PROCEEDINGS OF THE IEEE, FEBRUARY 2023 1

Artificial Intelligence-Driven Customized

Manufacturing Factory: Key Technologies,
Applications, and Challenges
Jiafu Wan, Member, IEEE, Xiaomin Li, Hong-Ning Dai, Senior Member, IEEE, Andrew Kusiak, Life
Member, IEEE, Miguel Martı́nez-Garcı́a, Member, IEEE, and Di Li

Abstract—The traditional production paradigm of large batch

T HE Industry 4.0 initiative is advocating smart manu-
arXiv:2108.03383v2 [cs.AI] 14 Apr 2023

production does not offer flexibility towards satisfying the facturing as the industrial revolution leading to global
requirements of individual customers. A new generation of economic growth [1]–[4]. Many countries, corporations, and
smart factories is expected to support new multi-variety and
small-batch customized production modes. For that, Artificial research institutions have embraced the concept of Industry
Intelligence (AI) is enabling higher value-added manufacturing 4.0, in particular the United States, the European Union, and
by accelerating the integration of manufacturing and information East Asia [5]. Some industries have begun a transformation
communication technologies, including computing, communica- from the digital era to the intelligent era. Manufacturing
tion, and control. The characteristics of a customized smart represents a large segment of the global economy, while the
factory are to include self-perception, operations optimization,
dynamic reconfiguration, and intelligent decision-making. The interest in smart manufacturing is expanding [6]. The progress
AI technologies will allow manufacturing systems to perceive the in information and communication technologies, for example,
environment, adapt to external needs, and extract the processing the Internet of Things (IoT) [7], [8], artificial intelligence
knowledge, including business models, such as intelligent produc- (AI) [9], [10], AI-Generated Content (AIGC) [11], and big
tion, networked collaboration, and extended service models. data [12], [13] for manufacturing applications, has impacted
This paper focuses on the implementation of AI in cus-
tomized manufacturing (CM). The architecture of an AI-driven smart manufacturing [14]. In the broad context of manu-
customized smart factory is presented. Details of intelligent facturing, customized manufacturing (CM) offers a value-
manufacturing devices, intelligent information interaction, and added paradigm for smart manufacturing [15], as it refers to
the construction of a flexible manufacturing line are showcased. personalized products and services. The benefits of CM have
The state-of-the-art AI technologies of potential use in CM, i.e., been highlighted by multinational companies.
machine learning, multi-agent systems, Internet of Things, big
data, and cloud-edge computing are surveyed. The AI-enabled Today, information and communication technologies are the
technologies in a customized smart factory are validated with base of smart manufacturing [16], [17], and intelligent systems
a case study of customized packaging. The experimental results driven by AI are the core of CM [18]. With the development
have demonstrated that the AI-assisted CM offers the possibility of AI technologies, new theories, models, algorithms, and
of higher production flexibility and efficiency. Challenges and applications - towards simulating, extending, and enhancing
solutions related to AI in CM are also discussed.
human intelligence - are continuously developed. The progress
Index Terms—Customized Manufacturing; Artificial Intelli- of big data analysis and deep learning has accelerated AI
gence; Industry 4.0; Smart Factory; Software-Defined Network.
to enter the 2.0 era [19]–[21]. AI 2.0 manifests itself as
a data-driven deep reinforcement learning intelligence [22],
network-based swarm intelligence [23], technology-oriented
I. I NTRODUCTION
hybrid intelligence of human-machine and brain-machine in-
This work was supported in part by the National Key R & D Program of teraction [24]–[26], cross-media reasoning intelligence [27],
China (Grant No. 2018YFB1700500), the Joint Fund of the National Natural [28], etc. Therefore, AI 2.0 offers significant potential to smart
Science Foundation of China and Guangdong Province (Grant No. U1801264),
and Macao Science and Technology Development Fund under Macao Funding manufacturing, especially, CM in smart factories [29].
Scheme for Key R & D Projects (0025/2019/AKP). (Corresponding author: Typically, AI solutions can be applied to several aspects
Hong-Ning Dai.) of smart manufacturing. AI algorithms can run the manu-
J. Wan and D. Li are with the School of Mechanical and Automotive
Engineering, South China University of Technology, Guangzhou, China (e- facturing of personalized products in a smart factory [30],
mails: [email protected], [email protected]). [31]. The AI-assisted CM is to construct smart manufacturing
X. Li is with the School of Mechanical Engineering, Zhongkai Uni- systems supported by cognitive computing, machine status
versity of Agriculture and Engineering, Guangzhou, China (e-mail: lixi-
[email protected]). sensing, real-time data analysis, and autonomous decision-
H.-N. Dai is with the Faculty of Information Technology, Macau University making [32], [33]. AI permeates through every link of CM
of Science and Technology, Macau SAR (email: [email protected]). value chains, such as design, production, management, and
A. Kusiak is with the Intelligent Systems Laboratory, Department of
Mechanical and Industrial Engineering, The University of Iowa, Iowa City, service [34], [35]. Based on these insights of CM and AI,
USA (email: [email protected]). the focus of this paper is on the implementation of AI in the
M. Martı́nez-Garcı́a is with the Dept. of Aeronautical and Auto- smart factory for CM involving architecture, manufacturing
motive Engineering, Loughborough University, UK (email: m.martinez-
[email protected]). equipment, information exchange, flexible production line, and
Manuscript received xx; revised xx. smart manufacturing services.
PROCEEDINGS OF THE IEEE, FEBRUARY 2023 2

The contributions of the research presented in this paper are • Smart interconnectivity. Smart manufacturing
as follows. embraces a cyber-physical environment, e.g.,
• The architecture of the AI-assisted CM for smart factories processing/detection/assembly equipment, and storage,
is developed by merging smart devices and industrial all operating in a heterogeneous industrial network.
networks with big data analysis. The Industrial IoT has progressed from the original
• The state-of-the-art AI technologies are reviewed and industrial sensor networks to the Narrow Band-Internet
discussed. of Things (NB-IoT), LoRa WAN, and LTE Cat M1 with
• The key AI-enabled technologies in CM are validated increased coverage at reduced power consumption [39].
with a prototype platform of a customized candy pack- Edge computing units are deployed to improve system
aging line. intelligence. Cognitive technology ensures the context
• The challenges and possible solutions brought by the awareness and semantic understanding of the industrial
introduction of AI into CM are discussed. IoT [40]. Intelligent industrial IoT as the key technologies
is widely used for intelligent manufacturing.
The remainder of the paper is organized as follows. In Sec-
• Dynamic reconfiguration. The concept of a smart factory
tion II, the relationship between the CM and AI is discussed.
aims at the rapid manufacturing of a variety of products
The general architecture of AI-assisted CM is presented in
in small batches. Since the product types may change
Section III. Section IV illustrates the implementation of AI
dynamically, system resources need to be dynamically
in intelligent manufacturing equipment. The intelligent infor-
reorganized. A multi-agent system [41] is introduced to
mation exchange process, flexible production line, and smart
negotiate a new system configuration.
manufacturing services in the AI-assisted CM are proposed in
• Massive volumes of data. An intelligent manufacturing
Section V and Section VI, respectively. A case study is pro-
system includes interconnected devices generating data
vided in Section VII. The challenges and possible solutions to
such as device status and process parameters. Cloud com-
the AI-assisted intelligent manufacturing factory are discussed
puting and big data science make data analysis feasible
in Section VIII. Section IX concludes the paper.
in failure prediction, active preventive maintenance, and
decision making.
II. C USTOMIZED M ANUFACTURING AND A RTIFICIAL • Deep integration. The underlying intelligent manufac-
I NTELLIGENCE turing entities, cloud platforms, edge servers, and up-
This section first summarizes the characteristics of cus- per monitoring terminals are closely connected. Data
tomized manufacturing in Section II-A and then discusses the processing, control, and operations can be performed
opportunities brought by AI-driven customized manufacturing simultaneously in the Cyber-Physical Systems (CPS),
in Section II-C. where the information barriers are broken down, thereby
realizing the deep integration of physical and information
environments.
A. Characteristics of customized manufacturing
Despite the progress made, manufacturing industry faces a
number of challenges, some of which are: traditional mass- B. Overview of AI technologies
production is not able to adapt to the rapid production of AI embraces theories, methods, technologies, and applica-
personalized products; and resource limitations, environmental tions to augment human intelligence. It includes not only AI
pollution, global warming, and an aging global population techniques such as perception, machine learning (ML), deep
have become more prominent. Therefore, a new manufac- learning (DL), reinforcement learning, and decision making,
turing paradigm to address these challenges is needed. The but also AI-enabled applications like computer vision, natural
customer-to-manufacture concept reflects the characteristics of language processing, intelligent robots, and recommendation
customized production where a manufacturing system directly systems, as shown in Fig. 1a. ML has outperformed traditional
interacts with a customer to meet his/her personalized needs. statistical methods in tasks such as classification, regression,
The goal is to realize the rapid customization of personalized clustering, and rule extraction [42]. Typical ML algorithms
products. The new generation of intelligent manufacturing include decision tree, support vector machines, regression
technology offers improved flexibility, transparency, resource analysis, Bayesian networks, and deep neural networks. Recent
utilization, and efficiency of manufacturing processes. It has advanced AI approaches like AIGC [11] can be used to intel-
led to new programs, e.g., the Factory of the Future in ligently direct the interaction between customers and product
Europe [36], Industry 4.0 in Germany [1], and Made in designers.
China 2025 [37]. Moreover, the United States has accelerated As a subset of ML algorithms, DL algorithms have superior
research and development programs [38]. performance than other ML algorithms. The recent success of
Compared with mass production, the production organi- DL algorithms mainly owes to three factors: 1) the availability
zation of CM is more complex, quality control is more of massive data; 2) the advent of computer capability achieved
difficult, and the energy consumption needs attention. In by computer architectures and hardware, such as Graphic
classical automation, the production boundaries were rigid to Processing Units (GPUs); 3) the advances in diverse DL
ensure quality, cost, and efficiency. Compared with traditional algorithms such as a convolutional neural network (CNN),
production, CM has the following characteristics. long short-term memory (LSTM) and their variants. Different
PROCEEDINGS OF THE IEEE, FEBRUARY 2023 3

Knowledge
Knowledge Machine
Machinelearning
learning Computer
Computer Customized
Customized
graph
graphanalysis
analysis Customer
Customer After-sales
After-sales
vision
vision product
product manage-
manage- service
service
design
design ment
ment
Reinforcement
Reinforcement
learning
learning Natural
Naturallanguage
language
Manufact-
Manufact-
processing
processing Customized
Customized
AI
AI uring
uring
product
product
Market
Market
analysis
analysis
manage-
manage-
Deep
Deeplearning
learning manufacturing
manufacturing
foster
foster ment
ment
Perception
Perception
Speech
Speech Customized
Customized
recognition
recognition Manufacturing
Manufacturing
maintenance
maintenance
product
product …
…
Decision
Decisionmaking
making logistics
logistics

Recommendation
Recommendation Customized
Customizedmanufacturing
manufacturing
system
system Intelligent
Intelligentrobot
robot

(a) (b)
Fig. 1. The AI and customized manufacturing. (a) AI technologies include perception, machine learning, deep learning, reinforcement
learning, and decision making as well as AI-enabled applications like computer vision, natural language processing, intelligent robots,
and recommendation systems. (b) AI can foster customized manufacturing in the aspects: customized product design, customized product
manufacturing, manufacturing maintenance, customer management, logistics, after-sales service, and market analysis.

from ML methods, which require substantial efforts in feature summary, AI-driven CM has the following advantages [47],
engineering in processing raw industrial data, DL methods [48].
combine feature engineering and learning process together, 1) Improved production efficiency and product quality. In
thereby achieving outstanding performance. CM factories, automated devices can potentially make
However, DL algorithms also have their disadvantages. decisions with reduced human interventions. Technolo-
First, DL algorithms often require a huge amount of data gies such as ML and computer vision are enablers
to train DL models to achieve better performance than other of cognitive capabilities, learning, and reasoning (e.g.,
ML algorithms. Moreover, the training of DL models requires analysis of order quantities, lead time, faults, errors, and
substantial computing resources (e.g., expensive GPUs and downtime). Product defects and process anomalies can
other computer hardware devices). Third, DL algorithms also be identified using computer vision and foreign object
suffer from poor interpretability, i.e., a DL model is like an detection. Human operators can be alerted to process
uncontrollable “black box”, which may not obtain the result deviations.
as predicted. The poor interpretability of DL models may 2) Facilitating predictive maintenance. Scheduled mainte-
prevent their wide adoption in industrial systems, especially in nance ensures that the equipment is in the best state.
critical tasks like fault diagnosis [43] despite recent advances Sensors installed on a production line collect data for
in improving the interpretability of DL models [44]. analysis with ML algorithms, including convolutional
neural networks. For example, the wear and tear of a
machine can be detected in real-time and a notification
C. AI-driven customized manufacturing can be issued.
As AI technologies have demonstrated their potential in 3) Developing of smart supply chains. The variability and
areas such as customized product design, customized prod- uncertainty of supply chains for CM can be predicted
uct manufacturing, manufacturing management, manufactur- with ML algorithms. Moreover, the insights obtained can
ing maintenance, customer management, logistics, after-sales be used to predict sudden changes in customer demands.
service, and market analysis as shown in Fig. 1b, industrial In short, the incorporation of AI and industrial IoT
practitioners and researchers have begun their implementation. brings benefits to smart manufacturing. AI-assisted tools
For example, the work [45] presents a Bayesian network- improve manufacturing efficiency. Meanwhile, higher value-
based approach to analyze the consumers’ purchase behaviour added products can be introduced to the market.
via analyzing RFID data, which is collected from RFID-tags However, we cannot deny that AI technologies still have
attached to in-store shopping carts. Moreover, a deep learning their limitations when they are formally adopted to real-
method is adopted to identify possible machine faults through world manufacturing scenarios. On the one hand, AI and ML
analyzing mechanic data collected from the real industrial algorithms often have stringent requirements on computing
environments such as induction motors, gearboxes, and bear- facilities. For example, high-performance computing servers
ings [46]. equipped with GPUs are often required to fasten the training
Therefore, the introduction of AI technologies can poten- process on massive data [49] while exiting manufacturing fa-
tially realize the customized manufacturing. We name such cilities may not fulfill the stringent requirement on computing
AI-driven customized manufacturing as AI-driven CM. In capability. Therefore, the common practice is to outsource (or
PROCEEDINGS OF THE IEEE, FEBRUARY 2023 4

Wireless link
Wired link
Router
Base station

Access point Logistics Customers

Suppliers Markets Computing center

Management

Smart
interation
Customer Visualization
Production
analysis
line

Edge Cloud
computing computing
Sensors

Smart Smart Sensors

Robots Motors devices services

AI models Data

Machines Maintenance Profits

Fig. 2. The architecture of AI-assisted customized manufacturing includes smart devices, smart interaction, AI layer, and smart services.

upload) the manufacturing data to cloud computing service to maximize their effectiveness. At the same time, the cor-
providers who can conduct the computing-intensive tasks. responding AI algorithms should be redesigned to match the
Nevertheless, outsourcing the manufacturing data to the third corresponding computing paradigm. Cloud intelligence is re-
party may lead to the risk of leaking confidential data (e.g., sponsible for making comprehensive, time-insensitive analysis
customized product design) or exposing private customer data and decisions, while the edge and local node intelligence
to others. On the other hand, transferring the manufacturing are applicable to the context or time-aware environments. In-
data to remote clouds inevitably leads to high latency, thereby telligent manufacturing systems include smart manufacturing
failing to fulfill the real-time requirement of time-sensitive devices, realize intelligent information interaction, and provide
tasks. intelligent manufacturing services by merging AI technologies.
As shown in Fig. 2, an AI-assisted CM framework that
III. A RCHITECTURE OF AN AI-A SSISTED C USTOMIZED includes smart devices, smart interaction, AI layer, and smart
M ANUFACTURING FACTORY services. We then explain this framework in detail as follows.
1) Smart devices: include robots, conveyors, and other ba-
This section first presents an AI-Assisted customized manu-
sic controlled platforms. Smart devices serve as “the physical
facturing (AIaCM) framework in Section III-A and then gives
layer” for the entire AIaCM. Specifically, different devices and
a brief comparison of the proposed AIaCM framework with
equipment, such as robots and processing tools are controlled
the state-of-the-art literature in Section III-B.
by their corresponding automatic control systems. Therefore, it
is crucial to meet the real-time requirement for the device layer
A. AI-Assisted Customized Manufacturing Factory in an AIaCM system. To achieve this goal, ML algorithms can
be implemented at the device layer in low power devices such
Different frameworks have been presented towards the
as FPGAs. The interconnection of the physical devices, e.g.,
increased interactivity and resource management [50]–[52].
machines, conveyors, is implemented at the device layer [59],
Most studies have focused on information communica-
[60] using edge computing servers.
tions [53] or big data processing [54]–[56]. So far, research
proposing generic AI-based CM frameworks is limited. Sys- 2) Smart interaction: links the device layer, AI layer, and
tem performance metrics, e.g., flexibility, efficiency, scalabil- services layer [61], [62]. It represents a bridge between differ-
ity, and sustainability, can be improved by adopting AI tech- ent layers of the proposed architecture. The smart interaction
nologies such as ML, knowledge graphs, and human-computer layer is composed of two vital modules. The first module
interaction (HCI). This is especially true in sensing, inter- includes basic network devices such as access points, switches,
action, resource optimization, operations, and maintenance routers and network controllers, which are generally supported
in a smart CM factory [57], [58]. Since cloud computing, by different network operating systems, or equipped with dif-
edge computing, and local computing paradigms have their ferent network functions. The basic network devices constitute
unique strengths and limitations, they should be integrated the core of the network layer [63], [64]. Different from the first
PROCEEDINGS OF THE IEEE, FEBRUARY 2023 5

TABLE I. Summary of most relevant state-of-the-art literature

Refs. Smart devices Smart interaction AI Smart services Pros Cons Applications
[50] X × X X Integration sensor with No edge computing Machine status monitor-
cloud services considered ing (primitive ML meth-
ods were used)
[51] X X X X Service-oriented smart No edge computing Milk production from
manufacturing considered buffalo pasture
[52] X X × × Integration CPS with No edge computing and Several cases from prod-
smart manufacturing in-depth analysis of AI uct design to manufac-
algorithms turing control
[53] X × × X Comprehensive consid- No AI as well as edge No specific application
eration of the entire in- computing considered
dustrial network
[55] X × X X Integration sensor with No edge computing Light gauge steel pro-
cloud services considered duction line
[56] X X X × Integration CPS with No edge computing Production line and fac-
AI considered tory management
[57] × × X × Diverse AI algorithms No consideration Cold spray additive
were used of smart devices, manufacturing,
interactions and augmented reality-
services guided inspection and
surface stress estimation

module which is fixed or static, the second module consists data. Similarly, a cloud-based manufacturing equipment [55] is
of the dynamic elements, including network/communications proposed to provide users with on-demand services. However,
protocols, information interaction, and data persistent or tran- outsourcing manufacturing data to cloud services providers
sient storage. These dynamic elements are essentially infor- who are often owned by third parties can also bring the risks
mation carriers to connect different manufacturing processes. of leaking customers’ private data and exposing confidential
The dynamic module is running on top of the static one. manufacturing data (e.g., product design models). Despite
AI is utilized in the prediction of wireless channels, op- most of the aspects being considered, the work [51] ignores
timization of mobile network handoffs, and control network the critical issues such as the edge computing paradigm and
congestion. Recurrent Neural Networks (RNN) or Reservoir advanced AI technologies.
Computing (RC) are candidate solutions due to their advan- In contrast, our AIaCM framework includes all the aspects
tages of them in analyzing temporal network data. in CM, including smart devices, smart interaction, AI tech-
3) The AI layer: includes algorithms running at different nologies, and smart services. Meanwhile, our AIaCM frame-
computing platforms such as edge or cloud servers [55], work also considers the advent of edge computing, software-
[65]. The computing environment consists of cloud and edge defined networks, and advanced AI technologies. Moreover,
computing servers running MapReduce, Hadoop, and Spark. we also present a full-fledged prototype to further demonstrate
AI algorithms are adopted at different levels of computing the effectiveness of the proposed framework (please refer to
paradigms in the AIaCM architecture. For instance, training a Section VII for more details). The implementation details of
deep learning model for image processing can be conducted the AIaCM architecture are discussed next.
in the cloud. Then, edge computing servers are responsible for
running the trained DL model and executing relatively simple
algorithms for specific manufacturing tasks. IV. I NTELLIGENT M ANUFACTURING D EVICES
4) Smart manufacturing services: include data visualiza-
A. Edge computing-assisted intelligent agent construction
tion, system maintenance, predictions, and market analysis.
For example, a recommender system can provide customers In the customized production paradigm, manufacturing de-
with details of CM products, and the information including the vices should be capable of rapid restructuring and reuse for
performance of a production line, market trends, and efficiency small batches of personalized products [66], [67]. However,
of the supply chain. it is challenging to achieve elastic and rapid control over the
massive manufacturing devices. The agent-based system was
B. Overview on state-of-the-art manufacturing methods considered a solution to this challenge [68], [69]. Agents can
Recently, substantial research efforts have been made to im- autonomously and continuously function in a collaborative
prove the interactivity and elasticity of exiting manufacturing system [70]. A multi-agent system can be constructed to
factories [50]–[58]. Table I summarizes most relevant state- take autonomous actions. Different types of agents have been
of-the-art literature. We can observe from Table I that most of constructed in [71]–[73].
the references only concentrate on a single or several aspects Although a single agent may have sensing, computing, and
in CM. For example, the work [50] presents a cloud man- reasoning capabilities, it alone can only accomplish relatively
ufacturing framework to analyze and process manufacturing simple tasks. Smart manufacturing may involve complex tasks,
PROCEEDINGS OF THE IEEE, FEBRUARY 2023 6

Motor

AI layer is-a
PLC is-a Conveying is-a Function is-a Computing
AI library
Data storage Computing facilities
Edge computing agent is-a
is-a is-a
Conveying_unit
Intelligent
Edge layer Warehouse
Cutting is-a ID_device 2
is-a is-a
is-a is-a
Edge computing node Agent Warehouse Working is-a
Management quality
Current
System ID_device 1
Agent is-a state
Neighbor
Working is-a
Agent layer is-a
quality
Current
Neighbor
state
Camera Motor
Conveyor belt Fig. 4. Manufacturing resources from the device function perspective.
The CM resources of a product can be mapped into computing,
cutting, conveying, and other functions.
Device layer
AGV
Robot Forklift
3D printer Robot arm
the process, the master-slave or auction mode can be adopted
Fig. 3. Edge computing-assisted manufacturing devices. This archi- for coordination, according to the status analysis of each edge
tecture includes the device layer, agent layer, edge computing layer, server.
and AI layer.
Additionally, with the help of edge computing, it is possible
to establish a quantitative energy-aware model with a multi-
agent system for load balancing, collaborative processing
for instance, the image-based personalized product recogni-
of complex tasks, and scheduling optimization in a smart
tion, expected from the emerging multi-agent systems [74],
factory [79]. The above procedure can also optimize the
[75]. However, the multiple agents are deficient in processing
production line with better logistics while ensuring flexibility
massive data. Recent advances in edge computing can meet
and manufacturing efficiency.
this emerging need [76]–[78]. As shown in Fig. 3, a variety
of decentralized manufacturing agents are connected to edge
computing servers via high-speed industrial networks. The B. Manufacturing resource description based on ontology
edge computing assisted manufacturing agents embrace the Intelligent manufacturing will be greatly beneficial to the in-
device layer, agent layer, edge computing layer, and AI layer. tegration of distributed competitive resources (e.g., manpower
An agent is equipped with a reasoning module and a and diverse automated technologies), so that resource shar-
knowledge base, offering basic AI functionalities such as ing between enterprises and flexibility to respond to market
inferencing and computing. Moreover, with the support of changes are possible (i.e., CM). Therefore, in smart manu-
new communication technologies (e.g., 5G mobile networks facturing, it is imperative to realize dynamic configurations of
and high-speed industrial wired networks), all agents and edge manufacturing resources [80], [81]. CM can optimize lead time
computing servers can be interconnected. and manufacturing quality under various real-world constraints
Agents run on edge computing servers to guarantee low- of dynamic nature (resource and manpower limitations, market
latency services for data analytics. The agent edge servers are demand, etc.).
connected by high-speed industrial IoT to achieve low latency. There are several strategies in describing manufacturing
Generally, edge computing servers support a variety of AI resources, such as databases, object-oriented method [82], and
applications. the unified manufacturing resource model [83]. In contrast to
An example of such a system is a personalized product the conventional resource description methods, the ontology-
identification based on deep learning image recognition. First, based description is one of the most prominent methods. An
a multiple agent subsystem is constructed for producing per- ontology represents an explicit specification of a conceptual
sonalized products. Then, a single agent records image or model [84], by way of a classical symbolic AI reasoning
video data at different stages of the CM process. Next, the method (i.e., an expert system). Modeling an application
edge computing server runs the image recognition algorithms, domain knowledge through an expert system provides a con-
such as a convolutional neural network (CNN), R-CNN, Fast ceptual hierarchy that supports system integration and inter-
R-CNN, Faster R-CNN, YOLO, or Single Shot Detection operability via an interpretable way [85], [86].
(SSD), all of which have demonstrated their advantages in In our previous work [87], the device resources of smart
computer vision tasks. The identification results are rapidly manufacturing were integrated by the ontology-based inte-
transferred to the devices. When the single edge computing gration framework, to describe the intelligent manufacturing
server cannot meet the real-time requirements, the multiple resources. The architecture consisted of four layers, namely,
agent edge servers may work collaboratively to complete the the data layer, the rule layer, the knowledge layer, and the
specific tasks such as product identification. Indeed, during resource layer. The resource layer represented the entity of
PROCEEDINGS OF THE IEEE, FEBRUARY 2023 7

intelligent manufacturing equipment (e.g., manipulators, con- Wireless link Information extraction process
Wired link
veyor belt, PLC), which was essentially the field device. Edge computing node Data Information knowledge
The knowledge layer was essentially the information model
composed of intelligent devices, which was integrated into the AI

domain knowledge base through the OWL language [88]. The Computing facilities

rule layer was used to gather the intelligent characteristics of

intelligent equipment, such as decision-making and reasoning.
Sensors
The data layer included a distributed database for real-time Sensors Sensors
Sensors
data storing, and the relational database was used to associate
the real-time data.
Due to the massive amount of data generated from man-
ufacturing devices, it is nearly impossible to consider all
Sensors
the manufacturing device resources. Thus, it is important to
construct a new manufacturing description model to realize
the reconfiguration of various manufacturing resources. In this
model, the resources can be easily adjusted by running the Fig. 5. Intelligent sensing based on edge AI computing. Sensor nodes
collect ambient data while edge computing notes can preprocess and
model. Therefore, ontology modeling is conducted on a device cache the collected data, which can be further transferred to remote
and related attributes of an intelligent production line in CM. cloud servers for in-depth data analysis.
The manufacturing resources are mapped to different functions
with different attributes. For instance, the time constraint of
a product manufacturing is divided into a number of time to the edge server or remote data centers. The edge computing
slots with consideration of features of processes and devices. servers (nodes) include the stronger processing units, larger
Then, the CM resources of a product can be mapped into memories, and storage space. These servers are connected
computing, cutting, conveying, and other functions with the to different sensors nodes and deployed in approximation
limited time slot, as shown in Fig. 4. Next, a customized to the devices, with the provision of the data storage and
product can be produced by different devices with different smart computing services by running different AI algorithms.
time constraints. Accordingly, a product can be represented Meanwhile, the edge computing servers are interconnected
by ontology functions. with each other to exchange information and knowledge.
Meanwhile, after making a reasonable arrangement of dif- Especially, the sensing parameters can be adjusted in a flex-
ferent manufacturing functions at different time slots, a DL ible monitoring subsystem in the manufacturing environment,
algorithm can forecast time slots of working states. The time according to different application requirements and the task
slots of working states are important for the reconfiguration of priority. To achieve a rapid response high priority system, the
manufacturing resources. Therefore, in actual applications, a edge AI servers should have access to the sensing data, and
different attribution of a device and customized products can capability to categorize the status of the CM environment.
be employed as a constraint condition. This can be done by processing the data features through ML
classification algorithms such as logistic regression, SVM, and
classification trees. When the data is out of the safety range, a
C. Edge Computing in Intelligent Sensing certain risk may exist in the manufacturing environment. For
The concept of ubiquitous intelligent sensing is a corner- instance, if an anomalous temperature event would happen in
stone of smart manufacturing in the Industry 4.0 framework. the CM area, the edge server could drive the affected nodes to
Numerous research studies have been conducted in monitoring increase their temperature sensing frequency, in order to obtain
manufacturing environments [89]–[91]. Most published results more environmental details and to make proactive forecasts
adopt a precondition-sensing system that only accepts a static and decisions.
sensing parameter. Obviously, this results in inflexibility and The environmental sensing data delivery is another im-
the sensing parameters are difficult to be adjusted to fulfill portant component in CM. With the development of smart
different requirements. Second, although some studies claim manufacturing, a sensing node not only performs sensing but
dynamic parameter tuning, the absence of a prediction function also transmits the data. With the proliferation of massive
is still an issue. Existing environment sensing (monitoring) sensing data, sensor nodes have been facing more challenges
cannot adjust the sensing parameters in advance to achieve a from the perspectives of data volume and data heterogeneity.
more intelligent manufacturing response. In order to collect environment data effectively, it is needed
As shown in Fig. 5, the manufacturing environment intel- to introduce new AI technologies. The sensor nodes can
ligent sensing based on the edge AI computing framework realize intelligent routing and communications by adjusting
includes two components: sensors nodes and edge computing the network parameters, assigning different network loads and
nodes [92]. Generally, smart sensor nodes are equipped with priorities to different types of data packets. With this optimized
different sensors, processors, and storage and communication sensing transfer strategy, the AI methods can make adequate
modules. The sensors are responsible for converting the phys- forecasts with reduced bandwidth usage.
ical status of the manufacturing environment into digital sig- Discussion. We present intelligent manufacturing devices
nals, and the communication module delivers the sensing data from edge computing-assisted intelligent agent construction,
PROCEEDINGS OF THE IEEE, FEBRUARY 2023 8

access points, network gateways, network switches, network

Coordination node
AI models routers, and terminals), the CM equipment and devices are
closely interconnected with each other and can be supported
Computing center
by edge or cloud computing paradigms [97]. Taking full
advantage of AI-driven software-defined industrial networks,
and relevant networking technologies is an important method
AI models
to achieve intelligent information sharing in CM [98], [99].
In conventional industrial networks, network control func-
SDN controller tions have been fixed at network nodes (e.g., gateways,
routers, switches). Consequently, industrial networks cannot
SDN router
be adapted to dynamic and elastic network environments,
especially in customized manufacturing. The software-defined
Control flow
Data center networking (SDN) technology can separate the conventional
Data flow network into the data plane and the control plane [100]. In
Fig. 6. Software-defined industrial networks consist of network this manner, SDN can achieve flexible and efficient network
coordinated nodes, SDN routers, SDN controllers, data centers, and control for industrial networks. It has been reported that a
cloud computing servers which can support intensive computing tasks software-defined industrial network can increase the flexibility
of AI algorithms. of a dynamical network system while decreasing the cost of
constructing a new network infrastructure [101].
The introduction of AI technologies to SDN can further
manufacturing resource description based on ontology, and bestow network nodes with intelligence. As demonstrated in
edge computing in intelligent sensing. It is a challenge to Fig. 6, AI technologies are introduced into traditional SDN
upgrade the existing manufacturing devices to improve the so as to form a novel software-defined industrial network
interoperability and the inter-connectivity. Retrofitting instead (SDIN). The proposed SDIN contains a number of mapping
of replacing all the legacy machines may be an alternative network nodes, SDIN related devices, data centers, and cloud
strategy in this regard. The legacy manufacturing equipment computing servers to support intensive computing tasks of
can be connected to the Internet by additively mounting sen- AI algorithms. Manufacturing devices are connected by their
sors or IoT nodes in approximation to existing manufacturing communication modules, and they are mapped to different
devices [93], [94]. Moreover, monitors can be attached to network terminal nodes. On the SDIN level, key devices
existing machinery to visualize the monitoring process. It is such as coordinated nodes and SDIN controllers construct the
worth mentioning that retrofitting strategies may apply for the SDIN layer. First of all, coordinated nodes are linked with the
sensing or monitoring scenarios while they are not suitable or ordinary nodes, and deliver network control messages from
less suitable for the cases requiring to make active actions other SDN devices. Second, the SDN routers are the key
(like control or movement). Furthermore, a comprehensive devices that realize the separation of data flow and control
plan should be made in advance rather than arbitrarily adding flow of the entire manufacturing network. In addition, the
sensors to the existing production line [95]. Retrofitting strate- SDIN controller is directly connected to the AI server, and
gies also have their limitations, such as a limited number of the AI server provides network decisions directly to the SDN
internal physical quantities can be monitored in a retrofitted controller.
asset with respect to a newly-designed smart machine. In the network information process, AI algorithms, such as
deep neural networks, reinforcement learning, SVM, and other
V. I NTELLIGENT I NFORMATION I NTERACTION I N A ML algorithms can be executed in a server according to the
S MART FACTORY state of the network devices, such as load information, commu-
In the CM domain, the information exchange system needs nication rate, received signal strength indicator, and other data.
to fulfill the dynamic adjustment of network resources so Then, the AI server returns the optimized results to the SDN
as to produce multiple customized products in parallel. In controller, and the results are divided into different instructions
order to obtain optimal strategies, many studies have focused for different network devices in the light of a specific CM task.
on this topic, and proposed insightful algorithms as well as Following the above steps, the SDN controllers send a set of
strategies [96]. However, there are still two open issues: a instructions to the routers and the coordination nodes. Finally,
network framework to dynamically adjust network resources, network terminals readjust the related parameters, (e.g., com-
and the end-to-end (E2E) data delivery. In this section, we munication bandwidth, transmitted powers) to complete the
present software-defined industrial networks and AI-assisted data communication process.
E2E communication to tackle these two challenges. Intelligent optimization algorithms (e.g., ant colony or
particle swarm optimization) can find optimal data transfer
strategies – based on the network parameters provided by
A. Software-defined industrial networks the SDIN, or given by the constraints of data interaction.
Industrial networks are a crucial component in CM, and cus- These algorithms can adjust the latency and energy consump-
tomized product manufacturing groups can be understood as tion requirements. Thus SDIN can improve the information
subnets. Via an industrial network (consisting of base stations, management processes within a CM industry framework,
PROCEEDINGS OF THE IEEE, FEBRUARY 2023 9

reducing the cost of dynamically adjusting or reconfiguring these methods still lack flexibility, and do not fulfill the
network resources. Moreover, it can improve and propel the emerging requirements of industrial applications. Generally,
whole manufacturing intelligence. Additionally, by adopting industrial E2E communications can be divided into two cat-
an AI-assisted SDIN, the production efficiency can be further egories: periodic communications and aperiodic communica-
improved. tions. Similarly, AI plays an important role in the MAC layer.
An example is a hybrid approach that combines the CSMA and
TDMA, with an intelligent optimization method, to improve
B. End-to-End communication
the efficiency of the E2E communication. In particular, the two
End-to-end (E2E) or device-to-device communication be- categories of communication requirements (high and low real-
tween manufacturing entities is a convenient communication time or periodic and aperiodic communications) are classified
strategy in industrial networks [102], [103]. E2E communica- by the AI-based method (e.g., naı̈ve Bayes). Next, an improved
tion provides communication services with lower latency and hybrid MAC is constructed on top of the CSMA and TDMA.
higher reliability, as compared to a centralized approach [104]. TDMA and CSMA schemes deal with the periodic and aperi-
With effective information interaction via E2E communication, odic data flows of the E2E communications. The size of this
the entire system can achieve full connectivity. In the context proposed mechanism can be adjusted in accordance with the
of CM, data transmission with different real-time constraints AI-optimized results of a real application.
has become a critical requirement [105]. The E2E industrial The network routing is also another key component of
communication approach optimizes the usage of network E2E communications. The key node of the routing path
resources (e.g., network access and bandwidth allocation) plays an important role in the E2E communications as well.
through data communication of varying latency [106], [107]. However, the performances of routing key nodes are impacted
Meanwhile, in order to realize the E2E communication in the by the workload; for instance, the amount of forwarded data.
industrial domain, a hybrid E2E communication network – Similarly, AI plays a significant role in the routing layer. The
based on the AI technology and SDIN – is here constructed predicted state parameters, such as communication rate and
by exploiting different media, communication protocols, and network loads of key nodes, can be obtained by using historical
strategies. The hybrid E2E-based communication mechanism data from the network node status by algorithms, such as deep
with the AI assistance can be divided into three layers: the neural networks or deep reinforcement learning (e.g. deep Q-
physical layer, the media access controlling (MAC) layer, and learning).
the routing layer.
In the physical layer, according to the advantages and VI. F LEXIBLE M ANUFACTURING L INE
disadvantages of the involved communication technologies,
A flexible manufacturing production line realizes customiza-
different communication media include optical fiber [108],
tion. AI-driven production line strategies and technologies,
network cable [109], and wireless radio [110]. Generally, in-
such as collective intelligence, autonomous intelligence, and
dustrial communications can be divided into wired or wireless
cross-media reasoning intelligence, have accelerated the global
communications. On the one hand, wired communications typ-
manufacturing process. Therefore, the subjects of cooperative
ically exhibit high-stability and low-latency. A representative
operation between multiple agents, dynamic reconfiguration
case is an industrial Ethernet, which is based on a common
of manufacturing, and self-organizing scheduling based on
Ethernet and runs improved Ethernet protocols, such as Ether-
production tasks are presented in this section.
CAT [111], EtherNet/IP [112], and Powerlink [113]. On the
other hand, wireless networks have been adopted in applica-
tions with relatively high flexibility [114], [115]. Nowadays, A. Cooperative multiple agents
an increasing number of mobile elements have been incor- Cooperation among multiple agents is necessary to dynam-
porated in manufacturing systems; therefore, wireless media ically construct collaborative groups for the completion of
has been widely exploited in mobile communications [116]. customized production tasks [120]. As discussed in Section IV,
Conventional strategies on fixed and static industrial networks multiple agents with edge computing provide a better option
may not fulfill the emerging requirements on flexible network than a single device to build a collaborative operation to realize
configurations. The AI and related technologies, such as deep CM [79], [121]. Therefore, by combining the edge computing-
reinforcement learning, optimization theory and game theory, assisted intelligent agents and different AI algorithms, a novel
can play significant roles in improving the communication cooperative operation can be constructed as shown in Fig. 7.
efficiency in the physical layer, e.g., determining the optimal The strategy of cooperative operation by multiples agents can
communication between wired and wireless networks while be divided into the order of submission, task decomposition,
achieving a good balance between network operational cost cooperative group, and subgroup assignment.
and network performance. The working process of a flexible manufacturing produc-
In the MAC layer, different devices have different require- tion line can be described as follows. First, according to
ments for E2E communications according to their specific the customers’ requirements, the CM product orders are is-
functions. Although many different MAC protocols have been sued to the manufacturing system through the recommender
proposed (e.g., CSMA–Carrier Sense Multiple Access) [117], system. After receiving the product orders, the AI-assisted
CDMA–Code Division Multiple Access) [118], TDMA–Time task decomposition algorithms take the product orders as the
Division Multiple Access) [119] and their improved versions, input, the device working procedure as the output, and the
PROCEEDINGS OF THE IEEE, FEBRUARY 2023 10

Working step

Product Edge agent

Users order

tep
i n gs g
rk in
Wo merg Fine group

Management
Computing server Wo
rk
me ing AI methods
Product task rgi ste
ng p

Customers

Process libary
Process analysis Edge agent Fine group
Order submmission Task decomposition Main cooperative group Sub cooperative group

Fig. 7. Cooperative multiple agents. The strategy of cooperative multiple agents can be divided into i) the order of submission, ii) task
decomposition, iii) cooperative group and iv) subgroup assignment.

product manufacturing time as a constraint; these algorithms ment. Then, according to the manufacturing task attributes,
are mainly executed at the remote cloud server. A product multiple agents complete the producing task. During this
order can be divided into multiple subtasks, which are sent to period, the corresponding device agents send their status data
all the agents via the industrial network. After the negotiation, to edge servers timely, and the manufacturing process can
agents return the answers to the edge server, which handles the be monitored by analyzing these data in the entire system.
working subtasks according to corresponding conditions and In contrast to the AI-driven cooperative operation between
constraints. Next, the AI-assisted cost-evaluation algorithm multiple agents, conventional methods often rely on human
calculates the cost of a producing group (i.e., cooperative operators who participate in the whole process or computer-
manufacturing group) from the historical data. Then, the edge assisted operators also requiring human interventions. These
agents intelligently select suitable device agents to finish the methods inevitably result in huge operational expenditure.
product order after considering the whole cooperative group
performances, such as producing time and product quality. B. Dynamic reconfiguration of manufacturing Systems
Moreover, the edge agents send the selection result to the
With the scientific development of the industrial market
device agents, which are chosen to take part in the producing
and manufacturing equipment, different industrial devices
order. The main cooperative group is constructed based on the
present different performance requirements representing mul-
working steps.
tiple function trends [123]. For instance, the latest Computer
The main cooperative group may not be well suited for real Numerical Control (CNC) machine tool can complete a wide
applications, especially for complicated CM tasks. Therefore, range of tasks, from lathing to milling functions. On the
an AI-based method for constructing a suitable-size coop- other hand, a dedicated manufacturing line does not meet
erative subgroup is an important step for dealing with the new industrial requirements, especially for customized produc-
mentioned problem. A possible strategy is to use cognitive ap- tion [124]. The trend today is towards reconfiguration and re-
proaches such as the Adaptive Control of Thought—Rational programmability of manufacturing processes [125]. Although
(ACT-R) model [122]. These subtasks cooperative groups can several studies have investigated the problem and presented
be mapped to the digital space (i.e., edge agent) and form meaningful results [126], [127], most of them lack intelligent
even lower level subgroups, all interconnected by the conveyor, design to fulfill the emerging requirements of dynamic recon-
logistics systems, and industrial communication systems. Each figuration of manufacturing systems, especially for customized
subgroup can delegate the same edge agent, to provide the manufacturing. In particular, the work [126] focuses on the
management and customers with manufacturing services. The communications between agents while [127] investigates the
characteristics of the subgroups are partly derived from the relationship between manufacturing flexibility and demands.
process constraints and the physical constraints of the plant. Thus, AI technologies have seldom been adopted in these
In principle, the higher the constrains the deeper the task tree studies. At present, ontology (as shown in Section IV) of-
will expand, from more abstract tasks to particular atomic fers insights into dynamic reconfiguration of manufacturing
targets achievable by the present devices. This structure can resources [104], [128].
be replicated with a probabilistic graphical model or with a A schematic of the dynamic reconfiguration process based
fuzzy tree. on the ontology inference is shown in Fig. 8. Each customized
After all the agents have been assigned with subtasks, product invokes several processing procedures. First, a person-
they form two level-cooperative groups. The formation of alized product manufacturing-related device (such as cutting,
these cooperative groups is beneficial to resource manage- materials handling device) is selected by ontology reasoning
PROCEEDINGS OF THE IEEE, FEBRUARY 2023 11

Feature 1 Algorithm 1
Computing AI
AGV Machine Monitor Machine Robot Robot Task set plaform methods
Knowledge abstraction 1 Feature Algorithm

Function Function Function Function Function Function

Production Task1 Task2 Task3 Task4

Device 2 Device Device Device Device Device tasks

Re

Re
3

pla
Period slots Slot Slot Slot Slot Slot Slot

pla
e c

e c
Device Device Device Device Idle Device Device
Idle Idle Busy
Idle Idle Busy Idle
Idle
Broken down

Idle
Device Device Device Device Device Idle Busy Idle Busy
Idle Busy Busy
Idle

Ontology inference Heavy load

Busy
Idle Busy Busy
Idle Busy Busy
Idle Busy
Fig. 8. Dynamic reconfiguration of manufacturing resources: 1)
device selection by ontology reasoning based on the device function;
2) CM production line is constructed; 3) automatic switching to other
Busy Busy
Idle Idle Busy
Idle Idle Busy
Idle
devices from heavy-load devices or broken devices.

Busy
Idle Idle Busy
Idle Idle Busy
Idle Idle

based on the device function. Then, the second selection of the

devices involved in the manufacturing is finished according Fig. 9. Self-organization of schedules of multiple production tasks
consists of three steps: task analysis, task decomposition and task
to ontology results with respect to the related manufacturing execution.
process, the manufacturing time, manufacturing quality, and
other parameters of a device. Finally, a CM production line
is constructed. Specifically, when the production line receives
As shown in Fig. 9, in terms of initialization, when a new
a production task, the raw material for a specific type of
production task is processed by the multi-tasking production
products is delivered from an autonomous warehouse. Then,
line, the new production tasks are divided into multiple steps
the production line completes the manufacturing tasks in the
by an AI-based method executed at the cloud. Additionally,
process sequence. Furthermore, when one of the manufac-
according to the process lead time, the producing period can
turing devices breaks down during the process, automatic
be decomposed into time slots of different lengths. Moreover,
switching of the related machining equipment by ontology
for one working step in a time slot, edge agents select all idle
inference is conducted. Meanwhile, the reasoning mechanism
device agents by comparing the mapping relationship between
reflects the reconstruction function of a flexible production of
working steps and device agent functions. This processed
the production line.
time slot information is then broadcasted to all the agents
The presented approach leads to optimal process planning
simultaneously. Then, idle device agents choose the working
and functional reconstruction. Besides, it shows the strengths
step by price bidding or negotiating with others, according
of ontology modeling and reasoning. In practice, only ontology
to the manufacturing requirements and self-conditions (e.g.,
and constraints need to be established according to the above
manufacturing time and quality). These results are broadcasted
description. According to Jena syntax1 , the corresponding API
to other agents, including different servers. Next, the edge
interface can be invoked to meet the task requirements of this
agents update the working state of the idle device agent in
scenario. In the future, other AI algorithms are expected to be
the corresponding time slot. These procedures are repeated
integrated with ontology inference.
until the new task steps are allocated within a certain or fixed
time. Lastly, multiple agents finish the scheduling of the new
C. Self-organizing Schedules of Multiple Production Tasks production task in a self-organization manner.
Self-organization of schedules with multiple agents and
Product orders generally have stochastic and intermittent time slots can effectively complete simultaneous production
characteristics as the arrival time of orders is usually un- tasks using a flexible production. Furthermore, production
certain [129]. This may result in having to share produc- line efficiency is improved. Consequently, all manufacturing
tion resources among multiple tasks. Therefore, creating self- resources, including different devices and subsystems, are
organizing schedules with a time slot based on multiple agents more intelligent to finish the multiple production tasks au-
for multiple production tasks is paramount [72]. The mech- tonomously. In contrast, conventional methods often require
anism of self-organizing schedules for multiple production huge human resources in scheduling and planning production
tasks can be divided into three steps: task analysis, task tasks [130]. Despite the recent advances in computer-aid meth-
decomposition, and task execution. ods [131], they still require substantial human interventions
1 Jena syntax defines a set of rules, principles, and procedures to specify
and cannot meet the flexible requirements.
the semantic web framework of Apache Jena (https://jena.apache.org/getting However, we have to admit that AI-driven self-organization
started/index.html). of schedules does not get rid of humans in the loop of
PROCEEDINGS OF THE IEEE, FEBRUARY 2023 12

Logistics system Camera Cloud server

Producing group
Grasping robot
Moving track

Producing Product Data base RFID

group
Sensor Optoelectr-
onic switch Box delivery robot
Robot
Sensor
ES (g1)
Feeding robot
ES(g3) Industrial
cloud
AGVs
Grasping
Feeding robot
Machine robot
ES(g2) Sensor
AGV
ES(g4) Edge server Candy Robot
Customers
Sensor arm
Conveyor belt
Procuct controller PLC
Wireless module
Producing group Mobile element Producing group Non-packed box

(a) Framework for case study (b) Implementation scenario of CM

Fig. 10. The framework of customized candy wrapping smart production. (a) The framework includes CM devices, the industrial network,
a conveyor, and a cyber-physical system. (b) The implementation of customized candy wrapping line consists of diverse devices.

the entire production process. The main goal of AI-driven layers, all of which are connected by the industrial IoT with
methods is to save unnecessary human resource consumption different link functions.
and mitigate other operational expenditures. In this manner, The first layer is the device layer, including five robots, two
human workers can concentrate on planning and optimizing AGVs, a conveyor system, and a warehouse. The device layer
the overall production procedure instead of conducting te- performs the basic functions of an intelligent production line,
dious and repetitive tasks. Meanwhile, an appropriate human such as carrying, clipping, loading raw material, and unloading
intervention is still necessary when full automation is not final products. Cognitive robots can be vertically integrated
achievable or is partially implemented. In this sense, AI-driven into a cyber-physical system in smart manufacturing [132].
methods can also assist human workers to give intelligent
The industrial network layer (second layer) plays a key
determinations.
role in the information interaction and intelligent connection
of different communication technologies – e.g., industrial
VII. C ASE S TUDY
wireless local networks (Wi-Fi, ZigBee), industrial Ethernet,
In the section, a case study is presented, showcasing the industrial NFC (Near Field Communication), and mobile
following aspects: prototype platform construction, big data communications. There are three sub-networks for finishing
analysis using AI technology for preventive maintenance, and different latency communication functions [133]. Specifically,
cloud-assisted customization service. wired industrial networks are employed as the inner equipment
to achieve higher real-time performance. In this aspect, the
A. Prototype platform construction wireless industrial networks were mainly adopted in the mon-
We implement a prototype of the AI-assisted CM frame- itoring system while the mobile wireless local networks also
work, namely a customized candy wrapping production line. helped to achieve higher-level flexibility [134]; for instance,
As shown in Fig. 10a, the framework includes the following mobile wireless nodes were dynamically deployed to monitor
components: CM devices, the industrial network, a conveyor, the industrial environment status.
and a cyber-physical system. All components are connected The third layer is the computing layer, which is mainly
through the industrial network, i.e., OPC Unified Architecture involved with the analysis, computing, and knowledge mining
(OPC UA) and Data Distribution Service (DDS). Fig. 10b of big data. A commercial solution has been adapted to build a
illustrates the implementation of customized candy wrapping cloud platform. XenServer developed by Citrix is used to real-
line. The candy packing line mainly includes the production ize the virtualization of the server cluster consisting of multiple
stations and the logistics transmission system. In the logis- virtual machines and the management of virtual machines.
tics transmission system, the packing box is continuously Meanwhile, we also establish a big data analytics framework,
transferred by the conveyor belts or AGVs. The production which is a software architecture based on a cloud platform for
stations are distributed discretely between the mainline and big data storage and distributed computing. Apache Hadoop,
the branch line, and RFID tags are adopted to obtain the an open-source solution, is used to provide the non-relational
operation information. The equipment types of the production database HBase and the computing architecture of YARN
stations include the materiel feeding, candy grasping, box (Yet Another Resource Negotiator). On top of the big data
delivery, and finished goods storage. The presented system framework, the AI-assisted optimization algorithms (such as
meets the requirements of small-batch production. In partic- deep learning models) have been deployed to realize intelligent
ular, the packaged candy followed the taste, flavor and color applications. To meet different latency requirements in the
preferences given by the customers. The system includes four platform, a hybrid computing paradigm, orchestrating the
PROCEEDINGS OF THE IEEE, FEBRUARY 2023 13

cloud and edge computing paradigms, is adopted. Explicitly, maintenance. The data pipeline consisted of three main steps,
edge computing is used to deal with real-time tasks, while data collection, data processing, and data mining and analysis
cloud computing was focused on completing time-insensitive using AI.
tasks, such as historical data processing. The deployment of From the perspective of preventive maintenance, the related
edge computing enables cloud service characteristics such as data collection represented the fundamental step in the fol-
mobile computing, scalability, and privacy policy [135]. lowing analysis. Different data, including the environmental
The fourth layer is the service layer. In this level, a large data, product processing data, device working status, and
number of manufacturing resources are stored at the cloud device logs were collected and transmitted to the computing
platform, which offers different AI services. Pattern recogni- servers, such as cloud and edge computing servers, through
tion, accurate modeling, knowledge discovery, reasoning, and the industrial network or industrial IoT. Edge computing and
decision-making capabilities are provided. cloud computing paradigms presented in [145] were employed
The working process of the platform is as follows. First, to address elastic and virtual manufacturing resources, which
customers select candy products according to their preferences, provided opportunities for real-time monitoring of produc-
which included the color, taste, quantity, and variety of the can- tion Key Performance Indicators (KPIs) and smart inven-
dies in an AI recommender web service system. Then, these tory management. The computing servers were responsible
proposed schemes and candy order parameters are delivered for data processing and for device maintenance. The data
to the manufacturing cloud through the web service, and the includes different types of manufacturing device-related data.
web server was connected to the cloud via the Internet. The First, redundant and misleading data were removed during
related product orders are created according to the submitting the data collection process. Then, the abstract data for real-
information. These orders were decomposed into different time or historical big data analysis was used for equipment
working steps by the ontology-based manufacturing system. maintenance. The AI-based techniques (e.g., deep learning)
Next, the multiple agents completed the production tasks in have been regarded as the most effective way for system and
a self-organized way. After obtaining the working steps, the equipment fault recognition based on big data analysis, so
manufacturing devices are assembled into collaborative groups these techniques were adopted in this step. Consequently, the
to finish all tasks. Thereafter, the platform finished the candy manufacturing maintenance knowledge base was built on top
wrapping task. of big data. Note that AI techniques play an important role in
During the product manufacturing process, the manufactur- extracting knowledge from massive and heterogeneous manu-
ing data is collected by sensors and is then transmitted to the facturing data, consequently achieving the intelligence [146].
cloud or nearby edge servers. The analyzed results provide key The manufacturing data includes 1) structured data such as
information for product monitoring. More importantly, these data stored in a rational database and 2) non-structured data
results can be used to adjust the processes and procedures to such as text, documents, sound, image, and video. Diverse
ensure higher quality and increase the production efficiency AI algorithms such as ontology learning, natural language
of the whole system. The model-driven method with ontology processing, and deep learning can be leveraged.
proposed in [136] was used to achieve interoperability and Furthermore, in our previous work [144], a big data solution
knowledge sharing in a manufacturing system across multiple for active preventive maintenance in manufacturing environ-
platforms in the product lifecycle. When multiple tasks were ments was proposed. This approach essentially combined a
needed to be finished in the platform, the manufacturing re- real-time active maintenance mechanism with an offline pre-
source reconstruction methods were employed for production diction method. The real-time performance was considered as
scheduling. The cloud-based manufacturing semantic model the main feature of a manufacturing system, especially equip-
proposed in [137] was used to obtain general task construc- ment maintenance. Therefore, to achieve the system mainte-
tion and task matching. After implementation, three candy- nance tasks, the hybrid edge and cloud computing paradigm
wrapping tasks with ten different candies were processed in for big data analysis represented a better option for preventive
the AI-assisted platform at the same time, which represented maintenance. Specifically, the equipment maintenance tasks
a typical production line model for mass wrapping, and the can be divided into online and offline tasks. On the one
first-in-first-out (FIFO) scheme was adopted accordingly. hand, as edge servers being deployed close to the equipment
can provide low-latency (real-time) service, edge computing
was adopted for dealing with the data online. On the other
B. Big data analysis using AI technology for preventive main- hand, cloud servers have powerful computing capabilities,
tenance and offline big data processing was executed at the cloud
Preventive maintenance for smart manufacturing has re- layer. Moreover, the analysis result was delivered to the man-
ceived attention in the literature [138]–[143]. A system ar- agement or data visualization system. The segmented model
chitecture for an active preventive maintenance system was provided in [147] for preventive maintenance of semiconductor
proposed in [144]. Based on this architecture, an improved manufacturing equipment, including both parametric and non-
preventive maintenance mechanism was constructed by merg- parametric models, was used for preventive maintenance.
ing cloud computing and edge computing with deep learn- Consequently, big data with AI is a key technology for
ing, as shown in Fig. 11. The smart preventive maintenance equipment maintenance in smart manufacturing. Big data helps
system was composed of CM devices, industrial networks, to build comprehensive condition monitoring and prediction
big data processing centers, and applications for preventive systems, which can provide preventive maintenance scheduling
PROCEEDINGS OF THE IEEE, FEBRUARY 2023 14

Equipment
Equipment Condition
Condition Intelligent Intelligent
Fault prediction
Fault prediction
diagnosis
diagnosis monitoring
monitoring maintenance Scheduling

Smart preventive maintenance

Knowledge
Data Data library
Data analysis
Big data mining collection processing

Offline big data

Online big data Data Information knowledge Cloud

Intelligence

AI

Environment data Process data Working status Device logs Product data Customer data

Industrial networks

Manufacturing devices

Fig. 11. Big data analysis using AI technology for smart preventive maintenance. The smart preventive maintenance system consists of CM
devices, industrial networks, big data processing center, and applications for preventive maintenance. The data pipeline consisted of three
main steps, data collection, data processing, and data mining and analysis using AI.

according to the equipment status. Integrated with the AI- CS

based methods, big data analysis can construct a maintenance
knowledge library and decrease the cost of operation and Product Assisted Remote Experiences Social
maintenance management of a smart manufacturing system. recommendation design monitoring Feedback network

C. Cloud-assisted customization service AI Data Information Knowledge

Unlike traditional manufacturing, CM can provide cus- Collection Analysis
tomized services. In other words, customers can participate in Edge nodes Cloud servers
the process of intelligent manufacturing [148], [149]. Recently,
CM
cloud computing has been proven to provide support for
customers taking part in the production process and drive the
customization services in a seamless manner [133], [150], Product Processing Product Product Logistics Market
allowing different data services to be quickly accessed by design analysis manufacturing checking trends
customers. Fig. 12. The cloud-assisted customization services cover the entire
Therefore, the integration of cloud computing with cus- produce life cycle: early stage, middle stage and later stage.
tomization services can improve the user experience of cus-
tomization services. We name such integration of cloud com-
puting with customization services as cloud-assisted cus- customers can also design the manufacturing products without
tomization services. As shown in recent studies [151], [152], a professional background in a digital simulation environment,
cloud-assisted customization services are well suited to of- virtual reality (VR), and/or augmented reality (AR). In the
fer a better user experience to the customer. In particu- middle stage of the production, CM can provide other personal
lar, cloud-assisted customization services are user-centric, customization services. For instance, customers can remotely
demand-driven, and service-oriented. monitor the details of the production process via a digital twin
Based on the multiple different services provided by cloud or a virtual production line system. Then, logistics information
computing, the customization service can be achieved from the of different stages can be sent to specific users via cloud
production cycle perspective, as shown in Fig. 12. The entire computing and mobile Internet. At the later stage, customers
production cycle was divided into three stages: early stage, can provide feedback on their user experience; this feedback
middle stage, and later stage. In the early stage of product can be used to improve the manufacturing process.
production, customers can select more suitable products by An important difference between intelligent manufactur-
the intelligent recommender system. As for the intelligent ing and traditional manufacturing lies in the fact that users
recommender system, big data analysis is adapted to integrate can participate in the production process via cloud-assisted
the order data, the production data, and the packing line technologies in intelligent manufacturing. Through industrial
status. Spark MLlib is adapted to realize the personalized cloud computing, CM can provide customers and users with
recommendation, which has advantages on algorithm, experi- customization services. This type of customization service
ence, and performance. Moreover, using cloud-assisted design, is to motivate participation and construct a new production
PROCEEDINGS OF THE IEEE, FEBRUARY 2023 15

shopping experience. be made in AI and in AI/robotics integration. However, as

In terms of the market trend prediction, the more the shown in Section VII, CM is already realizable in restricted
market data gathered, the more accurate the results. Differ- environments.
ent algorithms can assist in the data-driven decision-making Extensive studies have been conducted in the respective
process. For example, shape mining is a framework based technologies to support CM [156]–[158]: intelligent connec-
on engineering design data [153] and has been applied to tivity, data-driven intelligence, cognitive Internet of things, and
passenger car design. Further, the apriori and C5.0 algorithms industrial big data. Unfortunately, most of these studies have
for data mining are able to extract rules from databases [154]. mainly focused on single-device intelligence. Although these
The platform used in the experiment, cannot only finish on- studies provide useful suggestions towards smarter devices,
demand candy production based on the customization param- there is a still huge gap between realistic smart devices and
eters but also adapt to changes in the market. The constructed the current solutions. We consider that an ideal method for
platform can increase the efficiency of the entire system realizing smart devices in manufacturing should integrate a
of candy wrapping. This methodology, in combination with series of methods to ensure the CM with edge intelligence that
the aforementioned decision-making algorithms, constitutes a offers adaptability and response to a wide range of scenarios.
proof of concept of a seamless CM production pipeline. These different methods or algorithms can be orchestrated by
Discussion. It is worth mentioning that the number of a central AI unit, which decides which algorithm is more
interactions between customization and production should also useful for a particular task (combining a symbolic AI approach
be limited although cloud-assisted customization services can with machine learning). Meanwhile, a good balance between
greatly improve user experience [155]. In practice, there is intelligence and cost is still a challenge that needs to be
a trade-off between customization interactions and product addressed in the future.
manufacturing. From the user experience perspective, the more
customization interactions lead to a better user experience. B. Information interaction in CM
However, the increased number of customization interactions Most manufacturing frameworks represent the distributed
may also increase the expenditure and prolong the produc- systems, where high-efficient communications between differ-
tion time of products, consequently affecting user experience. ent components are needed [159], [160]. On the one hand,
Therefore, it can enhance customization services to provide different components of CM need not only effective connec-
users with some well-designed product samples accompanied tions but also highly efficient information interactions. As
by tutorials in advance. shown in early parts of this paper, the CM systems must
incorporate different information interaction technologies and
VIII. C HALLENGES AND A DVANCES algorithms. On the other hand, due to the device hetero-
The convergence of AI technologies, IoT, SDN, and other geneity and the increased demands of communications, the
new ICTs in smart factories, while it can significantly increase CM information interactive systems have to transmit massive
the flexibility, intelligence, and efficiency of CM systems data with different latency requirements. Consequently, the
it also poses challenges. In this section, we discuss these networks of CM need to be optimized to fulfill the diverse
challenges as well as recent advances. requirements of different applications, such as media access,
moving handover, and congestion control. In other words, CM
needs as a basis of a highly efficient information interaction
A. Smarter devices in customized manufacturing system, which integrates multiple information technologies,
In a CM environment, the equipment does not only per- including the absorbing DL and other AI algorithms. This
form the basic functions of automation but also needs to be system particularly depends on efficient and reasonable in-
intelligent and flexible. From the point of view of stand-alone formation interaction. With the development of manufacturing
equipment, the devices should have the following functions: techniques, an increased number of studies have focused on
parameter sensing, data storage, logical inference, information the massive volume of communications and a large number of
interaction, self-diagnostics, hybrid computing support, and connections [161]. Based on network performance optimiza-
preventive maintenance. From the perspective of the physical tion methods, researchers have developed several industrial
layer, the realization of CM is increasingly complex, as stan- networks [162], [163]. As different information transfer pro-
dard devices or equipment cannot accomplish this complicated tocols exist with different requirements, such as time-sensitive
task. Therefore, devices need to have both collaborative ability and time-insensitive data transfer, the construction of a more
and swarm intelligence to collaborate with each to complete efficient interaction industrial network represents another im-
complex tasks. Devices with the above functions can meet the portant challenge of future research. Existing studies mostly
requirements of personalized production and adapt to the trend focus on high-speed communications, rather than efficient
of achieving the automation and intelligence of manufacturing. interaction with intelligent networks. These methods usually
Therefore, CM devices must be based on the integration only meet the basic requirements of industrial networks, such
and deep fusion of advanced manufacturing technologies, as bandwidth and delay.
automation technologies, information technology, image tech- Recently, wireless mobile communications (e.g., 5G sys-
nology, communication technologies, and AI technologies to tems) and optical fiber communications have achieved great
realize the local and swarm intelligence. In this aspect, to development [164], [165]. The 5G technology for smart fac-
be realizable in a generic context, further progress needs to tories is still in its infancy, and there are still a number of issues
PROCEEDINGS OF THE IEEE, FEBRUARY 2023 16

such as deployment, accessing and spectrum management to solution for SMEs is to outsource manufacturing data to cloud
be addressed. A hybrid industrial network with the latest services providers or Machine learning as a service (MLaaS)
communication technologies and AI (including deep learning, provider who can offer on-demand computing services. Never-
integrated learning, transfer learning, etc.) can be a promising theless, outsourcing confidential data to untrusted third parties
solution for information interaction, which should consider may increase the risks of security and privacy leakage. Thus,
different data flows according to different applications of it is a prerequisite to enforce privacy and security protection
intelligent CM. schemes on manufacturing data before outsourcing. Moreover,
the expenditure of system operating and training personnel
C. Dynamic reconfiguration of manufacturing resources should not be ignored in practical deployment.
Another challenge is the effective technology transfer from
Intelligent manufacturing, especially CM, involves a dy-
research institutions to enterprises. Technology transfer in-
namic reorganization of the available resources and extreme
volves many non-technical factors and multiple parties. The
flexibility. The essence of CM is to provide customized
non-technical issues of technology transfer include marketing
products, which have the characteristics of small-batch, short
analysis, intellectual property management, technical invention
processing cycle, and flexible production. Therefore, manufac-
protection, commercialization, and financial returns. Many
turing resources need constant readjustment and reorganiza-
frontier technological innovations often end up with unsuc-
tion [166]. In addition, customized products are thought of as
cessful technology transfer due to ignorance of the non-
a variety of processing crafts. Moreover, the increasing com-
technical factors [176]. One of the main obstacles in tech-
plexity of the industrial environments makes the management
nology transfer lies in the technology readiness level (TRL)
of the resources even more difficult. A factory nowadays is
gap between research and industrial practice. In particular,
a system consisting of many sub-systems, which can produce
research institutions often focus on research results at TRL 1-
emergent behaviors through the interaction of the subsystems.
3 implying basic feasibility and effectiveness while industrial
Therefore, the dynamic reconstruction of manufacturing re-
enterprises often require transferred technologies at TRL 7-8
sources is one of the main challenges towards achieving a
or even higher levels meaning prototype demonstration and
generic CM facility. In this aspect, a number of strategies have
real deployment [177]. We admit that there is still a long way
been proposed [167], [168] in this domain.
for the AI-assisted CM framework before reaching TRL 7-
Recently, knowledge reasoning, knowledge graph, trans-
8. The investigation of the technology transfer of AI-assisted
fer learning, and other AI algorithms have attained great
CM will be a future direction. Both researchers and industrial
progress [169]–[171]. In our opinion, hybrid AI methods
practitioners are expected to work together to realize AI-
combining the latest knowledge reasoning technologies with
assisted CM.
swarm intelligence can be a promising solution for dynamic
reconstruction, which may consider different application sce- IX. C ONCLUSION
narios of CM. Manufacturing-based process optimization with
Recent advances in AI technologies have had an impact
ML technologies may be one of the effective methods to
on the manufacturing industry, especially within customized
reorganize resources. Moreover, digital twin technologies may
smart manufacturing. In this article, AI-assisted customized
be the driving technologies to improve resource reconfigura-
manufacturing architectures – incorporating IoT, edge intelli-
tion [172]–[174].
gence, and cloud computing paradigms – have been proposed.
These key AI-enabled technologies have been validated in
D. Practical deployment and knowledge transfer an industrial packaging scenario. Further, each of the aspects
Although the AI-assisted CM framework is promising to composing these architectures have been carefully reviewed.
foster smart manufacturing, several challenges in industrial The fusion of AI and manufacturing provides a potential
practice arise before the formal adoption of this framework. solution for customized manufacturing. Future research will
The first non-technical challenge mainly lies in the cost of be directed towards tackling the challenges related to smart
upgrading existing manufacturing machinery, digitizing man- manufacturing devices, effective information interaction, dy-
ufacturing equipment, and purchasing computing facilities as namic reconstruction of manufacturing resources, and practical
well as AI services. This huge cost may be affordable for small deployment issues.
and medium-sized enterprises (SMEs). With respect to up-
grading manufacturing production lines for SMEs, retrofitting R EFERENCES
legacy machines can be an economic solution as discussed [1] H. Lasi, P. Fettke, H.-G. Kemper, T. Feld, and M. Hoffmann, “Industry
in Section IV-C. In particular, diverse sensors and IoT nodes 4.0,” Business & information systems engineering, vol. 6, no. 4, pp.
239–242, 2014.
can be attached to existing manufacturing equipment to collect [2] V. Roblek, M. Meško, and A. Krapež, “A complex view of industry
diverse manufacturing data. Those sensors and IoT nodes can 4.0,” Sage Open, vol. 6, no. 2, p. 2158244016653987, 2016.
be collected with the Internet so as to improve the interconnec- [3] A. Kusiak, “Smart manufacturing,” International Journal of Production
Research, vol. 56, no. 1-2, pp. 508–517, 2018.
tivity of legacy machines. For example, Raspberry Pi models [4] S. Wang, J. Wan, M. Imran, D. Li, and C. Zhang, “Cloud-based
mounted with sensors can be deployed in workrooms to collect smart manufacturing for personalized candy packing application,” The
ambient data [93], [175]. Besides hardware upgrading, soft- Journal of Supercomputing, vol. 74, no. 9, pp. 4339–4357, 2018.
[5] L. Li, “China’s manufacturing locus in 2025: With a comparison of
ware tools as well as AI services should be also purchased and “Made-in-China 2025” and “Industry 4.0”,” Technological Forecasting
adopted by manufacturing enterprises. Similarly, the economic and Social Change, vol. 135, pp. 66–74, 2018.
PROCEEDINGS OF THE IEEE, FEBRUARY 2023 17

[6] Y. Lu, “Industry 4.0: A survey on technologies, applications and open eration artificial intelligence and further on,” in 2017 5th International
research issues,” Journal of Industrial Information Integration, vol. 6, Conference on Enterprise Systems (ES). IEEE, 2017, pp. 311–318.
pp. 1–10, 2017. [30] J. Wan, J. Yang, Z. Wang, and Q. Hua, “Artificial intelligence for
[7] L. Atzori, A. Iera, and G. Morabito, “The internet of things: A survey,” cloud-assisted smart factory,” IEEE Access, vol. 6, pp. 55 419–55 430,
Computer networks, vol. 54, no. 15, pp. 2787–2805, 2010. 2018.
[8] C. El Kaed, I. Khan, A. Van Den Berg, H. Hossayni, and C. Saint- [31] M. A. K. Bahrin, M. F. Othman, N. N. Azli, and M. F. Talib, “Industry
Marcel, “Sre: semantic rules engine for the industrial internet-of-things 4.0: A review on industrial automation and robotic,” Jurnal Teknologi,
gateways,” IEEE Transactions on Industrial Informatics, vol. 14, no. 2, vol. 78, no. 6-13, pp. 137–143, 2016.
pp. 715–724, 2017. [32] M. Dopico, A. Gomez, D. De la Fuente, N. Garcı́a, R. Rosillo, and
[9] N. J. Nilsson, “Principles of artificial intelligence morgan kaufmann,” J. Puche, “A vision of industry 4.0 from an artificial intelligence point
LOS Altos, California, 1980. of view,” in Proceedings on the International Conference on Artificial
[10] S. J. Russell and P. Norvig, Artificial intelligence: a modern approach. Intelligence (ICAI). The Steering Committee of The World Congress
Malaysia; Pearson Education Limited,, 2016. in Computer Science, Computer ¡, 2016, p. 407.
[11] Y. Cao, S. Li, Y. Liu, Z. Yan, Y. Dai, P. S. Yu, and L. Sun, “A Compre- [33] N. Mircică et al., “Cyber-Physical Systems for Cognitive Industrial
hensive Survey of AI-Generated Content (AIGC): A History of Gen- Internet of Things: Sensory Big Data, Smart Mobile Devices, and Au-
erative AI from GAN to ChatGPT,” arXiv preprint arXiv:2303.04226, tomated Manufacturing Processes,” Analysis and Metaphysics, no. 18,
2023. pp. 37–43, 2019.
[12] M. Chen, S. Mao, and Y. Liu, “Big data: A survey,” Mobile networks [34] V. Vasiliou and D. Milner, “Computer-Integrated Manufacture for Cold
and applications, vol. 19, no. 2, pp. 171–209, 2014. Roll Forming,” in Advances in Manufacturing Technology. Springer,
[13] A. McAfee, E. Brynjolfsson, T. H. Davenport, D. Patil, and D. Barton, 1986, pp. 79–85.
“Big data: the management revolution,” Harvard business review, [35] I. El Hassani, C. El Mazgualdi, and T. Masrour, “Artificial intelligence
vol. 90, no. 10, pp. 60–68, 2012. and machine learning to predict and improve efficiency in manufactur-
[14] B. Chen, J. Wan, L. Shu, P. Li, M. Mukherjee, and B. Yin, “Smart ing industry,” arXiv preprint arXiv:1901.02256, 2019.
factory of industry 4.0: Key technologies, application case, and chal- [36] E. C. (EC), “EUROPE 2020: A strategy for smart, sustainable and
lenges,” IEEE Access, vol. 6, pp. 6505–6519, 2017. inclusive growth,” Working paper {COM (2010) 2020}, 2010.
[15] V. Modrak, Z. Soltysova, P. Semanco, and P. R. Sudhakara, “Produc- [37] Z. Ji, “Intelligent manufacturing - main direction of “made in china
tion scheduling and capacity utilization in terms of mass customized 2025”,” China Mechanical Engineering, vol. 26, no. 17, pp. 2273–
manufacturing,” in Advances in Manufacturing II. Springer, 2019, pp. 2284, 2015.
295–306. [38] J. Holdren, T. Power, G. Tassey, A. Ratcliff, and L. Christodoulou,
[16] J. R. Baldwin and D. Sabourin, “Impact of the adoption of advanced “A national strategic plan for advanced manufacturing,” US National
information and communication technologies on firm performance in Science and Technology Council, Washington, DC, 2012.
the canadian manufacturing sector,” OECD Science, Technology and
[39] L. Da Xu, W. He, and S. Li, “Internet of things in industries: A survey,”
Industry Working Papers, 2002.
IEEE Transactions on industrial informatics, vol. 10, no. 4, pp. 2233–
[17] H. Lightfoot, T. Baines, and P. Smart, “Examining the information 2243, 2014.
and communication technologies enabling servitized manufacture,”
[40] B. Chen, J. Wan, Y. Lan, M. Imran, D. Li, and N. Guizani, “Improving
Proceedings of the Institution of Mechanical Engineers, Part B: Journal
Cognitive Ability of Edge Intelligent IIoT through Machine Learning,”
of Engineering Manufacture, vol. 225, no. 10, pp. 1964–1968, 2011.
IEEE Network, vol. 33, no. 5, pp. 61–67, 2019.
[18] B.-h. Li, B.-c. Hou, W.-t. Yu, X.-b. Lu, and C.-w. Yang, “Applica-
tions of artificial intelligence in intelligent manufacturing: a review,” [41] S. Wang, J. Wan, D. Li, and C. Zhang, “Implementing smart factory of
Frontiers of Information Technology & Electronic Engineering, vol. 18, industrie 4.0: an outlook,” International Journal of Distributed Sensor
no. 1, pp. 86–96, 2017. Networks, vol. 12, no. 1, p. 3159805, 2016.
[19] W. Li, W.-j. Wu, H.-m. Wang, X.-q. Cheng, H.-j. Chen, Z.-h. Zhou, and [42] H.-N. Dai, H. Wang, G. Xu, J. Wan, and M. Imran, “Big data analytics
R. Ding, “Crowd intelligence in AI 2.0 era,” Frontiers of Information for manufacturing internet of things: Opportunities, challenges and
Technology & Electronic Engineering, vol. 18, no. 1, pp. 15–43, 2017. enabling technologies,” Enterprise Information Systems, 2019.
[20] Y. Pan, “Heading toward artificial intelligence 2.0,” Engineering, vol. 2, [43] H. Wang, Z. Liu, D. Peng, and Y. Qin, “Understanding and Learning
no. 4, pp. 409–413, 2016. Discriminant Features based on Multiattention 1DCNN for Wheelset
[21] Y.-t. Zhuang, F. Wu, C. Chen, and Y.-h. Pan, “Challenges and opportu- Bearing Fault Diagnosis,” IEEE Transactions on Industrial Informatics,
nities: from big data to knowledge in AI 2.0,” Frontiers of Information vol. 16, no. 9, pp. 5735–5745, 2020.
Technology & Electronic Engineering, vol. 18, no. 1, pp. 3–14, 2017. [44] X. Zhang, F. Chung, and S. Wang, “An Interpretable Fuzzy DBN-Based
[22] M. Mohammadi and A. Al-Fuqaha, “Enabling cognitive smart cities Classifier for Indoor User Movement Prediction in Ambient Assisted
using big data and machine learning: Approaches and challenges,” Living Applications,” IEEE Transactions on Industrial Informatics,
IEEE Communications Magazine, vol. 56, no. 2, pp. 94–101, 2018. vol. 16, no. 1, pp. 42–53, 2020.
[23] Z. Zhang, K. Long, J. Wang, and F. Dressler, “On swarm intelligence [45] Y. Zuo, “Prediction of consumer purchase behaviour using Bayesian
inspired self-organized networking: its bionic mechanisms, designing network: an operational improvement and new results based on RFID
principles and optimization approaches,” IEEE Communications Sur- data,” International Journal of Knowledge Engineering and Soft Data
veys & Tutorials, vol. 16, no. 1, pp. 513–537, 2013. Paradigms, vol. 5, no. 2, pp. 85–105, 2016.
[24] M. Martinez-Garcia and T. Gordon, “A new model of human steering [46] S. Shao, S. McAleer, R. Yan, and P. Baldi, “Highly accurate machine
using far-point error perception and multiplicative control,” in 2018 fault diagnosis using deep transfer learning,” IEEE Transactions on
IEEE International Conference on Systems, Man, and Cybernetics Industrial Informatics, vol. 15, no. 4, pp. 2446–2455, 2019.
(SMC), 2018, pp. 1245–1250. [47] J. Lee, H. Davari, J. Singh, and V. Pandhare, “Industrial artificial intel-
[25] F.-Y. Wang, X. Wang, L. Li, and L. Li, “Steps toward parallel ligence for industry 4.0-based manufacturing systems,” Manufacturing
intelligence,” IEEE/CAA Journal of Automatica Sinica, vol. 3, no. 4, letters, vol. 18, pp. 20–23, 2018.
pp. 345–348, 2016. [48] T. Hayhoe, I. Podhorska, A. Siekelova, and V. Stehel, “Sustainable
[26] M. Martı́nez-Garcı́a, Y. Zhang, and T. Gordon, “Memory pattern iden- Manufacturing in Industry 4.0: Cross-Sector Networks of Multiple
tification for feedback tracking control in human–machine systems,” Supply Chains, Cyber-Physical Production Systems, and AI-driven
Human factors, p. 0018720819881008. Decision-Making,” Journal of Self-Governance and Management Eco-
[27] Y. Peng, X. Zhai, Y. Zhao, and X. Huang, “Semi-supervised cross- nomics, vol. 7, no. 2, pp. 31–36, 2019.
media feature learning with unified patch graph regularization,” IEEE [49] V. Sze, Y. Chen, T. Yang, and J. S. Emer, “Efficient processing of deep
transactions on circuits and systems for video technology, vol. 26, no. 3, neural networks: A tutorial and survey,” Proceedings of the IEEE, vol.
pp. 583–596, 2015. 105, no. 12, pp. 2295–2329, 2017.
[28] Y.-x. Peng, W.-w. Zhu, Y. Zhao, C.-s. Xu, Q.-m. Huang, H.-q. Lu, Q.-h. [50] A. Caggiano, T. Segreto, and R. Teti, “Cloud manufacturing framework
Zheng, T.-j. Huang, and W. Gao, “Cross-media analysis and reason- for smart monitoring of machining,” Procedia Cirp, vol. 55, pp. 248–
ing: advances and directions,” Frontiers of Information Technology & 253, 2016.
Electronic Engineering, vol. 18, no. 1, pp. 44–57, 2017. [51] F. Tao and Q. Qi, “New IT driven service-oriented smart manufacturing:
[29] X. Yao, J. Zhou, J. Zhang, and C. R. Boër, “From intelligent manu- framework and characteristics,” IEEE Transactions on Systems, Man,
facturing to smart manufacturing for Industry 4.0 driven by next gen- and Cybernetics: Systems, vol. 49, no. 1, pp. 81–91, 2017.
PROCEEDINGS OF THE IEEE, FEBRUARY 2023 18

[52] P. Zheng, Z. Sang, R. Y. Zhong, Y. Liu, C. Liu, K. Mubarok, [72] S. Wang, J. Wan, D. Zhang, D. Li, and C. Zhang, “Towards smart
S. Yu, X. Xu et al., “Smart manufacturing systems for industry 4.0: factory for industry 4.0: a self-organized multi-agent system with big
Conceptual framework, scenarios, and future perspectives,” Frontiers data based feedback and coordination,” Computer Networks, vol. 101,
of Mechanical Engineering, vol. 13, no. 2, pp. 137–150, 2018. pp. 158–168, 2016.
[53] M. Wollschlaeger, T. Sauter, and J. Jasperneite, “The future of industrial [73] I. Kovalenko, D. Tilbury, and K. Barton, “The model-based product
communication: Automation networks in the era of the internet of agent: A control oriented architecture for intelligent products in multi-
things and industry 4.0,” IEEE industrial electronics magazine, vol. 11, agent manufacturing systems,” Control Engineering Practice, vol. 86,
no. 1, pp. 17–27, 2017. pp. 105–117, 2019.
[54] S. Ren, Y. Zhang, Y. Liu, T. Sakao, D. Huisingh, and C. M. Almeida, [74] B. Park and J. Jeong, “Knowledge-based multi-agent system for smart
“A comprehensive review of big data analytics throughout product factory of small-sized manufacturing enterprises in korea,” in Inter-
lifecycle to support sustainable smart manufacturing: A framework, national Conference on Computational Science and Its Applications.
challenges and future research directions,” Journal of cleaner produc- Springer, 2019, pp. 81–93.
tion, vol. 210, pp. 1343–1365, 2019. [75] B. Chen, J. Wan, A. Celesti, D. Li, H. Abbas, and Q. Zhang,
[55] Y. Lu and X. Xu, “Cloud-based manufacturing equipment and big data “Edge computing in IoT-based manufacturing,” IEEE Communications
analytics to enable on-demand manufacturing services,” Robotics and Magazine, vol. 56, no. 9, pp. 103–109, 2018.
Computer-Integrated Manufacturing, vol. 57, pp. 92–102, 2019. [76] W. Shi, J. Cao, Q. Zhang, Y. Li, and L. Xu, “Edge computing: Vision
[56] J. Lee, B. Bagheri, and H.-A. Kao, “A cyber-physical systems archi- and challenges,” IEEE internet of things journal, vol. 3, no. 5, pp.
tecture for industry 4.0-based manufacturing systems,” Manufacturing 637–646, 2016.
letters, vol. 3, pp. 18–23, 2015. [77] V. S. Shah, “Multi-agent cognitive architecture-enabled iot applications
[57] K. S. Aggour, V. K. Gupta, D. Ruscitto, L. Ajdelsztajn, X. Bian, of mobile edge computing,” Annals of Telecommunications, vol. 73, no.
K. H. Brosnan, N. C. Kumar, V. Dheeradhada, T. Hanlon, N. Iyer 7-8, pp. 487–497, 2018.
et al., “Artificial intelligence/machine learning in manufacturing and [78] C. Yang, S. Lan, W. Shen, G. Q. Huang, X. Wang, and T. Lin,
inspection: A GE perspective,” MRS Bulletin, vol. 44, no. 7, pp. 545– “Towards product customization and personalization in IoT-enabled
558, 2019. cloud manufacturing,” Cluster Computing, vol. 20, no. 2, pp. 1717–
[58] A. Kusiak, “Intelligent manufacturing: bridging two centuries,” Journal 1730, 2017.
of Intelligent Manufacturing, vol. 30, no. 1, pp. 1–2, 2019. [79] J. Wan, B. Chen, S. Wang, M. Xia, D. Li, and C. Liu, “Fog computing
[59] X. Li, D. Li, J. Wan, A. V. Vasilakos, C.-F. Lai, and S. Wang, “A for energy-aware load balancing and scheduling in smart factory,” IEEE
review of industrial wireless networks in the context of Industry 4.0,” Transactions on Industrial Informatics, vol. 14, no. 10, pp. 4548–4556,
Wireless networks, vol. 23, no. 1, pp. 23–41, 2017. 2018.
[60] L. Liao, C.-F. Lai, J. Wan, V. C. Leung, and T.-C. Huang, “Scalable [80] G.-J. Cheng, L.-T. Liu, X.-J. Qiang, and Y. Liu, “Industry 4.0 devel-
distributed control plane for on-line social networks support cognitive opment and application of intelligent manufacturing,” in 2016 inter-
neural computing in software defined networks,” Future Generation national conference on information system and artificial intelligence
Computer Systems, vol. 93, pp. 993–1001, 2019. (ISAI). IEEE, 2016, pp. 407–410.
[61] S. Zoppi, A. Van Bemten, H. M. Gürsu, M. Vilgelm, J. Guck, and [81] S. Lemaignan, A. Siadat, J.-Y. Dantan, and A. Semenenko, “MASON:
W. Kellerer, “Achieving hybrid wired/wireless industrial networks with A proposal for an ontology of manufacturing domain,” in IEEE
WDetServ: Reliability-based scheduling for delay guarantees,” IEEE Workshop on Distributed Intelligent Systems: Collective Intelligence
Transactions on Industrial Informatics, vol. 14, no. 5, pp. 2307–2319, and Its Applications (DIS’06). IEEE, 2006, pp. 195–200.
2018. [82] Y. Zhang, S. C. Feng, X. Wang, W. Tian, and R. Wu, “Object oriented
[62] J. Wan, S. Tang, Z. Shu, D. Li, S. Wang, M. Imran, and A. V. Vasilakos, manufacturing resource modelling for adaptive process planning,”
“Software-defined industrial internet of things in the context of industry International Journal of Production Research, vol. 37, no. 18, pp.
4.0,” IEEE Sensors Journal, vol. 16, no. 20, pp. 7373–7380, 2016. 4179–4195, 1999.
[63] J. Wan and M. Xia, “Cloud-assisted cyber-physical systems for the [83] P. Vichare, A. Nassehi, S. Kumar, and S. T. Newman, “A Unified Man-
implementation of industry 4.0,” Mobile Networks and Applications, ufacturing Resource Model for representing CNC machining systems,”
vol. 22, no. 6, pp. 1157–1158, 2017. Robotics and Computer-Integrated Manufacturing, vol. 25, no. 6, pp.
[64] A. H. Sodhro, S. Pirbhulal, and V. H. C. de Albuquerque, “Artificial 999–1007, 2009.
intelligence-driven mechanism for edge computing-based industrial [84] Y. Lu, H. Wang, and X. Xu, “Manuservice ontology: A product data
applications,” IEEE Transactions on Industrial Informatics, vol. 15, model for service-oriented business interactions in a cloud manufactur-
no. 7, pp. 4235–4243, 2019. ing environment,” Journal of Intelligent Manufacturing, vol. 30, no. 1,
[65] Q. Zhao, L. Wei, Z. Sheng, and W. Liu, “Research on service-oriented pp. 317–334, 2019.
application support software platform of manufacturing informatiza- [85] H.-K. Lin and J. A. Harding, “A manufacturing system engineering
tion,” in 2016 2nd IEEE International Conference on Computer and ontology model on the semantic web for inter-enterprise collaboration,”
Communications (ICCC). IEEE, 2016, pp. 1402–1406. Computers in Industry, vol. 58, no. 5, pp. 428–437, 2007.
[66] E. Puik, D. Telgen, L. van Moergestel, and D. Ceglarek, “Assessment [86] Y. Alsafi and V. Vyatkin, “Ontology-based reconfiguration agent for
of reconfiguration schemes for reconfigurable manufacturing systems intelligent mechatronic systems in flexible manufacturing,” Robotics
based on resources and lead time,” Robotics and Computer-Integrated and Computer-Integrated Manufacturing, vol. 26, no. 4, pp. 381–391,
Manufacturing, vol. 43, pp. 30–38, 2017. 2010.
[67] D. Calvaresi, M. Marinoni, A. Sturm, M. Schumacher, and G. Buttazzo, [87] J. Wan, B. Yin, D. Li, A. Celesti, F. Tao, and Q. Hua, “An
“The challenge of real-time multi-agent systems for enabling IoT ontology-based resource reconfiguration method for manufacturing
and CPS,” in Proceedings of the international conference on web cyber-physical systems,” IEEE/ASME Transactions on Mechatronics,
intelligence, 2017, pp. 356–364. vol. 23, no. 6, pp. 2537–2546, 2018.
[68] Weiming Shen, Lihui Wang, and Qi Hao, “Agent-based distributed [88] C.-h. Liao, Y.-f. Wu, and G.-h. King, “Research on learning owl
manufacturing process planning and scheduling: a state-of-the-art sur- ontology from relational database,” in Journal of Physics: Conference
vey,” IEEE Transactions on Systems, Man, and Cybernetics, Part C Series, vol. 1176, no. 2. IOP Publishing, 2019, p. 022031.
(Applications and Reviews), vol. 36, no. 4, pp. 563–577, 2006. [89] W. Li and S. Kara, “Methodology for monitoring manufacturing
[69] Z. A. Khan, M. T. Khan, I. Ul Haq, and K. Shah, “Agent-based fault environment by using wireless sensor networks (WSN) and the internet
tolerant framework for manufacturing process automation,” Interna- of things (IoT),” Procedia CIRP, vol. 61, no. Supplement C, pp. 323–
tional Journal of Computer Integrated Manufacturing, vol. 32, no. 3, 328, 2017.
pp. 268–277, 2019. [90] Z. Sheng, C. Mahapatra, C. Zhu, and V. C. Leung, “Recent advances
[70] A. Giret, D. Trentesaux, M. A. Salido, E. Garcia, and E. Adam, in industrial wireless sensor networks toward efficient management in
“A holonic multi-agent methodology to design sustainable intelligent IoT,” IEEE access, vol. 3, pp. 622–637, 2015.
manufacturing control systems,” Journal of cleaner production, vol. [91] Y. Chen, G. M. Lee, L. Shu, and N. Crespi, “Industrial internet of
167, pp. 1370–1386, 2017. things-based collaborative sensing intelligence: framework and research
[71] S. Kumari, A. Singh, N. Mishra, and J. A. Garza-Reyes, “A multi- challenges,” Sensors, vol. 16, no. 2, p. 215, 2016.
agent architecture for outsourcing smes manufacturing supply chain,” [92] X. Sun and N. Ansari, “Edgeiot: Mobile edge computing for the
Robotics and Computer-Integrated Manufacturing, vol. 36, pp. 36–44, internet of things,” IEEE Communications Magazine, vol. 54, no. 12,
2015. pp. 22–29, 2016.
PROCEEDINGS OF THE IEEE, FEBRUARY 2023 19

[93] K. Zhang, J. Long, X. Wang, H.-N. Dai, K. Liang, and M. Imran, [115] V. C. Gungor and G. P. Hancke, “Industrial wireless sensor networks:
“Lightweight searchable encryption protocol for industrial internet of Challenges, design principles, and technical approaches,” IEEE Trans-
things,” IEEE Transactions on Industrial Informatics, pp. 1–10, 2020. actions on industrial electronics, vol. 56, no. 10, pp. 4258–4265, 2009.
[94] J. Huang, L. Kong, H. Dai, W. Ding, L. Cheng, G. Chen, X. Jin, [116] X. Li and J. Wan, “Proactive caching for edge computing-enabled
and P. Zeng, “Blockchain-based mobile crowd sensing in industrial industrial mobile wireless networks,” Future Generation Computer
systems,” IEEE Transactions on Industrial Informatics, vol. 16, no. 10, Systems, vol. 89, pp. 89–97, 2018.
pp. 6553–6563, 2020. [117] M. Bertocco, G. Gamba, A. Sona, and S. Vitturi, “Performance
[95] M. Pueo, J. Santolaria, R. Acero, and J. Sierra-Pérez, “Design method- measurements of CSMA/CA-based wireless sensor networks for in-
ology for production systems retrofit in SMEs,” International Journal dustrial applications,” in 2007 IEEE Instrumentation & Measurement
of Production Research, pp. 1–19, 2019. Technology Conference IMTC 2007. IEEE, 2007, pp. 1–6.
[96] F. Tao, J. Cheng, and Q. Qi, “IIHub: An industrial internet-of-things [118] E. E. Petrosky, A. J. Michaels, and D. B. Ridge, “Network scalability
hub toward smart manufacturing based on cyber-physical system,” comparison of IEEE 802.15. 4 and receiver-assigned CDMA,” IEEE
IEEE Transactions on Industrial Informatics, vol. 14, no. 5, pp. 2271– Internet of Things Journal, vol. 6, no. 4, pp. 6060–6069, 2018.
2280, 2017. [119] D. Yang, J. Ma, Y. Xu, and M. Gidlund, “Safe-wirelesshart: A novel
[97] D. Bruckner, M.-P. Stănică, R. Blair, S. Schriegel, S. Kehrer, M. See- framework enabling safety-critical applications over industrial WSNs,”
wald, and T. Sauter, “An introduction to OPC UA TSN for industrial IEEE Transactions on Industrial Informatics, vol. 14, no. 8, pp. 3513–
communication systems,” Proceeding of the IEEE, vol. 107, no. 6, pp. 3523, 2018.
1121–1131, 2019. [120] C. L. Tan and M. Tracey, “Collaborative new product development
[98] A. Tsuchiya, F. Fraile, I. Koshijima, A. Ortiz, and R. Poler, “Software environments: Implications for supply chain management,” Journal of
defined networking firewall for industry 4.0 manufacturing systems,” Supply Chain Management, vol. 43, no. 3, pp. 2–15, 2007.
Journal of Industrial Engineering and Management (JIEM), vol. 11, [121] J. Queiroz, P. Leitão, J. Barbosa, E. Oliveira, and G. Garcia, “An
no. 2, pp. 318–333, 2018. agent-based industrial cyber-physical system deployed in an automobile
[99] M. Ojo, D. Adami, and S. Giordano, “A SDN-IoT architecture with multi-stage production system,” in International Workshop on Service
NFV implementation,” in 2016 IEEE Globecom Workshops (GC Wk- Orientation in Holonic and Multi-Agent Manufacturing. Springer,
shps). IEEE, 2016, pp. 1–6. 2019, pp. 379–391.
[100] Y. Wu, H.-N. Dai, H. Wang, Z. Xiong, and S. Guo, “A Survey of [122] N. Taatgen, C. Lebiere, and J. Anderson, Modeling Paradigms in ACT-
Intelligent Network Slicing Management for Industrial IoT: Integrated R. Cambridge University Press, 2005, p. 29–52.
Approaches for Smart Transportation, Smart Energy, and Smart Fac- [123] H. A. ElMaraghy, “Flexible and reconfigurable manufacturing systems
tory,” IEEE Communications Surveys & Tutorials, vol. 24, no. 2, pp. paradigms,” International journal of flexible manufacturing systems,
1175–1211, 2022. vol. 17, no. 4, pp. 261–276, 2005.
[101] K. Kaur, S. Garg, G. S. Aujla, N. Kumar, J. J. Rodrigues, and [124] Y. Koren and M. Shpitalni, “Design of reconfigurable manufacturing
M. Guizani, “Edge computing in the industrial internet of things systems,” Journal of manufacturing systems, vol. 29, no. 4, pp. 130–
environment: Software-defined-networks-based edge-cloud interplay,” 141, 2010.
IEEE communications magazine, vol. 56, no. 2, pp. 44–51, 2018. [125] S. M. Saad, “The reconfiguration issues in manufacturing systems,”
[102] V. Galetić, I. Bojić, M. Kušek, G. Ježić, S. Dešić, and D. Huljenić, Journal of Materials Processing Technology, vol. 138, no. 1-3, pp.
“Basic principles of machine-to-machine communication and its impact 277–283, 2003.
on telecommunications industry,” in 2011 proceedings of the 34th [126] J. W. Park, M. Shin, and D. Y. Kim, “An extended agent communica-
international convention MIPRO. IEEE, 2011, pp. 380–385. tion framework for rapid reconfiguration of distributed manufacturing
[103] S. Mascia, P. L. Heider, H. Zhang, R. Lakerveld, B. Benyahia, P. I. systems,” IEEE Transactions on Industrial Informatics, vol. 15, no. 7,
Barton, R. D. Braatz, C. L. Cooney, J. M. Evans, T. F. Jamison pp. 3845–3855, 2019.
et al., “End-to-end continuous manufacturing of pharmaceuticals: inte- [127] Z. Wei, X. Song, and D. Wang, “Manufacturing flexibility, business
grated synthesis, purification, and final dosage formation,” Angewandte model design, and firm performance,” International Journal of Pro-
Chemie International Edition, vol. 52, no. 47, pp. 12 359–12 363, 2013. duction Economics, vol. 193, pp. 87–97, 2017.
[104] J. Wan, B. Chen, M. Imran, F. Tao, D. Li, C. Liu, and S. Ahmad, “To- [128] E. M. Sanfilippo, S. Benavent, S. Borgo, N. Guarino, N. Troquard,
ward dynamic resources management for IoT-based manufacturing,” F. Romero, P. Rosado, L. Solano, F. Belkadi, and A. Bernard, “Mod-
IEEE Communications Magazine, vol. 56, no. 2, pp. 52–59, 2018. eling manufacturing resources: An ontological approach,” in IFIP
[105] T. Stock and G. Seliger, “Opportunities of sustainable manufacturing International Conference on Product Lifecycle Management. Springer,
in industry 4.0,” Procedia Cirp, vol. 40, pp. 536–541, 2016. 2018, pp. 304–313.
[106] Z. Lin, S. Pearson et al., “An inside look at industrial ethernet [129] D. Mourtzis, E. Vlachou, C. Giannoulis, E. Siganakis, and V. Zo-
communication protocols,” White Paper Texas Instruments, 2013. gopoulos, “Applications for frugal product customization and design
[107] E. Sisinni, A. Saifullah, S. Han, U. Jennehag, and M. Gidlund, “Indus- of manufacturing networks,” Procedia CIRP, vol. 52, no. 1, pp. 228–
trial internet of things: Challenges, opportunities, and directions,” IEEE 233, 2016.
Transactions on Industrial Informatics, vol. 14, no. 11, pp. 4724–4734, [130] K. N. McKay and V. C. Wiers, “Integrated decision support for
2018. planning, scheduling, and dispatching tasks in a focused factory,”
[108] X. Li, M. Luo, Y. Qiu, A. Alphones, W.-D. Zhong, C. Yu, and Q. Yang, Computers in Industry, vol. 50, no. 1, pp. 5 – 14, 2003.
“Independent component analysis based digital signal processing in co- [Online]. Available: http://www.sciencedirect.com/science/article/pii/
herent optical fiber communication systems,” Optics Communications, S016636150200146X
vol. 409, pp. 13–22, 2018. [131] K. Biel and C. H. Glock, “Systematic literature review of
[109] H. Boyes, B. Hallaq, J. Cunningham, and T. Watson, “The industrial decision support models for energy-efficient production planning,”
internet of things (IIoT): An analysis framework,” Computers in Computers & Industrial Engineering, vol. 101, pp. 243 – 259, 2016.
Industry, vol. 101, pp. 1–12, 2018. [Online]. Available: http://www.sciencedirect.com/science/article/pii/
[110] D. Li, X. Li, and J. Wan, “A cloud-assisted handover optimization S0360835216303254
strategy for mobile nodes in industrial wireless networks,” Computer [132] A. Valente, “Reconfigurable industrial robots¡ªan integrated approach to
Networks, vol. 128, pp. 133–141, 2017. design the joint and link modules and configure the robot manipulator,”
[111] D. Jansen and H. Buttner, “Real-time Ethernet: the EtherCAT solution,” in Advances in Reconfigurable Mechanisms and Robots II. Springer,
Computing and Control Engineering, vol. 15, no. 1, pp. 16–21, 2004. 2016, pp. 779–794.
[112] A. Moldovansky, “Utilization of modern switching technology in [133] J. Wan, M. Yi, D. Li, C. Zhang, S. Wang, and K. Zhou, “Mobile
ethernet/IP networks,” in Proccedings of 1st Workshop on Real-Time services for customization manufacturing systems: An example of
LANs in the Internet Age, 2002, pp. 25–27. industry 4.0,” IEEE Access, vol. 4, pp. 8977–8986, 2016.
[113] G. Cena, L. Seno, A. Valenzano, and S. Vitturi, “Performance analysis [134] K. Gai and M. Qiu, “Reinforcement learning-based content-centric
of Ethernet Powerlink networks for distributed control and automation services in mobile sensing,” IEEE Network, vol. 32, no. 4, pp. 34–
systems,” Computer Standards & Interfaces, vol. 31, no. 3, pp. 566– 39, 2018.
572, 2009. [135] M. Satyanarayanan, “The emergence of edge computing,” Computer,
[114] W. Liang, X. Zhang, Y. Xiao, F. Wang, P. Zeng, and H. Yu, “Survey and vol. 50, no. 1, pp. 30–39, 2017.
experiments of WIA-PA specification of industrial wireless network,” [136] N. Chungoora, R. I. Young, G. Gunendran, C. Palmer, Z. Usman, N. A.
Wireless Communications and Mobile Computing, vol. 11, no. 8, pp. Anjum, A.-F. Cutting-Decelle, J. A. Harding, and K. Case, “A model-
1197–1212, 2011. driven ontology approach for manufacturing system interoperability
PROCEEDINGS OF THE IEEE, FEBRUARY 2023 20

and knowledge sharing,” Computers in Industry, vol. 64, no. 4, pp. 58th Annual Conference of the Society of Instrument and Control
392–401, 2013. Engineers of Japan (SICE). IEEE, 2019, pp. 1349–1354.
[137] T. Wang, S. Guo, and C.-G. Lee, “Manufacturing task semantic model- [158] J. Wan, J. Li, Q. Hua, A. Celesti, and Z. Wang, “Intelligent equipment
ing and description in cloud manufacturing system,” The International design assisted by cognitive internet of things and industrial big data,”
Journal of Advanced Manufacturing Technology, vol. 71, no. 9-12, pp. Neural Computing and Applications, pp. 1–10, 2018.
2017–2031, 2014. [159] G. B. Grant, T. P. Seager, G. Massard, and L. Nies, “Information
[138] M. C. Elish and D. Boyd, “Situating methods in the magic of Big Data and communication technology for industrial symbiosis,” Journal of
and AI,” Communication monographs, vol. 85, no. 1, pp. 57–80, 2018. Industrial Ecology, vol. 14, no. 5, pp. 740–753, 2010.
[139] J. Li, F. Tao, Y. Cheng, and L. Zhao, “Big data in product lifecycle [160] S. Jeschke, C. Brecher, T. Meisen, D. Özdemir, and T. Eschert,
management,” The International Journal of Advanced Manufacturing “Industrial internet of things and cyber manufacturing systems,” in
Technology, vol. 81, no. 1-4, pp. 667–684, 2015. Industrial internet of things. Springer, 2017, pp. 3–19.
[140] R. Y. Zhong, S. Lan, C. Xu, Q. Dai, and G. Q. Huang, “Visualization of [161] D. Mourtzis, E. Vlachou, and N. Milas, “Industrial big data as a result
RFID-enabled shopfloor logistics Big Data in Cloud Manufacturing,” of iot adoption in manufacturing,” Procedia cirp, vol. 55, pp. 290–295,
The International Journal of Advanced Manufacturing Technology, 2016.
vol. 84, no. 1-4, pp. 5–16, 2016. [162] W. Liang, M. Zheng, J. Zhang, H. Shi, H. Yu, Y. Yang, S. Liu,
[141] Y. Zhang, S. Ren, Y. Liu, and S. Si, “A big data analytics architecture W. Yang, and X. Zhao, “WIA-FA and its applications to digital factory:
for cleaner manufacturing and maintenance processes of complex A wireless network solution for factory automation,” Proceedings of
products,” Journal of Cleaner Production, vol. 142, pp. 626–641, 2017. the IEEE, vol. 107, no. 6, pp. 1053–1073, 2019.
[142] P. O’Donovan, K. Leahy, K. Bruton, and D. T. O’Sullivan, “An [163] A. E. Kalør, R. Guillaume, J. J. Nielsen, A. Mueller, and P. Popovski,
industrial big data pipeline for data-driven analytics maintenance “Network slicing in industry 4.0 applications: Abstraction methods
applications in large-scale smart manufacturing facilities,” Journal of and end-to-end analysis,” IEEE Transactions on Industrial Informatics,
Big Data, vol. 2, no. 1, p. 25, 2015. vol. 14, no. 12, pp. 5419–5427, 2018.
[143] J. Daily and J. Peterson, “Predictive maintenance: How big data analy- [164] J. Cheng, W. Chen, F. Tao, and C.-L. Lin, “Industrial IoT in 5G environ-
sis can improve maintenance,” in Supply Chain Integration Challenges ment towards smart manufacturing,” Journal of Industrial Information
in Commercial Aerospace. Springer, 2017, pp. 267–278. Integration, vol. 10, pp. 10–19, 2018.
[144] J. Wan, S. Tang, D. Li, S. Wang, C. Liu, H. Abbas, and A. V. Vasilakos, [165] E. Udd, “Overview of fiber optic sensors,” in Fiber Optic Sensors.
“A manufacturing big data solution for active preventive maintenance,” CRC Press, 2017, pp. 1–34.
IEEE Transactions on Industrial Informatics, vol. 13, no. 4, pp. 2039– [166] J. Zhang, X. Yao, J. Zhou, J. Jiang, and X. Chen, “Self-organizing
2047, 2017. manufacturing: Current status and prospect for Industry 4.0,” in 2017
[145] D. Georgakopoulos, P. P. Jayaraman, M. Fazia, M. Villari, and 5th International Conference on Enterprise Systems (ES). IEEE, 2017,
R. Ranjan, “Internet of things and edge cloud computing roadmap for pp. 319–326.
manufacturing,” IEEE Cloud Computing, vol. 3, no. 4, pp. 66–73, 2016. [167] J. Guozhu, “Customers-focused strategy of manufacture and the recon-
[146] M. H. Frické, Data-Information-Knowledge-Wisdom (DIKW) struction of manufacture systems,” Chinese Journal of Management,
Pyramid, Framework, Continuum. Cham: Springer Inter- vol. 1, 2006.
national Publishing, 2018, pp. 1–4. [Online]. Available: [168] V. R. S. Kumar, A. Khamis, S. Fiorini, J. L. Carbonera, A. O. Alarcos,
https://doi.org/10.1007/978-3-319-32001-4 331-1 M. Habib, P. Goncalves, H. Li, and J. I. Olszewska, “Ontologies for
[147] I. Regev and D. Benson-Karhi, “Segmentation models for the duration Industry 4.0,” The Knowledge Engineering Review, vol. 34, 2019.
of expected preventive maintenance in semiconductor fabs,” IEEE [169] E. Rajangam and C. Annamalai, “Graph models for knowledge repre-
Transactions on Semiconductor Manufacturing, vol. 29, no. 3, pp. 223– sentation and reasoning for contemporary and emerging needs–a sur-
229, 2016. vey,” International Journal of Information Technology and Computer
[148] A. Bonfanti, M. Del Giudice, and A. Papa, “Italian craft firms between Science (IJITCS), vol. 8, no. 2, pp. 14–22, 2016.
digital manufacturing, open innovation, and servitization,” Journal of [170] A. R. Zamir, A. Sax, W. Shen, L. J. Guibas, J. Malik, and S. Savarese,
the Knowledge Economy, vol. 9, no. 1, pp. 136–149, 2018. “Taskonomy: Disentangling task transfer learning,” in Proceedings of
[149] Q. Wang, X. Zhao, and C. Voss, “Customer orientation and innovation: the IEEE Conference on Computer Vision and Pattern Recognition,
A comparative study of manufacturing and service firms,” International 2018, pp. 3712–3722.
Journal of Production Economics, vol. 171, pp. 221–230, 2016. [171] F. K. Došilović, M. Brčić, and N. Hlupić, “Explainable artificial
[150] H. Wang, P. Shi, and Y. Zhang, “Jointcloud: A cross-cloud cooperation intelligence: A survey,” in 2018 41st International convention on
architecture for integrated internet service customization,” in 2017 information and communication technology, electronics and microelec-
IEEE 37th international conference on distributed computing systems tronics (MIPRO). IEEE, 2018, pp. 0210–0215.
(ICDCS). IEEE, 2017, pp. 1846–1855. [172] F. Tao, J. Cheng, Q. Qi, M. Zhang, H. Zhang, and F. Sui, “Digital twin-
[151] M. Bahrami and M. Singhal, “A dynamic cloud computing architecture driven product design, manufacturing and service with big data,” The
for cloud-assisted internet of things in the era of big data,” in Big Data International Journal of Advanced Manufacturing Technology, vol. 94,
and Computational Intelligence in Networking. CRC Press, 2017, pp. no. 9-12, pp. 3563–3576, 2018.
107–124. [173] Q. Qi and F. Tao, “Digital twin and big data towards smart manufac-
[152] Z. Zhang, X. Wang, X. Zhu, Q. Cao, and F. Tao, “Cloud manufacturing turing and industry 4.0: 360 degree comparison,” IEEE Access, vol. 6,
paradigm with ubiquitous robotic system for product customization,” pp. 3585–3593, 2018.
Robotics and Computer-Integrated Manufacturing, vol. 60, pp. 12 – [174] W. Kritzinger, M. Karner, G. Traar, J. Henjes, and W. Sihn, “Digital
22, 2019. [Online]. Available: http://www.sciencedirect.com/science/ twin in manufacturing: A categorical literature review and classifica-
article/pii/S0736584519300900 tion,” IFAC-PapersOnLine, vol. 51, no. 11, pp. 1016–1022, 2018.
[153] L. Graening and B. Sendhoff, “Shape mining: A holistic data mining [175] Y. Wang, T. L. Nguyen, Y. Xu, Z. Li, Q. Tran, and R. Caire, “Cyber-
approach for engineering design,” Advanced Engineering Informatics, physical design and implementation of distributed event-triggered sec-
vol. 28, no. 2, pp. 166–185, 2014. ondary control in islanded microgrids,” IEEE Transactions on Industry
[154] J. K. Bae and J. Kim, “Product development with data mining tech- Applications, vol. 55, no. 6, pp. 5631–5642, 2019.
niques: A case on design of digital camera,” Expert Systems with [176] J. Pan, S. Ding, D. Wu, S. Yang, and J. Yang, “Exploring behavioural
Applications, vol. 38, no. 8, pp. 9274–9280, 2011. intentions toward smart healthcare services among medical practi-
[155] A. Fels, B. Falk, and R. Schmitt, “User-driven customization tioners: a technology transfer perspective,” International Journal of
and customer loyalty: A survey,” Procedia CIRP, vol. 60, pp. Production Research, vol. 57, no. 18, pp. 5801–5820, 2019.
410 – 415, 2017, complex Systems Engineering and Development [177] J. Rybicka, A. Tiwari, and G. A. Leeke, “Technology readiness level
Proceedings of the 27th CIRP Design Conference Cranfield assessment of composites recycling technologies,” Journal of Cleaner
University, UK 10th – 12th May 2017. [Online]. Available: Production, vol. 112, pp. 1001 – 1012, 2016. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S2212827117301038 http://www.sciencedirect.com/science/article/pii/S0959652615012056
[156] A. K. Tyagi, S. Sharma, N. Anuradh, N. Sreenath et al., “How a user
will look the connections of internet of things devices?: A smarter
look of smarter environment,” in Proceedings of 2nd International
Conference on Advanced Computing and Software Engineering, 2019.
[157] J. Zhang, J. Du, and X. Shi, “Data-driven intelligent system for
equipment management in automotive parts manufacturing,” in 2019
PROCEEDINGS OF THE IEEE, FEBRUARY 2023 21

Jiafu Wan (M’11) is a Professor in School of Me- Andrew Kusiak is a Professor in the Department of
chanical & Automotive Engineering at South China Industrial and Systems Engineering and Director of
University of Technology, China. He has directed the Intelligent Systems Laboratory at The University
20 research projects, including the National Key of Iowa, Iowa City. He has chaired two departments,
Research and Development Project, and the Joint Industrial Engineering (1988-95) and Mechanical
Fund of the National Natural Science Foundation and Industrial Engineering (2010-15). His current
of China and Guangdong Province. Thus far, he has research focuses on applications of artificial intelli-
published more than 160 scientific papers, including gence and big data in smart manufacturing, product
100+ SCI-indexed papers, 50+ IEEE Trans./Journal development, renewable energy, sustainability, and
papers, 20 ESI Highly Cited Papers and 4 ESI Hot healthcare. He has published numerous books and
Papers. According to Google Scholar Citations, his technical papers in journals sponsored by profes-
published work has been cited more than 12,000 times (H-index = 49). His sional societies, such as the Association for the Advancement of Artificial
SCI other citations, sum of times cited without self-citations, reached 3500 Intelligence, the American Society of Mechanical Engineers, Institute of
times (H-index = 36) according to Web of Science Core Collection. He is Industrial and Systems Engineers, Institute of Electrical and Electronics
an Associate Editor of IEEE/ASME Transactions on Mechatronics, Journal Engineers, Nature, and other societies. He speaks frequently at international
of Intelligent Manufacturing and Computers & Electrical Engineering, and meetings, conducts professional seminars, and consults for industrial corpo-
Editorial Board for Computer Integrated Manufacturing Systems. He is a rations. Dr. Kusiak has served in elected professional society positions as
Leading Guest Editor for several SCI-indexed journals, such as IEEE Systems well as various editorial roles in over fifty journals, including five different
Journal, IEEE Access, Elsevier Computer Networks, Mobile Networks & Ap- IEEE Transactions. Professor Kusiak is a Fellow of the Institute of Industrial
plications, Computers and Electrical Engineering, Wireless Communications and Systems Engineers and the Editor-in-Chief of the Journal of Intelligent
and Mobile Computing, Journal of Internet Technology, and Microprocessors Manufacturing.
and Microsystems. His research interests include Cyber-Physical Systems,
Intelligent Manufacturing, Big Data Analytics, Industry 4.0, Smart Factory,
and Cloud Robotics. He is listed as a Clarivate Analytics Highly Cited
Researcher in 2019 and 2020.

Miguel Martı́nez-Garcı́a is a Lecturer in Human-

machine systems at Loughborough University, UK.
He received a BSc degree in Mathematics from the
Polytechnic University of Catalonia (UPC), Spain in
Xiaomin Li is currently an Associate Professor at 2013, a MSc in Advanced Mathematics and Math-
Zhongkai University of Agriculture and Engineer- ematical Engineering (MAMME) from the same
ing, Guangzhou, China. He received B.S. degree university in 2014, and a PhD in Engineering from
from the Air Force Engineering University, Xi’an the University of Lincoln, UK, in 2018. He also
China (2004); M.S. degree from the South China worked as a researcher since 2017, both at the
Agricultural University, Guangzhou, China (2014); University of Lincoln and at the Advanced Virtual
and Ph.D. degree from the South China Univer- Reality Research Centre (AVRRC) at Loughborough
sity of Technology, Guangzhou, China (2018). His University. His research interests include human-machine integration, machine
research interests include Cyber-Physical Systems, learning, artificial intelligence, intelligent signal processing, and complex
Smart manufacturing, Big data, Wireless Network. systems, with particular focus in the analysis of non-linear signals representing
phenomena of interest between humans and machines.

Hong-Ning Dai (M’05–SM’16) is currently with

Faculty of Information Technology at Macau Uni- Di Li is a Professor of School of Mechanical and
versity of Science and Technology (MUST) as an Automotive Engineering, South China University of
associate professor. He obtained the Ph.D. degree Technology, Guangzhou, China. Di Li was born in
in Computer Science and Engineering from Depart- Fu yang City, Anhui Province, China, in 1965. She
ment of Computer Science and Engineering at the received the B.S. and M. S. degrees in aerospace
Chinese University of Hong Kong. His current re- engineering from the Nanjing University of Aero-
search interests include Data Analytics, Blockchain nautics and Astronautics, Nanjing City, Jiangshu
Technology, Cyber-Physical System and industrial Province, China, in 1988 and the Ph.D. degree
Internet of Things. He is a holder of 1 U.S. patent in automation from the South China University of
and 1 Australia innovation patent. He is also the Technology, Guangzhou City, Guangdong Province,
winner of Bank of China (BOC) Excellent Research Award of MUST. He has China, in 1993. From 1988 to 1990, she was an
3 ESI highly-cited papers awarded by Clarivate Analytics and 1 IEEE Out- Assistant Engineer with the Qingdao Beihai Shipyard of the China State
standing Paper Award conferred by IEEE International Conference on Cyber Shipbuilding Corporation, Qingdao City, Shandong Province, China. From
Physical and Social Computing. He holds visiting positions at Department of 1993 to 1998, she was an Associate Professor with the Guangdong University
Computer Science and Engineering of the Hong Kong University of Science of Technology, Guangzhou City, Guangdong Province, China. Since 1999,
and Technology, School of Electrical Engineering and Telecommunications she has been a Professor with the Mechanical and Electrical Engineering
of the University of New South Wales, Hong Kong Applied Science and Department, South China University of Technology. She is the author of
Technology Research Institute (ASTRI). He has served as associate editors of more than 150 articles and holds eight patents. Her research interests include
IEEE Systems Journal, IEEE Access, and Connection Science, an editor of Ad high-performance embedded control system, numerical control, and machine
Hoc Networks, guest editors for IEEE Transactions on Industrial Informatics, vision. Ms. Li was a recipient of the First Prize for Science and Technology
IEEE Transactions on Emerging Topics in Computing. He is also an ACM Development of Guangdong Province in 2009 and Ding Ying Science and
senior member. Technology Awards in 2011.