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

net/publication/333567029

From Intelligence Science to Intelligent Manufacturing

Article in Engineering · August 2019

DOI: 10.1016/j.eng.2019.04.011

CITATIONS READS

109 12,693

1 author:

Lihui Wang
KTH Royal Institute of Technology
824 PUBLICATIONS 25,823 CITATIONS

SEE PROFILE

All content following this page was uploaded by Lihui Wang on 26 August 2019.

The user has requested enhancement of the downloaded file.

 Engineering 5 (2019) 615–618

Contents lists available at ScienceDirect

Engineering
journal homepage: www.elsevier.com/locate/eng

Views & Comments

From Intelligence Science to Intelligent Manufacturing

Lihui Wang
KTH Royal Institute of Technology, Stockholm 10044, Sweden

1. Introduction performed through computer communication involving one

examiner, one human, and one machine (i.e., computer) in separate
The aim of intelligent manufacturing is to establish flexible and rooms. The examiner can ask any questions. If the examiner cannot
adaptive manufacturing operations locally or globally by using distinguish the machine from the human on the basis of their
integrated information technology (IT) and artificial intelligence answers, the machine passes the test. In 1951, Marvin Minsky
(AI) that can combine advanced computing power with manufac- and Dean Edmonds, two graduate students from Princeton Univer-
turing equipment. Intelligent manufacturing depends on the sity, built the first neuron computer to simulate a network of 40
timely acquisition, distribution, and utilization of real-time data neurons [5].
from both machines and processes on manufacturing shop floors An important milestone in AI development was the first AI
[1] and even across product life-cycles. Effective information shar- workshop [6], which was held in 1956 at Dartmouth College by
ing can improve production quality, reliability, resource efficiency, John McCarthy. This workshop marked the end of the ‘‘dark age”
and the recyclability of end-of-life products. Intelligent manufac- and the beginning of ‘‘the rise of AI” in AI history. The term ‘‘artifi-
turing built on digitalization is also intended to be more sustain- cial intelligence”, suggested by McCarthy, was agreed upon at that
able and to contribute to the factories of the future. However, time and is still in use. McCarthy later moved to Massachusetts
intelligent manufacturing depends extensively on AI. To better Institute of Technology (MIT); in 1958, he defined the first AI lan-
grasp the future of intelligent manufacturing, it is necessary to guage, LISP, which is still used today. One of the most ambitious
understand AI. This paper provides the author’s perspective on AI projects in this area was the General Problem Solver (GPS) [7],
from intelligence science to intelligent manufacturing. which was created in 1961 by Allen Newell and Herbert Simon
of Carnegie Mellon University. The GPS is based on formal logic
2. A brief history of AI and can generate an infinite number of operators attempting to
find a solution; however, it is inefficient in solving complicated
AI is a branch of intelligence science. The field of intelligence problems. In 1965, Lotfi Zadeh of UC Berkeley published his famous
science broadly covers two areas: natural intelligence and artificial paper ‘‘Fuzzy sets” [8], which is the foundation of fuzzy set theory.
intelligence. Natural intelligence is the science of discovering the The first expert system, DENDRAL [9], was developed at Stanford
intelligent behaviors of living systems, while artificial intelligence, University in 1969 in a project that was funded by the National
or AI, is both the science and the engineering of making intelligent Aeronautics and Space Administration (NASA) and led by Joshua
software systems and machines. These two research areas have con- Lederberg, a Nobel Prize laureate in genetics. At that time, how-
tributed to each other over decades. Advancements in natural intel- ever, because most AI projects could only handle toy problems
ligence have laid a solid foundation for AI research on artificial rather than real-world ones, many projects were canceled in the
neural networks (ANNs), genetic algorithms (GAs), ant colony opti- United States, United Kingdom, and several other countries. AI
mization (ACO), etc., while advanced AI tools have helped to speed research entered into a so-called ‘‘AI winter.”
up discoveries in natural intelligence [2]. Because of the relatively Despite these funding cuts, AI research continued. In 1969,
short history of AI, research in this field is still active, promising, Bryson and Ho [10] proposed the basis of back-propagation for
and yet to be discovered further, such as in the context of neural network learning. Furthermore, the first GA was proposed
manufacturing. in 1975 by John Holland of the University of Michigan, who used
Before discussing intelligent manufacturing, it is necessary to selection, crossover, and mutation as genetic operators for
briefly review the history of AI, as summarized in Fig. 1. The history optimization [11]. In 1976, MYCIN [12] was developed by the same
of AI can be traced back to the early 1940s. The first AI was a binary group as DENDRAL at Stanford University. MYCIN, which is a rule-
ANN model created by Warren McCulloch and Walter Pitts of the based expert system for blood disease diagnosis using 450 if-then
University of Illinois in 1943 [3]. Although their model only consid- rules, was found to perform better than a junior doctor.
ered the binary state (i.e., on/off for each neuron), it served as a After 30 years, work on neural networks was taken up again in
basis for rapid ANN research in the late 1980s. In 1950, British the field of AI. A new period—in which AI became a science—began
mathematician Alan Turing proposed the well-known Turing Test in 1982, when John Hopfield published his Hopfield networks [13],
[4] to determine whether machines can think. The Turing Test is which remain popular today. In 1986, back-propagation became a

https://doi.org/10.1016/j.eng.2019.04.011
2095-8099/Ó 2019 THE AUTHOR. Published by Elsevier LTD on behalf of Chinese Academy of Engineering and Higher Education Press Limited Company.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
616 L. Wang / Engineering 5 (2019) 615–618

Fig. 1. A brief history of AI.

real implemented learning algorithm [14] in ANN, 16 years after it of DeepMind beat the world Go champion Lee Sedol in four out
was proposed. It also triggered the start of distributed AI (DAI) of five games using cloud computing, reinforcement leaning, and
through parallel distributed processing. After 22 years, fuzzy set a Monte Carlo search algorithm combined with a deep neural net-
theory or fuzzy logic was successfully built into dishwashers and work for decision-making. Its newer version, AlphaGo Zero [18],
washing machines in 1987 by Japanese companies. In 1992, surpassed the ability of AlphaGo in just three days through self-
genetic programming [15] was proposed by John Koza to manipu- learning from scratch. Today, AI techniques and systems can be
late a symbolic code representing LISP programs. Based on the found in every field from chess playing to robot control, disease
ideas of DAI and artificial life, intelligent agents gradually took diagnosis to airplane autopilots, and smart design to intelligent
shape in the mid-1990s. In the late 1990s, hybrid systems of fuzzy manufacturing. In addition to the AI techniques summarized in
logic, ANN, and GA became popular for solving complex problems. Fig. 1, machine learning and deep learning show a great deal of
More recently, various new AI approaches have come into being, promise for intelligent manufacturing.
including ACO, particle swamp optimization (PSO), artificial Table 1 classifies typical machine learning models based on
immune optimization (AIO), and DNA computing. The potential whether they are supervised or unsupervised, discriminative or
of AI in the future—such as in manufacturing—remains generative, and deep learning or non-deep learning.
unpredictable.
The first popular AI tool was probably the AI-based chess- 3. Representative examples of AI in manufacturing
playing computer program Deep Blue [16], which was created by
International Business Machines Corporation (IBM). When Garry In the context of manufacturing, intelligence science—or, more
Kasparov, the world chess champion at that time, played with specifically, AI in the form of machine learning models—
Deep Blue in 1997 in an exhibition match, he lost the match to contributes to intelligent manufacturing. Fig. 2 depicts one
Deep Blue by 2.5 to 3.5. Another early example is the Honda ASIMO scenario of human–robot collaboration (HRC) in which data from
robot in 2005, which was able to climb stairs. For a robot to move sensors and field devices are transformed to knowledge after the
in an unstructured environment and be commanded by a human, it application of appropriate machine learning models [19].
requires the abilities of natural language processing, computer Knowledge is further transformed into actions using domain-specific
vision, perception, object recognition, machine learning, and HRC decision modules. Consequently, human operators can work
motion control at runtime. More recently, in 2016, AlphaGo [17] with robots safely in an immersive environment, while the robots

Table 1
Typical machine learning models.

Machine learning models Supervised/unsupervised/semi-supervised Discriminative/generative Deep learning/non-deep learning

K-means clustering Unsupervised Generative Non-deep learning
K-nearest neighbors Supervised Discriminative Non-deep learning
Support vector machine Supervised Discriminative Non-deep learning
Hidden Markov model Supervised Discriminative Non-deep learning
Random forest Supervised Discriminative Non-deep learning
XGBoost Supervised Discriminative Non-deep learning
Ensemble methods Supervised Discriminative Non-deep learning
Convolutional neural network Supervised Discriminative Deep learning
Recurrent neural network Supervised Discriminative Deep learning
Long short-term memory Supervised Discriminative Deep learning
Naive Bayes Supervised Generative Non-deep learning
Gaussian mixture model Supervised Generative Non-deep learning
Generative adversarial nets Semi-supervised Generative Deep learning
 L. Wang / Engineering 5 (2019) 615–618 617

Fig. 2. Machine learning in intelligent manufacturing.

can predict what the humans will do next and provide in situ assis-  Better horizontal and vertical integrations in five years may
tance as needed [20,21]. remove the gaps between automation islands by 80% in gen-
Brain robotics [22] is another example of adaptive robot control eral, mainly enabled by the IoT and mobile Internet.
using the brainwaves of experienced human operators. Rather than
following the data–knowledge–action chain, a brainwave–action
progression can be realized by mapping human brainwave pat-
terns to robot control commands through proper training, as
shown in Fig. 3. A 14-channel EMOTIV EPOC+ device (EMOTIV,
USA) is used in this case to collect human brainwave signals. The
matching commands after signal processing are then passed on
to the robot controller for adaptive execution.

4. Opportunities and challenges

Enabled by AI and the latest IT technologies such as cloud com-

puting, big data analytics, the Internet of Things (IoT), and mobile
Internet/5G, numerous opportunities for intelligent manufacturing
lie ahead. These new technologies will facilitate real-time informa-
tion sharing, knowledge discovery, and informed decision-making
in intelligent manufacturing, as follows:
 The IoT provides better connectivity of machines and field
devices for data collection, thereby making real-time data
collection possible.
 Mobile Internet/5G makes it practical to transmit a large
amount of data in ultra-low latency for real-time informa-
tion sharing.
 Cloud computing offers rapid and on-demand data analysis;
it also helps store data, which can be easily shared with
authorized users.
 Big data analytics can reveal hidden patterns and meaningful
information in data so as to convert data into information
and further transform information into knowledge.
For example, new opportunities in intelligent manufacturing
may include: ① remote real-time monitoring and control with lit-
tle delay, ② defect-free machining by means of opportunistic pro-
cess planning and scheduling, ③ cost-effective and secure
predictive maintenance of assets, and ④ holistic planning and con-
trol of complex supply chains. Moreover, intelligent manufacturing
in the near future will benefit from the aforementioned technolo- Fig. 3. Brain robotics HRC. Reproduced from Ref. [22] with permission of Elsevier,
gies in different temporal scales, as follows: Ó 2018.
618 L. Wang / Engineering 5 (2019) 615–618

 In ten years, experience-driven manufacturing operations [5] Minsky ML. Theory of neural-analog reinforcement systems and its application
to the brain model problem [dissertation]. Princeton: Princeton University;
may become data-driven with prior knowledge support,
1954.
mainly enabled by cloud computing and big data analytics. [6] Brook R. The relationship between matter and life. Nature 2001;409:409–11.
 In 20 years, numerous small and medium-sized enterprises [7] Newell A, Simon HA. GPS, a program that simulates human thought. In: Billing
(SMEs) may gain a competitive edge in the global market H, editor. Lernende automaten. München: R. Oldenbourg; 1961. p. 109–24.
[8] Zadeh LA. Fuzzy sets. Inf Control 1965;8(3):338–53.
by being powered by cloud manufacturing and made avail- [9] Buchanan B, Sutherland G, Feigenbaum EA. HEURISTIC DENDRAL: a program
able to all. for generating explanatory hypotheses in organic chemistry. In: Meltzer B,
Nevertheless, complexity and uncertainty will remain major Michie D, Swann M, editors. Machine intelligence 4. Edinburgh: Edinburgh
University Press; 1969. p. 209–54.
challenges in manufacturing in the years to come. AI and machine [10] Bryson Jr AE, Ho YC. Applied optimal control: optimization, estimation, and
learning can provide opportunities to relax or even resolve these control. Waltham: Blaisdell Publishing Company; 1969.
challenges to a large extent. For example, deep learning can be used [11] Holland JH. Adaptation in natural and artificial systems: an introductory
analysis with applications to biology, control, and artificial intelligence. Ann
to better understand the manufacturing context and more accu- Arbor: University of Michigan Press; 1975.
rately predict a future problem or failure in a manufacturing pro- [12] Shortliffe E, editor. Computer-based medical consultations: MYCIN. New
cess before it happens, thus leading to defect-free manufacturing. York: Elsevier; 1976.
[13] Hopfield JJ. Neural networks and physical systems with emergent collective
Safe HRC is another challenge in the progression toward intelli- computational abilities. Proc Natl Acad Sci USA 1982;79(8):2554–8.
gent and flexible automation that includes humans in the loop. Such [14] Rumelhart DE, McClelland JL; PDP Research Group. Parallel distributed
collaboration is useful and necessary, especially in manufacturing processing: explorations in the microstructures of cognition.
Cambridge: MIT Press; 1986.
assembly operations, where deep learning can help make robots
[15] Koza JR. Genetic programming: on the programming of the computers by
intelligent enough to assist human operators while providing means of natural selection. Cambridge: MIT Press; 1992.
much-improved context awareness toward absolute human safety. [16] Deep Blue [Internet]. Armonk: IBM [cited 2019 Jan 7]. Available from: https://
Finally, cybersecurity and new business models must be ade- www.ibm.com/ibm/history/ibm100/us/en/icons/deepblue/.
[17] AlphaGo Korea [Internet]. London: DeepMind Technologies Limited; c2019
quately addressed before intelligent manufacturing can be put into [cited 2019 Jan 7]. Available from: https://deepmind.com/research/alphago/
practice in the factories of the future. alphago-korea/.
[18] Silver D, Schrittwieser J, Simonyan K, Antonoglou I, Huang A, Guez A, et al.
Mastering the game of Go without human knowledge. Nature 2017;550
References (7676):354–9.
[19] Liu H, Fang T, Zhou T, Wang L. Towards robust human–robot collaborative
[1] Wang L. Machine availability monitoring and machining process planning manufacturing: multimodal fusion. IEEE Access 2018;6:74762–71.
towards cloud manufacturing. CIRP J Manuf Sci Technol 2013;6(4):263–73. [20] Liu H, Wang L. Human motion prediction for human–robot collaboration. J
[2] Negnevitsky M. Artificial intelligence: a guide to intelligent systems. 2nd Manuf Syst 2017;44(Pt 2):287–94.
ed. Harlow: Addison–Wesley; 2005. [21] Wang P, Liu H, Wang L, Gao RX. Deep learning-based human motion
[3] McCulloch WS, Pitts W. A logical calculus of the ideas immanent in nervous recognition for predictive context-aware human–robot collaboration. CIRP
activity. Bull Math Biophys 1943;5(4):115–33. Ann 2018;67(1):17–20.
[4] Turing AM. Computing machinery and intelligence. Mind 1950;LIX [22] Mohammed A, Wang L. Brainwaves driven human–robot collaborative
(236):433–60. assembly. CIRP Ann 2018;67(1):13–6.

View publication stats