon et al. [51] proposed the concept of ‘‘module drivers’’,
which are a number of different criteria behind modularization,
throughout the entire product life cycle, including product
development, variance, manufacturing, quality, purchasing and
after sales. ‘‘Module drivers’’ enable the setup of independent
assembly units within each module that can be adapted precisely
to the requirements of that module.
G Model
CIRP-1406; No. of Pages 23
18 Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
As can be inferred independence, more specifically, functional
independence, dominates the modularization process. The theo-
retical basis of modularity originates from Suh’s [13,171]
independence axioms. Therefore, if possible, each function that
a product performs should be independent of all other functions,
realizing a one-to-one correlation between form and function. This
axiom has led to a search for a connection between physical
independence and functional independence [88,148].
Besides, Hillstrom [82] applied Suh’s axioms to evaluate the
interfaces among modules. Mistree et al. [119] demonstrated with
an example that there is a relationship between the independence
among modules and reducing interactions between them. Kusiak
and Huang [100] proposed an approach for modularizing a product
by considering the costs and performance of a product [73].
Eppinger and Pimmler [49] used a heuristic swapping algorithm to
cluster components into modules. Newcomb et al. [122] used the
clustering algorithm developed by Kusiak and Chow [98] to
identify modules in a product. Moreover, Kusiak and co-workers
have developed several clustering algorithms, which can be used
for component clustering [73,99,101,102].
Despite of all these achievements, the coming trend of
mechatronic product design has been transferring from the
independence notion into the desire of highly compact, or
integrated product with good versatility. Thereby, sophisticated
structure is required during embodiment design, as illustrated in
bio-inspired design. To follow this need, the functions within a
structure have to become more or less coupled, and the number of
the ‘‘interfaces’’ as presented in modular design would greatly
decrease to enhance robustness. Then, the modular design method
might only be suitable for providing a guidance of clustering, i.e.,
what functions are better to be integrated into one structure.
Currently, it becomes urgent to establish a novel coupled design
and optimization methodology, which inevitably requires the
intervention of human intelligence due to technology limitations
at present.
3.4.1.4. Detailed design. This is the phase of the design process in
which the arrangement, forms, dimensions and surface properties
of all of the individual parts are finally laid down. After
embodiment design where the materials are specified, now
production possibilities need to be assessed, costs should be
estimated, and all the drawings and other production documents
are validated [131]. Particularly, this is the stage where all the
information and specification of the product including the results
of geometric optimization, electromechanical coupled optimiza-
tion, as well as the improvement of assemblies, components and
costs [131], etc. are holistically verified.
3.4.2. Coupled design and coupled optimization
As the prototype of automating design, Gero’s FBS model treats
design as the transformation between different states. Originating
from this prototype, the most recent framework in Fig. 4 shares
similar characteristics, i.e., the design process is modelled as
sequential and iterative activities. However, with the increasingly
deeper understanding of the nature of automating design, the
researchers find that almost no modern product design can be
accomplished in such strict step-by-step manner, which actually
only serves as a conceptual or principle description. In practical
designs, the design stages or transitions among states are usually
entangled, and this can be seen from three perspectives: (1) as
mentioned earlier, automating design can be regarded as
knowledge operation, while different domain knowledge entities
are certainly interconnected; (2) from the ontological view, both
design knowledge and product knowledge can be represented by
global ontologies consisting of local ones, and the corresponding
semantic-network shows that the concept entities often have
extensive semantic relations with one another; (3) for any design,
there must exist versatile parts. Moreover, bio-inspired design is
recognized as a future trend for the reason that there is growing
need for highly compact, integrated, and multi-functional
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
Fig. 16. Epistemology process of automating design.
products. Consequently, absolute functional independence [169]
as proposed in the axiomatic design theory is difficult to achieve.
On the contrary, in most cases, design ideation is a coupled process,
e.g., the details can be considered during the conceptual design
phase, and such ‘‘idea jumps’’ might be common. The previously
introduced analogical design and bio-inspired design are actually
excellent examples as well as effective means for coupled design.
As shown in Fig. 16, the design stages (R, F, C, S, D) in sequential
descriptions are analogous to the local ontologies, while with
coupled concept, C, S, and D can be generally denoted as
‘‘morphology’’. Thus, in coupled design, the whole design process
is actually operated in a global ontology, and it has been
implemented by designers consciously or unconsciously. For
automating design, coupled ideation process is typically a highly
challenging task for computers, which as described previously
have to obey the sequential procedures as depicted in Fig. 18. This
again stresses the importance of human–machine integration.
Symbiotic with coupled design is the concept of coupled
optimization, which is a similar concept with multi-objective
optimization (MOO). The principal goal of MOO is to model a
decision maker’s preference [113], and in usual cases, the solution
of MOO is not very definite but more like a concept. Traditional
nonlinear MOO methods like the ones with priori, posteriori or no
articulation of preferences are essentially based on the single
objective optimization engine when implemented, while the
intelligent methods, such as genetic algorithm [95] or simulated
annealing algorithm are global and can be used to solve MOO
problems directly. Nevertheless, all these methods need a
prerequisite of mathematical expression or quantification of the
problem. Unfortunately, a designer will find it impossible to
always ‘‘measure’’ a problem precisely, e.g., in multi-disciplinary
optimization the fundamental theories are even more diverse.
For coupled optimization, quite often the qualitative and
quantitative optimizations come in a mixed manner. For example,
the design of a complex product would involve the determination
of optimal functional grouping, selection of optimal material and
principal mechanism, and optimal cybernetic configuration, while
within each category usual MOO has to be implemented. For
automating design, decoupling of the qualitative dilemmas has
always been a challenge for machine. The proposal of modular
design [72] introduced earlier greatly contributed to the develop-
ment of automated coupled optimization of machine, however,
currently it depends extensively on human–machine integration.
From the perspective of design science, three types of general
optimizations need to be particularly considered in kinematic,
dynamical and geometric levels, respectively. These optimizations
are conducted with the continuous comparisons between the
expected behaviour and the physical behaviour of the design, as
shown in Fig. 4.
(a) Kinematic optimization. Kinematic optimization is preceded
among configuration, expected behaviour and physical behav-
iour (Fig. 4). In this process, kinematic characteristics of the
potential topological structures are analysed, i.e., it is at the
topological level, thus more qualitative optimizations are
involved, requiring more human labour. If the kinematic
requirements fail, a designer has to loop back into the conceptual
design stage or even modify the functional specification.
G Model
CIRP-1406; No. of Pages 23
Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
(b) Dynamical optimization. Dynamical optimization is preceded
among structure, expected behaviour and physical behaviour.
At this stage, physical properties (like material, mass) have
already been assigned to the system, so it is possible for
computers to become increasingly involved, while the
designers’ job leans towards the ordinary MOO, leaving the
computer to perform automatic dynamic computation with the
established knowledge base mentioned earlier [4,23,143].
(c) Geometric optimization. Geometric optimization is preceded
among details, expected behaviour and physical behaviour.
At this phase, detailed geometric parameters of the design would
be specified. Thus this type of optimization is highly mathemat-
ical and programmable. The powerful computational capability
of a machine is best embodied at this stage, since the tasks left
are usually pure MOO or even single objective optimization.
4. Implementation
4.1. Integrated human–machine interface for designing turbine
engines and aircrafts
Design of turbine engines and aircrafts is very complex, because
optimal design solutions need to be attained over multi-disciplinary
domains, for example, the pipe routing problems for aircraft engine
design [208,209]. Thus, it calls for comprehensive and collaborative
engineering decision making across disciplinary boundaries. There-
fore, a unified human–machine integrated design environment is
necessary to provide engineers with a systematic spectrum of
techniques for flexibly concerting inter-disciplinary interactions,
settling discipline conflicts, and reaching at global optimums. In
order to maximize the strengths of both human and computer, an
image-based integrated design interface should be developed.
According to the imaginal thinking frame, visualization is
essential to help designers gain insightful understanding of
numerous physical behaviours, such as heat transfer, mass
advection, fluid turbulence, structural deformation and wave
propagation. Design visualization constitutes a rudimentary com-
ponent of an image-based human–machine integrated environ-
ment, which enables human to intuitively understand the
burgeoning amount of data resulting from science, engineering,
business, environment and other domains. As shown in Fig. 17,
design visualization processes data in two ways [193]. The first way
is to display data in elaborately structured tables, charts and
diagrams in order to manage the process systematically and
efficiently, which can be divided into tasks and workflow manage-
ment. The second way is to display data in graphics or pictures with
standardized symbol systems and semantic relations. In this way,
discrete data on two-dimensional (2D) or three-dimensional (3D)
domains are illustrated by continuous representations that
Fig. 17. Image-based human–machine integrated design interface for aeroengine
design.
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
19
graphically depict the states of the concerned physical systems,
which are often modelled using partial differential equations (PDE).
The image-based representation or visualization of design
process provides great help to increase the efficiency of the design
of turbine engines and aircrafts, which is greatly influenced by the
coherence of collaborators. As shown in Fig. 17, the design
collaboration system is mainly composed of project management
and workflow management [194]. The former is the core
competency that configures engineering sources in accord with
each product development project, and supports collaboration
among team members. The later allows various applications to be
seamlessly deployed and run in a continuous sequence without in-
process intervention from engineers. Image-based collaborative
design tools for project and workflow management make better
utilization of engineering resources, enforce effective planning and
coordination of myriad engineering tasks, keep close track of
engineering processes, and provide engineers with up-to-date
information.
Physical data of turbine engines or aircrafts generated by PDE
solvers are bound with discrete nodes, vertices of meshed domains
or scattered spatial points. Unstructured data can be transformed
into structured data through meshing and mapping to reduce
computation time and cost [194]. As shown in the upper part of
Fig. 17, although there are a large variety of physical fields, the
visualization of physical systems can be divided into the graphical
illustration of scalar, vector and tensor fields through the image-
based interface. The field visualization techniques are implemen-
ted on the basis of a range of linear and non-linear meshes, which
flexibly adapt to disparate data representation schemas adopted
by all kinds of multi-disciplinary problem solvers (e.g., ANSYS,
PATRAN, etc.). Both vector and tensor field visualization methods
are subsequently reduced into scalar field visualization, which
depicts a spatial distribution of an undirected physical quantity.
The field visualization methods joint discrete outcomes of PDE
solvers into continuous images that permit intuitive observations
and insightful cognitions. In this way, humans can make effective
decisions based on their superior capability in imaginal thinking.
As both of the process and physical systems for the design of
turbine engines and aircrafts are depicted as general images, the
merits of human experience, intuition and creativity can be
maximized for design evaluation and more effective decision
making, which largely dominates the success of a design. It can be
seen that the visualization method of turbine engine and aircraft
design follows the human–machine interactive principle of
automating design, demonstrating the effectiveness of the
proposed automating design methodology.
4.2. Ultra-precision grinding machine
An implementation of the human–machine integrated auto-
mating design paradigm is presented for the design of an ultra-
precision grinding machine. As shown in Fig. 18, the whole design
framework can be divided into three sections: ontology-based
knowledge management, virtual machine tool and VR/AR-based
human–machine integrated environment. The detailed informa-
tion flow of human–machine integrated multi-agent design
process is also shown, and human intelligence and machine
intelligence are treated as equal on the same platform. Two
important features arise from the VR/AR-based human–machine
integration: time-saving, and higher design efficacy and creativity.
The ultra-precision grinding machine is composed of several
sub-systems, such as bearings, spindle, drivers, etc., each of which
has its own knowledge structure and contents. It is almost
impossible to construct a uniform knowledge model in the system
level directly. Therefore, meta-ontology, as a more fundamental
level is used to establish a mutual basis for the integration and
sharing of different domain knowledge. As introduced earlier,
three classes of ontologies can be derived from the global ontology
to support the automating design of the ultra-precision grinding
machine. Product ontology is the knowledge base for product
G Model
CIRP-1406; No. of Pages 23
20 Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
Fig. 18. Design framework of a large ultra-precision grinding machine.
modelling and optimization, as well as the visualization of the
machine tool. Meanwhile, process ontology and resource ontology
are utilized to support the epistemology of design. On the other
hand, virtual machine tools are used in the design process because it
can dramatically reduce the cost and time of hardware testing and
iterative improvements of the physical prototypes [3]. Initial
requirements modelled by requirement ontology are clarified in
detail and put into the computer by the designers. The computer
integrating Suh’s axiomatic design [169] principles is able to conduct
mapping from requirement ontology to function ontology, which are
visualized as the basic motion patterns required for finishing all the
machining jobs. Meanwhile, non-functional requirements are
expressed as expected behaviours for evaluating design results in
the latter design phases. After that, configuration of the machine tool
is determined through conceptual design. Two sets of motion types
for tools and workpieces can be described by the mathematical tools
of Lie subgroup or submanifold that are suitable for expressing the
motions of rigid bodies. Furthermore, quotient kinematics, which
specifies topological relationships between a Lie subgroup or
submanifold and their quotient spaces, is found to be a promising
tool to enable the computers to solve the aforementioned
topological synthesis problem. More than ten topological structures
are determined as the potential solutions after initial optimization.
In order to promote the stiffness of the machine tool, an intuitive
configuration which decomposes the five-axis topology into the
three-axis topology is adopted to substantially guarantee dynamic
stiffness. In addition, the two-DOF virtual motion of the toric wheel
with an inclined spindle has been determined to fulfil the ultra-
precision requirement. Subsequently, the 3D CAD ‘‘virtual proto-
type’’ model of the machine tool is created, and various optimization
methods and simulation software are applied to refine the virtual
prototype in order to realize the expected behaviour, including
stiffness, stability and vibration mode, etc., through the coupled
optimization in terms of kinematic, dynamic and geometric
perspectives. Deduced from the product ontology, the CAD and
CAE models can be bilaterally transformed according to the protocol
of STEP. The multi-body simulation, for example, provides an easy
way to analyse the kinematic behaviour over the complete
workspace of a machine tool as well as to determine load histories
of components or joints [84,156,198]. Eventually, Finite-Element-
Analysis (FEA) is applied to define the dimensions of kinematics and
optimize the structural behaviour [84]. For any design, there exist
several key coupled aspects. For complex mechatronic product, the
two most significant optimizations, mechanical and electrical
optimizations are coupled, because the performance of a system
is co-determined by both mechanical structures and cybernetic
strategies. For a large ultra-precision grinding machine, the most
demanding part might be the design of the spindle box. Aiming at
this issue, Wang et al. [195] had provided a very enlightening
method for electromechanical coupled optimization, as shown in
the middle part of Fig. 18. The optimization was implemented by
building a global dynamical model encompassing mechanical and
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
electrical systems based on ontology, and the model was uniformly
expressed with state-space analysis.
After the refinement and optimization, the remaining task is to
select the optimal structures. The design of the ultra-precision
grinding machine is more of a multi-agent collaboration than an
individual work. Therefore, COM/DCOM technologies are applied
to keep efficient information exchange among the optimization
processes, meanwhile supporting real-time transmission of design
decisions and associate modifications. Consequently, to achieve a
seamless human–machine interface, the system is built under
collaborative VR/AR environments. TransVision [144], remote
user’s appearance [15], synchronized turntables [150] and
SeamlessDesign [94], etc. are some mature VR/AR systems.
With the proposed human–machine integrated automating
design paradigm, the complete design and optimization of the
ultra-precision grinding machine with a 1.2-m production capacity
takes only nine months to finish. The total mass of the machine is
limited to 12 t by utilising granite. The overall dimensions are
determined to be 2800 mm  2600 mm  2400 mm. The grinding
spindle is designated at a fixed 20̊ angle with a 320 mm diameter
toric grinding wheel. Optimization results show that maximal
static deformation of the tool centre point is limited to 6.5 mm and
that the first mode frequency can achieve over 108 Hz.
5. Conclusion and prospect
This paper is an attempt to present a comprehensive overview
on the studies related to human–machine integrated automating
design paradigm, which aims at combining human’s creativity and
machine’s powerful computational capability. Ultimately, the goal
is to alleviate human from traditional design burden. Despite the
fact that modern design cannot be implemented without
computers, if the lead time from function to configuration
(conceptual design) can be effectively shortened with sophisticat-
ed human–machine integration, faster decision, lower cost and
shorter development period can be accomplished.
As a matter of fact, the conclusions made so far are not only the
abstraction of the ideas presented in this review, but also hold
potential in guiding the implementation of modern automating
design, enabling manufacturing industries to respond quickly to
dynamic market changes, as well as providing new inspirations to
the studies in this field. Although promising achievements have
been made, there remain plenty of challenges, which are also
opportunities mainly coming from three aspects:
(1) Cognitive science. The decoding of cognitive process and design
ideation mechanism of human has become a significant trend
in the design realm.
(2) Computer science. The biggest impedance to get machine
involved in conceptual design might be the prevailing von
Neumann’s serial machine architecture [207,209]. It can be
expected that new architecture computers and new machine
languages will become a mainstream in the near future.
(3) AI technology and supporting tools. New AI theories and
intelligent algorithms will become new research highlights,
also acting as the technical basis for future automating design.
Acknowledgements
The authors want to explicitly acknowledge the input received
from numerous CIRP colleagues. The authors want to acknowledge
the support of the National Basic Research Program of China (grant
no. 2011CB013203).
References
[1] Ahmed S, Kim S, Wallace KM (2007) A Methodology for Creating Ontologies
for Engineering Design. Journal of Computing and Information Science in
Engineering 7(2):132–140.
[2] Altintas Y (2000) Manufacturing Automation, Cambridge University Press,
Cambridge.
G Model
CIRP-1406; No. of Pages 23
Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
[3] Altintas Y, Brecher C, Weck M, Witt S (2005) Virtual Machine Tool. CIRP
Annals—Manufacturing Technology 54(2):651–674.
[4] Altintas Y, Cao Y (2005) Virtual Design and Optimization of Machine Tool
Spindles. CIRP Annals—Manufacturing Technology 54(1):379–382.
[5] Altuna A, Cabrerizo A, Laresgoiti I (2004) Co-Operative and Distributed
Configuration. Object-Oriented and Internet-Based Technologies, Springer, Ber-
lin/Heidelberg69–80.
[6] Ando K, Yoshikawa H (1989) Generation of Manufacturing Information in
intelligent CAD. CIRP Annals—Manufacturing Technology 38(1):133–136.
[7] Atzori L, Iera A, Morabito G (2010) The Internet of Things: A Survey. Computer
Networks 54(15):2787–2805.
[8] Bach C (1920) Die Maschinen Elemente, Arnold Bergstrasser Verlagsbuch
handlung, Stuttgart. (1. Aufl 1880, 12. Aufl. 1920).
[9] Bernard A, Tichkiewitch S (2008) Methods and Tools for Effective Knowledge
Life-Cycle-Management, Springer, Berlin.
[10] Berners-Lee T (2000) Semantic Web on XML.
[11] Berners-Lee T, Hall W, Hendler JA (2006) A Framework for Web Science.
Foundations and trends in Web Science 1(1):1–130.
[12] Berners-Lee T, Hendler J, Lassila O (2001) The Semantic Web. Scientific
American 284(5):28–37.
[13] Bhatta SR (1995) Model-Based Analogy in Innovative Device Design, College of
Computing, Georgia Institute of Technology, Georgia. (Ph.D. Dissertation).
[14] Bilda Z, Demirkan H (2003) An Insight on Designers’ Sketching Activities in
Traditional Versus Digital Media. Design Studies 24(1):27–50.
[15] Billinghurst M, Kato H (2000) Out and about—Real World Teleconferencing.
BT Technology Journal 18(1):80–82.
[16] Börner K, Pippig E, Tammer EC, Coulon C (1996) Structural Similarity and
Adaptation. Advances in Case-Based Reasoning, Springer, Berlin/Heidel-
berg58–75.
[17] Borzo J (2004) Get the Picture: In the Age of Information Overload, Visuali-
zation Software Promises to Cut through the Clutter. The Wall Street Journal
;(January (12))R4.
[18] Braisby N, Gellatly A (2012) Cognitive Psychology, Oxford University Press,
New York.
[19] Brooks RA (1991) New Approaches to Robotics. Science 253(5025):1227–
1232.
[20] Brooks RA (2001) The Relationship between Matter and Life. Nature
409(5818):409–411.
[21] Burks A (1994) The Collected Papers of Charles Sanders Peirce, vols. VII–VIII.
Harvard University Press, Cambridge, MA.
[22] Buxton B (2010) Sketching User Experiences: Getting the Design Right and the
Right Design: Getting the Design Right and the Right Design, Morgan Kaufmann,
Burlington, MA.
[23] Cao Y, Altintas Y (2004) A General Method for the Modelling of Spindle-
Bearing Systems. Journal of Mechanical Design 126(6):1089–1104.
[24] Cao Y, Liu Y, Fan H, Fan B (2013) SysML-Based Uniform Behaviour Modeling
and Automated Mapping of Design and Simulation Model for Complex
Mechatronics. Computer-Aided Design 45(3):764–776.
[25] Carbonell J (1983) Learning by Analogy: Formulating and Generalizing Plans
from Past Experience. Machine learning: An Artificial Approach, Tioga, Palo
Alto, CA.
[26] Carroll JM (1997) Human–Computer Interaction: Psychology as a Science of
Design. International Journal of Human-Computer Studies 46(4):501–522.
[27] Cavallucci D, Weil R (2001) Integrating Altshuller’s Development Laws for
Technical Systems into the Design Process. CIRP Annals—Manufacturing Tech-
nology 50(1):115–120.
[28] Chakrabarti A, Shea K, Stone R (2011) Computer-Based Design Synthesis
Research: An Overview. Journal of Computing and Information Science in
Engineering 11(2):021003.
[29] Chandrasegaran SK, Ramani K, Sriram RD, Horváth I, Bernard A (2013) The
Evolution, Challenges, and Future of Knowledge Representation in Product
Design Systems. Computer-Aided Design 45(2):204–228.
[30] Chandrasekaran B, Josephson JR, Benjamins RV (1990) What are Ontologies
and Why do We Need Them? IEEE Intelligent Systems 14(1):20–26.
[31] Chang X, Rai R, Terpenny J (2010) Development and Utilization of Ontologies
in Design for Manufacturing. Journal of Mechanical Design 132(2):1–12.
[32] Chen YJ (2010) Development of a Method for Ontology-Based Empirical
Knowledge Representation and Reasoning. Decision Support Systems
50(1):1–20.
[33] Chen YJ (2010) Knowledge Integration and Sharing for Collaborative Molding
Product Design and Process Development. Computers in Industry 61(7):659–
675.
[34] Christophe F, Bernard A, Coatanéa É (2010) RFBS: A Model for Knowledge
Representation of Conceptual Design. CIRP Annals—Manufacturing Technology
59(1):155–158.
[35] Chryssolouris G (1996) Flexibility and Its Measurement. CIRP Annals—
Manufacturing Technology 45(2):581–587.
[36] Chütze M, Sachse P, Römer A (2003) Support Value of Sketching in the Design
Process. Research in Engineering Design 14(2):89–97.
[37] Ciocoiu M, Nau DS, Gruninger M (2001) Ontologies for Integrating Engineer-
ing Applications. Journal of Computing and Information Science in Engineering
1(1):12–22.
[38] Corbett J, McKeown PA, Peggs G, Whatmore R (2000) Nanotechnology:
International Developments and Emerging Products. CIRP Annals—
Manufacturing Technology 49(2):523–545.
[39] Coyne R, Park H, Wiszniewski D (2002) Design Devices: Digital Drawing and
the Pursuit of Difference. Design Studies 23(3):263–286.
[40] Coyne R, Rosenman MA, Radford AD, Gero JS (1987) Innovation and Creativity
in Knowledge-Based CAD. Expert Systems in Computer Aided Design, North-
Holland, Amsterdam435–465.
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
21
[41] Craig S, Harrison D, Cripps A, Knott D (2008) BioTRIZ Suggests Radiative
Cooling of Buildings Can Be Bone Passively by Changing the Structure of Roof
Insulation to Let Long Wave Infrared Pass. Journal of Bionic Engineering
5(1):55–66.
[42] Cross N (1999) Natural Intelligence in Design. Design Studies 20(1):25–39.
[43] Darlington MJ, Culley SJ (2008) Investigating Ontology Development for
Engineering Design Support. Advanced Engineering Informatics 22(1):112–
134.
[44] Deldin JM, Schuknecht M (2014) Biologically Inspired Design, Springer, Lon-
don. (Chapter 2).
[45] Deo HV, Suh NP (2004) Mathematical Transforms in Design: Case Study on
Feed-Back Control of a Customizable Automotive Suspension. CIRP Annals—
Manufacturing Technology 53(1):125–128.
[46] Dori D, Shpitalni M (2005) Mapping Knowledge about Product Lifecycle
Engineering for Ontology Construction via Object-Process Methodology. CIRP
Annals—Manufacturing Technology 54(1):117–122.
[47] Dreyfus H, Dreyfus S (1986) Why Expert Systems Do Not Exhibit Expertise.
IEEE Expert 1(2):86–90.
[48] Effendi I, Henson B, Agouridas V, de Pennington A (2002) Methods and Tools
for Requirements Engineering of Made-to-Order Mechanical Products. ASME
DETC/CIE 2002, ASME, Montreal, Canada.
[49] Eppinger SD, Whitney DE, Gebela DA (1992) Organizing the Tasks in Complex
Design Projects: Development of Tools to Represent Design Procedures. NSF
Design and Manufacturing Systems Conference, Atlanta, GA.
[50] Eppinger SD, Whitney DE, Smith RP, Gebala DA (1994) A Model-Based
Method for Organizing Tasks in Product Development. Research in Engineering
Design 6(1):1–13.
[51] Erixon G, von Yxkull A, Arnastroem A (1996) Modularity—The Basis for
Product and Factory Reengineering. CIRP Annals—Manufacturing Technology
47(1):159–162.
[52] Erkens A (1928) Beitragezur Konstruktionser. Zeitschrift des Vereins Deutscher
Ingenieure für Maschinenbau und Metallbearbeitung 72:17–21.
[53] Feng X, Li X, Matthias E (1987) Intelligent CAD of Shafts, Sleeves and
Cylindrical Gears by Micro Personal Computer. CIRP Annals—Manufacturing
Technology 36(1):69–72.
[54] Fodor JA, Pylyshyn ZW (1988) Connectionism and Cognitive Architecture: A
Critical Analysis. Cognition 28(1):3–71.
[55] Franke R (1951) VomAufban der Getriebe, VDI-Verlag, Dusseldorf.
[56] Gal-Tzur Z, Shpitalni M, Malkin S (1986) Integrate CAD/CAM System for
CAMs. CIRP Annals—Manufacturing Technology 35(1):99–102.
[57] Garner S (1992) The Undervalued Role of Drawing in Design. in Thistlewood
D, (Ed.) Drawing Research and Development, Longman, London98–109.
[58] Ge P, Lu SCY, Suh NP (2002) An Axiomatic Approach for ‘‘Target Cascading’’ of
Parametric Design of Engineering Systems. CIRP Annals—Manufacturing Tech-
nology 51(1):111–114.
[59] Gero JS (1990) Design Prototypes: A Knowledge Representation Schema for
Design. AI Magazine 11(4):26–36.
[60] Gero JS, Kannengiesser U (2004) The Situated Function–Behaviour–Structure
Framework. Design Studies 25(4):373–391.
[61] Gero JS, Kannengiesser U (2007) A Function–Behaviour–Structure Ontology
of Processes. Artificial Intelligence for Engineering Design Analysis and
Manufacturing 21(4):379–391.
[62] Gershenson JK, Prasad GJ, Zhang Y (2003) Product Modularity: Definitions
and Benefits. Journal of Engineering Design 14(3):295–313.
[63] Gibson K (2000) Divergent and Convergent Thinking with CAD. Journal of
Design Communication 2.
[64] Gleitman LR (1995) An Invitation to Cognitive Science: Visual Cognition, MIT
Press, Cambridge, MA267–296.
[65] Goel AK (1997) Design, Analogy and Creativity. IEEE Expert 12(3):62–70.
[66] Goel AK, Vattam S, Wiltgen B, Helms M (2012) Cognitive, Collaborative,
Conceptual and Creative—Four Characteristics of the Next Generation of
Knowledge-Based Cad Systems: A Study in Biologically Inspired Design.
Computer-Aided Design 44(10):879–900.
[67] Goel V (1995) Sketches of Thought, MIT Press, Cambridge, MA.
[68] Gruber T (1992) Ontolingua: A Mechanism to Support Portable Ontologies,
Knowledge Systems Laboratory, Stanford University, Stanford.
[69] Gruber T (1993) IA translation Approach to Portable Ontologies. Knowledge
Acquisition 5(2):199–220.
[70] Gruber T (1995) Toward Principles for the Design of Ontologies Used for
Knowledge Sharing. International Journal of Human-Computer Studies
43(4):907–928.
[71] Gu P, Hashemian M, Nee AYC (2004) Adaptable Design. CIRP Annals—
Manufacturing Technology 53(2):539–557.
[72] Gu P, Slevinsky M (2003) Mechanical Bus for Modular Product Design. CIRP
Annals—Manufacturing Technology 52(1):113–116.
[73] Gu P, Sosal S (1999) Product Modularization for Life Cycle Engineering.
Robotics and Computer Integrated Manufacturing 15(5):387–401.
[74] Guha S, Rastogi R, Shim K (1998) Cure: An Efficient Clustering Algorithm for
Large Database. ACM SIGMOD Record, 27ACM. 73–84. 2.
[75] Gundlach C (2006) Innovation mit TRIZ: Konzepte, Werkzeuge, Praxisanwen-
dungen, Symposion Publishing GmbH.
[76] Hansen F (1956) Konstruktionssystematik, VEB-Verlag Technik, Berlin.
[77] Hansen F (1965) Konstruktionssystematik, 2. Aufl, VEB-Verlag Technik, Berlin.
[78] Hartshorne C, Weiss P (1994) The Collected Papers of Charles Sanders Peirce,
vols. I–VI. Harvard University Press, Cambridge, MA.
[79] Harutunian V, Nordlund M, Tate D, Suh NP (1996) Decision Making and
Software Tools for Product Development Based on Axiomatic Design Theory.
CIRP Annals—Manufacturing Technology 45(1):135–139.
[80] Helms ME, Vattam SS, Goel AK (2009) Biologically Inspired Design: Process
and Products. Design Studies 30(5):606–622.
G Model
CIRP-1406; No. of Pages 23
22 Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
[81] Hernandez NV, Shah JJ, Smith SM (2007) Cognitive Models of Design Ideation.
ASME 2007 International Design Engineering Technical Conferences and
Computers and Information in Engineering Conference. American Society of
Mechanical Engineers 135–144.
[82] Hillstrom F (1994) Applying Axiomatic Design to Interface Analysis in
Modular Product Development. Advances Design Automation 69(2):363–371.
[83] Hirtz J, Stone RB, McAdams DA, Szykman S, Wood KL (2002) A Functional
Basis for Engineering Design: Reconciling and Evolving Previous Efforts.
Research in Engineering Design 13(2):65–82.
[84] Höhener K, Wzmo PP (1998) Simulation im Werkzeugmaschinenbau. Simu-
lationstechnik: 12. Symposium in Zürich, September, Tagungsband. Vdf
Hochschulverlag AG(12:15).
[85] Hong HB, Yin YH (2014) Ontological Modelling of Knowledge Management
for Human–Machine Integrated Design of Ultra-Precision Grinding Machine.
IEEE Conference on Enterprise System, Shanghai, China.
[86] Horvath I (2004) A Treatise on Order in Engineering Design Research.
Research in Engineering Design 15(3):155–181.
[87] Horváth I (2005) On some Crucial Issues of Computer Support of Conceptual
Design. Product Engineering, Springer, The Netherlands123–142.
[88] Huang C, Kusiak A (1997) Design of Modular Digital Circuits for Testability.
IEEE Transactions on Components Packaging and Manufacturing Technology
20:48–57.
[89] Israel JH, Wiese E, Mateescu M, Zöllner C, Stark R (2009) Investigating Three-
Dimensional Sketching for Early Conceptual Design—Results from Expert
Discussion and User Studies. Computers and Graphics 33(4):462–473.
[90] Jonson B (2005) Design Ideation: The Conceptual Sketch in the Digital Age.
Design Studies 26(6):613–624.
[91] Kamsu Foguem B, Coudert T, Béler C (2008) Knowledge Formalization in
Experience Feedback Processes: An Ontology-Based Approach. Computers in
Industry 59(7):694–710.
[92] Kaufmann H, Schmalstieg D (2003) Mathematics and Geometry Education
with Collaborative Augmented Reality. Computers & Graphics 27(3):339–345.
[93] Kimura F, Suzuki H (1989) A CAD System for Efficient Product Design Based
on Design Intent. CIRP Annals—Manufacturing Technology 38(1):149–152.
[94] Kiyokawa K, Takemura H, Yokoya N (1999) SeamlessDesign: A Face-to-Face
Collaborative Virtual/Augmented Environment for Rapid Prototyping of
Geometrically Constrained 3-D Objects. IEEE International Conference on
Multimedia Computing and Systems 2:447–453.
[95] Konak A, Coit DW, Smith AE (2006) Multi-Objective Optimization Using
Genetic Algorithms: A Tutorial. Reliability Engineering and System Safety
91(9):992–1007.
[96] Kong H, Hwang M, Kim P (2005) A New Methodology for Merging the Hetero-
geneous Domain Ontologies Based on the Word Net, Next Generation Web
Service Practices.
[97] Kruth JP, Detand J (1992) A CAPP System for Nonlinear Process Plans. CIRP
Annals—Manufacturing Technology 41(1):489–492.
[98] Kusiak A, Chow WS (1987) Efficient Solving of the Group Technology Prob-
lem. Journal of Manufacturing Systems 6(2):117–124.
[99] Kusiak A, He DW (1996) Performance Analysis of Modular Products. Interna-
tional Journal of Production Research 34(1):253–272.
[100] Kusiak A, Huang C (1996) Development of Modular Products. IEEE Transac-
tions on Components Packaging Manufacture Technology 19(4):523–538.
[101] Kusiak A, Larson N (1995) Decomposition and Representation Methods in
Mechanical Design. Journal of Vibration and Acoustics 117(B):17–24.
[102] Kusiak A, Szczerbicki E (1992) A Formal Approach to Specifications in
Conceptual Design. Journal of Mechanical Design 114(4):659–666.
[103] Larkin JH, Simon HA (1987) Why a Diagram is (Sometimes) Worth Ten
Thousand Words. Cognitive Science 11(1):65–100.
[104] Lawson B (1990) How Designers Think: The Design Process Demystified, Butter-
worth Architecture, Kent.
[105] Lee KD, Suh NP, Oh JH (2001) Axiomatic Design of Machine Control System.
CIRP Annals—Manufacturing Technology 50(1):109–114.
[106] Leu MC, Zhang W (2008) Virtual Sculpting with Surface Smoothing Based on
Level Set Method. CIRP Annals—Manufacturing Technology 57(1):167–170.
[107] Li Y, Yan R, Jian J (2010) A Semantics-Based Approach for Collaborative
Aircraft Tooling Design. Advanced Engineering Informatics 24(2):149–158.
[108] Lin J, Fox MS, Bilgic T (1996) A Requirement Ontology for Engineering Design.
Concurrent Engineering 4(3):279–291.
[109] Lu Y, Chen Y (1994) Humachine—A New Word for the 21st Century. Chinese
Journal of Mechanical Engineering 30(5):1–7.
[110] Lu Y, Chen Y (1995) Academic Architecture and Key Techniques of the Human
System. Chinese Journal of Mechanical Engineering 31(1):1–7.
[111] Lurie NH, Mason CH (2007) Visual Representation: Implications for Decision
Making. Journal of Marketing 71(1):160–177.
[112] Makris S, Rentzos L, Pintzos G, Mavrikios D, Chryssolouris G (2012) Semantic-
Based Taxonomy for Immersive Product Design Using VR Techniques. CIRP
Annals—Manufacturing Technology 61(1):147–150.
[113] Marler RT, Arora JS (2004) Survey of Multi-Objective Optimization Meth-
ods for Engineering. Structural and Multidisciplinary Optimization
26(6):369–395.
[114] Matousek R (1957) Konstruktionslehre des all gemeinenMaschinenbaus,
Springer Repringt, Berlin.
[115] Meijer J, Du K, Gillner A, Hoffmann D, Kovalenko VS, Tasuzawa T, Ostendorf A,
Poprawe R, Schulz W (2002) Laser Machining by Short and Ultrashort Pulses,
State of the Art and New Opportunities in the Age of the Photons. CIRP
Annals—Manufacturing Technology 51(2):531–550.
[116] Melkanoff MA, Kamvar E, Kops L (1985) A Manufacturing Oriented Intelligent
CAD System. CIRP Annals—Manufacturing Technology 34(1):159–162.
[117] Melvin JW, Suh NP (2002) Simulation within the Axiomatic Design Frame-
work. CIRP Annals—Manufacturing Technology 51(1):107–110.
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
[118] Miller G (1956) The Magical Number Seven, Plus or Minus Two: Some Limits
on Our Capacity for Processing Information. Psychological Review 63(2):81.
[119] Mistree F, Chen W, Rosen D, Allen JK (1994) Modularity and the Indepen-
dence of Functional Requirements in Designing Complex System. Concurrent
Product Design 74:31–38.
[120] Mizuno S, Akao Y (1993) QFD: The Customer-Driven Approach to Quality
Planning & Deployment, Asian Productivity Organization, Tokyo.
[121] Nee AYC, Ong SK, Chryssolouris G (2012) Augmented Reality Applications in
Design and Manufacturing. CIRP Annals—Manufacturing Technology 61(2):
657–679.
[122] Newcomb PJ, Bras B, Rosen DW (1996) Implications of Modularity on
Product Design for the Life Cycle. Proceedings of the 1996 ASME Design
Engineering Technical Conferences and Computers in Engineering Conference,
Irvine, CA, 18–22.
[123] Newell A (1980) Physical Symbol Systems. Cognitive Science 4(2):135–183.
[124] Newell A (1982) The Knowledge Level. Artificial Intelligence 18(1):87–127.
[125] Newell A, Simon HA (1972) Human Problem Solving, Prentice-Hall, Englewood
Cliffs, NJ.
[126] Ng LX, Ong SK, Nee AYC (2010) ARCADE: A Simple and Fast Augmented
Reality Computer-Aided Design Environment Using Everyday Objects. Pro-
ceedings of IADIS Interfaces and Human Computer Interaction Conference,
227–234.
[127] Ng LX, Oon SW, Ong SK (2011) GARDE: A Gesture-Based Augmented Reality
Design Evaluation System. International Journal on Interactive Design and
Manufacturing 5(2):85–94.
[128] Ong SK, Shen Y (2009) A Mixed Reality Environment for Collaborative
Product Design and Development. CIRP Annals—Manufacturing Technology
58(1):139–142.
[129] Owen R, Horváth I (2002) Towards Product-Related Knowledge Asset Ware-
housing in Enterprises. Proceedings of the Fourth International Symposium on
Tools and Methods of Competitive Engineering (TMCE 2002), Hubei, China, 155–
170.
[130] Pache M (2005) Sketching for Conceptual Design, Dr. Hut, München.
[131] Pahl G, Beitz W, Feldhusen J, Grote KH (2007) in Wallace K, Blessing L, (Eds.)
Engineering Design—A Systematic Approach. third ed. Springer, Berlin(Wallace
K., Blessing L. (Trans.)).
[132] Pappas M, Karabatsou V, Mavrikios D, Chryssolouris G (2006) Development
of a Web-Based Collaboration Platform for Manufacturing Product and
Process Design Evaluation Using Virtual Reality Techniques. International
Journal of Computer Integrated Manufacturing 19(8):805–814.
[133] Park J (2008) Augmented Reality Based Re-Formable Mock-Up for Design
Evaluation. IEEE International Symposium on Ubiquitous Virtual Reality 17–20.
[134] Paukner A, Suomi SJ, Visalberghi E, Ferrari PF (2009) Capuchin Monkeys
Display Affiliation toward Humans Who Imitate Them. Science
325(5942):880–883.
[135] Perkunder H, Israel JH, Alexa M (2010) Shape Modeling with Sketched
Feature Lines in Immersive 3D Environments. Proceedings of the Euro Graphics
Symposium on Sketch-based Interfaces and Modeling.
[136] Pugh S (1991) Total Design: Integrated Methods for Successful Product Engi-
neering, Prentice Hall, New York, NY.
[137] Qian L, Gero JS (1992) A Design Support System Using Analogy. Artificial
Intelligence in Design, Springer, The Netherlands795–813.
[138] Qian L, Gero JS (1996) Function–Behaviour–Structure Paths and their Role in
Analogy-Based Design. AI EDAM 10(4):289–312.
[139] Rahmani K, Thomson V (2012) Ontology Based Interface Design and Control
Methodology for Collaborative Product Development. Computer-Aided Design
44(5):432–444.
[140] Rajurkar KP, Zhu D, McGeough JA (1999) New Developments in Electro-
Chemical Machining. CIRP Annals—Manufacturing Technology 48(2):567–
579.
[141] Redtenbacher F (1952) Prinzipien der Mechanik und des Maschinenbaus. Mit
fünf lithographirten Tafeln.
[142] Reich T (1995) A Critical Review of General Design Theory. Research in
Engineering Design 7(1):1–18.
[143] Reinhart G, Weissenberger M (1999) Multibody Simulation of Machine Tools
as Mechatronic Systems for Optimization of Motion Dynamics in the Design
Process. Proceedings of IEEE/ASME International Conference on Advanced Intel-
ligent Mechatronics, IEEE, 605–610.
[144] Rekimoto J (1996) Transvision: A Hand-Held Augmented Reality System for
Collaborative Design. Proc. Virtual Systems and Multimedia, 85–90.
[145] Reuleaux F (1861) Konstrukteur, Vieweg und Sohn, Braunschweig.
[146] Reuleaux F (1875) Theorische Kinematic, Vieweg und Sohn, Braunschweig.
[147] Reuleaux F, Moll C (1854) Konstruktionslehre fur den Maschinenbau, Braunsch-
weig.
[148] Rosen DW (1996) Design of Modular Product Architectures in Discrete
Design Spaces Subject to Life-Cycle Issues. Proceedings of the 1996 ASME
Design Engineering Technical Conferences.
[149] Roy R, Kerr CIV, Sackett PJ, Corbett J (2005) Design Requirements Manage-
ment using an Ontological Framework. CIRP Annals—Manufacturing Technol-
ogy 54(1):109–112.
[150] Sakong K, Nam T (2006) Supporting Telepresence by Visual and Physical Cues
in Distributed 3D Collaborative Design Environments. CHI’06 Extended
Abstracts on Human Factors in Computing Systems, ACM: 1283–1288.
[151] Sauder P (1998) Design for Assembly and Schedule Optimization for Large
Structure Construction, Department of Mechanical Engineering, University
of Saskatchewan, Saskatoon, Canada. (Unpublished MSc Thesis).
[152] Schenk P (2005) Before and after the Computer: The Role of Drawing in
Graphic Design. Visual: Design: Scholarship 1(2):11–20. (The Research Journal
of the Australian Graphic Design Association).
[153] Schon D (1983) The Reflective Practitioner, Temple-Smith, London.
G Model
CIRP-1406; No. of Pages 23
Y.H. Yin et al. / CIRP Annals - Manufacturing Technology xxx (2015) xxx–xxx
[154] Schooler J, Ohlsson S, Brooks K (1993) Thoughts beyond Words: When
Language Overshadows Insight. Journal of Experimental Psychology: General
122(2):166–183.
[155] Schreiber G, Wielinga B, Jansweijer W (1995) The KACTUS View on the ‘O’
word, JCAI Workshop on Basic Ontological Issues in Knowledge Sharing.
International Joint Conference on Artificial Intelligence, Montreal, Canada.
[156] Schwertassek R, Wallrapp O (1999) Dynamik Flexibler Mehrkörpersysteme,
Vieweg, Braunschweig42.
[157] Shah JJ (1998) Experimental Investigation of Progressive Idea Generation
Techniques. ASME DETC Design Theory and Methodology Conference, Atlanta, GA.
[158] Shah JJ, Smith MS, Vargas-Hernandez N, Gerkens DR, Wulan M (2003)
Empirical Studies of Design Ideation: Alignment of Design Experiments with
Lab Experiments. Proceedings ASME DTM Conference, Chicago, IL.
[159] Shirai K, Murakami H, Twata K (1985) Development of a CAD/CAM System for
Progressive Dies. CIRP Annals—Manufacturing Technology 34(1):187–190.
[160] Shu LH (2010) A Natural-Language Approach to Biomimetic Design. Artificial
Intelligence for Engineering Design Analysis and Manufacturing 24(4):507–519.
[161] Shu LH, Ueda K, Chiu I, Cheong H (2011) Biologically Inspired Design. CIRP
Annals—Manufacturing Technology 60(2):673–693.
[162] Sim SK, Duffy A (2003) Towards an Ontology of Generic Engineering Design
Activities. Research in Engineering Design 14(4):200–223.
[163] Sohlenius G (1992) Concurrent Engineering. CIRP Annals—Manufacturing
Technology 42(2):645–656.
[164] Soininen T, Tiihonen J, Männistö T (1998) Towards a General Ontology of
Configuration. AI EDAM 12(4):357–372.
[165] Sriram R (1997) Intelligent Systems for Engineering: A Knowledge-Based Ap-
proach, Springer Verlag, Berlin.
[166] Stark R, Israel JH, Wohler T (2010) Towards Hybrid Modeling Environments-
Merging Desktop-CAD and Virtual Reality-Technologies. CIRP Annals—
Manufacturing Technology 59(1):179–182.
[167] Stones C, Cassidy T (2007) Comparing Synthesis Strategies of Novice Graphic
Designers Using Digital and Traditional Design Tools. Design Studies
28(1):59–72.
[168] Suh NP (1990) The Principles of Design, Oxford University Press, New York.
[169] Suh NP (2001) Axiomatic Design: Advances and Applications, Oxford University
Press, New York, NY.
[170] Suh NP (2005) Complexity in Engineering. CIRP Annals—Manufacturing Tech-
nology 54(2):46–63.
[171] Suh NP (2005) Complexity: Theory and Applications, Oxford University Press,
New York, NY.
[172] Suh NP, Do SH (2000) Axiomatic Design of Software Systems. CIRP Annals—
Manufacturing Technology 49(1):95–100.
[173] Suh NP, Sekimoto S (1990) Design of Thinking Design Machine. CIRP Annals—
Manufacturing Technology 39(1):145–148.
[174] Sure Y, Erdmann M, Angele J, Staab S, Stueder R, Wenke D (2002) OntoEdit:
Collaborative Ontology Development for the Semantic Web, Springer, Berlin/
Heidelberg.
[175] Suwa M, Tversky B (1997) What Do Architects and Students Perceive in their
Design Sketches?. A Protocol Analysis. Design Studies 18(4):385–403.
[176] Tatray P, Sohlenius G (1992) Probabilistic Aspects of Axiomatic. CIRP Annals—
Manufacturing Technology 41(1):173–176.
[177] Tomiyama T (1994) From General Design Theory to Knowledge Intensive
Engineering. Artificial Intelligence for Engineering Design Analysis and
Manufacturing 8(4).
[178] Tomiyama T, Gu P, Jin Y (2009) Design Methodologies: Industrial and Educa-
tional Applications. CIRP Annals—Manufacturing Technology 58(2):543–565.
[179] Tomiyama T, Kiriyama T, Takeda H, Xue D, Yoshikawa H (1989) Metamodel: A
Key to Intelligent CAD Systems. Research in Engineering Design 1(1):19–34.
[180] Tomiyama T, Yoshikawa H (1986) Extended General Design Theory. Technical
Report CS-145604, Centre for Mathematics and Computer Science, Amster-
dam, The Netherlands.
[181] Tschochner H (1954) Konstruieren und Gestalten, Girardet, Essen.
[182] Turban E, Aronson JE (2001) Decision Support Systems and Intelligent Systems,
6th ed. Prentice Hall, Book, MA.
[183] Tversky B (1999) What Does Drawing Reveal about Thinking? in Gero JS,
Tversky B, (Eds.) Visual and Spatial Reasoning in Design, University of Sydney,
Australia93–101.
[184] Ullman D, Wood S, Craig D (1990) The Importance of Drawing in the
Mechanical Design Process. Computers and Graphics 14(2):263–274.
[185] Uschold M, Gruninger M (1996) Ontologies: Principles, Methods and Appli-
cations. The Knowledge Engineering Review 11(02):93–136.
[186] Van Brussel H, Peirs J, Reynaerts D (2000) Assembly of Microsystems. CIRP
Annals—Manufacturing Technology 49(2):451–472.
[187] Van Houten F, Kimura F (2000) The Virtual Maintenance System: A Com-
puter-Based Support Tool for Robust Design, Product Monitoring, Fault
Diagnosis and Maintenance Planning. CIRP Annals—Manufacturing Technol-
ogy 49(1):91–94.
Please cite this article in press as: Yin YH, et al. Automating design with intelligent human–machine integration. CIRP Annals -
Manufacturing Technology (2015), http://dx.doi.org/10.1016/j.cirp.2015.05.008
23
[188] Vattam SS, Helms ME, Goel AK (2008) Compound Analogical Design: Inter-
action between Problem Decomposition and Analogical Transfer in Biologi-
cally Inspired Design. Design Computing and Cognition 08:337–396.
[189] VDI (1993) VDI 2221: Systematic Approach to the Development and Design of
Technical Systems and Products, BeuthVerlag, Düsseldorf, Germany.
[190] VDI (1997) VDI 2222: Part 1. Methodic Development of Solution Principles,
BeuthVerlag, Düsseldorf, Germany.
[191] VDI (2004) VDI 2206, Design Methodology for Mechatronic Systems, BeuthVer-
lag, Düsseldorf, Germany.
[192] Verhaegen PA, D’hondt J, Vertommen J, Dewulf S, Duflou JR (2009) Relating
Properties and Functions from Patents to TRIZ Trends. CIRP Journal of
Manufacturing Science and Technology 1(3):26–130.
[193] Wang C (2014) DataView: A Computational Visualisation System for Multi-
disciplinary Design and Analysis. Enterprise Information Systems. http://
dx.doi.org/10.1080/17517575.2014.914248.
[194] Wang C (2014) Insights from Developing Multidisciplinary Design and
Analysis Environment. Computers in Industry 65:786–795.
[195] Wang K, Takahashi A (2012) Semantic Web Based Innovative Design Knowl-
edge Modeling for Collaborative Design. Expert Systems with Applications
39(5):5616–5624.
[196] Wang L, Shen W, Xie H, Neelamkavil J, Pardasani A (2002) Collaborative
Conceptual Design—State of Art and Future Trends. Computer-Aided Design
34(13):981–996.
[197] Welbourne E, Battle L, Cole G (2009) Building the Internet of Things
Using RFID: The RFID Ecosystem Experience. Internet Computing 13(3):
48–55.
[198] Weule H, Albers A, Haberkem A (2002) Computer Aided Optimisation of the
Static and Dynamic Properties of Parallel Kinematics. Proceedings of the Third
Chemnitz Parallel Kinematic Seminar, 527–546.
[199] Wiendahl H, Lutz S (2002) Production in Networks. CIRP Annals—Manufactur-
ing Technology 51(2):573–586.
[200] Wiese E, Israel JH, Zöllner C, Pohlmeyer AE, Stark R (2009) The Potential of
Immersive 3D-sketching Environments for Design Problem-solving. Proceed-
ings of 13th International Conference on Human–Computer Interaction HCI
2009:485–489.
[201] Witherell P, Krishnamurty S, Grosse IR (2007) Ontologies for Supporting
Engineering Design Optimization. Journal of Computing and Information
Science in Engineering 7(2):141–150.
[202] Wogerbauer H (1943) Die Technik des Konstruierens. 2. Aufl, Oldenbourg,
Munchen.
[203] Yang D, Dong M, Miao R (2008) Development of a Product Configuration
System with an Ontology-Based Approach. Computer-Aided Design 40(8):
863–878.
[204] Yang J, Peng G, Liu W (2010) Research on Ontology-Based Integration of
Product Knowledge for Collaborative Manufacturing. The International Jour-
nal of Advanced Manufacturing Technology 49(9–12):1207–1221.
[205] Ye H, Liang M (2006) Simultaneous Modular Product Scheduling and
Manufacturing Cell Reconfiguration Using a Genetic Algorithm. Journal of
Manufacturing Science and Engineering 128(4):984–995.
[206] Yin YH, Xie JY (2011) Reconfigurable Manufacturing Execution System for
Pipe Cutting. Enterprise Information Systems 5(3):287–299.
[207] Yin YH, Xie JY, Xu LD (2012) Imaginal Thinking-based Human–machine
Design Methodology for the Configuration of Reconfigurable Machine tools.
IEEE Transactions on Industrial Informatics 8(3):659–668.
[208] Yin YH, Zhou C, Chen H (2012) A Pipe Route System Design Methodology for
the Representation of Imaginal Thinking. Advances in Knowledge Representa-
tion 77–116.
[209] Yin YH, Zhou C, Zhu JY (2010) A Pipe Route Design Methodology by Imitating
Human Imaginal Thinking. CIRP Annals—Manufacturing Technology 59(1):
167–170.
[210] Yong C, Chen Y (2000) Study on the Humachine Intelligent System and Its
Application. Chinese Journal of Mechanical Engineering 36(6):42–47.
[211] Yoshikawa H, Ando K (1987) Intelligent CAD in Manufacturing. CIRP Annals—
Manufacturing Technology 36(1):77–80.
[212] Yoshikawa T (1994) From General Design Theory to Knowledge-Intensive
Engineering. Artificial Intelligent in Engineering Design Analysis and
Manufacturing 8(4):319–333.
[213] Yoshikawa Y, Uehara K (1985) Design Theory for CAD/CAM Integration. CIRP
Annals—Manufacturing Technology 34(1):173–178.
[214] Zhan P, Jayaram U, Kim OJ (2010) Knowledge Representation and Ontology
Mapping Methods for Product Data in Engineering Applications. Journal of
Computing and Information Science in Engineering 10(2):021004.
[215] Zhang WY, Yin JW (2008) Exploring Semantic Web Technologies for Ontolo-
gy-Based Modeling in Collaborative Engineering Design. The International
Journal of Advanced Manufacturing Technology 36(9–10):833–843.
[216] Zhou C, Yin YH (2010) Pipe Assembly Planning Algorithm by Imitation
Human Imaginal Thinking. Assembly Automation 30(1):66–74.