From Intelligent Manufacturing to Smart Manufacturing for Industry 4.0 Driven by Next Generation Artificial Intelligence and Further On

Full text

From Intelligent Manufacturing to Smart
Manufacturing for Industry 4.0 Driven by Next
Generation Artificial Intelligence and Further On*
Xifan Yao, Jiajun Zhou, Jiangming Zhang
School of Mechanical and Automotive Engineering
South China University of Technology
Guangzhou, Guangdong, China
mexfyao@scut.edu.cn
Claudio R. Boër
University of Applied Sciences and Arts of Southern
Switzerland
Manno, Switzerland
claudio.boer@supsi.ch
Abstract—Smart manufacturing (SM) is emerging as a new
version of intelligent manufacturing (IM), reflecting the
magnitude and impact of smart technologies such the Internet of
Things, Cloud Computing, Cyber-Physical Systems and Big Data
on Industry 4.0. This paper addresses how IM evolves to SM
along with artificial intelligence (AI) evolution. To this end, this
study first summarizes how the symbolic AI (called AI 1.0)
characterized by structured contents and centralized control
structures evolves into the next-gene rati on AI (called AI 2.0)
characterized by unstructured contents, decentralized control
structures and machine learning (especially deep learning), and
expla in show IM enabled by AI 1.0 evolves into SM by AI 2.0
accordingly. Then, the comparison of IM and SM is discussed in
detail. Finally, the further development of smart manufacturing
for Industry 4.0 is given.
Keywords—Artificial intelligence; intelligent manufacturing;
smart manufacturing; cyber-physical system; social-cyber-physical
system; big data; Industry 4.0
I. INTRODUCTION
Big Data has become an emerging field of AI (Artificial
intelligence) for both academic and industrial communities
over the past years. More specifically, Chen et al. divided the
development of business intelligence and analytics (BI&A)
into 3 stages: the first centered on DBMS (database
management systems)-based structured content, called BI&A
1.0, the second on text and web analytics for unstructured web
content, called BI&A 2.0, an d the third on mobile and sensor-
based content, called BI&A 3.0 [1]. The amount of data
generated continues to grow exponentially along with the
digitalization of and use of the Internet of Things (IoT) in
factories, and manufacturing is identified as one of the five
domains in which Big Data has transformative potential [2].
Meanwhile, emerging is a new manufacturing model,
called smart manufacturing, which attracts a huge amount of
interest in both academic and industrial communities. The
Smart Manufacturing Leadership Coalition (SMLC) defined it
as "the intensified application of advanced intelligence
systems to enable rapid manufacturing of new products,
dynamic response to product demand, and real-time
optimization of manufacturing production and supply-chain
networks"[3].
As we know, there has existed so-called intelligent
manufacturing (IM), which can be viewed as the
implementation of AI in manufacturing. Then, what
relationship exists between SM and IM as well as Big Data
and AI? How do they evolve? This paper will try to answer
these questions with aims to clarify SM and IM by addressing
the current state of manufacturing intelligence/smart
pertaining to next-generation AI, and to identify fields where
intended efforts should be put on the next-generation
intelligent manufacturing--smart manufacturing.
II. AI EVOLUTION
Since AI was coined in 1956, the field of AI went through
ups and downs in the ensuing decades, and experienced two
major winters in 1974–80 and 1987–93 [4]. AI, can be roughly
classified into symbolic and sub-symbolic approaches (e.g.,
neural networks, fuzzy sets, and evolutionary algorithms).
At the first wave, in the 1960s symbolic intelligence had
won tremendous success at mimicking high-level thinking in
small demo programs for solving "toy problems". Meanwhile,
approaches based on neural networks or cybernetics wer e
given up or pushed into the background [5]. Especially, the
abandonment of connectionism in 1969 when Minsky and
Papert published the book Perceptrons that outlined the limits
of perceptrons [6], the large decrease in AI research in
response to the Lighthill Report of 1973, and DARPA's
funding cuts of the early 1970s, led AI to enter the first winter
in 1974–80 [4].
In 1980s, expert systems (ESs) or knowledge-based
systems (KBSs) got big success, in which there were
knowledge bases that involve high -level, domain knowledge
extracted from experts and expressed in specific structured
formats. Applications were developed also in manufacturing
[7] to solve complex problems of machine programming with
multiple tools working simultaneously on the same workpiece.
Such a need led to the emergence of knowledge
engineering/knowledge representation that is central to classic
AI research. Meanwhile, the "winter" of connectionist
*Project suppor t
ed by the National Natural Science Foundation of China
(51175187
, 51675186),
the Science & Technology Foundation of Guangdong
Province
(2016B090918035), and the Science & Technology Program of
Nansha, Guangzhou (2015CX005)..
2017 5th International Conference on Enterprise Systems
2572-6609/17 $31.00 © 2017 IEEE
DOI 10.1109/ES.2017.58
311

research came to an end in the middle 1980s, when the work
of Hopfield, Rumelhart and others revived large scale interest
in neural networks. Thus, sub-symbolic approaches without
specific representations of knowledge, e.g. neural networks
(NNs), fuzzy sets, statistical and genetic algorithms (GA),
began to get researchers' attention. As a result, sub-symbolic
methods came into AI systems. Unfortunately, the second AI
winter was triggered by the breakdown of the Lisp machine
market in 1987, the revocation of new expenditure on AI by
the Strategic Computing Initiative in 1988, the fizzle of the
fifth generation computers, and the fall of expert systems in
the early 1990s [4].
In the 1990s, distributed AI (DAI) got increasingly focused
along with the emergence of multi agents (MAs), which made
previous AI systems in centralized, hierarchical control
structures become flat, and substituted by a set of loosely
coupling agents collaborating with each other, interoperating
about messages and mutually learning from experience.
In the 2000s, Web 2.0, web services and web intelligence
emerged in AI systems. The prevalence of the Internet resulted
in large volumes of data available online. In addition,
individuals or organizations produce a great deal of structured
and unstructured data which are in need for processing. The
storing and processing of such lar ge volume of data reflects a
urgent need and a great challenge in mining, and processing
this data as knowledge [8].
The on-going wave of interest for AI started around 2010,
promoted by three factors associated with each other [9]: the
sources of Big Data including e-commerce, social media,
research community, organization, and government; the
machine learning approaches and algorithms which have been
dramatically improved based on raw material provided by Big
Data; and the powerful computers which support the
computing of Big Data.
Such a disruptive progress is shifting the traditional AI
(called AI 1.0), which emphasized symbolic approaches
characterized by structured contents and cen tralized control
structures, to enter a new version, called Artificial Intelligence
2.0 (AI 2.0), which emphasizes machine learning (especially
deep learning) characterized by unstructured contents and de-
centralized (distributed) control structures. Figure 1 briefly
illustrates AI evolution from the perspective of content and
control: from AI 1.0 (Symbolic AI), through either DAI
(Distributed AI) called AI 1.5D or Web AI called AI 1.5W, to
AI 2.0.
In the process of AI evolution, there were more and more
sub-symbolic approaches were added into. Although neural
networks were revived in 1980s, AI was still dominated by
symbolic approaches at that time. Later, the integration of
conventional expert experience, artificial neural networks,
evolutionary algorithms and fuzzy sets in various
configuration resulted in the so-called hybrid expert systems.
Neural networks facilitated the process of knowledge gaining
in some extent by training patterns instead of loading rules.
Evolutionary algorithms were used as efficient tools for
addressing complicated practical engineering optimization
tasks such as production scheduling problems. In 2010s, deep
learning (DL) originated form neural networks has emerged,
which manages to approach intelligence from large amounts
of unstructured data such that overcomes, in some extent, the
symbolic AI bottleneck problem that depends on knowledge
extracting, which might be the most hard part of building an
expert system. Due to the diversity of AI approaches and
solutions, there is not a clear distinction between what AI
approaches are applied. For example, in fuzzy systems
knowledge is preserved in styles of symbolic feature, while
implemented as a neuron-like numerical procedure in neural
networks. And intelligent agents can use both symbolic and
sub-symbolic approaches. Therefore, between AI 1.0 and AI
2.0 as shown in Fig. 2(a), there is a transition, called 1.5X,
which includes 1.5D and 1.5W as shown in Fig. 1.
Structured
Distributed
Symbolic
1950-60s
Agent
1990s
Sub-symbolic
(NN/FL/GA)
1980s
Web2.0
2000s
DL
Intelligent system
Smart system
IoT
L
IoS/Cloud
Big data
2010s
AI 1.0
AI 2.0
AI 1.5D
AI 1.5W
Unstructured
Centerlized
DAI
Web AI
Content
Configuration
Fig. 1. AI evolution from the perspectives of content and control
AI 1.0 AI 2.0
Symbolic
Expert systems
Big data
1 computer,
many users
1 computer,
1 user Many computers,
1user
(a)
(b)
Symbolic domination
x
Structured contents
x
Centralized control architectures
Sub-symbolic domination
x
Unstructured contents
x
De-centralized control
architectures
AI 1.5X
1960 Time1970 1980 1990 2000 2010
Agent Web
DL
For toy problems
NN/FL/GA
First winter
x Abandonment of connectionism
x Lighthill Report
Second winter
x Collapse of the Lisp machine market
x Fall of expert systems
First wave Second wave
Third wave
Things+clouds
Things+clouds
1960 Time1970 1980 1990 2000 2010
Fig. 2. AI evolution versus Computing's: (a) AI evolution; (b) Computing
evolut ion
As AI can be termed as a highly computerized system
whose behaviors require intelligence [9], computing evolution
as shown in Fig. 2 (b), also has big impact on AI. The first
312

wave of computing began when mainframe computers were
used in 1950-60s to manage companies' operations. Then in
1980-90s, personal computers rose and carried over the jobs
such as expert systems that had been deployed in expensive
special-purpose machines, which might lead to a false
impression of the so-called "the fall of the Lisp machine
market" [10]. Now, in 2010s, we are entering a new age of the
IoT, cloud computing, pervasive computing or ubiquitous
computing, which results in the emergence of Big Data. There
is a need of new AI tools to collect and analyze such Big Data.
To summarize the above discussion, AI evolution is given
in Table I.
TABLE I. AI EVOLUTION
Age
1950s -1960s
1980s
1990s
2000s
2010s-
AI Foc us
Symbolic
Expert system &Sub-symbolic
Agent
Web
Smart
Computat ion
Mainframes
PCs
PCs
Networks
Things+clouds
Processing
content/focus
DBMS- based stru cture d
content/ Knowledge
representation
Computational intelli gence/soft
computing/Data
analytic &
statistical methods
Distributed
computing
intelligence
Unstru ctured user crea ted
content/Web analytics and
web intelligence
IoT- based big data/
C
ontext-
aware analysis/
Deep learning
Control
structure
Centralized
Centralized
Distributed
Web-service ba sed
CPS-based distributed
III. FROM INTELLIGENT MANUFACTURING TO SMART
MANUFACTURING ALONG AI EVOLUTION
A. Intelligent manufacturing
For simplicity, intelligent manufacturing (IM) can be
viewed as the intersection of AI and manufacturing. Thus, IM
progresses along with the development of AI as shown in Fig.
3. In IM field, the first book, Manufacturing Intelligence, was
published in 1988 [11], and then we witnessed applications of
the methods, techniques and paradigms of AI in
manufacturing, resulting in the emergence of many specific
IM systems such as those in design, scheduling, production,
inspection, diagnosis, modeling, and control [12], during the
second AI waves shown in Fig. 2(a).
Structured
Distributed
Intelligent manufacturing
(Systems enabled by AI 1.0)
Smart manufacturing
(Systems enabled by AI 2.0)
Unstructured
Centralized
Distributed AI systems
For example, Multi-agent
systems
Web AI systems
For example, shopper
recommendation systems
Fig. 3. Intelligent manufacturing evolution along with AI
There are research papers tackled with the implementation
of AI in manufacturing industry. For example, Teti and
Kumara [12] surveyed the relevant AI methods introduced in
manufacturing before 1997, and grouped them as follow:
Knowledge-based/Expert systems (KBSs/ESs), Neural
Networks (NNs), Fuzzy Logic (FL), Multi Agents (MAs), and
others such as Evolutionary Algorithms and Simulated
Annealing (SA). The AI applications resulted in intelligent
components for CIM such as intelligent CAX (e.g. CAD, CAP,
CAM, and CAQ) as well as in intelligent robots [13]. In the
early years of intelligent manufacturing system (IMS)
development, KBSs/ESs had attracted much more attention,
and later NNs, case-based reasoning, GA and FL attracted
attention, too [14]. KBSs/ESs were introduced efficiently in
CIM components, while IMS was partially introduced in
industry but mainly for large companies [13]. The most
famous IMS research was the international scheme of joint
research, called Intelligent Manufacturing System found in
1995 and originally dated back to 1989 from Japan [15],
whose members come from Japan, US, EU and other
industrial countries.
In 1990s, Agent-based systems for intelligent
manufacturing [16] had emerged, followed by web-services
based systems for manufacturing [17] as well as Enterprise 2.0
[18] and crowdsourcing [19] in 2000s. The agent-based
method seemed to be the potential solution as it offered a
proper paradigm for the intelligent CIM components and IMS,
and more detail can be seen in [20-22]. Intelligent agents are
used in distributed AI (DAI), and such an agent-based DAI
approach has the ability to handle the issues of the present
software applications, specifically those working conditions
that are highly dynamic and uncertain [23]. However, most
agent-based systems are still at a research and prototype stage
in labs and not widely adopted in manufacturing.
B. Smart manufacturing
In 2010s, instead of intelligent technologies (Symbolic AI)
in manufacturing, we are seeing a similar convergence of
"smart" technologies (called "smart AI" in contrast to
Symbolic AI) in manufacturing with the potential to radically
improve the management of manufacturing enterprises in the
product life cycle so as to provide more options for customers
[24], as shown in Fig. 3.
313

The technologies used for the implementation of smart
manufacturing span a wide spectrum of domains, which are
initially referred as the IoT technologies [25], and then many
other related techniques such as Internet of Services (IoS),
Cyber-Physical Systems (CPS), Big Data, and advanced
robotics are added into. These smart technologies are taking
center stage in the second generation intelligent manufacturing
(IM 2.0.), i.e. smart manufacturing. The rise of IoT/CPS and
smart objects (phones) has made products become more
networked and accessible, from which the wealth of data
generated allows accurate targeting and further enabling
proactive management of enterprises through informed, timely,
in-depth decision execution [24]. Moreover, the fusion of
human, data and smart/intelligent algorithms has far-reaching
effects on manufacturing efficiency.
Of these smart terms, Big Data emphasizes data analysis,
while CPS cover s a larger scope compared to IoT or IoS, and
becomes increasingly important in manufacturing context [26].
Figure 4 exemplifies smart manufacturing as a cyber-physical
production system, which is viewed as the combination of IoT
and IoS. In the cyber space such as in cloud, manufacturing
related resources are virtualized and packaged as cloud
services that can be shared and utilized on demand via IoS. As
there exists large amount of manufacturing services, and a
single service usually cannot satisfy complicated task
requirements, it is necessary to solve such a so-called service
composition and optimal selection problem in order to form a
business process for manufacturing by using intelligent
optimization algorithms such as particle swarm optimization,
differential evolution, and bee colony algorithm. Then the
optimized business process created in the cyber system is
delivered to the physical shop floor for execution by linking
each cloud manufacturing service (virtual machin e) to its
corresponding physical machine. Meanwhile, the status in the
physical workshop is sensed via IoT to the cyber system to
track if the business process is fulfilled or not.
Internet of Things (IoT)
Physical system
Cyber system
Virtual Machine 1
Virtual Machine 2
Virtual Machine ...
Virtual Machine n
Machine ...
Internet of Services (IoS)
Smart manufacturing
(Cyber-physical production system)
Machine 1
Machine n
Fig. 4. Smart manufacturing exemplified a s a cyber-physical production
system
IV. COMPARISON OF IM AND SM
The traditional IM systems are centralized in configuration
with structured contents such as in database, knowledge base,
intelligent CAD (ICAD), ICAP, and ICAM, and usually used
in departments of an enterprise on a relatively small scale.
Meanwhile, agent technology promotes distributed AI in
manufacturing. With the development of the Internet,
manufacturing enterprises have shifted to the Web to sell and
promote their products, which leads to the emergence of
unstructured data of social media. Later, advances in the
Internet of Things and Services, smart technologies are
becoming increasingly utilized by industry, which results in
the emergence of smart manufacturing, and makes enterprises
face the challenge of exponential growing Big Data. As such,
enterprises need to utilize Big Data related techniques for
predicting, proactive maintenance, and production.
Nevertheless, such Big Data analytics is not available in
traditional, even in agent- or web- based manufacturing
systems, due to their limitation of data acquisition and
processing capabilities [27].
As stated above, traditional IM is based on symbolic AI,
which seeks to integrate human experience and knowledge in
manufacturing, usually extracted from experts in production.
In traditional IM systems, human experience/knowledge is
expressed by the IF-THEN production systems or other
formatted ways (e.g., frames, objects, and semantic). However,
Big Data generated in SM is out of the pr ocessing capability
of traditional database syst ems and software tools. Thus, Big
Data analytics becomes critical to enterprises in order to turn
raw data to actionable information, and/or knowledge that
helps decision making [28].
As shown in Fig. 5, IM is knowledge-based while SM is
data-driven and knowledge-enabled. Data and knowledge are
related in the hierarchy of Data, Information, Knowledge and
Wisdom (DIKW hierarchy). The emergence of "Big Data
(analytics)" makes decision makers in enterprises shift their
focus from knowledge back to "data" [28]. Owing to huge data
volumes being processed, predictive models and decision
emerging there from are based on machine learning [28],
especially deep learning, which can be used to abstract high-
level r epresentations from lar ge amounts of data [29].
Interviews
Intelligent manufacturing
Smart manufacturing
Expert system/
Knowledge-
based system
End
user
Domain
expert
Tests
Tests
Builds Based on
Modeling
Learning
Prediction
Action
Data
Information
Knowledge
Wisdom
DIKW
ERP
MES CRM
Web
IoT
Driven
by
Smart
AI
Others
(GA, NN,Ă)
AI
Cyber-physical production system
Symbolic
AI
Knowledge
engineer
Fig. 5. Intelligent manufacturing versus smart manufacturing
314

Therefore, in Big Data era, decision making is driven by
predictions - learning from data (experience) to predict, and
actions are taken in response to predictions [30]. Machine
learning, which learns from data and uses statistical
approaches to assist decision making that operates well in
practice, contrasts with the older "expert system" approach
that aims to mimic the rules from human experts with the help
of programmers translatin g the explicit rules into software
code.
Now, there is an increasing growing of literature on Big
Data in manufacturin g as be seen from surveyed in the last
few years especially after 2012 [31]. There is a profound
distinct between Big Data and structured data in DBMS. Big
Data is unstructured large scale data sets that is hard to be
processed by traditional software tools within a tolerable
elapsed time [32], characterized by a high Volume, Velocity
and Variety to need new emerging approaches for its
conversion into Value [33]. Although BI&A is divided into 3
stages from the viewpoint of data structures/sources as stated
above, both BI&A 2.0 and 3.0 are related to Big Data, and
they differ only in the data sources.
As stated above, we address AI or intelligent
manufacturing in two stages. Before 2000 we usually referred
the application of AI in manufacturing as intelligent
manufacturing (IM), but now smart manufacturing is
increasingly used, instead, as next-generation manufacturing
model with the smart sensing and control techniques [34]. At
first, the smart factory was initially studied with the
application of IoT in production [35], and later becomes a key
part of Industry 4.0 (Industrie 4.0) [36]. Then, more and more
"smart" techniques such as the IoS, cloud computing, and CPS
were also introduced to the smart factory or smar t
manufacturing [37].
Now, the term "smart" involves the creation and use of
data throughout the entire product cycle for the purpose of
more flexible manufacturing processes that respond quickly to
on-demand changes at low cost without harming the
environment [38]. It is the introduction of those technologies
that makes manufacturing become "smart" and different from
the "older" intelligent man ufacturing. Among those, IoT an d
CPS are, in a sense, acronymous, both of which try to connect
the cyber world and the physical world, and result in large
scale data set - Big Data. SM's strength lies in implementing
manufacturing intelligence (MI) from a comprehensive global
viewpoint under the support of ICT technologies including
IoT, IoS, CPS and Big Data that were not available previously.
As such, Davis et al. defined SM as "the use of data-driven MI
in multiple real-time applications deployed throughout all
operating layers across the factory and supply chain"[39],
and the future enterprise as "data driven, knowledge enabled,
and model rich with visibility across the enterprise such that
all operating actions are executed proactively by applying the
best information and performance metrics"[34].
Such ubiquitous use or access of mined information/
knowledge out of Big Data throughout the entire product value
chain, from product lines to the demand-supply networks,
supports new services and business models such as
"Everything-as-a-Service" and "Pay-per-use" as in cloud-
based design and manufacturing [27]. So "Design-as-a-
Service" and "Product-as-a-Service" can be accessible on
demand. As such, we are entering the next-generation
intelligent manufacturing –smart manufacturing, in which
factories have the ability to sense, understand, think, and
respond to our needs. The comparison of IM an d SM is given
in Table II, from which we can see that SM provides
significantly more benefits than IM.
TABLE II. THE COMPAR ISON OF IM AND SM
Charact eristics
IM
SM
Structure
Centralized
Distributed
Optimal scale
Usually local
Global
Structured content (data)


Big data (unstructured
content )

IoT/CPS

IoS/Cloud computing

Deep learning

Entire value chain support

Ubiquitous a ccess

Virtualization

Everyt hing-as-a -Servi ce

Visibility

Proactivity

Adaptability

Self-organization

Self-predictiveness

Context-awareness

Syste m of syste ms

V. FURTHER DEVELOPMENT OF SMART MANUFACTURING FOR
INDUSTRY 4.0
The term "Industry 4.0" originates from the high-tech
program of the German government, which derives from
"smart factories" [40]. Following the first Industrial
Revolution "Mechanization", the second "Mass production",
and the third "Automation", Industry 4.0 emerges through the
utilization of CPS, IoT and IoS [41, 42].
Smart factories (manufacturing) and Industry 4.0 are
empowering each other, both often described in CPS
architectures [41, 43-46]. However, the CPS architecture is not
sufficient for Industry 4.0 or a manufacturing system, which is,
by its very nature, socio-technical. Similar to Industry 4.0
proposed by Germany, "Made in China 2025 Strategy" also
focuses on intelligent (smart) manufacturing [47]. Moreover,
there is an increasing need of customized/personalized
products and sustainable manufacturing [48], as well as the
emergence of Enterprise 2.0, socialized enterprises,
crowdsourcing, social manufacturing, and open innovation, so
social dimension should be as well considered in smart
manufacturing/smart factories/ Industry 4.0 [49], as illustrated
in Fig. 6. To meet such needs, wisdom manufacturing (or wise
manufacturing) in the form of social CPS (SCPS) was
proposed [50, 51].
315

The previous three industrial revolutions tried to promote
or focus ed on mass production, while Industry 4.0 focuses on
mass customization/personalization. Therefore, Industry 4.0
somehow represents a socio-technical revolution , converting
the previous technical revolution into a socio-technical one
through the introduction of smart manufacturing technologies
and the convergence with social intelligence and human
wisdom such as social computing, collective intelligence,
crowdsourcing, and innovation. And SCPS-based
manufacturing can be viewed as the extension of CPS-based
manufacturing (including smart manufacturing and 3D
printing, which naturally revives craft production in the CPS
form) by adding the social dimension [49] , as shown in Fig. 6.
Time
Pre-industrial
revolutions
~2010
~1780
Technical
effects
Social
effects
Smart manufacturing
3Dprinting
SCPS-based manufacturing
(Wisdom manufacturing)
CPS-based manufacturing
Industry 1.0
(Mechanization)
Industry 2.0
(Electrification)
Industry 3.0
(Automation)
Productivity
Industry 4.0
(CPS/Wisdom)
Volume per model
Craft
production
Mass production
Mass personalization
Mass customization
(a)
(b)
Time
Fig. 6. Industry 4.0 as a social-technical revolution for producing
custom ized/ personalized products. ( a) Manufactu ring pa radi gm shi ft; (b)
Industr ial Re voluti ons
Wisdom manufacturing integrated IoT, IoS, Internet of
Contents and Knowledge (IoCK), and Internet of People (IoP)
in manufacturing context [51]. IoCK can be seen as DIKW,
Internet of Knowledge, or Internet of data (Big Data). Thus,
IoT, IoS, IoCK, and IoP can grouped as IoTSKP (Internet of
Things, Services, Knowledge, and People). Most new
emerging manufacturing models focus on one or two aspects
of IoTSKP, for example, smart factory on IoT, cloud
manufacturing on IoS (cloud computing), Enterprise 2.0/
crowdsourcing on IoP, predictive manufactur ing on Big Data
[52], and CPS-based smart manufacturing [53]. Wisdom
manufacturing is such a model that integrates the ideas of
these emerging models together, as shown in Fig. 7.
The integration of IoT, IoS, and IoP in manufacturing
leads to the rapid rise of available data sets, with which
enterprises are overwhelmed. Big Data is what these emerging
manufacturing models have in common. On the one hand, Big
Data as data-intensive computing, provides us a new paradigm
beyond exper imental and theoretical research and computer
simulations of natural phenomena [54] to rethink what it
means to AI or intelligent (smart) manufacturing. But on the
other hand, Big Data is considered as one of the greatest
challenges that 21st-century enterprises have to face. So it is
an urgently demand and predictive challenge to turn Big Data
into actionable information/knowledge for these emerging
manufacturing models.
Wisdom manufacturing
(Socio-cyber-physical production system)
Internet by and for people
Internet of Contents and Knowledge
Internet of Things
Internet of Services
Manufacturing Infrastructure
IoS-enabled manufacturing
(e.g., cloud manufacturing)
IoT-enabled manufacturing
(e. g. smart factory)
IoP-enabled
Enterprise 2.0/crowdsourcing
IoCK-enabled manufacturing
(e. g. predictive manufacturing)
Cyber-physical production system
(Smart manufacturing)
Data
Data
Data
Fig. 7. Wisdom manufacturing vs. other emerging manufacturing models
with big data in comm on
As shown in Fig. 8, SCPS-based manufacturing (wisdom
manufacturing) integrates the physical, cyber, and social
systems as a whole, which covers 6 semiotic levels - from
physical to social, and generates data from [55]: (1) the social
system that consists in the IoP including social media
networks, Web 2.0, crowdsourcing communities and mobiles;
(2) the cyber system that consists in the IoCK, and IoS
including digitalization of manufacturing such as NC/CAD/
CAM/CAE/CAPP/PDM/ERP, simulation, and virtual
manufacturing; and (3) the physical system that consists in the
IoT including sensor s and smart objects.
Now, this is a trend to blend "symbolic" AI with "smart"
AI, which will result in hybrid AI called "wise" AI or
Artificial Wisdom. As such, AI evolves from "symbolic" (AI
1.0) to "smart" (AI 2.0), and further to "wise" (AI 3.0). In a
sense, AI 3.0 can be viewed as the combination of symbolic
AI, smart AI and others. Namely, we have " wise AI (AI 3.0)
= smart AI (AI 2.0) + symbolic AI (AI 1.0) + others".
Similarly, manufacturing is evolving from intelligent to smart,
and will to wise (wisdom).
As the further development of smart manufacturing, wise
(wisdom ) manufacturing integrates not only symbolic AI and
smart technologies (smart AI) but also human
intelligence/wisdom in manufacturing, more specifically
integrating humans, computers and machines/things,
ubiquitous/artificial/collective intelligence, as well as human
knowledge/experience as a whole [50].
316

Social system
Cyber system
Internet of Contents and
Knowledge
Data generated from IoP,
social media, and mobiles
Information
Data
Knowledge
Control
Sensing
Service discovery,
composition &d provision
Internet of Things (IoT)
Internet of People (IoP)
Internet of Services
Physical system
Data sources
Information and
knowledge/models
extracted from big
data/ Deep learning
Big data Analytics Decision
SOA
Cloud
Physical Empiric Syntatic Semantic Pragmatic Social
Data generated from
NC/CAD/CAM/CAE/PDM/ERP/Virtrual
manufacturing/simulation
Data generated from IoT
Sensors, and Smart objects
Decision
making &
control
Ubiquitous/Pervasive computing
Ubiquitous/Pervasive intelligence
Explicit knowledge
Cloud computing
Intelligent algorithms
Machine learning
AI
Tacit knowledge
Knowledge sharing
Social computing
Open innovation
Crowdsourcing
Collective intelligence
SCPS-based manufacturing (Wisdom manufacturing)
CPS-based manufacturing (Smart manufacturing)
Fig. 8. A framework for SCPS-based manufacturing
REFERENCES
[1] H. C. Chen, R. H. L. Chiang, a nd V. C. Storey, "Business Intelli gence and
Analyti cs: Fro m Big D ata t o Big Impact," Mis. Quart., vol. 36, pp. 1165-
1188, Dec 2012.
[2] J. Manyika, M. Chui, B. Brown, J. Bughin, R. Dobbs, C. Roxburgh, et al.,
"Big Data: The next frontier for innovation, competition, and
productivity.
http://www.mckinsey.com/insights/business_technology/big_data_the_ne
xt_frontier_for_innovation.," McKinsey Global Institute2011.
[3] SMLC. (2011, July 24). Implementing 21st Century Smart
Manufacturing Workshop Summary Report.
[4] Wikipedia. (2017, Mar. 22). AI winter. Available:
https://en.wikipedia.org/wiki/AI_winter
[5] Wikipedia. (2017, Mar. 20). Artificial intel ligence. Available:
https://en.wikipedia.org/wiki/Artificial_intelligence
[6] M. Minsky and S. Papert, Perceptrons: An Introduction to Computational
Geometry: MIT Press, 1969.
[7] C. R. Boër, M. P etitti, F. Lombardi, and J.-P. Simon, "A CAPP/CAM
expert sy stem for a hi gh produ ctivit y, high flexibility CNC turni ng
center," Annals of the CIRP, vol. 39, pp. 481-483, 1990.
[8] F. Berma n, "Go t data? : a gu ide to data pr eservation i n the i nfor mation
age," Commun. ACM, vol. 51, pp. 50-56, 2008.
[9] National Science and Technology Council. (2016). Preparing for the
future of artificia l intelligence. Available:
https://www.whitehouse.gov/sites/default/files/whitehouse_files/microsit
es/ostp/NSTC/preparing_ for_the_future_of_ai.pdf
[10] J. Hendler, "Avoiding Another AI Winter," IEEE. Intell. Syst., vol. 23, pp.
2-4, 2008.
[11] P. Wright and D. A. Bourne, Manufacturing intelligence. Boston, MA:
317

Addison-Wesley, 1988.
[12] R. Teti and S. R. T. Kumara, "Intelligent Computing Methods for
Manufacturing Systems," CIRP Ann-Manuf. Technol., vol. 46, pp. 629-
652, 1997.
[13] P. Kopacek, "Intelligent Manufacturing:Present State and Future Trends,"
J. Intell. Robot. Syst., vol. 26, pp. 217-229, 1999.
[14] F. Meziane, S. Vadera , K. Kobba cy, and N. Prou dlove, "Intel ligent
systems in manufacturing: current developments and fu ture prospect s,"
Integr. Manuf. Syst., vol. 11, pp. 218-238, 2000.
[15] H. Setoya, "History and review of the IMS (Intelligent Manufacturing
System)," in International Conference on Mechatronics and Automation,
2011, pp. 30-33.
[16] W. Shen and D. H. Norrie, "Agent-Based Systems for Intelligent
Manufacturing: A State-of-Art Survey," Knowl. Inf . Syst., vol. 1, pp. 129-
156, 1999.
[17] A. Mostafaeipour and N. Roy, "Implementation of Web based Technique
into the Intelligent Manufacturing System," Int. J. Comput. Appl., vol. 17,
pp. 38-43, 2011.
[18] A. M cafee , "Enterprise 2.0: the da wn of e mergent collabor ation ," IEEE
Eng. Manag. Rev., vol. 34, pp. 38-38, 2006.
[19] E. Estellés-Arolas, "Towards an integrated crowdsourcing definition," J.
Inf. Sci., vol. 38, pp. 189-200, Apr 2012.
[20] J. Madejski, "Survey of the Agent-Based Approach to Intelligent
Manufacturing, " Journal of Achievements in Materials and
Manufacturing Engineering, vol. 21, pp. 67-70, 2007.
[21] L. Monostori, J. Vancza, and S. R. T. Kumara, "Agent -based systems for
manufa cturing," CIRP Ann-Manuf. Technol., vol. 55, pp. 697-720, 2006.
[22] P. Leitao, "Agent-based distributed manufacturing control: A state-of-the-
art survey," Eng. Appl. Artif. In tel., vol. 22, pp. 979-991, Oct 2009.
[23] H. A. Abbas, S. I. Shaheen, and M. H. Amin, "Simple, Flexible, a nd
Interoperable SCADA System Based on Agent Technology," Intelligent
Control and Automation, vol. 06, pp. 184-199, 2015.
[24] C. Leiva. (2015). On the Journey to a Smart Manufacturing Revolution.
http://www.industryweek.com/systems-integration/journey-smart-
manufacturing-revolution.
[25] A. Louchez and B. Wang. (2014). From Smart Manufacturing to
Manufacturing Smart. https://www.automationworld.com/sensors-
discrete/smart-manufacturing-manufacturing-smart.
[26] L. H. Wang, M. Tor ngren, and M. Onori, "Current status and
advancement of cyber-physical systems in manufacturing," J. Manuf.
Syst., vol. 37, pp. 517-527, 2015.
[27] D. Z. Wu, D. W. Rosen, L. H. Wang, and D. Schaefer, "Cloud-based
design and manufacturing: A new paradigm in digital manu facturing and
design innovation," Comput. Aided. Design., vol. 59, pp. 1-14, Feb 2015.
[28] S. Batra, "Big Data Analytics and its Reflections on DIKW Hierarchy,"
Rev. Manage., vol. 4, pp. 5-17, 2014.
[29]M. M. Najafabadi, F. Villanustre, T. M. Khoshgoftaar, N. Seliya, R. Wald,
and E. Muhar emag ic, "Dee p learni ng appli catio ns and cha lleng es in big
data analytics," J. Big Data , vol. 2, p. 1, 2015.
[30] E. Siegel, Predictive Analysis. New J ersey: J ohn Wil ey & So ns, Inc.,
2013.
[31] P. O'Donovan, K. L eahy, K. Bruton, and D. T. J. O 'Sullivan, "Big data in
manufa cturi ng: a sy stematic mapping st udy," J. Big Data, vol. 2, pp. 1-22,
2015.
[32] C. Snijders, U. Matzat, and U. D. Reips, ""Big Data" : Big Gaps of
Knowledge in the Field of Internet Science," Int. J. Internet Sci, vol. 7,
pp. 1-5, 2012.
[33] A. D. Mauro, M. Greco, and M. Grimaldi, "A formal definition of Big
Data base d on its essential feature s," Libr. Rev., vol. 65, pp. 122-135,
2016.
[34] J. Davi s, T. Edgar, R. G raybi ll, P. Korambath, B. Schott, D. Swink, e t al.,
"Smart manu factur ing," Annu. Rev. Chem. Biochem. Eng., vol. 6, pp.
141-160, 2015.
[35] D. Zu ehlk e, "S martFa ctor y-Towards a fa ctor y-of-thing s," Annu. Rev.
Control., vol. 34, pp. 129 -138, Apr 2010.
[36] M. Hermann, T. Pentek, and B. Otto. (2015). Design principles for
Industrie 4.0 scenarios: a literature review. Available:
http://www.snom.mb.tu-dortmund.de/cms/de/forschung/Arbeitsberichte/
Design-Principles-for-Industrie-4_0-Scenarios.pdf
[37] H. S. Kang, J. Y. Lee, S. Choi, H. Kim, J. H. Park, J. Y. Son, et al.,
"Smart Manufacturing: Past Research, Present Findings, and Future
Directions," Int. J. Pr. Eng. Man-Gt., vol. 3, pp. 111-128, Jan 2016.
[38] Wikipedia. (2017, Mar. 20). Smart manufacturing. Available:
https://en.wikipedia.org/wiki/Smart_ manufacturing
[39] J. Davis, T. Edgar, J. Porter, J. Bernaden, and M. Sarli, "Smart
manufacturing, manufacturing intelligence and demand-dynamic
performance ," Comput. Chem. Eng., vol. 47, pp. 145-156, Dec 20 2012.
[40] Wikipedia. (2017, Mar. 20). Industry 4.0. Available:
https://en.wikipedia.org/wiki/Indu stry_4.0
[41] N. Jazdi, "Cyber physical systems in the context of Industry 4.0," in 2014
IEEE International Conference on Automation, Quality and Testing,
Robotics, 2014, pp. 1-4.
[42] H. Kagermann, W. Wahlster, and J. Helbig. (2013). Recommendations for
implementing the strategic initiative I NDUSTRIE 4.0. Available:
http://www.acatech.de/filea dmin/user_upload/Baumstruktur_nach_Websi
te/Acatech/root/de/Material_fuer_Sonderseiten/Industrie_4.0/Final_repor
t__Industrie_4.0_accessible.pdf
[43] R. F. Babiceanu and R. Seker, "Big Data and virtualization for
manufacturing cyber-physical systems: A survey of the current status and
future ou tlook, " Comput. Ind., vol. 81, pp. 128-137, Sep 2016.
[44] J. Lee, B. Bagheri, and K. Hung-An, "A cyber-physical systems
architecture for Industry 4.0-based manufacturing systems," Manuf. Lett.,
vol. 3, pp. 18-23, 2015.
[45] P. J. Mosterman and J. Zander, "Industry 4.0 as a Cyber-Physical System
study," Software and Systems Modeling, vol. 15, pp. 17-29, Feb 2016.
[46]F. Shrouf, J. Ordieres, and G. Miragliotta, "Smart factories in Industry 4.0:
A review of the concept a nd of energy management approached in
production ba sed on the I ntern et of T hing s paradi gm," i n 2014 IEEE
International Conference on Industrial Engineering and Engineering
Management, Selangor, Malaysia, 2014, pp. 697-701.
[47] J. Zhou, "Intelligent manufacturing - Main Direction of "Made in China
2025"," China Mechanical Engineering, vol. 26, pp. 2273-2284, 2015.
[48] C. R. Boër, P. Pedrazzoli, A. Bett oni, and M. S orlini, Mass Customization
and Sustainability: An Assessment Framework and Industrial
Implementation: Springer, 2013.
[49] X. F. Yao and Y. Z. Lin, "Emerging ma nufacturing paradi gm shi fts for t he
incoming indu strial revolution," Int. J. Adv. Manuf. Tech., vol. 85, pp.
1665-1676, Jul 2016.
[50] X. F. Yao, H. Jin, and J. Zhang, "Towards a wisdom manufacturing
vision," Int. J. Comp ut. Integ. M., vol. 28, pp. 1291-1312, Dec 2 2015.
[51] X. Yao, Z. Lia n, Y. Ya ng, Y. Zhang, a nd H. J in, "Wi sdom manufac turin g:
new humans-comput ers-t hing s collaborati ve manufacturing model ,"
Comput. Integr. Manuf. Syst., vol. 20, pp. 1490-1498, 2014.
[52] J. Lee, E. Lapira, B. Bagheri, and H. A. Kao, "R ecen t advances a nd
trends in pr edicti ve manu factu ring sy stems in big data environ ment,"
Manuf. Lett., vol. 1, pp. 38–41, 2013.
[53] J. Zh ou, X. Yao, M. Liu, J. Zhan g, and T. Tao, "A state-of-th e-art review
on new emerging intelligent manufacturing paradigms," Integr. Manuf.
Syst, vol. 23, pp. 624-639, 2017.
[54] T. Hey, S. Tansley, and K. Tolle, The Fourth Paradigm: Data-Intensive
Scientific Disc overy. Redmond, VA: Microsoft Research, 2009.
[55] X. Yao, J. Zhou, C. Zhang, and M. Liu, "Proactive manufacturing - a big-
data driv en emerg ing manu factu ring paradig m," Integr. Manuf. Syst., vol.
23, pp. 172-185, 2017.
318