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

net/publication/338224253

An Evolution of Innovations Protocols and Recent Technology in Industrial IoT

Chapter · January 2020

DOI: 10.1007/978-3-030-32530-5_11

CITATIONS READS

2 12,335

3 authors, including:

Saranya Vasanthamani G. R. Kanagachidambaresan

Sri Krishna College of Engineering and Technology Vel Tech - Technical University
10 PUBLICATIONS 91 CITATIONS 99 PUBLICATIONS 638 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Optimization of sensor nodes View project

Energy Efficiency in Wireless Sensor Networks using Cluster Allocation and Routing Algorithm View project

All content following this page was uploaded by Saranya Vasanthamani on 11 June 2020.

The user has requested enhancement of the downloaded file.

EAI/Springer Innovations in Communication and Computing

G. R. Kanagachidambaresan
R. Anand
E. Balasubramanian
V. Mahima Editors

Internet of
Things for
Industry 4.0
Design, Challenges and Solutions
EAI/Springer Innovations in Communication
and Computing

Series Editor
Imrich Chlamtac, European Alliance for Innovation, Ghent, Belgium
Editor’s Note
The impact of information technologies is creating a new world yet not fully
understood. The extent and speed of economic, life style and social changes already
perceived in everyday life is hard to estimate without understanding the technological
driving forces behind it. This series presents contributed volumes featuring the
latest research and development in the various information engineering technologies
that play a key role in this process.
The range of topics, focusing primarily on communications and computing
engineering include, but are not limited to, wireless networks; mobile communication;
design and learning; gaming; interaction; e-health and pervasive healthcare; energy
management; smart grids; internet of things; cognitive radio networks; computation;
cloud computing; ubiquitous connectivity, and in mode general smart living, smart
cities, Internet of Things and more. The series publishes a combination of expanded
papers selected from hosted and sponsored European Alliance for Innovation (EAI)
conferences that present cutting edge, global research as well as provide new
perspectives on traditional related engineering fields. This content, complemented
with open calls for contribution of book titles and individual chapters, together
maintain Springer’s and EAI’s high standards of academic excellence. The audience
for the books consists of researchers, industry professionals, advanced level students
as well as practitioners in related fields of activity include information and
communication specialists, security experts, economists, urban planners, doctors,
and in general representatives in all those walks of life affected ad contributing to
the information revolution.
About EAI
EAI is a grassroots member organization initiated through cooperation between
businesses, public, private and government organizations to address the global challenges
of Europe’s future competitiveness and link the European Research community
with its counterparts around the globe. EAI reaches out to hundreds of thousands of
individual subscribers on all continents and collaborates with an institutional member
base including Fortune 500 companies, government organizations, and educational
institutions, provide a free research and innovation platform.
Through its open free membership model EAI promotes a new research and
innovation culture based on collaboration, connectivity and recognition of excellence
by community.

More information about this series at http://www.springer.com/series/15427

G. R. Kanagachidambaresan • R. Anand
E. Balasubramanian • V. Mahima
Editors

Internet of Things
for Industry 4.0
Design, Challenges and Solutions
Editors
G. R. Kanagachidambaresan R. Anand
Department of CSE Department of EEE
Vel Tech Rangarajan Dr Sagunthala R&D Amrita Vishwa Vidyapeetham University
Institute of Science and Technology Bangalore, India
Chennai, India
V. Mahima
E. Balasubramanian Department of ECE
Department of Mechanical Engineering Vel Tech Rangarajan Dr Sagunthala R&D
Vel Tech Rangarajan Dr Sagunthala R&D Institute of Science and Technology
Institute of Science and Technology Chennai, India
Chennai, India

ISSN 2522-8595     ISSN 2522-8609 (electronic)

EAI/Springer Innovations in Communication and Computing
ISBN 978-3-030-32529-9    ISBN 978-3-030-32530-5 (eBook)
https://doi.org/10.1007/978-3-030-32530-5

© Springer Nature Switzerland AG 2020

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the
editors give a warranty, express or implied, with respect to the material contained herein or for any errors
or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims
in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
To our students, parents, and the almighty
Preface

Internet of Things (IoT) has become the inevitable component in smart city
­development. The decline in the workforce of developing countries is presently
solved through cyber-physical system (CPS) technology. The industries are enabled
with CPS to monitor their production, enhancing their productivity through Industry
4.0 standards. The main objective of Industry 4.0 is to integrate all the machines as
a single integrated network for M2M communication. The CPS is enabled in auton-
omous robots, UAVs, cybersecurity, cloud storage systems, additive manufacturing,
augmented realities, and big data computations. The production became noticeable
during Industry 3.0 standards; however, the microcomputer interface and machine-
to-machine interaction in Industry 4.0 have made a tremendous increase in produc-
tion. Industry 4.0 has decentralized analytics, critical decision making, and increase
in response time during productions. The machine-to-machine interaction has
become possible through wireless and wired communication. Robust and foolproof
system design is required to ensure workers safety. The existing companies with 3.0
standards are facing a formidable challenge to meet 4.0 standards. The existing
CNC machines cooperation with CPS is tedious, and researchers are working
toward fault-tolerant algorithm design for the same. Creating transparency in pro-
duction between joint ventures has brought greater impact on profit and production.
An efficient venture makes the system prone to cyber-based attacks. Researchers are
focusing on lightweight cryptography algorithm to prevent the CPS from attacks
and provide high security. These connected machines in industries provide a high
volume of data, informing the nature of the machine and its present state of work-
ing. The machine about to fail can easily be predicted through prognostics and arti-
ficial intelligence. The system has the capability to identify the broken parts and
their faults to the user. Deep learning and machine learning algorithms can ensure

vii
viii Preface

workers safety and can avoid industrial accidents. The role of robotics and additive
manufacturing through 3D printing has attracted a lot of young researchers in real-
izing their ideas through rapid prototyping. These chapters provide an insight for
the development of algorithms in Industrial 4.0 scenario. The book concentrates on
providing an overview in areas such as robotics, RFID, image processing, workers
safety, software-defined network, and smart grid applications.

Chennai, Tamil Nadu, India G. R. Kanagachidambaresan

Bangalore, Karnataka, India R. Anand
Chennai, Tamil Nadu, India E. Balasubramanian
Chennai, Tamil Nadu, India V. Mahima
Acknowledgment

We would like to thank Imrich the Series Editor and Eliska the Managing Editor at
EAI for giving us the opportunity to edit a book in Innovations in Communication
and Computing (Springer). We thank the authors who have put tremendous efforts
in this book project. We would also like to thank the reviewers who give an unbiased
review on time for successful completion. The generous support from Vel Tech
Rangarajan Dr. Sagunthala R&D Institute of Science and Technology, Amrita
Vishwa Vidyapeetham, Bangalore, allowed contributing an exciting interesting
material to new researchers and students. We hope this book would be a necessary
material for engineers and scientists.

ix
Contents

  1 A Survey on RFID in Industry 4.0���������������������������������������������������������    1

Ermal Elbasani, Pattamaset Siriporn, and Jae Sung Choi
  2 Data Fusion-Based AI Algorithms
in the Context of IIoTS����������������������������������������������������������������������������   17
Raluca Maria Aileni, Suciu George, Sever Pasca,
and Valderrama Sukuyama Carlos Alberto
  3 Presents the Technology, Protocols, and New Innovations
in Industrial Internet of Things (IIoT)��������������������������������������������������   39
Quanxin Zhao
  4 Decision Making Models Through AI for Internet of Things��������������   57
E. P. Ephzibah, S. Sree Dharinya, and L. Remya
  5 Internet of Things and Additive Manufacturing:
Toward Intelligent Production Systems in Industry 4.0����������������������   73
A. Suresh, R. Udendhran, and G. Yamini
  6 Deep Learning Enabled Smart Industrial Workers
Precaution System Using Single Board Computer (SBC)��������������������   91
S. Pradeep Kumar, S. Selvakumari, S. Praveena, and S. Rajiv
  7 A Novel Design Augmentation of Bio-­Inspired Artificial
Immune Technique in Securing Internet of Things (IOT) ������������������ 103
G. Usha, P. Madhavan, and M. V. Ranjith Kumar
  8 Role of AI and Bio-Inspired Computing in Decision Making�������������� 115
Surekha Paneerselvam
  9 Smart Recognition System for Business Predictions
(You Only Look Once – V3) Unified, Real-Time
Object Detection�������������������������������������������������������������������������������������� 137
Allumallu Veera Venkata Susmitha

xi
xii Contents

10 IoT-Based Monitoring and Management of Electric

Vehicle Charging Systems for DC Fast Charging Facility ������������������ 147
R. Suresh Kumar, P. K. Rajesh, J. Joys Nancy, S. Abirami,
and K. Vishnu Murthy
11 An Evolution of Innovations Protocols and Recent
Technology in Industrial IoT������������������������������������������������������������������ 161
V. Saranya, M. J. Carmel Mary Belinda,
and G. R. Kanagachidambaresan
12 Smart Meter: A Key Component for Industry 4.0
in Power Sector���������������������������������������������������������������������������������������� 177
M. Kathiresh and A. Mahaboob Subahani
13 Role of Software-Defined Network in Industry 4.0������������������������������  197
Vakaavinash Reddy and K. Venkatesh
14 Integrating IoT and Machine Learning – The Driving
Force of Industry 4.0��������������������������������������������������������������������������������  219
A. Suresh, R. Udendhran, and M. Balamurugan
15 Machine Intelligence and Automation: Deep Learning
Concepts Aiding Industrial Applications ���������������������������������������������� 237
S. Sree Dharinya and E. P. Ephzibah

Index������������������������������������������������������������������������������������������������������������������ 249
About the Editors

G.R. Kanagachidambaresan had completed his

PhD in Information and Communication Engineering
at Anna University, Chennai. He has successfully com-
pleted numerous consultancy projects and obtained
patents. He has authored many research articles in
leading journals and conferences. He is serving as an
editorial board member in many leading journals. He is
presently working in projects funded from Indian
Space Research Organisation (ISRO). He has edited
several books and also guest edited many special
issues. He is a software solution provider for compa-
nies and industries. He holds numerous copyrights and
patents on software developed and product designed.
He is presently working on Internet of Things applica-
tions and remote health monitoring.
R. Anand completed his PhD in EEE at Anna
University, Chennai. He has successfully authored arti-
cles in leading journals and conferences. He has also
carried out many consultancy works and projects for
MNCs. He is presently an assistant professor in the
Department of EEE, Amrita Vishwa Vidyapeetham
University, Bangalore, India. He has successfully
hosted many special issues in leading journals.

xiii
xiv About the Editors

E. Balasubramanian had completed PhD in the field

of Robotics and Control at Concordia University,
Montreal, Canada. He has more than 15 years of expe-
rience in carrying out numerous research projects. He
executed consultancy projects for national agencies
such as BARC, IGCAR, BEL, ISRO, and automobile
sector. He collaborated with Taiwan scientists under
Indo-Taiwan schema for the development of Micro
Aerial Vehicles and SWARM control through image
processing techniques. He is presently working on
Indo-Canada collaborative research project aiming at
inspection of railway bridges and heritage structures
using UAVs and also cooperating with Korean scien-
tists (Indo-Korea Collaboration) to develop unmanned
amphibious vehicle to collect water samples for water
quality monitoring. His team developed unmanned
aerial vehicles (UAVs) for diverse applications includ-
ing power line inspection, telecom tower inspection
and radiation measurement, environmental monitoring,
surveillance, and aerial view inspection. His team won
the first prize of INR 5 lakhs in a national-level compe-
tition organized by Power Grid Corporation Limited
and the prestigious telecom sector Aegis Graham Bell
award in the category of best innovative business model
2014. His research interests are UAVs, robotics, and
control and sensors with online data acquisition
systems.
V. Mahima is presently an assistant professor in the
Department of ECE in Vel Tech Rangarajan Dr
Sagunthala R&D Institute of Science and Technology,
Avadi, Chennai. She has successfully authored many
articles and projects. Her areas of interest include
Wireless Sensor Network, Wireless Body Sensor
Network, Internet of Things, and Industry 4.0.
Chapter 1
A Survey on RFID in Industry 4.0

Ermal Elbasani, Pattamaset Siriporn, and Jae Sung Choi

Introduction

The advanced and rapid achievements of information and communication technol-

ogy have brought in a new industrial revolution, so-called Industry 4.0. This has
come with new challenges in science and industry, but specific efforts need to be
done to demonstrate that new industrial systems can be applied to factory environ-
ments. Mentioning important topics of Industry 4.0 is RFID system, and Internet of
Things (IoT) technology, with the function to generate data from target objects and
then access an Internet. Moreover, RFID system can be a role as a link between
process flow data and physical asset data. IoT assists accumulate a lot of data
through RFID tags and readers and sensors. And it improves transparency and effi-
ciency of production management with big data analysis and artificial intelligence
technologies.
Despite RFID as main part of IoT, in Industry 4.0, Internet of Services (IoS) and
cyber-physical systems (CPS) play a key role in production process. Also, much
research is being done on RFID applications in industry such as inventory manage-
ment system [1], object location detection [2], RFID using MES [3], planning and
scheduling decision support system [4–6], material flow monitoring system [7, 8],
and quality assurance system [9].
This research aims to review the Industry 4.0 studies on the viewpoint of RFID,
since on our good knowledge does not exist any similar review with particular focus
on the significance of RFID in the Industry 4.0.

E. Elbasani · P. Siriporn · J. S. Choi (*)

Department of Computer Engineering, Sun Moon University, Asan, South Korea
e-mail: [email protected]

© Springer Nature Switzerland AG 2020 1

G. R. Kanagachidambaresan et al. (eds.), Internet of Things for Industry 4.0,
EAI/Springer Innovations in Communication and Computing,
https://doi.org/10.1007/978-3-030-32530-5_1
2 E. Elbasani et al.

History of RFID in Industry 4.0

RFID in History

RFID technology has evolved rapidly over years, particularly widely applied with
the advancements of web technology. The initial appearance of RFID technology
was in World War II, used in a British project that developed a system to detect
and identify whether fighter aircraft belong to enemy or their ally, using RF
technology.
And the fourth industrial revolution is when through Internet or the web tech-
nology and the physical industrial technologies combined into one. Since the
early 1990s, Internet has had revolutionary impact on living, culture, and com-
merce. In fourth industrial revolution, RFID can be the main source of data gen-
eration in smart environments, such as smart factory, with utilization of Internet
technologies.

RFID Technology

RFID system itself is simple. RFID is found in long-range and low-range format.
Long-range RFID referred also as RAIN RFID, based on the UHF band, has the
fastest recognition speed and can read tags up to 15 m, and it is suitable for fac-
tory usage.
The low-range RFID uses the HF band, provides low tag recognition, and reads
till 1.2 m. The low-range RFID reader is suitable for smartphone users and has
found a great usage in consumer-trade shopping activities and also other areas like
transport, etc. With the development of technology, there are being invented a lot of
ideas of using it.
In manufacturing, RFID can be an important tool to help manufacturers effi-
ciently produce more products, and also machines can communicate with each other
and perform an immediate action if a mistake occurs.
RFID systems are ideal for any company that is required or wants to utilize auto-
mated processes and interconnect modules and devices. Industry 4.0 has the oppor-
tunity to be easily embraced in the manufacturing industry, and RFID has the ability
to realize everything. Passive RFID has been used massively in the past years, and
it is easily expected the passive RFID system will be continuously and widely
applied in our industries, especially in the ultrahigh-frequency range, and also active
RFID not as widely, but with the advantage for real-time location system (RTLS)
applications [10].
In order to make active RFID usefully and popularly the same, it is required
standardization of active 433 MHz RFID to drive growth. Also, mentioning, ultra-­
wideband (UWB) and Bluetooth low-energy (BLE) RTLS solutions will further
enhance the solution [11].
1 A Survey on RFID in Industry 4.0 3

Fig. 1.1 Development of the RFID toward the industrial IoT and Industry 4.0

A significant concept of RFID is generally shown in Fig. 1.1, that describe how
the main components of RFID develop to be embedded in smart Industrial technol-
ogy. Over the past decade, remote monitoring, used by RFID and IoT to manage
multiple sensor inputs and using RFID in localization, has been an attractive issue.
Industry 4.0 is moving to focus on smart products, analyze RFID/IoT data, and
research on network functions to improve system performance.

RFID and Big Data

Nowadays, many factories have begun involving advanced information technolo-

gies and knowledge discovering methods to facilitate information flow, with the
main need to deal with the large volume of data which are generated and collected
from manufacturing processes and its related operations. The generated and col-
lected data from processes in factory such as production, transporting, surveillance,
etc. is growing exponentially [12].
Particularly in Industry 4.0, data integration and analyzing is applied to improve
in decision-making stage, increasing performance in scalable data flow adaptation
processes and systems [13]. Big data issue deals with large volume that need to be
processed in time as quick as possible and with the ability to combine various data
formats.
One another challenge of passive RFID tags is to adapt to complex, dynamic
manufacturing environments since they generate relatively small storage, compared
to large number of real-time data record processes. Also for a high decision-making
standard, it requires a universal and effective network connection to arrange data on
the network. This challenges the data transferring and network bandwidth when
dealing large number of RFID tags emitting signals to be recorded and analyzed.
Active RFID tags, which occupy large storage space, make able for higher inter-­
communication among machines, and also workers have more ease to detect and
cope with different events and make the proper decision.
The big data technologies are a powerful and useful way to address variety and
huge-capacity data. Next there are explained in sequentially five steps the concept
of RFDI data processing from signal capturing to decision-making stage, also illus-
trated in Fig. 1.2.
4 E. Elbasani et al.

Fig. 1.2 RFID data processing phase

First is collection phase: it has to do with capturing data signals from RFID
readers and RFID tags and also from smart sensors distributed. RFID raw data is
stored in a database in tables format. In this stage it is challenging to make the most
of knowledge to support advanced decision.
Second is data cleaning: RFID devices and IoT devices generate large-amount
raw data. These data are automatically correlated with process flow. However, due
to the environment effect, the random nature of RFID events, there may be many
incomplete, and inaccurate, data. And in this stage, the data goes in a process of
organizing based on some programmed standards from the upper-level stages.
The third step is data integration. In the case of massive data, systems need to
configure real-time and non-real-time data and structured and unstructured data.
The integration process, which adopted data mining methods, is by clustering, clas-
sification, association, and prediction using decision trees, genetic algorithms, and
supported vector machines techniques.
This step helps you re-configure valid data according to process flow order and
time sequence to translate valid data into meaningful pattern information and
knowledge. Inputs of this stage are cleaned and organized data, and outputs are data
that flows according to process order.
The fourth step consists of knowledge extraction. The purpose of this step is
to transform the organized data into meaningful information to support control
and decision-making for output control. Inputs are organized data according to
process, and outputs are information derived from useful attributes and data fea-
tures. SVM (support vector machine) [14], k-means clustering algorithm, and
deep neural network (DNN) can be used to process large RFID-based production
data. You can discover knowledge, patterns, and rules based on the processing
results.
The final stage of decision-making and influential work is to interpret the infor-
mation, knowledge, patterns, and rules achieved in a useful and simple way.
When IoT and RFID are applied to the manufacturing process, heterogeneous
RFID-based data, real time and in huge amount data, are generated and recorded,
and that is called industrial big data [15, 16].
In [17, 18] researches provide a broad overview and studies about variety of
management of captured data by RFID and related mining, analyzing, and process-
ing techniques. The large amounts of data generated from sensing devices, mobile
devices, and RFID devices are termed as big data.
Mining techniques for RFID data help to find a pattern regarded to process when
it is applied. This means that this pattern matches the respective frequent process,
with the purpose to provide concrete knowledge to user/owner and assure also a
visual information [19].
1 A Survey on RFID in Industry 4.0 5

In the analyzing process, RFID data mining algorithms, which are utilized with
SVM (support vector machines), DT (decision tree), and NN (neural networks)
[14], have been adopted for cases of clustering, classification, and deep learning.
In addition, the mining method is used in operational research methods, combining
mixed integer and probability-reduction programming. The most highlighted
methods of data mining are explained as follows:
• Associative Rules: In mining methods in data science, associative rules are one
of the most studied methods. In [11], associative rules include both positive and
negative cases. For instance, a positive association is when spatiotemporal data
are shown together, and where the target is not displayed in the location is a
negative association. However, as the volume of RFID data increases, the cur-
rent associative methods require some modification in order to gather more use-
ful data. In [20] is explained a combined ontology-based rules with hierarchical
clustering associative rules for reducing duplication of data during RFID data
mining process. In addition, [9] describe a new approach of multiple associative
rules to reduce RFID data abstraction level. The advantages and disadvantages
of the given algorithm are that it provides a new research direction for RFID
path data mining.
• Cluster Analysis: Clustering algorithm is also mainly used for RFID data mining.
It creates groups of data with high similarity or for various features of data [11].
The primary purpose of this rule is to reduce repetition and uselessly categorize
the gathered path data into the DB. While the data tags are clustered, every man-
ufacturing work is associated with data about space, time, and instruction of
operation. Thus, the flow of RFID data can be related to statistics or machine
learning model for analysis [21].
• Frequent Path Data Mining: Since path tracking and tracing are always useful
information to management and control center, RFID has been used to provide
an effect with relatively cheap cost to monitor location and position. Furthermore,
its efficiency is based on mining tools for pattern recognition [22]. Many algo-
rithms have been introduced over the period [23], since problem formulas and
path data mining algorithms vary depending on the application situation. As the
usefulness of RFID technology increases, frequent path mining methods will
draw more attention specifically in industry.
Data includes useful information with great value or also useful pattern recogni-
tion solutions. Generating efficient information from the RFID data can help with
task management, planning, and error detection, position and localization, and
tracking and tracing.
6 E. Elbasani et al.

RFID in Internet of Things

About RFID with IoT

Internet of Things was briefly described as an integration of network which collects

and exchanged data acquired through embedded sensors, actuators, or other devices
in things or physical objects in the world [24, 25].
The evolution of IoT over the three generations, as shown in Fig. 1.3, is charac-
terized by enabling key technologies, major enhancing of architectural solutions
and available products [26]. The objectives founding IoT in the first generation are
to identify and combine data from objects (tagged things) using the capabilities of
RFID and sensor connected through wireless sensor network systems (WSNs),
which have led the use of the Electronic Product Code (EPC) and machine-to-
machine communication (M2M) in industry. The objectives of the second genera-
tion, the full interconnection of things and Web of Things, focus on directly
connecting the objects to the Internet using technologies for integrating devices into
IP network and Web of Thing technologies. The third IoT generation is the era of
Social Internet of Things, cloud computing, information-centric networking for
IoT. This generation focuses on making intelligent objects community, exploiting
cloud computing technologies for storage, communications and processing capa-
bilities, and shifting from host-centric to content-centric networking.
Different IoT architectures have been proposed by different researchers. For
example, Atzori et al. [27] proposed to develop an architecture with three layers
for IoT such as the sensing, the network, and the application layers. Jia et al. [28]
divided the architecture into perception, network, and service or application lay-
ers. Liu et al. [29] proposed to design an IoT application infrastructure that con-
sists of physical, transport, middleware, and application layers. Yaqoob et al. [30]
divided IoT architecture into four layers of sensors and connectivity, gateway and

Fig. 1.3 Evolution of IoT

1 A Survey on RFID in Industry 4.0 7

network, management and security services, and smart applications. Da et al. [31]
described the four-layer architectures for service-oriented IoT that contains sens-
ing layer, networking layer, service layer, and interface layer, and the whole layers
of architecture must enforce security. In specific for industrial IoT applications,
the design architecture needs to consider energy, latency, throughput, scalability,
topology, security, and safety. Mashal et al. [32] proposed five layers of IoT archi-
tecture which contains perception layer, network layer, middleware layer, applica-
tion layer, and business layer and that architecture is added two new layers of the
processing layer and business layer because three-layer architecture is not enough
due to the development in IoT. Kumar and Mallick [33] summarized the different
IoT architecture layers as three-layer, five-layer, modified five-layer, six-layer,
seven-layer, and FOG architecture. The modified five-layer architecture consists
of application layer, sensing layer, communication layer, service layer, and infra-
structure layer. The six-layer architecture comprises focus layer, cognizance layer,
transmission layer, infrastructure layer, and competence business layer. The
seven-layer composes of application layer, ASM layer, service layer, communica-
tion layer, network layer, hardware layer, and environment. The FOG computing
layer, where smart sensors and smart network gateways do a part of the data pro-
cessing and analytics, consists of physical layer, monitoring layer, preprocessing
layer, storage layer, security layer, and transport layer, and that inserts monitoring,
preprocessing, storage, and security layers between the physical and trans-
port layers.
Furthermore, industry has utilized the advantage of IoT to enhance industrial
systems and applications. This rises to the industrial IoT (IIoT) and enables to inter-
connect humans, machines, and devices. In order to apply IoT in the industry, it
needs to consider technology requirement and architecture design needed for
numerous devices coordination system as trillion connected devices in the future of
the IoT system.
The Industrial Internet Consortium (IIC) [34] has provided Industrial Internet
Reference Architecture (IIRA) with the concept of industrial Internet system (IIS).
The IIRA drives IIoT systems through Industrial Internet Architecture Framework
(IIAF) based on ISO/IEC/IEEE 42010:2011.
For Industry 4.0 as subset of IIoT, the Working Group for Industry 4.0 [35] have
provided the Reference Architecture Model for Industry 4.0 (RAMI4.0), which con-
sists of a three-dimensional model and six layers of IT to representation of
Industry 4.0.
Industry has been taking advantage of IoT which is tending to be pervasively
used in various industries. The foundational technology of the IoT, such as RFID,
has been widely used in industry since 1980. RFID, which refers to radio-frequency
identification, provides the contributions of identifying and tracking objects to
gather data. Electronic Product Code (EPC) was developed by Auto-ID Centre at
MIT in 1999. EPC is electronic code stored data from RFID and serves as a univer-
sal RFID system. The use of RFID for sharing information has EPCglobal as stan-
dardization of EPC. According to the previous described architectures of IoT, RFID
and IoT can incorporate to embed RFID and gather information from RFID-enabled
physical objects.
8 E. Elbasani et al.

IoT makes objects sense physical real world through Internet protocol and wire
or wireless networks connecting and embedded sensors such as RFID, which can be
embedded into the objects and acquires data from “things” or physical objects. IoT
and RFID can be applied to production processes in order to control and improve
productivity in manufacturing.
Additionally, one of the technologies in automated identification and data cap-
ture (AIDC) is RFID, which can be used to automatically identify objects and
directly access data processing systems. Integrating RFID and IoT will provide for
monitoring objects embedded tags.
RFID technology is not only enabled to embed on physical objects with purpose
to gathering the physical information but also enabled to monitoring of chemicals,
gas, temperature, and humidity, during production or shipment. RFID was imple-
mented in the form of interdigitated chemo-capacitance provided with humidity-­
sensitive polymers. Oprea, A. et al. presented the development of novel
polymer-based capacitive humidity sensors integrated onto flexible RFID label sub-
strates [36]. In the perishables supply chain, RFID-based sensors play the important
role to detect the quality of perishables goods and examine the condition of perish-
able product during shipment through RFID-based temperature sensors [37].
Localization is one of RFID applications that has been used with the benefit of
passive RFID tags, no battery and cheap cost. RFID localization can be categorized
by different techniques as tag and reader localizations. For tag localization tech-
nique, an object or human needs to carry the tag, which will be detected through
multi-reader in environment, while for reader localization technique, the single
reader will read signal of multi-tags deployed in the environment instead of carrying
the tag on the object or human.
Additionally, Logistics 4.0 can use RFID technologies for resource planning,
warehouse management systems, transportation management systems, intelligent
transportation systems, and information security [38].

RFID-Based IoT in Industry 4.0

One of the core technologies, which have been used in Fourth Industry Revolution,
is the Internet of Things. That is the reason why Industry 4.0 is also known as indus-
trial Internet. IoT is able to connect physical objects, systems, and services, enable
object-to-object communication, and share physical information objects [39].
Among IoT technologies, RFID is widely used as a cheap cost with powerful capa-
bility to identify, trace, and track physical objects, providing real-time data about
the relative devices, decreasing labor, simplifying process, increasing the accuracy
about storages, and improving efficiency [31]. The integration of IoT and RFID into
Industry 4.0 enables a supply chain management, smart factory, as well as a smart
city. RFID was used to identify the good and monitor the stocks in warehouse
management.
1 A Survey on RFID in Industry 4.0 9

Supply Chain Management

Supply chain management runs through among supply chain partners supplier,
manufacturer, distributor, retailer, and consumer. The integration of IoT in the sup-
ply chain provides performance in every single of the processes in manufacturing
environments; man-to-man, man-to-machine, and machine-to-machine connections
are provided for intelligent perception [40]. IoT-enabled manufacturing character-
izes real-time data gathering and sharing among various devices, operators, and
other manufacturing resources [41].
IoT technologies help to monitor industrial operations and provide supply chain
visibility and traceability for the movements of resource and related information
flows in manufacturing operations [4, 42]. Supply chain visibility is to track and
trace a product from origin to destination along the logistics promises to enable real-­
time tracking of material flows, improved transport handling, as well as accurate
risk management, lay of the supply chain. It means all of the plays in supply chain
are able to be linked to each together and easily achieve procedures from supplier to
customer service. The important need is communication and exchanging informa-
tion with each other.
The technology of RFID can give information of physical objects and data col-
lected on several points in the supply chain with longer range of read better than bar
code technology and easier use for manufacturer. The integration of RFID and IoT
aims to be efficient sharing of resources controlling quality can be enabled by the
application of RFID in manufacturing shown in [7, 43–47].
The goal of logistic is not only transportation products in the right place, at the
right time, but also in the right quality. Thus, the term “Logistics 4.0” has been
highlighted in the past years. Also, Logistics 4.0 is concerned to the same condi-
tions as smart services and products, which are used to define “Smart Logistics” [48].
RFID is capable of identifying and capturing shipping products [49] described
logistics-oriented Industry 4.0 application model in two dimensions as physical
supply chain dimension and digital data value chain dimension. The physical supply
chain dimension is a level of automatic interaction along with physical thing; RFID
and IoT system can be used among two dimensions, for example, gathering data,
tracking and tracing, and making services along the supply chain automatically.
In 2003, the world’s leading retailers, Metro Group in Germany, Marks &
Spencer in the UK, and Walmart in the USA, utilize RFID in tracking supplies [50].
Fashion retailers also use RFID in supplies [51, 52].
Passive RFID tags can be exploited to detect and record customer’s activity or
shopping behaviors. For example, ShopMiner, RFID-Based Customer Shopping
Behavior Mining System, is able to identify customer shopping behaviors with high
accuracy and low overhead and is robust to interference [53]. RFID is not only able
to collect customer shopping behavior but also enhance customer shopping experi-
ence [54]. The requested services of a customer need to be trustable and make the
customer use the service and share their data without any concerns [26].
10 E. Elbasani et al.

Fig. 1.4 Smart factory

Smart Factory

Industry 4.0 is the link of IoT and the digital supply chain partners around a
smart factory that build up the business. A general concept of the main contribu-
tors is shown in Fig. 1.4. Each digital supply chain links applications to the busi-
ness process and each other as well. In addition, the smart factory relies on smart
manufacturing. The smart manufacturing connects all machines, sensors, or
devices in the manufacturing to intercorporate each other. The goal of smart man-
ufacturing is to run production and product transactions using advanced informa-
tion and manufacturing technologies smoothly and consistently. The approaches
of smart manufacturing, such as cyber-physical systems (CPS) and IoT frame-
works can be used for data collected and analyzed in real time through IoT [55].
The smart factory connects physical and digital world based on CPS. Wang
et al. [56] presented a conceptual design of smart factory framework consisting of
four layers, namely, physical resource layer, industrial network layer, cloud layer,
and terminals for supervision and control layer. The smart things may communi-
cate through the industry network to collect physical resource information in the
cloud and allow human to interact through the terminals. The smart factory leads to
optimize production in order to improve flexibility, productivity, resource and
energy efficiency, transparency, promoting integration, profitability, and friendli-
ness to labor. RFID can develop along with the smart factory through the advan-
tages of RFID tags. By using RFID tags embedded on the products, the raw
products in smart factory processes will be read and written so as to update infor-
mation during production.
1 A Survey on RFID in Industry 4.0 11

RFID in Communication and Networking for Industry 4.0

Industry 4.0 uses both a smart object and network for the integration of physical
devices into the network during producing and manufacturing processes.
Communications and networking can be evaluated as a comprehensive in-depth
technology for capabilities of Industry 4.0 features, such as analytics of large
data, embedded systems, and simulations also. RFID and RTLS (Real-Time
Location Tracking System) are one of the most supported technologies in Industry
4.0. Systems based on RFID and RTLS are achieving: [57]. Process-optimized
product production with a large number of versions, improved function and
flexible assembly, and high level of data transparency.
Mobile technology has shown significant strides since they first came out, and
now they are more than just communication devices. These tools allow the Internet
to receive and process large amounts of data and provide in-built sensors such as
high-quality cameras and microphones to record and transmit the recorded. Mobile
service is a feature provided by mobile device software. The software uses network,
integrated hardware, sensors, storage, and multimedia applications to deliver these
capabilities. Most devices (smart devices) these days come with some of the manu-
facturer’s own built-in services. Other services can be provided by the network
operator. Smartphone devices can be integrated with third-party service providers or
system integrators to enable connectivity on the network.
Specific case of mobile RFID method is when mobile services use RFID as a
technology to access information on objects tagged with RFID tag [58]. The mobile
RFID service simply provides information about objects tagged with RFID. Mobile
RFID services use services in relation with RFID reader and back-end network
infrastructure, but network and back-end servers are not always required to deliver
services. When it comes to mobile RFID, it’s a different technique from traditional
RFID. In mobile RFID, the tag is fixed and the reader is mobile and not the other
way around. There are certain advantages over traditional RFID. It’s enough if it is
not required a wire connected and only needs a mobile reader that can cover a wide
area instead of several of fixed readers.
Also, people are always carrying cell phones. Smart mobiles are devices with a
sense of intelligence, embedded with a lot of tiny sensors [59]. They can see, hear,
identify, track, and measure perfectly. People frequently are in mobile state and
continuously in need to interact with their surroundings every day. In an environ-
ment when objects or devices become smarter, by utilizing various sensors and
communication facilities, will bring in creating a smarter environment around people.
Mobile RFID services needed a mobile RFID platform that can be easily deliv-
ered using smart mobile devices. Cell phones integrated with RFID readers are
called mobile RFID readers. It is able to create a hybrid reader by combining UHF
readers with other identification technologies such as NFC and barcode. This com-
bination creates a unique identification platform. Smartphones with an RFID-­
integrated platform will be a key driver of the IOT service. Ubiquitous services are
available in a ubiquitous environment. Mobile RFID services are one driving force
to achieve these goals.
12 E. Elbasani et al.

Conclusions

Roles of RFID Technology in Retail

RFID technology in the supply chain is to transfer information in digital format and
then collect data from a variety of objects available to everyone. RFID technology
also has a great advantage in tracking and positioning individuals and objects in
factories or in other locations within the supply chain.
Meanwhile collecting massive amounts of data from enterprise operating sys-
tems can be a serious factor. Reading multiple tags at the same time can cause
conflicts and ultimately result in data loss. Therefore, studies are continuing on new
ways to prevent this phenomenon while anti-collision algorithms are applied. RFID
can be useful such as operational efficiency, enhanced visibility, reduced costs,
increased security, improved customer service, improved information accuracy,
and increased revenue. Table 1.1 lists the main tasks that include RFID that is most
relevant to real industry.

Roles of RFID Technology in Logistics

RFID technology has shown great benefits of noncontact, non-line-of-sight, auto-

matic identification and convenient operation, making it ideal for material tracking
and vehicle and shelf identification. The application of RFID technology effectively
implements logistics management automation and intelligence. This automation
and intelligence is primarily implemented in logistics operations management and
supply chain management.
RFID in the logistics operations management application is primarily performed
in transportation and distribution management, warehouse storage and material

Table 1.1 Advantages and disadvantages of RFID applications

Application Advantages Disadvantages
Retail Localization Tag collisions
Preventing the return of expired products in the Privacy for consumer can be
supply chain violated
Unique ID Metal-liquid signal absorption
Data integrations
Logistic Identification of goods Security
Improve the quality of inventory operations Tag collision
Improve the meticulous management Multi-RFID reader needed
Anti-counterfeit
Enhance the scope of information sharing
Manufacturing Streamline asset tracking Data management
Tracking with higher accuracy Not all companies use RFID
1 A Survey on RFID in Industry 4.0 13

management, and distribution processing management. RFID optimizes operations

management by quickly and accurately tracking each homework link in real time
through the entire logistics operations management process [60, 61].

Roles of RFID Technology in Manufacturing

RFID in manufacturing completes the operation of automated production and iden-

tification, tracks resources, and traces the position of products in production lines to
decrease labor costs and error rates and to maximize efficiency. Information in
RFID tags can be read first from the RFID reader, and product processing compa-
nies can create necessary information from relevant jurisdictions, such as technol-
ogy for integrated processing, unit date of processing, processing stage, additives
used, packing processing, accurate selection by weight on electronic tags, etc.
Sources of information are stored on electronic tags after the processing com-
pany’s data is faithful. The end user of the retailer or wholesale vendor can then
easily recognize the relevant information by looking up the product information
through the query terminal. The quality and transparency of the product make it
easy to track accidents. Table 1.1 shows the benefits and problems of RFID across
applications.

References

1. Ozguven, E. E., & Ozbay, K. (2015). An RFID-based inventory management framework for
emergency relief operations. Transportation Research Part C: Emerging Technologies, 57,
166–187.
2. Cai, H. B., Andoh, A. R., Su, X., & Li, S. (2014). A boundary condition based algorithm for
locating construction site objects using RFID and GPS. Advanced Engineering Informatics,
28(4), 455–468.
3. Zhong, R. Y., Dai, Q. Y., Qu, T., Hu, G. J., & Huang, G. R. (2013). RFID-enabled real-time
manufacturing execution system for mass customization production. Robotics and Computer-­
Integrated Manufacturing, 29(2), 283–292.
4. Zhong, R. Y., Li, Z., Pang, L. Y., Pan, Y., Qu, T., & Huang, G. Q. (2013). RFID-enabled real-­
time advanced planning and scheduling shell for production decision making. International
Journal of Computer Integrated Manufacturing, 26(7), 649–662.
5. Wong, W. K., Guo, Z. X., & Leung, S. Y. S. (2014). Intelligent multi-objective decision-­making
model with RFID technology for production planning. International Journal of Production
Economics, 147, 647–658.
6. Guo, Z. X., Ngai, E. W. T., Yang, C., & Liang, X. D. (2015). An RFID-based intelligent deci-
sion support system architecture for production monitoring and scheduling in a distributed
manufacturing environment. International Journal of Production Economics, 159, 16–28.
7. Qu, T., Yang, H. D., Huang, G. Q., Zhang, Y. F., Luo, H., & Qin, W. (2012). A case of imple-
menting RFID-based real-time shop-floor material management for household electrical appli-
ance manufacturers. Journal of Intelligent Manufacturing, 23(6), 2343–2356.
14 E. Elbasani et al.

8. Zhang, F. Q., Jiang, P. Y., Zheng, M., & Cao, W. (2013). A performance evaluation method
for radio frequency identification-based tracking network of job-shop-type work-in-process
material flows. Proceedings of the Institution of Mechanical Engineers, Part B: Journal of
Engineering Manufacture, 227(10), 1541–1557.
9. Wu, C. H., Ho, G. T. S., Yung, K. L., Tam, W. W. Y., & Ip, W. H. (2018). An RFID-based fallen
object detection system: A case study of Hong Kong’s light rail system. IEEE Journal of Radio
Frequency Identification, 2(2), 55–67.
10. Tang, T., Lin, T., Jie, W., & Ming, C. (2018). Full realization of digitalization is the only way
to intelligent manufacturing. Chinese Journal of Mechanical Engineering, 29, 366–377.
11. Alfian, G., Rhee, J., Ahn, H., Lee, J., Farooq, U., Ijaz, M. F., & Syaekhoni, M. A. (2017).
Integration of RFID, wireless sensor networks, and data mining in an e-pedigree food trace-
ability system. Journal of Food Engineering, 212, 65–75.
12. Wang, S., Wan, J., Zhang, D., Li, D., & Zhang, C. (2016). Towards smart factory for Industry
4.0: A self-organized multi-agent system with big data based feedback and coordination.
Computer Networks, 101, 158–168.
13. Blanchet, M., Rinn, T., von Thaden, G., & de Thieulloy, G. (2014). Industry 4.0: The new
industrial revolution. How Europe will succeed. In Think act. Munich: Roland Berger Strategy
Consultants GmbH.
14. Ma, H., Wang, Y., & Wang, K. (2018). Automatic detection of false positive RFID readings
using machine learning algorithms. Expert Systems with Applications, 91, 442–451.
15. Wang, C., Jiang, P., & Ding, K. (2017). A hybrid-data-on-tag-enabled decentralized control
system for flexible smart workpiece manufacturing shop floors. Proceedings of the Institution
of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science, 231(4),
764–782.
16. Hashem, I. A. T., Yaqoob, I., Anuar, N. B., Mokhtar, S., Gani, A., & Khan, S. U. (2015). The
rise of big data on cloud computing: Review and open research issues. Information Systems,
47, 98–115.
17. Ertek, G., Chi, X., & Zhang, A. N. (2017). A framework for mining RFID data from schedule-­
based systems. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 47(11),
2967–2984.
18. Feng, J., Li, F., Xu, C., & Zhong, R. Y. (2018). Data-driven analysis for RFID-enabled smart
factory: A case study. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 1–8.
https://doi.org/10.1109/TSMC.2018.2882838.
19. Bu, F. (2018). A data mining framework for massive RFID data based on Apriori algorithm.
Journal of Physics: Conference Series, 1087, 022020.
20. Hao, H., & Zhao, B. (2015). Application of RFID technology in development of warehouse
logistics management system by internet of things. International conference on education,
management, commerce and society (EMCS-15). Atlantis Press. https://doi.org/10.2991/
emcs-15.2015.89.
21. Ertek, G., Chi, X., & Zhang, A. (2017). A framework for mining RFID data from sched-
ule-based systems. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 47(11),
2967–2984
22. Liu, Y., et al. (2012). Mining frequent trajectory patterns for activity monitoring using
radio frequency tag arrays. IEEE Transactions on Parallel and Distributed Systems, 23,
2138–2149.
23. Shen, B., Zheng, Q., Li, X., & Xu, L. (2015). A framework for mining actionable navi-
gation patterns from in-store RFID datasets via indoor mapping. Sensors (Basel), 15(3),
5344–5375.
24. Kang, H. S., Lee, J. Y., Choi, S., et al. (2016). Smart manufacturing: past research, pres-
ent findings, and future directions. International Journal of Precision Engineering and
Manufacturing-Green Technology, 3, 111–128.
25. Xia, F., Yang, L. T., Wang, L., & Vinel, A. (2012). Internet of things. International Journal of
Communication Systems, 25(9), 1101–1102.
1 A Survey on RFID in Industry 4.0 15

26. Atzori, L., Iera, A., & Morabito, G. (2017). Understanding the internet of things: Definition,
potentials, and societal role of a fast evolving paradigm. Ad Hoc Networks, 56, 122–140.
27. Atzori, L., Iera, A., & Morabito, G. (2010). The Internet of things: A survey. Computer
Networks, 54(15), 2787–2805.
28. Jia, X., Feng, O., Fan, T., & Lei, Q. (2012). RFID technology and its applications in inter-
net of things (IoT). In Proceedings of the 2nd IEEE International Conference on Consumer
Electronics, Communications and Networks (CECNet), Yichang, China, Apr. 21–23, 2012,
pp. 1282–1285.
29. Liu, C. H., Yang, B., & Liu, T. (2014). Efficient naming, addressing and profile services in
internet-of-things sensory environments. Ad Hoc Networks, 18, 85–101.
30. Yaqoob, I., Ahmed, E., Hashem, I. A. T., Ahmed, A. I. A., Gani, A., Imran, M., & Guizani, M.
(2017). Internet of things architecture: Recent advances, taxonomy, requirements, and open
challenges. IEEE Wireless Communications, 24(3), 10–16.
31. Da Xu, L., He, W., & Li, S. (2014). Internet of things in industries: A survey. IEEE Transactions
on Industrial Informatics, 10(4), 2233–2243.
32. Mashal, I., Alsaryrah, O., Chung, T. Y., Yang, C. Z., Kuo, W. H., & Agrawal, D. P. (2015).
Choices for interaction with things on internet and underlying issues. Ad Hoc Networks, 28,
68–90.
33. Kumar, N. M., & Mallick, P. K. (2018). The internet of things: Insights into the building
blocks, component interactions, and architecture layers. Procedia Computer Science, 132,
109–117.
34. IIC Working TWG and SWG. (2015). Industrial internet reference architecture. Needham:
Industrial Internet Consortium.
35. Adolphs, P., Bedenbender, H., Ehlich, M., Epple, U., Hankel, M., Hoffmeister, M., et al.
(2015, July). GMA status report: Reference Architecture Model Industrie 4.0 (RAMI 4.0).
VDI/VDE-Gesellschaft Mess- und Automatisierungstechnik. Retrieved from https://www.
zvei.org/fileadmin/user_upload/Presse_und_Medien/Publikationen/2016/januar/GMA_
Status_Report__Reference_Archtitecture_Model_Industrie_4.0__RAMI_4.0_/GMA-
Status-Report-RAMI-40-July-2015.pdf
36. Oprea, A., Barsan, N., Weimar, U., Bauersfeld, M. L., Ebling, D., & Wöllenstein, J. (2008).
Capacitive humidity sensors on flexible RFID labels. Sensors and Actuators B: Chemical,
132(2), 404–410.
37. Delen, D., Sharda, R., & Hardgrave, B. C. (2011). The promise of RFID-based sensors in the
perishables supply chain. IEEE Wireless Communications, 18(2), 82–88.
38. Barreto, L., Amaral, A., & Pereira, T. (2017). Industry 4.0 implications in logistics: An over-
view. Procedia Manufacturing, 13, 1245–1252.
39. Zhong, R. Y., et al. (2017). Intelligent manufacturing in the context of industry 4.0: A review.
Engineering, 3(5), 616–630.
40. Tao, F., Cheng, Y., Xu, L. D., Zhang, L., & Li, B. H. (2014). CCIoT-CMfg: Cloud comput-
ing and internet of things-based cloud manufacturing service system. IEEE Transactions on
Industrial Informatics, 10(2), 1435–1442.
41. Bi, Z., Xu, L. D., & Wang, C. (2014). Internet of things for enterprise systems of modern
manufacturing. IEEE Transactions on Industrial Informatics, 10(2), 1537–1546.
42. Lu, B. H., Bateman, R. J., & Cheng, K. (2006). RFID enabled manufacturing: Fundamentals,
methodology and applications. International Journal of Agile Systems and Management, 1(1),
73–92.
43. Wang, Y. M., Wang, Y. S., & Yang, Y. F. (2010). Understanding the determinants of RFID
adoption in the manufacturing industry. Technological Forecasting and Social Change, 77(5),
803–815.
44. Huang, G. Q., Zhang, Y. F., Chen, X., & Newman, S. T. (2008). RFID-enabled real-time
wireless manufacturing for adaptive assembly planning and control. Journal of Intelligent
Manufacturing, 19(6), 701–713.
45. Liu, W. N., Zheng, L. J., Sun, D. H., Liao, X. Y., Zhao, M., Su, J. M., et al. (2012). RFID-­
enabled real-time production management system for Loncin motorcycle assembly line.
International Journal of Computer Integrated Manufacturing, 25(1), 86–99.
16 E. Elbasani et al.

46. Zhang, Y., Zhang, G., Wang, J., Sun, S., Si, S., et al. (2014). Real-time information cap-
turing and integration framework of the internet of manufacturing things. International
Journal of Computer Integrated Manufacturing, 28(8), 1–12. https://doi.org/10.1080/09511
92X.2014.900874.
47. Dai, Q. Y., Zhong, R. Y., Huang, G. Q., Qu, T., Zhang, T., & Luo, T. Y. (2012). Radio frequency
identification-enabled real-time manufacturing execution system: A case study in an automo-
tive part manufacturer. International Journal of Computer Integrated Manufacturing, 25(1),
51–65.
48. Zhong, R. Y., et al. (2015). A big data approach for logistics trajectory discovery from RFID-­
enabled production data. International Journal of Production Economics, 165, 260–272.
49. Hofmann, E., & Rüsch, M. (2017). Industry 4.0 and the current status as well as future pros-
pects on logistics. Computers in Industry, 89, 23–34.
50. Tajima, M. (2007). Strategic value of RFID in supply chain management. Journal of Purchasing
and Supply Management, 13(4), 261–273.
51. Landmark, A. D., & Sjøbakk, B. (2017). Tracking customer behaviour in fashion retail using
RFID. International Journal of Retail & Distribution Management, 45(7/8), 844–858.
52. Madhani, P. M. (2015). Enhancing customer lifetime value in fast fashion retailing with RFID
initiatives. International Journal of Business and Globalisation, 15(2), 205–237.
53. Zhou, Z., Shangguan, L., Zheng, X., Yang, L., & Liu, Y. (2017). Design and implementation
of an RFID-based customer shopping behavior mining system. IEEE/ACM Transactions on
Networking, 25(4), 2405–2418.
54. Choi, S. H., Yang, Y. X., Yang, B., & Cheung, H. H. (2015). Item-level RFID for enhancement
of customer shopping experience in apparel retail. Computers in Industry, 71, 10–23.
55. Niggemann, O., Biswas, G., Kinnebrew, J. S., Khorasgani, H., Volgmann, S., & Bunte, A.
(2015). Data-driven monitoring of cyber-physical systems leveraging on big data and the
internet-of-things for ciagnosis and control. In CEUR Workshop Proceedings, Vol. 1507, pp.
185–192.
56. Wang, S., Wan, J., Li, D., & Zhang, C. (2016). Implementing smart factory of Industrie 4.0: An
outlook. International Journal of Distributed Sensor Networks, 12(1), 3159805.
57. Rüßmann, M., Lorenz, M., Gerbert, P., Waldner, M., Justus, J., Engel, P., & Harnisch, M.
(2015). Industry 4.0: The future of productivity and growth in manufacturing industries.
Boston Consulting Group, 9(1), 54–89.
58. Temidayo, O., Oke, A. E., & Aigbavboa, C. (2018). Fusing RFID with mobile technology
for enhanced safety of construction project team members. Proceedings of the International
Conference on Industrial Engineering and Operations Management, 2018, 312–321.
59. Jiang, P., Sun, P., Kai, D., & Wang, C. (2015). Radio frequency identification-based graphi-
cal formalized deduction method for process tracking and its industrial applications. Chinese
Journal of Mechanical Engineering, 51(20), 9–17.
60. Alyahya, S., Wang, Q., & Bennett, N. (2016). Application and integration of an RFID-enabled
warehousing management system – A feasibility study. Journal of Industrial Information
Integration [Internet], 4, 15–25.
61. Wang, Q., Alyahya, S., Bennett, N., & Dhakal, H. (2015). An RFID-enabled automated ware-
housing system. International Journal of Materials, Mechanics and Manufacturing, 3(4),
287–293.
Chapter 2
Data Fusion-Based AI Algorithms
in the Context of IIoTS

Raluca Maria Aileni, Suciu George, Sever Pasca,

and Valderrama Sukuyama Carlos Alberto

I ndustrial Internet of Things: Overview in the Context

of Industry 4.0 and Massive Data Exploitation

In general, the IoT [1] concept enables Internet access to any type of electronic
device (thing). Thus, each “thing” will have an IP (Internet Protocol) address and will
be addressable using the standard Internet technology like DNS (domain name sys-
tem). In an automation context, the concept of the Internet of Things will be a device
or functionality that we can find in standard ISA-95 automation implementations.
Automation applications of the Internet of Things typically include sensors,
actuators, programmable logic controllers (PLCs), control loops, data analytics,
control models, simulators, optimizers, etc. Thereby, the concept can be viewed as
a physical device or as a functionality that is computed in a software system executed
on any type of device having sufficient computational resources. Currently, there is
no consistent technology that can be attributed to a device belonging to the Internet
of Things. Several standardization efforts are underway. For example, ITU (the
International Telecommunication Union) has a standardization working group
SG20 on “IoT and its applications including smart cities and communities” [2], and
IEEE (the Institute of Electrical and Electronics Engineers) has an ongoing IoT
Ecosystem study and the IEEE P2413 draft standard for an architectural framework
for the Internet of Things [2].
Nowadays, IoT (Internet of Things) has developed into a core technology of
hyper-connected society that connects and interacts with all things (elements/
objects), spaces, and environments in the real world. In accordance with this trend,

R. M. Aileni (*) · S. George · S. Pasca

Politehnica University of Bucharest, Bucharest, Romania
V. S. Carlos Alberto
University of Mons, Mons, Belgium

© Springer Nature Switzerland AG 2020 17

G. R. Kanagachidambaresan et al. (eds.), Internet of Things for Industry 4.0,
EAI/Springer Innovations in Communication and Computing,
https://doi.org/10.1007/978-3-030-32530-5_2
18 R. M. Aileni et al.

most of the industrial fields such as electricity, transport, medical, retail, and
manufacturing are trying to create advanced services and new businesses through
fusion with IoT technology. Especially in the manufacturing industries, there are an
increased number of smart factories that realize that intelligent production
automation can be achieved by using IoT technology.
The Industrial Internet of Things [3] gives us a way to obtain proper visibility
and insight into the operation of the company and assets by integrating software,
backend cloud computing, and storage systems. Shortly, we can say that the
Industrial Internet of Things offers a method of business operational processes
transformation using as feedback the results given by big data sets interrogation
through advanced analytics.
The Internet of Things (IoT) in Industry 4.0 [4] can support collaboration and
connection between objects automatically. Therefore, with the increasing number of
involved devices, IoT systems may consume large amounts of energy, and, taking it
into consideration, relevant energy efficiency issues have recently been attracting
much attention from both academia and industry.
Recently, IoT has evolved [5] into edge computing technology that enhances
data processing and analysis by decentralizing the computing power that was
performed in the central cloud to an end device. Since edge computing is a structure
suitable for a manufacturing environment requiring real-time processing of data and
fast response, many studies have considered edge computing as an optimized
platform technology for constructing a smart factory.
Industry 4.0 has as a reference the 4th Industrial Revolution, and the introduction
of Internet technology has as a reference the manufacturing industry, having the
purpose to obtain more intelligent factories, increased adaptability, and the efficiency
of the resources.
The future of Industry 4.0 is to go beyond new computer integrated manufactur-
ing techniques. Here, we can talk about nine technologies that are transforming
industrial production: simulation, autonomous robots, augmented reality, the
Industrial Internet of Things (IIoT), Digital Twin, the cloud, cybersecurity, additive
manufacturing, horizontal/vertical system integration, and, in the end, big data ana-
lytics. The industrial IoT links all of the components, equipment, and products with
embedded computing and sensors of the company. One of the essential components
of Industry 4.0 is the fusion of the physical and virtual world. This fusion can only
be made possible by cyber-physical systems (CPSs).
To continue with Industry 4.0, the most important feature is represented by smart
factories. A smart factory is, in fact, a factory where CPSs and IoT help people and
machines in the achievement of their tasks.
Some sub-processes [6] for a smart factory:
M2M (machine-to-machine) communication via IoT
Consistent communication between the sensor and the cloud
Robotics and innovative drive technologies integration
2 Data Fusion-Based AI Algorithms in the Context of IIoTS 19

Radio-frequency identification (RFID) as the basis for parts tracking and intelligent
products
The smart factory aims to [7] build an intelligent collaborative system in which
a variety of detecting devices (sensors) and machines that have been physically
connected to the manufacturing site is objectified and interconnected via the
telecommunication network to share the production process freely and control it
autonomously.
Thus, to accomplish this [8], it is important to help the platform structure and
communication environment in which data, produced from production devices or
IoT sensors, can be vertically and horizontally exchanged, integrated, and analyzed
with other devices and upper systems.
The rest of the paper is organized as follows: In Sect. 2 the IIoT systems are
presented, which are based on a modular architecture in Industry 4.0, followed by
the section portraying the cybersecurity layers of software and hardware. Section
4 depicts notions about big data fusion in the context of IoT systems based on
smart sensors, pursued by the fifth section introducing concepts about IIoTS
communication protocols and cloud computing services. Thus, Sect. 6 enounces
ideas about data fusion based on advanced AI algorithms, and henceforth, the last
section illustrates the conclusion of this paper and envisions the future work.

IIoT Systems Based on Modular Architecture for Industry 4.0

Industry 4.0 is related to the IoT (Internet of Things) (Fig. 2.1), CPSs (cyber-­
physical systems), ICT (information and communications technology), EA
(enterprise architecture), and EI (enterprise integration). The main objective of
Industry 4.0 is to accomplish a large amount of operational effectiveness and
productiveness and also a more elevated level of automatization [9]. The attributes
of the Industry 4.0 [10, 11] are digitization, optimization, customization of
production, automation and adaption, HMI (human-machine interaction), value-­
added services and businesses, and automatic data exchange and communication.
Such attributes are exceptionally related to technologies and modern algorithms
and show that the Industry 4.0 represents an industrial activity which has the
purpose to add significant value and information management.
IIoT (Industrial Internet of Things) semantically depicts an applied science
movement, whereas the Industry 4.0 is highly connected with the anticipated
financial impact. IIoT leads to the Industry 4.0 [12]. IIoT is not exclusively the
network of the physical objects in the industry; yet, it additionally incorporates the
computerized representations of the products, procedures, and factories, such as the
3D model or physical behavior models of the machines. In the final report of
Industry 4.0 Working Group, published in 2013 [13], industry experts and
academicians have been attempting to completely understand the Industry 4.0
20 R. M. Aileni et al.

CYBERSECURITY

INTERNET AUGMENTED
OF THINGS REALITY

CLOUD COMPUTING BIG DATA

INDUSTRY 4.0

SYSTEM AUTONOMOUS
INTEGRATION ROBOTS

SIMULATION ADDITIVE
MANUFACTURING

Fig. 2.1 Industry 4.0 [14]

repercussions for the manufacturing process. The most important repercussions in

this context are for manufacturing IT systems.
The word “Industry 4.0” relates to [15] production’s growing digitalization.
Experts look forward to a change from centrally managed manufacturing systems
with fixed process flows to cyber-physical system (CPS) manufacturing systems for
the future.
The possibilities to customize [16] the production process are quite limited due
to the static structure of factories. Based on this shortcoming, the idea of a modular,
decentralized, and digital system arose. The modularity shall provide the flexibility
to design and implement customized system structures. Another benefit would be
the possibility to extend the system with other modules directly.
Advanced digital technology [17] is used in the production process. Industry 4.0
is going to transform the entire production domain and it will lead to increased
efficiency. The traditional production is going to change the relationships among
producers, suppliers, and customers. The connection between humans and machines
is going to be changed as well. At this point, there are nine technology trends that
design the building blocks of Industry 4.0:
• Big Data and Analytics – In the context of Industry 4.0, the collection and com-
plete evaluation of data from numerous distinct sources – the equipment used in
production and the enterprise- and customer management-related systems – are
going to become standard in order to support taking real-time decision.
2 Data Fusion-Based AI Algorithms in the Context of IIoTS 21

• Autonomous Robots – Robots will finally interact with one another. They are
going to work safely with humans, learning from them simultaneously. These
robots won’t be expensive, and they will have a greater potential compared to the
robots currently used in the manufacturing industry.
• Simulations – They are going to be utilized furthermore in plant operations in
order to take advantage of real-time data, mirror the physical world, and create a
virtual model which can incorporate machines, products, and humans. This will
allow workers to test and develop the machine settings used in the next product
in line in the virtual created world before the physical modification, thereby
decreasing the setup time of the machines and increasing their quality.
• Horizontal and Vertical System Integration – Using the Industry 4.0, companies,
their departments, functions, and different capabilities will be more cohesive, as
cross-company, common data-integration networks emerge and facilitate entirely
automated value chains.
• Cybersecurity – Alongside the increased connectivity and usage of standard
communication protocols that work with Industry 4.0, the demand of protecting
crucial industrial systems and manufacturing lines from cybersecurity hazards is
completely growing. As a result, there are some essential aspects, such as secure
and decent communications, complex identity, and access management of users
and machines.
• The Cloud – More production-related undertakings are going to necessitate
increased data sharing across sites and company bounds. Simultaneously, the
efficiency of cloud technologies will advance, achieving reaction times of only
several milliseconds. As a result, machine data and performance will increasingly
be deployed to the cloud, enabling more data-driven services for production
systems.
• Additive Manufacturing – Companies have just begun to adopt additive manu-
facturing. A good example of this concept would be 3D printing, which is used
to produce individual components and to prototype. With Industry 4.0, these
additive-manufacturing methods can be widely adopted to produce small batches
of customized products that offer construction advantages.
Three frameworks can further specify key elements of Industry 4.0: the smart
product, the smart machine, and the augmented operator.
The smart product’s guiding concept is to extend the workpiece’s function to an
active aspect of the scheme. The products obtain a memory on which operational
information and specifications are stored as an employee construction plan
directly.
Smart machine defines the cyber-physical production systems (CPPS) process of
machines. Decentralized self-organization enabled by CPS will replace the
traditional manufacturing hierarchy. They portray autonomous parts with local
control intelligence that can interact through open networks and semantic
descriptions to other field systems, manufacturing modules, and products. This
enables machines to organize themselves within the manufacturing network.
22 R. M. Aileni et al.

Production lines [18] will become so flexible and modular, that under situations
of highly flexible mass manufacturing, it can be designed even in the smallest
lot size.
The augmented operator, in the challenging setting of fully modular manufactur-
ing systems, aims at the worker’s technological support. Industry 4.0 does not tend
to migrate to evolving manufacturing without the worker.
In the Industry 4.0 [19], IoT is required to offer auspicious solutions for the
operation and role of the industrial systems. Industry 4.0 is an initiative with
technological innovations such as cloud computing, electric vehicles (EV), the
Internet of things, artificial intelligence, and cyber-physical systems. Industry 4.0
has attracted attention from governments, researchers, and industries.

 ybersecurity Aspects Correlated with the Software

C
and Hardware Layers

An IIoT system [20] is designed to exist for a long time. In an IIoT infrastructure, a
significant number of devices are monitoring and controlling a process control
entity. More IoT-based industrial entities or plants may interact with each other
involving these IoT smart devices. In real life, IIoT can be applied in smart cities for
transportation, smart homes, wearables, emergency services, air and energy
monitoring systems, and oil and gas industry. As the number of IIoT devices
increases, the number of open connections increases as well and so the vulnerability
to attacks. As a known example of weakness, devices placed in the exterior are
exposed to tampering. Regarding devices which monitor a specific area, someone
can control them to track something else.
The NIST (the National Institute of Standards and Technology) created the
Cyber Security Framework (NIST, 2014), which organizes the following categories
of cybersecurity activities:
• Identify – the growth of the organizational knowledge on how to control cyberse-
curity risks of the systems, data, and assets.
• Protect – the development and the implementation of the appropriate protection
in order to ensure delivery of critical services.
• Detect – the development and the implementation of the appropriate activities in
order to identify the circumstance of a cybersecurity event.
• Respond – the development and the implementation of the appropriate activities
in order to act when a cybersecurity event is detected.
• Recover – it enables separation of this framework from the others.
These categories [21] can provide features for an organization to handle risk
decisions and address threats and incident management. The Cyber Security
Framework considers the importance of recovery planning and suggests the
following for recovery of any attacks caused by cyberattacks. The IIoTS has building
support for the sensor-cloud. The sensors are integrated into this sensor-cloud, being
organized as wireless sensor networks (WSNs). For better wireless network security,
2 Data Fusion-Based AI Algorithms in the Context of IIoTS 23

Application and control system: information

Application layer management center, various end users,
computing, processing and display platform, etc.

Processing layer Information integration category and directional

information transmission

Main bearing network: internet, mobile

Transmission layer communication network, computer, local area
network, etc.

Analysis layer Analyze and select information and delete

information junk.

Information perception and collection: RFID

Perception layer reader, bar code read-only device, sensor video
camera, etc.

Fig. 2.2 Software layers [23]

the solution is a network node trust with several detection routes. In an IIoT
environment, the sensor-cloud assures communication of the WSN. The information
gathered from the sensors is sent to the cloud. Next, data is stored and processed in
more data centers, which are mainly created for this purpose. For users, this
processed data can be sent on demand from the cloud.
Too many devices included in an IIoT infrastructure [22] can slow the update of
software defenses, as well as hardware security measures. Known defensive
solutions in the IIoT domain are replacing the system components or shutting down
the compromised system. This is not enough for IIoTS security.
An IIoT system uses wireless communication, having a sensor/device layer; a
communication layer, i.e., the medium where wireless signals propagate; and the
user application layer (Fig. 2.2), where data is received. Changing the position of an
IIoT surveillance camera or taking (a camera) and placing it somewhere else.
Stealing components of a device results in data disclosure. As a solution, an IIoT
device can be equipped with motion or temperature sensors, e.g., gyroscopes.
Unauthorized access has the main purpose of finding vulnerabilities, including
passive activities like probe computers or network, capturing, analyzing, and
disclosure of packet content.

Attacks on the Perception Layer

Several sensors such as WSN, Zigbee, RFID, infrared, QR code, and other kinds are
used to collect information (temperature, humidity, force, pH levels, pressure,
vibration, etc.) which is then transmitted through the network/communication layer
to the central information processing unit.
24 R. M. Aileni et al.

This layer is the most susceptible to hardware attacks. The attacker is required to
be in the network or physically close to the nodes of the system. Here are some
common attacks on this layer:
• Fake node injection – an infected node is placed between the nodes of the net-
work, making an opening for the attacker into the network and hence gaining the
ability to control all the data flow. This intervention could cause the node to stop
transmitting data and, in consequence, destroy the whole network.
• Hardware tempering – the attacker changes parts of the node’s hardware or
replaces the whole node and therefore gains information regarding the network,
like communication key, cryptographic key, routing table, etc. This puts into
danger even the higher layers of the network.
• Malicious code injection – the transmission is halted because of DoS attacks in
WSN or virus on the nodes. The network loses resources and makes the services
unavailable.
• Sleep denial attack – nodes located in more remote places in the network gener-
ally run on batteries and enter a sleep mode when they are not in use in order to
increase their battery life. This type of attack implies that wrong input is fed to
the nodes to prevent them from entering the sleep mode, which leads to power
being consumed until the nodes shut down.
• WSN node jamming – wireless sensor networks perform on radio frequency. In
order to deny service and communication between nodes, the attacker sends
noise signals over the network or jams the signals of WSN.
• RF interference of RFIDs – RFIDs work on radio frequency as well. The attacker
interferes with the RFID signal by sending noise over the network and creating a
distortion in the node communication.

Attacks on the Communication Layer

The information is transmitted successfully in this layer through various mediums

like RFTD, satellite, infrared, and Wi-Fi units, with regard to the type of sensors and
the sensitivity of data.
From a [24, 25] network point of view, these are some examples of known active
attacks in IIoTS:
• Packet dropping.
• RFID spoofing – intrusion in the network and manipulation of data.
• RFID unauthorized access – the data in nodes can be easily modified (read, write,
delete), as there is no secured authentication system.
• Fraudulent packet injection attacks – i.e., replay captured packet, sending false
packets that cause malfunctioning of specific tasks.
• Sinkhole attacks – they guide all signals from WSN nodes to a single common
point. Such attacks nullify the safety of data and drop all the packets rather than
deliver them to the destination.
2 Data Fusion-Based AI Algorithms in the Context of IIoTS 25

• Routing information attacks – instantaneous attacks where the intruder can coil
up the system and make routing loops, send false alerts, shorten or amplify
source courses, and permit or drop movement, all by spoofing, replaying, or
changing routing data.
• Man-in-the-middle attacks – the intruder meddles with two sensor nodes in order
to gain access to confined information. This kind of assaults violates the privacy
of the nodes and doesn’t require the intruder to be physically present on the site;
using the communication protocol of the IoT is enough.

Attacks on the Processing Layer

This layer blends the communication layer and the application layer. Here is
where all the cloud and intelligent computing are done. The performance of this
layer includes storing the data from lower-level layers into the database and
managing the service. It can as well compute information and process data
automatically.
The most common type [25] of attacks that occur in this layer are cloud attacks
because of the fact that, during this phase, the data is sent to the cloud. These are
some frequent attacks that affect the network:
• Shared resources – the attacker is able to monitor all the resources being shared
between virtual machines using convert channels; thus the information risks to
be compromised.
• Third-party relationships – third-party web services combine more than one
source, leading to an increase in the number of security issues like network and
data security.
• Underlying infrastructure security – lower-level layers can’t be accessed by the
developer, so the underlying layer’s security becomes the responsibility of the
provider. Though the developer could implement a secure application, its security
is still vulnerable due to lower layers.

Attacks on the Application Layer

This layer grants services as per usual demand. The processed information coming
from lower layers is involved in producing services that are useful for end users. It
provides a platform for such applications from which users could benefit in
many ways:
• Malware attacks (viruses, worms, Trojan Horse, spyware) – they destroy system
files, tamper with data, and control the computer. Either it is disguised in software
or the attacker injects it.
26 R. M. Aileni et al.

• Denial-of-service (DoS) attacks – they cause depletion on the system’s energy,

slow processing, until the computer blocks or shuts down. Distributed DoS
(DDoS) is designed to attack more devices which will send a high volume of
traffic to networks that will consume computer resources.
• Phishing attacks – leakage of data.
• Malicious scripts – they cause the system to close up and, by running executable
active-x scripts, the data is stolen.
There are measures deployed in IIoTS as following:
• Embedded security: device authentication and authorization to access the net-
work for certain operations using authentic software, specially designed for that
network.
• Data integrity: applying cryptographic hash functions on the data so it isn’t tem-
pered with when it reaches the receiving end.
• Fragmentation redundancy scattering (FRS): the sensitive data on the cloud is
separated into fragments and stored on distinct servers. The fragments don’t
contain any meaningful information by themselves, fact which minimizes the
risk of leakage.
• In IIoTS, confidentiality and integrity could be solved with data encryption.
• An IIoTS firewall includes rules and attacks signatures for filtering the traffic;
they are always updated to repair new issues regarding vulnerabilities or attacks
(i.e., patches). Unexpected behavior in IIoTS.
• Anomalies in datasets can [25] be classified as such:
–– Contextual anomaly – a command with its data value predefined range is exe-
cuted in a no-show time duration.
–– Collective anomaly – multiple TCP (Transmission Control Protocol) requests.
–– In networks: attack patterns or signatures – a new attack can be found if it
corresponds with the pattern; this method is integrated into firewalls;
analysis of behaviors that differ from previous ones sometimes brings false
positives.
IIoT comes [23] with the sensor-cloud paradigm and the trust-based mecha-
nisms: independent sensor-cloud (ISC), collaborative sensor-cloud (CSC), and
mutual sensor-cloud (MSC).
In ISC, a sensor network assigns to each node a trust value and a trust value
threshold; a cloud assigns to the data center a trust value and a trust value
threshold.
In CSC, a sensor node is assigned to a trust value only by its network, while a
trust value threshold is assigned by both the sensor network and the cloud.
In MSC, sensor nodes’ trust values and trust value thresholds are assigned by
their sensor network. Data centers’ trust values and trust value thresholds are
assigned by the cloud. However, here, apart from the other mechanisms, trust value
thresholds for data centers are mutual between WSN, cloud, and users.
2 Data Fusion-Based AI Algorithms in the Context of IIoTS 27

 ig Data Fusion in the Context of IIoT Systems Based

B
on Smart Sensors

In the last few years, sensors [26] have been utilized all over the place, and nowa-
days, they have the potential to impersonate human beings very tightly. Data fusion
enables this to happen, which influences a microcontroller to fuse the individual
information gathered from different sensors in order to obtain a more precise and
valid perspective on the information than the one collected from each discrete
sensor all alone.
A basic example of sensor fusion is represented by an e-compass, in which the
mix of a 3D magnetometer and a 3D accelerometer gives functionality to the
compass. Furthermore, more elaborated fusion technologies give clients an
innovative and more improved experience, by consolidating 3D accelerometers, 3D
gyroscopes, and 3D magnetometers (which estimate the constituent elements of the
magnetic field in a specific direction, in comparison with the spatial orientation of a
given device). Every one of these sensor types provides proper functionality and has
the following constraints:
• Accelerometers: X, Y, and Z axis linear movement detection, however sensible to
vibrations
• Gyroscope: presents a rotational sensor (pitch, roll, and yaw), but zero bias drift
• Magnetometer: X, Y, Z axis magnetic field detection, but sensitive to magnetic
interference
Another example of data fusion is a gaming platform that can recognize feelings
electronically by keeping under observation and gathering data from physiological
factors and states:
• Muscle relaxation (MR), through a pressure sensor
• Heart rate variability (HRV), through a two-electrode ECG on a chip
• Sweat (S), through capacitive sensor
• Attitude (A), through an accelerometer that monitors the state of relaxation of a
human being
• Muscle contraction (MC), through a pressure sensor
Information [27] collected from the sensors can be utilized for various services,
giving several advantages and simplifying people’s lives. Another utilization is for
data mining and other use cases that increase security and privacy problems
regarding IoT technologies. More capabilities can be assembled with data fusion
and REC technologies. IoT will cover each part of our lives in under 10 years,
sensor fusion having the main impact of this spreading phenomenon. Big data
processing has a significant impact nowadays, with mechanisms and tools that
produce various services.
IoT creates a large amount of information which would be less valuable except
the case in which people can learn by utilizing them. Data fusion, one of the critical
28 R. M. Aileni et al.

challenges in IoT, creates innovative services. In IoT environments, data fusion

represents a big challenge that should be tackled to create innovative services. In the
particular case of smart cities communities, when 50–100 billion gadgets start
sensing [28], a fusion will be fundamental. Data fusion can be applied at two levels:
cloud and network levels.
Data fusion can be described as a processing method that connects, compounds,
reunites, and incorporates information from various sources. Moreover, it contributes
to developing a context-awareness model that permits comprehension of the
situational context. The sensor information filter emphasizes the necessity of
avoiding vast volumes of information to be transmitted over the network.
Recent buyers of the IIoT [29] have recognized competitive upper hands and
innovative business models for expanding the income, reduce prices, and also
enhance the customer service and support. Expressions such as artificial intelligence,
predictive maintenance, and augmented and virtual reality are not anymore just
favorite expressions. The ideas, innovations, and concepts provided by the adopters
are being embraced and connected to these industries each day. Nevertheless, IIoT
adoption challenges continue to exist.
IIoT (Industrial Internet of Things) represents a system which associates and
incorporates OT (operational technology) environments, including ICS (industrial
control systems) with business frameworks, processes, and analytics. IIoT systems,
besides ICS and OT, are associated widely to different frameworks and individuals.
Also, they utilize sensors and actuators that interface with the physical world;
uncontrolled changes can produce precocious circumstances. IIoT has the following
advantages: a large number of sensors can gather and process the information; based
on what the information reveals, the system will react. The precise connectivity
increases the risk of cyberattacks by the individuals who might need to cut down the
framework.
IIoT systems present [30] a framework of the Industry 4.0, in which research
topics are categorized into the smart design, smart machines, smart monitoring,
smart control, and smart scheduling. Smart design is evolving with the rapid
development of new technologies such as virtual reality (VR) and augmented reality
(AR), so the traditional design will be upgraded. Design software such as computer-­
aided design (CAD) and computer-aided manufacturing (CAM) can interact with
physical smart prototype systems in real time, enabled by three-dimensional (3D)
printing integrated with CPSs and AR. In Industry 4.0, intelligent machines are
made using smart robots and other types of smart objects capable of perceiving and
interacting in real time. Smart monitoring is an essential aspect for the operations,
maintenance, and optimal scheduling of Industry 4.0 manufacturing systems. The
widespread deployment of various types of sensors makes it possible to achieve
smart monitoring. For example, data and manufacturing factors such as temperature,
electricity consumption, vibrations, and speed can be obtained in real time. In
Industry 4.0, high-resolution and adaptive production control (i.e., smart control)
can be achieved by developing cyber-physical production-control systems. Smart
control is mainly executed to physically manage various smart machines or tools
through a cloud-enabled platform. End users can switch off a machine or a robot via
2 Data Fusion-Based AI Algorithms in the Context of IIoTS 29

their smartphones. The smart scheduling layer mainly includes advanced models
and algorithms to draw on the data captured by sensors. Data-driven techniques and
advanced decision architecture can be used for intelligent scheduling. For example,
to achieve real-time, reliable scheduling and execution, distributed smart models
using a hierarchical interactive architecture can be used.
The replacement of human [30] resources with intelligent systems requires the
adaptation of these systems so that they can safely monitor the industrial process,
which implies the development of specialized IIoT systems. Although there are
many techniques to identify the issues that can be selected, intelligent production
puts a strong emphasis on predictive capacity development across the factory. The
ability to accurately predict problems can ensure the availability of equipment and
availability for demand-side supply offers while optimizing energy consumption
and the cost of life equipment by supervising sensor systems and depending on the
sensor parameters to make individual decisions.
The topology of an industrial network contains standard components, such as
devices, systems, and databases, which facilitate the factory data flow. This topology
describes a particular type of network that is used in production to orchestrate the
production process, known as an automation network. Since the adoption of
standards is a significant issue for these types of industrial systems, the topology
presented is abstracted at a level that allows hierarchical data flow to be displayed
without being burdened with low-level details. Data transmission begins with
instruments at the level of low-level recording measurements (e.g., temperature)
and culminates in end-user access to relevant information.
The sensors transmit continuous measurements (e.g., room temperature, micro-
phone signal, robot actuator status) to programmable logic controllers (PLCs).
PLCs are digital computers logically programmed to automate the production
process. The PLC is programmed by automation engineers to evaluate the parameters
measured by each sensor and to initiate appropriate actions according to their state.
Measurements transmitted to PLCs are persistent in memory at established intervals.
This type of data transmission is standard in industrial IT systems and is referred to
as serial data, measured or timed data (Fig. 2.3). Subsequently, persistent PLC time
data is transferred to an archive at a specified interval (e.g., every 2 hours), the file
typically taking the form of a relational database. Once the data persists in the file,
computer systems can consume data from the warehouse and generate reports for

Server Database Application End-user

Sensors

Fig. 2.3 General mode of data transmission

30 R. M. Aileni et al.

end users. Examples of this type of systems are building management systems, air
quality monitoring systems, and industrial production systems. Access to data in the
archive can be done using standard database and I/O interfaces. However, big data
may require some processing to be used; for example, the noise of a microphone
recorded signal will be eliminated. While accessing archive data allows it to be used
in some industrial applications, the high latency characteristics of archiving in
automation networks means that it is not inherently appropriate to real-time
industrial applications. If real-time sensor data are to be accessed, there must be a
direct communication with the devices (e.g., PLC usage) on the lower levels of an
automation network. This can be done using industry protocols and interfaces such
as Modbus, LonWorks, BACnet, OLE Process Control (OPC), and MTConnect or
where smart sensors are using MQTT, COAP, and HTTP. However, as mentioned
above, the adoption of intelligent sensors and emerging technologies in large-scale
industrial plants can be restricted by quality regulation and control, as well as the
high risks associated with adopting new technologies. Therefore, initial data
management and infrastructure requirements for intelligent manufacturing may
need to consider how old and emerging technologies can work. An alternative to
storing data is also their storage in the cloud.
In order to minimize human intervention in industrial processes, applications
should reduce the workload of personnel by making calculations, process the data
and display them in a way that the supervisor understands, or implement notification
modules considering the data processed to alert authorized personnel or the
technical team.
Figure 2.4 illustrates an example of an IIoT system as a real-world network of
digital counterparts. Product life cycle management (PLM) provides data about
the production line status. Digital factory is a comprehensive model of the real
factory that can be used for communication, simulation, and optimization during
its life cycle. Software products in the domain of digital factory typically come
with different modules enabling functions such as material flow simulation,
robot programming, and virtual commissioning. In the context of IIoT, the digital

Fig. 2.4 An IIoT system

as a real-world network of
digital counterparts [28]
2 Data Fusion-Based AI Algorithms in the Context of IIoTS 31

factory can be considered as a complement to PLM. While PLM aims to integrate

data along the product life cycle, the digital factory incorporates the data of
production resources and processes. For the IIoT both are necessary, high-fidelity
models of the product and its products based on data from sensors subsequently
processed.

I IoTS Communication Protocols and Cloud Computing

Services

IIoTS communication [31] was developed for many years. A good example of this
could be the vehicular networks where two or more cars can communicate between
them with a wireless network connection (WSN). This dynamic communication
was named ad hoc networks (VANET), and it was made for customers who wanted
a safer car with different functionality such as traffic monitoring system, update,
emergency warning, and road assistance. Another part in [32] the automotive
domain is cloud computing, based on network solutions. The data collected from
external factors was stored in a cloud platform for safety reasons. Part of IIoTS
communication [33] is used in the medical domain, more precisely in the
healthcare sector, which can help doctors to monitor and control patient’s
healthcare.
SCADA (Supervisory Control and Data Acquisition) Systems [34] are tasked
with monitoring and controlling a network of PLCs (programmable logic controllers)
and RTUs (remote terminal units) that measure the performance of local operations,
with the use of sensors, and acquire information. It offers real-time updates on the
status of its physical processes within its network. A secure communication needs
to be established between the devices on the field and the SCADA host in order to
transmit the data, to register the updates, and to gain control over the physical
processes successfully. This concept is accomplished through an array of specific
communication protocols, which carry the information coming from devices on the
field all the way to a cloud or local central control center. Vendors started coming up
[35] with their own communication protocols before certain organizations began
developing open standards. The merging with IIoTS has led to an increase in the
number of protocols. In spite of the considerable number of both proprietary and
nonproprietary protocols, some are more used than others: Modbus, PROFIBUS
(Process Field Bus), Wi-Fi, HTTP (Hypertext Transfer Protocol), CoAP (Constrained
Application Protocol), DNP3 (Distributed Network Protocol-3), AMQP (Advanced
Message Queuing Protocol), and WiMAX (Worldwide Interoperability for
Microwave Access).
From a communication standpoint [36], IoT and CPSs depend heavily
on mobile Internet (i.e., telecommunication networks), which haven’t played too
big of a part in industrial communication thus far. Both IT and telecom
networks couldn’t manage to meet the requirements for a trustworthy and adept
32 R. M. Aileni et al.

communication. However, things stand differently now as, on the one hand,
Ethernet TSN (Time-­Sensitive Networking) assures hard real-time capabilities
and, on the other hand, the telecom industry sees the industrial automation field as
a promising application for their products. Both technological advancements have
the chance to alter the structure of industrial networks and even might be a
prerequisite for implementing IIoTS and CPSs.
In the early stages of IIoTS communication [36, 37], in order to overpower the
constraints brought in by parallel cabling between sensors, actuators, and
controllers, as well as to bring to an end the gap on the lower levels of the
automation hierarchy, dedicated automation networks named fieldbus systems
were created. Table 2.1 shows the chronological evolution of industrial
communication and notes milestones in diverse technology fields which are
relevant to the field of automation.
Up until a few years ago, industrial communication relied on a combination of
fieldbus systems, Ethernet-based approaches, and some wireless solutions. But all
of these clashed hard with the market. The introduction of IoT and CPS concepts
into the automation world put the request of integrating information flows in a wider
context: the technological breakthroughs in communication make possible the inter-
connection on a larger and more fine-grained scale [38].
From an application perspective, the trend is to move the business logic into
cloud-based applications. Cloud computing [38, 39] is an element of the system that
combines both hardware and software parts in the data centers. There is a series of
clouds in this domain: public cloud, created as a pay-as-you-go manner for the
general public, and private cloud, which corresponds to internal data centers of a
business that are not available for other organizations or people.
The architecture of the cloud can be divided into four layers: hardware (data
center), infrastructures, platform, and application. By themselves, the layers can be
seen as consumers for the layer below and as services for the layer above. The cloud
platform can be shared into other sections, more accurately in three sections [40]:
• Popular service as a service (SaaS) – it concerns the supplying of applications
running in cloud environments. Applications usually can be accessed through a
web browser or a thin client.
• Platform as a service (PaaS) – it refers to platform-layer resources (e.g., software
development frameworks, operating system support, etc.).
• Infrastructure as a service (IaaS) – it provides processing, storage, and network
resources and grants the customer the control of the operating system, the
storage, and the applications.
Cloud computing model has the economic advantages of freeing the business
owner from the obligation of investing in the infrastructure, renting resources to
needs, and paying only for the usage. Furthermore, it favors the decrease in operating
costs because of the fact that service providers don’t need to supply capacities
according to peak load. As far as technical advantages, it ensures certain benefits
like energy efficiency, elasticity, and flexibility, optimization of hardware and soft-
ware resource utilization, and performance isolation [40].
Table 2.1 The evolution of industrial communication and related technology fields
2

–1970 1970–1980 1980–1990 1990–2000 2000–2010 2010–2020

Computer ARPANET Ethernet Internet WWW Bluetooth TSN
networks (Advanced ISO/OSI MAP (World Wide Web) SOAP (Time-
Research (International (Manufacturing Automation WLAN (Simple Object Access Protocol) Sensitive
Agency Organisation for Protocol) (Wireless Local Area ZigBee Networking)
Network) Standardization) MMS Network) 6LoWPAN
(Manufaturing Messaging (IPV6 over Low – Power WPAN)
Specification) UWB
(Ultra Wide Band)
Mobile internet 2G: GSM 3G: UMTS 4G: LTE+
(Global System for (Universal Mobile 5G
Mobile Telecommunications System)
Communication) 3G: HSPA
(High Speed Packet Access)
4G: LTE
(Long-Term Evolution)
IT trends Ubiq. Comp. IoT CPS Ind. Internet
(Internet Of Things) (Cyber Physical System) Industry 4.0
Industrial Modbus PROWAY SDS Powerlink
Data Fusion-Based AI Algorithms in the Context of IIoTS

communication (Process Data Highway) (Smart Distributed Modbus/TCP

HART System) (Transport Control Protocol)
(Highway Addressable DeviceNet PROFINET
Remote Transducer) ControlNet (Process Field Net)
CAN TTP IEC61784
(Controller Area network) (Time Triggered EtherCAT
INTERBUS Protocol) (Ethernet for Control Automation
PROFIBUS LIN Technology)
(Process Field Bus) (Local Interconnect OPC UA
LON Network) Wireless HART
(Local Operating Network) EN50170 IEC61784-2
AS-I Ethernet/IP ISA 100.11a
33

(Actuator Sensor Interface) IEC61158

34 R. M. Aileni et al.

Data Fusion Based on Advanced AI Algorithms

The data fusion community is in continuous evolution. Applications of data fusion

range from DoD (Department of Defense) applications such as automatic target
recognition for smart weapons and battlefield surveillance to non-DoD applications
such as improved medical diagnosis and condition-based maintenance.
Data fusion was [41] created initially for the military (initiative promoted by
DoD). A few applications include automatic target recognition for smart weapons,
guidance for autonomous vehicles, or remote sensing. Data fusion implies collecting
data from multiple sources (sensors) and combining them for improved performance.
It is proven that this technique is better than the standard single source data
acquisition. A better example to understand data fusion is the ability of humans and
animals to recognize objects. It is not sufficient to tell the quality of an item (e.g.,
food) just by seeing it, and such, a combination of sight, touch, smell, and even taste
provides a much better evaluation of the specific item and suggests a definition of
data fusion based on human senses.
In the military domain, an [42] easy example of data fusion would be combining
the information obtained from an infrared imaging sensor and a radar. We can use
these to determine the exact location of an object, a determination that could not be
done by each sensor individually. Different types of sensors are implemented to
obtain data fusions, meaning there are many techniques we can use in our system
depending on the data that is combined and the nature of the information.
In the military field, there are also included applications such as guidance for
autonomous vehicles, target recognition, remote sensing, battlefield surveillance,
and automated threat recognition. Military applications have also extended from
cognition monitoring of weapons and machinery to monitoring the health status of
individual soldiers and assistance to logistics [42].
Surface-to-air and air-to-air defense systems have been developed by the military
to detect, track, and identify hostile aircraft, missiles, and antiaircraft weapons.
Sensors such as passive electronic support measures, radar, identification-friend-foe
sensors, infrared, electro-optic image sensors, and visual (human) sightings are
used by these defense systems. Challenges to these data fusion systems include
enemy countermeasures, the need for rapid decision-making, and potentially large
combinations of target-sensor pairings [41].
These techniques involving data [42] fusion can be performed using advanced AI
algorithms (neural networks, genetic algorithms, artificial intelligence), probabilistic
approaches (Kalman filter, Bayesian approach with Bayesian network and
­state-­space models), and statistical approaches (multiple variants statistical analysis
and its corresponding data mining engine, weighted combination). AI algorithms
(Fig. 2.5) are used mainly to derive estimators or classifiers and act as a fusion
framework of classifiers/estimators.
If we refer to the nonmilitary field, we have different areas in which data fusion
has a significant impact: medical applications, smart buildings, autonomous vehi-
cles, robots, industrial applications, etc. Automatic incident detection can be pre-
2 Data Fusion-Based AI Algorithms in the Context of IIoTS 35

Fig. 2.5 List of AI algorithms [43]

sented as an application of data fusion based on AI algorithms. Koppelman and Ivan

developed this application. One of the data sources is a probe vehicle, while the
other is a fixed detector (inductive loop detectors). Based on a neural network, two
strategies were tested. The first one implies observing the traffic directly and deter-
mines whether an incident occurs or not. In the second strategy, different incident
detection algorithms preprocess data from each source, resulting in information
which is combined by the neural network. It was determined that the latter strategy
provided better results and improved performance. Data fusion techniques were
also applied in the road network control issues, advanced driver assistance, advanced
traveler information systems, and traffic demand estimation [43].
The area of medical diagnosis is another example of data fusion system for non-
military applications. Nowadays, in order to provide improvements in medical diag-
nostic capability, there are being developed increasingly sophisticated sensors for
medical applications. The purpose behind fusing these data together is to reduce the
chance of false diagnoses and to improve the diagnostic capability [43].

Conclusions

The generation of IIoTS-based platform will favor the creation of a reliable, tangi-
ble, and safe open platform which will further allow the integration and develop-
ment of new services, control software, and applications. In addition, a testing
framework for IIoT-based platforms will be developed in order to grant security and
reliability to the structure and communication environment so that the data collected
from sensors and devices can be exchanged bilaterally and assimilated by other
devices and systems.
In the context of production, the evolutionary process will soon get to the point
of networked manufacturing systems, which hold a high degree of autonomy.
36 R. M. Aileni et al.

Systems like these depict autonomous components that come in contact through
open networks and semantic descriptions to other field systems, manufacturing
modules, and products. Such a thing allows the machines to organize themselves
within the network.
Data fusion can connect, reunite, and incorporate data coming from various
sources. The scope behind using data fusion is to improve performance by reducing
the network traffic and energy consumption, as well as increasing the accuracy of
the results. By involving advanced AI algorithms, numerous techniques become this
way available, leading to the development of a great number of applications with
significant impact.

Acknowledgments This work has been supported in part by UEFISCDI Romania and MCI
through projects PARFAIT, OPTIMUM, PAPUD, and Smart-PDM and funded in part by
European Union’s Horizon 2020 research and innovation program under grant agreement No.
777996 (SealedGRID project), No. 787002 (SAFECARE project), and No. 826452 (Arrowhead
Tools).

References

1. Delsing, J. (2017). IoT automation: Arrowhead framework. Boca Raton/London/New York:

CRC Press.
2. Zuehlke, D. (2018). SmartFactory – Towards a factory-of-things. Annual Reviews in Control,
34, 129–138.
3. Gilchrist, A. (2016). Industry 4.0: The industrial internet of things. Berkeley: Apress.
4. Wang, K., Wang, Y., Sun, Y., Guo, S., & Wu, J. (2016). Green industrial internet of things archi-
tecture: An energy-efficient perspective. IEEE Communications Magazine, 54(11), 48–54.
5. Lee, S.-J., Lee, Y.-K., & Yoon, D. (2018). Design considerations for automated crane assign-
ment. IOP Conference Series: Materials Science and Engineering, 417(1), 012019.
6. Wang, L., & Wang, G. (2016). Big data in cyber-physical systems, digital manufacturing and
Industry 4.0. International Journal of Engineering and Manufacturing (IJEM), 4, 1–8.
7. Jeschke, S., Brecher, C., Meisen, T., Özdemir, D., & Eschert, T. (2016). Industrial internet of
things and cyber manufacturing systems (Springer Series in Wireless Technology) (pp. 3–19).
Cham: Springer International Publishing.
8. Lee, Y.-K., Lee, S.-J., & Yoon, D. (2018). Architecture of middleware system for intelligent fac-
tory automation based on industrial internet of things. International Conference on Big Data,
IoT, and Cloud Computing (BIC).
9. Thames, L., & Schaefer, D. (2016). Software-defined cloud manufacturing for Industry 4.0.
Procedia CIRP, 52, 12–17.
10. Posada, J., Toro, C., Barandiaran, I., Oyarzun, D., Stricker, D., de Amicis, R., & Vallarino, I.
(2015). Visual computing as a key enabling technology for Industry 4.0 and industrial internet.
IEEE Computer Graphics and Applications, 35(2), 26–40.
11. Roblek, V., Meško, M., & Krapež, A. (2016). A complex view of Industry 4.0. SAGE Open,
6(2), 1–11.
12. Song, H., Rawat, D., Jeschke, S., & Brecher, C. (2016). Cyber-physical systems: Foundations,
principles and applications. London: Academic Press.
13. Kagermann, H., Helbig, J., Hellinger, A., & Wahlster, W. (2013). Recommendations for imple-
menting the strategic initiative INDUSTRIE 4.0: Securing the future of German manufacturing
industry. Final report of the Industrie 4.0 Working Group.
2 Data Fusion-Based AI Algorithms in the Context of IIoTS 37

14. Wolf, A. (2017, May 19). Industry 4.0 – Digitisation of manufacturing. Bosch India Blog
[Online]. Available: https://blog.boschindia.com/iot/industry-4-0-digitisation-of-manufactur-
ing/. Almada-Lobo, F. (2016). The Industry 4.0 revolution and the future of manufacturing
execution systems (MES). Journal of Innovation Management, 3(4), 16.
15. Lang, S., Reggelin, T., Motasem, J., & Hofmann, W. (2018). Towards a modular, decentral-
ized and digital Industry 4.0 learning factory. The 6th International Conference on Enterprise
Systems (ES 2018), At Limassol (Cyprus).
16. Scalabre, O. (2018). Embracing Industry 4.0 – and rediscovering growth. Boston Consulting
Group. Retrieved from https://www.bcg.com/capabilities/operations/embracing-industry-
4.0-rediscovering-growth.aspx
17. Weyer, S., Schmitt, M., Ohmer, M., & Gorecky, D. (2015). Towards Industry 4.0 –
Standardization as the crucial challenge for highly modular, multi-vendor production systems.
IFAC-PapersOnLine, 48(3), 579–584.
18. Ivanov, D., Dolgui, A., Sokolov, B., Werner, F., & Ivanova, M. (2016). A dynamic model
and an algorithm for short-term supply chain scheduling in the smart factory Industry 4.0.
International Journal of Production Research, 54(2), 386–402.
19. Stankovic, J.-A., Le, T., Abdeltawab, H., & Tian, Y. (2019). Hardware/software security
patches for internet of trillions of things. arXiv:1903.05266.
20. Barrett, M.-P. (2019). Framework for improving critical infrastructure cybersecurity (Version
1.1).
21. Zhou, L., & Guo, H. (2018). Anomaly detection methods for IIoT Networks. IEEE International
Conference on Service Operations and Logistics, and Informatics.
22. Sarma, M. (2018). The global race for artificial intelligence: Weighing benefits and risks. New
Delhi: Institute for Defence Studies & Analyses.
23. Ahemd, M.-M., Shah, M. A., & Wahid, A. (2017). IoT security: A layered approach for attacks
& defenses. International Conference on Communication Technologies (ComTech).
24. Zhu, C., Rodrigues, J. J. P. C., Leung, V. C. M., Shu, L., & Yang, L. T. (2018). Trust-based
communication for the industrial internet of things. IEEE Communications Magazine, 56(2),
16–22.
25. Suciu, G., Scheianu, A., Bălăceanu, C. M., Petre, I., Dragu, M., Vochin, M., & Vulpe, A. (2018).
Sensors fusion approach using UAVs and body sensors. Advances in Intelligent Systems and
Computing, 747, 146–153.
26. Wang, M., Perera, C., Jayaraman, P. P., Zhang, M., Strazdins, P., Shyamsundar, R. K., &
Ranjan, R. (2016). City data fusion: Sensor data fusion in the internet of things. International
Journal of Distributed Systems and Technologies (IJDST), 7(1), 15–36.
27. Suciu, G., Suciu, V., Martian, A., Craciunescu, R., Vulpe, A., Marcu, I., Halunga, S., & Fratu,
O. (2015). Big data, internet of things and cloud convergence – an architecture for secure
E-health applications. Journal of Medical Systems, 39(11), 141.
28. Newton, M. (2017). The state of the industrial internet of things. https://www.sciencedirect.
com/science/article/pii/S0166361517307285.
29. Eden, P., Blyth, A., Jones, K., Soulsby, H., Burnap, P., Cherdantseva, Y., & Stoddart, K. (2017).
SCADA system forensic analysis within IIoT. In Cybersecurity for Industry 4.0 (pp. 73–101).
Cham: Springer International Publishing.
30. He, W., Yan, G., & Li Da, X. (2014). Developing vehicular data cloud services in the IoT envi-
ronment. IEEE Transactions on Industrial Informatics, 10(2), 1587–1595.
31. Al-Sarawi, S., Anbar, M., Alieyan, K., & Alzubaidi, M. (2017). Internet of things (IoT) com-
munication protocols: Review. 8th International Conference on Information Technology
(ICIT).
32. Sajid, A., Abbas, H., & Saleem, K. (2017). Cloud-assisted IoT-based SCADA systems secu-
rity: A review of the state of the art and future challenges. IEEE Access, 4, 1375–1384.
33. Janicke, H., Nicholson, A., Webber, S., & Cau, A. (2015). Runtime-monitoring for industrial
control systems. Electronics, 4(4), 995.
34. Babu, B., Ijyas, T., Muneer, P., & Varghese, J. (2017). Security issues in SCADA based indus-
trial control systems. 2nd International Conference on Anti-Cyber Crimes (ICACC).
38 R. M. Aileni et al.

35. Wollschlaeger, M., Sauter, T., & Jasperneite, J. (2017). The future of industrial communica-
tion: Automation networks in the era of the internet of things and Industry 4.0. IEEE Industrial
Electronics Magazine, 11(1), 17–27.
36. Zhang, Y., Qian, C., Lv, J., & Liu, Y. (2016). Agent and cyber-physical system based self-­
organizing and self-adaptive intelligent shopfloor. IEEE Transactions on Industrial Informatics,
13(2), 737–747.
37. Armbrust, M., Fox, A., Griffith, R., Joseph, A., Katz, R., Konwinski, A., Lee, G., Patterson,
D., Ariel Rabkin, A., Ion Stoica, I., & Zaharia, M. (2010). A view of cloud computing.
Communications of the ACM, 53(4), 50–58.
38. Botta, A., de Donato, W., Persico, V., & Pescape, A. (2016). Integration of cloud computing
and internet of things: A survey. Future Generation Computer Systems, 56, 684–700.
39. Xiong, K., & Perros, H. G. (2009). Service performance and analysis in cloud computing.
2009 IEEE Congress on Services – I, Los Angeles, CA.
40. Jing, L., Wang, T., Zhao, M., & Wang, P. (2017). An adaptive multi-sensor data fusion method
based on deep convolutional neural networks for fault diagnosis planetary gearbox. Sensors,
17(2), 414.
41. Faouzi, N.-E. E., Leung, H., & Kurian, A. (2011). Data fusion in intelligent transportation
systems: Progress and challenges – A survey. Information Fusion, 12(1), 4–10.
42. Il, M. L., Hall, D., & Llinas, J. (2017). Handbook of multisensor data fusion: Theory and
practice (2nd ed.). Boca Raton: CRC Press.
43. Ahmadi, S. (2014). LTE-advanced: A practical systems approach to understanding 3GPP LTE
releases 10 and 11 radio access technologies. Waltham: Elsevier.
Chapter 3
Presents the Technology, Protocols,
and New Innovations in Industrial Internet
of Things (IIoT)

Quanxin Zhao

Innovation and Application

Application field and industrial product are fundamental for IIoT. The primary
IIoT application fields include cyber-physical systems, Industry 4.0, machine-to-
machine communications, multi-agent systems, and wireless sensor networks.
Moreover, the primary industrial products in the applications of IIoT include car
and truck, car part, electronic power, electronics, foodstuff, heating, and oil
and gas.
Reference architecture is the guidance for designing and producing industrial
products. RAMI 4.0, the reference architecture model for Industry 4.0, can be
analyzed from the aspects of layers and hierarchy levels. On the one hand, the lay-
ers include business, function, information, communication, integration, and asset.
On the other hand, the hierarchy levels include connected world, enterprise, work
center, station, control device, field device, and product [1–3].
It is heating up to discuss about the Industrial Internet of Things (IIoT) and how
it brings benefits to businesses. IIoT can help companies maximize interconnected
real-time operational data. IIoT will not only give enterprises quick and flexible
decision-making ability, but help companies optimize their operating costs. Making
the digital transformation of enterprises possible, industrial IoT technology helps
achieve below benefits in a step-by-step manner:
• Automate predictive analysis.
• Better connecting people, assets, and processes.
• Fully optimized enterprise performance.
• Immediately realize the collaborative operation of personnel.

Q. Zhao (*)
Sichuan Netop Telecom Co., Mianyang, Sichuan, China
e-mail: [email protected]

© Springer Nature Switzerland AG 2020 39

G. R. Kanagachidambaresan et al. (eds.), Internet of Things for Industry 4.0,
EAI/Springer Innovations in Communication and Computing,
https://doi.org/10.1007/978-3-030-32530-5_3
40 Q. Zhao

Technology and Protocol

The protocol is important for connecting the products. The IIoT communication
protocols can be classified into some kinds, which are as follows: First, the wireless
sensor network includes WirelessHART, IEC 62591 (WirelessHART), ISA 100.11a,
and Zigbee. Second, the M2M communication includes CoAP, OPC-UA, DDS, and
Modbus. Third, the messaging includes MQTT, AMQP, and XMPP. Fourth, the low-­
power wide-area network (LPWAN) includes NB-IoT, Sigfox, LoRa, and
LoRaWAN. Fifth, the cellular network includes 5G, 4G, 3G, LTE, WiMax, GPRS,
and GMS. Sixth, the wireless local area network (WLAN) includes IEEE 802.11.
Seventh, the wireless personal area network (WPAN) includes IEEE 802.15.4.
Above protocols are described below [4–6].

WLAN

Wireless local area network (WLAN), a data transmission system, which uses the
technology of radio frequency (RF) to substitute the existing local network of
twisted copper wires (i.e., coaxial) by electromagnetic waves to make the LAN with
wirelessness. With a simple access architecture, the network enables users to achieve
the ideal realm of “information and convenience.”

WPAN

The communication technology of wireless personal area network (WPAN) utilizes

wireless communication connection. Recommended by IEEE, there exist three
technologies, which are 802.15 on the basis of Bluetooth technology, high-­
frequency 802.15.3 (i.e., UWB), and low-frequency 802.15.4 (i.e., Zigbee). WPAN
is designed to achieve activities within tiny range, business emerging wireless com-
munication network technology with rich types, specific groups, and wireless
seamless connections [7–10].

5G

In the technology of mobile cellular communication, the fifth-generation mobile

communication technology (i.e., 5G) is the up-to-date. Compared with previous
mobile cellular communication, such as 4G (LTE-A, WiMax), 3G (UMTS, LTE),
and 2G (GSM), there exists significant change in the performance, such as
­throughput and system capacity, jitter and latency, lower power consumption and
costs, and massive attached devices with massive connectivity [11–15].
3 Presents the Technology, Protocols, and New Innovations in Industrial Internet… 41

There are two phases for 5G. In Release-15, the first phase is to accommodate
deployments of existing commercial. In Release-16, the second phase can be a can-
didate technology of IMT-2020 for the International Telecommunication Union
(ITU). In the IMT-2020 specification of ITU, its speed can be up to 20 Gbit/s con-
tributed by both high-capacity Multi-Input-Multi-Output (MIMO) and large chan-
nel bandwidth.

1G/2G/3G/4G

In 1978, the first large-capacity mobile cellular communication system in the

world, named the Advanced Mobile Phone Service (AMPS) system, is designed by
Bell Labs. In order to access the public telephone network automatically for the
mobile terminals, frequency reuse technology is adopted by AMPS within the
entire service coverage area of the system. Therefore, it has better voice quality and
greater system capacity and further solves the large capacity faced by the public
mobile communication system. It requires conflicts with spectrum resource limita-
tions. The beginning to deploy on a large-scale deployment of the AMPS systems
is started by the United States in the late 1970s. AMPS has won unanimous praise
from users for its excellent network performance and service quality, which has
facilitated on the global scale of researching cellular mobile communication tech-
nologies starting after AMPS’s fast development in the United States.
The ETACS system, the NTT/JTACS/NTACS system, and the NMT-450 system
are developed by Europe and Japan by the mid-1980s. These cellular mobile com-
munication networks are analog-based frequency division duplex (FDD) systems
(known as 1G systems) [16–20].

900/1800 MHz GSM Mobile Communication

GSM mobile communication network (known as 2G), working in the 900/1800 MHz
frequency band, can offer voice and data services. In the GSM mobile communica-
tion system, the TDMA technology is adopted by the wireless interface, and the
MAP protocol is adopted by the core network mobility management.
The 900/1800 MHz GSM second-generation mobile cellular communication
service includes types of services:
• End-to-end double-way voice services
• Mobile short-message service, using the mobile station initiated by the GSM
network and the message platform, and the message service received by the
mobile station
• Mobile data services
• Mobile bearer services
• Mobile supplementary services, for example, call forwarding service, display of
calling numbers
42 Q. Zhao

• Mobile intelligent network services, for example, prepaid services

• Domestic roaming and international roaming services
Operators which have the 900/1800 MHz GSM second-generation mobile cel-
lular communication service establish their GSM networks, and the provided type
of mobile communication service may be part or all. A mobile communication ser-
vice is provided by a network of different operators. It is necessary to establish state
approved international communication gateways for mobile network accessing ser-
vices of international communications.

800 MHz CDMA Mobile Communication

800 MHz CDMA mobile communication, operating in the 800 MHz frequency
band, is also the second-generation mobile cellular communication service provid-
ing the voice and data services. Narrowband code division multiple access (CDMA)
technology is adopted in wireless interface. CDMA, as well as IS-41 protocol, is
adopted in the core network mobility management.
The CDMA working in 800 MHz acts as the second-generation mobile cellular
communication service which includes types of services:
• End-to-end double-way voice
• Mobile message service, using the mobile station initiated by the CDMA net-
work and the message platform, and the message service for mobile stations
• Mobile bearer services
• Mobile data services
• Mobile supplementary services, for example, call forwarding service and display
for calling number
• Mobile intelligent network services, for example, prepaid services
• Domestic roaming services
• International roaming services
Operators of the 800 MHz CDMA second-generation mobile cellular commu-
nication service should establish their mobile CDMA communication networks;
besides, the provided type of mobile communication service may be part or all.
The network through which a mobile communication service is provided may be
the same operator’s network or may be completed by a network of different
operators.
The third-generation mobile cellular communication (i.e., 3G) service provides
voice services, data services, and video image services.
3G mobile communication services provide services of broadband multimedia,
including 144 kbps in high-speed mobile circumstance, 384 kbps in walking/slow-­
moving circumstance, and 2 Mbps data in indoor circumstance. Transmission and
guarantee high reliability of service quality (QoS). The third generation of mobile
cellular communication services includes all kinds of mobile services including
multimedia.
3 Presents the Technology, Protocols, and New Innovations in Industrial Internet… 43

The operator of the 3G mobile communication service must establish a 3G

mobile communication network by itself, and the type of mobile communication
service provided may be part or all of the end-to-end service. The network through
which a mobile communication service is provided may be the same operator net-
work facility or may be completed by network facilities of different operators.
International communication services must be approved and established by the state
for international communication services.
Although the 3G system had solved the shortcomings of 1G and 2G systems, its
actual speed was far from the expected value. Then the international organizations
3GPP and 3GPP2 started a new round of 3G evolution plan, and LTE stood out
among many candidate standards. The 3GPP organization launched the “LTE Plan,”
which achieves a smooth transition from 3G to 4G, so LTE is also known as the
quasi-4G standard. The ultimate goal of the program is to provide a low-latency,
high-throughput, large-scale coverage wireless communication network. LTE has
two working modes: FDD and TDD. Among them, LTE-TDD has its own intellec-
tual property rights. It was commercialized in China at the end of 2013, and its
high-speed bandwidth capability has brought a new experience to users. Currently,
more than 4 million LTE base stations have been deployed worldwide, and this
number is expected to grow with future growth.

NB-IoT

In Internet of Everything, NB-IoT (i.e., the Narrowband Internet of Things) is one

of the important areas. NB-IoT developed on cellular network technology usually
work together with UMTS networks, GSM networks, or LTE networks consuming
180 kHz of bandwidth.
For the emerging technology of low-power wide-area network (LPWAN),
NB-IoT provides wireless data connection in WAN for massive low-power devices.
For those low-power devices, they have the characteristics of long standby time and
coverage of indoor wireless data connection.

XMPP

XML is the standard universal markup language. As a subset environment of XML,

XMPP protocol has flexible to development and scalable application. To develop
applications or add new functionality to existing customer systems easily, software
protocol included in XMPP on the server side enables the server talking to
another one.
XMPP defines roles of client, server, and gateway. The server has following
functionalities, such as recording the client information, connecting management,
and routing information. To connect heterogeneous instant messaging systems, the
gateway is used.
44 Q. Zhao

AMQP

As open standard advanced message queuing protocol in an application layer, the

Advanced Message Queuing Protocol (AMQP) can give services of unified messag-
ing for message-oriented middleware. Based on AMQP, clients and message mid-
dleware transmit messages, no matter the client/middleware different products,
different development languages, and so on. Implementations in Erlang include
RabbitMQ and more. It supports the architecture of various message exchanges:
store and forward, distributed transactions, publish subscriptions, content-based
routing, file transfer queue, and point-to-point connection.

MQTT

Message Queuing Telemetry Transport (MQTT) is designed on the basis of para-

digm of the publish/subscribe according to ISO/IEC PRF 20922. MQTT, a publish/
subscribe messaging protocol, is running on the TCP/IP protocol suite, which is
suitable for low-end hardware remote devices with conditions of poor network con-
nection. A message middleware is required to do this.
The MQTT has features as follows:
1. One-to-many message publishing is applied by publish/subscribe message mode.
2. Payload content shield during message transmission.
3. TCP/IP network connection provision.
4. Message publishing service classified for three kinds:
• “At most once” where message loss/duplication is permitted, as which
depends on the underlying TCP/IP network.
• “At least once” to permit message duplication for ensuring the arrival of
message.
• “Only once” ensuring the arrival of message only once.
5. Small transmission, low overhead, and network traffic are reduced to minimum
during protocol exchange.

Modbus

For communication with programmable logic controllers (PLCs), a serial commu-

nication protocol (Modbus) is published in 1979. In the area of industrial communi-
cation protocols (De facto), the industry standard Modbus can connect industrial
electronic devices. The main reasons why Modbus is used more widely than other
communication protocols are:
3 Presents the Technology, Protocols, and New Innovations in Industrial Internet… 45

1. Published and no copyright requirements.

2. Easy to deploy and maintain.
3. There are not many restrictions for the vendor to modify the mobile local bits or
bytes.
Modbus allows multiple (approximately 240) devices to be connected to com-
municate on the same network, for example, a device that measures temperature and
humidity and sends the results to a computer. In the Data Acquisition and Monitoring
Control System (SCADA), Modbus is typically used to connect monitoring com-
puters to remote terminal control systems (RTUs).

DDS

DDS (Data Distribution Service) is a distributed real-time data distribution service

middleware protocol, which is “TCP/IP” in distributed real-time network and used
to solve network protocol interconnection in real-time network, whose function is
equivalent to “bus on bus”:
What is a distributed real-time network system?
For a real-time system, it focuses on the correctness of calculation, which usually
relies on the result of correctness of program logical and time. There will be system
error if system time constraints are not met. Therefore, it requires the system to
timely respond to external events requests, to finish the event processing with time
constraint, and timely control and coordinate tasks operation.
The system is the normal state of reality, from the defense system to the tempera-
ture and humidity remote control system of the warehouse, which is common in all
walks of life.

OPC-UA

The core of the OPC Unified Architecture (OPC UA) communication standard is
interoperability and standardization. The traditional OPC technology solves the
problem of interoperability between hardware devices at the level of control;
besides, communication standardization at the level of enterprise is also required.
Prior to OPC UA, the access specifications were based on Microsoft’s COM/DCOM
technology, which would give new levels of communication a weakness that cannot
be eradicated. In addition to the inflexibility of traditional OPC technology and the
limitations of platforms, the OPC Foundation has released the latest unified method
of data communication, covering timely OPC data access specifications, different
aspects of the OPC Historical Data Access Specification, the OPC Alarm Event
Access Specification, and the OPC Security Protocol, but with functional extensions.
46 Q. Zhao

OPC UA is another breakthrough after the success of traditional OPC technol-

ogy, making data collection, information modeling, and communication reliable
and secure.

CoAP

In the CoAP network, there are massive resource-constrained devices; therefore,

devices with small space of memory and power with limited value cannot afford
traditional large HTTP protocol application. Then, a CoAP protocol is proposed on
the basis of the REST architecture by the IETF’s CoRE working group.
In the 6LowPAN protocol stack, CoAP protocol is designed on the layer of appli-
cation. After describing the content, features, and interaction model of the CoAP
protocol in detail, this article uses the Contiki embedded operating system on the
uIPv6 START KIT wireless network development kit, which both implements
browser a CoAP protocol and uses its own client. The CoAP protocol implemented
program increases function of database interaction, thereby realizing web interface
the function of real-time and historical data view.

Zigbee

Zigbee is a low-speed short-distance communication protocol in wireless networks.

In the layers of the media access and physical, Zigbee applied the IEEE 802.15.4
standard, which has the features of low speed and power consumption, low cost and
complexity, reliable and secure, supporting massive online nodes, and multiple
online topologies.

ISA100.11a

ISA100.11a is one of the international standards for industrial wireless sensor net-
works. On the basis of IEEE 802.15.4, but only using its 2.4 GHz ISM band (with-
out sub-1 GHz band). It has been approved by the International Electrotechnical
Commission (IEC) in September 2014 and has become the official international
standard; the standard number is IEC 62734.
The protocol defines the following five types of device roles:
1. The upper computer control system is a platform for users and engineers to inter-
act with the ISA100.11a system.
2. Gateway defines host computer interfaces to the network and is also the interface
with other factory-level networks. Multiple gateways can exist in an ISA100.11a
network system.
3 Presents the Technology, Protocols, and New Innovations in Industrial Internet… 47

3. The backbone router (BBR) is the infrastructure of the ISA100.11a backbone

network and can provide data routing in the network of backbone. The commu-
nication protocol on the ISAl00.11a backbone network can be a wireless proto-
col, such as Wi-Fi, or a wired protocol, such as a standard Ethernet. That is, the
backbone of ISA100.11a is another high-performance network.
4. Field devices, which have one terminal node device and two field routers.
Terminal node devices typically have sensors/actuators. In addition to the sensor/
actuator, the field router also has a routing function to route data of the terminal
node device within the ISA100.11a subnet.
5. Handheld device is a device that accesses the ISA100.11a system for on-site
maintenance and configuration.

WirelessHART

WirelessHART is designed for the first wireless communication standard with open
interoperable characteristics to satisfying process industry plant applications critical
needs such as stable, secure, and reliable during wireless communications.
Each WirelessHART network has the following components:
1. Field device attached to the plant equipment or process.
2. Gateways enabling wireless field devices to communicate with host applications
with the connection of the backplane and plant-level communication networks.
3. Network management software is to configure network, schedule devices com-
munication, manage packet routing, and monitor health of network.

Standards

Standard and stack can guide the company for developing products. The IIoT com-
munication standard and technology stack can be classified into layers and groups.
On the one hand, layers are consisting of a framework layer, a transport layer, a
network layer, a link layer, and a physical layer. On the other hand, main groups
include WPAN, WLAN, cellular network, LPWAN, satellite network, and tradi-
tional industrial computer network (fieldbus), which will be described in detail.

Fieldbus

Fieldbus was developed internationally in the late 1980s and early 1990s for field
intelligent device interconnection communication networks in the fields of process
automation, manufacturing automation, and building automation. Fieldbus serves
as the basis for the plant’s digital communication network and communicates the
48 Q. Zhao

links between the production process site and control equipment with high control
and management level. Fieldbus is a fully control system as well as a distributed
grassroots network. This integrated technology with intelligent sensing, control,
computer, digital communication, and other technologies has attracted worldwide
attention and will lead to profound changes in the structure and equipment of auto-
mation systems. Many internationally powerful and influential companies have
developed fieldbus technologies and products from varying degrees.
The fieldbus system (known as the fieldbus control system (FCS)) is the 15th-­
generation control system. People refer to the FCS as the first generation of the
Pneumatic Signal Control System before the 1950s. The electric analog signal con-
trol system such as 4–20 mA is called the second generation, and the digital com-
puter centralized control system is called the third generation. The distributed
control system (DCS) since the middle of the decade is called the fourth generation
The FCS breaks through the limitations of the DCS using a communication-­
dedicated network and adopts an open and standardized solution to overcome the
closed system defects. Meanwhile, the centralized and decentralized system of dis-
tributed systems has become fully distributed structures, and the functions of con-
trol have been fully decentralized to the scene. It can be said that fieldbus system
prominent features are openness, digital, and dispersion communication.
Fieldbus technology has gone through the initial stages of fragmentation and
separation, and the current standard is IEC 61158. Among them, there are strong
strengths and influences: Foundation Fieldbus (FF), LonWorks, Profibus, HART,
CAN, Dupline, etc. They have their own characteristics and have formed their own
advantages in different application areas. Here, it will briefly describe the character-
istics of fieldbus technology, closely related to the reliability and practicability of
the system, and introduce the key technologies such as fieldbus network structure
and architecture and the current status of several popular fieldbus technologies.
Finally, the development trend and technology prospect of fieldbus are explained.
First, the technical characteristics of the fieldbus are as below:
1. The openness of the system. An open system means that the communication
protocol is open, and devices of different manufacturers can be interconnected
and exchange information. Fieldbus developers are committed to establishing an
open system of a unified factory underlying network. For standards, openness
here refers to the consistency and openness. A fieldbus network open system
gives users the right to integrate the system. Users can make products from dif-
ferent suppliers into random systems according to their own needs and objects.
2. Interoperability and interoperability. Interoperability here refers to the realiza-
tion of information transmission and communication between interconnected
devices and systems. It supports digital communication of point to point and
point to multipoint. Interoperability refers to devices from different manufactur-
ers with similar performance support interchanging for interoperability.
3. Functional autonomy and intelligent field devices. It distributes the functions of
sensing measurement, compensation calculation, and processing and controlling
engineering quantity. The automatic control functions are completed only by the
field device and diagnosing equipment operation status.
3 Presents the Technology, Protocols, and New Innovations in Industrial Internet… 49

4. Highly dispersed system structure. Since the field device itself can complete the
basic functions of automatic control, the fieldbus has formed a new architecture
of a fully distributed control system. It fundamentally changed the system of
distributed control system combining the concentration and dispersion of exist-
ing DCS, improving reliability, and simplifying system structure.
5. Adaptability to the on-site environment. Working at the front end of the field
equipment, as the fieldbus on the bottom of the factory network, designing for
the environment of field. The capability of interference can be realized by two-­
wire system for power transmitting and communicating for satisfying require-
ments of intrinsic safety.
Second, the advantages of fieldbus are below.
Due to the above characteristics of the fieldbus, especially the simplification of
the structure of the fieldbus system, the design, installation, and commissioning of
the control system to the normal production operation and maintenance and repair
thereof all show superiority:
1. Save hardware quantity and investment. Because the smart devices scattered in
the front end of the fieldbus system device process kinds of functions such as
alarm, control, calculation, and sensing. Therefore, transmitter number can be
reduced, eliminating the need for separate controllers, computing units, etc. In
addition, it is required for DCS with the technology of signal conversion, isola-
tion and conditioning, and their complex wiring. Industrial PCs can also be used
as operating stations, with savage investment of hardware. As reducing control
equipment, the control room can be reduced.
2. Save installation costs. The wiring of the fieldbus system is very simple. Because
a pair of twisted pairs or a cable can usually be connected to multiple devices, the
amount of cables, terminals, slots, and bridges is greatly reduced. The wiring
design and the proofreading work of the joints are also decreased very much.
When it is necessary to add on-site control equipment, it is not necessary to add
a new cable, and it can be connected to the original cable nearby, which saves
investment and reduces the workload of design and installation. According to the
calculation data of the typical test project, the installation cost can be saved by
more than 60%.
3. Save maintenance overhead. Because the field control equipment has the ability
of self-diagnosis and simple fault handling and through the digital communica-
tion to send relevant diagnostic maintenance information to the control room, the
user can query the operation of all equipment and diagnose the maintenance
information, so as to analyze the cause of the fault early and quickly eliminate it.
Maintenance downtime is reduced, and maintenance is reduced due to simplified
system structure and simple wiring.
4. The user has a high degree of system integration initiative. Users are free to
choose the devices provided by different vendors to integrate the system.
Avoiding the choice of a certain brand of products is “framed dead” equipment;
it will not be incompatible with the incompatible protocols and interfaces in the
system integration, so that the initiative in the system integration process is com-
pletely in the hands of users.
50 Q. Zhao

5. Improve system accuracy and reliability. Due to the intelligent and digitized
fieldbus devices, it fundamentally improves measurement and control’s accuracy
and reduces transmission errors compared with analog signals. At the same time,
due to system structure which is simplified, it reduces equipment as well as wir-
ing and strengthens internal field instrument functions: the round-trip signal
transmission is reduced, and the operational reliability of the system is improved.
In addition, due to its equipment standardization and functional modularization,
it also has the characteristics such as easy reconstruction as well as simple
design.

AMPS

The AMPS (Advanced Mobile Phone System), as well as NMT and NTT, is one of
typical representatives of mobile communication systems in the United States.
AMPS’s analog cellular transmission works on the 800 MHz band which is used in
America, as well as parts of the Pacific Rim.
From the mid-1970s to the mid-1980s, the first generation was a simulated
mobile cellular communication network. Compared to previous mobile communi-
cation systems, the breakthrough was that Bell Labs proposed cellular networks
concept by in the 1970s. The cellular network, that is, the cell system, greatly
increases the system capacity due to frequency reuse.
Total Access Communication System (TACS), which is divided into the ver-
sion of both ETACS (Europe) and NTACS (Japan), occupies the 900 MHz band.
This standard is widely used in some Asian countries, Japan, and the United
Kingdom.
The first-generation mobile communication system has the main feature which is
the frequency division multiplexing. Moreover, analog modulation is applied for the
voice signal, with 30/25 kHz used for and each analog user channel. Great commer-
cial success has been achieved by the first-generation system; however, it leaves
some drawbacks in the areas of:
1. Spectrum utilization
2. Business types
3. Data service speed
4. Confidentiality
5. Cost of equipment
6. Size and weight
As the analog system has such fundamental technical defects, people further
developed and emerged digital mobile communication technology (i.e., the
second-­generation mobile communication system) from the mid-1980s. First, the
Pan-­European Digital Mobile Telecommunications Network (GSM) system was
first introduced by Europe. Next, Japan and the United States designed their
mobile digital communication systems. Compared with analog mobile communi-
cation, digital mobile communication network improves utilization of spectrum
3 Presents the Technology, Protocols, and New Innovations in Industrial Internet… 51

and supports many services. The second-generation mobile communication

system (i.e., narrowband digital communication system) aims to transmit low-
speed data and voice services. In the second generation of digital cellular mobile
communication systems, typical representatives are the IS-95, US DAMPS, and
the European GSM system:
1. GSM is designed for global digital cellular communications, which is a DMA
standard and originated in Europe. Interconnected with ISDN, it supports
64 kbps data rate. DCS1800 used by GSM represents the 900 MHz band and the
1800 MHz band. FDD duplex and TDMA multiple-access modes are adopted by
GMS, supporting eight channels per carrier frequency with 200 kHz signal band-
width. The is relatively complete, and the technology is relatively mature. The
shortcoming of GSM standard system is analog system capacity which is limited
and cannot be further increased.
2. Digital Advanced Mobile Phone System (DAMPS, known as IS-54) also uses
the 800 MHz band with TDMA multiple access, which is one of the two early
introductions of the digital cellular standard of North America.
3. Another North American digital cellular standard, IS-95, uses the 800/1900 MHz
frequency band with CDMA multiple access.
Since the second generation of mobile communication aims to transmit low-­
speed data and voice services, since 1996, there have been 2.5 generations of mobile
communication systems (i.e., IS-95B and GPRS) for the medium-speed data trans-
mission problem. The main mobile communication services currently are still ser-
vices of low-rate data and voice. Due to network development, the multimedia and
data communication develops very fast. So, broadband multimedia mobile commu-
nication is becoming the goal for the third-generation mobile communication. From
the perspective of development, in the third-generation mobile communication, the
core technology is CDMA due to its own technological advantages. The following
requirements are imposed on 3G wireless transmission technology (RTT, radio
transmission technology):
1. High-speed supporting services of multimedia transmission. The environment of
indoor has above 2 Mbps transmission data rate. The indoor and outdoor walking
environment is at least 384 kbps. The outdoor vehicle movement is at least
144 kbps, and the satellite mobile environment is at least 9.6 kbps.
2. The transmission rate can be allocated as needed.
3. The links of uplink and downlink can adapt to the asymmetric demand.
In 1985, the International Telecommunication Union (ITU) proposed the third-­
generation mobile communication system, which was named FPLMTS (the Future
Public Land Mobile Telecommunication System) and renamed IMT-2000
(International Mobile Telecommunication-2000) in 1996. It is because the system
achieves maximum service rate up to 2000 kbps operating at the 2000 MHz fre-
quency band. It is expected to be commercially available around the year 2000.
The main systems are cdma2000, WCDMA, and TD-SCDMA. In 1999, ITU-R
TG8 18th meeting adopted the “IMT-2000 Radio Interface Technical Specification”
proposal.
52 Q. Zhao

WLAN

It can be easily networked, because WLAN can easily and quickly accept new
employees without having to make too many changes to the user management con-
figuration of the network. WLAN is easier to implement in places where wired
network cabling is difficult. The WLAN solution eliminates the need for perforated
wiring and doesn’t cause any damage to the building.
For every user using a WLAN, WLAN performance is extremely important. But
under some special cases, it is especially important than other features, so it has to
talk about this first. Regarding WLAN performance, we usually refer to the opti-
mal transmit and receive throughput of a WLAN over a wireless signal. The IEEE
standard always provides the theoretical throughput limit for each 802.11 protocol.
For example, 802.11g has a theoretical maximum physical rate of 54 Mbps. In
comparison, the latest second-wave 802.11ac protocol can achieve a maximum
theoretical rate of 2.34 Gbps. But it is worth noting that these bit rates have never
been true.
The distance of the wireless access point, the number of terminals connected to
the AP, wireless signal congestion, and physical obstacles all greatly reduce the
actual throughput.
To minimize performance loss, WLAN vendors use a variety of techniques to
improve wireless chips and antennas. If your organization requires very high per-
formance for a particular application or if your deployment environment has a
significant disruption that can severely degrade the performance of an enterprise
WLAN, then you need to choose a vendor that provides the best WLAN
performance.
In 1990, the IEEE 802.11 WLAN standard working group is established by the
IEEE 802 Standardization Committee. IEEE 802.11 (i.e., Wi-Fi) is an approved
standard in June 1997 that designs the physical layer as well as media access con-
trol layer specifications. The physical layer designs the characteristics of signal as
well as data transmission modulation. Further, it defines two methods of RF trans-
mission as well as one method of infrared transmission. The standard of RF trans-
mission uses direct sequence spread spectrum as well as frequency-hopping
spread spectrum and operates in the frequency band of 2.4000–2.4835 GHz. IEEE
first developed a wireless LAN standard named IEEE 802.11. IEEE 802.11 is
applied solving user wireless access in campus networks as well as in office. The
service rate reaches only 2 Mbps. IEEE 802.11b replaces the IEEE 802.11 stan-
dard because 802.11 cannot have good transmission distance and speed
performance.
In 1999, the officially approved IEEE 802.11b specifies the WLAN with oper-
ating 2.4–2.4835 GHz frequency band with 11 Mbps transmission rate and
50–150 ft transmission distance. This standard can supplement to IEEE 802.11. It
adopts coded modulation mode with two modes of operation: modes of basic and
3 Presents the Technology, Protocols, and New Innovations in Industrial Internet… 53

point to point. For the rate of data transmission, it can be at different rates of 11,
5.5, 2, and 1 Mbps according to actual conditions. Automatic switching changes
the design of WLAN and expands ​WLAN application areas. IEEE 802.11b
adopted by many manufacturers has becoming WLAN standard mainstream. The
widely used products are introduced in airports, hotels, homes, offices, stations,
etc., industry concerned IEEE 802.11a as well as IEEE 802.11g due to many
latest WLAN standards emergency.
It developed the IEEE 802.11a standard in 1999, which specifies the frequency
band of WLAN operating on 5.15–5.825 GHz, with 54/72 Mbps (Turbo) data
transmission rate, and 10–100 m controlled transmission distance. This standard
supplements IEEE 802.11, which expands the standard layer of physical, uses
technology of OFDM’s spread spectrum with QFSK modulation, and provides
10 Mbps wired Ethernet and 25 Mbps wireless ATM. The interface of wireless
frame structure supports many services including data, images, and voice, and
one sector can support many mobile terminals and any terminal. The standard
follow-up IEEE 802.11b is IEEE 802.11a. The original design of IEEE 802.11a
aims at substituting 802.11b. However, operating in 2.4 GHz band requires no
license. This band belongs to the special frequency bands of industry, education,
medical, etc. and is public. It requires a license to operate in the
5.15–8.825 GHz band.
At present, IEEE has introduced the latest version of the IEEE 802.11g authenti-
cation standard, which proposes a transmission rate of IEEE 802.11a, which is bet-
ter than IEEE 802.11b. It adopts two modulation modes, including OFDM and
IEEE 802.11b used in 802.11a. CCK is compatible with both 802.11a and 802.11b.
It is more suitable to deploy 802.11a for WLAN operators; WLAN enterprises seem
very likely choosing 802.11g in order to balance the investment in existing 802.11b
devices.
The IEEE 802.11i standard has authentication of device as well as user port in
IEEE 802.1x to modify and integrate the WLAN MAC layer and defines a mecha-
nism of authentication with encryption format for improving security of WLAN. The
IEEE 802.11i new revised standard mainly includes “technology of Wi-Fi Protected
Access (WPA)” as well as “Strong Security Network (SSN).” In 2004, the Wi-Fi
Alliance adopts the 802.11i standard as the second version of WPA and began
implementation. It is important to construct WLAN with the IEEE 802.11i standard.
The data security is the first priority that WLAN equipment manufacturers and
WLAN operators should consider first.
The IEEE 802.11e standard proposes improvements to the MAC layer protocol
of WLAN supporting multimedia data transmission to support the quality of service
(QOS) guarantee mechanism for all WLAN radio broadcast interfaces. IEEE
802.11f, defining access nodes’ communication, supports IEEE 802.11 Access
Point Interoperability Protocol (IAPP). IEEE 802.11h is for managing spectrum of
802.11a.
54 Q. Zhao

IEEE 802.15.4

Specifications, such as Wi-Fi, WirelessHART, and ZigBee, have the basis of IEEE
802.15.4. It defines low-rate wireless personal area networks’ layer of physical and
protocol of media access control. The rate of data transmission will be up to
250 kbps in the ISM band of both 868/915 MHz and 2.4 GHz.

WPAN

With the communication technology developing rapidly, there exists the require-
ment for communicating in meters around person’s itself. Therefore, the concept
of wireless personal area network (WPAN) as well as personal area network (PAN)
is proposed. The WPAN can help wireless devices within a close geographical area
to set up wireless connections to communication and connect each other. In 2002,
it established the IEEE 802.15 working group, which concentrates on the WPAN
standardization in the area of the physical layer (PHY) as well as media access
layer (MAC). Wireless devices provide communication standards.
In IEEE 802.15 working group, different standards of applications are developed
by four different task groups (TGs). These standards of applications are different
from the following aspects, such as consuming power, services provision, and trans-
mitting rate. The primary targets of above task groups are as below:
1. Task group TG1 focuses on developing the standard for Bluetooth wireless
personal area network, numbered IEEE 802.15.1. This standard concentrates
on short-range communication and proposes a close-range, medium-speed
WPAN.
2. Task group TG2 focuses on developing coexistence solution between IEEE
802.11 and IEEE 802.15.1, numbered IEEE 802.15.2.
3. Task group TG3 focuses on developing multimedia applications in the wireless
personal area network for achieving high data rate as well as quality of service,
numbered IEEE 802.15.3.
4. Task group TG4 focuses on developing low-cost, low-rate transmission and uni-
form standard for low-speed interconnection for low-rate wireless personal area
network (LR-WPAN), numbered IEEE 802.15.4.
The LR-WPAN is a wireless communication network with low-cost, enabling
applications under wireless connections with low energy and low cost. In compari-
son with WLAN, LR-WPAN require no infrastructures. For the LR-WPAN, the
PHY layer as well as MAC sublayer protocols are defined in IEEE 802.15.4.
In IEEE 802.15.4 standard, the LR-WPAN should be of the characteristics
as below:
1. Different carrier frequencies give diverse transmission rates of 20, 40, and
250 kbps.
3 Presents the Technology, Protocols, and New Innovations in Industrial Internet… 55

2. Support star as well as peer-to-peer network topologies.

3. There are two address formats of 16-bit as well as 64-bit.
4. CSMA-CA (carrier sense multiple access with collision avoidance).
5. Support mechanism of acknowledgment (ACK) ensuring the reliability of
transmission.

Conclusion

Along the life cycle, product with embedded smart electronics can definitely form
an internal even global information network, making IoT with industrialization.
Such trend has made the vision of Industry 4.0 closer to the reality, i.e., a fully
context-sensitive, controllable, describable, manageable, self-regulating manufac-
turing system.
There can be drawn a conclusion that IIoT has mandatory requirement for
Industry 4.0, which is because IIoT is an emerging approach concentrating on the
area of “Computer Networks and Communications” for delivering “a Comprehensive
Broadband Infrastructure.”
Giving the reader a systematical view of IIoT including the innovation and appli-
cation, the technology, the protocol, and standards, this chapter shows that smart
devices are changing in the world, in which significant trend has already been
extended from people’s daily life to the world industry.
In the upcoming Industry 4.0, the connected smart devices all around the world
via the Internet will provide secure, real-time, and reliable services of sensing, com-
municating, and computing, making effects on the following aspects, such as chips,
terminals, base stations, networks, software tools, testing devices, operating sys-
tems, and APPs.

References

1. Alcaraz, C. (2019). Security and privacy trends in the industrial internet of things. Cham:
Springer.
2. Wan, J., Lai, C. F., Song, H., et al. (2019). Software-defined industrial internet of things.
Wireless Communications and Mobile Computing, 2019, 1–2.
3. Aazam, M., Harras, K. A., & Zeadally, S. (2019). Fog computing for 5G tactile industrial
internet of things: QoE-aware resource allocation model. IEEE Transactions on Industrial
Informatics, 15(5), 3085–3092.
4. Matthyssens, P. (2019). Reconceptualizing value innovation for Industry 4.0 and the indus-
trial internet of things. Journal of Business & Industrial Marketing. https://doi.org/10.1108/
jbim-11-2018-0348.
5. Vitturi, S., Sauter, T., & Pang, Z. (2019). Real-time networks and protocols for factory automa-
tion and process control systems. Proceedings of the IEEE, 107(6), 939–943.
6. Jaloudi, S. (2019). Communication protocols of an industrial internet of things environment: A
comparative study. Future Internet, 11(3), 66.
56 Q. Zhao

7. Salonikias, S., Gouglidis, A., Mavridis, I., et al. (2019). Access control in the industrial inter-
net of things. In Security and privacy trends in the industrial internet of things (pp. 95–114).
Cham: Springer.
8. Ludwig, S., Karrenbauer, M., Fellan, A., et al. (2018). A5G architecture for the factory of the
future. In 2018 IEEE 23rd International Conference on Emerging Technologies and Factory
Automation (ETFA) (Vol. 1). IEEE, pp. 1409–1416.
9. Liao, Y., Loures, E. F. R., & Deschamps, F. (2018). Industrial internet of things: A systematic
literature review and insights. IEEE Internet of Things Journal, 5(6), 4515–4525.
10. Unverdorben, S., Böhm, B., & Lüder, A. (2018). Reference architectures for future production
systems in the field of discrete manufacturing. In 2018 IEEE 14th International Conference on
Automation Science and Engineering (CASE). IEEE, pp. 869–874.
11. Zhang, Y., & Chang, J. (2016). Application of fieldbus control system on AC/DC microgrid.
In Advances in energy, environment and materials science: Proceedings of the International
Conference on Energy, Environment and Materials Science (EEMS 2015), Guanghzou, PR
China, August 25–26, 2015 (p. 15). Boca Raton: CRC Press.
12. Huang, B. W., & Li, J. Y. (2014). Design of digital teaching platform of the internet of
things. In Applied mechanics and materials (Vol. 670, pp. 1671–1674). Zug: Trans Tech
Publications.
13. Liu, P., Liu, X. P., & Wang, B. P. (2012). Application of the measurement and control network
on rewinder. In Advanced materials research (Vol. 463, pp. 1509–1512). Stafa-Zurich: Trans
Tech Publications.
14. Liang, Q. (2012). Communications, Signal Processing, and Systems. Proceedings of the
International Conference on Communications, Signal and Processing, Springer.
15. Ho, Q. D., Gao, Y., Rajalingham, G., et al. (2014). Wireless communications networks for the
smart grid. Berlin: Springer.
16. Li, Z. D., Feng, G. H., Feng, X., et al. (2015). Design of gas density monitor system based on
wireless sensor network. In Applied mechanics and materials (Vol. 716, pp. 888–891). Zug:
Trans Tech Publications.
17. Wu, X., & Tang, D. (2009). Hierarchical software architecture of real-time monitoring and
management system for area water resources based on OPC. In 2009 International Conference
on Information Technology and Computer Science (Vol. 2). IEEE, pp. 287–291.
18. Grangel-González, I., Halilaj, L., & Auer, S., et al. (2016). An RDF-based approach for imple-
menting Industry 4.0 components with administration shells. In 2016 IEEE 21st International
Conference on Emerging Technologies and Factory Automation (ETFA). IEEE, pp. 1–8.
19. Hankel, M., & Rexroth, B. (2015). Industrie 4.0: The reference architectural model Industrie
4.0 (RAMI 4.0). ZVEI: Die Elektroindustrie.
20. Kagermann, H., Helbig, J., & Hellinger, A., et al. (2013). Recommendations for implementing
the strategic initiative INDUSTRIE 4.0: Securing the future of German manufacturing indus-
try; final report of the Industrie 4.0 Working Group. Forschungsunion.
Chapter 4
Decision Making Models Through
AI for Internet of Things

E. P. Ephzibah, S. Sree Dharinya, and L. Remya

Biologically inspired computation had taken its first step some decades ago. Still
there is a need for optimization of many problems in real-time scenarios. When
found in nature, there are algorithms available that have more possibilities of opti-
mization and can help humans to solve naturally hard problems using limited time
and space. Some of those techniques are discussed in this chapter as they provide
valuable and time-conserving methods that are observed from the nature. This par-
ticular area of study binds the subareas like connectivity, communication, and social
performance with emergence and existence. This field can be well flourished with
the help of artificial intelligence (AI) and machine learning as it involves training
combined with decision-making for optimization. This field of study requires
knowledge in biology as it monitors the behavioral aspects of the insects, animals,
and birds. Also a mathematical background is mandatory as it deals with mathemati-
cal axioms and proofs for the formation of the solution. The problems that are found
in the nature require solutions that need lot of computation time and space. The
natural environments containing the organisms like algae, ant, bee, fish, cat, wolf,
and elephant are already equipped with sensational information that helps them to
solve the heterogeneous problems around them. The activities of these organisms
when imitated by computers can be treated to be one of the major subjects in this
field. Biologically inspired computing is a major subset of natural computation.
Artificial bee colony algorithm is a combination of stochastic genetic/evolu-
tionary algorithm and bee colony optimization technique that helps in solving
multi-­objective optimization problems. Another type of algorithm is called

E. P. Ephzibah · S. Sree Dharinya (*)

School of Information Technology and Engineering, Vellore Institute of Technology,
Vellore, Tamilnadu, India
e-mail: [email protected]
L. Remya
School of Mechanical Engineering, VIT, Vellore, India

© Springer Nature Switzerland AG 2020 57

G. R. Kanagachidambaresan et al. (eds.), Internet of Things for Industry 4.0,
EAI/Springer Innovations in Communication and Computing,
https://doi.org/10.1007/978-3-030-32530-5_4
58 E. P. Ephzibah et al.

swarm intelligence that takes into account the artificial intelligence and the
behavioral aspects of cat, whale, chicken, and other animals for solving the opti-
mization problems [1].

Why Do Birds Fly, Bees Swarm, and Fish School?

They do this because the groups are smarter when thinking together than the indi-
vidual would be on their own. Scientists call this swarm intelligence and this com-
bines the power of many minds to one. Researchers at unanimous AI have developed
an innovative technology to allow an online grid to form artificial swarm intelli-
gence. They quickly and actively enter questions, reach decisions, make predic-
tions, generate forecasts, and optimize opinions. A combination of AI algorithms in
real-time human input has the knowledge, wisdom, insight, and intuitions and
amplifies the collective intelligence empowering them to be smarter together.
A very smart intelligent optimization technique that is new and effective in com-
parison to the existing swarm intelligent algorithms is called grey wolf optimization
algorithm. It is also a swarm intelligent-based algorithm for optimization that mim-
ics the leadership quality of a wolf for group hunting. It is a hierarchy-based tech-
nique that is followed strictly to obtain or reach the global maxima [2]. The elephant
search algorithm is an algorithm that was built with the idea of searching an ele-
phant, for example, in the case of selecting best gene expression from the microar-
ray data available in large volume [3]. A stochastic feature selection method for
reduction of features is also possible with the help of firefly search algorithm. Thus
exploring the different AI-based optimization techniques that are biologically
inspired for solving problems using machine learning is the objective of the chapter.
To get a deeper insight on the study area, it is mandatory to understand the fun-
damentals of biological computing or bio-inspired computing.

Neural Networks

Neural networks are data processing algorithms which are adaptive and nonlinear in
nature [4]. The network combines the different processing units connected in a sys-
tem with diverse layers. The characteristics of the networks include its self-­
adaptability, self-organizing, and capacity to learn from the inputs or response. The
neural networks work like neural system in human beings or any higher-level organ-
isms as the way they take input from surroundings. Like neurons in the organisms,
neural network replicates to provide output by processing the input. The output
from the network is again given back to the network for improvement in the current
system if required.
4 Decision Making Models Through AI for Internet of Things 59

Though there are many approaches to implement neural networks, a widely used
and simple one is perception. The perception networks were developed around the
1960s by Frank Rosenblatt. Perception takes several binary inputs and produces a
single binary output. The number of input varies from one to any numbers.
For example, a simple perception network will be as shown below:

x1

x2 output

x3

The output of the perception network can be calculated as

0 if ∑ w x ≤ threshold
 j j j
Output = 
1 if ∑ j w j x j > threshold

where w1, w2, … are the weights which are real numbers which express importance
of respective input to the rate of output. The output label is evaluated based on the
value of the ∑j(wjxj). If the value is less than or equal to threshold, the output will
be 0, and if it is greater than threshold, output will be 1. Threshold is also a real
number which is a parameter of neuron.
We can illustrate it with a real-time application. If the output is buying behavior
of a particular product among customers, we can consider input variables as:
• Brand loyalty to the particular product
• Promotion policies that product has currently
• Customer need (whether the customer required the product or not)
Suppose the customer is highly loyal to the brand, x1 = 1. Also there are abundant
promotion policies such as free coupons, get one buy one, etc., the second variable
x2 = 1. The third is if the customer is in need to buy the product, the third variable
x3 = 1. And when it comes to the weight, if we choose weights as w1 = 2, w2 = 3, and
w3 = 5. The larger-value w3 indicates that customer need is important rather than
brand loyalty and promotion policies. Finally, suppose we choose a threshold of 5
for the perception neutron. With these choices, the perception will be providing the
desired output on the buying behavior of the individual, whether he will buy or not.
The characteristic of perception network is that it can be considered for both
linear and nonlinear systems. It can also deal with multiple inputs and multiple
outputs. It also includes multiple layers.
60 E. P. Ephzibah et al.

Artificial Life

Artificial life or Alife means life made by human being not by the Mother Nature. It
consists of the study of the systems similar to the natural system, evolution of such
akin systems with the help of computer simulation or algorithms, robotics, or bio-
technology [5]. Artificial life acts as a powerful tool for understanding how the
ecological communities are evolved [6].
Artificial life researches are mainly divided into three based on its domain:
• Soft ALife
• Hard Alife
• Wet ALife

Soft Alife

It refers to the computer or mathematical simulation. It is a self-organized system

and linked to many subdomains. One of the examples for this type is Reynolds’
boids model [7]. In it, self-propelled agents move according to kinetic rules.

Hard Alife

It refers to physical robots. Robots are mechanical or artificial agents, guided by the
computer program or circuits. Such system has the advantage of the physical reality.
Physical experiments are conducted nowadays in many areas including behavior
observed in living system, collective navigation, self-assembly, etc.

Wet Alife

It refers to chemical or biological research. It consists of the synthesis of life like

behavior using the principle of self-organization. A recent example is liquid robot,
i.e., dynamic behaviors of microscopically visible chemical droplets.

Cellular Automata

Cellular automata is a field of study that was developed by von Neumann and Ulam
in 1960. They are discrete models for complex natural systems that exist in all the
areas of life. It is the mathematical representation of physical systems in which
4 Decision Making Models Through AI for Internet of Things 61

space and time are distinct. Also in type of systems, physical measures take on finite
set of discrete values [8]. Cellular automata consist of arrays and cells. An array is
a regular uniform lattice infinite in nature, whereas cell is the discrete variable at
each site. These are also known as cellular spaces, cellular structures, etc.

Learning Classifier Systems

These are the framework that uses genetic algorithms to study learning in rule-based
systems. These are multifaceted learning algorithm. It has evolved as the pioneer in
evolutionary biology and artificial intelligence [9]. LCS receive feedback from the
environment with which it interacts and learns from that (Fig. 4.1).
Learning classifier systems mainly consist of four components:
• Classifier
• Performance component
• Reinforcement component
• Discovery component

Classifier

These are the finite population of condition-action rules. It is the repository of the
knowledge of the system.

Biology Computer science

Artificial intelligence

Evolutionary
biology Machine learning

Evolutionary Reinforcement Supervised

computation learning learning

Evolutionary
algorithm

Genetic algorithm Learning classifier system

Fig. 4.1 Foundations of LCS community. (Source: Urbanowicz and Moore [10])
62 E. P. Ephzibah et al.

Performance Component

It directs the communication with the system and the environment.

Reinforcement Component

It is also known as credit assignment component. It transfers the credit or reward

received from the environment to the respective classifier.

Discovery Component

It finds out better rules, and it tries to improve the existing system through genetic
algorithm (Fig. 4.2).

Discovery
Component

Performance
Component

Match Set match Population

classifiers state s
classifier1
Environment

that apply in ...

the current classifiern
state s.

action a
Evaluation of Action
actions utility Selection

reward r
Reinforcement
Component

Learning Classifier System

Fig. 4.2 LCS and environment. (Source: Holmes et al. [11])

4 Decision Making Models Through AI for Internet of Things 63

The main factors that lead to the popularity and acceptability of LCS are its
adaptability, generalization, and scalability. While considering adaptability, ­learning
classifier systems are adaptive systems which can handle online learning even in
drastically changing environment. It is highly useful in both predicting and describ-
ing the drastic changes such as epidemics, etc.
Representation of what it learned and application of learning even if it is unseen
situations are the feature of a properly generalized system. In learning classifier
system, the evolution of general rules gives way to generalize things especially situ-
ations like environmental issues. Scalability is the rapid growth of the system even
though it becomes more complex in nature. The methodologies in LCS such as train
autonomous robots, theoretical approach for learning complexity, etc. help to
increase the scalability of the system.
Models of LCS include Wilson’s XCS [12], Stolzmann’s Anticipatory Classifier
System (ACS) [11], Holmes’ EpiCS [13], etc. Its applications are mainly in the
areas of autonomous robotics, knowledge discovery, and computational economics.

Membrane Computing

Membrane computing is a group of computation models which are derived from

taking inputs and information from the natural aspects such as living cell function-
ing, cell organization in tissues, organ structures, etc. [14]. Membrane computing
was introduced by Păun G in 1998. Several classes of models were defined and were
called as P systems. The components of a P system are membrane structure and
delimiting compartments. The main issues a P system may face are computing
power and computational efficiency.
There are three types of P systems:
• Cell-like P systems
• Tissue-like P systems
• Neural-like P systems

Cell-Like P Systems

It represents the eukaryotic cell. It includes membrane structure and delimiting

compartments. The objects are represented by symbols from a given alphabet. The
most common type rules are multiset rewriting rules.
64 E. P. Ephzibah et al.

Tissue-Like P Systems

Here a common environment consists of many one-celled membranes. It contains

huge number of objects, even the environment also. The environment acts as a
medium for communication of all cells. Even among them too, few can communi-
cate each other due to the presence of channel. If the cells consist of less rules, it’s
called P colony.

Neural-Like P Systems

It is of two types. One is akin to tissue-like P systems and cells placed in nodes and
contains lot of objects with a state controls evolution. The other is spiking neural P
systems which use only one type of object, and information works with distance
between the spikes.
The P systems are easily understandable. Its scalability and programmability are
welcoming features. Its applications not only restricted to biological process, but to
computer graphics, cryptography, optimization, etc.

Sensor Networks

Sensor networks are a group of sensor nodes which are widely dispersed either
within the phenomenon or very close to it. The position of such need not be pro-
grammed before [15]. Such arrangement helps for deployment in inaccessible areas
such as terrains, disaster-prone areas, etc. Self-organizing capability is the main
feature to be possessed by the protocol or algorithm. It also has cooperative efforts
by the sensor nodes mechanism.
The sensor nodes are scattered in a sensor field. These nodes can collect data and
route data back (Fig. 4.3).
The factors influencing the design of the sensor networks are fault tolerance,
scalability, production costs, operating environment, sensor network topology, and
hardware constraints.

Fault Tolerance

Fault tolerance is the capacity of the sensor network to sustain without any failures
in sensor nodes. The reliability Rk(t) or fault tolerance within the time interval
(0, x) is
4 Decision Making Models Through AI for Internet of Things 65

Internet and
D C B
satellite Sink
E
A

Task manager
node

User

Sensor field Sensor nodes

Fig. 4.3 Sensor network. (Source: Akyildiz et al. [15])

Rk ( x ) = e – λ kx,

where λk is the failure rate of sensor node k and x is the time period.

Scalability

The sensor nodes are huge in number used for deployment. The new schemes must
ensure that they are able to work with these high numbers of nodes and can utilize
the density properly. The density μ can be calculated as

µ ( R ) = ( N ·π R2 ) / A,

where N is the number of scattered sensor nodes in region A and R is the radio trans-
mission range. Basically, μ(R) gives the number of nodes within the transmission
radius of each node in region A.

Production Costs

Since the sensor nodes deal with large number of sensor nodes, cost of each sensor
will be an important factor which justifies the overall price of the product. If the
overall cost of sensor networks is higher than the traditional network, it won’t be
cost justified. And hence the cost of each sensor node needs to be kept low.
66 E. P. Ephzibah et al.

Operating Environment

They can operate in all the inaccessible areas such as high terrains, bottom of the
ocean, contaminated fields due to physical and biological intervention, interior of
large machine, and also battlefield.

Sensor Network Topology

Thousands of nodes are deployed throughout the sensor field. They are densely
deployed and with high density. Such deployment needs careful handling of topol-
ogy maintenance. Issues are checked and corrected during three phases such as
redeployment and deployment phase, post-deployment phase, and redeployment of
additional node phase.

Hardware Constraints

Sensor node has four components – a sensing unit, a processing unit, a transceiver
unit, and a power unit. It has additional application-dependent components such as
location finding system, power generator, and mobilizer. Sensor unit consists of sen-
sors and analog to digital convertors (Fig. 4.4).

Location finding system Mobilizer

Processing
Sensing unit unit
Processor
Sensor ADC Transceiver
Storage

Power
Power unit generator

Fig. 4.4 Sensor node components. (Source: Akyildiz et al. [15])

4 Decision Making Models Through AI for Internet of Things 67

All these subunits need to fit in very small-sized module smaller than even cubic
centimeters. There are constraints apart from size. The nodes must consume
extremely low power, operate in very high densities and low production cost, be
autonomous, operate in inaccessible areas, and be adaptive to environment.
Sensor network has many applications. They are part of military command, con-
trol, communications, computing, intelligence, surveillance, reconnaissance, and
targeting (C4ISRT) systems. The rapid deployment, self-organization, and fault tol-
erance characteristics of sensor networks make them a very promising sensing tech-
nique for military C4ISRT. It has environmental applications such as forest fire
detection, bio-complexity mapping of the environment, flood detection, and agricul-
tural application. While coming to health sector, it can be used for telemonitoring of
human physiological data, monitoring of people inside hospital, and drug adminis-
tration in hospitals. In home applications, its service extends to smart home and
home automation. Other applications include vehicle theft detection, vehicle track-
ing and detection and interactive museums, etc.

Biodegradability Prediction

Biodegradability prediction tries to predict biodegradability of anthropogenic mate-

rials in the environment. As it became very essential to predict the impact of the
chemicals in the environment, different computational models are taken into con-
sideration for the same [16]. Modelling techniques include linear and nonlinear
regression, expert system, neural networks, and artificial intelligence.

Regression Models

It consists of liner, nonlinear, and multilinear correlation of biodegradation rates

with parameters such as molecular weight, structure, solubility, etc.

Expert System or Survey Models

This model stakes expert opinion in the respective area by conducting surveys, and
based on that models predict degradation.
68 E. P. Ephzibah et al.

Machine-Based Learning

Including neural networks, these models process the chemical structure of the
products based on it.
Biologically inspired computing which is commonly referred to as bio-inspired
computing is a close relation to the field of artificial intelligence which has a link
with evolutionary algorithms. Evolutionary algorithms form the basis for a very
impressive and beneficial evolutionary theory with a badge called survival of the
fittest. That means answers to a composite problem can be obtained by using the
computer as a means to get inexpensive ways to deal with complex problems.
Particularly they are used for search and optimization problems.
Swarm intelligence is a special phenomenon that groups all the different types of
algorithms that take into consideration the term called communication as its key
mechanism. The living creatures together put their efforts and cooperate and suc-
cessfully solve problems because independently they will find it tough to elucidate.
But if these organisms join their efforts, they can solve these types of problems.
Swarm intelligence algorithms are very good and efficient in solving optimization
problems. Moreover, they are useful in classification, clustering, robotics, and other
types of biologically inspired computing applications.
Some examples of swarm intelligence algorithms are particle swarm optimiza-
tion which is inspiring some behavior of bird flocks, ant system, the bacterial forag-
ing optimization algorithm which is based on the foraging behavior of the E. coli
bacterium, and bee colony optimization.
Dynamic constrained optimization problems are interesting to solve as the solu-
tion to be obtained is dynamic or movable in nature. So the algorithm needs ele-
ments to track the moving optimum. The problems become quiet complex in some
steps of the search. Robotics is a field started in industrial settings where very pre-
cise and fast motion or dynamic characteristic of the machines is needed. Those
technologies are not suited for interacting with people.
Bio-robotics is a field of artificial intelligence where thinking is involved in the
combination of both body and the brain and the group. There exist a lot of examples
in biology where we can see how that can be achieved and how we can leverage the
same kind of ideas. It is a new discipline connecting living creatures and robots. By
looking at the nature, how nature does things. Looking at on the design principles of
nature and trying to build robots that are either implemented through those princi-
ples or act like those principles. Looking at the materials that exist in nature and the
algorithm that nature uses, the behavior and the activities of the animals can be used
to learn from them and use it applying those techniques to robots.
Almost all the locomotive robots are inspired by these soft animals that live.
Compared to more traditional robots, something that are used for welding the doors
of cars, robots are made to enhance safer technology. Because they are soft and
small, because of their size, and because of their compliance body to adapt to natu-
ral environments, they can perform tasks that are difficult and impossible for other
classes of robots that are conventional in nature. Many of the people in this field are
4 Decision Making Models Through AI for Internet of Things 69

inspired by the incredible ability of the biological systems to cooperate. From the
nature with swarm of insects like termites that build mount or ant colony that forge
over large regions, their quick and wise decision-making skills inspire to build the
robot systems that could do all tasks efficiently. Could I make robots that build, and
could I make robots that could go through coral reefs and monitor the house of a
system? Consider the traffic on the highways and roads, and self-driving cars, a lane
full of self-driving cars, there are a lot of robots on the road.
An algorithmic understanding can be built and large number of skilled laborers
can work together in a decentralized way. And they can be slightly faulty, and they
can have low capabilities as opposed to perfect robots that navigate through com-
plex environment. That’s really amazing manufacturing technology because mass
manufacture of simple things is possible and these simple things will cooperate and
create something of much higher complexity. What we are aiming at in robotics is
much really where humans and robots can coexist, can coordinate, and can distrib-
ute tasks to the joint benefit of all. As you know there are many big differences
between driving mechanism of current robots and human body. Work on building
soft robots is being carried out to enhance and augment healthy human individuals.
These robots will give people with disabilities the super human strength. There are
a lot of people with some kind of disability that need human strength but just need
a small system. New robotic components are manufactured that are using soft mate-
rial able to apply to person with mask and associated with digital devices.
Many animals have the ability to move to newer and dynamic spaces that are
unpredictable in ways that engineered systems cannot achieve even with human
operator. But if we did have these systems that can achieve these capabilities, we
will have much better ability to address needs such as rescue, emergency response,
and collecting information from environmental monitoring. Using fundamental bio-
logical principles, some of these challenges can be addressed in order to improve
engineering design and operation.
The fish robots are one of serious applications. Consider the task of environmen-
tal monitoring on a global scale to collect information to make better models of
what’s going on in the ocean. Given information on temperature, humidity, chemi-
cal content in the ocean, and species growth and decay and to come up with better
ways of knowing what’s going to happen in terms of seismic activity, species track-
ing, weather currents, and transportation are safe. The ways of data being collected
urgently involve sending the research crews with number of people, and they go to
specific locations. Collecting the data using autonomous vehicles is the need of the
hour. Fish robots were developed to operate in dynamic spaces, not very well known
and changing very quickly to be sent for a period of months. Migrating to the remote
location to fix parts or physically collect the data that come out with them has
become troublesome. This can be done with energy-efficient sensors that have lot of
power to run, with high-power camera to show what is going on in a large space.
The engineering characteristic to achieve with similar capabilities in biological sys-
tems could be a fairly large characterization here. The robots should be fast and
agile. One of the fastest swimmers in nature is tuna fish. The key thing about them
is that they have a fairly narrow body, the first part is fairly long and rigid, and then
70 E. P. Ephzibah et al.

there is a fairly rigid tail fin at the back. So a model of a fish with a rigid body in
water is taken as the model, and the characteristics are noted for lift and drag opera-
tions and activities. This can give the information about how to make the vehicles
go quickly in a large remote and unknown space, whether it is under the ocean,
inside a dense forest, or in the space.
Role of ant colony optimization technique in decision-making for robots:
Decision-making using ant colony optimization is an emerging trend in the field of
robotics that tries to analyze and enhance the path distance between any two posi-
tions. The robots mimic the behavior of ants for motion. The ants move in search of
their food from their resting place to the spot where the food source is available.
Randomly, ants take the first move. When they walk on the floor, they leave phero-
mone trail, a chemical substance that is capable of getting evaporated in due course
of time. Subsequently, when other ants get into that space/location, they find these
pheromone trails, and based on the intensity of these trails, they try to learn the
distance to the food particle. The chemical substance reduces the intensity when
there is less number of ants or no ants in the path. The intensity is more and vigorous
if the path is less and there are more ants going and coming back from the source to
the destination in that path. Obviously the ants try to follow the way where the
intensity of the trail is more, thereby educating the ants in this phase and helping
them to take decision. This gives the major idea or information to the robots in the
case of robotic movements as well. Robots, in which many in number work in the
same workspace, in parallelism and simultaneous manner, are made to perform their
tasks without colliding with each other. Robots in the workspace are expected to
move communicating each other like ants in an efficient and effective manner.
Communication among robots plays a very important role in building the system.
They learn from each other to achieve the target. When the robots get the informa-
tion about the path travelled by the other robot, they receive information in terms of
time, energy, etc. Hence decision-making becomes effective and the shortest route
is also identified. The data received influences the other robots to follow the path
travelled adequately by their predecessors who have created a pattern-like structure
for efficient building of the system. The ant colony optimization (ACO) method
firstly has the arbitrary path in the starting of the method, and they get converged
into an operative one with minimum energy so the robots will follow it most effec-
tively using ACO, an algorithm that takes advantage of the functioning of the
ant colony.
Genetic bee colony optimization for charging unmanned cars:
Bees are very smart and probably the smartest insects, and they are the source of
solution for lot of real-time problems. Bees are extraordinary explorers and naviga-
tors. They orient themselves in a radius of 6 kilometers around their house. They
remember numerous resource locations like food, water supply, and their hives, and
they commute between those locations in astounding accuracy and that with a brain
that is of a very small size of a peanut. But the individual feet of the bees are the
only one of the many facets of bee intelligence. It is together how they solve com-
plex tasks; they act collectively as a network of tiny brains looped together to vari-
4 Decision Making Models Through AI for Internet of Things 71

ous forms of communication. They share information and also energy. The process
of sharing turns one centralized static energy store into many thousand small mobile
energy store. Electric cars need careful planning and strict discipline. Substantial
infrastructure changes have to be made for charging them to meet the energy hunger
for tomorrow’s cars.
How are we going to make our electric cars convenient, cost-efficient, and
environmentally less harming?
The obvious answer is that we charge them like bees. We need three components.
Firstly, cars need to ask another car for energy, secondly cars need to be charged,
and thirdly self-driving cars have all the sophisticated sensors in them with comput-
ing equipment. There will not be a necessity to stop the car to charge between them.
We can charge our neighbor’s car while driving, and this is how it’s going to work
in the near future. Cars can communicate between each other and do a quick and
safe energy transfer between moving cars. There could be many potential donors
thousands of charging possible along the road. Cars need not be charged to full, or
even some charge can be added or received as much as consumed. As long as the
target is reached and as long as the other one arrives at his or her target location,
everyone would be happy to help because the swarm has become energy safety net.
The swarm is like a virtual power cable extending from the wall with the plug
through many cars to the ones that actually need the charge. Cars are sort of mobile
power outlet, and if traffic jams are common, they also can now be reduced to some
extent. This could even be the business concept. The car manufacturers are informed
to put this technology into the cars. Lot of money can also be generated by this kind
of technology and people can be helped. By the year 2022, 50% of the car equipped
with this technology is expected, and if just charge 10% of the energy that is being
transferred as a feed for this service, and an average number of ten energy snacks
per month would yield the market of seven million euros in Germany alone, and
now think of Europe, china, or the USA. This would help the customers who would
like to operate their cars all the time without any stops. Those people that charge
their cars regularly provide service to one another. Think of self-driving, taxi com-
panies, or delivery companies, they would be able to operate their cars throughout
the day and night; this technology would definitely be more beneficial.

References

1. Darwish, A. (2018). Bio-inspired computing: Algorithms review, deep analysis, and the scope
of applications. Future Computing and Informatics Journal, 3(2), 231–246.
2. Sharma, P., ShirshSundaram, M. S., Sharma, A., & Gupta, D. (2019). Diagnosis of Parkinson’s
disease using modified grey wolf optimization. Cognitive Systems Research, 54, 100–115.
3. Panda, M. (2017). Elephant search optimization combined with deep neural network for micro-
array data analysis. Journal of King Saud University - Computer and Information Sciences, In
press.
72 E. P. Ephzibah et al.

4. Graves, A., Wayne, G., Reynolds, M., Harley, T., Danihelka, I., Grabska-Barwińska, A., et al.
(2016). Hybrid computing using a neural network with dynamic external memory. Nature,
538(7626), 471.
5. Langton, C. G. (Ed.). (1997). Artificial life: An overview. Mit Press United States of America.
6. Maes, P. (1995). Artificial life meets entertainment: Lifelike autonomous agents.
Communications of the ACM, 38(11), 108–114.
7. Reynolds, C. W. (1987). Flocks, herds and schools: A distributed behavioral model. ACM
Siggraph Computer Graphics, 21(4), 25–34.
8. Codd, E. F. (2014). Cellular automata. Academic Press Academic Press, Inc. Orlando, FL,
USA.
9. Lanzi, P. L. (2000). Learning classifier systems: from foundations to applications (No. 1813).
Springer Science & Business Media, Berlin Heidelberg.
10. Urbanowicz, R. J., & Moore, J. H. (2000). Learning classifier systems: A complete introduc-
tion, review, and roadmap. Journal of Artificial Evolution and Applications, 2009, 209–238,
Berlin Heidelberg.
11. Holmes, J. H., Lanzi, P. L., Stolzmann, W., & Wilson, S. W. (2002). Learning classifier sys-
tems: New models, successful applications. Information Processing Letters, 82(1), 23–30.
12. Wilson, S. W. (1987). Classifier systems and the animal problem. Machine Learning, 2(3),
199–228.
13. Holmes, J. H. (1996). Evolution-assisted discovery of sentinel features in epidemiologic sur-
veillance. Drexel University.
14. Păun, G., & Rozenberg, G. (2002). A guide to membrane computing. Theoretical Computer
Science, 287(1), 73–100.
15. Akyildiz, I. F., Su, W., Sankarasubramaniam, Y., & Cayirci, E. (2002). A survey on sensor
networks. IEEE Communications Magazine, 40(8), 102–114.
16. Peijnenburg, W. J., & Damborský, J. (Eds.). (1996). Biodegradability prediction (Vol. 23).
Kluwer Academic Publishers, Petra Sprado.
Chapter 5
Internet of Things and Additive
Manufacturing: Toward Intelligent
Production Systems in Industry 4.0

A. Suresh, R. Udendhran, and G. Yamini

Introduction

The Internet of things, Industry 4.0, and smart sensors can be combined to make a
heady cocktail that suggests an exciting future with effective and improved perfor-
mance. Security is very essential in every field form in industrial sector. Through an
intelligent system, we could able to detect major challenges existing in Industry 4.0
and address it.
A sensor is an electronic device that senses the physical quantity and converts it
into electrical quantity. They can be used to sense pressure, temperature, humidity,
etc. For the purpose of interpretation and processing, the sensor data could be taken
and then displayed in the computer for the human reference purposes. Nowadays,
sensors are widely used for the protection of facility control and industries as shown
in Fig. 5.1.
Traditional manufacturing process works on the principle in producing certain
number of products by holding the reserve in case of unexpected damage or short-
ages at certain set of periods [1]. There are four main categories involved in the
traditional or commercial technology. These are as follows:
Injection Molding: As the name refers, a plastic material is softened, and then it is
injected during the fabrication process inside a mold [2]. This process comes
under a mechanism, and the other mechanism involves the removal of the mate-
rial from the mold after which it could be cured and solidified. This is highly
used in the plastic goods fabrication process that gives a good finishing surface.

A. Suresh (*)
Department of Computer Science and Engineering, Nehru Institute of Engineering and
Technology, Coimbatore, India
R. Udendhran · G. Yamini
Department of Computer Science and Engineering, Bharathidasan University, Trichy, India

© Springer Nature Switzerland AG 2020 73

G. R. Kanagachidambaresan et al. (eds.), Internet of Things for Industry 4.0,
EAI/Springer Innovations in Communication and Computing,
https://doi.org/10.1007/978-3-030-32530-5_5
74 A. Suresh et al.

Fig. 5.1 IoT sensors in industry

The disadvantage among these processes is that the cost of the injection mold is
very high and it is not sufficient for the low-budget series.
CNC Machining: The material should be machine as a part in any field that involves
drilling, turning, or milling [3]. In this process, the machine is fitted with the
material particle, and several tools could be used to remove the material until a
desired format is obtained. Similar to the previous process, this process could not
be used for the small-scale applications due to their cost. But the accuracy of this
process is very high with the tolerance rate of 25 μm. But the internal features
could be fabricated with this process and are very tedious for the machining pro-
cess [4]. It is best suited for the engine components.
Plastic Forming: Several techniques such as pressure forming, thermoforming, and
vacuum forming are involved in this process. Each type is specific in their func-
tionality, but the process of heating a plastic sheet and draping it around the mold
is similar. Then the main plug and the sufficient air pressure could be used to
make a shape out of the sheet. The forming process utilizes thermoplastics virtu-
ally, and only one side of the plastic could be modulated by this process [5].
Compared to the above two processes, the installation and by-product costs are
very low.
Plastic Joining: This process involves in combining the semifinished parts.
Welding, adhesive bonding, and fastening are the process involved here [6].
The incorporation of latches along with the snap and hinges that fits into the
design part or with the screws and bolts as the external fasteners comes under
the process of fastening. Epoxy is said to be an adhesive that could be used in
joining the parts together involved under the category of adhesive bonding [7].
Finally, welding is the stage that involves the application of heat under suffi-
cient pressure to combine the material together. The semifinished parts and
their property play a vital role in choosing the sufficient technique [8]. However,
this process is quite time-­consuming, and the cost for the labor charges seems
to be very high [9].
5 Internet of Things and Additive Manufacturing: Toward Intelligent Production… 75

 omparison Made Among Traditional and Additive

C
Manufacturing (AM) in Industry 4.0

Based on all the abovementioned process regarding the AM and traditional

manufacturing process, a comparison table has been given that could help in choos-
ing the right and adaptable one for the process [10] (Table 5.1).
Based on the application, the best suitable method should be adapted, and iden-
tifying the requirements is the most necessary matter of all [11]. These are the fol-
lowing requirements:
Quantity: Large-scale manufacturing process could be adapted by the injection
molding and foaming. On the other hand, AM methods [12] are best adapted for
the small-scale process. The cost estimation had been made for the GoPro handle
that has been carried out with the injection molding (Quickparts) and SLS
(Sculpteo) process, and it shows the SLS technique is the most effective valid
option at 486 units as shown in Fig. 5.2.

Table 5.1 Comparison of different IoT-enabled processes with the advantages and disadvantages
in industry
Process Advantages Disadvantages
Selective laser sintering Very complex and strong Rough surface with the grainy
Layers of 0.06–0.15 mm with Possible snap features and finish
the fast print speed and rough the parts could be stacked in
surface the volume
Stereolithography Finished end product will The parts are very weak and less
Layers of 0.06–0.15 mm with be very smooth sustainable to heat and sunlight
the average print speed and
smooth surface
Binder jetting Print speed is very fast with Surface finish is very rough with
Layers of 0.089–0.12 mm the multicolor printing the weak parts and quality
with the very fast print speed facility
and rough surface
Fused deposition modeling Strength of the whole setup Poor-quality surface and printing
Layers of 0.1–0.3 mm with components are very high speed is very low
the low print speed and very with the minimization of
rough surface cost
Injection molding Reasonable turnover rate Equipment cost is very high and
Excellent surface finish with with the high tolerance parts of the equipment are really
the tolerance level of 50 μm thin
CNC machining Materials are very compact Equipment cost is very high and
Smooth surface finish with with the reasonable turnover complex design
the tolerance level of 25 μm rate and high tolerance
Plastic forming The price is very nominal Single-sided control with the
Smooth surface with the for the large body parts equipment components really thin
1 mm tolerance value and uses only thermoplastic as
typically their material
Plastic joining All the materials could be Labor cost is very high and huge
Based on the finished product used in this process time-consuming process
76 A. Suresh et al.

Fig. 5.2 An example of AM in Industry 4.0 with injection molding

Lead time: In case of the traditional manufacturing process, the number of days for
the units to be manufactured could be increased due to the arrangement of molds,
but this is not the case in 3D printing. In case of the GoPro handle, 25–1000 units
could be done at a period of 15 days, whereas 25 units could be done within
3 days, and 1000 units could be completed within 7 days with AM process.
Complexity and shapes: Customizing the complex parts is the essential one that
could be done effectively by 3D printer using a professing [13]. This could not
be effective when it comes under molding, forming, and machining.
Selection of material: Traditional manufacturing leads to the selection of high-­
quality material, but when it comes to the 3D printer, the material selection could
be scattered or thinly dispersed [14]. In certain cases, SLS uses powder made up
on thermoplastic material that is specific to that particular machine, FDM uses
molten thermoplastic, etc.
Other considerations: Several applications that use the 3D printing technology with
their tensile strength [15].

The Role of Additive and IoT in Industry 4.0

The creation of digital representation of the object is extremely important with its
geometrical attributes, texture, and color. This role is successively played with the
help of additive manufacturing (AM) and subtractive manufacturing (SM) [16]. The
5 Internet of Things and Additive Manufacturing: Toward Intelligent Production… 77

traditional way of trimming of materials with the help of high-end tools is known as
SM [1]. AM process effectively used the software platform that is commonly found
in visual aids, tooling components, functional models, direct manufacturing, etc.
[2]. The process of cutting away the material from the certain block of material to
bring that block to a required shape is termed as the subtractive manufacturing pro-
cess [3]. This traditional process could be done manually, but accurately this could
be done through the CNC machine [4]. As per the requirement of the designers, the
CNC machine could perform the cutting task, which is concentrated on the three
axes (say x, y, and z plane) of the material. In doing so there could be a less chance
for the designers to flip the material. There is a major advantage of adapting subtrac-
tive manufacturing since this process has the ability to extremely machine a minia-
ture piece of material into a living hinge [5]. This process could not be achieved
with the 3D printing. Certain parts could be created with the help of the additive
manufacturing that needs living hinge component, while special components could
be created with the help of a living hinge [17]. The most commonly used devices in
the process of Industry 4.0 manufacturing have been discussed below:

Roland 3D CNC Milling Machine MDX-40A15

The Roland 3D CNC Milling Machine had been especially designed for the process
of rapid prototyping as a benchtop CNC machine [7]. This particular 4-axis machine
could handle several types of materials and is said to be G-code compatible. A
smooth finish could be provided as an end product. The cost of the material is
extremely low [8]. This milling machine is used moderately in industries. Digital
caliper finds its place in the majority of the measurement purposes. This could pro-
mote maximum accuracy with the precise distance of measuring small objects.
These calipers permit for “zeroing” at any point along the slide, which could permit
differential measurements [9]. The Roland 3D CNC Milling Machine MDX-40A15
along with the digital calipers has been shown in Fig. 5.3.

Fig. 5.3 (a) Roland 3D CNC Milling Machine MDX-40A15. (b) Digital caliper
78 A. Suresh et al.

Additive Manufacturing Process in Industry 4.0

Additive manufacturing popularly said as rapid prototyping or 3D printing has

gained its attention recently that uses the principle of addition of materials in a layer
with the particular processes [18]. This requires an additional material to be fitted
which seems to be designed with the help of certain software in a three-dimensional
structure. This can be directly implemented in the real-time application purpose,
thereby decreasing its time and the complication. Instead of removing the materials,
the addition of materials serves the system as a better head conductor and makes
them lighter [11]. This could remove certain constraints that are created by the con-
ventional manufacturing by adding the materials layer by layer [12]. At first, the
materials are added in a thin powdered layer to the building platform, and with the
help of the laser, the powder diffuses in certain points by a computer-generated
component design data. Then another layer is placed that fuses the material to form
a predefined structure. The additive manufacturing technology can be used in wide
variety of fields starting from the medical and marine application to the electronics
and consumer accessories [19] as shown in (Fig. 5.4).
Several processes are involved in a metal additive manufacturing and they are as
follows [14].
Material Deposition Process
This method transports the material from the powder jet or the filament into the
surface. The depth of the layer could vary in a certain micrometers range. The

Fig. 5.4 Block diagram of the stages involved in the metal additive manufacturing
5 Internet of Things and Additive Manufacturing: Toward Intelligent Production… 79

important factor monitored over here is the flow rates of the material either a powder
or a polymer [15]. Depending on the powder size, the flow rate should be estimated.
The material is supplied in 2 axes although it’s a 3D axis. With the help of this pro-
cess, we could eliminate the usage of any support material that does not cause a
viable geometry [16]. Material deposition could be done with the help of cold spray,
extrusion process, and blown powder process.
Power Bed Process
(a) Laser Beam Melting
Additive manufacturing involves several ways in the deposition of the layers.
Selective laser melting (SLM) and selective laser sintering (SLS) process melts and
softens the layer settled in the material. Certain industries use the metal powder or
the polymer in the production of the layers. In the process of laser beam, the high-­
intensity laser is used to melt the metal. Melting process involves the material of
powdered form [20]. The powder is spread equally with the help of re-coater arm,
and the laser power of about 50–2000 W has been used to solidify the process.
When this is done, then another layer of powder is coated, thereby lowering the
depth of the material. This particular process is carried out in an inert temperature
of argon or helium atmosphere since the powder can easily oxidize. When the depth
of the layer is small, then the manufacturing part is well defined, although the time
of manufacturing increases. A heat posttreatment is necessarily carried out to
remove the deformation of the materials or polishing the surface [21]. This should
be done in a controlled environment in order to increase its precision rate. Similarly,
the material used in the laser polymerization is the photosensitive resin. This pro-
cess is cured with the help of the UV radiation from the low intense laser source.
(b) Electron Beam Melting
This process is more productive when compared to the laser in its features. This
process uses an electron beam of maximum power of about 3 kW. This is also car-
ried out with the certain temperature in an inert gas environment. The powder used
should be preheated that neglects the thermal gradient. Although the temperature
used here is very high, the cooling procedure is also carried out to avoid the internal
stress of the material. The powder could move at a faster rate that works well with
the huge depth of the layers. But due to its broad beam width, this method is not
precise, and it is not sufficient when compared to the laser beam method [22].
Liquid Bed Process
This involves the process of stereolithography and thermosetting resin. The prin-
ciple of photo-polymerization is involved in the stereolithography procedure. The
lower-power helium-cadmium UV radiation has been used in this process [23]. This
UV radiation coagulates to form a thin layer in the surface of the material. A SLA
machine is built in such a way that the whole process is engrossed in a liquid resin
with the laser source and the required hardware and software platform is also built.
With the CAD model, the layer is being scanned, and once this process is done, the
80 A. Suresh et al.

Lenses

X-Y scanning mirror

Laser Laser beam
Vat
Elevator

Liquid
photopolymer

Sweeper

Layered part

Build platform

Fig. 5.5 Parts involved in IoT-based stereolithography

inner portion is engraved and undated below a layer [24]. A blade is used to remove
the extra fitted layer to ensure the flatness of the surface. After this process the next
layer is deposited, and again the procedure is carried out like the previous one. With
this method we can able to produce a highly accurate polymer parts [25]. The pro-
cess of stereolithography is shown in Fig. 5.5.

IoT-Based Additive Manufacturing Versus Traditional

Manufacturing in 3D Printing

For comparing both the IoT-based AM and the traditional manufacturing process,
we have defined a part that has been printed by both the methods. Figure 5.6 shows
the block printed by additive manufacturing process. The block has a well-defined
feature and finish. The CNC milling machine encompasses a much diverse process
than the 3D printer [26]. Sufficient supervision should be provided for carrying out
the process at CNC machine though similar files are being used for making the part.
This is because the materials are removed from the block during the process. This
part is sufficiently made with the help of wax. Previous studies have stated that the
CNC machine could provide best finish or end product when compared to the 3D
printer [27]. Figure 5.7 shows the block created by the CNC machine. This particu-
lar block has certain problem when this case is closely compared. The inner crevice
of the M has not properly curved due to the miniature dimension and the size of the
tool used for cutting. We could change the drill-bit of the machine by replacing it
with the smaller one. But this could increase the complexity of the machine, thus
increasing the processing time [28].
5 Internet of Things and Additive Manufacturing: Toward Intelligent Production… 81

Fig. 5.6 Block created by

the 3D printing

Fig. 5.7 Block created by

the CNC milling machine

After the manufacturing process, it is necessary to compare the measurements

with the certain parameters [29]. There is a necessary usage of the scale factor that
compares with the original model, and this could be due to the occurrence of size
restrictions. The end product of the CNC machining material could be slight thicker
in their dimensions, whereas with the 3D printing, the end product could shrink
from their original size due to their plastic material shrinking after the process of
printing and cooling. Several parameters that are sufficiently measured for both the
SM and AM process are as follows:
• Setup – Number of steps and setup time for each machine
• Ease of use – how easy is it to use the machine
• Percentage of wastage of the material (%) – wasted material due to the process
of printing and cutting
• Machining speed – Manufacturing the parts rapidly at certain time period
• Accuracy – Comparison of accuracy made with the model
• Surface finish – smoothness of the part surface
Discussing about these parameters, it is said that the surface finish of AM is
comparatively low when compared to the CNC method [30]. In all the other param-
eters, additive manufacturing leads its place that could be modeled with the help of
CAD software platform. In fact, several trials could be possible only by the AM
82 A. Suresh et al.

process. Surface quality is the most important feature to be considered in the

additive manufacturing. Depending on the particle size, the surface should be used
in manufacturing. When the particle size is huge of above 30 μm, then the surface
produced by the selective laser melting will be coarse. However, less particle size of
below 20 μm could produce a smooth and a polished surface. There is a drawback
in using the fine-sized particle as it could hinder the flow rate of the particle in the
layer, thereby increasing the cost. The surface quality is the most important concern
in the quality estimation. Certain process involves the fine powder layer, for instance,
binder jetting process (BJG). The tolerance level should also be taken into consid-
eration due to its heat treatment in its fabrication process. If this fails, then there
could be considerable dimensional changes in the end material. To avoid this fault,
additional necessary acceptable materials could be used which could be removed at
the later. Besides, the process of SLM and EBM is more accurate with a reduced
amount of tolerance rate. Hence, with the availability of the resource, product pro-
duction time, number of products produced, etc., we must select the method based
on these requirements. The process selection is done with the help of the type of the
material we use during fabrication. When the ceramic type of material is used, then
BJG is the most preferred process that could be used. When the fabricated material
is of metals or composite type, then we can adopt any three processes like SLM,
EBM, or BJG. For example, let’s consider the magnesium-based materials. Due to
its low boiling points, the process of EBM could not be adopted. Due to their high
intense beam used in their power bed, which results in the melting of certain materi-
als. Therefore, the best adopted method could be either SLM or BJG. In the fabrica-
tion of brittle materials such titanium aluminide, the three processes could also be
adopted. But due to the high cooling rate, SLM process could not be adopted since
this could lead to the internal stress of the material. The cracks could be eliminated
only when the cooling rate is very low. The material limitations and their property
are shown in Fig. 5.8.

Fig. 5.8 Process selection

criteria
5 Internet of Things and Additive Manufacturing: Toward Intelligent Production… 83

 dvantage and Faster Deployment of IoT-Based Additive

A
Manufacturing in Industry

In the aspect of the capabilities of manufacturing, a different trade-off between AM

and CM exists among the processes in which the additive manufacturing possesses
the capability of reducing materials as well as usage of energy, less wastage, innova-
tive acceleration, as well as reliable supply chains [31]. Therefore, it is evident from
the above statement that AM plays an important role in deciding the future of manu-
facturing. Let us discuss the important advantages of the additive manufacturing.
Creativity: By integrating CM processes, the AM leads to innovative frameworks
along with its effective geometrics which can be easily developed during the
course of the manufacturing process. This type of innovation enhances the per-
formance, environmental-friendly, as well as cost-effective solution in engineer-
ing and product application, for instance, AM systems which are programmed to
accommodate properties inside a component which enhances the performance
by employing conventional processing.
Design products with fewer, more complex parts: The unique ability to create com-
ponents using few but complex parts is a major advantage of AM. This feature is
so unique that it only exists in AM which upgrades the performance at a system
level excluding the manufacturability present at the subsystem level. In order to
reduce the overhead association concerning the production planning as well as
control and documentation, a minimum number of parts in an assembly are con-
sidered. By employed minimum number of parts, less labor time needed to man-
ufacture the product and the “footprint” of the assembly line are gained which in
return lessen the overall reduction in manufacturing expenses.
Preserve energy by obviating production process: Lower energy consumption is
achieved by deploying AM which results in less material usage, encouraging
recycle of by-products, and manufacturing lighter products.
Decrease in wastage: Creating objects layer by layer reduces materials needs and
less human error in production as well as the costs by 96% unlike the traditional
manufacturing working that increase usage of material. Eventually,
“cradle-­togate” present in the environmental footprints of parts manufacturing
through rejection of the tools diminishes as well as material scrap existing in the
CM processes.
Faster deployment to Industry 4.0: As soon as the Standard Tessellation Language
[STL] file is created by employing 3D digital description (3D scanning or 3D
imaging) to the parts, the need for more expensive, in-effective part tooling as
well as prototype fabrication is achieved.
Less weight: With less material, complex shapes applied to the specification of the
conventional parts with the help of the AM can be achieved.
AM enables rapid response to markets: With the ability to build fresh production
options outside of factories, for instance, mobile units that can be located close to
the source of local materials, thereby increasing the agility of manufacturing. AM
84 A. Suresh et al.

does not include high-cost tools unlike in traditional manufacturing which leads to
longer production process and slow production rate when there is need for high
volume of production. AM does exist in the manufacturing operation to fabricate the
molds, as well as tooling employed in the production.
In order to deploy the tooling faster to market with less cost, tooling is created as
well as printed directly to the market and not outsourced. Spare tooling parts can be
manufactured on demand, excluding the need for stockpiles as well as complex sup-
ply chains. However, one challenge encountered in AM tools in manufacturing
metal tooling is the experience of higher capital costs as well as steep learning
curves which should be addressed by increasing the rate of consumption by the
manufacturers.

Case Study

By employing injection molding, numerous manufacturing cost data for five unique
objects, namely, a cockerel, a GoPro handle, and representative models of products,
were gathered. The below table represents the five major objects along with their
bounding box dimensions: This section of case study presents a vivid understanding
of the Sculpteo batch control 3D printing which is more costly unlike the rapid
injection molding for industry representative parts.
The details regarding the cost of five parts in 3D printing production can be
found in the Sculpteo website. There are several algorithms which can assess the
part for 3D print-ability for any errors which lead to failure of 3D printing. Moreover,
“Solidity Check” can be organized by the Sculpteo which assures that all behavior
of the part possesses minimum feature size threshold of approximately 1 mm. The
calculation for cost per unit and its total batch cost can be determined by employing
the “Batch Control” option present in the selective laser sintering.
By producing effective production run, the batch control can be automated for
more than ten parts which optimizes the unit price by producing effective solutions
which include orientation, layer thickness, and finish quality which leads to reliable
control and well suited for separate products. By employing the “rapid injection
molding” method, the production components can be produced based on the quotes
from the manufacture and mainly employed for prototypes as well as short series
manufacturing, but to a limit of 10,000 units. But the parts which need hollow inte-
rior do not need this technique. However, for parts with hollow interior, “gas-­
assisted injection molding” can be employed which is considered as secondary
option for manufacturing.
The charts represent the comparison among the injection-molded technique and
the 3D printing techniques. The number of manufactured units is being mentioned
by the intersection point in the chart as shown in Fig. 5.9.
The data gathered presents the 3D printing which offers insight into the eco-
nomical needs to manufacture groups of approximately 500 units more cost-­
effectively unlike the rapid injection molding methods. The injection-molded parts
5 Internet of Things and Additive Manufacturing: Toward Intelligent Production… 85

Fig. 5.9 Comparison chart Remote Control

for various manufacturing 240

parts 192

144

Cost (€)
96

48

0
25 50 100 502 900
No. of Units
Sculpteo Protomold
Car Handle
240

192

144

Cost (€)
96

48

0
25 50 100 407 900
No. of Units
Sculpteo Quickparts

Sprocket
240

192

144
Cost (€)

96

48

0
25 50 100 294 900
No. of Units
Sculpteo Quickparts

Rooster
240

192

144
Cost (€)

96

48

0
25 50 100 492 900
No. of Units
Sculpteo Protomold
86 A. Suresh et al.

and a 3D printed part consist of different materials, but some of the properties do not
differ. However, the 3D printed parts possess various surface finish unlike the
injection-­molded parts. But surface finish and material properties of the 3D printed
part serve its need for any specific products or applications, and the economical
manufacturing method for up to 500 unit production runs does not depend on
unit size.

Influence of 3D Printing in Industry

In the section of case study, the influence linked with growing part size in both 3D
printing and injection molding manufacturing technique is presented and also
analyzed.
In order to complete this case study, quotes for 3D printing with a generic, hol-
low, five-sided box of increasing size as well as injection molding were produced.
By employing “Batch Control” option for selective laser sintering, the 3D print-
ing quotes were produced by the Sculpteo. The cost for the parts in the rapid injec-
tion molding was estimated by employing CustomPartNet, which is a free online
resource for determining the cost of manufactured parts.
With the help of extensive part gallery, a box was chosen to produce baseline
manufacturing cost which determines injection molding. The below table presents
the chosen settings for all manufacturing processes.
As mentioned earlier, by employing the CustomPartNet estimation tool, manu-
facturing cost estimations were produced for part volumes beginning at 10 units and
then at intervals of 50 from 50 units to 1000 units. As a result, the data was plotted
which enabled comparison of total cost with the number of manufactured units, and
the data was obtained from the Sculpteo’s batch control option with standard layer
thickness. The data reveals that 3D printing is cheap as soon as the size of the object
grows. In SLS, the high cost of the largest box offers better results for the cost cal-
culation algorithms. During the SLS process, the machine manufacturing time per
unit as well as quantity of material plays an important role such a way if there is an
increase in the Z-dimension which is the height of the part in the build volume of the
SLS machine, then the component consumes more time for production as well as
consumes more material in real quantity as well as manufacturing quantity.
Moreover, an increase in production is needed when a component increases in size
and few can suit in a single build volume.
However, the rapid injection molding does not experience this challenge provid-
ing 3D printing to gain more economic viability in the case of big parts if the pro-
duction trials are very diminutive, thus making it effective.
There are several advantages for additive manufacturing process in Industry 4.0
when compared with the traditional manufacturing technology. The neural networks
automate the manufacturing in almost all areas of sectors [32, 33]. But various
parameters tend to limit the usage of AM. In this chapter, the technology behind
the AM and traditional manufacturing had been mentioned. With this chapter, a
5 Internet of Things and Additive Manufacturing: Toward Intelligent Production… 87

customer could be able to find a suitable technology that fits with the process could
be identified. Thus, the working of the manufacturing process has been explained in
this chapter. The process variables such as the flow rate and temperature play a
major role in this additive manufacturing process. The dimensions and the surface
quality should be considered for the proper end product of the material. Although
there are certain limitations in the process selection, the best method should be
adapted depending on the materials used. New design possibilities and product par-
adigms along with the new products had bought advancement in the field of AM. To
bring the AM into the field, several business plans along with exploring and adopt-
ing AM in small and large scale should be made at a best rate.

Conclusion

Although several equipment have been used in the process of AM in Industry 4.0 by
quickly varying the parts, it is not necessary that they should have a convenient
usage with the enough equality. In case of prototyping, this has an ease usage, but
this could not be applicable for educating environment. IoT and AM have its own
pros and cons. It’s difficult to create a case of one type over the other as a broad
recommendation. This recommendation is based on the result and their application
purpose. If a product should be created at a faster rate that has the capability to cre-
ate numerous parts concurrently, then AM could be the best choice of selection. But
while considering the surface finish of the material, then AM should be avoided that
could shrink the material by size. Hence IoT-based AM could be adapted in this
case. It is intellectual to use the combination of both the IoT and AM. We could
design and create the parts by means of AM technology in Industry 4.0 with
enhanced performance.

References

1. Guan, K., Qin, W., Liu, Y., Yin, X. Q., Peng, C., Lv, M., Sun, Q., & Wu, J. Q. (2016). Evolution
of porosity, pore size and permeate flux of ceramic membranes during sintering process.
Journal of Membrane Science, 520, 166–175.
2. Yang, D. L., Yhang, Y., Song, X. Z., Chen, Y. Z., Shen, Z. Q., & Yang, C. (2016). Effects of
sintering temperature and holding time on porosity and shrinkage of glass tubes. Ceramics
International, 42, 5906–5910.
3. Wilkinson, S. K., Turnbull, S. A., Yan, Z., Stitt, E. H., & Marigo, M. (2017). A parametric
evaluation of powder flowability using a freeman rheometer through statistical and sensitivity
analysis: A discrete element method (DEM) study. Computers and Chemical Engineering, 97,
161–174.
4. Giuntini, D., Chen, I.-W., & Olevsky, E. A. (2016). Sintering shape distortions controlled by
interface roughness in powder compositions. Scripta Materialia, 124, 38–41.
5. Molla, T. T., Bjork, R., Olevsky, E., Pryds, N., & Frandsen, H. L. (2014). Multi-scale modeling
of shape distortions during sintering of bi-layers. Computational Materials Science, 88, 28–36.
88 A. Suresh et al.

6. Vasinonta, A., Beuth, J., & Griffith, M. (2008). Process maps for controlling residual stress
and melt pool size in laser-based SFF processes. In Solid freeform fabrication proceedings
(pp. 200–208). Austin: University of Texas at Austin.
7. Gockel, J., Beuth, J., & Taminger, K. (2014). Integrated control of solidification microstructure
and melt pool dimensions in electron beam wire feed additive manufacturing of Ti-6Al-4V.
Additive Manufacturing, 1–4, 119–126.
8. Shen, N., & Chou, K. (2012). Numerical thermal analysis in electron beam additive manu-
facturing with preheating effects. Proceedings of the 23rd annual international solid freeform
fabrication symposium, pp. 774–784.
9. Shen, N., & Chou, Y. K. (2012). Thermal modeling of electron beam additive manufactur-
ing process—powder sintering effects. Proc. the 7th ASME 2012 International Manufacturing
Science and Engineering Conference, pp. 287–295.
10. Ardila, L.C., Garciandia, F., Gonzalez-Diaz, J.B., Alvarez, P., Echeverria, A., Petite, M.M.
and Deffley, R. (2014). Ochoa, J. Effect of IN718 recycled powder reuse on properties of parts
manufactured by means of selective laser melting. Physics Procedia. 56, 99–107.
11. Prasad, Y. V. R. K., Rao, K. P., & Sasidhar, S. (Eds.). (2015). Hot working guide: A com-
pendium of processing maps. ASM International. Second Edition. ISBN: 978-1-62708-
091-0. https://www.asminternational.org/home/-/journal_content/56/10192/25536184/
PUBLICATION
12. Buffa, G., Campanella, D., Fratini, L., & Micari, F. (2016). AZ31 magnesium alloy recy-
cling through friction stir extrusion process. International Journal of Material Forming, 9(5),
613–618.
13. Shao, Y., Tang, T., Li, D., & Peng, Y. (2015). Simulation of the extrusion texture of magnesium
alloy AZ31 using crystal plasticity finite element method. Materials Science Forum, 817, 538.
14. Wang, X., Wu, M., Ma, W., Lu, Y., & Yuan, S. (2016). Achieving superplasticity in AZ31 mag-
nesium alloy processed by hot extrusion and rolling. Journal of Materials Engineering and
Performance, 25(1), 64–67.
15. Stráská, J., Janeček, M., Čížek, J., Stráský, J., & Hadzima, B. (2014). Microstructure stability
of ultra-fine grained magnesium alloy AZ31 processed by extrusion and equal-channel angular
pressing (EX–ECAP). Materials Characterization, 94, 69–79.
16. Wang, S. C., Zhang, X. H., Hu, X. L., Cai, Z. G., & Peng, B. (2017, March). Study on the hot
extrusion forming process of AZ31 magnesium alloy cylindrical shell. In Advanced materi-
als and energy sustainability: Proceedings of the 2016 international conference on advanced
materials and energy sustainability (AMES2016) (pp. 249–258). World Scientific.
17. Vasinonta, A., Beuth, J., & Griffith, M. (2000). Process maps for controlling residual stress
and melt pool size in laser-based SFF processes. In Solid freeform fabrication proceedings
(pp. 200–208). Austin: University of Texas at Austin.
18. Ardila, L.C., Garciandia, F., Gonzalez-Diaz, J.B., Alvarez, P., Echeverria, A., Petite, M.M.
and Deffley, R. (2014). Ochoa, J. Effect of IN718 recycled powder reuse on properties of parts
manufactured by means of selective laser melting. Physics Procedia. 56, 99–107.
19. Shao, Y., Tang, T., Li, D., & Peng, Y. (2015, April). Simulation of the extrusion texture of mag-
nesium alloy AZ31 using crystal plasticity finite element method. Materials Science Forum,
817, 538.
20. Yu, H., Park, S. H., & You, B. S. (2015). Die angle dependency of microstructural inhomo-
geneity in an indirect-extruded AZ31 magnesium alloy. Journal of Materials Processing
Technology, 224, 181–188.
21. Kim, B., Park, C. H., Kim, H. S., You, B. S., & Park, S. S. (2014). Grain refinement and
improved tensile properties of Mg–3Al–1Zn alloy processed by low-temperature indirect
extrusion. Scripta Materialia, 76, 21–24.
22. Tang, W., Huang, S., Li, D., & Peng, Y. (2015). Mechanical anisotropy and deep drawing
behaviors of AZ31 magnesium alloy sheets produced by unidirectional and cross rolling.
Journal of Materials Processing Technology, 215, 320–326.
5 Internet of Things and Additive Manufacturing: Toward Intelligent Production… 89

23. Lu, L., Liu, C., Zhao, J., Zeng, W., & Wang, Z. (2015). Modification of grain refinement
and texture in AZ31 Mg alloy by a new plastic deformation method. Journal of Alloys and
Compounds, 628, 130–134.
24. Yang, Q., Jiang, B., He, J., Song, B., Liu, W., Dong, H., & Pan, F. (2014). Tailoring texture and
refining grain of magnesium alloy by differential speed extrusion process. Materials Science
and Engineering: A, 612, 187–191.
25. Yang, Q., Jiang, B., Zhou, G., Dai, J., & Pan, F. (2014). Influence of an asymmetric shear
deformation on microstructure evolution and mechanical behavior of AZ31 magnesium alloy
sheet. Materials Science and Engineering: A, 590, 440–447.
26. Hu, H. J., Fan, J. Z., Zhai, Z. Y., Wang, H., Li, Y. Y., & Gong, X. B. (2014). Physical fields
evolution and microstructures for compound extrusion of AZ31 magnesium alloy. Russian
Journal of Non-Ferrous Metals, 55(3), 254–262.
27. Pan, F., Wang, Q., Jiang, B., He, J., Chai, Y., & Xu, J. (2016). An effective approach called the
composite extrusion to improve the mechanical properties of AZ31 magnesium alloy sheets.
Materials Science and Engineering: A, 655, 339–345.
28. Drozdenko, D., Bohlen, J., Yi, S., Minárik, P., Chmelík, F., & Dobroň, P. (2016). Investigating
a twinning–detwinning process in wrought Mg alloys by the acoustic emission technique. Acta
Materialia, 110, 103–113.
29. Kecskes, L. J., Hammond, V. H., Eichhorst, M., Herzig, N., & Meyer, L. (2016). Development
of rolling schedules for equal channel angular extrusion (ECAE)-processed AZ31 magnesium
alloy. US Army Research Laboratory Aberdeen Proving Ground United States.
30. Kondoh, K., Hotta, M., Liao, J., Kaneko, K., Fujii, N., Kametani, H., & Koshi, A., Kurimoto
Ltd. (2016). Production method of extrusion billet and production method of magnesium alloy
material. U.S. Patent 9,518,314.
31. Suh, J., Victoria-Hernandez, J., Letzig, D., Golle, R., Yi, S., Bohlen, J., & Volk, W. (2015).
Improvement in cold formability of AZ31 magnesium alloy sheets processed by equal channel
angular pressing. Journal of Materials Processing Technology, 217, 286–293.
32. Suresh, A., Udendhran, R., Balamurgan, M., et al. (2019). A novel internet of things frame-
work integrated with real time monitoring for intelligent healthcare environment. Springer-­
Journal of Medical System, 43, 165. https://doi.org/10.1007/s10916-019-1302-9.
33. Suresh, A., Udendhran, R., & Balamurgan, M. (2019). Hybridized neural network and decision
tree based classifier for prognostic decision making in breast cancers. Springer - Journal of
Soft Computing. https://doi.org/10.1007/s00500-019-04066-4.
Chapter 6
Deep Learning Enabled Smart Industrial
Workers Precaution System Using Single
Board Computer (SBC)

S. Pradeep Kumar, S. Selvakumari, S. Praveena, and S. Rajiv

Introduction

From the Industrial point of view, the safety measures for the workers are manda-
tory and must be able to be checked on a routine base. The evaluation of the IIoT-4.0
is given in the following diagram.
In this paper, the single board computer (SBC) is used to solve the mandatory
process. The SBC is used to control whether switching options should be enabled or
disabled to the power supply of a big machine after verifying that the worker is
wearing all the necessary kinds of safety equipment. Deep learning uses image pro-
cessing to detect the safety objects, such as helmet, gloves, vest coat, etc. The pro-
gramming language python is used to access the above process with the camera for
image processing. In each level of transformation, the input data will be slightly
more abstract and a composite mode of representation.
Figure 6.1 shows the evaluation of the IIoT from the Industry 1.0 to Industry 4.0.
In the beginning of Industrial 1.0, the mechanization of the weaving loom by steam
power was initiated. After that in the 1870s, IoT was widely used for mass produc-
tion and electrical energy. A huge change occurred in the evaluation of the IoT in
1969 with the marvelous invention of the machine called a computer. Ever since, the

S. Pradeep Kumar (*)

Sri Venkateswara College of Engineering and Technology, Thiruvallur, Tamil Nadu, India
S. Selvakumari
Sriram Engineering College, Chennai, Tamil Nadu, India
S. Praveena
Free Launcher, Thiruvallur, Tamil Nadu, India
S. Rajiv
VIT, Vellore, Tamil Nadu, India

© Springer Nature Switzerland AG 2020 91

G. R. Kanagachidambaresan et al. (eds.), Internet of Things for Industry 4.0,
EAI/Springer Innovations in Communication and Computing,
https://doi.org/10.1007/978-3-030-32530-5_6
92 S. Pradeep Kumar et al.

Fig. 6.1 Industrial internet of things 4.0

IoT has played a major role in the industrial process, which includes the communi-
cation process with the network and machine with cyber security [1].
This chapter is sorted as follows. Section “Prelims” talks about the framework
segments topics used in this project, including those of prelims and architecture of
the containing interfaces. Section “Proposed System” gives an outlined description
of the proposed method and model of mathematical execution. The graph and the
output are condensed in Section “Math Model”. Section “Results” gives recommen-
dations for future work, and the chapter is concluded in Section “Conclusion”.

Prelims

Image Processing

Imagining processing is the process of changing over ordinary images to digitalized

format, and it also carries out certain tasks on the image, such as a request to upgrade
a picture or to remove some valuable data from it [2, 3]. Another way of signal
agreement to get the input as image, such as video edge or photo yield, might be a
picture or specifications about the picture. Typically, the picture preparing frame-
work incorporates treating the image as a few dimensional sign, while previously
some type of methods may have been applied to them. The processing of an image
commonly includes the following steps:
• Importing the picture using an optical scanner or by computerized picture or by
camera.
• Analyzing and controlling the picture, which incorporates information compres-
sion and picture upgrade and spotting designs that are not visible to human eyes.
6 Deep Learning Enabled Smart Industrial Workers Precaution System Using Single… 93

• Output is the last stage wherein the result can be a modified picture or sifted
picture that depends on picture examination.
Picture preparation is the procedure of the examination and control of a digitized
picture, particularly so as to improve its quality. Observation has the capacity to
consequently recognize an individual through facial recognition and may improve
modern security strategies. Current progressions in man-made consciousness and
machine learning are improving the precision of such systems. Improved compo-
nent learning and progressed neural system configuration have made article discov-
ery significantly all the more fascinating and productive. The main purposes of the
picture preparing include the following: Representation is to watch the items that
are not noticeable; Image honing and rebuilding is to make superior pictures; Picture
recovery is to look for the picture of intrigue; Estimation of pattern is to quantify
different articles in a picture; and Picture Recognition is to recognize the items in a
picture.

Deep Learning

Deep learning is a widely used technique, and the “deep” in “deep learning” has
lead us to understand the quantity of layers through which the information is
changed to achieve the exact output that overcomes the drawbacks of the old algo-
rithm of the machine learning; more accurately called deep learning frameworks.
This learning has been applied to the fields of computer vision (CV) of various ver-
sions, speech recognition, natural language processing (NLP), audio-recognition
(AR), social network filtering (SNF), machine-translation, medical image analysis,
and other processes of machine learning concepts [4]. Deep learning designs are
regularly built with a ravenous layer-by-layer method of these deliberations and
choose which highlights improve execution.
Deep learning strategies hinder highlight designing by making an elucidation of
the data into negligible, widely appealing depictions similar to essential parts and
derive layered structures that oust abundance in depiction. Profound learning com-
putations can be associated with unsupervised learning assignments [5, 6]. This is a
basic bit of leeway in light of the way that unlabeled data are more rich than named
data. Cases of profound structures that can be set up in an unsupervised manner are
neural history blowers and profound conviction frameworks.
Deep learning is a sort of AI where a model figures out how to perform portrayal
assignments straight forwardly from pictures, substance or sound. Profound learn-
ing is ordinarily executed utilizing neural framework building. The term profound
implies the number of layers in the framework—the more layers, the more profound
the framework. Regular neural frameworks contain only a couple of layers, while
profound frameworks can have hundreds.
94 S. Pradeep Kumar et al.

Wireless Sensor Network (WSN)

A Wireless Sensor Network (WSN) is a network of various kinds of devices that

communicate the data and information collected from one supervised field through
wireless links [7–9]. The Gathered data is transformed through multiple nodes, and
the data is connected to other networks with a gateway. The WSN protocol selected
may depend upon the requirements of the application. In the WSN, the wireless
process may consist of base stations and a number of nodes called wireless sensors.
Sensors may vary according to their types, size, speed rate, bandwidth, capacity, and
memory. Some of the applications of WSNs are used to measure the environmental
physical conditions and to organize the gathered data at some point of location.

TensorFlow

The tool used to write machine learning programs is TensorFlow. TensorFlow is one
of the broadly utilized libraries for executing machine learning and different calcu-
lations including a huge number of scientific activities. TensorFlow was developed
by Google, and it is one of the most popular libraries for Machine Learning. The
core component of TensorFlow is the computational graph and Tensors that traverse
among all the nodes through edges [10, 11]. TensorFlow is a powerful open-source,
artificial intellect library using data flow graphs to build a myriad of models. It is
mostly used for classification, perception, understanding, discovering, predicting,
and making.
TensorFlow excels at numerical computing, which is critical for deep learning. It
has a rich set of application programming interfaces in most major languages and
environments needed for deep learning projects. A tensor flow may contain a set of
primitive values shaped into an array of any number of dimensions. These mon-
strous quantities of huge clusters are the reason that GPUs and different processors
intended to do skimming point arithmetic exceed expectations at accelerating these
calculations and algorithms.

Convolution Neural Network (CNN)

Image classification using CNN is most effectual. Primary and leading, we need a
set of imagery. In this case, we take images as our initial training data set [12]. The
most common reflection data input parameters are the number of images, image
size, number of channels, and the number of levels per pixel.
Figure 6.2 is a clear view of the convolutional neural network explained with the
working process and classification models. In the above diagram, the input is taken
through the camera, and the processes of convolution and pooling layers are carried
6 Deep Learning Enabled Smart Industrial Workers Precaution System Using Single… 95

Helmet
Gloves
vest cost
safety shoe

FULLY
INPUT CONVOLUTION + RELU POOLING CONVOLUTION + RELU POOLING FLATTEN CONNECTED SOFTMAX

HIDDEN LAYERS CLASSIFICATION

Fig. 6.2 Convolutional Neural Network (CNN)

out and the same processes repeated twice. Some of the algorithm is used to identify
the process of safety equipment based on the above method, and with the help of
this input, the other process will continue [13]. The output could be compared with
deep learning, and based on the output, the SBC will execute the rest of the process.

Proposed System

The working principle of the proposed methodology is in the deep learning concept,
and it will verify the occurrence of a helmet and other protection gadgets to ensure
the safety of factory workers [13]. The proposed line of attack incorporates the pro-
tection aspects confirmation of the labor behavior of cutting and grinding machines.
A camera is located before the workbench and connected with the single board
computer that may verify the helmet presence through image processing. The veri-
fication time of TensorFlow using the Raspberry pi, deep learning method is mea-
sured with different subjects and helmets [14]. The large machine will be switched
on after confirmation that the worker is wearing a helmet. A camera is located fac-
ing the workers and the feed from the camera connected with the single board com-
puter verifies the helmet occurrence through image dispensation (Fig. 6.3).
Deep learning and object detection serve as the main features of this work. The
python program code used for object detection is run inside the Raspberry pi type of
(single board computer) board to identify objects [14]. The camera sensor and
obstacle detection sensors are used to activate the mechanism. Input to the system is
from the camera sensor. The single board computer called Raspberry pi is con-
nected with the camera as a USB. Through Wi-Fi, the device is connected to a lap-
top or any monitor.
Figure 6.4 shows the process of identifying the safety equipment and allowing
the smart locking and unlocking of the big critical machine [15]. As per the above
diagram, the camera is used to identify the safety equipment of workers if they are
not wearing any one of the safety elements the big machine will not switch on and
96 S. Pradeep Kumar et al.

Fig. 6.3 Dataflow diagram

Fig. 6.4 Safety equipment

detecting

allow work on it [14, 16]. The camera is connected to the single board computer,
and that device is trained by the concept of deep learning with the image processing.
The Python programming language is used to access the above process.

Math Model

Dissimilar to YOLO, SSD embraces a multi-scale method, which implies that the
component maps that are utilized to identify various items are at various scales.
Since each element guide is created from convolution results at a similar level, the
convolution open fields of the various levels must be different in size [17]. In par-
ticular, the receptive fields of a high-level convolution layer are somehow larger
than those of lower layers [18–20], and the separated data compared to an abnormal
state highlight layer is more abstract. The more abstract the feature extraction infor-
mation is, the less detailed the information will be; in this manner, SSD discovery is
likewise insensitive toward small articles.
6 Deep Learning Enabled Smart Industrial Workers Precaution System Using Single… 97

The formula for computing the convolution responsive field is the following:

SCRF ( i ) = ( SCRF ( i − 1) − 1) N s + S f (6.1)

where SCRF(i) is the size of the convolution receptive field of the ith layer, Ns is the
step length, and Sf is the size of the filter.

Pb
Ca −
amin = 2 P =  i + 0.5 − pk  P
 Fk (6.2)
2 
img img
Pfeature 

hb
Cb −
bmin = 2 h =  j + 0.5 − hk 
 himg
img  Fk (6.3)
hfeature  2 

Pb
Ca −
amax = 2 P =  i + 0.5 − pk  P
 Fk (6.4)
2 
img img
Pfeature 

hb
Cb −
bmax = 2 h =  j + 0.5 − hk 
 himg
img  Fk (6.5)
hfeature  2 

where (cx, cy) denotes the center coordinates of the default bounding box, hb is the
height of the default bouncing box, wb is the width of the default jumping box, hfeature
is the tallness of the component map, wfeature is the width of the element map, |fk| is
the size in the kth highlight map, wimg is the stature of the first picture, wimg is the
width of the first picture, which is focused at (i + 0.5|fk|), (j + 0.5|Fsk|) and scaled to
a height of hk and a width of wk in the kth highlight map. In the event that the
SSD_300 × 300 model is embraced, with the end goal that the size of the informa-
tion picture is 300 × 300, the component maps of the model are for the most part
created from layers. The sizes of the convolution open field and the mapping area of
the default bouncing box on each element map are:

L overlap = ( L1 ∩ L 2 ) / ( L1 ∪ L 2 ) (6.6)

To assess the recovery capacities of the strategies, their capacity to recover ques-
tions in each picture was first evaluated, and the normal estimation of these out-
comes was then determined [21, 22]. Typically, the recovery capacity is surveyed
utilizing the F-measure, which is the weighted normal of the exactness and review
and is communicated as follows:

F=
(a 2
)
+1 × P × R
(6.7)
a × (P + R)
2
98 S. Pradeep Kumar et al.

Results

In this work, we have established a secure working environment for the workers by
using the deep learning method and image processing, which help ensure the safety
of workers in the industrial environment. Additionally, our work helps us to avoid
major accidents in industry and to reduce man power for monitoring, and we are
also able to achieve this method at minimum cost [23, 24]. An opportunity for future
work can be to extend the process for other wearable safety materials (Fig. 6.5).
Figure 6.6 shows the detection of the safety equipment helmet using our device.

Fig. 6.5 Detecting human

and other objects using
deep learning

Fig. 6.6 Safety equipment

detecting
6 Deep Learning Enabled Smart Industrial Workers Precaution System Using Single… 99

Relay

Critical Machine

Raspberry Pi

Camera

Fig. 6.7 Raspberry pi connected LED instead of critical machine

Figure 6.7 shows the smart locking system of the huge critical machine. After
verifying the safety equipment, the device enables the power supply. When the
trained device scans the worker who stands in front of the workbench, as soon as the
helmet is detected the green light glows. Only when the green light glows is the
worker able to power on the machine. If the red light glows, it means the worker is
not wearing the safety equipment, and if workers try to start the machine, it will
not start.
After ensuring the helmet is worn by the worker, the machine will start, and in
the meanwhile, if the worker tries to remove the helmet, the critical machine will
shutdown with safety mode. These precautions are for the workers safety so that a
major accident can be avoided.
Another expansion of the program is to recognize whether the laborers are wear-
ing their vests; a third augmentation includes adding a program to distinguish a
specialist dependent on the framework of a specialist and ordinary leg developments
while strolling. Moreover, the utilization of a database of the board for every one of
the laborers with the end goal that the program can record every one of the
­specialists’ chronicles in regard to security rule infringement is being investigated.
Further examination will proceed until the program can adequately distinguish all
the PPEs (vests, boots, globes, and so forth). It will be simpler to recognize vests
when utilizing this program than to distinguish boots and gloves on the grounds that
the cameras by and by cannot accurately catch the state of the boots and gloves on
account of their small size contrasted with vests.
100 S. Pradeep Kumar et al.

Conclusion

In this work, we have established a secure working environment for the workers by
using the deep learning method and image processing, which help to ensure the
safety of workers in the industrial environment. Additionally, our work helps us to
avoid major accidents in industry and to reduce man power for monitoring, and we
are also able to achieve this method at minimum cost. In future, the work can be
extended for other wearable safety materials.

References

1. Din, S., Paul, A., Ahmad, A., Gupta, B. B., & Rho, S. (2018). Service orchestration of optimiz-
ing continuous features in industrial surveillance using big data based fog-enabled internet of
things. IEEE Access, 6, 21582–21591.
2. Krüger, J., Lien, T. K., & Verl, A. (2009). Cooperation of human and machines in assembly
lines. CIRP Annals, 58(2), 628–646. [2].
3. Haddadin, S., Albu-Schäffer, A., & Hirzinger, G. (2009). Requirements for safe robots:
Measurements, analysis and new insights. The International Journal of Robotics Research,
28(11–12), 1507–1527.
4. Asadi, E., Li, B., & Chen, I. M. (2018). Pictobot: A cooperative painting robot for interior
finishing of industrial developments. IEEE Robotics & Automation Magazine, 25(2), 82–94.
5. Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). Imagenet classification with deep
convolutional neural networks. In Advances in neural information processing systems
(pp. 1097–1105).
6. Xia, F., Campi, F., & Bahreyni, B. (2018). Tri-mode capacitive proximity detection towards
improved safety in industrial robotics. IEEE Sensors Journal, 18(12), 5058–5066.
7. Martin, A., & Voix, J. (2018). In-ear audio wearable: Measurement of heart and breathing
rates for health and safety monitoring. IEEE Transactions on Biomedical Engineering, 65(6),
1256–1263.
8. Lazzerini, B., & Pistolesi, F. (2018). Multiobjective personnel assignment exploiting work-
ers’ sensitivity to risk. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 48(8),
1267–1282.
9. Lazzerini, B., & Pistolesi, F. (2018). An integrated optimization system for safe job assignment
based on human factors and behavior. IEEE Systems Journal, 12(2), 1158–1169.
10. Tsukada, K., Tomioka, T., Wakabayashi, S., Sakai, K., & Kiwa, T. (2018). Magnetic detection
of steel corrosion at a buried position near the ground level using a magnetic resistance sensor.
IEEE Transactions on Magnetics, 54(11), 1–4.
11. Rybczyński, A., Wolska, A., Wisełka, M., Matusiak, J., & Pfeifer, T. (2018). Welding arc
ignition and photobiological hazard evaluation. In 2018 IEEE International Conference on
Environment and Electrical Engineering and 2018 IEEE Industrial and Commercial Power
Systems Europe (EEEIC/I&CPS Europe) (pp. 1–6). IEEE.
12. Doan, D. R. (2018). Unrecognized hazards [electrical safety]. IEEE Industry Applications
Magazine, 24(5), 5–5.
13. Doan, D. (2017). Equipment testing safety [electrical safety]. IEEE Industry Applications
Magazine, 23(4), 6–7.
14. Doan, D. (2017). Dangers of distractions [electrical safety]. IEEE Industry Applications
Magazine, 23(3), 5–72.
6 Deep Learning Enabled Smart Industrial Workers Precaution System Using Single… 101

15. Spang, R. J. (2017). Integrated safety management: Creating an all-inclusive electrical safety
program. IEEE Industry Applications Magazine, 23(3), 64–70.
16. Mayer, S., Hodges, J., Yu, D., Kritzler, M., & Michahelles, F. (2017). An open semantic frame-
work for the industrial Internet of Things. IEEE Intelligent Systems, 32(1), 96–101.
17. Doan, D. (2017). Workplace hazards for the electrical worker’ [electrical safety]. IEEE
Industry Applications Magazine, 23(2), 6–6.
18. Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). Imagenet classification with deep
convolutional neural networks. In Advances in neural information processing systems,
pp. 1097–1105.
19. Lawrence, S., Giles, C. L., Tsoi, A. C., & Back, A. D. (1997). Face recognition: A convolu-
tional neural-network approach. IEEE Transactions on Neural Networks, 8(1), 98–113.
20. Chen, C. F., & Hsiao, C. H. (1997). Haar wavelet method for solving lumped and distributed-­
parameter systems. IEE Proceedings-Control Theory and Applications, 144(1), 87–94.
21. Henriques, V., & Malekian, R. (2016). Mine safety system using wireless sensor network.
IEEE Access, 4, 3511–3521.
22. Khan, M., Silva, B. N., & Han, K. (2016). Internet of things based energy aware smart home
control system. IEEE Access, 4, 7556–7566.
23. Li W., Zhao R., Xiao T., & Wang X. (2014). DeepReID: Deep filter pairing
neuralnetworkforpersonre-­identification. In Proceedings of the IEEE conference on computer
vision and pattern recognition (pp. 152–159).
24. Luo, P., Tian, Y., Wang, X., & Tang, X. (2014). Switchable deep network for pedestrian detec-
tion. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition
(pp. 899–906).
Chapter 7
A Novel Design Augmentation of Bio-­
Inspired Artificial Immune Technique
in Securing Internet of Things (IOT)

G. Usha, P. Madhavan, and M. V. Ranjith Kumar

Introduction

Various germs, viruses, and other pathogens are more harmful to the human immune
system. However, the human immune system has the ability to protect against vari-
ous attacks and threats to keep the body healthy. Various types of basic definitions
are existing to understand the concepts of self, nonself, and antigen identification
[1, 2]. The artificial immune systems (AIS) [3] are a type of computational intelli-
gent techniques. The AIS are based on rule-based machine-learning techniques that
are based on human immune systems. These artificial immune systems [3] are used
in various types of problem-solving techniques. These artificial immune systems
(AIS) are new bio-inspired model, which are used to solve information security. AIS
techniques [4] are akin to other problem-solving techniques such as genetic algo-
rithms, neural networks, evolutionary algorithm, and swarm intelligence techniques.
AIS techniques [5] consist of unique features such as distributed, dynamic in nature,
self-learning, self-adapting, and self-organizing in nature. The main unique features
for using AIS technique is that it uses an easy method to identify patterns. Various
types of immune algorithms [3, 6] are existing to security issues of immune sys-
tems. These are shown in Fig. 7.1.
The bio-inspired algorithm has been playing a vital role in the AIS. The simple
taxonomy of algorithms are depicted in Fig. 7.1.
The negative selection algorithms (NSA) has to be accessed by hypothesis-­
resistant techniques. These can be included with the self and nonself models to cre-
ate and identify arbitrary and counteracting elements. Self-accessing techniques
managed by the T cell can be measured at the development stage.

G. Usha (*) · P. Madhavan · M. V. R. Kumar

SRMIST, School of Computing, Kattanlathur, Kanchipuram District, India
e-mail: [email protected]

© Springer Nature Switzerland AG 2020 103

G. R. Kanagachidambaresan et al. (eds.), Internet of Things for Industry 4.0,
EAI/Springer Innovations in Communication and Computing,
https://doi.org/10.1007/978-3-030-32530-5_7
104 G. Usha et al.

Bio Inspired Algorithm

Clonal Negative Immune Deidrick

Selection Selection Network Cell Algo-
Algorithm Algorithm Algorithm rithm

Fig. 7.1 Types of bio-inspired algorithm

In a negative selection algorithm, clonal choice techniques are used to assess the
flexibility to wireless sensor network IDS plan. The resistant area of this work
attempts to recreate fragmentations of the hypothesis resistance techniques in a
wireless sensor network SN and that endeavors to determine an IDS scheme. This is
completed utilizing the properties of negative determination.
The negative determination is utilized in hypothesis techniques to show intru-
sions in sensor networks. Negative choice is commonly not favored in planning
intrusions for sensor networks, because of the quantity of finders it creates.
Nevertheless, this work uses highlights, for example, time span, time interim
between exchanges, information sent, and so on.
The negative selection algorithms with self and nonself mechanisms are used for
creating substances. To toggle it, danger theory can be used that rotates around risk
in dendritic cells when an interruption or an oddity is recognized.
The blue penciling techniques with NSA attempts to create various arbitrary
parallel strings. The strings are coordinated as self, and the remaining are named as
an identifier set. This blue penciling is utilized to coordinate the types in the system
when the identification stage is named as an irregularity cell.
Negative Selection Algorithm will undoubtedly have a huge number of contrast
data because it does not have a unified hazard gauging instrument and has a simi-
larly conflicting memory locator component. Similarly, negative determination is
not a thickly populated condition or a system with substantial traffic. The traffic
builds in a negative choice algorithm and the quantity of locator sets increments.
Since the indicator set is arbitrarily created, it builds the idleness and turns out to be
computationally invalid.
The negative determination calculation is not reasonable for thick situations, as
it will create the high false negatives. High false positives cannot adapt up to chang-
ing elements of the framework with respect to time. It cannot adapt up to evolving
elements, memory detectors, and faster response time,
Clonal selection algorithm is another type of algorithm that is additionally used
to provide safe framework. The clonal selection is fundamentally utilized to nega-
tive selection for the recognition system, which is dependent on intrinsic insuscep-
tible properties utilizing an unsupervised AI approach. This calculation depends on
distinguishing self and nonself designs, which is the center of negative ­determination.
The clonal selection can be utilized for negative determination along with clonal
7 A Novel Design Augmentation of Bio-Inspired Artificial Immune Technique… 105

choice administrator. The system conventions, for example, TCP and UDP are
grouped in the wake of doling out a quality worth and a lot of locator genotypes are
determined.
The clonal choice is typically utilized to a negative choice administrator or a
liking adding machine, which is utilized to ascertain the fondness between the
counteracting agent and the antigen. The locator set shaped from the negative
determination calculation is cloned, and a lot of indistinguishable clones are cre-
ated. These clones, in any case, are not by any stretch of the imagination indistin-
guishable. Hyper-transformation guarantees that the clones produced are of
different types of the parent clone; however, they are of a similar string size. This
has the impact of making the finder set bigger and increasingly complete. The
clone choice calculations are not appropriate for thick conditions, low false posi-
tives, and cannot adapt up to evolving elements, memory identifiers, and faster
reaction time,
The fundamental highlights of immune network algorithm are programmed
assurance of the populace estimate and blend of neighborhood with worldwide
hunt. An insusceptible calculation, named CLONALG, was created to perform pat-
tern acknowledgment and advancement. The creators exhibited exactly that this cal-
culation is fit for learning a lot of info designs by choosing, replicating, and changing
a lot of “counterfeit safe cells.” The creators demonstrated the reasonableness of the
calculation for multimodal search and introduced experimental outcomes where it
could beat a wellness sharing methodology.
The artificial immune system engaged with CLONALG is likewise found in a
developmental calculation, enabling it to be described as a transformative calcula-
tion motivated in the insusceptible framework. It is noted that there is a significant
calculated contrast between the clonal choice calculation and a transformative cal-
culation. In the previous, the hypothesis of advancement is utilized to clarify the
conduct of the framework, while in the last it motivated its improvement. Various
fascinating highlights of pick AINet are listed as follows.
1. It shows a deterministic and elitist determination system for each clone.
2. The cardinality of the populace is consequently controlled by the concealment.
3. The quantity of newcomers’ increments increase as the populace increments in
size. This is in such a case that the populace is consistently expanding in size,
which means that the issue has numerous nearby optima and the more optima it
finds, the more it is equipped for finding.
4. No encoding of the people of the populace is required.
There exist various contrasts between pick AINet and CLONALG. CLONALG
encodes the people of the populace utilizing parallel strings, whereas AINet depends
on genuine esteemed vectors. The select AINet incorporates the communication of
the system cells with the earth (wellness) and with one another (partiality), permit-
ting the dynamic control of the populace estimate. In the pick AINet case, newcom-
ers are just permitted to enter the populace after the present cells.
In CLONALG, the partiality proportionate transformation depends on a control
procedure where short blasts of high transformation rates are trailed by some
­breathing periods. Interestingly, AINet pursues a Gaussian transformation that is
106 G. Usha et al.

conversely relative to the standardized wellness of each parent cell. Select AINet
additionally introduces various likenesses with the advancement techniques (ES).
These algorithms are used to explain the mammalian adaptive immune system
behavior. Internet of Things (IoT) are used in various types of smart devices such as
smartphone and various types of sensing devices in order to transfer data locally and
to exchange various types of information [6]. Nowadays, the IoT technology is used
in mobile applications, entertainment, and other new context in order to realize the
better user interface, which provides cheaper, reliable, and robust communication
techniques. The important and most challenging concept in IoT is the design of
miniature things provided with security and privacy.
Security is a very important task for increasing the number of connected devices
in IoT. Every organization transfers the data faster than the content of the data.
Security plays a major role in data transfer between organizations. Data security and
privacy are very important issues in IoT to secure data. If a user needs to access
unauthorized data, it violates the security property of the complete network. In order
to use IoT in a broader range, several issues such as security and privacy are most
important aspects that need to be solved.
In order to provide authentication and access control, the security and privacy
control techniques need to be addressed. IoT applications need to be prevented from
security and privacy problems. The security solutions need to prevent illegitimate
access and should enable legitimate users to access resources in an authorized
manner [7].
While understanding security and privacy of the IoT application, we need to
address the intrusion detection and prevention systems. The main aim of the intru-
sion detection system is to distinguish between self and nonself patterns. Intrusion
detection techniques play major role in detecting and securing a network from mali-
cious activity [7]. The IoT technique faces various types of limitations in intrusion
detection technique [3]. The intrusion detection systems lack to provide standard-
ized protocol. The resources for computation are minimal. In order to overcome the
shortcomings, we provide an IoT-based design technique, which is used to detect
and prevent attacks in distributed environment with low computational and self-­
organization complexity. The proposed technique also provides a secured optimiza-
tion technique that provides security in IoT [8]. The organization of the work is
structured as follows. Section “Theoretical Background” describes the theoretical
background emphasizing negative selection and honeybee optimization algorithm
[5]. Section “Proposed System Architecture” describes the organization of the pro-
posed intrusion detection design technique [9]. Finally, this work is concluded with
the conclusion and with future work.

Theoretical Background

Various techniques are available to classify self/nonself classification methodolo-

gies. The techniques are based on statistical techniques or signature-based tech-
nique. The signature-based [3] technique needs predefined attack patterns to
7 A Novel Design Augmentation of Bio-Inspired Artificial Immune Technique… 107

properly identify attack. Additionally signature-based technique also requires large

amount of data. These types of signature-based attack techniques cannot detect new
type of attack patterns [1]. The statistical technique relies on the data with statistical
value. These techniques are not suitable for IoT intrusion detection technique [10].
In this work, we have proposed a dynamic optimized security technique to detect
and prevent attacks in IoT.
Neural Network–Based Negative Selection Algorithm
Much research work has been carried out in order to use neural networks in IoT
platform. In this approach, we use simple perceptron-based neural network tech-
nique, which uses supervised learning with binary classification [1]. This binary
classification algorithm correctly classifies the set into linear separable sets. The
input x is mapped to a single binary value. The input function f(x) is given as

1 if w. x + b > 0
f ( x) = 
0 otherwise (7.1)
“w” is known as weight vector, “b” is known as bias, and “w.x” is given in Eq. (7.2)

(7.2)
∑
m
i
wi ⋅ xi

Here “m” is number of inputs and “.” is known as the dot product. Similarly, a
large number of different classes are solved using feed-forward neural network.
This technique solves a number of neurons with different layers using the activation
function and the number of hidden layers. In order to simplify the proposed tech-
nique [1, 2], it is assumed that the input does not have any classification errors. In a
negative selection algorithm, it is inspired by self-nonself discrimination of mam-
malian immune systems [8].
This algorithm suggests the principle of self-nonself discrimination approach.
This technique is followed by discriminating anomaly and self-pattern detection
techniques. As discussed earlier, the input is classified as two sets. One is a normal
set denoted as N, and another set is anomaly data denoted as A. The complete data-
set is represented as C, where

C = NA (7.3)

Next, we analyzed the working of honeybee algorithm and the purpose of using
it in the proposed work [11].
In Eq. (7.3), the NA entity is known as normal and in order to classify NA as
attack or normal set, the following classification equation is used:

D = ( f ,M ) (7.4)

where “f” is known as classification function. “M” consists of multiple detector sets.
In order to further provide secured optimization technique, we use honeybee algo-
rithm to optimize.
108 G. Usha et al.

Honeybee Algorithm
The honey bee algorithm is an improvised technique for artificial immune systems
for intrusion detection. The honey bee must fly for long distances and run around to
different sources flowers so as to gather spore or nectar. In such a state of affairs, the
food supply has an additional variety of nectar. During a colony, the scout bees start
the method of hunting and search for the patch in a colony. The information contain-
ing food supply is not available for scout bees.
Toney honey bee calculation is primarily based on bunching structure teams. A
hub is chosen as cluster head obsessed on hub degree, neighbor’s conduct, skillful-
ness course, mutability speed, and remaining vitality. In addition, the planned strat-
egy evolved from the bumble bees and provides practiced and stable bunch
development.
The honey bee calculation is to be done for the process of collecting food infor-
mation. In this technique, each procedure has its terribly having their own qualities
and confinements. During this work, the honey bee calculation is done with the cal-
culation of to the grouping issue. The subsequent space controls the essential sys-
tem of the honeybee calculation.
The honey bee begins by flying for the rationale for scavenging food. The honey
bees gather food dirt or nectar from varied sources. On the off probability that the
sustenance supply contains a lot of nectar, a lot of honey bees can visit that site for
his or her assignment to assemble a lot of nectar. The scout honey bee begins
the method toward rummaging by haphazardly ransacking through the fix in an
exceedingly settlement. The scout has no earlier information of nourishment
sources.
The scout could be a type of honey bee that appears for nourishment source and
provides steering to operating drones. With that, the settlement announces a neigh-
borhood of its public as a scout honey bee for investigation throughout the gathering
season. The honey bee assesses various fix characteristics that supported the pres-
ence of sugar or nectar within the patch. The honey bee plays out a form of
move referred to as the “waggle move.” The presence of nectar and therefore
the heading of the sustenance supply area unit are then determined.
The bearing of sustenance is speculated smitten by the honey bee move therein spe-
cific heading. The speed of move demonstrates the presence of nectar within the sus-
tenance supply. Moderate move exhibits that the sustenance supply contains less
nectar and therefore the different method around. This move is very important for
correspondence between the state members. Through on these lines, the operat-
ing drones sent to various sustenance hotspots for nectar accumulation with
none guide, control, or another steering. The operating drones look out for the move
floor for the waggle move of scout honey bees.
The operating drones begin their journey smitten by the sort of move seen on the
move floor. On the off probability, if move is fast, a lot of operating drones area
unit sent there sustenance hotspot for nectar gathering. Also, the state accumulates a
lot of nectar in an exceedingly expert method. Succeeding waggle move is basic for
the scout, once coming back to the nectar. If there is still enough dirt within the sus-
tenance supply, the move is promoted, and therefore the manual laborer honey
7 A Novel Design Augmentation of Bio-Inspired Artificial Immune Technique… 109

bees area unit sent to the sustenance supply. The honey bee calculation has been
widely utilized within the writing for bunching in information to the mining.
The patch qualities’ area unit are determined by the quantity of sugar or nec-
tar within the patch. Whenever a food supply is found, the bees perform a
dance referred to as “waggle dance.” Based on the gap and the quantity of food sup-
ply, the direction of the food supply is calculated. Thus, the bee dance plays a
vital role in judging the gap of the food supply. The dance speed is employed to
understand the quantity of nectar within the food supply. Once the dance is slow, it
imposes that the supply has terribly less quantity of food supply. This dance plays a
major role in communicating between the members of the colony.
At the same time, the other bees (worker bees are selected for detecting different
food sources)collect the nectar from other food sources. This is achieved with the
help of the scout bees’ dance. As mentioned before, if the dance is faster, the worker
bees collect the food from the nectar. These types of algorithms are very useful in
path finding problems in data mining. In our technique, we use this honeybee algo-
rithm [11] to search for the secured optimized resource in IoT.
In honeybee calculation, mostly bunch honey bee calculation could be a multi-­
target sweetening procedure propelled from the looking conduct of nectar honey
bees. It is connected to varied streamlining problems, such as, the voyaging sales
rep issue, longest traditional subsequence, and framework chain duplication. In
our planned arrangement, the bumble bee calculation is utilized to isolate the sys-
tem into bunches. We expect numerous parameters during cluster development,
such as, hub movable, hub degree, hub remaining vitality, and neighbor quality.
Our planned arrangement begins with causing honeybees within the hunt area. In
HBAC (Honey bee calculation), the honey bees square is organized into three dis-
tinct gatherings: scout honey bees, footer honey bees, and utilized honey bees.
The worker honey bees handle investigation on the nourishment sources superior
to the previous one. The spectator honey bee appear out for the move floor for the
waggle move of the utilized honey bees. The spectator honey bee chooses the suste-
nance supply addicted to the type of move that utilized honey bee performs. The
sort of move demonstrates the heading of sustenance even as the presence of nec-
tar therein. The honey bees square measures the operating drones that play out the
nectar accumulation.
During this work, the looking conduct of honey bees calculations (HBAC) is
employed to settle on the choosing of cluster heads. Bunches square measure
formed addicted to the selected cluster heads. The task of CH is distributed to
totally different hubs that have additional nectar than the previous one.
The planned HBAC calculation works in two stages. The primary one is the bunch
arrangement and the second is cluster maintenance. In this stage, the honey bee
province is isolated into three sets; the principal 100% of the world is thought as
scout honey bees. The second forty-fifth is employed honey bees, whereas the
remaining forty-fifth is labeled as spectator honey bees. The scout honey bees speak
to all or any out bunches within the system. The hubs within the system square mea-
sure isolated into bunches utilizing HBAC calculation. The operating system of
arrangement stage is expressed in rule one.
110 G. Usha et al.

1. Initialize the preparation tests

2. Produce the opening populace Xi, i = 1, . . .,N utilizing condition (7.1)
3. Compute the wellness of the sustenance sources Q(Neci) utilizing condition (7.6)
4. Do (while set of utilized honey bee)
(i) Make new nourishment sources (arrangements) FSi utilizing condition (7.5)
(ii) Figure the worth Q(Neci)
(iii) Apply ravenous choice procedure
(iv) End
5. Figure the likelihood pi for the nourishment source Xi by condition (7.4)
6. Do (while set of spectator honey bee)
(i) Select an answer Xi relying upon pi
(ii) Produce sustenance sources (arrangements) FSi
(iii) Decide the worth Q(Neci)
(iv) Apply ravenous choice procedure
(v) End
7. On the off chance that there is a dismissed sustenance hotspot for the scout honey
bee
8. Substitute it with new sustenance source which will be unpredictable shaped by
condition (7.1)
Along these, spectator honeybees are equivalent number to utilize honey bees.
The calculation starts and focuses (called introductory CHs) in the search space.
The CHs are chosen arbitrarily and wellness is assessed. The wellness of a CH
depends on the measure of nectar. The CH having more nectar is a decent contender
for the CH job when contrasted with those having less nectar. The measure of nectar
in a hub relies upon a few parameters, Hub vitality (Hnode). In this parameter, the
hub having most extreme power is designated as a solid contender for CH. Also, the
hub having medium power is known as an ideal hub. The hub having a base power
is known as a frail hub.
Neighbor quality (Nnode) is the neighbor with one bounce separation and viewed
as a decent neighbor. The second neighbor that jumps away is viewed as the moder-
ate neighbor, and neighbors far away are viewed as an awful neighbor.
Hub degree (Anode) is the hubs having a degree over 10% are viewed as incred-
ible. A hub having a degree somewhere in the range of 5% and 10% is viewed as a
decent hub. The hub with a less degree is known as a terrible hub.
Hub portability (Pnode) is the hub with relative versatility and is known as
recom-repaired. Correspondingly, the hub with various versatilities and possessing
a similar course is known as incompletely prescribed. The hubs with various prob-
abilities and varying bearing are set apart as not suggested for CH.
The measure of nectar in a hub xi can be determined by

hi = H node + Nnode + Anode + Mnode (7.5)

7 A Novel Design Augmentation of Bio-Inspired Artificial Immune Technique… 111

where hi is that the hub to be investigated. The capability of a hub to show into the
CH is set (streamlined) in order that the geometer separation of every CH to alterna-
tive CH need to be round the same. The nectar sources are processed (optimized) in
order that the geometer separation of each nourishment space to alternative suste-
nance are – as need to be roughly the equivalent. Here, there is the difficulty of
gathering N specially appointed system hubs into a K range of noncovering bunches.
The problem of grouping network nodes into a K range of nonoverlapping clus-
ters is measured. The pursuit system of the scout, spectator, and used honey bees is
rehashed to deliver another people of the CHs (arrangements). The memory of the
used honey bees that stores CHs (solutions) have visual knowledge and also the
tests performed for checking the character of the new CHs (novel arrangements).
Q(Node)is extremely the quantity of nectar at hub “i.” The passer-by honey bees
watch the moving procedure of the used honey bee and disentangle to go to the CH
hub Xj with probability pi determined as

(7.6)
pi = Q ( Neci ) / ∑ ik=1 F ( Neck )
fs

FSi ( x + 1) = FSi ( x ) + aij

(7.7)
Here, when the measure of accessible nectar determined for novel CH is better
as thought about than the past, the honey bee memorizes the new nectar sum and
overlooks the past. If not, the area of the prior CH is kept up in her memory with no
modification.
The utilized honey bees share the amount of nectar of various CHs and their
directions with different honey bees on the moving floor when they come back to
the hive in the wake of finishing the pursuit procedure. At the point, when an on-­
looker honey bee watches the move of honey bees that are moving on the move
floor, she breaks down the nectar sum by watching the sort of move that utilized
honey bees perform.
The honey bees check the number of nectar in human positions. The new loca-
tion is preserved and overlooks this position place away in her memory (alter the
position cluster in their memory) subject to the wellbeing of the hopeful’s position.
The probable rationalization to the bunching issue is that the space of the nectar
positions and also the presence of nectar (its wellness) speak to its quality connected
with the conceivable arrangement and may be determined by the condition

Q ( Neci ) = (1) / (1 + cfi ) (7.8)

Once manufacturing of food, every candidate food location is compared with the
prevailing (old) food position. The previous one is preserved as long as the nectar
quantity of candidate food position is a smaller amount than the previous one which
is given by (7.9).

Y ( x ) = N 0 (Y ( x _ 1) )
(7.9)
112 G. Usha et al.

The choice between the previous and candidate answer is disbursed with the
greedy mechanism. The management parameters employed in this approach area
unit a most range of rounds and also the most CHs (clusters). The search method is
meant in such the way that each of the exploration and also the exploitation of honey
bees may be disbursed conjointly.
Cluster Maintenance and Communication
The cluster maintenance and communication continue to play a vital role in the
security issues. With that, the bunches area unit formed in cluster arrangement
stage, the subsequent stage is to primary teams if any adjustment in skilfulness and
vitality exhaustion of a hub happens. The bunches may be effectively primary sus-
tained, if the hubs and their neighbors area unit characterized.
The hub metal within the cluster that starts honey bees (short scrounging extent
honey bee) that may visit hubs within the system instating as of hub metal. The
quantity of hubs within the bunch is resembling its size. The set CH is in addition
characterized that stores the information of CHs after the effective bunch
development.

Proposed System Architecture

Data Preprocessing and Data Analysis

Initially the input is given which is classified as set of self patterns and set of nonself
patterns [12]. The number of hidden layers is calculated, the number of neurons in
each layer is also computed, and the rules are set for training. The abovementioned
algorithm is used to generate the detector technique [12]. For neural network train-
ing, the self patterns are recognized and for each pattern the corresponding bits are
matched. In this, the number of negative detector set [8] is calculated as the size of
the negative set. From this, the output is calculated as set of detectors, which are
capable of classifying nonself patterns.
The process of naïve detector [11] is that it generates the values randomly. After
that, the affinity value of naïve detector value is computed with each member in the
self-set. Based on the threshold if the naïve detector value does not match any ele-
ment in self-set that value is added to negative detector set. The abovementioned
steps are used with the help of single-layer feed-forward neural network [12]. This
method is simple, fast and used to classify the technique as self and nonself classi-
fication. In this step, finally we obtain the trained data.
Figure 7.2 represents the proposed system architecture in detail. Now we discuss
the design process in detail.
Data Optimization
After identifying the secured data, the information need to be traveled in optimized
manner. Hence, in order to optimize the proposed work, we use honeybee algorithm
[11]. In this technique, both local and global search techniques are used. After get-
ting the trained data, we are generating the initial population, which consists of size
7 A Novel Design Augmentation of Bio-Inspired Artificial Immune Technique… 113

Data Optimization
Data Pre-processing and Data
Analysis
Randomly Generate
initial population Evaluate the
Network Self-set and fitness value
Audit Negative set for the global
Data Data search
Generate Fitness
Evaluation
Determine
the hidden Sort the population Allocate
layers worker bees
in randomly
Local Search selected
patches
Compute no.of
neurons in each Select the best
layer locations from
neighbourhood

Apply the Determine the Define

rule to stop neighbourhood pattern
training size with
optimal
Evaluate the fitness solution
Trained Data value

Fig. 7.2 Proposed system architecture

n for worker bees [11]. The fitness evaluation function is known as the initial popu-
lation. The input values are calculated from maximum to minimum. First in optimi-
zation, local search technique is used. In this local search technique, the worker bees
calculate the neighborhood size based on the neighborhood search [11].
Finally, the fitness value is allotted based on best location. In the global search
technique, the evaluation value is calculated and the values are allotted randomly to
worker bees. In this way, the optimized path is selected. The abovementioned pro-
cess is used to transfer data from one node to another node in IoT in a secured man-
ner. The proposed technique is used to transfer data securely and in optimized way.

Conclusion

In this chapter, we proposed a novel design architecture to detect and prevent attacks
in IoT. The proposed design framework is used to detect and prevent attacks in
IoT. The proposed work consists of data preprocessing and data analysis and data
114 G. Usha et al.

optimization steps. In data preprocessing two algorithms are used. They are nega-
tive selection and neural network algorithm [8]. Negative selection algorithm is
used to detect attack patterns. Neural network algorithm is used to correctly classify
unseen patterns. Neural network technique is used to prevent attacks in IoT. Neural
network algorithm [8] is implemented in order to train the online data for IoT. Next,
honeybee algorithm [11] is used to detect optimized route solution. Honeybee
­algorithm is used to provide optimized solution when the network size is large.
This chapter provides a strong indication that for IoT environment, the proposed
work will outperform for both security and optimization issues.

References

1. Atzori, L., Iera, A., & Morabito, G. (2010). The Internet of Things: A survey. Computer
Network Journal, 54, 2787–2805.
2. Kawamoto, Y., et al. (2014). Internet of Things (IoT): Present Stateand future prospects. IEICE
Transactions on Information and Systems, E97-D(10), 2568–2575.
3. Hosseinpour, F., Bakar, K. A., Hardoroudi, A. H., & Kazazi, N. (2010, Nov). Survey on arti-
ficial immune system as a bio-inspired technique for anomaly based intrusion detection sys-
tems. In Intelligent networking and collaborative systems (INCOS), 2010 2nd international
conference on (pp. 323–324). IEEE.
4. Stout, W., & Urias, V. (2016). Challenges to securing the Internet of Things. In ICCST.
5. Jing Liu, Yang Xiao., & Philip Chen, C. L. (2012). Authentication and access control in the
Internet of Things. 32nd international conference on distributed computing systems workshops.
6. Kolias, C., Kambourakis, G., & Maragoudakis, M. (2011). Swarm intelligence in intrusion
detection: A survey. Computers & Security, 30(8), 625–642.
7. Alaparthy, V. T., & Morgera, S. D. (2018). A multi-level intrusion detection system for wire-
less sensor networks based on immune theory. IEEE Access, 6, 47364–47373.
8. Pamukov, M. E., Poulkov, V. K., & Shterev, V. A. (2019). Negative selection and neuralnet-­
work based algorithm for intrusion detection in IoT. In Proceedings of the 2018 41st interna-
tional conference on Telecommnications and Siganl Processing (tsp) (pp. 1–5). IEEE.
9. Shi, Y., Li, T., Li, R., Peng, X., & Tang, P. (2017). An immunity-based IOT environment secu-
rity situation awareness mode. Journal of Computer and Communications, 5, 182–197.
10. Bitam, S., Zeadally, S., & Mellouk, A. (2016). Bio-inspired cybersecurity for wireless sensor
networks. IEEE Communications Magazine, 54(6), 68–74.
11. Ahmad, M., Ikram, A. A., Leela, R., Masood, I., & Ulla, R. (2017). Honey bee algorithm-based
efficient cluster formation and algorithm scheme in mobile ad hoc networks. International
Journal of Distributed Sensor Networks, 13(6), 13–18.
12. Amouri, A., Alaparthy, V. T., & Morgera, S. D. Cross layer-based intrusion detection based on
network behavior for IoT. In Wireless and microwave technology conference.
Chapter 8
Role of AI and Bio-Inspired Computing
in Decision Making

Surekha Paneerselvam

Introduction

The basic theory of a human process is to establish an interaction with the outside
world. This interaction involves decisions, which in turn lead to an argument of
whether the decisions made are good or bad. Finally, people arrive at a conclusion,
perceiving that they have made a good decision. But, in the process of decision mak-
ing, they definitely seek help or support from someone to conclude whether the
decision made is good. The supportive person tries to conceive the decisions and
group them as structured, unstructured, and semi-structured. Structured decision
problems lead to an optimal solution, for example, the decision made in optimizing
the cost of travel between two cities. Unstructured decision problems do not have
any predicted solution, for example, to make a choice between two equivalently
good job offers or to make a decision to choose a life partner. Semi-structured deci-
sion problems definitely need a support system to develop alternate solutions and
then pick the best out of them. Such support is provided with artificial intelligence
(AI) techniques, and the unstructured decision-making process is simply referred as
Intelligent Support Systems for Decision Making (ISSDM).
To a wider extent, AI mimics the process of human decision making, and this has
also been proved by advanced AI researchers in several real-time situations. The
human decision-making process is summarized by Pomerol and Adam [1], who
claim that reasoning and recognition are the key parameters in decision making.
They also claim that ‘good’ decisions are characterized by reasoning, which involves
the process of weighing the alternatives and choosing the best decision. Most of the
reasoning-based decision-making problems can be implemented using analytical
techniques, and as a result, they can be implanted into ISSDM. On the other hand,

S. Paneerselvam (*)
Amrita School of Engineering, Amrita Vishwa Vidyapeetham, Bengaluru, India
e-mail: [email protected]

© Springer Nature Switzerland AG 2020 115

G. R. Kanagachidambaresan et al. (eds.), Internet of Things for Industry 4.0,
EAI/Springer Innovations in Communication and Computing,
https://doi.org/10.1007/978-3-030-32530-5_8
116 S. Paneerselvam

recognition is the process which makes a decision without spotting out a proper
reasoning. Such kind of recognition-based decision making has been implemented
in emergency response situations such as fire services, pattern recognition applica-
tions, and authentication based on fingerprint and face, and in situations where
immediate response is required. Decision making affects the neural system emo-
tionally, and research has demonstrated the capability of modelling the emotional
characteristics with decision making.
Decision support systems (DSS) were used since the early 1970s in order to
assist several industrial activities in decision making. They were designed to pro-
vide an organized set of tools for the concerned decision-making situation and to
improve the outcome for a chosen application [2]. A set of contradictory decisions
are made for the problem under consideration, and the DSS chooses the best deci-
sion from the set of decisions.
A decision support system (DSS) is a system under the control of one or more
decision-makers that assists in the activity of decision making by providing an orga-
nized set of tools intended to impart structure to portions of the decision-making situ-
ation and to improve the ultimate effectiveness of the decision outcome [2]. A
decision support system (DSS) is a system which lets one or more people to make
decision/s in a concrete domain, in order to manage it the best way, selecting at each
time the best alternative among a set of alternatives, which generally are contradic-
tory. Over the years, DSS evolved as a model-based set of procedures used to process
the data for a given problem and to make suitable judgements. These model-based
DSS were used to assist a human being in making decisions [3]. Later, the intellectual
capabilities of human beings were combined with DSS using personal computers
with a motive to improve the quality of the decisions made [4]. These computer-
based support systems were later extended with a variety of possible decisions so that
the best decision could be chosen for the problem under consideration. The DSS
helped decision-makers to frame data and models based on an entity to solve struc-
tured as well as unstructured problems [5]. The basic feature of DSS lies in the model
component for which quantitative techniques such as simulation, optimization, and
statistics are used to represent a decision model [6, 7].
During the 1990s, DSSs were applied to perform tasks such as data analysis,
monitoring and control, planning, prediction, recording data, and retrieval of data.
To carry out these tasks, decision models such as optimization models based on
linear programming, control algorithm models, statistical models based on regres-
sion, and simulation models were developed to assist humans in decision making.
The intelligent DSS were later evolved whose models were based on intelligent
algorithms. These decision-making models were more commonly used due to their
faster decision-making ability, consistency in the decisions made, and improved
quality of decisions [8]. The computer-based decision-making systems incorporate
an explicit decision procedure based on various theoretical principles to make deci-
sions in an intelligent way [9]. The process of decision making is affected emotion-
ally in the neural system either consciously and unconsciously. Research carried out
during recent decades in Intelligent Support Systems for Decision Making (ISSDM)
has shown the ability to handle emotion through decision making [10]. ISSDM use
several artificial intelligence techniques to assist in the decision-making process.
8 Role of AI and Bio-Inspired Computing in Decision Making 117

These tools such as the artificial neural networks, fuzzy logic, and bio-­inspired algo-
rithms have been serving as an influential means to solve real-world problems that
involve large amount of data. Some of the capabilities of these intelligent algorithms
which have made them intelligent decision-makers are as follows:
–– Understanding the situation and making sense out of the uncertainty or
ambiguity
–– Learn through experience
–– React in a timely manner to a new situation (adaptive)
–– Handling perplexing solutions
–– Use knowledge to recognize various factors in a decision
ISSDM have been growing and emerging as useful tools for several applications
by employing a variety of AI techniques. The basic concepts of AI techniques and
their role in decision making are discussed in the following section.

ANN for Intelligent Decision Making

Artificial neural networks (ANN) are a powerful group of learning algorithms

inspired from human nervous system. They are mainly used to approximate or esti-
mate functions which have a huge dependency on the number of inputs they handle.
ANN structure consists of interconnected neurons arranged in three layers, namely,
the input, output, and hidden layer. The activation functions govern the manner in
which the signals traverse between the layers through neurons. A weight is assigned
to the connection made between neurons. Initially, the weights are assigned in ran-
dom. The basic structure of a neural network is shown in Fig. 8.1. The input layer
consists of neurons that process the inputs obtained from the real world and the
output layer presents the results to the real world. The hidden layer lying between
the input layer and the output layer consists of neurons that process information
internally. The signals from the input layer are weighted and passed on to the hidden
layer and in turn the weighted information is traversed to the output layer. This
simple form of information processing network is referred to as feed-forward net-
work. Some networks send the output back to the hidden layer for further process-
ing, known as the feedback network. Once the architecture of the neural network is
formed, the network is prepared for the training and testing phases.
There are different ANN algorithms based on the learning concept, which are
broadly classified into two categories, supervised learning and unsupervised learn-
ing. In supervised learning, the performance of the network is graded manually and
the desired output is predicted, which in turn is presented to the network. In the
unsupervised class of learning, the network is trained with updation of weights and
tries to obtain the output based on the input information. This happens over a large
number of iterations governed by the learning rules. The learning rules are
­administered by a control factor known as the learning rate. Slower learning rate
results in a large convergence time, while faster learning rate results in inappropriate
output. Hence a better choice has to be made as far as the learning rate is concerned.
118 S. Paneerselvam

Feedback

Inputs Outputs

Competition
(or inhibition)

Feedback

Fig. 8.1 Basic ANN model

The learning process is based on several learning laws, to name a few, Hopfield law,
Hebb’s rule, delta rule, extended delta rule, competitive learning rule, outstar learn-
ing rule, memory-based learning law, and Boltzmann learning law.
Major application areas of ANN are (a) prediction, (b) classification, (c) data
association, and (d) data conceptualization. The perceptron network, back propaga-
tion network, and directed random search network belong to the group of prediction
networks. Networks used for classification are learning vector quantization, counter-­
propagation network, and probabilistic neural networks, while Hopfield network,
Boltzmann machine, Hamming network, and bi-directional associative memory are
applied to data association–related problems. For data conceptualization, adaptive
resonance network and self-organizing map have been more suitable networks.
In decision making, ANNs have been playing a vital role in making comprehen-
sive decisions. This involves the need to make decisions through evolution and as a
result consumes a lot of time. The evolution can be short term based on the course
of an assignment, which has been influenced through feedback from the decisions
obtained from the previous stages of the assignment. The evolution can also be long
term, which evolves the transition over a course of life time. Whether the evolution
is short term or long term, ANN should be capable of changing the current prefer-
ence decisions based on the feedback obtained from the previous evolved decisions.

Fuzzy for Intelligent Decision Making

Fuzzy logic (FL) was proposed by Lotfi A. Zadeh in 1965 [11]. Fuzzy logic (FL)
provides decision support based on the way humans reason them out. Fuzzy logic is
used to represent uncertainty by assigning the input to a range of values between 0
8 Role of AI and Bio-Inspired Computing in Decision Making 119

and 1. The binary 0 and 1 refer to completely false and completely true, respec-
tively. The range of values between the extremities is usually formulated using
mathematic tools and later processed by computers, to achieve human-like thinking.
The gap between qualitative and quantitative modelling can be bridged by applying
FL. The input-output mapping of a FL model is quantitative, but internally the sys-
tem is governed by a set of linguistic rules which are qualitative.
The fuzzy set is formed with elements in the universe, which contains all the
possible elements for a related problem. The universe is defined by a membership
function μA(x) where x is the number of elements in a base set. A variety of member-
ship functions are defined for various applications, namely, triangular function,
trapezoidal function, Γ-function, S-function, exponential function, and Gaussian
function.
The process of obtaining the mapping between input and output using fuzzy
logic is known as fuzzy inference. Decisions are made within the mapping based on
rules. Fuzzy inference involves the processes namely fuzzification and defuzzifica-
tion. The linking of input variables to output variables through expert knowledge
and experience is based on the fuzzy rules. The rules fall under two basic categories
such as Mamdani fuzzy rules and Takagi-Sugeno fuzzy rules. The inference system
simply referred as a fuzzy expert system consists of modules such as fuzzifier, the
inference engine, defuzzifier, and a fuzzy rule base.
The crisp inputs values are fuzzified and further sent to the rule base module. The
rules are constructed using conditional IF-THEN statements which are used to con-
trol the output variable. The rules are applied to the fuzzy values, and the outputs are
evaluated as fuzzy values. To send the fuzzy values to the real world, they are
defuzzified and converted back to equivalent crisp values.
FL has been identified as suitable tools for decision making due to their capabili-
ties such as:
–– More choices to define uncertainty
–– Consistency and redundancy of rule base and rules
–– Addressing nonlinear issues also due to universal approximation capabilities
–– Exact solutions can be obtained with rough approximation for a problem
–– Simple models for a problem whose mathematical model is complex

Bio-Inspired Algorithms

Bio-inspired algorithms are a growing group of algorithms capable of searching the

optimal solution from a search space at a quicker rate for a given optimization prob-
lem. Some of the existing conventional search algorithms follow a systematic pro-
cedure to obtain the optimal solution from the search space, and this takes a longer
time for convergence. The disadvantages of conventional optimization algorithms
are overcome by bio-inspired algorithms, in which biological behaviour of certain
species are analysed to perform the optimal search. Hence the group of algorithms
120 S. Paneerselvam

are also known as nature-inspired algorithms. These algorithms are framed based on
the evolution resulting from interaction between species and interaction within the
species in nature. The interaction between/within the species could be cooperative
or competitive. Bio-inspired algorithms possess the following capabilities:
• Applicable to wide range of problems
• Few control parameters to tune the algorithm
• Better convergence rate while reaching the optimum value
Until recent years, ant colony optimization and particle swarm optimization
inspired by colony of ants and swarm of bees were the basic bio-inspired algorithms
applicable to solve constraint-based optimization problems. Later, behaviour of
bacteria, fireflies, fruit flies, bats, and cuckoo birds were analysed to frame a wider
class of bio-inspired algorithms. In this section, the variants of bio-inspired algo-
rithms and their role in decision making is discussed.

Genetic Algorithms

Genetic algorithms (GA) were introduced as an optimization tool by John H. Holland

in the early 1960s [12]. GA works by selecting individuals from a population of
permissible solutions and then applying genetic operators, namely, crossover and
mutation on them to choose the fitter individuals. The group of fitter individuals
progresses to the next generation, and the evolution process continues iteratively
until the convergence criteria is met. At the end of each iteration, the old population
is rejected and the iteration continues using the new population. The basic pseudo-
code of a GA is illustrated as:

Initialize the population of individuals

Evaluate the population through the fitness function
While no best solution do
For every individual in the population
Generate new population from old population
Evaluate its fitness
Selection
Crossover
Mutation
End For
End While

Genetic algorithms have been used to solve a wide range of problems, namely,
maximization/minimization problems, multi-objective optimization, and constraint-­
based optimization. Some of the industrial applications where GA has been suc-
cessful are the job shop scheduling problem, vehicle routing problem, network
routing, and load dispatch problem. The major challenges involved while solving
8 Role of AI and Bio-Inspired Computing in Decision Making 121

problems using GA was in framing the fitness function, to understand the fitness of
the solutions obtained through evolution, especially when the search space is large.
This leads to scaling the solution space and thus leading to increase in the algorith-
mic complexity.
Thus GAs can be either applied to single or multi-objective problems whose fit-
ness function can be evaluated with ease in a well-defined solution space. When the
population size is increased for high dimension problems, then scaling is required
as a result of which the algorithmic efficiency decreases.
The probable areas where GA can be applied to obtain convincing results are
listed as follows
–– Maximization/minimization problems
–– Searching and sorting
–– Job allocation (job shop scheduling)
–– Parallel computation
–– Routing, classification, coordination

Particle Swarm Optimization

Particle swarm optimization (PSO) developed by Dr. Eberhart and Dr. Kennedy in
1995 is a population-based stochastic optimization technique, inspired by collective
social behaviour of fish schooling, bird flocking, or insect swarming [13]. A swarm
of homogenous agents interact among themselves in the population, and they
undergo the process of evolution to obtain the optimal solution. Unlike GA, PSO
does not perform crossover and mutation during the evolution process. But when
compared with GA, the merits of PSO are that the implementation process in PSO
is simpler since there are a few parameters to adjust.
The PSO algorithm is initialized with a set of random particles which are the
probable solutions to the defined problem. The algorithm then searches for the local
optima by updating the successive generations. At the end of each iteration, every
particle is updated based on two best values. One known as the pbest, this is the best
solution that has been identified so far based on the fitness. The second value known
as the gbest is the best value obtained so far in the population. PSO also defines
another value known as the lbest, which refers to the best value in the population’s
topological neighbours. Upon finding the pbest and the gbest values, the particle
updates its velocity and position using Eq. (8.1) and (8.2).

v[ ] = v [ ] + c1 ∗ rand ( ) ∗ ( pbest [ ] − present [ ]) + c2 ∗ rand ( )

∗ ( gbest [ ] − present [ ]) (8.1)

present [ ] = present [ ] + v [ ] (8.2)

122 S. Paneerselvam

v[] is the particle velocity, present[] is the current particle (solution). pbest[] and
gbest[] are defined as stated before. rand () is a random number between (0,1). c1,
c2 are learning factors. Usually, c1 = c2 = 2.
The pseudo-code of the procedure is as follows:

Initialize swarm of particles (population)

While number of iterations is not maximum Do
For each particle
Calculate fitness value
If the fitness value is better than the best fitness value
(pBest) in history set current value as the new pBest
End If
Choose the particle with the best fitness value of all the
particles as the gBest
For each particle
Calculate particle velocity according equation (a)
Update particle position according equation (b)
End For
End While

To summarize, PSO is a population-based multidimensional multi-objective

problem, with all the potential candidate solutions to the problem in a hyper-plane.
When compared to GA, there is a possibility of highly scaling the problem space in
case of complex problems.
The thrust areas where PSO is applied and proved to be performing better as an
optimization solution are as follows:
–– Job shop scheduling problems
–– Adaptive learning problems
–– Minimization/maximization problems
–– Chaotic and oscillatory systems
–– Searching and thresholding
–– Location identification
–– Resource management in a distributed environment

Ant Colony Optimization

Ant colony optimization (ACO) proposed by Dorigo in 1997 [14] is a popular opti-
mization algorithm which was biologically inspired based on the behaviour of ants.
ACO is a probabilistic technique which can be applied to real-world problems to
find better paths through graphs. The search process is based on the trailing behav-
iour of ants between their colony and source of food thus establishing a path. The
agents (artificial ants) in the algorithm locate the optimal solutions by moving
through a population consisting of all possible solutions. During their movement,
8 Role of AI and Bio-Inspired Computing in Decision Making 123

the agents record their positions and the quality of the food, analogous to the phero-
mone trailing in natural ants. This process of recording enables other agents in the
population to identify better solutions.
The potential solutions are identified for a problem from a pool of possible solu-
tions through probabilistic distribution. As the agents explore the possible solutions,
each agent is updated locally. Further these possible solutions update the values of
the trails such that the better quality solutions are selected in the subsequent evolu-
tions. The process continues until an optimal solution is identified. It has been
proved from several implementations that the algorithm completes the search pro-
cess in a reasonable time.
The ACO algorithm works as follows:

Initialize the population with ants and pheromone values

While termination condition not reached do
For every ant in the population
Position ant on a starting node;
For step size
Obtain current state based on probabilistic tran-
sition rules
End For
Update the intensity of pheromone trail
Compare and identify the best solution
End For
End While

The advantage of ACO lies in its convergence time mainly due to its property of
inherent parallelism. ACO is suitable for solving dynamic optimization problems,
continuous optimization problems, and NP-hard combinatorial problems. The prop-
erty of distributed computation in ACO avoids premature convergence, hence the
reason for faster convergence. The performance of the ACO algorithm reduces
when the problem dimension increases.
ACO has been applied to a variety of real-world problems such as the following:
–– Travelling salesman problem
–– Network analysis and clustering
–– Scheduling and routing
–– Quadratic assignment problem
–– Shortest path problem
–– Economic dispatch problems
–– Vehicle routing problem
–– Gaming theory
–– Graph colouring and set covering
–– Forecasting
–– Agent-based dynamic scheduling
–– Parameter estimation
124 S. Paneerselvam

Artificial Bee Colony

The artificial bee colony (ABC) algorithm defined by Karaboga is a bio-inspired

algorithm based on the foraging behaviour of honey bees [15]. Honey bees select
site for hive, communicate, allot tasks among themselves, reproduce, forage, navi-
gate, and lay pheromone. These capabilities of honey bees have been mimicked in
the ABC algorithm. The colony of artificial bees is organized into three groups,
namely, employed bees, onlookers, and scouts. The employed bees occupy one half
of the colony, while the onlooker bees occupy the other half. One employed bee is
allotted for each food source. In ABC algorithm, potential food sources are identi-
fied from the population initialized by the bees. Once the candidate solution (food
source) is identified, then the validity (fitness) of the candidate solution is evaluated.
In the successive stage, if a candidate solution of better fitness is discovered by the
employed bees, then the existing candidate solution is discarded. Employed bees
share all the information with the onlooker bees, and they try to improve the fitness
of the discarded candidate solutions. If the onlooker bees are also not able to improve
the fitness, then the candidate solutions are completely rejected from the population.
The rejected solution is taken care of by the scout and replaces the population with
another probable candidate solution. The process continues until the candidate solu-
tions with better fitness are evolved. The pseudo-code of the ABC algorithm is as
follows:

Initialize the population of candidate solutions

Evaluate the population of candidate solutions
While best solutions are not found so far do
Produce new solutions for the employed bees to evaluate
Apply selection based on Deb’s method
Calculate the probability values for the solutions
Produce the new solutions for the onlookers from the
solutions
Apply selection process based on Deb’s method
Determine the abandoned solution for the scout and replace
population
Store the best solution achieved so far
End While

A wide range of studies have been carried out in the ABC algorithm over the
years, based on their dance, communication, queen’s behaviour, mating behaviour,
navigation behaviour, and foraging behaviour. The ABC algorithms have been
applied to constraint-based and unconstrained-based optimization problems in com-
bination with differential evolution. The major application areas of ABC algorithm
in the field of decision making are as follows:
8 Role of AI and Bio-Inspired Computing in Decision Making 125

–– Numerical function optimization

–– Network routing and allocation
–– Benchmarking problems
–– Searching
–– Allocation and assignment
–– Probability distribution

Fish Swarm Algorithm

Li et al. proposed the fish swarm algorithm (FSA) in 2002 [16]. The FSA is a
population-­based intelligent bio-inspired technique that simulates the schooling
behaviour of fish under water. Usually, fish move close to each other in a group
called swarm in order to search for food, protect themselves from predators, and in
turn avoid collisions within the swarm. The behaviour in a swarm is characterized
by leaping, chasing, swarming, and searching. While applying the swarm algorithm
based on this behaviour to solve optimization problems, an individual fish within
the swarm is considered as a point and is called the candidate solution. Collectively
a population is formed which consists of several such candidate solutions or points.
The environment in which the artificial fish moves forms the search space, in which
the optimum solution is searched for.
Experiments have proved that the FSA has a capability of finding the global
optimization solution without evaluating the local minimum. FSA has a strong
ability to achieve global optimization by avoiding local minimum. The population
is formed based on the position of the fish in a multidimensional search space.
The behaviour of the fish during the food search is weighted by a term known as
the food satisfaction factor. The distance between two fish in the search space is
characterized by the Euclidean distance. The fish in the swarm searches for the
food in a random manner in the search space moving in a direction based on high
food concentration. The objective of the optimization technique is to minimize the
food satisfaction factor. Within the population, fish establishes a swarming behav-
iour to satisfy the food intake and during the process, they tend to attract new
swarm members.
Once a fish identifies a food location, the individuals in its neighbourhood fol-
low. Within the visual region of the fish, some fish in the population will be able to
find more amount of food when compared to the others. These fish will obviously
follow the best one to increase the food satisfaction percentage. The basic parame-
ters involved in FSA are the visual distance, maximum step length, and a crowd
factor. The algorithm efficiency is mostly dominated by the two parameters, namely,
visual distance and maximum step length.
The pseudo-code of the fish swarm algorithm is as follows:
126 S. Paneerselvam

Initialize the swarm of fish in a random manner.

While stop condition is reached do
for each fish
Evaluate the fitness
do step Follow (based on the food search behaviour)
if (Follow has failed) then do step Swarm
if (Swarm has failed) then do step Prey
end
end
end for
end while

To summarize, FSA is a population-based multi-objective problem, with the

ability to solve complex nonlinear high-dimensional problems and with all the
potential candidate solutions to the problem in a hyper-plane. The algorithm is capa-
ble of converging faster since it requires very few parameters to be tuned for opti-
mization. Moreover, FSA does not possess the crossover and mutation processes
used in GA, so the algorithm converges faster.
The common areas where FSA is applied and proved to be performing better as
an optimization solution are as follows:
–– Function optimization
–– Parameter estimation
–– Combinatorial optimization
–– Least squares support vector machine
–– Geotechnical engineering problems

Firefly Algorithm

Firefly algorithm (FA) proposed by Yang [17] was inspired by the flashing behav-
iour of fireflies and were used to address the non-convex objective functions in
NP-hard problems. These problems have equality- and inequality-based constraints,
and the firefly algorithm has been proved by providing suitable optimal solutions to
these problems. The algorithm employs a population-based search with agents that
are analogous to fireflies. The algorithm uses a learning mechanism that facilitates
good tuning of parameters thus balancing the exploration and exploitation. The
algorithm iterates with numerous agents that communicate with each other. The
bioluminescent glowing lights are used as a media through which the information is
shared. These glowing lights enable the fireflies to explore the search space more
effectively when compared to the distributed random search. Each fly is attracted by
the brighter firefly within its neighbourhood. The optimization algorithm has the
following rules based on the biological behaviour:
8 Role of AI and Bio-Inspired Computing in Decision Making 127

• The fireflies are attracted towards brighter ones irrespective of their sex.
• The degree of attractiveness of a firefly and brightness of a firefly are propor-
tional to each other.
• The brightness or light intensity of a firefly is determined by the value of the
objective function of a given problem.
The pseudo-code of the firefly algorithm is as follows:

Initialize population of fireflies

Determine the Light Intensity based on the fitness function
Define a light absorption coefficient
While (maximum numbers of Generations have not reached) do
For each firefly
Movement of flies towards attractive flies
Attractiveness varies with distance according
to the light absorption coefficient
Evaluate new solutions and update light intensity
End For
Determine the current best based on rank
End While

The behaviour based on attraction towards brightness and the ability of dealing
with multimodality has made the firefly algorithm to deal with multimodal func-
tions to solve NP-hard optimization problems.
Some of the potential areas in which the firefly algorithm has been applied suc-
cessfully are as follows:
–– Multi-objective optimization
–– Dispatch problems
–– Digital image compression
–– Feature selection
–– Chaotic problems
–– Nonlinear problems – structural optimization
–– Multimodal design problems
–– Forecasting problems – stock exchange
–– Antenna design optimization
–– Nonlinear optimization
–– Load dispatch problems
–– NP-hard scheduling problems
–– Dynamic environment problems
–– Classifications and clustering
–– Training ANN
128 S. Paneerselvam

Bacterial Foraging Algorithm

Bacteria foraging optimization (BFO) algorithm proposed by Passino in 2002 [18]

is a stochastic global search technique inspired biologically by the foraging behav-
iour of E. coli. The algorithm is based on the biological behaviour in which the
organisms with less capabilities are eliminated and the others survive through the
process of natural selection. The process involves three stages, namely, locating the
food, handling the population, and consuming the food. Such foraging strategy hap-
pens at an individual level or at a group level. During the food search (evolutionary)
process, the bacteria try to maximize their energy consumption per unit time spent
in search of food. All bacteria in the population establish communication among
themselves through signals. The foraging decisions are based on the energy con-
sumption and the communication through signals. The agents are the bacterium,
and they are governed by two actions: tumbling and swimming.
The tumble action involves the movement of the agents in a random direction in
search of an optimal solution, while in the swimming action the bacterium move
around in the population by taking small steps in search of food. The process con-
tinues, and as a result of reproduction during the chemotactic movement, the best
agents are retained in the population, while the inefficient ones are eliminated. The
best agents in the population are those which have better objective function and
maximum energy per unit time. The algorithm uses an elimination-dispersal opera-
tor which comprehends the environmental changes due to the elimination of agents.
Though the algorithm is simple and easy to implement, it suffers from poor con-
vergence rates. The working of the BFO algorithm is as follows:

Initialize population
Evaluate the fitness
While number of iterations not reached do
For every chemotactic process
Evaluate the fitness function
Perform tumbling and swimming
End For
While chemotaxis operation do
Evolution through Reproduction
End While
Elimination Dispersal operator to retain the best
individuals
End For
End While

BFO has been found application to a variety of real-world problems due to its
excellent global searching capability and proved its effectiveness over variants of
GA and PSO. On the other hand, BFO has also failed with poor convergence crite-
ria. The algorithm is also sensitive to the step size value required for the tumbling,
8 Role of AI and Bio-Inspired Computing in Decision Making 129

and hence it is difficult to generalize an explicit constraint-handling mechanism for

the algorithm. Certain improvements in the mathematical modelling and adaptation
strategies of the BFO algorithm might make the algorithm more efficient.
Some of the application areas where BFO has been successful are as follows:
–– Multi-objective function optimization
–– Load forecasting and compensation
–– Harmonic analysis in power systems
–– Optimal tuning of PID controller
–– Load dispatch and unit commitment
–– Machine learning
–– Maximization and minimization
–– Job shop scheduling

Cuckoo Search Optimization

The cuckoo search optimization (CSO) algorithm for single and multi-objective
nonlinear optimization problems was proposed by Yang and Deb in 2002. The algo-
rithm draws its inspiration from the biological breeding behaviour of cuckoo birds
which they lay their eggs in the nest of host birds. They remove the eggs of the host
bird and try to increase the hatching probability of the cuckoo eggs. The individuals
for evolution in this search algorithm are generated through Levy flight mechanism,
which is a special class of random walk with irregular step lengths based on proba-
bility distribution. There are three types of brood parasitism adopted by the cuckoos
namely the intraspecific, cooperative, and nest takeover mechanisms. In intraspe-
cific brood parasitism, the cuckoos lay eggs in the host bird’s (same species) nest
and later does not care for the eggs [19]. Cooperative breeding refers to the pairing
of two or more females with the same male, and they lay their eggs in the same nest
in a cooperative manner and are mutually cooperative in providing parental care
[20]. The nest takeover [21] breed mechanism refers to a cuckoo occupying another
host birds’ nest to lay its eggs.
The cuckoo search algorithm is based on three basic rules:
(i) Each cuckoo lays an egg one at a time and puts it in nest chosen at random.
(ii) The nests with good quality will progress to the next generation.
(iii) The host bird in the host nest is capable of discovering a foreign egg based on
probability.
The Levy flight mechanism provides the random walk in which the step length is
obtained based on Levy distribution. The new solutions are generated through the
random walk process. The movement is always around the best solution obtained so
far; hence, the local search is always faster. Some fraction of new solutions are gener-
ated by far-field randomization, so that they are lying far from the current best solu-
tion. This enables the algorithm from avoiding getting trapped in the local optimum.
130 S. Paneerselvam

Based on the biological behaviour of the cuckoo birds and the above-mentioned
rules, the algorithm for finding an optimal solution is framed as follows:

Initialize a population of host nests

While (stopping condition not reached) do
Evaluate the fitness of a randomly chosen cuckoo
Select a nest from the population
if(fitness of cuckoo is higher than the fitness of the nest)
{
Replace the fitness of the nest with the fitness of
the cuckoo
}
Discard the worse nests and build new ones
Retain the best nests (solutions);
Rank the nests
Obtain the current best
End While

The application of Levy flights in CSO has proved to be an efficient approach to

solve optimization problems. The random walk property of Levy flight mechanism
helps the CSO algorithm to converge much faster, thus improving the algorithmic
efficiency. In order to obtain the optimal solution, a large number of iterations are
required, which is one of the drawbacks of the CSO algorithm. This increased
amount of iterations is obtained when the discovery rate is small and the Levy step
size is high. On the other hand, when the discovery rate is high and the Levy step
size is small, the algorithm converges faster, but the best solution may not be
obtained. Hence, a trade-off has to be maintained between the discovery rate and the
Levy step size.
The CSO algorithm has been successful in the following problem areas:
–– Optimization of ANN parameters
–– Gradient-based optimization
–– Multi-objective scheduling and allocation
–– Phase equilibrium problems
–– Optimization in cluster centres
–– Reliability optimization
–– Economic load dispatch
–– Path identification for network analysis

Fruit Fly Algorithm

Fruit fly optimization algorithm (FFA) is an evolutionary algorithm introduced by

Pan [22]. The algorithm is inspired by the biological behaviour of fruit flies. The
fruit flies have excellent smell and sight senses, using which they find their food.
8 Role of AI and Bio-Inspired Computing in Decision Making 131

It has been proved in biological sciences that these group of flies can smell food
even at a distance of 40 km. The foraging behaviour in search of food is performed
in two stages. During the first stage, the flies identify the food through their smell
(osphresis organ), and in the second stage they get near the food through their sensi-
tive vision. Usually, the smell phase is shorter than the vision phase.
The procedure of the FFA is summarized as follows:

Initialize the fruit fly swarm in a random manner

While (maximum number of iterations not reached) do
Each individual randomly searches the food source and distance
is obtained
Smell concentration value is calculated for every fruit fly
Identify the location with best concentration value
Position the swarm for successive iterations
End While

The fruit fly algorithm is capable of finding the global optimum at a faster rate
with improved accuracy due to the smell concentration parameter. Due to this
parameter, the algorithm converges without falling into the local minima, hence
improving the robustness. The update strategy is very simple in the fruit fly algo-
rithm, but it is difficult to define a standard update strategy during the initialization
process.
The algorithm is still growing and finding its roots in several applications. Some
of the areas where the algorithm has found scope are as follows:
–– Military applications
–– Medicine
–– Management and finance
–– In combination with data mining techniques

Bat Algorithm

The bat algorithm (BA) was introduced by X.S Yang in 2010 [23] for continuous
problem domain. Later a binary version of the bat algorithm was introduced by
Mirjalili et al., in 2014 for discrete problem domain. The algorithm evolves on the
basis of echolocation behaviour of bats. Bats use echoes of sounds (echolocation)
emitted by them and then they navigate through their surroundings. Bats use this
approach to find their prey during nights. Using the echolocation process as the
basis, the bat algorithm was proposed to obtain the optimal solution for a multi-­
objective optimization problem. Bats have the capability to adjust their flight veloc-
ity and the frequency of the sound (pulse emission) while searching for food. These
parameters, namely, the flight velocity and the pulse emission are tuned adaptively
so that their intensity decreases once the bat has identified a potential prey.
132 S. Paneerselvam

The algorithm is governed by three important rules.

1. The distance between the bat and the food is governed by the echolocation capa-
bility of bats. This is in turn used for sensing barriers in the darkness.
2. The flight movement of bats during food search is random based on its charac-
teristic features such as velocity, frequency, and loudness.
3. Bats also have the capability to change the frequency of the sound (loudness).
The algorithm evolves around a population of bats. An initial position is assigned
to each bat in random which forms the initial population. The bats move within the
population and try to obtain the local optimum solution and from where it tries to
obtain the global optimum solution. During the process, the parameters, namely,
pulse emission and loudness are updated. The process continues until the best solu-
tion is obtained.
The procedure of bat algorithm (BA) is shown as follows:

Initialize population of the bats

Initialize parameters with random values
While (Stopping criterion not met) do
Generate new solutions
Compare pulse rate of current with the random one
Select solution among the best solution
Generate a new solution based on random flight
Compare loudness, pulse frequency of current with the
random
Accept the new solutions
Update pulse rate and reduce loudness
End while

The bat algorithm has been found suitable for nonlinear and multimodal prob-
lems. It is mostly suitable for high-dimensional problems where convergence crite-
ria is a major challenge.
Bat algorithm has been successful in the following applications:
–– Structural design optimization
–– Unit commitment and economic load dispatch
–– Adaptive learning problems
–– Path planning and scheduling
–– Network routing and analysis
–– Constraint-based optimization
–– Multi-objective optimization
8 Role of AI and Bio-Inspired Computing in Decision Making 133

Role of ISSDM in Industry 4.0

During the recent years, the word ‘Industry 4.0’ has become so common and is try-
ing to establish its roots in the future of industrial production. The initiative was
started by the Germans, and then the concept has extended universally to under-
stand and sort out means for the forthcoming industrial revolution [24]. In a smart
industry environment, almost all the processes are integrated. Raw material,
machines, products, human beings, logistics, communication, and IT-related opera-
tions are all interlinked with each other with the objective to improve the overall
production. The intelligent support systems for decision making are modelled to
increase the mass production in smart industries by decentralizing the control oper-
ations and decisions. This facilitates amendments in the production process thus
increasing the demand for mass customization. Most of the industries have already
established the Industry 4.0 drive and have been successful in the market. On the
other hand, some of the industries are still looking for prospects and advantages to
join the Industry 4.0 drive. In spite of the various dilemmas, companies are trying
to get into the Industry 4.0 drive due to the increase in competitiveness, fault diag-
nosis, fault reduction, need to use currently growing technologies, and skilled
manpower.
In Industry 4.0, decisions have to be taken such that the efficiency of the manu-
facturing process is maintained throughout the several phases of operation. There is
a large amount of data to be handled which has paved the way for the need for suit-
able decision models. To enhance the error analysis and prediction of a particular
situation to take decisions, intelligent tools such as AI and bio-inspired algorithms
are used in decision support systems, thus improving the automation process in an
industry. In Industry 4.0 paradigm, the decision-making models are decentralized,
in which the production entities are broken down into numerous sub-entities. Each
sub-entity works individually trying to make its decisions for its problem definition.
All the sub-entities are collaborated and through certain mechanisms which co-­
ordinate and exchange information between themselves. To handle such complex
tasks, ISSDM is applied so that all the entities work in an autonomous fashion based
on the knowledge of the models and knowledge of human decisions.
In any industrial process, there are a wide range of parameters that have to be
tuned and controlled very often. This is a challenging task for the AI or bio-inspired
algorithm chosen for the purpose, since the major drawback of these algorithms lies
in tuning of its parameters. Therefore depending upon the industrial process, the
intelligent algorithm is chosen wisely. The general algorithm to apply ANN and bio-­
inspired algorithm to an Industry 4.0 application is as follows:
134 S. Paneerselvam

Step 1: Define the objective function and a set of process vari-

ables for the problem under consideration
Step 2: Train the ANN to simulate the relation between the pro-
cess variables and the objective function.
Step 3: Define the constraints related to the process variable
and the convergence criteria
Step 4: Select a new set of process variables based on the opti-
mization technique
Step 5: Apply ANN to simulate the process behaviour and the
objective function
Step 6: Check for stopping condition otherwise repeat Steps 3
to 5

The above algorithm is a hybrid combination of ANN and bio-inspired algo-

rithm. The need for hybridization is to reduce the time taken to tune the algorithmic
parameters and improve the convergence rate.

Conclusion

Recent advances in high computing power have led to handling large amount of data
and information. As a result, the automation in several industrial applications have
become more complex especially in the areas of finance, and health care. Thus
AI-based tools for decision making has becoming an emerging research during the
recent years. Most of the real-world problems related to the Industry 4.0 paradigm
are complex, uncertain, and dynamic. AI and bio-inspired algorithm or a hybrid
combination of these algorithms has proved their capability in solving such prob-
lems. They have also proved to handle situations in changing environments, when-
ever a real-time decision is required.
Humans are excellent decision-makers – they take decisions based on past expe-
rience, skill set, and knowledge. Human-machine interaction is again evolving to
train the machines to take decisions based on human experience. Such communica-
tion between man and machine is based on autonomy, interaction, and personifica-
tion. Such intelligent interfaces have been emerging with the aid of AI and
bio-inspired algorithms. Manyika et al. [25] stated that ‘sophisticated analytics can
substantially improve decision making’. Using ISSDM, improvement can be done
in decision making by collecting relevant information, providing specific informa-
tion for better performance, conducting experiments to validate the performance,
forecasting, and casting the present scenario.
Several challenges are still remaining in intelligent decision making for future
researchers to study and propose better solutions. There are a wide range of unsolved
complex real-world problems due to the following characteristics:
–– Difficult for humans to understand and obtain the relation between the opera-
tional variables
8 Role of AI and Bio-Inspired Computing in Decision Making 135

–– Variables are dynamic in nature

–– Involve events that are difficult to understand and perceive
–– Involve large amount of data
The challenge for ISSDM is to provide solution for such kind of problems lead-
ing a new path for innovation thus providing efficient and effective decisions.

References

1. Pomerol, J.-C., & Adam, F. (2008). Understanding human decision making – A fundamental
step towards effective intelligent decision support. In Intelligent decision making: An AI-based
approach (pp. 3–40). Berlin, Heidelberg: Springer.
2. Marakas, G. M. (1999). Decision support systems in 21st century – US edition. Upper Saddle
River, London: Prentice Hall.
3. Little, J. D. C. (1970). Models and managers: The concept of a decision calculus. Management
Science, 16(8), B-466.
4. Keen, P. G. W., & Scott Morton, M. S. (1978). Decision support systems: An organizational
view. Boston: Addison Wesley.
5. Beulens, A. J. M., & Van Nunen, J. A. E. E. (1988). The use of expert system technology in
DSS. Decision Support Systems, 4(4), 421–431.
6. Castle, E. N., Becker, M. H., & Smith, F. J. (1972). Farm business management: the decision
making process. Farm business management: the decision making process. 2nd edn.
7. Bell, J. (1992). Discretionary decision-making: A jurisprudential view. The uses of discretion,
89–111.
8. Haagsma, I. G., & Johanns, R. D. (1994). Decision support systems: An integrated and distrib-
uted approach (Vol. 01821, pp. 205–212). Billerica: Computational Mechanics, Inc..
9. Fox, J., & Das, S. (2000). Safe and sound: Artificial intelligence in hazardous applications
(p. 307). Menlo Park: AAAI Press/MIT Press.
10. Santos, M. F., Portela, F., & Vilas-Boas, M (2011). INTCARE: Multi-agent approach for real-­
time intelligent decision support in intensive medicine. (pp. 364–369).
11. Zadeh, L. A. (1965). Information and control. Fuzzy Sets, 8(3), 338–353.
12. Goldberg, D. E., & Holland, J. H. (1988). Genetic algorithms and machine learning. Machine
Learning, 3(2), 95–99.
13. Eberhart, R., & Kennedy, J. (1995). A new optimizer using particle swarm theory. In MHS’95.
Proceedings of the Sixth International Symposium on Micro Machine and Human Science.
IEEE.
14. Dorigo, M., & Gambardella, L. M. (1997). Ant colony system: A cooperative learning approach
to the traveling salesman problem. IEEE Transactions on Evolutionary Computation, 1(1),
53–66.
15. Karaboga, D., & Basturk, B. (2007). A powerful and efficient algorithm for numerical function
optimization: Artificial bee colony (ABC) algorithm. Journal of Global Optimization, 39(3),
459–471.
16. Li, X.-l. (2002). An optimizing method based on autonomous animats: Fish-swarm algorithm.
Systems Engineering-Theory & Practice, 22(11), 32–38.
17. Yang, X.-S. (2010). Firefly algorithm, stochastic test functions and design optimisation.
International Journal of Bioinspired Computation, 2(2), 78–84.
18. Passino, K. M. (2002). Biomimicry of bacterial foraging for distributed optimization and con-
trol. IEEE Control Systems Magazine, 22(3), 52–67.
19. Ruxton, G. D., & Broom, M. (2002). Intraspecific brood parasitism can increase the number
of eggs that an individual lays in its own nest. Proceedings of the Royal Society B: Biological
Sciences, 269(1504), 1989–1992.
136 S. Paneerselvam

20. Gibbons, D. W. (1986). Brood parasitism and cooperative nesting in the moorhen, Gallinula
chloropus. Behavioral Ecology and Sociobiology, 19, 221–232.
21. Payne, R. B., Sorenson, M. D., & Klitz, K. (2005). The Cuckoos, Oxford University Press.
22. Pan, W.-T. (2012). A new fruit fly optimization algorithm: Taking the financial distress model
as an example. Knowledge-Based Systems, 26, 69–74.
23. Yang, X.-S. (2010). A new metaheuristic bat-inspired algorithm. In Nature inspired coopera-
tive strategies for optimization (NICSO 2010) (pp. 65–74). Berlin, Heidelberg: Springer.
24. Schlaepfer, R. C., Koch, M., & Merkofer, P. (2015). Challenges and solutions for the digital
transformation and use of exponential technologies. Deloitte. http://deloitte.com/content/dam/
Deloitte/ch/Documents/manufacturing/ch-en/manufacturing-industry-4-0-24102014.pdf
25. Brown, B., Chui, M., & Manyika, J. (2011). Are you ready for the era of ‘big data’. McKinsey
Quarterly, 4(1), 24–35.
Chapter 9
Smart Recognition System for Business
Predictions (You Only Look Once – V3)
Unified, Real-Time Object Detection

Allumallu Veera Venkata Susmitha

Introduction

Object recognition is viewed as a standout amongst the most difficult issues in this
field of PC vision, as it includes the mix of item arrangement and object confine-
ment inside a scene. As of late, Deep Neural Network systems (DNNs) have been
shown to accomplish better item discovery execution through different methodolo-
gies, with YOLOv2 (an enhanced You Only Look Once model) being one of the
cutting edges in DNN-based object location strategies in both speed and precision.
Even though YOLOv2 can accomplish constant execution on a ground-breaking
GPU, it stays exceptionally trying to utilize this methodology for continuous item
identification in video on installed figuring gadgets with constrained computational
power and restricted memory. In this project, we propose another system called Fast
YOLO V3, a quick You Only Look Once structure that quickens YOLOv2 to have
the capacity to perform object location in video on inserted gadgets in a continuous
way. To begin with, we use the developmental profound knowledge system to
advance the YOLOv3 arrange engineering and deliver an improved design that has
2.8× fewer parameters with only an ~2% IOU drop. To additionally lessen control
utilization on installed gadgets while looking after execution, a movement versatile
surmising technique is brought into the proposed Fast YOLO structure to decrease
the recurrence of profound derivation with O-YOLOv3 dependent on transient
movement qualities.

A. V. V. Susmitha (*)
National Chiao Tung University, Hsinchu, Taiwan

© Springer Nature Switzerland AG 2020 137

G. R. Kanagachidambaresan et al. (eds.), Internet of Things for Industry 4.0,
EAI/Springer Innovations in Communication and Computing,
https://doi.org/10.1007/978-3-030-32530-5_9
138 A. V. V. Susmitha

Literature Survey

We have investigated a couple of updates to YOLO. We made a heap of little arrange-

ment changes to enhance it. It is more noteworthy than last time, yet dynamically
correct. It is still snappy, notwithstanding. We present YOLO v3 as another way to
deal with object discovery. Earlier work on item location repurposes classifiers to
perform recognition, while we outline object location as a relapse issue to spatially
isolated bouncing boxes and related class probabilities. A solitary neural system
predicts jumping boxes and class probabilities straightforwardly from full pictures
in a single assessment. Since the entire recognition pipeline is a solitary system, it
very well may be improved start to finish on identification execution.
We brought together engineering, and it is amazingly quick. Contrasted with best
in class location frameworks, YOLOv3 makes more confinement blunders yet is far
less inclined to anticipate false identifications where nothing exists. At long last,
YOLOv3 adapts extremely broad portrayals of items. It beats all other location tech-
niques, including DPM and R-CNN, by a wide edge while summing up from nor-
mal pictures to works of art on both the Picasso Dataset and the People-Art Dataset.

Motivation Behind Using YOLOv3

Why Yolov3?

Among cutting edge techniques for profound learning object location (Faster
R-CNN, SSD, YOLO, etc.), Yolov3 emerges because of its incredible harmony
among speed and exactness. In Yolov3, each info picture is partitioned into an S × S
lattice. For every cell in the network, some jumping boxes forecasts are created at
the same time with class probabilities/scores for foreseeing objects related with that
lattice cell. Each score reflects how sure the model is that the container contains a
specific class of item [1–5].

Comparison of YOLOv1, v2 & v3

In Fig. 9.1, we refer to the various parameters and their performances and took this
figure from the paper referred to in reference [6], and now we will see the compari-
son of YOLOv3 and why exactly we are using YOLOv3. One advantage of picking
this characterization is that when YOLO cannot recognize the sort of plane, it gives
a high score to the plane as opposed to compelling it into one of the sub-classes. At
the point when YOLOv3 sees the farther picture, it just back propagates arrange-
ment misfortune to prepare the classifier. YOLOv3 finds the bouncing box that
­predicts the most elevated likelihood for that class, and it figures the arrangement
misfortune just as those from the guardians [7].
9 Smart Recognition System for Business Predictions (You Only Look Once… 139

Fig. 9.1 Comparison of YOLOv1 & YOLOv2

Design of the Model

Bounding Box Prediction

Following YOLO9000, our framework predicts jumping boxes utilizing measure-

ment groups as stay boxes. The below figure is from the paper cited in reference [8].
Then it predicts four organizations for each bouncing box, tx, ty, tw, th, on the off
chance that the cell is counterbalanced from the upper left corner of the picture by
(cx, cy) and the bouncing box earlier has width and tallness pw, ph (Fig. 9.2).

Class Prediction

Every cell predicts the classes; the ricocheting box may be contained using multi-­
label portrayal. We do not use a softmax as we have found it is futile for good execu-
tion, rather we simply use self-ruling determined classifiers. During set up, we use
betray entropy disaster for the class desires. This arrangement causes movement to
progressively amazing territories such as the Open Images Dataset [9]. In this data-
set there are many covering marks. Using a softmax powers the assumption that
every cell has exactly one class, which is routinely not the circumstance. A multi-­
label approach is a superior model for getting precise data [10–14].

Feature Extraction

In our task, we have used another framework for performing Feature Extraction.
Our new framework is a blend approach between the framework used in YOLOv2,
Darknet-19, and that advanced waiting to orchestrate stuff. Our framework uses
140 A. V. V. Susmitha

Fig. 9.2 Bounding box

prediction. bx, by, bw, bh = x,
y center coordinates, width
and height of our
prediction. tx, ty, tw,
th = Network outputs. cx,
cy = Top left coordinates of
the grid. pw, ph = Anchor
dimensions for the box

dynamic 3 × 3, and 1 × 1 convolutional layers now have some backup course of

action affiliations as well and are on a very basic level greater. It has 53 convolu-
tional layers, so we call it Darknet-53. Every framework is set up with indistinct
settings and attempted at 256 × 256 single item precision. Thus, Darknet-53 per-
forms on standard with best in class classifiers with less coasting point assignments
and more speed. Darknet-53 is better than ResNet-101 and 1.5× speedier. Darknet-53
has equivalent execution to ResNet-152 and is 2× faster. Darknet-53, moreover,
achieves the most amazing evaluated skimming point exercises each second. This
infers the framework structure better uses the GPU, making it progressively able to
evaluate and thus snappier. That is for the most part because ResNets currently have
such numerous layers and are not very profitable [15–22].

Training

We train on full pictures with no hard-negative mining or any of that stuff. We use
multi-scale preparing, bunches of information growth, clump standardization, all
the standard things. We utilize the Darknet neural system structure for preparing
training and testing.
• End-to-End Convo-Net
• Input = Images from coco dataset (common objects in context)
• Metafile: Indicating the coordinates and the class of each of the objects in image
• Size of input = 416 × 416
• Minimum confidence threshold = 0.5
• Non-max suppression threshold = 0.3
• Load pre-trained Darknet-53 and add 53 detection layers to it
• Output of Network: Position of bounding box (x, y, w, h) class of object,
confidence
9 Smart Recognition System for Business Predictions (You Only Look Once… 141

Network Architecture

The above Fig. 9.3 demonstrates the system design of YOLOv3 alluded to in [8].
YOLO might be a convolutional system, and its definitive yield is created by apply-
ing a 1 × 1 portion on a component outline. In YOLOv3, the recognition is done by
applying 1 × 1 discovery pieces on highlight maps of three totally extraordinary
sizes at three better places inside the system. The state of the location bit is 1 × 1×
(B × (5 + C)). Here B is that the assortment of bouncing boxes a cell on the compo-
nent guide will foresee, “5” is for the four jumping box traits and one item certainty,
and C is that the assortment of classifications. In YOLOv3 prepared on coco palm,
B = 3 and C = 80, in this manner the portion estimate is one × one × 255. The ele-
ment outlined by this portion has indistinguishable tallness and measurement of the
past component delineate has location properties on the profundity as portrayed
higher than.
YOLOv3 makes forecasts at three scales that square measure precisely by down
examining the size of the information picture by 32, 16, and 8, severally. The pri-
mary recognition is made by the 82nd layer. For the essential 81 layers, the picture
is down inspected by the system, such that the 81st layer includes a walk of 32. If
we have an image of 416 × 416, the resultant component guide would be of size
13 × 13.
At that point, the element outline layer 79 is exposed to numerous convolutional
layers before being up inspected by 2× to measurements of 26 × 26. This element
outline is then profundity connected with the element delineate layer 61. A compa-
rable method is pursued yet again, wherever the component outline layer 91 is
exposed to a few convolutional layers before being profundity linked with an ele-
ment delineate layer 36. Here similar examination objects might be picked inside a
similar article by various layers.

Fig. 9.3 YOLOv3 network architecture

142 A. V. V. Susmitha

Results

Real Time Image Testing (Fig. 9.4)

Input: Real image

Output: Position of bounding boxes (X, Y, W, H) Score, Class ID, Confidence

Art Work Image Testing (Fig. 9.5)

Input: Art work image

Output: Position of bounding boxes (X, Y, W, H) Score, Class ID, Confidence

Sample Video Testing (Fig. 9.6)

Input: Sample video

Output: Position of bounding boxes (X, Y, W, H) Score, Class ID, Confidence

Fig. 9.4 Real time image testing

9 Smart Recognition System for Business Predictions (You Only Look Once… 143

Fig. 9.5 Art work image testing

Fig. 9.6 Sample video testing

144 A. V. V. Susmitha

Summary

From the above results, we can say that we have a new approach to identify objects
with this advanced technology. Still, there are a few constraints in the performance
of YOLO as the model has to be detected as well as classified by using only Prime
neural network. The optimization loss can be trained on the end-to-end network for
better detection performance to avoid the localization errors. The limitation occurs
when the Faster RCNN can perform with fewer background errors, whereas our
model faces some struggles such as more background errors. Unlike other models,
our model collects the data from bounding boxes, which generalizes the objects
with improper configurations. Some of the main constraints to work on our model
are as follows:
1. Batch Normalization: In our model Batch Normalization has helped in regular-
ization, which resulted in suppressing the Overfitting cause.
2. High Resolution Classifier: Generally, the images that we have used in our model
are 256 × 256. Thus, we can say that it increased the input size while the model
I trained on the DarkNet ImageNet database. There is a 4% increment on MAP.
3. Anchor Boxes: There is a tremendous change in the accuracy by utilizing the
anchor boxes, whereas in YOLOv2, they are not used, but are replaced by the
means of input dimensions using K-means clustering to define the anchor boxes.
4. Multi-Scale Training: In the perspective of using different image sizes for the
purpose of detecting objects, the performance was not up to the level.
5. Fine-Grain features: The prediction detection in model is on 13 × 13 feature
map, which is much smaller than other YOLO models. This feature helps us
in localizing both larger and smaller objects efficiently.
6. DarkNet: The Architecture we used in our model is better compared with
YOLOv2. It has 19 convolutional layers, five Maxpooling layers, and a softmax
layer on the top of the convolutional layer for the purpose of classification.
Let us present a short conclusion of Improvements in YOLOv3:
Bounding Box Predictions YOLOv3 simply like YOLOv2 utilizes dimension
clusters to produce Anchor Boxes. Presently, as YOLOv3 is a solitary system, the
misfortune for objectiveness and classification should be determined independently,
yet from a similar system. YOLOv3 predicts the objectiveness score utilizing logis-
tic regression where 1 means total cover of bounding box earlier over the ground
truth object. It will anticipate just 1 bounding box earlier for one ground truth object
and any mistake in this would lead to both classifications just as detection misfor-
tune. There would likewise be other bounding box priors that would have an objec-
tiveness score more than the edge; however, not exactly the best one, for these
mistakes will develop for the detection loss and not for the classification loss.

Class Predictions YOLOv3 utilizes free logistic classifiers for each class rather
than an ordinary softmax layer. This is done to make the multi-label classification.
Take a model where a lady appears in the image and the model is prepared on both
9 Smart Recognition System for Business Predictions (You Only Look Once… 145

individual and lady, having a softmax here will prompt the class probabilities that
have been partitioned between these two classes with state 0.4 and 0.45 probabili-
ties. Be that as it may, autonomous classifiers tackle this issue and give a yes versus
no likelihood for each class, similar to what is the likelihood that there is a lady in
the image would give 0.8 and what is the likelihood that there is an individual in the
image would give 0.9, and we can name the item as both individual and lady.

Predictions over scales To help detect shifting scales, YOLOv3 predicts boxes at
three distinct scales. At that point, highlights are separated from each scale by utiliz-
ing a technique like that of feature pyramid systems.

Applications

Detection of objects is the primary task in Computer Vision

1. For detection of objects in real time, YOLO is a brilliant neural network.
2. Uses Non-Maximal Suppression for the purpose of best bounding box.
3. For the use of standard layers as convolutional with 3 × 3 kernel, Max-pooling
with 2 × 2 kernel.

Future Scope

What can be improved for the upcoming Model?

1. For smaller and medium objects, the average precision should be improved, so it
can perform better than Faster RCNN.
2. To decrease the localization errors, MAP must be increased efficiently.
3. Currently, for DarkNet implementation, we used C language, but Python imple-
mentation can be more efficient.
Because we know that YOLOv3 is better—stronger but not faster— future work
could be more concentrated on the perspective of making it faster. As we need the
most accurate algorithm, the speed should be traded off to boost the accuracy to
increase the complexity for the DarkNet architecture. The most important feature of
YOLOv3 is to make decisions at three different scales, and it is a completely convo-
lutional network and generates output by 1 × 1 detection kernel.

References

1. YOLOv1: https://arxiv.org/abs/1506.02640.
2. YOLOv2: https://arxiv.org/abs/1506.02640.
3. YOLOv3: https://pjreddie.com/media/files/papers/YOLOv3.pdf.
146 A. V. V. Susmitha

4. DarkNet: https://pjreddie.com/darknet/.
5. He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. In
Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 770–778.
6. https://arxiv.org/pdf/1612.08242.pdf.
7. https://medium.com/@jonathan_hui/real-time-object-detection-with-yolo-yolov2-
28b1b93e2088.
8. https://pjreddie.com/media/files/papers/YOLOv3.pdf.
9. Gordon, D., Kembhavi, A., Rastegari, M., Redmon, J., Fox, D., & Farhadi, A. (2017). Iqa:
Visual question answering in interactive environments. arXiv preprint arXiv:1712.03316.
10. Huang, J., Rathod, V., Sun, C., Zhu, M., Korattikara, A., Fathi, A., Fischer, I., Wojna, Z., Song,
Y., Guadarrama, S., & Murphy, K. (2017). Speed/accuracy trade-offs for modern convolutional
object detectors. In Proceedings of the IEEE conference on computer vision and pattern recog-
nition (pp. 7310–7311).
11. Krasin, I., Duerig, T., Alldrin, N., Ferrari, V., Abu-El-Haija, S., Haija, A., Kuznetsova, H.,
Rom, J., Uijlings, S., Popov, A., Veit, S., Belongie, V., Gomes, A., Gupta, C., Sun, G., Chechik,
D., Cai, Z., Feng, D. N., & Murphy, K. (2017). Openimages: A public dataset for large-scale
multi-label and multi-class image classification. Dataset available from https://github.com/
openimages.
12. Lin, T.-Y., Dollar, P., Girshick, R., He, K., Hariharan, B., & Belongie, S. (2017). Feature pyra-
mid networks for object detection. In Proceedings of the IEEE Conference on Computer Vision
and Pattern Recognition (pp. 2117–2125).
13. Parham, J., Crall, J., Stewart, C., Berger-Wolf, T., & Rubenstein, D. (2017). Animal population
censusing at scale with citizen science and photographic identification. In 2017 AAAI Spring
Symposium Series.
14. Redmon, J. & Farhadi, A. (2017). Yolo9000: Better, faster, stronger. In Computer Vision and
Pattern Recognition (CVPR), 2017 IEEE Conference on (pp. 6517–6525). IEEE.
15. Ren S., He K., Girshick R., & Sun J. (2015). Faster r-cnn: Towards real-time object detection
with region proposal networks. arXiv preprint arXiv:1506.01497.
16. Russakovsky, O., Li, L.-J., & Fei-Fei, L. (2015). Best of both worlds: human-machine col-
laboration for object annotation. In Proceedings of the IEEE Conference on Computer Vision
and Pattern Recognition (pp. 2121–2131.
17. Shrivastava, A., Sukthankar, R., Malik, J., & Gupta, A. (2016). Beyond skip connections: Top-­
down modulation for object detection. arXiv preprint arXiv:1612.06851.
18. Szegedy, C., Ioffe, S., Vanhoucke, V., & Alemi, A. A. (2017). Inception-v4, inception-resnet
and the impact of residual connections on learning. In Thirty-First AAAI Conference on
Artificial Intelligence.
19. Cai, H., Wu, Q., Corradi, T., & Hall, P. (2015). The crossdepiction problem: Computer
vision algorithms for recognising objects in artwork and in photographs. arXiv preprint
arXiv:1505.00110.
20. Dalal, N., & Triggs, B. (2005). Histograms of oriented gradients for human detection. In
Computer Vision and Pattern Recognition, 2005. CVPR 2005. IEEE Computer Society
Conference on (Vol. 1, pp. 886–893). IEEE
21. Dean, T., Ruzon, M., Segal, M., Shlens, J., Vijayanarasimhan, S., Yagnik, J., et al. (2013). Fast,
accurate detection of 100,000 object classes on a single machine. In Computer Vision and
Pattern Recognition (CVPR), 2013 IEEE Conference on (pp. 1814–1821). IEEE.
22. Donahue, J., Jia, Y., Vinyals, O., Hoffman, J., Zhang, N., Tzeng, E., & Darrell, T. (2013).
Decaf: A deep convolutional activation feature for generic visual recognition. arXiv preprint
arXiv:1310.1531.
Chapter 10
IoT-Based Monitoring and Management
of Electric Vehicle Charging Systems
for DC Fast Charging Facility

R. Suresh Kumar, P. K. Rajesh, J. Joys Nancy, S. Abirami,

and K. Vishnu Murthy

Introduction

The conventional powered internal combustion engine (ICE) emits toxic gases from
the tailpipe, which leads to emissions and health risks associated with the respira-
tory system. In Europe, Norway aims to transition 25% of its vehicle to electric
propulsion and 100% of passenger cars by the year 2030. Asian subcontinent coun-
tries such as India decided 100% electrification on two and three wheelers by the
year 2025. Besides, vehicle populations are growing the demand for the charging
infrastructure and energy storage requirements of facilities, which present the chal-
lenges faced by electrification.
However, with the current charging stations powered by alternating current (AC),
a slow charging station typically needs an 8–12 h duration to charge 0–100% SOC
of the EV’s storage devices. During charging, EVs primarily require proper authen-
tication and authorisation between the vehicle and charging station, security for
payment, load balancing during peak and low utilized conditions [1].
The prime setback for the xEVs charging duration is with existing AC chargers,
which hinders the implementation and adoption. To overcome these challenges, DC
charging stations have been introduced by EV manufactures. The challenges associ-
ated with DC chargers align with battery pack health and vulnerability of safety due
to quick charging. To overcome the above blockades, IoT integrated facilities need
to transform efficiency and safety leading to a new paradigm shift. Presently avail-
able chargers are modified with a monitoring process for charging xEVs, where

R. Suresh Kumar (*) · P. K. Rajesh

PSG College of Technology, Coimbatore, India
e-mail: [email protected]
J. Joys Nancy · S. Abirami · K. Vishnu Murthy
Sri Krishna College of Technology, Coimbatore, India

© Springer Nature Switzerland AG 2020 147

G. R. Kanagachidambaresan et al. (eds.), Internet of Things for Industry 4.0,
EAI/Springer Innovations in Communication and Computing,
https://doi.org/10.1007/978-3-030-32530-5_10
148 R. Suresh Kumar et al.

xEV vehicle parameters are articulated with a smartphone device assisted cloud
computing facility. The emerging IoT driven EV charging stations are manufacturer
assisted secured for the users against cybersecurity threats and issues [2–6].
The IoT bandwidth cloud promotes the offer to the applications to receive, anal-
yse, control and backup the stored parameters to assist EV users to enhance safe and
reliable operations. With the progression of IoT, insecurities and vulnerabilities of
the systems were rectified with communication protocols adopted by integrated IoT
fast chargers. Their implementation leads towards the adoption of IoT driven DC
fast chargers with secure interconnection nodes (Fig. 10.1).
A sensor multiplex with network topology incorporation of IoT enabled devices
in EV chargers abstracts the real-time data about the system supervision and perfor-
mance. While supervising integration and deployment of IoT enabled DC charging
infrastructure, it is commanding to recognise the challenges related to security
issues. The IoT protocols must be subjective to develop secure, compact and
decision-­making DC chargers for EVs for quick charging functionality [6–11]. This
chapter makes the effort to address an approach and a secure design for IoT inte-
grated DC charging infrastructure. The chapter focuses on architecture to encom-
pass safe-­keeping and confidentiality of bill payments. Also, the chapter discusses
how the IoT enabled DC fast chargers differ from the conventional AC charging
infrastructure facilities. The EVs charged at different locations with standard charg-
ing devices and emphasis on implementing the supervisory charging environment
between EVs and charging station with authentication integral is addressed.

Battery Management Systems On-board Charger

Lithium-Ion Battery Pack

AC- Charger Scoket

DC Fast Charging Station

Battery Management Systems

Lithium-Ion Battery Pack

Fig.10.1 (a) AC charging systems. (b) DC fast charging stations

10 IoT-Based Monitoring and Management of Electric Vehicle Charging Systems… 149

The abduction of information between chargers and electricity suppliers to the

power grid are not emphasised in this chapter due to policy variation of the power
transmission and distribution. The chapter discussions are made on analysis of the
current scenario of EV charging infrastructure, methods to design a secure EV
charging system, exploitation of relevant theory and risk assessment study. Finally,
conclusions and certain limitations of the smart IoT driven charging infrastructure
facilities are indicated [12–17].

Installed Present Charging Infrastructure Review

One of the topmost priorities in the smart city facility is to make efficient energy
utilisation sustainable. Plug-in hybrid electric vehicles (xEVs) and plug-in Electric
Vehicles (EVs) are the two main architectures manufactured by the EVs manufac-
turers. For the range and onboard utilisation, two powertrain deployment was the
prime focus. EV using countries have EVs charged with AC and DC chargers based
on controlled and adapter circuits with battery pack capacity. DC chargers stations
expressly reduce the time by 80% depending on battery capacity [18]. DC chargers
can charge the battery pack from zero to 100% state of charge (SOC) for an electric
vehicle from 650–1850 seconds [16, 19–26].
Also, installation cost, power electronics components and time for developing
IoT integration are the main challenges. Certain studies portray charging installa-
tions in the current charging stations lacked sacredness of personal information
about the users. The vulnerabilities ranging from the ID for the vehicle manufactur-
ers assigned by service providers might lead to encryption if the standard protocols
such as OCPP are used. Consequently, there is also the opportunity to come up with
new guideline protocols for the IoT fitted DC chargers.
The security process relevant to hacking and injection of malware being slow in
updating is also a vital concern for the IoT enabled DC chargers. Unsanctioned
admittance of fake EV user ID and illicit transactions from actual account holders
are also critical challenges for the xEVs manufactures deployment. These key chal-
lenges are the prime focus that should be addressed by the manufactures at the ini-
tial phase of DC chargers installation.
The exploitation of wireless communication with IoT leads to a creative engi-
neering industry on horizontal and different verticals. On the other hand, high-speed
grid and soft computing devices become cheaper, which enforces EV charging
infrastructure to proceed with intelligent chargers. Certain studies investigated IoT
testing standards, such as physical security, data transfer in cloud, certification and
firmware [26–31].
To overcome the above issues during the initial phase of product development,
the EV and DC fast charger manufacturers need to address them. Security and dis-
cretion considerations scrutinised and validated by standard protocol promote the
reliability of the charging infrastructure. This chapter provides insight into the IoT
compelled rapid charging of battery packs with the secured baseline to overcome
the threat of high steadfast DC chargers.
150 R. Suresh Kumar et al.

IoT-Based DC Fast Charging Deployment Requirements

The following factors are the requirements to deploy IoT-based DC fast char-
gers [32].
• The abstraction of xEVs user necessities with technological demands to estimate
the latent attainment of IoT-based charging integrated infrastructure facility.
• Technology-driven factors and tasks that affect the implementation of DC fast
chargers and with integrated IoT-based monitoring and control flow process.
• Validation of design and product development sequences by xEVs makers, EV
manufacturers, internet service providers, component suppliers and investors.
• The collaboration of cloud computing service providers with power grid
suppliers.
• Assessment of legislative guidelines, framework and investment activities influ-
encing towards quick charging DC chargers deployment.
• Scrutinising vulnerability of energy storage devices in the DC fast charging facil-
ity remotely through cloud computing facility.
Also, a systems-level fault diagnostics study will be carried out on the EVs and
the information will be shared with the service provider (service station) to detect
and change the defect devices at the systems level to enhance the reliability of EVs.
The communications to the grid should have sufficient intelligence to meet the dif-
ferent bandwidths of vehicle charging populations. The chargers are fused with data
of the vehicle, charger status and substation. It enhances the effective operating
mode of electric vehicles. Besides, if the demand is very minimal while the vehicle
is on the charger, the charges tariff can discount to the consumer, which promotes
electrification of transportation and simultaneously optimises the grid and generates
station power.
IoT context with numerous features and advancements are applied to EV charg-
ing infrastructure facilities as a core design of information and communication tech-
nology. Also, concurrent growth in artificial intelligence, neural network, big data
analytics, machine learning and cloud computing will permit controlling, process-
ing and making judgments based on the enormous amount of information needed.
A large number of devices and systems tangled will make improvements in machine-­
to-­machine communications and will empower the required real-time bidirectional
communications to meet the needs of security, reliability and quality of service
(Fig. 10.2).
Owing to the constraint of the limited range and the long charging times, several
techniques required to optimise the battery usage and to enhance the range of EVs
have been developed. The two important investigation encounters that need to be
addressed are: (1) Energy-efficient EVs routing charging to established search algo-
rithms for the adoption of the demand for EVs to calculate the route’s ability to
enhance the driving range; (2) Battery efficiency, battery parameters and faculty
condition, need to be monitored and data shared with the service and original
­equipment manufacturer for preventive maintenance to maximise battery health and
life (Fig. 10.3).
10 IoT-Based Monitoring and Management of Electric Vehicle Charging Systems… 151

Electric
Charger
Service and Vehicle
Dealership
Cloud Storage

Grid Substation

IoT Hub

User Information

Fig. 10.2 Attributes of IoT-based DC fast charger systems

Grid Substation

Service Station

Cloud Services
Power
Line

DC Fast Chargers

Fig. 10.3 Integrated service providers for DC fast charging systems

152 R. Suresh Kumar et al.

Electric Vehicle Charging

Battery Charging

The main challenges of EVs are charging and cost of the battery. Battery overcharg-
ing and deep discharge results in rapid ageing characteristics. The present energy
storage devices constructed with lithium-ion chemistries are more vulnerable to
various physical parameters, such as overcharged, temperature window, safety
issues, state of charges, cycle life, shelf life and cell balancing. To overcome these
challenges, sensors and estimators algorithms with actuators integrated with battery
management and supervisory control termed as battery management systems (BMS)
are needed. These BMS have many serviceable segments and key challenges, such
as measurement of cell voltage to correlate state of charge, cell balancing, safe oper-
ating temperature and fault diagnosis. Thousands of cells are connected in series
and parallel combinations as a module, and a few of the module integrations are said
to be a vehicle battery pack. The battery pack and a motor controller are the EVs
powertrain management systems. Proven literature demonstrates an internal resis-
tance equivalent circuit (Rint) model where the charging and discharging have a
similar trajectory. The equivalent circuit model (ECM) and the study of the Kalman
filter helps to estimate the state of charge shown in Fig. 10.4. By using the Coulomb
counting Ampere-hour integral method adopted to estimate SOC, the voltage of the
cell to battery pack measurements correlated with SOC.
The xEVs battery pack state of charge (SOC) and state of health (SOH) predica-
tion provides the input BMS to display the remaining charge as monitoring the
slowly varying battery ageing parameters. Most of the survey depicted that an adap-
tive extended Kalman filter (AEKF) is the best suited for the highly non-linear
systems.
Certain research depicts the capacity estimation technique determined by the
Q-estimation algorithm. The structure shown in Fig. 10.4 is a double layer estima-
tion structure integrated with an algorithm for combined SOC and Q evaluation, and

Fig. 10.4 Adaptive

extended Kalman filter
[33]
10 IoT-Based Monitoring and Management of Electric Vehicle Charging Systems… 153

the SOC estimation is determined from an adaptive extended Kalman filter (AEKF).
The Q-estimation algorithm provides the AEKF filter with an update value of the
battery capacity, given the actual level of ageing. During fast charging, the Q value
factor tunes the charger to set automatically based on the ageing.
Figure 10.5 depicts the charging and discharging profile of a lithium-ion cell
constructed with ion phosphate (LFP) chemistry. The plot shows the hysteresis of
the charging and discharging profile. The Q algorithm predicts the level of ageing
under different operating temperature conditions. The collected data about the his-
tory of operation will provide the information to the user about the remaining useful
life of the battery pack of the vehicle (Fig. 10.6).

3.40 Discharging
Charging
3.38

3.36

3.34
OCV(V)

3.32

3.30

3.28

3.26

3.24

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
SOC

Fig. 10.5 Charging and discharging profile vs state of charge for 18,650 lithium-ion phosphate
cell

Substation and Service Data Fusion

Mobile Data Data Otimization Rotate Vehicle

Architecture Access Fusion Algorithm Decision Architecture

Intellection Realization Expectancy

Fig. 10.6 Battery parameters observing system using IoT

154 R. Suresh Kumar et al.

The design facilitates the assimilation of the different sensors, regardless of

their communication technology. It is because the block outfits encapsulation
mechanisms for the data access received from different communication protocols
through the wireless topology. In this manner, the design features can obtain infor-
mation from different sensors connected to the system. The data obtained from the
vehicle decides the decision for the charging station based on the availability from
nearby charging stations. The next layer is to obtain information from any of the
sensory networks. Data collected by the different sensors in the mobile architecture
is pre-­processed for data access. As the data from assorted sources receives infor-
mation fusion, optimisation algorithms are performed to represent the knowledge
for prediction and estimation for the demands. The optimisation algorithms’ mech-
anisms depend on the implementation of the information superimposed in the data
access. The optimisation model combines information from derived routes, on the
geographic locations of the routes, earlier ingestion data, information from EVs
­sensors, instrument cluster devices and mobile navigation sensors. The optimisa-
tion models help sort levels of the traffic conditions, the route travelled, the battery
level and the physical strain of the user. Certain optimisation works on the particle
swarm optimisation (PSO) were synchronised by a neural network to optimise sev-
eral constraints in the routes. It is expected that the control strategy with real-time
optimisation estimates and predicts based on drive cycle demand and GPS data.
From the optimal torque demand by the powertrain, the battery pack parameters
adopt the actual torque by shifting torque demand to the road situation by limiting
the speed. However, it relies on a high reckoning effort to encounter the real-time
requirement. The fused sensor network deployed in the design is the most com-
monly used for electric vehicles and chargers the battery based on the optimised
route location. The battery SOC, temperature, distance travelled and network faults
in the battery pack will be crucial parameters to optimise battery stacking results as
a cost structure of the intelligent systems.

Electric Vehicle Battery Management Using IoT

The IoT grounded battery-conditioning system for EVs discussed in this section
comprises a communication protocol to the mobile architecture, data acquisition,
cloud storage and an intelligent vehicle instrument cluster. An integrated IoT
embedded architecture provides communication from the different data flow nodes.
Encapsulated data is connected via the Internet gateway for all battery system
parameters such that it enables one to back-up, analyse and control the critical
parameters through the cloud computing systems.
The integrated IoT-based battery conditioning system communicates digitally
with TCP/IP via Mat file (MATLAB) format for data acquisition purposes. The data
from battery measurement parameters from the mobile architecture are extracted for
further needs. Integrated protocol are optimised for connecting devices to vehicle
operators, a special case of the D2S pattern, since vehicles are connected to the serv-
10 IoT-Based Monitoring and Management of Electric Vehicle Charging Systems… 155

Fig. 10.7 Battery parameters observing system using IoT

ers through the internet data in the vehicle. Data is shared to the database and web
server and client on the cloud system through the Internet gateway. The data in the
cloud system is processed, analysed and a decision arrives at operators through the
xEVs touch-based instrument cluster panel. The IoT architecture for the battery
pack observing systems will be an embedded system utilised as an IoT to provide
suggestions via hardware integration, for the main function to acquire data acquisi-
tion and as an Internet gateway for the data access. This acquired data is processed
into an optimised algorithm estimation and a prediction is proposed using Kalman
filter with the layout shown in Fig. 10.7, and the same is done using a graphical user
interface to the data server for programming and executes in the vehicle instrument
cluster.
The web-based GUI executes the data acquisition functions to regain battery
measurement parameters from the MATFILE format. Transferred data through
structured query language will be stored in cloud computing stations.

IoT Charging Station Integration Requirements

In the designing process of IoT integrated xEV charger systems the following
requirements must prevail for safety and security. Guidelines and technical mea-
sures correlation expose threats in the present chargers. These procedures and tech-
nical scrutiny aim to alleviate extortions, liabilities and threats identified in the IoT
driven EV chargers (Table 10.1).
156 R. Suresh Kumar et al.

Table 10.1 Standard for DC charging station and EV connectors and the communication standard
between EV and charging standard [1]
Communication Standard Description
DC electric vehicle charging IEC 61581-23 Electric vehicle conduction-DC EV charging
station [34] station
Plugs, socket outlet, vehicle IEC 62916-3 Dimensional compatibility and
connectors and vehicle inlet for interchangeability requirement for DC and
DC charging [35] AC pin and tube type vehicle connectors
Communication between plug-in SAEJ2847/2 Requirement and specifications for
vehicles and off-board DC communication between plug-in vehicles
chargers and off-board DC chargers
Communication between EV ISO 15118 Vehicle to the grid communication interface
DC charging specification DIN Digital communication between xEVs and
communication 70121:2014:2 deployed DC chargers [1].

User Device

IoT Hub Database

IoT Device
Authentication

Payment Bill Procesing for the

Service energy Consumed

Fig. 10.8 A typical IoT charging station integration

Figure 10.8 depicts IoT modem connections to the cloud service, and it shows the
communication and power line for the rapid DC chargers. Mobile applications
installed in the xEVs’ user device ensure the security and authentication for the start
of charging. In this schematic, the user’s locations are identified via navigation sys-
tems with charger ID for the service provider to assess the vehicle required demand.
Also, the user’s requisition for charging station ID must be entered. If the user ID is
matched and further processed, a payment service on the cloud service initiates a
payment transaction with an external online payment service provider from wallet, a
credit card, a debit card, RFID-based detection systems or net banking payment.
After the payment transaction is authorised by the payment service, the cloud service
transmits a command to the IoT hub in fast DC chargers to authorise the connected
EV of the user and allow the use of the charging station. The payment is only for the
10 IoT-Based Monitoring and Management of Electric Vehicle Charging Systems… 157

consumed energy initially calculated based on the SOC of the battery. As per the
standard DIN SPEC 70121 and ISO 15118 bidirectional communications are initi-
ated between the vehicle and cloud computing to the user via the user display devices
(smartphone app). Also, the payment service is a bank account that is interlinked
with cloud service for the cost deduction authentication. If the users require a partial
charge to charge the battery and want to travel a shorter distance, the transaction
money refund to the user account is also one of the potential benefits from the IoT
enabled DC fast chargers.

Conclusion

In rapid development, IoT technology drives opportunities for xEV manufactures to

privilege their strong presence in the market. Besides, IoT still lacks platform stan-
dards, network standard protocols, privacy, interoperability and faces malware
threats. IoT manufacturers have provided solutions and consideration for the design
and development of highly secure authentication.
Also, the integrated IoT should possess security data on areas that are open to
exploitation and have it critically assessed before the integration. During the design
process itself, essential checkpoints in architecture are thoroughly analysed with
threats associated in the setup. Additionally, the charging infrastructure of xEVs
integrated with different xEVs manufactures is going to adopt the standard protocol.
The main factors in portraying successful deployment are the recognition, detec-
tion, access management, encryption, adaptability and safe transactions for bill pay-
ment, applied to every entity of the IoT driven charging infrastructure. A test of the
smart application in the user interface display or mobile enables the users to monitor
and control parameters of the EVs for the knowledge and information about the
charger systems utilisation. Cloud information of the individual EV database will be
helpful to the EV manufacturers for research and development details. Also, the
same details will provide a clear usage foundation of the batteries and charge/dis-
charge patterns.
During peak and low demand conditions, the energy providers can analyse the
consumption of energy. Also, the consumed energy and charging time through IoT
provides data about the best EVs for the customers to choose the right vehicle.
Through the IoT self-discharge and ageing characteristics of the energy storage
devices, information can be predicted for certain vehicles not operated for a couple
of days or longer duration.

References

1. Ranabhat, P. (2018). Secure design and development of IoT enabled charging infrastructure for
electric vehicles; using CCS standard for DC fast charging. Metropolia University of Applied
Sciences, Information Technology.
158 R. Suresh Kumar et al.

2. International Electrotechnical Commission. (2014). IEC 61851-23. Electric vehicle con-

duction system – Part 23 – D.C. electric vehicle charging station. Geneva: International
Electrotechnical Commission.
3. Shaffers, H., Komninos, N., Pallot, M., Trousse, B., Nilsson, M., & Oliveira, A. (2011). Smart
cities and the future internet: Towards cooperation frameworks for open innovation. In The
Future Internet (pp. 431–446). Berlin/Heidelberg: Springer.
4. Zhang, Q., Liu, M., Lin, X., Liu, Q., Wu, J., & Xia, P. (2019). Optimal resonant beam charging
for electronic vehicles in the internet of intelligent vehicles. IEEE Internet of Things Journal,
6(1), 6–14.
5. Wang, H., He, L., Qi, X., Wang, H., Li, G., & Dai, X. (2012). Simulation system of the perfor-
mance of power battery for electric vehicle based on internet of things. In Proceedings of 2012
International Conference on Measurement, Information and Control (Vol. 2, pp. 681–684).
6. Ouya, A., De Aragon, B. M., Bouette, C., Habault, G., Montavont, N., & Papadopoulos, G. Z.
(2018). An efficient electric vehicle is charging architecture based on LoRa communication.
In 2017 IEEE International Conference on Smart Grid Communications (SmartGridComm)
2017, vol. 2018–January, no. October, pp. 381–386.
7. Li, H., Zhang, W. S., Shi, J. Y., & Ma, X. Y. (2013). The sensing technologies in the elec-
tric vehicle charging and exchanging station. Applied Mechanics and Materials, 303–306,
1456–1459.
8. Kobeissi, A. H., Bellotti, F., Berta, R., & De Gloria, A. (2018). IoT Grid alignment assistant sys-
tem for dynamic wireless charging of electric vehicles. In 2018 Fifth International Conference
on Internet of Things: Systems, Management and Security. IoTSMS, 2018 (pp. 274–279).
9. Karpenko, A., Kinnunen, T., Framling, K., & Dave, B. (2018). Open IoT ecosystem for smart
EV charging. In 2018 Global Internet Things Summit (GIoTS) 2018.
10. Li, K., Xue, Y., Cui, S., & Niu, Q. (2014). Intelligent computing in smart grid and electri-
cal vehicles: International conference on life system modeling and simulation, LSMS 2014
and international conference on intelligent computing for sustainable energy and environment,
ICSEE 2014 Shanghai, China, September 20–23, 2014 Proceedings, Part III. In Commun.
Comput. Inf. Sci., Vol. 463, no. September 2014.
11. Hussain M., Alam M. S., & Beg M. M. S. (2019). Proceedings of 2nd International Conference
on Communication, Computing and Networking, Vol. 46. Springer Singapore.
12. Hariri A. O., El Hariri M., Youssef T., & Mohammed O. (2018). A decentralized multi-agent
system for management of en route electric vehicles. In Conf. Proc. – IEEE SOUTHEASTCON
(Vol. 2018, pp. 1–6).
13. Bahadoorsingh, S., Meetoo, C., Sharma, C., & Hosein, P. (2019). A unique approach to
demand side management of electric vehicle charging for developing countries. In 2018 IEEE
International Smart Cities Conference (ISC2) (pp. 1–2). IEEE.
14. Asaad, M., Ahmad, F., Alam, M. S., & Rafat, Y. (2018). IoT enabled monitoring of an opti-
mized electric vehicle’s battery system. Mobile Networks and Applications, 23(4), 994–1005.
15. Ruben, A., Vitor, M., Joao, C., Andres, A., Joao, L., & Jose, A. (2018). Development of an IoT
system with smart charging current control for electric vehicles. IECON 2018 - 44th Annual
Conference of the IEEE Industrial Electronics Society (pp. 4662–4667). IEEE.
16. Chaudhary, R., Jindal, A., Aujla, G. S., Aggarwal, S., Kumar, N., & Choo, K.-K. R. (2019).
BEST: Blockchain-based secure energy trading in SDN-enabled intelligent transportation sys-
tem. Computers & Security, 85, 288–299.
17. Hua, M., Li, C., Huang, Y., & Yang, L. (2017). Throughput maximisation for UAV-enabled
wireless power transfer in the relaying system. In 2017 9th International Conference on
Wireless Communications and Signal Processing (WCSP) (pp. 1–5). IEEE
18. Guide to EV charging Jun-2018. https://www.zap-map.com/charge-points/
19. Meng, Q., & Jin J. (2011). The terminal design of the energy self-sufficiency Internet of
Things. In 2011 International Conference on Control, Automation and Systems Engineering
(CASE) (pp. 1–5). IEEE.
10 IoT-Based Monitoring and Management of Electric Vehicle Charging Systems… 159

20. Zhang, R., Cheng, X., & Yang, L. (2016). Flexible energy management protocol for coopera-
tive EV-to-EV is charging. IEEE Transactions on Intelligent Transportation Systems, 31(3),
97–101.
21. Zeng, F., Zhang, R., Cheng, X., & Yang, L. (2017). Channel prediction based scheduling for
data dissemination in VANETs. IEEE Communications Letters, 21(6), 1409–1412.
22. Barbarisi, O., Vasca, F., & Glielmo, L. (2006). State of charge Kalman filter estimator for
automotive batteries. Advances in Automotive Control-Control Engineering Practice, 14(3),
267–275.
23. Huvio, E., Grönvall, J., & Främling, K. (2002). Tracking and tracing parcels using a distributed
computing approach. In O. Solem (Ed.), Proceedings of the 14th Annual conference for Nordic
researchers in logistics (NOFOMA’2002), Trondheim, Norway (pp. 29–43).
24. Bellavista, P., Cardone, G., Corradi, A., & Foscini, L. (2013). Convergence of MANET and
WSN in IoT urban scenarios. IEEE Sensors Journal, 13(10), 3558–3567.
25. Bellavista, P., Cardone, G., Corradi, A., & Foscini, L. (2013). Convergence of MANET and
WSN in IoT urban scenarios. IEEE Sensors Journal, 13(10), 3558–3567.
26. Shaffers, H., Komninos, N., Pallot, M., Trousse, B., Nilsson, M., & Oliveira, A. (2011). Smart
cities and the future internet: Towards cooperation frameworks for open innovation. In The
future internet assembly (pp. 431–446). Berlin, Heidelberg: Springer.
27. Hernandez-Munoz, J., Vercher, J., Munoz, L., & Galache, J. (2011). “Smart cities at the
forefront of the future internet”, the future internet, lecture notes computer. Science, 6656,
447–462.
28. Mulligan, C. E. A., & Olsson, M. (2013). Architectural implications of smart city business
models: An evolutionary perspective. IEEE Communication Magazine, 51(6), 80–85.
29. Meetoo, C., Bahadoorsingh, S., Balbadar, V., Jaglal, D., Sharma, C., Baboolal, K., & Williams,
M. (2018). Electric vehicle policy formulation framework for SIDS in the Caribbean. In 2018
IEEE Transportation Electrification Conference and Expo (ITEC) (pp. 152–157). IEEE.
30. Mahadeo, G. E., Bahadoorsingh, S., & Sharma, C. (2017). Analysis of the impact of bat-
tery electric vehicles on the low voltage network of a Caribbean island. In 2017 IEEE
Transportation Electrification Conference and Expo (ITEC) (pp. 364–369). IEEE.
31. Buratti, C., Conti, A., Dardari, D., & Verdone, R. (2009). An overview on wireless sensor
networks technology and evolution. Sensors, 9(9), 6869–6896.
32. Kobeissi, A. H., Bellotti, F., Berta, R., & De Gloria A. (2018). IoT grid alignment assistant sys-
tem for dynamic wireless charging of electric vehicles. In 2018 Fifth International Conference
on Internet of Things: Systems, Management and Security (pp. 274–279). IEEE.
33. Taborelli, C., Onori, S., Maes, S., Sveum, P., Al-Hallaj, S., & Al-Khayat, N. (2016). Advanced
battery management system design for SOC/SOH estimation for e-bikes applications.
International Journal of Powertrains, 5(4), 325–357.
34. Fluhr, J., & Lutz, T. (2011). Chapter 8 communication with and for electric vehicles. InTech.
35. Morris, B., Foiadelli, F., Longo, M., & Zaninelli, D. (2015). e-Mobility forecast for the trans-
national e-corridor planning. IEEE Transactions on Intelligent Transportation Systems, 17(4),
680–689.
Chapter 11
An Evolution of Innovations Protocols
and Recent Technology in Industrial IoT

V. Saranya, M. J. Carmel Mary Belinda, and G. R. Kanagachidambaresan

Introduction

Manufacturing industries experience complex difficulties in building supply chain

connection in this technological business world. To develop a successful system,
new concepts such as Internet of Things (IoT), Digital Technology, and Cyber
Systems have gained more importance in the industrial sector including manufac-
turing units. The evolution in the digital era defines the Fourth Industrial Revolution
(Industry 4.0) and it is also recognized as German high-tech strategy for future
manufacturing industries [1]. Industry 4.0 produces overwhelming results by trans-
forming the manufacturing and production process of industries. In addition,
Industry 4.0 will play a vital role in transforming well-established companies into
Smart industries with the help of Internet of Things (IoT) and Cyber Systems. A
distributed approach produces an immense value in Industry 4.0, which highlights
self-determining organization of processes and smart objects throughout the net-
work by collaborating on the processes of real and virtual world [2]. The develop-
ment of incorporated processes and human-machine communication motivate
involvedness and liveliness for data transmission between industrial chains [3].
With the assistance of Industry 4.0, industries gain lot of operational efficiency in
time, cost, and production. Designing and developing the infrastructure of IoT offer
common platforms via cloud systems between associates in supply chains; conse-
quently, business processes get optimized [4]. The notion of Industry 4.0 has gained
great significance in the recent years. The increase in the usage of computerized

V. Saranya (*)
Department of IT, Sri Krishna College of Engineering and Technology, Coimbatore, India
M. J. Carmel Mary Belinda · G. R. Kanagachidambaresan
Department of CSE, Vel Tech Rangarajan Dr Sagunthala R&D Institute of Science
and Technology, Chennai, India

© Springer Nature Switzerland AG 2020 161

G. R. Kanagachidambaresan et al. (eds.), Internet of Things for Industry 4.0,
EAI/Springer Innovations in Communication and Computing,
https://doi.org/10.1007/978-3-030-32530-5_11
162 V. Saranya et al.

Fig. 11.1 Evolution of

technology

systems after the Third Revolution, Industry 4.0 deals with developing more digi-
tized ­systems and network incorporation via smart systems. Through Industry 4.0,
smart systems would facilitate the substitution of human beings in assured tasks and
effortlessness in the working atmosphere.
As IIoT features are similar to that of IoT in the industrial sector, and Industry
4.0 concepts are effectively subsets of IIoT, as shown in Fig.11.1, in identifying the
difference between IoT and IIoT. Even though the fundamental concepts of both are
the same, i.e., interrelated smart devices that permit remote sensing, data collection,
processing, monitoring, and control are the parameters that recognize the IIoT as a
subset of IoT, there are the strong necessities for constant operation and safety as
well as the equipped technology employed in the industrial sector. Therefore, this
chapter includes an evolutional study of IIoT in Section “Evolution of IIOT”.
Section “Innovations and Recent Technologies in IIoT” evaluates the innovations
and recent technologies used, Section “Protocols in IIoT” discusses the protocols
used for designing this environment, and finally Section “Conclusion” concludes
this chapter.

Evolution of IIOT

The Industrial Internet of Things (IIoT) is a commercial technology, and because of

its major benefits, it occupies a major part in the development of an industry. These
technological developments have made innovations in the implementation of smart
factories and predictive technology. As the machines are equipped with sensors,
industries train the employees on the supply and delivery chains with the specific
industrial tools to monitor and reply the output from the sensors and companies can
be able to rationalize and restructure the industrial business operations. The
Industrial Internet of Things (IIoT) discusses how the interconnected sensors,
instruments, and other devices are interrelated together with workstations for indus-
trial applications, including manufacturing and energy management. This connec-
tivity permits data collection, exchange, and analysis, theoretically facilitating
developments in efficiency and productivity as well as other economic benefits. The
IIoT is an evolution of a distributed control system (DCS) that allows for a automa-
tion by using cloud computing to improve and enhance the method controls. The big
data and analytics are preliminary to influence the end users.
11 An Evolution of Innovations Protocols and Recent Technology in Industrial IoT 163

IIoT is flagging the method for a broad change of predictable business processes
from gathering to delivery to preservation to fabrication. Warehouse sensors permit
firms to monitor stock cleverly so that shares and supplies are well-organized when-
ever needed. This prevents overstocking and under stocking of essential items.
Sensors along the delivery chain make it likely to track shipments from the moment
they leave the factory premises to the moment they arrive at the customer premises.
Sensors along the production line lead to timely detection of potential breakdowns.
By depending on projecting care to fix problems before they happen, companies
sidestep costly interruptions and disruptions in production. All these exhibitions
improve efficiency, minimize unnecessary expenses, and maximize quality.
• Contemporary Responses and Mass Facilities
IIoT has designed an alteration in the organization and updated expectancies.
The smooth access up to the moment records, products, and services that our mobile
devices provide us has caused an improved demand for real-time systems and
responses, which is even more common in IIoT. As sensors provide present percep-
tion and statistics about warehouses and machines, a greater need for immediate, or
real-time, services has arisen. The purpose is obvious: IIoT, this connection of busi-
ness machines and sensors up the internet and different devices, is useless without
mechanisms in area that react up to the moment notifications and updates. There is
no experience in understanding about forthcoming issues or shortages if there is no
manner up to date respond with the necessary speed and accuracy that avoid poten-
tial issues.
This shortage has marked a new beginning to an entire different innovation:
crowd offerings. Businesses pool collectively all their assets—employees, partners,
subcontractors, freelance professionals, and specialists—to create a pool of avail-
able service vendors to respond to the elevated demand generated through actual-­
time services. This crowd is based on discipline provider control software program
and their mobile devices to live informed of service requests, product information,
consumer records, and more.

Industry 4.0

The industrial IoT 4.0 is a suitable method for processing data such as sensor out-
put, consumer input, identification of the service provider availability, knowledge
provider, and other various effective real-time responses given by mobile devices
and also gives an idea about the next step to industrial revolution. As the evolution
of IoT to IIoT after twenty years created a potential evident in the new industrial
technology. To imagine the continuous change in technology and redefine the daily
lifestyle and work over the course made a remarking foundation in the field of
IIoT. The connections between devices and accumulation of raw data with the help
of Big Data Analytics made the industries to rely on artificial intelligence (AI) and
edge computing for processing and analysis.
164 V. Saranya et al.

Revolution of IoT

In the revolution of IoT, Kevin Ashton created awareness that physical objects can
be linked to a network during 1999. By this it is known that information to comput-
ers no longer needed to be programmed and given in the form of functions but sense
the world around them by fingers without any added information given by the
humans this process created a new computational technology era. IoT has evolved
in all the perspectives from GPS to smart homes and smart cars to health and fitness
monitoring and other smart environments and also sets a stage for the next industrial
revolution by giving way to IIoT.
As an example, a smart home owner controls his home devices by working on an
app on his mobile by just making some clicks and swipes for switching on the AC
or heater and also checks the status of the refrigerator for pickup things from super-
market. Other apps from the mobile could be able to update the real-time status and
current scenario around their living environments. The smart application system
software is developed to perform the user-defined task much faster and controls the
devices automatically by connecting with internet. The inputs and outputs from the
devices can be viewed by using electronic gadgets, and also the overall process can
be analyzed. Built-in sensors are used to make a trustworthy connectivity between
devices for better determination of data and sequence of transmission while per-
forming the required task.

Versions of IIoT

In the past centuries, goods and other necessary products are manufactured by using
hand or by using animals. However, at the beginning of the nineteenth century man-
ufacturing and production drastically introduced industry 1.0 and other operations
rapidly developed based on the requirements [5]. The different versions of IIoT are
listed as follows.
Industry 1.0
During the 1800s, water- and steam-powered machines were developed to aid the
workers who started the industrial revolution 1.0 as the production capabilities
increased and also business grew to a next level where the owners, managers, and
other employees serving customers used steam-powered engines and water as a
source of power. This led to a lot of changes in textile industries and constituted a
major part of the British economy.
Industry 2.0
At the beginning of twentieth century, electricity became the primary source of
energy for functioning the machineries. The users felt easier to use when compared
with water- and steam-powered individual machines. Eventually, new machines
were designed to use electric power sources and also with portability. To increase
11 An Evolution of Innovations Protocols and Recent Technology in Industrial IoT 165

the efficiency and effectiveness in manufacturing units, this period created a drastic
development in major management programs where labors works for a part of job
by increasing the productivity. The concept of assembly lines increased mass pro-
duction. Lean manufacturing principles further enabled manufacturing industries to
improve the quality and also produce good products.
The electrified machines in factories massively increased the production rates.
The mass production from steel industries helped introduce railways into the sys-
tem. Innovations in chemical industries lead to the invention of the synthetic dye,
which marked a primitive state in the fabric design with various colors.
Industry 3.0
After a few decades of the twentieth century electronic devices using transistor,
integrated circuits chips made automated individual machines to supplement the use
of operating workers. Later, developments were made in the software side to control
the electronic hardware. By this embedded mechanism, the components are inte-
grated where material planning is made by using enterprise resource planning tools
where humans can plan, schedule, and track products from industry to the customer.
The concept of supply chain management formalized the result of extended geo-
graphic dispersion.
As the third revolution resulted in huge development in computers and informa-
tion and communication systems, the industries lean on the digital technology in
production that offered a Digital Revolution, which led to a change from analog and
mechanical systems to digital systems. The digital era is also known as
Information Age
Industry 4.0
In the twenty-firstt century, the industrial revolution 4.0 gets connected with IoT
using manufacturing techniques to enable systems for sharing information and ana-
lyze and use it for performing intelligent actions. It creates a cutting-edge technol-
ogy in additive manufacturing, robotics, artificial intelligence and other cognitive
technologies, advanced materials, and augmented reality, according to the article [6].
The development of new technology Industry 4.0 was first developed during the
later stages of the twentieth century. Manufacturing execution systems, shop floor
control, and product life cycle management were farsighted concepts that lacked the
technology needed to make their complete implementation possible. Although
“Industry 4.0” is the common term referring to the fourth industrial revolution, aca-
demics still argue to appropriately define the methodology (Fig. 11.2).

Applications of IIoT

1. Industrial Automation
Industrial automation is the most substantial and common application of
IoT. Automation systems make machines and tools to function in an efficient way
166 V. Saranya et al.

Fig. 11.2 Industrial evolution

by using sophisticated software tools that act as an observer and make developments
for succeeding process rehearsals. A greater level of improvement has been shown
by machine automation with accuracy. Automation tools such as PLC (Programmable
Logic Control) and PAC (Programmable Automation Control) use smart sensor net-
works associated with a central cloud system, which collect huge amount of data.
Application-oriented software are designed to analyze the data that improves accu-
racy, efficiency, error control, and easy access and reduces man power in specific
tasks in industrial automation. The industrial IoT provides solutions in connecting
industries with major components such as machines, tools, and sensors for network
connection. These components provide effective solutions to check the status, ship-
ment, schedule maintenance, and also monitor the process flow overtime.
2. Smart Robotics
Intelligent robots are developed by the industries based on the required specifica-
tion and are created by IoT-enabled industries. By utilizing these robots, manufac-
turing units ensure smooth handling of tools and materials with efficiency and
accuracy. Robots are designed to perform complex tasks with high embedded sen-
sors that analyze real-time environment as an interface for man-machine interac-
tion. Robotics networks are connected to cloud in a secure way for monitoring and
controlling.
Engineers and designers can be able to access and analyze the data to yield quick
actions for product enhancements and also prevent any unexpected failure due to
machine failure.
3. Predictive Maintenance
Modern industrial machines are equipped with smart sensors for continuous
monitoring of the current status of major components for detecting any critical
issues before the system gets completely failed. The smart sensors notify even the
maintenance warning to the centralized system and also send alert messages to the
11 An Evolution of Innovations Protocols and Recent Technology in Industrial IoT 167

responsible teams for further processing. For every maintenance procedure, engi-
neers are allotted with an analyzed schedule maintenance plans which do not affect
the routine task and also unnecessary problems in the production line. Some unex-
pected machine failure may happen which leads to damage of products, delay in
delivery, and also loss in business for producers. But in recent years, the develop-
ment of technology has made maintenance of machines on real-time basis by stor-
ing the overall performance to a cloud system. History of each machines helps to
process a predictive maintenance system.
4. Integration of Smart Tools/Wearables
Workforce can perform its task with improved accuracy and efficiency by inte-
grating smart sensors in tools and machines. Specially planned wearables and smart
glass support workforces to reduce error and expand safety at the functioning envi-
ronments. Smart wearables can be able to intimate the users by giving instant alert
messages to employees during critical situations such as gas leak or fire. Wearables
can be designed for monitoring health conditions of individuals continuously.
5. Smart Logistics Management
Logistics is one of the major areas in many industries which needs continuous
development to support increasing demands. Smart technology provides a perfect
solution to solve complex operations and manage the delivery of goods to the cus-
tomers efficiently. Marketing monsters such as Amazon uses drones to deliver goods
to their customers. Advanced technologies such as drones offer better efficiency,
accessibility, speed and require less manpower. However, initial investments are
huge compared to conventional methods and implementation has limitations.
Airline is an added major industry that uses IoT for performing its daily opera-
tions at the manufacturing and predictive maintenance of planes in service. At the
manufacturing plant, airline companies use IoT solutions to track thousands of com-
ponents required for every single day at work. Unified management of portfolios
helps to manage its supplies effortlessly.
6. Software Integration for Product Optimization
Smart analytics of any IoT system enhances the possibilities of the system for
optimization and improvement. Many industries and major companies implement a
customized software for deep analysis of huge data collected by the sensor networks
and machines. The analyzed data gives a better overview of process improvement
strategies for optimizing the products. The analysis of data and its behavior patterns
provide cost-effective solutions in a period of time. As though analysis of large
amount of data was very hard, inaccurate and more time consuming before the
introduction of software tools.
7. Smart Package Management
The manufacturing units deliver its products without any damage by using IoT
technology for package management, which gives a lot of convenience and
168 V. Saranya et al.

e­ fficiency. The smart sensors embedded on the packages could be able to detect the
vibrations, atmospheric conditions such as temperature and humidity, and also they
can be able to monitor the stages of packages and update the status periodically.
Finally, the product delivery to customers is confirmed by sending messages.
8. Autonomous Vehicles
Industries use automotive vehicles using the IoT technology by enabling self-­
driving system to carry goods within the company premises. These vehicles can be
able to detect obstacles along its path and find another route to reach its prescribed
destination using the GPS technology and wireless technology to communicate with
the control station [7].
Pros of Industrial Internet of Things
• Enhanced accuracy
• Optimization of product and process
• Analytical preservation and exploration
• Advanced competence
• Distant accessibility and observing
• Improved security
• Scalability of network
• Reduced energy consumption

Challenges in IIoT

Business becomes a constant battle for balancing the benefits of new technology
with increased risks using sophisticated data. However, things are different in the
case of critical infrastructure. Before adopting the IIoT with smarter components, it
is wise to think about some challenges to overcome with efficient benefits.
1. Security: Major Concern
Increased usage of digital connectivity in IIoT deployment sets to lot of security
risks. As the machines and other collected data are placed in a common cloud envi-
ronment, it leads to higher risks creating cyber attacks against the critical infrastruc-
ture. Even though the records are maintained in a common repository, all the security
measures have to be placed by using protocols.
2. Preserving Visibility of “Things”
Tracking of things make network engineers to build systems that show network
visibility where the intruders can be able to easily analyze the valuable goods by
getting a clue from either the customer or the suppliers. But the network could get
started from the initial zero to potentially thousands of connected sensors and
devices. It is really very hard to manage visibility for a single factory floor, but the
Industrial Internet of Things will add additional challenges as you scale it globally.
11 An Evolution of Innovations Protocols and Recent Technology in Industrial IoT 169

3. Protecting the Industrial Ethernet

The industrial Ethernet creates a highway for building an IIoT environment
where 100% internet connectivity is needed to perform each task. Hence, mainte-
nance of systems can be made only through online services, but to go offline for
maintenance can be guaranteed by using proper cables only for permanent usage
with a logic of no data loss even during the connectivity issues.
4. Unified Legacy Systems and IIoT Infrastructure
A unified legacy system environment using IoT infrastructure focuses less on
security and more attention on performing optimizing operations and third-party
integrations. But in case of IIoT environment, a new modernizing infrastructure is
developed by adding smart technologies, which focuses individually on security,
third-party intrusion, and on all stages of processing.
5. Data Storage
As large amount of data being generated from different sources for any analysis
to be made ensures rapid development of IIoT. All the raw data cannot be stored in
the IIoT devices due to its less memory size and limited energy level. If the network
is deployed in an unreachable or unreliable area, it may cause vulnerable threats
leading to computational problems. To avoid the challenging aspects such as data
processing, secure data storage, efficient data retrieval, and dynamic data collection
in IIoT can be solved by designing a flexible and economic framework using for
computing and cloud computing techniques.

Innovations and Recent Technologies in IIoT

The new innovations and implementation of recent technology have made the IIoT
industries to enhance their process by increasing the production, marketing, and
other business fields [8]. The new innovations can be able to provide some list ser-
vices to the industrial group.
Collaborative Enterprise
In the current scenario, companies undergo high pressure from customers, stake-
holders, and global competitors. To improve the performance, there must be an
operational improvement focusing on business process and collaborative concepts.
A collaborative management model (CMM) provides information on the complex
interactions, applications, collaborations, and processes that an enterprise demands.
This framework model is a basis for organizing and controlling the key business
processes in an enterprise (Fig. 11.3).
Cybersecurity for Industries and Smart Cities
Cybersecurity secures the information and conditions transmission through the net-
work. An effective integration of other technologies creates its own boundaries
increasing the reliability. Cybersecurity concerns are obstructing determinations to
170 V. Saranya et al.

BUSINESS
OPERATIONS
RETIRE

SUPPLIERS
LIFECYCLE

COLLABORATIVE
ENTERPRISE

CUSTOMER

DESIGN
SUPPLY CHAIN

AUTOMATION

Fig. 11.3 Collaborative enterprise

implement new approaches for lower costs and more-competitive business

approaches [9–13]. Cybersecurity Advisory Service can help to address the chal-
lenges through perceptive reports, individual consultations with experienced ana-
lysts, and peer networking opportunities. This provision provides an ultimate way
to learn what others are doing, how they are doing it, and the benefits they are
achieving. It also enables the consumers to keep updated on the several new devel-
opments occurring in cybersecurity technologies, practices, and principles
(Fig. 11.4).
Digital Transformation
A successful digital transformation promotes an organization in a process that trans-
forms its current state to a desired future state. In case of production operations,
there involves moving from a local, soloed, procedural operating model to a more
comprehensive model that is integrated to the enterprise, responsive to customers
and market changes, and optimized in the context of overall business performance.
The transformation is made based on the following (Fig. 11.5):
1. Customer-focused Innovation
2. Transformation framework to digital transformation
3. Technologies, security, and architecture
4. Business justification and solution selection
11 An Evolution of Innovations Protocols and Recent Technology in Industrial IoT 171

ICS Supplier / Cyber Part-time Plant Full-time Plant Full-time ICS

Service Provider ICS Staff ICS Staff Cyber Ops Group

Cybersecurity Management Solutions

Anomaly &
threat
Incident

Program cost
Level of Protection

intelligence
Zone management
Physical Unidirectional
Firewalls, ics
Security, gateways,
device Antivioate
Asset firewalls,
Manage
Inventory. anti-malware
Device
Contain
Defend
Secure

Program Maturity

Fig. 11.4 Cybersecurity management system

TO:
Customer-Focused Innovation
21ST Century Operation
Effective but Sub-
optimal Operation Identify New Boundary BUSINESS
PERFORMANCE
PERFORMANCE INCREASES

EXCELLENT Technologies, Security and Design

GOOD
SATISFACORY
MARGINAL Deploy, Build Expertise, Create
POOR

LOCAL,SILOED, TRANSFORMATION INTEGRATED,RESPONSIVE,

PROCEDURAL PROCESS OPTIMIZED

Fig. 11.5 Digital transformation

Protocols in IIoT

The IIoT infrastructure uses client/server protocols for transmitting data. These pro-
tocols are used to show how the request and response roles are executed and define
the technique that is to be monitored in a smart industrial environment.
• OPC UA
172 V. Saranya et al.

OPC Unified Architecture (OPC UA) is a machine-to-machine communication

protocol used for industrial automation designed and developed by the OPC
Foundation. Some distinguished characteristics are as follows:
• Emphasis on connecting with manufacturing tools and data gathering
• Open source and freely accessible and maintainable under GPL 2.0 license
• Cross platform not linked with the operating system or system language
• Service-oriented architecture (SOA)
• Essential complication
• Robust safety
The vendors and organizations model their complex data to a service-oriented
architecture of OPC UA protocol based on integral information model for forming
statistical infrastructure. This protocol is used by industries such as pharmaceuti-
cals, oil and gas, building automation, industrial robotics, security, manufacturing,
and process control for functioning and performing optimistically.
• HTTP (REST/JSON)
REST stands for Representational State Transfer that depends on a displaced
client-server and primary infrastructures. In most of the cases, the transmission is
made using the HTTP protocol. JSON (JavaScript Object Notation) is a lightweight
data-interchange format that makes easy and user friendly for reading and writing
for the users.
• MQTT [14]
MQTT (Message Queuing Telemetry Transport) is an ISO standard publish-­
subscribe-­based messaging protocol. It works on the upper layer of the TCP/IP
protocol. It is measured for connections with distant locations with limited network
bandwidth. An MQTT system contains customers connecting with a server fre-
quently called as “broker.” A client may be an originator of data or a subscriber. All
the clients can get connected with a broker for performing a task.
To access an information, the procedure is arranged in the form of a hierarchy
where a publisher can distribute a new data set and control messages are sent to the
broker. The brokers distribute the information to the clients who need to access it.
The publisher need not have the information about the data subscribers and vice
versa. While publishing the client first gets connected with a broker and sends a
default message as an acknowledgment to the subscribers if the brokers detect the
publishing client (Fig. 11.6).
• CoAP [6]
Constrained Application Protocol (CoAP) is a protocol that is used to specify
how the low-powered computing constrained devices can be operated in the IoT
environment. This protocol was designed by Internet Engineering Task Force
(ITEF) for enabling simple devices and controlling them to join IoT by using
constrained network with low bandwidth and availability. CoAP functions as
11 An Evolution of Innovations Protocols and Recent Technology in Industrial IoT 173

CLIENT C

Subscribe “temperature”

BROKER

Subscribe “temperature”

CLIENT A CLIENT C

Fig. 11.6 Example for MQTT protocol

REST

TP Client C
HT
CoAP
HTTP
Proxy C
Server
C
CoAP CoAP
C C

Server

Constrained Environments
The Internet

Fig. 11.7 CoAP architecture

aHTTP for constrained devices that perform as a sensor or an actuator used in

machine-to-­machine (M2M) communication. Finally, this protocol is designed
for better reliability with low bandwidth and high congestion with reduced over-
head (Fig. 11.7).
174 V. Saranya et al.

Fig. 11.8 AMQP system

• DDS:
The Data Distribution Service (DDS) is a networking standard that links the sen-
sor nodes to the cloud server. DDS uses a connectionless data model with the ability
to publish and subscribe to data with the desired quality of service (QoS). A DDS-­
based system does not interact with the application systems. A DataBus exists in the
structure that robotically determines and attaches publishing and subscribing applica-
tions. The DataBus contests and imposes QoS. DDS disables problems related with
system-to-system integration, such as lack of scalability, interoperability, and the
ability to evolve the architecture. It allows plug and play simplicity, scalability, and
especially high performance. Possibly the best way to understand DDS is to inspect
the systems that use it. Applications extend to the areas such as healthcare, energy,
defense, transportation, industrial automation, and communications industries.
• AMQP
AMQP stands for Asynchronous Message Queuing Protocol, which is an open
standard that uses binary application layer protocol designed for message-oriented
middleware for communicating producers, brokers, and consumers through mes-
sages that increase coupling and scalability. The AMQP enables a wide range of
submissions and assemblies to work together, irrespective of their inner designs,
regulating enterprise messaging on an manufacturing scale (Fig. 11.8).

Conclusion

Thus, the IoT ecosystem has been adopted by the existing industries for process
improvements, better quality management, cost–effectiveness, and improved effi-
ciency. The upcoming industries utilize the strength of IoT infrastructure for prod-
uct optimization and using big data analytics from thousands of tiny sensors for data
11 An Evolution of Innovations Protocols and Recent Technology in Industrial IoT 175

gathering. Industrial IoT is fast growing technology with limitless possibilities for
future industries and manufacturing units. This chapter briefly discusses the innova-
tions, challenges, and protocols used by the IIoT systems.

References

1. Hofmann, E., & Rüsch, M. (2017). Industry 4.0 and the current status as well as future
­prospects on logistics. Computers in Industry, 89, 23–34.
2. Ungurean, I., & Gaitan, V. G. (2014). An IoT architecture for things from industrial environ-
ment. In: Communications (COMM), IEEE 2014 10th International Conference on (pp. 1–4).
3. Gorecky, D., Schmitt, M., Loskyll, M. et al. (2014). Human-machine interaction in the Industry
4.0 era. In: The 12th IEEE International Conference on Industrial Informatics (INDIN), Porto
Alegre, Brazil (pp. 289–294).
4. Lu, Y. (2017). Industry 4.0: A survey on technologies, applications and open research issues.
Journal of Industrial Information Integration, 6, 1–10. Available at: http://www.sciencedirect.
com/science/article/pii/S2452414X17300043. Accessed May 14, 2017.
5. McKinsey Digital. (2015). Industrie 4.0. How to navigate the digitization of the manufacturing
sector, Berlin, Düsseldorf, and Munich. McKinsey.
6. “Industry 4.0 and manufacturing ecosystems” by Deloitte University Press
7. Gossman, W. E., & Hartmaier, P. J.. (2001). System and method for providing data to a wire-
less device upon detection of activity of the device on a wireless network. US Patent 6317594
B1, 13 Nov 2001.
8. Muhtaroglu, F. C. P., Demir, S., Obalı, M., & Girgin, C.. (2013). Business model canvas per-
spective on big data applications. In: Proceedings of the 2013 IEEE International Conference
on Big Data, Santa Clara, CA, USA (pp. 32–37).
9. Lee, E. A. (2008). Cyber physical systems: Design challenges. In: The 11th IEEE interna-
tional symposium on object oriented realtime distributed computing (ISORC), Orlando, Fl
(pp. 363–369).
10. Kang, W., Kapitanova, K., Son, S. H., et al. (2012). RDDS: A real-time data distribution ser-
vice for cyber-physical systems. IEEE Transactions on Industrial Electronics, 8(2), 393–405.
11. Zhang, Y., Xiang, Y., & Huang, X. (2017). A cross-layer key establishment model for wire-
less devices in cyber-physical systems. In: Proceedings of the 3rd ACM Workshop on Cyber-­
Physical System Security, CPSS@AsiaCCS 2017, Abu Dhabi, 2 Apr 2017 (pp. 43–53).
12. Hwang, J.-J., Chuang, H.-K., Hsu, Y.-C., & Wu, C.-H.. (2011). A business model for cloud
computing based on a separate encryption and decryption service. In: Proceedings of the 2011
International Conference on Information Science and Applications, Jeju Island, South Korea
(pp. 1–7).
13. Carter, C. R., & Ellram, L. M. (2003). Thirty-five years of the journal of supply chain
management: Where have we been and where are we going? The Journal of Supply Chain
Management, 39(2), 27–39.
14. https://info.smartindustry.com/si16-203-lp-kepware
Chapter 12
Smart Meter: A Key Component
for Industry 4.0 in Power Sector

M. Kathiresh and A. Mahaboob Subahani

Introduction

Electric power has become indispensable to human survival and progress. Grid is
the electricity system that comprises of power generation, power transmission,
power distribution, and power consumption. In conventional power grids, electric
power is transmitted from a few centralized power generating stations to a large
number of distribution centers with electricity users or customers [1]. A smart grid
is a new variant of electric power grid under development, which allows unconven-
tional energy flow and bidirectional data flow to create an advanced automatic and
decentralized power distribution network. Apart from the efforts to meet the grow-
ing demand, automation in the energy distribution is also necessary to enhance
people’s life standard [2].
There are basically two types of energy meters, namely, electromechanical and
electronic meters. The electromechanical induction meter operates by counting the
revolutions of an aluminum disc, which is made to rotate at a speed proportional to
the power consumed. The number of revolutions is proportional to the energy usage.
The voltage coil consumes a small and relatively constant amount of power, typi-
cally not greater than 2 W, which is not registered on the meter. The current coil
similarly consumes a small amount of power in proportion to the square of the cur-
rent flowing through it, typically up to a couple of watts at full load, which is regis-
tered on the meter [1–3].
The metallic disc is acted upon by two coils. One coil is connected in such a way
that it produces a magnetic flux in proportion to the voltage and the other produces
a magnetic flux in proportion to the current. The field of the voltage coil is delayed
by 90 degrees using a lag coil. This produces eddy currents in the disc, and the effect

M. Kathiresh (*) · A. M. Subahani

Department of Electrical and Electronics Engineering, PSG College of Technology,
Coimbatore, TN, India

© Springer Nature Switzerland AG 2020 177

G. R. Kanagachidambaresan et al. (eds.), Internet of Things for Industry 4.0,
EAI/Springer Innovations in Communication and Computing,
https://doi.org/10.1007/978-3-030-32530-5_12
178 M. Kathiresh and A. M. Subahani

is such that a force is exerted on the disc in proportion to the product of the instan-
taneous current and voltage. A permanent magnet exerts an opposing force propor-
tional to the speed of rotation of the disc. The equilibrium between these two
opposing forces results in the disc rotating at a speed proportional to the power
being used. The disc drives a register mechanism that integrates the speed of the
disc over time by counting revolutions, much like the odometer in a car, in order to
render a measurement of the total energy used over a period of time for a single-­
phase AC electricity meter. Additional voltage and current coils are required for a
three-phase configuration.
Horizontal aluminum rotor disc is visible in center of meter. The aluminum disc
is supported by a spindle which has a worm gear that drives the register. The register
is a series of dials which record the amount of energy used. The dials may be of the
cyclometer type, an odometer-like display that is easy to read where for each dial a
single digit is shown through a window in the face of the meter, or of the pointer
type where a pointer indicates each digit. With the dial pointer type, adjacent point-
ers generally rotate in opposite directions due to the gearing mechanism. The
amount of energy represented by one revolution of the disc is denoted by the symbol
which is given in units of watt-hours per revolution.
Electronic meters display the energy used on an LCD (liquid crystal display) or
LED (light-emitting diode) display, and can also transmit readings to remote places.
Apart from the amount of electricity used, the electronic electric meter type can also
record other parameters of the load and supply such as maximum demand, power
factor, and reactive power used. Some of the advanced kinds of electric meters also
include electronic clock mechanisms to compute a value, rather than an amount of
electricity consumed, with the pricing varying by the time of day, day of week, and
seasonally.
The technology used in most of the solid-state electric meter type is the use of a
current transformer to measure the current. The main current-carrying conductors
need to pass through the meter itself and so the meter can be located remotely from
the main current-carrying conductors, which is a particular advantage in large-­
power installations.
The modules involved in the conventional and smart metering system are shown
in Fig. 12.1. Traditional meter reading is inefficient to meet the future residential
and industrial developmental needs. Hence, there is an increased demand for auto-
matic meter reading systems, which collect meter readings without any human
intervention [3]. Collection of meter reading is also inefficient because a meter
reader has to physically be onsite to utility environment. Traditional electromechan-
ical meters used today are prone to drift over temperature. This method of collecting
of meter readings becomes more problematic and costly when readings have to be
collected from vast and often scattered rural areas. The older electromechanical and
present electronic meter readings are taken by persons involved in the job, and the
amount is noted down in the electricity board card. This may consume a lot of time
and doubtlessly require many human resources. Fraudulent customers could cheat
without being detected for several years. The tariff for the energy consumption is
12 Smart Meter: A Key Component for Industry 4.0 in Power Sector 179

Fig. 12.1 Components of conventional meter and smart meter

calculated based on the readings taken by people from the electricity board. Hence,
it is more prone to error.
Smart meters are the next generation of meters, which can replace existing
energy meters and offer a range of benefits, both for the individual electricity and
gas consumer and for the network systems in general. Through automated record-
ing, actual energy usage over short intervals and automated communicating of
metering data to the network operator, the smart meter eliminates the need for a
home visit to manually read the meter and facilitate more cost-efficient network
operations [4].
Electricity suppliers will be able to offer innovative pricing arrangements for
customers to support the efficient use of energy (i.e., time-of-use tariffs). But the
smart meter also offers potential advantages directly to customers. Smart meters can
be an alternative to the concept of estimated billing system, meaning customers will
only be billed for the electricity they actually use and smart meters allow easier
switching of energy supplier to get the best deals. However, the most immediate
consumer benefit of the smart meter is the potential to receive more frequent and
more detailed information on how much energy is being used (also called feed-
back). For instance, households with a smart meter will be offered to receive con-
sumption and cost statement from the supplier every 2 months, hereinafter referred
to as the home energy report. At the specific request of the consumer, the grid opera-
tor will forward the smart metering data on a more frequent (e.g., daily) basis to the
energy supplier – or another service provider – for additional analysis of the con-
sumer’s energy consumption over the previous days, weeks, and months.
180 M. Kathiresh and A. M. Subahani

For monitoring of real-time energy consumption, customers can add a wireless

communication interface, in-home display or online application to their smart
meter. Products and services are emerging on the market for this purpose. In order
to support product and market development, functional requirements for the com-
munication interface have been established in legislation. The grid operators have
also established technical requirements for this purpose.
Smart meters can read real-time energy consumption information including the
values of voltage, phase angle, and the frequency and securely communicates that
data. The ability of smart meters for bidirectional communication of data facilitates
the process of collecting information regarding the electricity fed back to the power
grid from customer premises. Smart meters can communicate and execute control
commands remotely as well as locally. They can be used to monitor and also to
control all home appliances and devices at customer premises.
Smart meters are usually capable of much more than this by including the ability
to communicate with in-home appliances, programmable communicating thermo-
stats, and other loads. Smart meters can record consumption data at intervals (as
frequently as every minute) and automatically transfer the information to the utility
over a secure network [5]. Various communication architectures, including point-to-­
multipoint and mesh networks, have been implemented. This network, often in con-
junction with a backhaul layer, provides two-way connectivity between the utility
and the meter. The network also supports pushing signals to the meter, which could
be used for an “on demand” reading to confirm power restoration after an outage, or
to notify the meter of an upcoming “super peak” event. When the meter is notified
of an event, the information is communicated to appliances inside the customer’s
home using the meter’s onboard GSM radio chip or other means. Based on pre-
defined preferences or rules, different appliances can act to reduce or shift con-
sumption in preparation for or during the “super peak” events.
They can measure electricity consumption from the grid, support decentralized
generation sources and energy storage devices, and bill the customer accordingly.
Smart meters can be programmed such that, only power consumed from the utility
grid is billed while the power consumed from the distributed generation sources or
storage devices owned by the customers are not billed. Smart meters can limit the
maximum electricity consumption, and can terminate or reconnect electricity sup-
ply to any customer remotely.
A smart meter system employs several control devices, various sensors to iden-
tify parameters and devices to transfer the data and command signals. In future
electricity distribution grids, smart meters would play an important role in monitor-
ing the performance and the energy usage characteristics of the load on the grid.
Collection of energy consumption data from all customers on a regular basis allows
the utility companies to manage electricity demand more efficiently and also to
advise the customers about the cost-efficient ways to use their appliances. Smart
meters can be programmed to maintain a schedule for operation of the home appli-
ances and control operation of other devices accordingly.
In addition, integration of smart meters helps utility companies in detecting
unauthorized consumption and electricity theft in view of improving the distribution
12 Smart Meter: A Key Component for Industry 4.0 in Power Sector 181

efficiency and power quality. Design of future electricity markets are aimed at pro-
viding their customers with highly reliable, flexible, readily accessible, and cost-­
effective energy services by exploiting advantages of both large centralized
generators and small distributed power generation devices. In addition, distributed
generation would be an essential integral part of future household energy systems.
Utility companies try to identify more profitable customers to provide optional
value-added services, as smart meters can identify such customers based on the
distributed generation sources and overall power consumption.

Smart Metering for Developing Countries

In many developing countries, conventional energy meters are used for billing the
energy consumed by customers. For ease of operation of the home appliances, mon-
itoring the grid, improving the power quality, improved load sharing, detecting non-
technical losses, and other implied advantages, smart meters are to be introduced in
developing countries. Power utility companies worldwide lose about 20 billion dol-
lars each year because of nontechnical losses. In addition, growing nontechnical
losses because of theft and billing irregularities force the utility companies to imple-
ment a transparent and genuine metering system. However, deployment of smart
grid and smart meter system involves huge budgets. It would be very difficult for
utility companies to invest billions of dollars on an infrastructural upgrade that has
no direct return on the investment. Hence, smart meters with minimum required, but
essential features may be designed for implementation in countries with weaker
economy. Moreover, smart meters might not be implemented for luxury in opera-
tion, but they must be introduced in order to fight the basic problems that power
utility companies and its customers face [6, 7].
Smart meters with great networking capability and advanced software tools are
difficult to tamper and hack, which improves the distribution efficiency. Integration
of smart meters enhances facilitation of decentralized generation and power storage
devices. In the near future, total energy demand is expected to become double the
current electricity demand. In view of this situation, many developing countries do
not have resources for the additional capacity addition. To fill this gap, apart from
increasing the installed generation capacity, controlling the electricity theft and
regularizing the existing electricity customers can manage the load within demands.

Benefits of Smart Meters over Conventional Meters

The smart meter installation brings quite evident benefits to the utilities in a medium
or short span of time. It provides a more timely and precise billing and reduces the
costs of interacting with customers by activating, closing, or suspending contracts
with no personnel displacement. It also offers significant help in avoiding
182 M. Kathiresh and A. M. Subahani

e­ lectricity-­related frauds, and finally it offers the potential for better power peak
control and distribution. The interest of electrical energy distributors is therefore
self-evident, and it is reasonable to assume that, even for other energy sources (gas,
water, and heat), the utilities could find it technologically sound and profitable to
extend a similar smart metering philosophy, even though some delay can be expected
due to the lack of such a preexisting physical interconnection as it is the low-voltage
power line for the electricity meter [8].
On the other hand, alternatives (e.g., wireless solutions) are available, and regu-
latory authority decisions can offer additional obligations or incentives. Nonetheless,
a complete exploitation of smart metering advantages is only achieved when an
efficient local interface between the smart meter and the household is accomplished.
Benefits can then be extended to the community and to the final user. As a matter of
fact, the interconnection of energy users and the ability to easily provide data related
to energy consumption are crucial opportunities for the management of such a criti-
cal area as the energy saving. After the Kyoto Protocol, even the most recent
20-20-20 agreements converge at implementing policies aimed at saving energy and
reducing CO2 emissions.

Design of a Smart Meter Prototype

In the smart meter, potential transformer (PT) and current transformer (CT) are used
for measuring current and voltage respectively. The sensed sinusoidal signal from
CT and PT is given to the signal conditioning unit where the signal is rectified or
clamped to a reference voltage of half of its peak magnitude in order to shift the
negative half into positive side. The rectified signal is then given to a microcon-
troller for processing. The microcontroller calculates the power consumption using
the voltage and current signals, and the power factor, which is obtained by sampling
the magnitude of the voltage and current.
Energy is calculated by counting the power consumption time using the inbuilt
timer and multiplying it with the power consumption. Apart from this, prescribed
peak load time from the consumer will be noted and alarm will be given to the con-
sumer. During the peak hours, consumers can turn ON or turn OFF the loads
remotely through the Internet Gateway module (ESP8266). Figure 12.2 shows the
various functional blocks of the smart meter system.

Current Measurement

The circuit shown in Fig. 12.3 is used for current measurement, which measures the
total current drawn from the supply. The secondary side of CT produces current
which is reduced and proportionally varying to the primary current to be measured.
This current is converted to voltage with the help of a shunt resistor called burden
12 Smart Meter: A Key Component for Industry 4.0 in Power Sector 183

Fig. 12.2 Structure of the smart meter prototype

Fig. 12.3 Current measurement circuit

resister. The shunt resister voltage is rectified by an Op-Amp precision rectifier in

order to have a circuit behaving like an ideal diode or rectifier. The Op-Amp preci-
sion rectifier is a combination of a half-wave precision rectifier and a summing
amplifier. It operates by producing an inverted half-wave-rectified signal and then
adding that signal to the original signal in the summing amplifier.
When the input is positive, the diode D1 is ON and D2 is OFF, so both the op-­
amp1 and op-amp2 act as an inverter. Hence the output voltage is equal to the input.
When the input is negative, the diode D1 is OFF and D2 is ON, and the differential
184 M. Kathiresh and A. M. Subahani

input to the op-amp2 is zero. Hence, the output voltage is equaled positive of the
input voltage. This circuit is also known as absolute value circuit as the output is
always positive for both positive input and negative input.
Then the output of the rectified voltage is adjusted to 0–5 V with the help of a
variable resistor. Then, output ripples are filtered by the capacitor connected across
the output. The output DC voltage is given to the microcontroller for further pro-
cessing. The microcontroller calculates the actual current value using Eq. 12.1.
(assuming that the ADC reference voltage is 5 V and the output resolution is 10 bit).

 311 
I ADC  
I rms =  1023  (12.1)
2

Voltage Measurement

Figure 12.4 shows the circuit diagram for voltage measurement which measures the
supply voltage. The secondary side of PT produces voltage which is reduced and
proportionally varying to the primary voltage to be measured. This voltage is recti-
fied and filtered by the same Op-Amp precision rectifier circuit that was discussed
in the above section.
Then the output of the rectified voltage is adjusted to 0–5 V with the help of a
variable resistor. Subsequently, output ripples are filtered by the capacitor connected
across the output. The output DC voltage is given to the microcontroller for further

Fig. 12.4 Voltage measurement circuit

12 Smart Meter: A Key Component for Industry 4.0 in Power Sector 185

processing. The microcontroller calculates the actual current value using Eq. (12.2)
(assuming that the ADC reference voltage is 5 V and the output resolution is 10 bit).

 311 
VADC  
Vrms =  1023  (12.2)
2

Power Factor Calculation

Essentially, power factor is a measurement of how effectively electrical power is

being utilized in a power system network. The higher the power factor, the more
effectively electrical power is being utilized. It is the relationship between active
power and the total power consumed or apparent power.
The power factor is also defined as the cosine of the phase angle difference
between the voltage and current waveforms. As resistive loads do not produce any
phasor difference between the voltage and current, all the useful power is delivered
directly and converted to heat. Hence, the power consumed by a resistive load is the
real power which is fundamentally the circuit’s average power, so the power factor
is unity [9].
On the contrary, the reactive loads such as inductor and capacitor produce a pha-
sor difference between the voltage and current. So the power factor will not be unity.
In addition, the utility loads are not pure resistive loads at any time. Hence before
the power measurement, it is very much needed to calculate the power factor [9, 10].
The idea of calculating power factor is to count the time difference (Td) between
the zero crossing instant of both the voltage (Tv) and current (Ti), and convert the
time into degree (θd) using the Eqs. (12.3) and (12.4), respectively. This is the pha-
sor difference between the voltage and current, taking the cosine value of it will give
the power factor as given in Eq. (12.5).

Td = Tv − Ti (12.3)

θ d = 360° fTd (12.4)

Where f is power supply frequency

PF = cosθ d (12.5)
186 M. Kathiresh and A. M. Subahani

Active Power and Energy Calculation

The active power is calculated in Watt, by multiplying the voltage, current, and
power factor that is given in Eq. (12.6). Active power measurement is made for
every short interval, and they are summed for a period of time as given in Eq. (12.7).

Pi = Vrms,i I rms,i PFi (12.6)

Pi +1 = Pi + Pi +1 (12.7)

Once the active power is calculated for a period of time, the energy is calculated
in kWh using Eq. (12.8), and the total energy consumption is calculated by the
cumulative sum of energy consumption over a period of time using the Eq. (12.9).
The total energy consumption in kWh can be used for calculating the number of
units consumed and in turn the tariff if the per unit cost is known.

Pi × time
E p,i = (12.8)
1000 × 3600

E p,i +1 = E p,i + E p,i +1 (12.9)

Reactive Power and Energy Calculation

The reactive power is calculated in VAR, by multiplying the voltage, current, and
sine of the phasor difference between them as given in Eq. (12.10). Reactive power
measurement is done for every short interval, and they are summed for a period of
time as given in Eq. (12.11).

Qi = Vrms,i I rms,i sin θ d (12.10)

Qi +1 = Qi + Qi +1 (12.11)

Once the reactive power is calculated for a period of time, the energy is calcu-
lated in VARh using Eq. (12.12), and the total energy consumption is calculated by
the cumulative sum of energy consumption over a period of time using the Eq.
(12.13). The total reactive energy consumption in VARh can be used for calculating
the penalty for industrial customers if it exceeds the limit.

Qi × time
E p,i = (12.12)
3600
12 Smart Meter: A Key Component for Industry 4.0 in Power Sector 187

EQ ,i +1 = EQ ,i + EQ ,i +1 (12.13)

Real-Time Clock (RTC)

RTCs are designed for ultra-low power consumption as they usually continue run-
ning when the main system is powered down. The low power DS3231 RTC module
is used as real-time clock which can work either in 24-hour mode or 12-hour mode
with AM/PM indicator. It automatically adjusts for months fewer than 31 days
including leap year compensation up to year 2100. It also comes with built-in power
sensing circuit which senses power failures and automatically switches to back up
supply of 3 V CMOS Battery. The address and data are transferred from RTC to the
microcontroller serially through an I2C, bidirectional bus.

Load Control Unit

A relay is a type of actuator that provides connection between two terminals or

points in association with a control signal. In simple terms, a relay is an electrical
switch in which a small electrical current is required to control the switching opera-
tion (ON and OFF). The relays, depending on the principle of operation, can be
classified as electromechanical relays and solid-state relays. Electromechanical
relay uses an electromagnet to achieve the mechanical action of the switch. A small
current given to the coil wound on an iron core will energize the coil, and the con-
tacts of the relay move to on position. When the coil is de-energized, the contacts
move back to off position.
In contrast to electromechanical relays, which consist of moving parts and mag-
netic flux, a solid-state relay or SSR does not consist of any moving parts. A solid-­
state relay consists of semiconductor devices such as transistors, thyristors, SCRs,
or TRIACS to achieve the switching operation. So in the prototype developed,
800 V, 10 A, rated SFS2004/59 TRIAC is used for connecting and disconnecting the
lamp load with the power supply.

Working of the Smart Meter Prototype

The smart meter prototype consists of potential and current transformers,

PIC19F877A microcontroller, IoT gateway (ESP8266), relay control unit, signal
conditioning unit, RTC module, and LCD display. Once the smart meter is powered
up, it initializes its modules. The rms value of voltage and current is calculated by
reading the values from PT and CT, respectively. The power factor is calculated by
188 M. Kathiresh and A. M. Subahani

measuring the phase difference between the voltage and current. Then the instanta-
neous power is calculated by multiplying all the parameters that are calculated. This
results in the instantaneous power consumption of the load. This process is repeated
for few minutes and the cumulative power is calculated in watts. Total energy
­consumption is calculated from the total power and the time till which the power is
calculated in kWh. The energy consumption data is sent to smart grid server through
an IoT gateway which can help to monitor the total power consumption or demand
in any particular area/region. The tariff is calculated by multiplying it with the per-­
unit cost at the smart grid server, and the same is sent to the customers’ mobile phone.
The parameters such as voltage, current, power consumption, and total tariff are
displayed in the LCD display. The same is sent to the customers’ mobile through
SMS on request. The workflow of the prototype is shown in Fig. 12.5. The load
demand data from the server is fed to the smart meter continuously to enable the
meter for optimizing the load pattern of the customer. With the help of real-time
clock, the peak hours at which the power consumption from the grid is high is iden-
tified and the customer is given alert about the start time of peak load. Hence, the

Fig. 12.5 Workflow of the

smart meter prototype
12 Smart Meter: A Key Component for Industry 4.0 in Power Sector 189

customer can turn OFF the loads based on his need which will save him from higher
tariff rate. In addition, suggestions are given to the customers for task scheduling for
optimum usage of energy to save tariff.

Simulation Results

Using Proteus 8 simulator, the prototype model of smart meter is simulated. In the
simulation, PIC microcontroller is used which is interfaced with voltage and current
measurement unit, signal conditioning unit, and load control devices. All the param-
eters such as voltage, current, power factor, and power consumption data are dis-
played in 20x4 LCD as shown in Fig. 12.6.
As mentioned in the above section, the prototype developed is given with the total
load demand data at periodic intervals, by which it identifies the peak hour and alerts
the customer. The same is displayed in the LCD as shown in Fig. 12.7a. Once the
peak hour is identified, the smart meter starts optimizing the loads that are currently
connected to the grid. During this optimization, power consumption of the daily
routine task of the customer is also taken into account along with the individual load
demand data. The smart meter then prioritizes the tasks based on all the abovemen-
tioned parameters and the suggestions are given to the customer to reduce the load
either by turning OFF any load or by rescheduling the task. In the simulation results,
the same condition is also verified during the peak hour, and the suggestion made by
the smart meter is highlighted in Fig. 12.7b. For simulation purpose, random data are
used for load demand and the power consumption of daily routine tasks.

Fig. 12.6 Simulation result of the smart meter prototype

190 M. Kathiresh and A. M. Subahani

Fig. 12.7 Simulation result of the smart meter prototype. (a) When peak hour is identified. (b)
Suggestions given to the customer for optimum energy use
12 Smart Meter: A Key Component for Industry 4.0 in Power Sector 191

Hardware Results

The prototype of smart meter is developed in the laboratory which is shown in

Fig. 12.8. It consists of power supply module, measuring unit (PT and CT), zero
crossing detector (ZCD), signal conditioning unit, microcontroller, real-time clock,
IoT gateway, and LCD display.
For experimental purpose, lamp loads are used, and TRIAC is used for load con-
trol. The power consumption and the other parameters are displayed in LCD as
shown in Fig. 12.9a. The load demand data is given to the prototype to verify the
response in peak hour. The smart meter identified the peak load demand and dis-
played the warning command in the LCD which is shown in Fig. 12.9b. Further, the
power consumption and tariff details are sent to the customers’ mobile as SMS
along with the peak hour alert as shown in Fig. 12.10.

Fig. 12.8 Experimental setup of the smart meter prototype

192 M. Kathiresh and A. M. Subahani

Fig. 12.9 Display command of the smart meter prototype. (a) Power consumption. (b) Peak our
alert

Fig. 12.10 Intimation to

the customers’ mobile
12 Smart Meter: A Key Component for Industry 4.0 in Power Sector 193

Implementation of Smart Meter with Prepaid Facility

An attempt is made to develop a smart meter with prepaid facility like the facilities
available with mobile networks and direct-to-home (DTH) services. In the prepaid
smart meter model, the customer has to recharge the balance by purchasing the
recharge cards. The smart meter calculates the total energy consumption and the
equivalent tariff for every 1 minute, and the tariff amount is debited from the bal-
ance. In this scheme, a minimum balance is fixed, and as long as the balance is
greater than the minimum balance, the process continues. If the balance is less than
the minimum balance, the customer is given alert and prompted to recharge. If the
balance becomes zero, then a warning message is sent to the customer giving him a
grace period of 30 minutes to recharge. If the meter is not recharged even after the
grace period, the supply is disconnected for the customer with an intimation mes-
sage. Once this happened, the customer has to contact the energy provider for fixing
the issue. He will be given power supply with a penalty that is derived by the supply
provider. Algorithm for the prepaid smart meter is given below.
Algorithm for prepaid smart meter
Start
Step 1: Clear all registers
Step 2: Initialize Timers T0 (interval in min.), T1
Step 3: Data acquisition
Voltage conversion:
Start Conversion
Read Digital value and convert it to its equivalent analog value
Find the sum of the squares of the voltage samples ∑ 1 vn 2
n

Increase the count ‘n’

Current conversion:
Start Conversion
Read Digital value and convert it to its equivalent analog value
Find the sum of the squares of the current samples ∑1 in 2
n

Increase the count ‘m’

Go to: Step 3
Step 4: Detect positive zero crossing of Voltage:
Start timer T1 n

Find Voltage rms (Vrms= ∑vn n )

2

Increase cycle count ‘cv’

Return
194 M. Kathiresh and A. M. Subahani

Step 5: Detect positive zero crossing of Current:

Stop timer T1
360°
Calculate phase angle difference and power factor ( θ = × t d ; PF = cos θ)
n
0.02
Calculate current rms (Irms= ∑in n )
2

1
Increase cycle count ‘ci’
Step 6: If timer T0 = 0
Calculate the following
Average voltage rms=Vrms/cv
Average current rms=Irms/ci
Average PF = cos θ/ci
Real Power = VrmsIrms cos θ in kW
Number of units is equal to [Real Power × Time period (interval)]
Calculate the tariff from the number of units
Display all the values calculated above
Step 7: Check the available balance if > tariff
Check the available balance if < Rs. 10/-
Send the customer a warning message to recharge
Subtract the tariff from the prepaid balance
Go to: Step 1
Step 8: If the balance < tariff
Intimate the customer that the supply will be cut off in 5 minutes
Set the timer (T1) to 30 min.
Calculate the cumulative tariff
Step 7: If the timer T1 = 0
Check if the balance < tariff
Turn off the supply
Stop
Else go to: step 7
The flowchart of the smart meter with prepaid facility is shown in Fig. 12.11. The
customer is supplied with energy till the available balance is greater than the tariff
amount calculated. If not, the smart meter will execute the subroutine A till the
grace period. Here the grace time is set to 30 min. If the customer recharges the bal-
ance within the grace period, then the smart meter does nothing, but continues to
supply energy to the customer by entering into the point B. But if he does not
recharge the balance, the customer is disconnected from the grid immediately and
intimation is sent to the customer.
12 Smart Meter: A Key Component for Industry 4.0 in Power Sector 195

Fig. 12.11 Flowchart of the smart meter with prepaid facility

Smart Meter Applications

The function of smart meter can be further extended to the following applications
such as outage detection (OD), Automated Control of Home Appliances (ACHA),
demand response (DR), distributed generation systems (DGs), and electronic vehi-
cle charging (EVC).
196 M. Kathiresh and A. M. Subahani

References

1. Gao, C., & Redfern, M. A. (2011, September 5–8). A review of voltage control in smart grid
and smart metering technologies on distribution networks. In Proceedings 46th International
Universities’ Power Engineering Conference, Soest.
2. Barai, G. R., Krishnan, S., & Venkatesh, B. (2015). Smart metering and functionalities of
smart meters in smart grid – A review. In 2015 IEEE Electrical Power and Energy Conference
(EPEC), London, ON (pp. 138–145). https://doi.org/10.1109/EPEC.2015.7379940.
3. Bansal, P., & Singh, A. (2016). Smart metering in smart grid framework: A review. In 2016
Fourth International Conference on Parallel, Distributed and Grid Computing (PDGC),
Waknaghat (pp. 174–176). https://doi.org/10.1109/PDGC.2016.7913139.
4. Barman, B. K., Yadav, S. N., Kumar, S., & Gope, S. (2018). IOT based smart energy meter
for efficient energy utilization in smart grid. In 2018 2nd International Conference on Power,
Energy and Environment: Towards Smart Technology (ICEPE), Shillong, India (pp. 1–5).
5. Luan, W., Peng, J., Maras, M., Lo, J., & Harapnuk, B. (2015). Smart meter data analytics
for distribution network connectivity verification. IEEE Transactions on Smart Grid, 6(4),
1964–1971. https://doi.org/10.1109/TSG.2015.2421304.
6. Al-Ali, A. R., Landolsi, T., Hassan, M. H., Ezzeddine, M., Abdelsalam, M., & Baseet, M.
(2018). An IoT-based smart utility meter. In 2018 2nd International Conference on Smart
Grid and Smart Cities (ICSGSC), Kuala Lumpur (pp. 80–83). https://doi.org/10.1109/
EPETSG.2018.8658501.
7. Hafez, O. (2017). Time of use prices considering smart meters and their future implementation
in Saudi Arabia smart grid. In 2017 Saudi Arabia Smart Grid (SASG), Jeddah (pp. 1–5). https://
doi.org/10.1109/SASG.2017.8356504, doi: https://doi.org/10.1109/ICSGSC.2018.8541314
8. Morello, R., De Capua, C., Fulco, G., & Mukhopadhyay, S. C. (2017). A smart power meter
to monitor energy flow in smart grids: The role of advanced sensing and IoT in the elec-
tric grid of the future. IEEE Sensors Journal, 17(23), 7828–7837. https://doi.org/10.1109/
JSEN.2017.2760014.
9. Emanuel, A. E. (1993). Apparent and reactive powers in three-phase systems. European
Transactions on Electrical Power, 3, 7–14.
10. Haben, S., Singleton, C., & Grindrod, P. (2016). Analysis and clustering of residential custom-
ers energy behavioral demand using smart meter data. IEEE Transactions on Smart Grid, 7(1),
136–144. https://doi.org/10.1109/TSG.2015.2409786.
Chapter 13
Role of Software-Defined Network
in Industry 4.0

Vakaavinash Reddy and K. Venkatesh

Introduction

Dueto the innovative advances in the industry, the present IoT gadgets are
­progressively tending to convert smart by furnishing them with a facility to con-
nect the internet to accumulate real-time data. This advancement has created a
revolution in most of the verticals in the industry, for example, Agriculture, Health
Care, Manufacturing, Energy and Minerals, and furthermore Automobile by
empowering them to equip IoT sensors for easy access information from any place
on the planet. The development of Cloud Computing made it easy to coordinate
them on a solitary common shared platform not only to log the information but
also to control these gadgets remotely by providing Over the Air (OTA) software
package updates. These smart devices can be incorporated and controlled utiliz-
ing smartphones as well and it was anticipated that the total number of associated
IoT gadgets hits up to 25 billion by 2020 [1]. As the number of associated gadgets
is extending progressively and becoming smart, the current underlying network
infrastructure should also be improved to dodge delay in data packet transmission
and make the network even smarter by knowing the importance of data carried by
each packet. The priority for critical data packets must be recognized ahead and
routed so that the Quality of Service (QoS) will not be undermined for critical IoT
operations.
In an Internet of Things (IoT) system, the connected end devices are named as
‘motes’. These motes accumulate the information from the encompassing environ-
ment by the means transducers and sensors that are associated and transmit the
information to the cloud platform or some other feedback mechanisms with the goal

V. Reddy (*) · K. Venkatesh

Department of Information Technology, SRM IST, Kattankulatur, Chennai, India
e-mail: [email protected]

© Springer Nature Switzerland AG 2020 197

G. R. Kanagachidambaresan et al. (eds.), Internet of Things for Industry 4.0,
EAI/Springer Innovations in Communication and Computing,
https://doi.org/10.1007/978-3-030-32530-5_13
198 V. Reddy and K. Venkatesh

that the response can be passed to the servo motors and actuators to control the
overall system. These motes are further associated with the border router for s­ ending
and receiving information over the internet. Most of the organizations that provide
cloud service, such as Google, Microsoft, and Amazon, have propelled an innova-
tive service called IoT as a service that allows clients to associate and work with
their motes on top of the online Cloud platform [2]. These service providers have
additionally presented their IoT Development boards with inbuilt sensors and
Artificial Intelligence (AI) gadgets like Google Home, Echo, and Alexa, which
empower the users to direct their gadgets through voice commands and speech,
which apparently indicates that the future revolution in technology is going to occur
in the IoT.
Software-Defined Networking (SDN) application development in IoT is an
approach to manage wireless sensor networks and distributed computing, which
supports the administration and enables automatic effective framework designing
practices to ensure the overall aim to optimize performance and monitoring [3].
Incorporating SDN is expected to deliver the way in which the static design of tra-
ditional frameworks is decentralized and complex while existing systems need all
the simpler troubleshooting, monitoring, and versatility. SDN introduces a unified
framework insight on a central system by decoupling the routing method (control
plane) of the system from the packet forwarding process (data plane) and uniting it
with IoT-enabled devices [4]. The control plane includes a minimum of one control-
ler which operates as a brain for SDN network where the entire intelligence is
united. Yet at the same time, the centralization has its own specific drawbacks con-
cerning security and elasticity.
This chapter mostly focuses on complications that ascend in IoT while forward-
ing critical traffic and probabilities to include Software-Defined networking in
IoT. It also explains device configuration in the network to acquire appropriate
topology and routing table in the SDN controller. The process to incorporate one of
the sensors like Temperature, LDR, PIR – Motion, and Humidity sensors into motes
will also be discussed. The information that was gathered will be stored in the cloud
platform or passed to the feedback system to regulate the encompassing environ-
ment of the motes, i.e., for example, if there occurs a fire accident near environment
of the mote, then it generates critical data packets that are given higher preference
than the additional motes that are producing less prior traffic so that the condition
can be controlled and neutralized with least delay.
The essence of this chapter can be addressed as follows. Section “Technologies
Incorporated” presents concise information on the hardware equipment and soft-
ware that are incorporated. Section “Related Works” gives a brief on how SDN
application implementation visualizes. The foremost problem is represented in
Section “The Problem Definition”. Section “Architecture and Test Environment”
gives information on the architecture for applying SDN Technology in IoT. The
intended solution is explained in Section “The Proposed Solution”. The obtained
results for the proposed solution are highlighted in Section “Result and Discussion”.
The conclusion followed by the future work is shared in Section “Conclusion”.
13 Role of Software-Defined Network in Industry 4.0 199

Technologies Incorporated

To include Software-Defined Networking (SDN) in the Internet of Things (IoT)

there are a few open-source hardware and software requirements that are claimed
for information collection from sensors and to route the data packets to the server
running on top of the cloud platform. The facts about those components are dis-
cussed below.

Raspberry Pi Model 3

Raspberry Pi v3 is a low-cost pocket-sized portable mini-computer that runs with

5 V micro USB power supply. It can run the Linux-based Debian operating system
in which the hardware comes with 1GB of RAM and a quad-core processor of
1.2Ghz with built-in wireless network adapter that can be utilized in both client
mode and Access Point mode [5]. It can be used in any type of embedded applica-
tion system wirelessly. Raspberry Pi comes with multiple versions, but version 3
series or further is recommended for installing OpenWRT base ovs package.

OpenWRT OVS

OpenWRT is an open-source project, it is a customizable router firmware based on

Linux kernel that can be easily deployed on any Linux system platform and it pro-
vides a variety of application-specific bundles. Open vSwitch (OVS) is a package
inside the Unified Configuration Interface (UCI) that can be downloaded and
installed through CLI or GUI [6]. A Raspberry Pi 3 SD memory is flashed with
OpenWRT firmware obtained from OpenWRT compatible device downloads page
on the website and accessed GUI through LAN default gateway IP address in the
device associated machines web browser. As Raspberry Pi 3 comes with an inbuilt
wireless network adapter, it can be configured either in client mode or Access Point
mode. Initially, to download the required packages it will be configured in client
mode and later changed to AP mode to allow IoT devices to connect with the inter-
net. In the case of Raspberry Pi-2 model B, a USB wireless network adapter is
inserted in any one of the available USB ports and installed in the adapter-specific
wireless-supported drivers from the UCI package repository (Fig. 13.1).
To perform SDN operations, the OpenWRT router control plane is decoupled
from the forwarding plane by creating a virtual bridge through the ovs-vsctl com-
mand. The location and port of the SDN controller are specified through virtual
switch set-controller command and switch connectivity status changes to true. The
interfaces that are targeted to control with SDN are associated virtually in ovs and a
static IP address is given to each device connected in the network. The Backbone
200 V. Reddy and K. Venkatesh

Fig. 13.1 Connection Setup of OpenWRT OVS in Raspberry Pi 3

internet connects to Raspberry Pi LAN port, and the IoT devices are associated with
Wlan0 interface that was already set to AP mode and the devices start appearing on
the ONOS controller GUI. Initially, when the devices are connected to the virtual
bridge, they lose internet connectivity as the virtual network is not yet created and
the connected devices are decoupled from the IP Stack. Once the reactive forward-
ing in the SDN controller is activated, the devices will start communicating with
each other and after assigning the IP address, subnet, and default gateway the end
devices can communicate over the internet.

ESP-32 LWS

ESP32 is reliable and trustworthy in industrial contextures, it is an open-source IoT

Development board built on ESP-8266 wireless module with operating calefaction
extending from −40 to +125 degrees centigrade. It can abolish external circuit
defects and adjust to changes in external conditions embedded in Wearable, mobile
electronic gadgets, and the Internet of Things applications, ESP32, achieve very low
consuming power with a combination of various dedicated packages [7]. It also
includes many features like overclocking, different modes, and wide ranges of
power. ESP32 is immensely integrated with embedded network adapters. IoT appli-
cations using ESP32 require minimal circuit board designs.
13 Role of Software-Defined Network in Industry 4.0 201

LibWebSockets (LWS) is an easy-to-use, light-weight 1 MB package developed

in C for executing advanced network protocols. By the integration of lws with
HTTP/2, it generates a self-signed certificate when the device is booted for the first
time supports secured data transmission over SSL and allows more intelligent man-
agement of packet streaming, compression of headers, priority and multiplexing
queries [8]. The LWS allows to add up to four access points after flashing with esp-­
idf and connects automatically to available AP that is configured on boot up when
other APs are down or unreachable. Once the devices are connected to the internet,
it is possible to update the applications in the device Over the Air (OTA). The appli-
cations that run inside the device are different from the factory image and it is an
isolated package that parts the internal disk into individual sectors for factory image
and the applications. It is possible to implement the HTTP/2-based secure tunneling
in LWS package [9], which adds extra layer security to the IoT devices. This also
provides multiple benefits such as enhanced performance and memory
consumption.
The esp-idf is a package that can be obtained from GitHub repository and
flashed in ESP-32 device. The installation requires a 64-bit Linux toolchain, an
integrated bundle that consists of all the necessary software tools. In case of win-
dows, it is recommended to download and install MSYS2, which provides Linux
environment in windows machine. It consists of a package manager (Pacman) to
obtain all the required toolchain related to esp-idf. Before using the make in Linux
it is required to install CMake version greater than 3.0 from the Pacman and gen-
romfs package placed in the root directory and added to the system variables and
upgrade the git repository. To flash the esp-idf, it is required to connect the device
to the machine using USB and specify the /dev/ttyUSB0 or respectively connected
USB port and execute the flashing command. Once the device boots up, it creates
an access point named ESP to configure the device to deploy application and OTA
update settings to monitor the serial port it is required to install termite RS232
terminal software.
This chapter mainly concentrates on priority-based packet stream management
using HTTP/2 protocol in LWS framework. User code of LWS not be customized
upon the using event loop by providing a generic timer of high resolution. It is
accessible for Unix style platforms such as ESP32, Windows, Linux, and even
BSD. LWS provides scalability to several thousands of connections at a time.

Open Network Operating System (ONOS)

Open Network Operating System (ONOS) was built on top of Java with Apache
Karaf license and gives control panel to software network, network managing parts
like links and switches, and software running programs to deliver communication
services to adjoining networks and end hosts. It consists of management services for
software-defined networks and customized communication routing [10]. Software
that runs on this can be found in Use cases and Apps. Over multiple servers, it runs
202 V. Reddy and K. Venkatesh

as a distributed system which allows using it in memory resources and CPU [11].
There are two types of membership for this, one is Partner and another one is
Collaborator, with wavering stages of commitment. This software has been pre-­
owned as a platform, in which applications have been integrated into neighbouring
projects or have been mentioned on top. It resolves all the disruptions in the soft-
ware stack without failure by controlling the networking operation [12]. ONOS
provides greater flexibility in handling data packets from the IoT devices and vali-
dating the data first by understanding the importance of the data. Also, the data path
is controlled using SDN to route the data packets and manages the large-scale IoT
devices in the SDN-enabled switch. In this project, an application was built on top
of ONOS to enable priority-based critical traffic forwarding in IoT using LWS and
HTTP/2 protocol in ESP-32 devices.

Sensors

IoT system has sensors or devices, which through some type of connectivity com-
municate with the cloud. Once the information reaches the cloud, it is processed by
software and decides to function on actions like showing an alert or adjusting the
sensors automatically without the user’s need. Internet of Things also refers as the
Internet of Everything, which consists of devices enabled with internet that gather,
convey, and react on information which is given. Sensor data is the device’s output
that notices and answers to some kinds of input from the encompassing environ-
ment. Sensors are used to measure physical quantities. There are several types of
sensors that are divided based on the quantities. They are Potential or radio or mag-
netic or Electric current sensors, Humidity sensor, Flow or fluid velocity sensors,
Pressure sensors, Heat or Temperature or Thermal sensors, Proximity sensors,
Optical sensors, and Position sensors.
As there are a wide-range of sensors, we can gauge all the physical changes in
the external environment around us. In this project, we use a couple of elementary
sensors that are broadly incorporated in each consistent day to day life such as
DHT111, a temperature, and Humidity sensor, for producing real-time traffic in this
testbed. It operates on power ranges from 3.5 to 5.5 V. It can determine 0–50 degrees
centigrade of temperature with some precision of ±2 degrees centigrade and humid-
ity which relatively ranges from 20–95% with the perfection of ±5%. It gives fully
measured digital outputs for two measurements. It has its proprietary protocol of
1-wire and interaction between sensor and microcontroller, which is impossible
with a direct interface of its peripherals. The protocol should be executed in the
MCU firmware with accurate required timing by the sensor. This sensor consists of
4 I/O pins that are connected to Esp-32 wireless module General Purpose Input
Output pins.
13 Role of Software-Defined Network in Industry 4.0 203

Related Works

In this chapter [13, 14], an architecture was predicted to support the small-scale
proving ground to put SDN in the (IoT) Internet of Things. As it is pricier for acquir-
ing the programming development boards like NetFPGA, the author builds up a
testbed with Raspberry Pi and achieved Open vSwitch (OVS) with definitive proto-
col OpenFlow as a substitute productive setup. The products obtained are correlated
with the outcome with both Raspberry Pi switch and NetFPGA-1G and proven that
the network throughput is approximately the equivalent in both development boards.
The advancement of data transmission technologies, smartphones, Nanoelectronics,
cloud-based networking, things to the internet any place and at any moment, and
physical devices enable connecting people. This is known as the (IoT) Internet of
Things. At present, so many applications are being advanced for all round of IoT
scenarios, like traffic control, industrial process monitoring, health care, home auto-
mation, civilian, earthquakes warning, and military operations. These applications
in various domains claim their own committed systems and platforms. Most of the
IoT services are supported by various systems and platforms, and a system or plat-
form is correlated with specific application domains. This concludes in dedicated
systems and platforms, individually consisting of many actuators and sensors, a
group of gateways, a computing center, etc. In addition to this, IoT service providers
reconfigure or design the protocols for communication and devices for every appli-
cation and for every user. Also, they must manage a huge amount of inessential data.
Software-defined network (SDN) technologies promote the building of network ser-
vices. The reasonably integrated controller and the consistent southbound interfaces
bring a key but dynamic architecture to provide basic functions and interfaces which
can rapidly develop the user applications. Furthermore, incompatibility of data in
various formats and models can be dynamically controlled over and easily achiev-
able. Consequently, the applications in the various domains can be provided conve-
niently and accurately by the same system.
IoT structure consists of four components such as service, platform, network,
and device; the service that supports the interface and communicates with the users,
which connected with the platform to provide customized services to the users.
There are several types of platforms in IoT which are data analysis, service develop-
ment, device, and service platforms. For example, the device platform supports the
execution of the services environment and development for users. The act of IoT
platform is to support and execute IoT services. In the past, the platform was devel-
oped in a closed manner. However, industries and governments are refining assorted
IoT platforms as open-source platforms. One of the reasons for this tendency is to
improve the speed to get in the market, decrease the development cost and enhance
the quality of software. By opening the software platform, many developers are
given the opportunity to provide their effort. This will ultimately accelerate the
development process and reduce cost. Moreover, the service development platform
supports developing toolkits to users for them to develop IoT services easily. In the
204 V. Reddy and K. Venkatesh

end, it provides the production and implementation of different applications.

Including service and platform, one of the important components of IoT is the net-
work. It helps to transmit data among devices, contents, services, and users. The
network should have the capability to process, control and manage huge amounts of
network traffic. In addition, security and privacy play a crucial role in IoT. Each
stage of data transmission should have a security solution to secure it from errors.
The basic components within the IoT service domains are apprehended behavior,
the accomplishment of real-time awareness of the physical environment, and help
with a human in charge through a wide range of determination and data visualiza-
tion. IoT not only connects sensor devices and produce the data for an intention but
also targets the computerization and expansion within the contemporary systems.
In addition, the data should be examined in a manner of acceptable modelling
approach by understanding the significance of the data. This, the expected IoT ser-
vices, such as health monitoring, smart homes, smart grids, smart cities, and smart
traffic systems, have already integrated these fundamental models of computeriza-
tion and resource expansion for any environment. The act of IoT devices is not
defined to collect data, they should connect with assorted networks to provide a
wide range of services. For example, IoT devices require the capability to interact
either node-to-node or node-to-Web, which depends on whether the target node is
in their own network. The flow for developing an IoT device is in open-source or
open-hardware approaches for developers and manufacturers. For instance, IoT
device production supplies basic standard architecture. With the basic standard or
level architecture, users and developers can develop their own IoT devices.
In [15], an architecture with the horizontal model for developing real-time ready-­
to-­use applications on the Internet of Things (IoT) using a software-defined network
(SDN) technology was proposed. The author assesses the practicability to execute
real-time IoT applications on top of the Cloud platform using Software-Defined
Networking (SDN) in intermediate routers and gateways. The testbed was built up
using Beacons (Bluetooth Radio Transmitters) that are associated with the Raspberry
Pi switch operated by SDN controller in both north and southbound interfaces. The
data exchange and the packet generation are made between smartphones that are
connected. The information related to data exchange in SDN northbound and south-
bound interface through JSON format was explained clearly. In the JSON format
packet data unit (PDU), priority is one of the elements which will be used as a base
to route critical traffic in this chapter. An IoT (Internet of Things) architecture based
on Software-defined Network (SDN) technologies facilitates allocating different
resources, along with data, devices, and software, between different applications at
various points. Incomparable data can be provided internally in the network, and
advanced services and applications in various domains can be supplied accurately
and rapidly.
OpenFlow is an approved interface which admits researchers to precisely control
how packets are routes in actual Software-defined network (SDN). It is based on
Ethernet switch but controls an open protocol which can be pre-owned to program
the flow table in different routers and switches. And it also organizes three factors
13 Role of Software-Defined Network in Industry 4.0 205

such as Flow table, OpenFlow protocol, and secure channel. Flow table contains the
flow access that determines how data flows are handled in the network. Through
flow entries, data flows can actively be adapted and broadcast in the network frame-
work. A secure channel is pre-owned as communication means between the SDN
switch and controller to organize a protective connection. OpenFlow protocol sup-
ports a basic interface that can be characterized externally by researchers, therefore
eradicating additional programming in switches. The achievable ways to reduce the
abeyance of critical traffic between two End-to-End IoT nodes in the identical net-
work such as remote command robotic arm used during tele-surgeries as the traffic
will be a crucial thing during the live operations. The simulation was accomplished
using OpenDayLight SDN Controller, Mininet, and OpenFlow vSwitch. The results
are appraised based on the ping and probe packets, but when a real-time critical data
packet generates through the internet, many issues such as delay in MQTT Server
response, congestion in border router due to other nodes, and Packet drop will arise
due to connectivity errors.
In [16–18], it is explained that Wireless Sensor networks have a hidden motive
of the Industrial IoT process and casting industrial automation-based applications.
The development of IIoT accesses high potential for exfoliating device network
and acquiescence with combined and regulated planning from low-level device
activities to high-level significant communications of application data. This chapter
delivers an industrial development desires program definability in the basic archi-
tectural components of IIoT, along with sensor cloud services, network infrastruc-
tures, wireless field devices, and IIoT gateways. Software-defined network is to
deal with necessary challenges in an integrated IIoT system, such as accuracy,
security, well-­timed scalability, and quality of service. Advanced IIoT system
design is expected based on the current networking technology like SDN scheduled
to solve the critical traffic issues [16]. With the addition of IoT, communicability
will not be barred to only smart devices. Based on Raspberry-Pi architecture, all the
technical details about the IoT network in problems raised in industries are thor-
oughly analyzed.
The industrial growth in WSNs and installed systems has authorized several
applications in various domains along with IoT systems. A distinctive class of
Internet of Things authorized industrial management systems was proposed in IIoT,
which maintains adequacy and economic advantages to install the system, maintain-
ability, inseparability, and scalability. The complete design of IIoT systems and
wide-ranging of IIoT applications desire an adjustable software definable efficiency
for solving necessary challenges [17]. A new Software-defined IIoT design with its
main segments depends on low power technical WSNs. Also, an expected architec-
ture for software-defined activities in field devices, gateways, and sensor to send
data to the cloud platform is not proposed. The quality of service can individually
rule out the data flow programming equivalent to the active network conditions and
adopt the SDN controller bond to asset the finest route for the actual time data com-
munication. The network can be dynamically handled on critical data packet routing
procedures and prioritization of service will be covered as part of this chapter.
206 V. Reddy and K. Venkatesh

The Problem Definition

In a very few years, the Internet of Things (IoT) has gone from innovation or set of
advances that were cutting-edge to the circumstance today where incorporated in all
the fields. If in case the IoT ecosystem has an issue or is presented to shortcomings,
at that point the industries that are associated with it are similarly weakened. Truth
be told while handling critical traffic is one of the serious issues affecting the
improvement, there are various issues that stem specifically from this. IoT is trans-
forming into its advancements as associated gadgets wind up more intelligent and
progressively vivid, and desires to change over IoT information to a stream of
knowledge to perceptions that boost economic value [19]. Also, the rapid adoption
of IoT is driving the data acquisition and sensor costs down, allowing larger feasible
business use cases that are been too costly earlier.
In IoT, connecting such a diverse number of devices is one of the greatest chal-
lenges and that will contradict the basic structure of current correspondence models
lacks essential improvements. While the IoT yet possesses the ability to evolve into
something in the prospects to improve the demand for businesses likewise, for those
beforehand adopted this next generation technology still required to achieve some-
what in underlying network infrastructures. IoT had broadened its outskirts from the
automated industrial sensors to wearable devices where the fundamental connecting
network base is yet the same. The gadgets are intelligent enough to recognize
changes in the encompassing environment however, the network devices are not
quite intellectual to recognize the significance of the information they were routing.
This makes a boundary between connecting network and innovation progression.
The Systems administration of extensive robotized machines is an ongoing con-
centration for modern computerization and one of the challenges is the availability
with conventional network equipment that isn’t intended to help more than connec-
tivity of the local network. Modern systems can be exceptionally decentralized,
inflexible and complex to oversee because of the tight coupling of the mechaniza-
tion information and control plane that is regularly inserted inside the hardware. The
registering and correspondence nodes are regularly arranged separately when the
plant is set up and interconnections stay static from that point. The conventional
mechanical interchanges progressive structure comprises of three system levels
with different organizing advancements and conventions that limit what can be
accomplished and includes multifaceted nature due to limited setup.
In the present IoT framework, we depend on centralized, server and client
approach [20] using the SDN-enabled switches in which the servers are deployed on
top of the Cloud platform and the client devices are IoT gadgets that are connected
wirelessly and placed physically in local areas from where the data need to be gath-
ered. The IoT client gadgets are associated with servers utilizing the internet.
Figure 13.2 demonstrates a manufacturing industry containing different zones are
furnished with multiple real-time sensors that are connected to the cloud server by
the border gateway router to increase the overall productivity [21].
13 Role of Software-Defined Network in Industry 4.0 207

Fig. 13.2 Critical Nodes in an Industrial IoT Ecosystem

As we clearly observe the preceding smart industry there exist a few nodes in
which the overall productivity is based on sensors that have a critical role and sen-
sible towards significant data acquisition. Any suspension or delay of data move-
ment from these sensors and the applications that are associated with it are thus
impaired. The Hydraulic pressure sensor, temperature sensor fitted in carving
machine, piezometer sensor in robot welding arm, and the smoke detector equipped
in the warehouse are the crucial nodes that can arise a critical data packet at any
moment. Alternatively, the surveillance camera, humidity, and carbon dioxide sen-
sors stay broadly general-purpose nodes which are not that significant but connected
to the same router.
At the point, if there is an emergency state in any one of the nodes in any zone,
for instance, a decrease in the hydraulic pressure due to leakage or overheating
208 V. Reddy and K. Venkatesh

problem in the carving machine or a fire accident in the warehouse, these sensors
generate a critical data packet to the monitoring center in the remote location. Even
though there is a criticalness in the real-time environment, the border router handles
this critical traffic as ordinary data and forwards it to the cloud server this also cre-
ates the delay as the other non-critical nodes also overflow the border router simul-
taneously with their traffic [20]. These significant data packets from critical nodes
need to be given a prior consideration by the border router so that the latency [19]
can be reduced.

Architecture and Test Environment

To achieve the overall goal to reduce the latency of critical traffic, the OpenWRT
control plane is separated from the IP Stack that is data plane in the border router
[22] by utilizing the Software-Defined Networking (SDN) controller tool Open
Network Operating System (ONOS) (Fig. 13.3).
OpenWRT-based Open vSwitch installed in Raspberry Pi remains as the south-
bound interface for SDN by following the steps to install in Linux that are displayed
in OVS website.

Fig. 13.3 Testbed Architecture

13 Role of Software-Defined Network in Industry 4.0 209

Fig. 13.4 OpenWRT Open vSwitch configuration

The Raspberry Pi’s wlan0 was connected to a virtual port created in ovs-bridge
named as br-ovswlan configurated in the Access point (AP) mode. The eth0 port is
connected to backbone internet connectivity using RJ45 ethernet port in rasp-
berry pi.
The virtual switch is interfaced with SDN controller using ovs-vsctl set-­controller
br-ovswlantcp:ipaddress:port as shown in Fig. 13.4 and once the controller is up
and running the status can be seen through ovs-vsctl show command. The physical
ports are added to the virtual bridge using virtual [22] port using the ovs-vsctl add-­
port br-ovswlan wlan0. When the ESP-32 devices are arranged in the wireless sig-
nal extent on boot up they are programmed to connect Pi access point automatically.
If the esp devices are away from AP they search for other nearby APs that are con-
figured through lws-based esp-idf in their range to transfer the data with the AP to
the cloud server.
The ONOS is installed in a Linux PC or a VM operating on the Linux platform
by following the steps mentioned in the ONOS website [6]. In case of controller
installed on virtual machine, an additional virtual network adapter was attached in
hypervisor and configured in bridged mode with external Wi-Fi adapter to allow
OpenFlow standard protocol for packet delivery by enabling reactive forwarding
application in onos controller and specify the flow rules so that the OVS running on
OpenWRT Pi will start associating with onos and the topology of connected ESP
devices in southbound devices will start to appear on the onos graphical user inter-
face (GUI) (Fig. 13.5).
210 V. Reddy and K. Venkatesh

Fig. 13.5 Topology of IoT devices in SDN Controller

The onos controller requires the built-in ovs device driver application to be
installed and started to interface with ovs-db. The GUI topology viewer displays the
ovs switch and the number of devices connected to it. The devices are provided with
a static IP address, as there is no concept of routing and switching in SDN, all the
virtual devices are considered same to forward packets from the source to the desti-
nation [22] but onos also allows to use the remote DHCP server in the network using
built-in DHCP Server app. The ovs switch is assigned an IP address and specify the
default gateway to connect the southbound IoT devices to the internet. To allow the
devices to start communication with the outside network the reactive forwarding
application in onos is installed and started.
OpenFlow is the essential southbound protocol in the SDN control plane.
OpenFlow works at time scales near the information arrange a round trip time
(RTT), typically milliseconds, and is essentially used to program stream tables that
control packet sending utilizing the match-activity display [23]. ONOS’s sending
observations principally use OpenFlow to program switch packet stream tables.
OpenFlow likewise gives some constrained configuration capacities, (for exam-
ple, bringing a port up or down, or designing meters); however, many required setup
activities [24], advising a change which controllers associate with, empowering the
OpenFlow specialist on a switch, or empowering OpenFlow on explicit ports are
unequivocally out of extension for OpenFlow.
13 Role of Software-Defined Network in Industry 4.0 211

Virtual switches that require arrangement past what OpenFlow gives may need to
utilize other southbound conventions, for example, NetConf, ovsdb, SNMP, and so on
[25]. ONOS right now gives drivers to a few southbound arrangement conventions.
Later, ONOS may give setup reflections which enable gadgets to be designed in a non-
exclusive, without agonizing over explicit conventions and their fluctuating semantics
crosswise over gadgets. In this testbed we implemented OpenFlow v1.4 and the flow
table obtained during the packet forwarding as shown in Figs. 13.6 and 13.7.
To monitor the real-time critical traffic from the IoT devices, a monitoring dash-
board was created with some parameters which can measure the latency of critical
traffic that has been generated from sensitive nodes in the IoT network. This dash-
board also plots a real-time graph across time taken in milliseconds across time.

Fig. 13.6 OpenFlow-based OVS Flows in SDN Controller

Fig. 13.7 Critical Traffic monitoring dashboard on Cloud Platform

212 V. Reddy and K. Venkatesh

The IoT devices are capable to obtain precisely synchronized time in millisec-
onds using Network Time Protocol (NTP) servers to make sure the accrued time at
which the request was posted to a cloud database we have accessed the NTP Server
services from ESP-32 devices with Unix EPOCH timestamp. The NTP services
provide an option to specify offset time to match the local time zone. It was also
observed that there is a slight delay about 1500 ms between the NTP and current
time due to the UDP services in NTP causes a minor difference. To balance this
variance the offset was increased to 1500 ms in IoT devices. The server is also
capable of obtaining Unix time and the time taken is calculated from the difference
between Sent Time and Received Time in milliseconds.
The real-time sample data was captured from IoT sensors and presented it on the
Monitoring Dashboard. The parameters considered are Sent Time, Received Time,
Sensor value, and Priority. Time Taken is calculated based on these parameters. As
shown in Fig. 13.8, the parameters are obtained from the Non-SDN nodes and it was
observed that the Time Taken in milliseconds for High priority data is the same
compared to non-priority data.
In case of data from SDN-incorporated devices, the Time Taken in milliseconds
is reduced even on same internet speed.
As shown in Fig. 13.9, the Time Taken for high priority data is less compared to
Non-SDN device critical traffic.

The Proposed Solution

The main intention is to reduce the abeyance of critical traffic. First, we will analyze
the central nodes in the network that can trigger critical state at any time (Fig. 13.10).

Fig. 13.8 Sample Data of Non-SDN Traffic on Monitoring Dashboard

13 Role of Software-Defined Network in Industry 4.0 213

Fig. 13.9 Sample SDN Data Captured on Monitoring Dashboard

Every IoT node in the network will have inferior and superior sensor values. For
example, the inferior for heart rate monitoring sensor is 55 beats per minute and
superior is 105 beats per minute. When the sensor reports a value that is external of
this range, then that will be deliberated as a critical state [26]. At first, the priority
value will be set to zero and when there is a critical state the priority did replace one
in the packet header. The sensors are graded to the limits to set priority depending
on the obtained values and move onward to the SDN controller.

functiondispatchPacketData ()
priority = 0
inputsensordata
loop while sensordata # 0
ifsensordata>= 105 OR sensordata<= 55 then
priority = 1
end If
if priority > 0 then
path = MinimumTimeDelay
else
path = Regular
end If
outputCloudDatabase
end while
end function

The controller oversees the packet priority and determines the path in which the
data packet requires to be routed. When the priority of a data packet is exceeding
zero, then the SDN controller will appreciate the imperativeness of that packet
214 V. Reddy and K. Venkatesh

Fig. 13.10 Flowchart to

Route Critical Traffic

requires to be paid distinguishing attention and routes in a path with the least defer-
ment to the cloud database where the users will get real-time notifications in appli-
cations [27]. In case of priority zero, the SDN controller will be ahead of the data
packet in the usual path. In this manner, by outlining the critical nodes in the IoT
network, latency of the critical traffic can be declined.
In HTTP/2 the data is transmitted in streams and bit 5 in the header is an unsigned
8-bit entity staging a PRIORITY frame that can be posted on a stream in multiple
states; however, it is not possible to send within continuous frames that include one
header frame [9]. There is a possibility of reaching those frames post-processing of
data, which would make it have no impact on the distinguished stream which doesn’t
13 Role of Software-Defined Network in Industry 4.0 215

influence outline transmission on that stream. It is functional for a stream to end up

shut while prioritization data that makes an impact on that stream is in transmission.
Since the stream could be given a need that is not the same as what is planned. To
maintain a strategic distance from these issues, an endpoint should hold stream
prioritization state for a period after streams become shut. The more drawn out state
is held, the lower the opportunity that streams are selected wrong priority or default
priority of the streams.
The PRIORITY of frame can be sent for a stream in the shut or idle state. This
considers the re-prioritization of a grouping of child streams by adjusting the need
of an unused or shut parent stream. The payload of a frame contains a header block
fragment. Prioritization of data in a HEADERS outline is sensibly equal to a differ-
ent PRIORITY outline; however, incorporation in HEADERS avoids the potential
for agitating in stream prioritization when fresh streams are made. Prioritization
fields in HEADERS outlines resulting to the first on a stream re-prioritize the stream.

Result and Discussion

By implementing the suggested testbed architecture with the proposed solution, the
latency of traffic in the critical IoT nodes can be minimized. To observe these
changes, a real-time critical state can be created in any of the IoT sensors connected
to the southbound interface of the SDN Controller, and the packet propagation time
can be traced using the packet monitoring tools like Wireshark. Also the Unix
EPOCH of the packet sent is differentiated with time reached in the Cloud server
network interface, and the average time taken to propagate critical traffic to Cloud
database can be observed in the created monitoring and compared with the time
taken without using SDN controller by means of different network speeds.
As presented in the Fig. 13.11, the graph compares the latency of critical traffic
in IoT with SDN and without using SDN. The latency using SDN traffic is observed
to reduce with respect to time contrasted to normal traffic. This result was posted on
an average speed of 30 mbps. It was also tested across various internet speeds on an
average scale in which the latency is constantly less compared to Non-SDN Critical
Traffic. It was also observed that packet stream using HTTP/2 priority states does
not affect the regular traffic produced by the other nodes in the network.

Fig. 13.11 Flowchart to Route Critical Traffic

216 V. Reddy and K. Venkatesh

Conclusion

In the current system, an eternally rising number of difficulties assorted real-time

applications in IoT that are operating on top of the cloud platform are addressed.
SDN innovation is employed to accomplish an excellent concept among the possi-
ble number of open-source techniques to send critical data from one point to another
in the system. This chapter discussed the adversities that arise in IoT while manag-
ing emergency traffic during the critical state. By having those arguments in mind,
the SDN technology is included in the system assuming that it is one of the decisive
among the existing solutions. When the priority label in the HTTP/2 packet header
can be changed with an account to time and the sensor value, it is easy for an SDN
controller to route the packet with least latency. The procedure that is expected in
the system to determine the critical nodes in the IoT network is taken as an advance
by implementing practically on a testbed as the following work. The expected way
does not supply any excess additional security mechanism to the IoT system in SDN
which can be introduced as a forthcoming assignment [28, 29]. Our upcoming
intention is to implement major algorithms and functions and the security-related
issues [30] in the SDN controller on an IoT network.

Acknowledgments I would like to express my wholehearted thanks to my educator (Mr.

K. Venkatesh) who gave me this elegant chance to explore new trending innovations and to the
organization of this research work that helped me gain knowledge in cutting-edge technologies.
Besides, I would like to thank my companions for giving intense support to conclude this task
within a limited period.

References

1. Gartner, Inc., Gartner Says 13.6 Billion Connected “Things” Will Be in Use in 2018, Up 30
Percent From 2015. Press Release, Stamford, Conn., November 10, 2015.
2. Dores, C., Reis, L., & Lopes, N. (2014) Internet of things and cloud computing. In: 2014 9th
Iberian Conference on Information Systems and Technologies (CISTI).
3. Venkatesh, K., & Nagarajan, V. (2018). A framework to support dynamic routing and traf-
fic provisioning in optical networks using virtualization and software-defined networking.
International Journal of Pure and Applied Mathematics, 118(11), 733–737, ISSN: 1311–8080.
4. Bera, S., Misra, S., & Vasilakos, A. (2017). Software-defined networking for internet of things:
A survey. IEEE Internet of Things Journal, 4(6), 1994–2008.
5. Setup – Raspberry Pi Documentation. Raspberrypi.org, 2018. [Online]. Available: https://
www.raspberrypi.org/documentation/setup/. Accessed 09 Sept 2018.
6. Open vSwitch Documentation – Open vSwitch 2.10.90 documentation. Docs.openvswitch.
org, 2018. [Online]. Available: http://docs.openvswitch.org/en/latest/. Accessed 09 Sept 2018.
7. Get Started – ESP-IDF Programming Guide v4.0-dev-141-g106dc0590 documentation. Docs.
espressif.com, 2019. [Online]. Available: https://docs.espressif.com/projects/esp-idf/en/latest/
get-started/. Accessed: 25 Mar 2019.
8. warmcat/lws-esp32-factory. GitHub, 2019. [Online]. Available: https://github.com/warmcat/
lws-esp32-factory. Accessed 25 Mar 2019.
13 Role of Software-Defined Network in Industry 4.0 217

9. Hypertext Transfer Protocol Version 2 (HTTP/2). Http2.github.io, 2019. [Online]. Available:

https://http2.github.io/http2-spec/. Accessed 25 Mar 2019.
10. ONOS – Home – Wiki. Wiki.onosproject.org, 2018. [Online]. Available: https://wiki.onospro-
ject.org/. Accessed 09 Sept 2018.
11. Kandah, F., & Schmitt, S. (2018). SAND: Smart and adaptable networking design using vir-
tual slicing over software-defined network. EAI Endorsed Transactions on Internet of Things,
4(13), 155333. https://doi.org/10.4108/eai.6-4-2018.155333.
12. Anand, N., Babu, S., & Manoj, B. (2018). On detecting compromised controller in soft-
ware defined networks. Computer Networks, 137, 107–118. https://doi.org/10.1016/j.
comnet.2018.03.021.
13. Kim, H., Kim, J., & Ko, Y. (2014). Developing a cost-effective OpenFlow testbed for
small-scale Software Defined Networking. In: 16th International Conference on Advanced
Communication Technology.
14. Hu, P., Ning, H., Chen, L., & Daneshmand, M. (2019). An open internet of things system
architecture based on software-defined device. IEEE Internet of Things Journal, 6(2), 2583–
2592. https://doi.org/10.1109/jiot.2018.2872028.
15. Li, Y., Su, X., Riekki, J., Kanter, T., & Rahmani, R. (2016). A SDN-based architecture for hori-
zontal internet of things services. In: 2016 IEEE International Conference on Communications
(ICC).
16. Ahmed, K., Blech, J., Gregory, M., & Schmidt, H. (2018). Software defined networks in indus-
trial automation. Journal of Sensor and Actuator Networks, 7(3), 33. https://doi.org/10.3390/
jsan7030033.
17. Muciaccia, T., & Passaro, V. (2017). Future scenarios for software-defined metro and access
networks and software-defined photonics. Photonics, 4(4), 1. https://doi.org/10.3390/
photonics4010001.
18. Li, J., Yang, L., Fu, X., Chao, F., & Qu, Y. (2017). Dynamic QoS solution for enterprise net-
works using TSK fuzzy interpolation. In: 2017 IEEE International Conference on Fuzzy
Systems (FUZZ-IEEE).
19. Meseguer Llopis, J., Pieczerak, J., & Janaszka, T. (2016). Minimizing latency of critical traf-
fic through SDN. In: 2016 IEEE International Conference on Networking, Architecture, and
Storage (NAS).
20. Banerjee, A., Sufyanf, F., Nayel, M., & Sagar, S. (2018). Centralized framework for controlling
heterogeneous appliances in a smart home environment. In: 2018 International Conference on
Information and Computer Technologies (ICICT).
21. Srinivas, L. N. B., & Ramasamy, S. (2017). An analysis of outlier detection techniques for
wireless sensor network applications. International Journal of Pure and Applied Mathematics,
117(16), 561–564, ISSN: 1311–8080..
22. Zhou, W., Li, L., Luo, M., & Chou, W. (2014). REST API design patterns for SDN north-
bound API. In: 2014 28th International Conference on Advanced Information Networking and
Applications Workshops.
23. Sayee Kumar, M., & Purusothaman, T. (2015). Hybrid broadcast group management protocol
for secure, scalable and efficient group communication. Journal of Computer Science, 11(2),
344–350, ISSN: 1549–3636..
24. A framework of controller with flow table cache and performance analysis in software defined
industrial networks. RevistaTecnica De La Facultad De Ingenieria Universidad Del Zulia
(2016). https://doi.org/10.21311/001.39.3.27.
25. Yu, H., Ng, B., & Seah, W. (2018). TTL-based efficient forwarding for nanonetworks with
multiple coordinated IoT gateways. IEEE Internet of Things Journal, 5(3), 1807–1815. https://
doi.org/10.1109/jiot.2018.2812868.
26. Venkatesh, K., Srinivas, L., Mukesh Krishnan, M., & Shanthini, A. (2019). QoS impro-
visation of delay sensitive communication using SDN based multipath routing for medical
applications. Future Generation Computer Systems, 93, 256–265. https://doi.org/10.1016/j.
future.2018.10.032.
218 V. Reddy and K. Venkatesh

27. Bisio, I., Lavagetto, F., & Luzzati, G. (2016). Cooperative application layer joint video coding
in the internet of remote things. IEEE Internet of Things Journal, 3(6), 1418–1426. https://doi.
org/10.1109/jiot.2016.2610103.
28. Karthik, G. M., Sayee Kumar, M., & Aruna, A. (2017). Fully Homomorphic public key encryp-
tion based on arbitrary key aggregation cryptosystem for secured data communication in cloud
infrastructure. International Journal of Pure and Applied Mathematics, 117(15), 1143–1149.
29. Flauzac, O., Gonzalez, C., & Nolot, F. (2015). Original secure architecture for IoT based on
SDN. In: 2015 International Conference on Protocol Engineering (ICPE) and International
Conference on New Technologies of Distributed Systems (NTDS).
30. Vishwesh, J., & Rajashekar, M. B. (2017). Internet of Things (IoT): Security analysis & secu-
rity protocol CoAP. International Journal of Recent Trends in Engineering and Research, 3(3),
417–425. https://doi.org/10.23883/ijrter.2017.3126.6ibua.
Chapter 14
Integrating IoT and Machine Learning –
The Driving Force of Industry 4.0

A. Suresh, R. Udendhran, and M. Balamurugan

Explosive Growth of Intelligent Manufacturing in Industry 4.0

Intelligent factories are created by the initiative action taken by the German
­strategic Industry 4.0 in which the cyber-physical systems (CPSs) play a role in
updating and transforming the manufacturing technologies by cloud computing
and Internet of Things (IoT). In the era of Industry 4.0, the physical process is
being monitored by the manufacturing system called as “digital twin” in the physi-
cal world, which communicates with the machines, sensors and humans and make
smart decisions by communicating in real time. Industry 4.0 is a combination of
intelligent production processes and the embedded production system, which could
make transformation from the production and industry value chains as well as the
business frameworks.
In Industry 4.0, updating the manufacturing systems to an intelligent level takes
place and that makes possible the availability of manufacturing technologies and
sophisticated information in order to achieve a smart, flexible and reconfigurable
process in manufacturing to look over the global and the dynamic market as shown
in Fig. 14.1. Large companies and industries, small- and medium-sized enterprises
(SMEs), and supply chains for holistic manufacturing require certain information
and physical process flows that are enabled by Industry 4.0. Based on the varying

A. Suresh (*)
Department of Computer Science and Engineering, Nehru Institute of Engineering
and Technology, Coimbatore, India
R. Udendhran · M. Balamurugan
Department of Computer Science and Engineering, Bharathidasan University, Trichy, India
e-mail: [email protected]

© Springer Nature Switzerland AG 2020 219

G. R. Kanagachidambaresan et al. (eds.), Internet of Things for Industry 4.0,
EAI/Springer Innovations in Communication and Computing,
https://doi.org/10.1007/978-3-030-32530-5_14
220 A. Suresh et al.

Fig. 14.1 Development of intelligent industry as presented in AI business

situations and necessities, certain keystone technologies are required for the intelli-
gent manufacturing process, based on the previous behaviour of the changing nature.
Therefore, adaptive decisions could be made by solving the problem through direct
communication among the manufacturing systems in a timely manner. In certain
cases, Artificial Intelligence (AI) could be adapted by certain technologies that learn
from the past experience with the intelligence, co-­operated and ever-present practice
in industries [1].

Machine Learning Is Expected to Drive Growth in Industry

Intelligent manufacturing also includes IoT-enabled services with the cloud com-
puting technology for storage purposes. Nowadays, the amount of data generated
and its use are increasing rapidly day by day [2]. The data generated contains
­various format, qualities and semantics, for instance, the data collected from the
environment, production and machine lines. This is mentioned in different forms,
namely Industry 4.0 at Germany, Smart factory in South Korea and Smart
Manufacturing in USA. The quality-related data possibility has the potential to
enhance the quality of the product and process sustainability [3]. The surface rough-
ness could be determined with the help of the statistical features of the machine
learning approach. However, machine learning possess certain negative impacts as
well as challenges, since in certain cases this could distract the system from the
main issues or challenges and lead to conclusions that in turn lead to incorrect
actions [4, 5].
14 Integrating IoT and Machine Learning – The Driving Force of Industry 4.0 221

Fig. 14.2 Machine learning in automobile industry

For the transformation in manufacturing domain could be made by the collabora-

tion of developments made in mathematics and computer science (e.g. statistical
learning) with the easy–to-use tools offered freely to grasp and ­sustain the reposito-
ries of increasing manufacturing data in automobile industry as shown in Fig. 14.2.

Machine Learning in Industry 4.0

Predictions on machine learning technologies can be made based on the wide range
of data available [6]. This could draw knowledge from the experience independently
and construct pattern recognition. But in terms of the future, AI has no scope in the
long term [7]. Internet of things are believed to be distant in concepts due to their
enormous space for storage and the branches of AI, machine learning and deep
learning makes use of big data for the optimization process, gain novel solutions and
obtain novel insights. Companies and industries ranging from small to large interna-
tional sectors all require data that could be used by them. Predictions could be made
with the help of the available software that deals with evaluation and consolidation
[8]. Process efficiency could be obtained with the new solution by properly analyz-
ing the data obtained from the sensor, customers and logs [9]. Moreover, an IT
infrastructure is required due to the addition of huge amount of data, which is attuned
to machine learning and AI workloads and process as shown in Fig. 14.3. The main
tasks of the machine learning are as follows: (1) recognizing patterns and (2) arriv-
ing at solutions from them. Further, these findings could be utilized at their best [10].
222 A. Suresh et al.

Fig. 14.3 Machine learning as enabling technology in Industry 4.0

 urrent Applications of Machine Learning in Smart

C
Industries

Image recognition is the commonly used learning method at present. Certain other
processes include intelligent bots or digital assistants, speech processing, speech
recognition, face recognition, analysis of text and video, and automatic translation
and transcription.

Artificial Intelligence in Industries 4.0

In smart factories, all the components are interconnected, i.e. the machines, compo-
nents and the interfaces could communicate with one another. The manufacturing
process could be optimized with the huge amount of data. With the help of image
recognition and analysis, optimization of big data could be made, which serves as
the best example.

Digital Monitoring and Control

In production industries, the system can identify the objects on the conveyor belt
and can sort them all as shown in Fig. 14.4. In quality control, this intelligent system
works well detecting flaws and colour changes in the product.
14 Integrating IoT and Machine Learning – The Driving Force of Industry 4.0 223

Fig. 14.4 Monitoring with high-performance HMI/SCADA

Artificial Intelligence and Predictive Maintenance

Machine learning techniques are highly used in the field of support services and
maintenance of companies [11]. The maintenance of cycle, energy consumed from
the individual machines could be captured by AI and these information deals with
optimization of the process in the concurrent stages. The need for replacement of
the parts and the defective areas could be indicated by the operating data. The
increasing amount of data makes the optimization process good with more accurate
predictions.

Brief Application of Deep Learning in Industry

Deep learning (DL) gains the attention over commercial machine learning process
by two factors in the field of prediction as well as at the training [12]. (1) For the
process of training, the features needed for engineering and handcrafts requirements
are highly reduced [13]. (2) Extraction could be done the easiest way by DL tech-
niques, which are not caught by the human view. Moreover, accuracy is also
enhanced by the DL techniques. Based on the statement given by experts, the deep
learning process tends to increase the growth of the economy. The maximum ben-
efits are accepted in the field of IT, telecommunications, finance sectors and in man-
ufacturing companies.
224 A. Suresh et al.

 pplications of Human Machine Interface (HMI)

A
in Industry 4.0

Applications of HMI gain attention in consumer use cases like the virtual assistants
as shown in Fig. 14.5. Necessary applications are focused rather than the repetition
of messages that tends to impact the industries. The control of machines in the
industries is said to gain a lot of attention. Plague or a mistake that happens at the
user interfaces in the gas and oil industries could abruptly decrease the production
rate and may have an environmental impact. Integration of state of the art quality in
HMI could raise the efficiency level at the field with new levels [14].
Safety is an important factor that should be concerned as the highest priority
rather than the efficiency. Obtaining physical feedback from the haptic technologies
is beneficial for machine operators. Enable to increase concentrate on the primary
task by freeing the hands by AR or wearable glasses. Real-time status from the
machine could be obtained by the operators so that the prioritization of workloads
with the requirement of material and anticipating tool could be gathered at every
individual task. Off-site monitoring could also be done by the specialized agents
that could enhance the safety around the environment or else it could require the

Fig. 14.5 Combined HMI in industry

14 Integrating IoT and Machine Learning – The Driving Force of Industry 4.0 225

labor or c­ oncerned person to walk to that particular locality. Surveying the status
could be done by the promoted managers of that specific plant, and when a bottle-
neck around the field occurs, the immediate notifications could be received from the
user end [15].
New HMI technology also includes automotive sector as a rapid adopter. In the
high-end model of the infotainment systems, new HMI technologies are determined,
and the application also includes the redesign of mundane features like door open-
ers. The technologies most desired by the automakers include the gesture recogni-
tion and touch screen technologies. Touch screen in the car will not be same as the
one present in the smart phones. This is due to the fact that capacitive screens are
being replaced by the resistive screens in most of the mobile phones. This is done to
lower the cost, increase the longevity etc., enable to increase the adaption of new
technology. Advanced Driver-Assistance Systems include infrared sensors that
could help in monitoring the driver’s attentiveness [16].
The car includes gesture recognition that is fitted at the outside. When drivers who
carry held load of groceries in their hands approach their Ford Escape car parked,
they could open the tailgate by passing their foot underneath the rear bumper. In
Digital Out of Home (“DOOH”) advertising, advertisers tend to use the concept of
gesture recognition and eye-tracking technology. Deployment of digital displays is
highly expensive. For sale purposes, ad displays could be fitted in the form of wall-
mounted advertisement. In certain other cases, through downloading the application
other information could be obtained. The combinations of new analytics with the
higher engagement that enable advertisers to aim specific group of users have
resulted in new revenue for both media companies and their customers alike.
In recent advancements, the connected devices (“Things”) with their virtual rep-
resentation have gained more prominence than ever, creating effective applications,
services like smart healthcare, intelligent industry 4.0, smart transportation for less
pollution with environmental monitoring. These areas have been of focus for many
researchers and organizations as well as standardization bodies among the commu-
nities in semantic web and insurance companies. The main success behind IoT is
that it provides flexibility to applications that demand communication, storage,
accessibility and service platforms and create opportunities for important sectors
such as retail, green energy, manufacturing, smart homes as well as personalized
end-user applications. Therefore, it is evident that IoT plays an important role in
daily routine. Generally, the huge volume of data generated from industries, internet
is overwhelming and reaches 2.5 quintillion bytes of data and in past several years,
around 90% of data was generated from various sources. Among these data, the data
generated from the sensors known as “Sensory data” provide meaningful informa-
tion when analyzed by employing algorithms and they can be converted into a better
understanding about real human world leading the machines to act smartly and
learning by themselves which is known as “Artificial Intelligence”. Artificial
Intelligence offers automation to application, products and knowledge hierarchy
and provides effective data transformation in which we can adapt meanings to the
stages based on the context of semantics and IoT as shown in Fig. 14.6.
226 A. Suresh et al.

Fig. 14.6 Knowledge hierarchy in the context of IoT

 achine-Learning-Based Human Machine Interface

M
in the Context of Industry 4.0

Machine-learning-based HMIs are the recent trend that has been going on in all the
process industry 4.0. This HMI could have an access through the web, which means
that any device could be connected once they gain their access through the internet.
These HMIs are a great solution to the traditional HMI, and we do not need any
human intervention [17]. User would be able to monitor and control the devices
from a remote location. It is not necessary for the user to be physically present. With
the rising use of mobile application, this could be made easier with the go on pro-
cess and get the notifications by hand [18].
Since these HMIs rely on internet access, updates could be done automatically
and ensure the system is using the latest notifications. As soon as the user logs in to
the application, it could send the most recent updates for the users, relevant of
accessing the usual webpage [19].
One of the major problems that occur in these HMIs is security issues. Privacy is
the greatest threat for internet-based applications. A proper connection should be
maintained among the components of the HMI connection in order to enhance the
reliability. A hacker could attempt to gain access over the component and change
the configuration, which could interrupt all the other systems connected to it [20].
This is extensively done in DoS, since hacking the server could interrupt all other
nodes connected to it. During the protocol development, it is necessary for the pro-
tocol designers to pay attention to this threat.
Multiple industries are involved in this machine-learning-based human machine
Interface for the effective operation of the HMI systems as shown in Fig. 14.7. An
issue known as federated identity management occurs, which needs the access from
14 Integrating IoT and Machine Learning – The Driving Force of Industry 4.0 227

Fig. 14.7 Example of machine-learning-based human machine interface in Industry 4.0

the remote organization for the multiple components involved in the HMI since
several equipment need certain authorization (person or equipment involves in get-
ting permission from other entities). This issue could be solved by Security Assertion
Markup Language (SAML), Web Services Trust (WS-Trust), and PKI [21]. It is
necessary to have interoperability, since no conflicts should happen among the
involved industries.
This web-based HMI is not required in every field. Local HMIs could pay a
severe attention where there is a necessity of a person to be physically present in
that environment. Each organization should finely balance the needs [22]. If we opt
to gain mobility, convenience and monitoring, then we should use only the web-­
based application. Otherwise the traditional HMIs are the best.

In-Depth Analysis of Machine Learning

The Internet consists of thousands of machines, instruments, equipment, drones,

consumer electronics, wares, utilities, robots and then millions of sentients, intelli-
gent and digitized objects. That is, the current Internet, which simply consists of
computer servers, desktops, laptops, smartphones and hand-held, is bound to scale
exponentially to have all kinds of devices and things. Precisely speaking, the current
Internet is the network of networked computers [23]. But the future is the network
of networked computers, devices and digitized entities. In other words, the Internet
of devices, services, energy and things is to emerge and evolve rapidly to empower
228 A. Suresh et al.

not only businesses but also people in their daily deals, decisions and deeds. When
such kinds of connected and integrated devices, services, data sources, applications,
and things communicate with one another, there will be massive amounts of inter-
actional, transactional, commercial, operational and analytical data. Therefore, IT
professors and professionals across the world are striving hard and stretching fur-
ther to bring forth pioneering data analytic products, platforms, processes, patterns
to meticulously capture, cleanse and crunch the growing data volumes using the
batch as well as real-time processing methods to extricate actionable insights in
time. Thus, IoT data along with its analytic can result in smarter and sophisticated
applications not only for worldwide businesses but also every individual in the
increasingly connected countries, counties and cities [24]. In this technological
world computers have made their footprints in every field. Without the computer,
technology does not exist, and hence with the evolution of computers we are able to
communicate at a faster rate, for instance e-commerce and on-line marketing have
found their success in many developing countries in which internet is the backbone
in powering these technologies, in the case of industries, which needs computer to
operate their machines as well as store and retrieve data. A lot of equipment are cali-
brated with the help of the computers. Nowadays storage of large amount of data is
possible with the help of big data technologies [25]. In the present scenario, machine
learning seems to be trending, which makes the machine to learn by itself and inte-
grating it with the Internet of Things. For instance, IoT-based healthcare provides
direct connection to equipment (MRIs) and uses Artificial Intelligence for the initial
triage to diagnose problems and reduce the requirement of putting someone on site
by 40%, thereby reducing response time and improving customer satisfaction.
Healthcare companies are streaming EMR data and using deep learning to predict
the health of patients that are potential for code blue or sepsis shock and send a
nurse to visit. Technology has been increasing day by day. Machines play a vital
role in reducing the effort of humans. It is necessary to have a connection to all those
machines. Machine learning algorithms have three distinct phases: learning, valida-
tion and classification [26]. The first phase is when the algorithm learns to classify
data points by creating a knowledge model with the most relevant data features.
Machine learning algorithms take annotated data as input and create a model, which
serves as input for the next phase. In this system the data features vary from the set
of statistical operations applied to each of the smartphone’s inertial sensors data.
After creating the model from example inputs with a certain classifier, the model
needs to be validated with different sets of annotated data. This process tells us the
ratio of correctly and incorrectly classified data points in a confusion matrix. The
last phase is to feed non-annotated data to machine learning algorithm in order to be
classified. The algorithm will compare the non-annotated data points with the model
built in the learning phase and decide on the activity represented by this new
data points.
For producing a complete description of the classification function in a learning
algorithm, a cluster of candidate applications are consistently provided. Nevertheless,
it is predominantly the case where major candidates are inconsequential to the
14 Integrating IoT and Machine Learning – The Driving Force of Industry 4.0 229

e­ rudition assignment, leading to the problem of over fitting, and this will depreciate
the efficiency of the engaged learning algorithm. The indoctrination speed as well
as learning veracity may be momentously retrograded by these excessive features.
Hence it is of basic concern to exquisite the accordant and significant features in the
processing step. This chapter delineates fundamental feature selection issues and
current research objectives. Machine learning is an evolving field today due to an
increase in big data. It aids to progress observations from a huge quantity of data,
which was very ponderous to humans and at times also inaccessible. Machine learn-
ing is a specialization of computer, which is an expeditiously trending topic in
today’s context and is anticipated to bang more in the upcoming years. In a specula-
tion, more discriminating power can be attained by aggregating the size of the fea-
ture vector. However, the learning procedure and the model generalization are
slow-paced and compromised due to extensively corpulent feature vector. Feature
selection is specifically necessary when one is manipulating a vast dataset with
dimensions up to thousands.
The prime applications of feature selection are as follows: (i) abbreviating the
measurement cost and accommodation requirements (ii) due to the finiteness of
training sample sets, coping with the degradation of the classification performance
(iii) minimizing exploitation and training time and (iv) promoting data understand-
ing and data conception. It may be visualized that evaluating the significance of
distinctive description may not be crucial; nevertheless, the actual remonstrance is
to evaluate the consequence of subsets of features.
Data is being created continuously every day throughout the world. According to
the Solutions Company CSC, a big data analytic company, the data evolved at the
year 2020 could be 44 times larger than the year 2009. Hence, it is essential to rec-
ognize data and add observation for the better indulgent of the human world. The
quantity of data is so large these days that conventional procedures cannot be used.
Evaluating data or constructing extrapolative models by hand is almost infeasible in
some consequences and may consume more time and take away productivity [27].
Machine learning produces consistent, imitable results and educates itself from for-
mer computation.
There are two types of data used in machine learning. One is labelled data and
the other is unlabelled data. In labelled data the attributes are essential and neces-
sary that could imply a tag to the information used in supervised learning, which
could be categorical or numerical. To estimate a value of regression, numerical data
could be used, and categorical data could be utilized for classifying. Unlabelled data
are implemented in unsupervised learning since they have only data points. Due to
this implementation the machine can find the structures and patterns present in the
data set. Hence these two data types of machine learning can be utilized with super-
vised as well as unsupervised learning correspondingly. Supervised learning carries
out a learning map flanked by a set of variables, say X and Y, which are the Input
and Output variables. This mapping is enforced to anticipate the output for invisible
data. After gaining the set of data, algorithms simplify the data and articulate the
refutable value H for the given sets of data as shown in Fig. 14.8.
230 A. Suresh et al.

Fig. 14.8 Feature selection process

There are further two types of classifications in supervised learning. They are
known as Regression and Classification. A regression could be used within certain
variables for a statistical relationship between two or more variables. Classification
arises very commonly in day to day life.
It contains partitioning up objects. Each object is accredited to one of the several
mutually exhaustive and exclusive kinds known as classes. Each object should be
authorized to specifically one class, i.e. more than one object shouldn’t be attached
to a single class. Unsupervised learning examines how systems could acquire to
signify meticulous input patterns in a path that redirects the statistical structure of
the general collection of input patterns. A contradiction to supervised learning is
that there are no specific objective outputs or environmental estimation accompany-
ing every input.
Feature selection:
This framework comprises two parts: (1) utilization of search engine to obtain
the subset of the feature and (2) obtaining the best candidate from the given criteria.
The inspection for a variable subset is a NP-hard enigma. Therefore a better solution
could not be assured till an exhaustive search is affected in the solution space. The
filter method could be the best for the classifier, while a wrapper method utilizes
a classifier to calculate feature subsets. This could form the selection scheme, and
these methods are plentiful and they could be utilized effectively for the classifier.
Filtering methods such as bilateral information, autonomous component analysis,
class detaching measure, or variable ranking could be effectively used for clas-
sification. In most of the datasets, all the attributes may not help in defining or
­determining class labels. Enhancing the size of an attribute vector may draw more
14 Integrating IoT and Machine Learning – The Driving Force of Industry 4.0 231

d­ iscriminating power. The excessive attribute vectors affects the learning process,
which cause the classifier closely bounded with the training data and compromise
model generalization. An attribute congruity can be measured with two strategies:
(1) univariate approach and (2) multivariate approach. Univariate strategies are flex-
ible in terms of speed and structure. Most probably inter-­relationship and dependen-
cies among the attributes are not counted in univariate, whereas it is well concerned
in multivariate search techniques. Multivariate searches also have some restrictions.
Initially, they are exposed beyond their limit, mainly in settings. Furthermore, while
handling a large attribute arena, it becomes too expensive.
Though ranking of attribute is undefined, its computational and statistical scal-
ability makes it desirable to various variable subset selection strategies. Since it
claims only the computation and sorting of n scores it becomes more efficient. The
main disadvantage of attribute ranking is that it leads to the assortment of iterative
subset. This problem can be resolved by using a smaller subset of equivalent attri-
butes. Attribute ranking can also be achieved by ranking metric and discarding all
features that do not reach an enough value.
Due to the high predictability in neural network, it is referred to as feed-forward
network. In the current automated industry, we use the backward propagation that
comes under the concept of supervised neural networks as shown in Fig. 14.9. The
development in this rule has been done that arouses an interest to know more regard-
ing the artificial neural network. In this rule we follow the method of computation
efficiency, which could provide changes in the differential function activation units.
With this method we would be able to learn the input and output data very ­effectively.
This method of backward propagation was first adopted in the Harvard University.
Several procedures have been adopted to improve the backward propagation speed,

Input layer First hidden Second hidden Output layer

layer layer

Fig. 14.9 Two-dimensional neural network

232 A. Suresh et al.

tolerance effect, local minima etc. This could be used to solve the diversity problem
that could be classifying patterns, function approximation, prediction of time-series
as well as developing the non-linear systems.

 achine Learning Influence in Control Systems-Based

M
Industries

Control systems act as the constituent division in this industrial world. The control sys-
tem is designed to automate the process without the need for human intervention. This
could avoid some human errors during the process. This automation process could be
greatly supported with the PLCs, which could be used in the process of designing,
testing and optimizing the process. The basic advantages of automating a system are
increased production, accuracy, safety to the environment and flexibility [28].
There are four main features which make the automation process to be in the lead
of the control system engineering:
• Transfer of information from the analogue system into the digital format, which
could be supported by the computer-based systems.
• The control methods for the process of automation are carried out with the help
of the language-based variable, with the exception of the heavy and tedious
mathematical calculations [29].
• Amendment of communication is done by the human intervention with less
effort for processing the information [30].
• Utilization of multifaceted automation processed system had increased the pro-
ductivity in the recent times.
Nowadays, the automatic detection of the technology around the environment
is possible that could detect the happenings in real time, which takes all the neces-
sary information to enhance the performance [31]. Meanwhile, when there is less
­possibility of real-time response, certain malevolent activities, human errors and
operational deficiencies occur [32]. Several such IoT-based sensors and machine
learning applications are available on the shelf. For example, e-Gauge energy
meters connect to the building’s electric panel and measure its aggregate electric-
ity consumption every second. These energy meters can upload the measured data
to cloud, which stores, analyses and displays the data for the users, thereby help-
ing them monitor and curtail their usage [33]. Nest’s smart learning thermostat is
an excellent example of Internet-enabled smart home appliance. The Nest ther-
mostat can learn home occupancy patterns and program itself. If the user forgets
to turn off heat before leaving, the thermostat will take care of it. The thermostat
can be programmed from anywhere over the Internet. The manufacturers claim, by
doing all these smart things, that it can cut heating and cooling costs by up to 15%.
Besides, there are several other smart home appliances and sensors available in the
market. Besides employing Internet-enabled smart appliances, there has been work
on programmatically regulating domestic electricity consumption profile.
14 Integrating IoT and Machine Learning – The Driving Force of Industry 4.0 233

Conclusion

Initiatives made for the improvement of the quality, estimation of the manufacturing
costs, customer requirements and optimization of the process should be best under-
stood by the manufacturing industries to obtain the maximum benefits. Moreover,
for complex dimensions and dynamics, support is required to handle them all. Other
improvements made in the field of machine learning include knowledge discovery,
data mining and artificial intelligence. Still the wide field of machine learning
includes various algorithms and methods that are still evolving.

References

1. Alpaydin, E. (2010). Introduction to machine learning (2nd ed.). Cambridge, MA: MIT Press.
2. Apte, C., Weiss, S., & Grout, G. (1993). Predicting defects in disk drive manufacturing: A case
study in high dimensional classification. In IEEE Annual Computer Science Conference on
Artificial Intelligence in Application (pp. 212–218). Los Alamitos, CA.
3. Azadeh, A., Saberi, M., Kazem, A., Ebrahimipour, V., Nourmohammadzadeh, A., & Saberi, Z.
(2013). A flexible algorithm for fault diagnosis in a centrifugal pump with corrupted data and
noise based on ANN and support vector machine with hyper-parameters optimization. Applied
Soft Computing, 13, 1478–1485. https://doi.org/10.1016/j.asoc.2012.06.020.
4. Çaydaş, U., & Ekici, S. (2010). Support vector machines models for surface roughness predic-
tion in CNC turning of AISI 304 austenitic stainless steel. Journal of Intelligent Manufacturing,
23, 639–650. https://doi.org/10.1007/s10845-010-0415-2.
5. Chand, S., & Davis, J. F. (2010, July). What is smart manufacturing? Time Magazine.
6. Cherkassky, V., & Ma, Y. (2009). Another look at statistical learning theory and regularization.
Neural Networks, 22, 958–969. https://doi.org/10.1016/j.neunet.2009.04.005.
7. Chinnam, R. B. (2002). Support vector machines for recognizing shifts in correlated and other
manufacturing processes. International Journal of Production Research, 40, 4449–4466.
https://doi.org/10.1080/00207540210152920.
8. Cohn, D. (2011). Active learning (p. 10). In C. Sammut & G. I. Webb (Eds.),
Encyclopedia of machine learning (p. 1058). New York, NY: Springer. https://doi.org/
10.1007/978-0-387-30164-8.
9. Cook, D. F., Zobel, C. W., & Wolfe, M. L. (2006). Environmental statistical process control
using an augmented neural network classification approach. European Journal of Operational
Research, 174, 1631–1642. https://doi.org/10.1016/j.ejor.2005.04.035.
10. Corne, D., Dhaenens, C., & Jourdan, L. (2012). Synergies between operations research and data
mining: The emerging use of multi-objective approaches. European Journal of Operational
Research, 221, 469–479. https://doi.org/10.1016/j.ejor.2012.03.03.
11. Udendhran, R. (2017). A hybrid approach to enhance data security in cloud storage. In
ICC’17 Proceedings of the Second International Conference on Internet of things and Cloud
Computing at Cambridge University, United Kingdom. March 22–23, 2017, ISBN: 978-1-­
4503-4774-7. https://doi.org/10.1145/3018
12. Suresh, A., Udendhran, R., Balamurgan, M., et al. (2019). A novel internet of things frame-
work integrated with real time monitoring for intelligent healthcare environment. Springer-­
Journal of Medical System, 43, 165. https://doi.org/10.1007/s10916-019-1302-9.
13. Suresh, A., Udendhran, R., & Balamurgan, M. (2019). Hybridized neural network and decision
tree based classifier for prognostic decision making in breast cancers. Springer - Journal of
Soft Computing. https://doi.org/10.1007/s00500-019-04066-4.
234 A. Suresh et al.

14. Operator Screen (HMI) Design Guidelines. Retrieved from www.hexatec.com

15. Anderson, R. J., & Davies, B. (2010). Using virtual objects to aid underground storage tank
teleoperation. Albuquerque, NM: Sandia National Labs.
16. Backes, P. G. (2008). JPL space station telerobotic engineering prototype development:
Advanced telerobotics system technology. Pasadena, CA: Jet Propulsion Lab.
17. Bajcsy, R., Enciso, R., Kamberova, G., Nocera, L., & Sara, R. (1998). 3D Reconstruction
of Environments for Virtual Collaboration. Proc. 4th IEEE Workshop on Applications of
Computer Vision, Princeton, NJ.
18. Bourhis, G., & Agostini, Y. (1998). Task sharing and trading for an intelligence powered
wheelchair: Problematics. Symposium on Robotics and Cybernetics. In CESA '96 IMACS
Multiconference. Computational Engineering in Systems Applications Journal of Intelligent
and Robotic Systems archive,22(3–4), 269–287. Norwell, MA: Kluwer Academic Publishers.
https://doi.org/10.1023/A:1007943113785.
19. Bourhis, G., & Agostini, Y. (2008). The Vahm robotized wheelchair: System architecture
and human-machine interaction. Journal of Intelligent and Robotic Systems: Theory and
Applications, 22(1), 39–50.
20. Bourhis, G., & Agostini, Y. (2013). Man-machine cooperation for the control of an intelligent
powered wheelchair. Journal of Intelligent and Robotic Systems: Theory and Applications,
22(3–4), 269–287.
21. Brutzman, D. (2011). Virtual world visualization for an autonomous underwater vehicle.
Challenges of Our Changing Global Environment. In Conference Proceedings. California,
USA.
22. Chen, P. C., & Hwang, Y. K. (2006). Generating natural-language like robot motion command
set through machine learning. In Proceedings of IEEE International Conference on Systems,
Man and Cybernetics (Vol. 1, pp. 114–119). Piscataway, NJ: IEEE.
23. Al-Fuqaha, M., Guizani, M., Mohammadi, M., & Aledhari, M. A. (2015). Internet
of things: A survey on enabling technologies, protocols, and applications. IEEE
Communication Surveys & Tutorials, 17(4). Fourth Quarter 2015 (PDF) Internet of
things in marketing and retail. Available from: https://www.researchgate.net/publication/
312290899_Internet_of_things_in_marketing_and_retail.
24. Andersson, P., & Mattsson, L. (2015). Service innovations enabled by the “Internet of Things”.
IMP Journal, 9(1), 85–106. Friedrich, H., Muench, S., Dillmann, R., Bocionek, S., & Sassin,
M. (2009). Robot programming by demonstration (RPD): Supporting the induction by human
interaction. Machine Learning, 23(2–3):163–189.
25. Graves, A. R. and Czarnecki, C. (2007). Distributed generic control for multiple types of
telerobot. In International Conference on Robotics and Automation. Detroit, MI, USA.
26. Berndt, H. (2015). Creating an IoT Ecosystem through scenarios. Lisbon: IEEE.
27. Caputo, A., Marzi, G., & Pellegrini, M. (2016). The internet of things in manufacturing inno-
vation processes: Development and application of a conceptual framework. Business Process
Management Journal, 22(2), 383–402.
28. Jagersand, M. and Nelson, R. (2010). Visual space task specification, planning and control.
Proceedings of IEEE International Conference on Computer Vision, IEEE, Piscataway, NJ. Pp.
521-526.
29. Krauss, C., Do, X. A., & Huck, N. (2017). Deep neural networks, gradientboosted trees, ran-
dom forests: Statistical arbitrage on the S&P 500. European Journal of Operational Research,
259(2), 689–702.
30. Kshetri, N. (2017). The economics of the Internet of Things in the Global South. Third World
Quarterly, 38(2), 311–339.
31. Davenport, T. H., & Lucker, J. (2015). Running on data. Deloitte Review, 16, 5–15.
32. Sawaragi, T. (2009). Design of human-artifacts symbiotic systems under resource bounded-
ness. Journal of Japanese Society for Artificial Intelligence, 13(3), 371–378.
33. Sklyar, V. (2016). Top ten technologies for IoT [Online]. Available at: https://habrahabr.ru/
post/308892/
14 Integrating IoT and Machine Learning – The Driving Force of Industry 4.0 235

A. Suresh is currently working as the Professor & Head,

Department of the Computer Science and Engineering in
Nehru Institute of Engineering & Technology, Coimbatore,
Tamil Nadu, India. He has nearly two decades of experience
in teaching, and his areas of specializations are Data Mining,
Artificial Intelligence, Image Processing, Multimedia and
System Software. He has published 82 papers in international
journals. He has published more than 40 papers in National and
International Conferences. He has one patent in the name of Rear
view optimised safety helmet. He has published five book chap-
ters for IGI Global Publisher. He has served as a guest editor/
reviewer for Springer, Elsevier, and Inderscience journals, etc. He
is a member of IEEE, ISTE, IACSIT, IAENG, MCSTA, MCSI,
and Global Member of Internet Society (ISOC). He has organized
several National Workshops, Conferences and Technical Events.
He is regularly invited to deliver lectures in various programmes
for imparting skills in research methodology to students and
research scholars. He has published four books, namely ‘Data
structures & Algorithms’, ‘Computer Programming’, ‘Problem
Solving and Python Programming’ and ‘Programming in ‘C’ for
DD Publications, Excel Publications and Sri Maruthi Publisher,
Chennai, respectively.

R. Udendhran pursued Master of Technology in Computer

Science and Engineering from Bharathidasan University. He has
presented his research work in the University of Cambridge at the
United Kingdom, which is available in ACM digital Library. He
has his research publications included in SCI indexed database.
Currently, he is pursuing PhD in deep learning in Bharathidasan
University.

M. Balamurugan is the Professor and Head of the Department

of Computer Science and Engineering at Bharathidasan
University. He has completed his PhD and guided 10 research
scholars. His area of research includes cognitive science and data
science. He has completed a UGC-funded research project, has
two patents and has published 30 papers in reputed journals.
Chapter 15
Machine Intelligence and Automation:
Deep Learning Concepts Aiding Industrial
Applications

S. Sree Dharinya and E. P. Ephzibah

Introduction

Nowadays artificial intelligence with deep learning is a thriving area in which more
research and better practical tools and application are built. Earlier, the complex
problems were approached with solutions based on mathematical concepts and now
that they are looked at with a more intuitive approach a simpler and practical way is
adopted. To bring about solutions to be simpler and also to avoid complicated con-
cepts, computers build on concepts with many layers called deep learning with the
use of artificial intelligence. The most difficult part is the hard-coded technology
that the AI uses for the patterns. The solutions to such problems which are more
subjective with the knowledge of real-world technology have been the evolution of
machine learning [1].
Deep learning is used in various applications such as image analysis and text iden-
tification. It uses supervised and unsupervised learning to represent data in a hierar-
chical format. Although big data has great opportunities, it had some constraints on
data mining and information processes due to the characteristics of its volume [2].
The growth of sensor networks and communications has posed a great insight into
deep learning for the usage and classification and applications of machine learning.
The digital transformation of industry which includes the smart systems is towards
the Industry 4.0. Smart systems and factories with cyber physical system are growing
towards Industry 4.0. Deep learning techniques are capable of identifying the com-
plex signal which are sent and received [3]. Digital manufacturing, smart factories
and applications with speech and text recognition are some of the perspectives that
drive the deep learning techniques along with IoT to develop and get into Industry 4.0.

S. Sree Dharinya (*) · E. P. Ephzibah

School of Information Technology and Engineering, Vellore Institute of Technology,
Vellore, Tamilnadu, India
e-mail: [email protected]

© Springer Nature Switzerland AG 2020 237

G. R. Kanagachidambaresan et al. (eds.), Internet of Things for Industry 4.0,
EAI/Springer Innovations in Communication and Computing,
https://doi.org/10.1007/978-3-030-32530-5_15
238 S. Sree Dharinya and E. P. Ephzibah

The data is stored in large volume and thus the big data has an explosive amount
of data. This volume of large data is both in structured and unstructured forms.
Nearly three-fourths of the data are unstructured, which is generally the data from
internet and mobile applications [4]. Beyond the volume of the data, the large exact-
ness aids in inaccurate objects and blaring data and replicated data objects. Analyzing
real-time data, like an online data becomes very critical in case of big data. Handling
the massive data becomes difficult with applications in health care or industry, as
the data to be extracted is to be intuitive and precise. The major constraint in data
mining is the large data and is very difficult to analyse the large variation of data.
Deep learning plays a vital role in harvesting the targeted data from the huge vol-
ume of data and also from difficult systems [5]. Deep learning takes back to the
early techniques where the data is pre-trained and perfected for effective data mining.

Deep Learning Models

The deep learning concept arises when a set of data is to be retrieved from a com-
plex set of data or environment. Figure 15.1 represents a deep learning model.
According to the concept of artificial intelligence, it is difficult for the computer to
understand raw data given by sensors which is similar to the image given in the

Vehicle Animal Person Output Entity

Third hidden layer

Second hidden layer

First hidden layer

Input visible layer

Fig. 15.1 Deep learning model

15 Machine Intelligence and Automation: Deep Learning Concepts Aiding Industrial… 239

figure. The way in which the pixels of the picture are mapped is quite complex and
it seems next to impossible once we think of solving it directly [6]. Using the tech-
nique of deep learning it is possible to identify the object with a sequence of nested
mapping, and the layers of each mapping depict the different attributes of the object.
Here the input data is shown in the observable layer and is also termed likewise as
it has all the features of the object which is to be identified. Then consecutively the
set of hidden layers are available, which reveal the abstract properties of the image.
The layers are termed as hidden layers since the data to be extracted is not based on
the values but on the underlying concept of mapping the object to be established.
The pictures here are conceptions of the type of data which is hidden. The first layer
compares with the limits related to the brightness of the picture and the second layer
identifies the angles. They are collectively searched as a data to view the picture.
The next layer finds the precise objects related to angles of the picture. The final
layer is the one which describes the parts of the object which it has to identify in the
picture [7]. Deep learning does not invoke the factors of each layer but it helps to
run a program with sensible input [8].
Deep learning thereby helps to organize the data rather than the exact content of
the data. The data is not visible at the moment the object is identified. After layer of
identification and identification of edges and contours, the data is visualized as an
object in the last layer. Such techniques evolve the types of deep learning with vari-
ous approaches based on the output parameters. For a large data set where the iden-
tification of the object is complex, deep learning uses the method of identifying the
concept as an outcome. Such techniques aid with the machine learning algorithms
and also they are used in various sectors as industries, smart home and smart city
developments with the help of IoT. The category of learning from a complex set of
data is done using deep learning concepts.
Industries make use of the data available in large volume and indulge in data-­
driven applications for their benefits like easy access to information, knowledge to
upgrade and manufacture the products in demand, increase the income and so on.
They like to optimise the manufacturing operations using machine learning tools
and techniques. A company that manufactures steel pipes for oil and natural gas
industries makes use of the data science team to combine artificial intelligence in
the area of automation solutions. Different data sources are integrated and the
system becomes flexible with the help of machine learning algorithms that are far
easier and faster [9]. Deep learning helps to test different models and chooses the
best model in limited time; thus, predictions are done in seconds. Result analyses
and performance reports are efficiently done faster so as to obtain customer satis-
faction. In a real-time system, it allows to control energy consumption, produce
better products, save cost, deliver to the customers on time. Thus, they provide
numerous ways for the industries to be a leader among their competitors with
traditional approaches [10].
Deep learning neural networks are one of the principal machine learning tech-
niques in the recent years. Many of the well-known and commonly used applica-
tions that use deep learning neural networks are speech recognition, face recognition,
autonomous or self-driving cars, image tracking and computer vision [11].
240 S. Sree Dharinya and E. P. Ephzibah

Deep learning is widely used in industries, as it takes very complex concepts and
composes them out of smaller and smaller concepts to capture something very com-
plicated [12].
According to the medical industry, drug discovery is a major area where deep
learning can be applied to find a fit in terms of the drug with the proper disease [13].
It is a flourishing field in all the countries of the world, as it captures the special
benefits of the drugs and their demand among the people in curing the diseases.
Using artificial intelligence and deep learning, it is more practical to cure any kind
of disease.
Deep learning uses large volume of data especially in medical field where a tre-
mendous increase in the volume of data over a period of time is possible. For exam-
ple, if the heart rate of a patient is to be monitored for every second and the data is
to be recorded, imagine the volume of data for a period of 10 days or 1 month [14].
Deep learning is the state-of-the-art machine learning method especially for medi-
cal imaging and image analysis.
Deep learning and high-performance computing are enabling the new industrial
revolution –an era of smart creation and streamlined production [15]. Manufacturers
have been quick to adopt deep learning, artificial intelligence and machine learning
using their ability to improve processes across supply chains, asset management,
predictive maintenance pricing and more with such podium-packed automation,
cloud technology and streamlined IT (Information Technology) are more important
than ever before. With deep learning, manufacturers are already exploring and sim-
ulating designs in virtual environment. Intelligently managing logistics and printing
3D prototypes their optimizing costs at each step by speeding up design and testing
and keeping machine to be running to minimize down time. Soon advances in deep
learning driven by powerful GPU accelerators will make virtual products develop-
ment even more accurate by simulating models with unprecedented accuracy.
Environments in various fields make sophisticated digital twining a reality for man-
ufacturers. Minimizing expensive designing process at production stage is again
challenging. Robotic machinery, however, contributes to cost-saving by identifying
downtime opportunities and shutting itself down. Service operations are revolution-
ized by predictive maintenance that optimizes the equipment operational life time
and stops failures before they even occur by embedding and managing AI in new
infrastructure. Many industries work for this kind of operations. Especially compa-
nies like HPE allow AI to work for the business and for IT. The data scientist, inno-
vators and big data architects and those leading product engineering companies
support digital evolution within manufacturing.
Artificial intelligence and high-performance computing are changing the land-
scape in both e-commerce and physical retail shops. Retailers are using artificial
intelligence, machine learning and deep learning to attract, to engage and to main-
tain relationships with customers improving both efficiency and loyalty. In such a
highly competitive environment, automation cloud technology and streamlined IT
are important than ever before. Online retailers rely on AI recommendation engines
to thoroughly analyse the activities of the potential customers and drive traffic to
their sites, and in physical stores AI is planning and guiding shoppers’ experience
15 Machine Intelligence and Automation: Deep Learning Concepts Aiding Industrial… 241

with smart layouts and inventories checkout free technology and video analytics.
With deep learning, retailers would soon be exploiting the potential of personalized
digital assistance to make recommendations. Visual search in stores define the right
item and even virtual reality to fit in for a quicker buying process. The future poten-
tial of AI in retail is huge with only minimum percent of companies confident in
responding to far changing demands. Embedding and managing AI in the infrastruc-
ture makes the business faster and more productive with lots of profit [16–18].
AI and deep learning also have a very great impact on the dairy farms and milk
products industries. Cows in the form require more attention and technology help
the owners to make decisions for the better growth in cows and for good income.
Artificial intelligence acts as a great tool to get more insights. It helps to keep track
of the cows and makes the farmers life easier. Sensors are connected to the neck of
the cows [19]. By analysing the movement of the sensors, the machine learning
algorithm figures out the movement of the cow and how it is doing. It is also pos-
sible to distinguish the behaviours of the cow like eating, drinking, standing, lying
and walking. The software identifies the disease the cow is prone to and can warn
the farmer. Using tensor flow and open-source tools it learns the behaviour of the
animals and it gets better over time. It improves the efficiency of milk production,
animal welfare to keep the cows healthy.
In any business anticipating and planning what is next is crucial for success.
When caring for someone’s health, predicting what is ahead has the potential to save
lives. There are so many AI tools and techniques along with machine learning to
achieve the task. An initiative that will make machine learning and predictive ana-
lytics widely available across all of health care is on the move. Today it is observed
that only the largest health systems have the resources to build and play the predic-
tive analytics [20]. Demand for data scientist having the right people on board is
either cost prohibitive and sustainable or possible. These lifesaving technologies
should be made available everywhere. Machine learning models are designed and
built into each product that is delivered. Caregivers can now easily make predictive
analytics routine and actionable. Imagine how valuable access to data-driven
machine learning could be for caregivers like data mutineers. Utilizing machine
learning, data mutineers can discover patterns in their own patient’s data and they
can identify the best interventions to reduce the patient’s length of stay and predict
the next deadly infection. Based on patients’ specific risk factors the deep learning
and AI techniques could even prevent the next onset of diabetes; these benefits
could be made available to many in health care who are not just as customers. There
are companies that take this technology and learning in creating healthcare systems
using deep learning and artificial intelligence [21–24].
There are lots of applications that make use of the deep learning algorithm to find
solutions to the real-time complicated and time-consuming problems. Figure 15.2
given below depicts a few of them.
Deep learning has become a well-known ignition technology behind many
devices such as amazon alexa, siri on i-phone and many other voice-enabled devices.
At the core of this learning is the use of artificial neural networks which are com-
puter programs that enable machines to learn. They are inspired by some of the
242 S. Sree Dharinya and E. P. Ephzibah

Fraud Detection Vehicle classification

Pedestrian Detection Deep learning Gesture Detection

Forest fire Face recognition

Detection

Fig. 15.2 Applications of deep learning

computation that goes on in our brain, real neural networks. Consider a situation of
building a machine that can read and write information. At the core of artificial
neural networks are the artificial neurons that are connected. The neuron can be con-
nected to pixels of the image. The job of the artificial neuron is to detect patterns
from its incoming connections. Like human brain that has many neurons, artificial
neural networks have many neurons, each one performing a different operation on
technically different types of data. In deep artificial neural networks there are mul-
tiple layers, and this is the core idea behind deep learning.

Types of Deep Learning Models

The various types of deep learning techniques are classified under categories such
as supervised, unsupervised and semi-supervised. Supervised learning is grouped as
regression and classification problems. The classification problem is based on the
category which is attribute-based and the regression is based on the original value
like predicting the time series.
Unsupervised learning is based on the input of the data and has no relevant out-
put parameters. The dissemination of data is more focused as there are no exact
answers and no specific mentors. New algorithms are devised and are classified and
identified for clustering and association problems. Clustering helps to identify the
group behavior, and association helps to discover large data. They tend to solve the
learning problems associated with unsupervised learning algorithms. Industry uses
communication systems to implement the artificial intelligence concepts. The tradi-
tional type of communication system is shown in Fig. 15.3. They work in blocks and
in a linear format where there are responsibilities assigned to each block. This type
of communication becomes more difficult and more prediction-based. The robust-
ness of the system is the thought process when the blocks undergo certain flaws.
A simulation-based communication system can be correctly interpreted using
machine learning algorithms.
15 Machine Intelligence and Automation: Deep Learning Concepts Aiding Industrial… 243

SENDER

SOURCE CHANNEL TRANSMITTER

SOURCE MODULATOR
ENCODER ENCODER ANTENNA

C
H
A
N
N
E
RECIEVER L

SOURCE CHANNEL DETECTOR/ RECIEVER

DESTINATION
DECODER DECODER DEMODULATOR ANTENNA

Fig. 15.3 Traditional communication system

 pplications of Deep Learning in Industries and Smart

A
Applications

Industries use different artificial intelligence techniques. In warehouses one such air
abnormality detection in industries is identified through deep learning techniques.
Warehouses are an integral component of the transport and logistics industry, and
tremendous amounts of operational efficiencies can be unlocked by introduction of
intelligent IoT solutions for Warehouse Management. One such solution is real-time
hazardous chemical or gas detection that ensures safety of warehouse by detecting
hazardous, inflammable chemicals/gases that could get leaked from items stored in
a warehouse. The deep learning method called HMT algorithm for detection of
hazardous gas and chemicals in warehouse is used widely.
Toxic gases have huge environmental effects. People are suffering several dis-
eases due to them. Proper detection of toxic gas level is important for us. Toxic
gases have huge environmental effects. Sometimes when highly inflammable and
hazardous gas/chemicals leak it leads major accidents, life loss, and severe health
issues. The protected transportation of these chemicals and gases are to be concen-
trated in high range, because only during transportation large quantity of them are
transferred. Minute leakage also may lead to major accidents such as life and prop-
erty loss, air pollution and affects health from outer skin to micro DNA in living
things. These effects last for generations to generations.
The existing systems use different set of sensors like MQ2 and MQ7; MQ9 and
MQ4, MQ2 and MQ7 gas sensors for detection which do not cover some hazardous
gas. These systems use machine learning algorithm which was trained by some
predefined set of training data set by rule-based data analysis for finding anomaly
and release of toxic gases. It also uses “Deep Belief Network” (DBN) as an unsu-
pervised learning algorithm to find the generic features of the input. That output is
fed into SVM trained in a way to detect whether there is abnormality present in the
data with single class output.
Prediction is processed online using the algorithm, as it identifies state xt as usual
or anomalous earlier getting the following xt + 1. This procedure studies ­continuously
244 S. Sree Dharinya and E. P. Ephzibah

without a requirement to store the entire stream. This procedure runs in an unsuper-
vised, automatic fashion without data tags or manual parameter alteration and adapts
to vibrant surroundings and idea drift, as the underlying statistics of the data stream is
often non-stationary. It also makes anomaly detections as early as possible and mini-
mizes false positives and false negatives which are true for batch scenarios as well.

Environmental Gas Sensors

Detecting principles of solid-state smoke sensors and gas sensors which are used to
detect the air pollution level must be able to work in any harmful condition irrespec-
tive of thermal or chemical attack.
Types of sensors are Semi-conductor and capacitor-type gas sensor, Non-
Dispersive Infrared Method using Pyroelectric Infrared Sensors, Solid-Electrolyte.
All other techniques which involve detection finished from outdoor the pipeline by
way of visual observation or portable detectors are capable of discover very small
leaks and the leak place, however the detection time is very long. Although the
charges involved for detection seems to be high,the precision obtained is remark-
ably accountable. This downside is vulnerable to disappear for a number of these
techniques due to forthcoming technological improvements [25].
It gives an overview of the pollution caused to the environment. The solid state
sensors which may cause hazardous environmental pollution are taken into consid-
eration and conventional instruments are used which in turn is costly and time con-
suming. Addition to this as we are living in a rapidly developing environment rapid
development in the gas sensors are important to live a safe and comfortable life.
Unsupervised real-time abnormality detection for running data explains the
problems in detecting anomaly in streaming data and proposed a machine learning
algorithm to deal with data. Normally “Machine Learning” and “Deep Learning”
models are trained in batches using data which are collected for that particular
model. But in case of streaming data its different and we should use real time data’s
to detect anomaly in our data. For this they proposed Hierarchical Temporal Memory
(HTM) and they also present their results using Numenta Anomaly Benchmark
(NAB), a benchmark containing real-world data streams with labelled anomalies.
The algorithm learns continuously and does not require the data to be stored like in
static batch anomaly detection. Even “Time Series Analysis” which is a highly
researched area in data science is not applicable for real-time streaming data.
Usually HTM learns continuously and the model’s spatio temporal characteristics
analyses the input and does not detect anomalies. The computations on the output of
HTM, provides a better prediction error and there is a likelihood that the system is
in an anomalous state. The threshold on this likelihood is used to detect whether an
anomaly is found. The number of sensor data and flow of streams of data are increas-
ing and anomaly detection in real time is getting more important and unavoidable.
The algorithm framed is capable of detecting spatial and temporal anomalies in
predictable and noisy domains.
15 Machine Intelligence and Automation: Deep Learning Concepts Aiding Industrial… 245

High-dimensional and Large-scale Anomaly Detection Using a Linear

One-class SVM with Deep Learning [26, 27].

Anomaly detection in low-dimensional data is fairly straightforward and easy com-

pared to high-dimensional data. In high-dimensional data, all the data won’t be
relevant to anomaly and this will conceal the presence of the anomaly present in the
data, and this is commonly known as “Curse of Dimensionality” in data science
communities. To handle this problem, this chapter proposed a hybrid algorithm
which is a combination of unsupervised feature extractor and an anomaly detector.
First the most relevant features for anomaly are extracted from the given high-­
dimensional data and then they are fed into an anomaly detector to find the presence
of anomaly in the data. The feature extractor used is “Deep Belief Network” (DBN),
which is an unsupervised learning algorithm to find the generic features of the input.
That output is fed into SVM trained in a way to detect whether there is anomaly
present in the data with single class output. This hybrid algorithm after implementa-
tion gave good results with better accuracy of anomaly detection, as it addressed the
complexity and scalability issues of 1SVM, especially in large-scale datasets. DBN
is used as a dimensionality reduction algorithm and helped solve the problem of
scalability of 1SVM model.
Hierarchical Temporal Memory is a machine learning algorithm first proposed in
2005 by a Brain Scientist called Jeff Hawkins. This model is inspired by neocortex
which is a part of biological brain. It is a part of the cerebral cortex concerned with
sight and hearing in mammals, regarded as the most recently evolved part of
the cortex.
The HMT algorithm concept is given such that, in brain, the neocortex cells are
very similar to one other, and they do not have any relation to the sensory input
organs such as eyes and ears. But the same neocortex handles different inputs and
trains itself to gain useful information and predict using this input. This is why
HMT algorithms can be used for wide range of applications and not restricted to one
particular set of problems. HMT algorithms show promise in building intelligent
systems. It can be used to solve problems which are easy for humans but difficult for
machines (Vision, Audio). Hierarchical Temporal Memory consists of nodes in a
hierarchical tree structure. All these nodes are using same algorithm called as Naïve
Bayes. These nodes form the networks in Hierarchical structure with many inputs
and few outputs. HMT uses the spatial-temporal concept. This concept acts as the
basic and absolutely necessary component of human intelligence. Temporal means
with relation to time. In a sequence of inputs the time input acts as key to identify
the dynamic changes in the input. Spatial means the space. The static position of the
inputs handling both space and time simultaneously might not be easy and so hier-
archical structure is used. HMT is different from other machine learning algorithms
as it is completely self-learning.
The deep learning techniques use the application IoT for smart applications. As
given by Sunil Raj Thota et al. [20], the devices interconnected and controlled
through internet are called Internet of Things [3]. There are many methods of solid
waste collection and management. Something like using wireless sensor networks,
246 S. Sree Dharinya and E. P. Ephzibah

IoT, Microcontroller-enabled senor networks and many more. Among all the IoT-­
based sensor network with sensors working in the real time is understood to be the
best and can grab the current status of the bin and helps in communicating it via
some Wi-Fi module. Also real-time analytics and route optimization algorithms
have added up a lot more advantage for the methodology and provide a solution
which is more accurate, optimistic and efficient when compared to others. The
smart dust bin is connected to the internet to provide real-time information on waste
management necessary to avoid diseases by managing smart dustbins. The system
can monitor the level of trash can and shows the status of the dust bin level through
the mobile app from anywhere through Internet. It is useful and can be fixed in the
trash cans at public places, smart cities and other places. For the implementation of
smart dustbins, microcontrollers and sensors are used to reduce the cost and to make
it an efficient device. To monitor the trash in the dust bin sensors are attached to
trash can to detect the level. The system consists of sensor circuitry used for moni-
toring the smart trash can. The project module is divided into two parts: sensor sec-
tion and Node MCU Wi-Fi module section. Sensors are attached to the dust bin.
Sensors are used to detect the level in the dust bin. The sensor senses the content of
the dust bin and sends the signals or the data to the Arduino. Wi-Fi Module helps to
send the information of the trash to the receiver side through internet [28].
The waste management in industry is also managed as given by Mohan et al. [29]
as the major challenge faced by urban India in current context of urbanization is
managing solid waste. The waste generation in 2025 is expected to be double the
amount of waste generated in that of 2012. The complexity of solid waste manage-
ment in India is added by unavailability of infrastructure and local municipal bod-
ies. The customized model emphasizes on door to door collection of waste for
6 days in a week. The waste collector is equipped with hand-pulled carts with two
separate bags for recyclable and non-recyclable waste. This model provides daily
services around 250 households. This model works as decentralized waste manage-
ment system. The collected waste is transported by waste collectors for secondary
segregation. After that the waste is separated from the waste to make compost in
non-biodegradable. Sustainability, sale of recyclable waste and compost are the
major sources of revenue. Of the total revenue generated in past 3 years, around
48% comes from user charges. It is a positive indicator for the sustainability of
the model.

References

1. Unnikrishnan, S., Razil, M., Benny, J., Varghese, S., & Hari, C. V. (2017). LPG monitoring
and leakage detection system. In 2017 International Conference on Wireless Communications,
Signal Processing and Networking (WiSPNET), Chennai (pp. 1990–1993). IEEE.
2. Shi, L., Wang, X., Zhang, T., Hu, C., Luo, K., & Bai, B. (2016, August). Hazardous gas detec-
tion four-rotor UAV system development. In Mechatronics and Automation (ICMA), 2016
IEEE International Conference on (pp. 2461–2465). IEEE.
15 Machine Intelligence and Automation: Deep Learning Concepts Aiding Industrial… 247

3. Li, S., Wen, Y., Li, P., Yang, J., & Yang, L. (2012, October). Leak detection and location for
gas pipelines using acoustic emission sensors. In Ultrasonics Symposium (IUS), 2012 IEEE
International (pp. 957–960). IEEE.
4. Ajith, S., Harivishnu, B., Vinesh, T. K., Sooraj, S., & Prasad, G.. (2017). Automated gas
pollution detection system. In 2017 Second International Conference for Convergence in
Technology (I2CT), Mumbai (pp. 483–486). IEEE.
5. Wang, C., Olson, M., Sherman, B., Dorjkhand, N., Mehr, J., & Singh, S. (2018, October).
Reliable leak detection in pipelines using integrated DdTS temperature and DAS acous-
tic fiber-optic sensor. In 2018 International Carnahan Conference on Security Technology
(ICCST) (pp. 1–5). IEEE.
6. Spachos, P., Song, L., & Hatzinakos, D. (2014, January). Gas leak detection and localization
system through Wireless Sensor Networks. In Consumer Communications and Networking
Conference (CCNC), 2014 IEEE 11th (Vol. 1130, No. 1131, pp. 10–13). IEEE.
7. Shu, L., Chen, Y., Sun, Z., & Tong, F. (2017). Detecting the dangerous area of toxic gases
with wireless sensor networks. IEEE Transactions on Emerging Topics in Computing. http://
ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=7917279&isnumber=6558478.
8. Varma, A., Prabhakar, S., & Jayavel, K. (2017, February). Gas leakage detection and smart
alerting and prediction using IoT. In Computing and Communications Technologies (ICCCT),
2017 2nd International Conference on (pp. 327–333). IEEE.
9. Alotaibi, M., Lo’ai, A. T., & Jararweh, Y. (2016, November). Integrated sensors system based
on IoT and mobile cloud computing. In Computer Systems and Applications (AICCSA), 2016
IEEE/ACS 13th International Conference of (pp. 1–5). IEEE.
10. Mhatre, L., & Rai, N. (2017, February). Integration between wireless sensor and cloud. In
I-SMAC (IoT in Social, Mobile, Analytics and Cloud) (I-SMAC), 2017 International Conference
on (pp. 779–782). IEEE.
11. Bokefode, J. D., Bhise, A. S., Satarkar, P. A., & Modani, D. G. (2016). Developing a secure
cloud storage system for storing IoT data by applying role based encryption. Procedia
Computer Science, 89, 43–50.
12. Ahmad, S., Lavin, A., Purdy, S., & Agha, Z. (2017). Unsupervised real-time anomaly detection
for streaming data. Neurocomputing, 262, 134–147.
13. Erfani, S. M., Rajasegarar, S., Karunasekera, S., & Leckie, C. (2016). High-dimensional and
large- scale anomaly detection using a linear one-class SVM with deep learning. Pattern
Recognition, 58, 121–134.
14. Chen, X., Wang, W., & Li, W. (2012, June). An overview of hierarchical temporal memory: A
new neocortex algorithm. In Modelling, Identification & Control (ICMIC), 2012 Proceedings
of International Conference on (pp. 1004–1010). IEE.
15. Brugo, T., Palazzetti, R., Ciric-Kostic, S., Yan, X. T., Minak, G., & Zucchelli, A. (2016).
Fracture mechanics of laser sintered cracked polyamide for a new method to induce
cracks by additive manufacturing. Polymer Testing, 50, 301–308. https://doi.org/10.1016/j.
polymertesting.2016.01.024.
16. Rai, R. K., Bhattarai, D., & Neupane, S. (2019). Designing solid waste collection strategy
in small municipalities of developing countries using choice experiment. Journal of Urban
Management. https://doi.org/10.1016/j.jum.2018.12.008.
17. Lee, D. D., & Lee, D. S. (2001). Environmental gas sensors. IEEE Sensors Journal, 1(3),
214–224.
18. Ramírez, C., & Gonzalez, E. (2019). Methodological proposal for the interinstitutional man-
agement of wastes in health care centers in Uruguay. MethodsX, 6, 71–81.
19. Yusof, N. M., Zulkfli, M. F., Mohd Yusof, N. Y. A., & Azman, A. A. (2018). Smart waste
bin with real time monitoring system. International Journal of Engineering & Technology,
7(2), 29.
20. Thota, S. R., Neelima, S., Lalitha Pruthvi, K. V. N., Mouika, K., Pravallika, M., & Sowmya,
N. (2018). Smart trash can monitoring system using IoT – Creating solutions for smart cities.
Waste Management, 29, 2625–2643.
248 S. Sree Dharinya and E. P. Ephzibah

21. Selvaraj, K., & Chakrapani, A. (2017). Smart dustbin monitoring system using LAN Server
and Arduino. International Journal of Advances in Computer and Electronics Engineering,
2(4), 20–23.
22. Jino Ramson, S. R., & Jackuline Moni, D. (2016). Wireless sensor networks based smart bin.
Computers & Electronics Engineering, 64, 337–353.
23. Nelles, M., Grünes, J., & Morscheck, G. (2016). Waste management in Germany – Development
to a sustainable circular economy? Procedia Environmental Sciences, 35, 6–14.
24. Ghosh, S. K. (2016). Swachhaa Bharat Mission (SBM) – A paradigm shift in waste manage-
ment and cleanliness in India. Procedia Environmental Sciences, 35, 15–27.
25. Capel, A. J., Edmondson, S., Christie, S. D., Goodridge, R. D., Bibb, R. J., & Thurstans,
M. (2013). Design and additive manufacture for flow chemistry. Lab on a Chip, 13(23),
4583–4590.
26. Choi, S., Jun, C., Zhao, W. B., & Noh, S. D. (2015a). Digital manufacturing in smart manu-
facturing systems: Contribution, barriers, and future directions. In Advances in production
management systems: Innovative production management towards sustainable growth (Vol.
460, pp. 21–29).
27. Choi, S., Kim, B., & Noh, S. (2015b). A diagnosis and evaluation method for strategic plan-
ning and systematic design of a virtual factory in smart manufacturing systems. International
Journal of Precision Engineering and Manufacturing, 16(6), 1107–1115. https://doi.
org/10.1007/s12541-015-0143-9.
28. Samir, K. C. (2014). An age distribution story: Reading into the population pyramid of Nepal.
Asian Population Studies, 10(2), 121–124.
29. Mohan, G., Sinha, U. K., & Lal, M. (2016). Managing of solid waste through public private
partnership model. Procedia Environmental Sciences, 35, 158168.
Index

A Artificial bee colony (ABC)

Access point (AP) mode, 209 algorithm, 124
Adaptive extended Kalman filter (AEKF), 152 artificial bees, 124
Additive-manufacturing methods, 21 Deb’s method, 124
Adopted data mining methods, 4 decision making, 124
Advanced Driver-Assistance Systems, 225 pseudo-code, 124
Advanced Message Queuing Protocol Artificial immune systems (AIS)
(AMQP), 44 AINet, 105
Advanced metering infrastructure (AMI), see bio-inspired algorithm, 103
Energy management blue penciling techniques, 104
Advanced Mobile Phone System (AMPS) clonal choice techniques, 104
European GSM system, 51 clonal selection algorithm, 104
first-generation, 50 CLONALG, 105
FPLMTS, 51 data optimization, 113–114
NMT and NTT, 50 honey bee algorithm (see Honey bee
second generation, 51 algorithm)
TACS, 50 immune network algorithm, 105
US DAMPS system, 51 intrusion detection, 106, 107
Ant colony optimization (ACO), 70 IoT, 106
algorithm, 123 negative selection, 104, 106
artificial ants, 122 neural network–based NSA, 107
dynamic optimization problems, 123 NSA, 104
real-world problems, 123 problem-solving techniques, 103
values of trails, 123 proposed system architecture
Anticipatory Classifier System (ACS), 63 (see Proposed system
Applications, IIoT architecture)
autonomous vehicles, 168 security, 106
industrial automation, 165 Artificial intelligence (AI), 22, 28, 34, 220,
predictive maintenance, 166 222, 223, 225, 228, 240
smart logistics management, 167 behavioral aspects, 58
smart package management, 167 biodegradability prediction, 67
smart robotics, 166 bio-inspired computing, 68
smart tools/wearables, integration, 167 bio-robotics, 68
software integration, product evolutionary biology, 61
optimization, 167 machine learning, 57

© Springer Nature Switzerland AG 2020 249

G. R. Kanagachidambaresan et al. (eds.), Internet of Things for Industry 4.0,
EAI/Springer Innovations in Communication and Computing,
https://doi.org/10.1007/978-3-030-32530-5
250 Index

Artificial neural networks (ANN) FA (see Firefly algorithm (FA))

application, 118 FFA (see Fruit fly optimization algorithm
feedback network, 117 (FFA))
learning FSA (see Fish swarm algorithm (FSA))
algorithms, 117 GA, 120–121
concept, 117 PSO, 121
laws, 118 Biological computing/bio-inspired computing,
rate, 117 58, 68
model, 118 Bio-robotics, 68
neurons, 117 Blue penciling techniques, 104
Asynchronous Message Queuing Protocol Bluetooth low-energy (BLE), 2
(AMQP), 174 Burden resister, 182–183
Augmented reality (AR), 28
Automated Control of Home Appliances
(ACHA), 195 C
Automated identification and data capture Carrier sense multiple access with collision
(AIDC), 8 avoidance (AMPS), 55
Automatic meter reading systems, 178 Challenges, IIoT
Automotive vehicles, 168 data storage, 169
industrial Ethernet, 169
security, 168
B things, visibility, 168
Backbone router (BBR), 47 unified legacy system environment, 169
Bacteria foraging optimization (BFO) Charging infrastructure
application, 129 challenges, 149
foraging strategy, 128 EVs architectures, 149
GA and PSO, 128 issues, 149
working, 128 malware, 149
Bat algorithm (BA), 131, 132 protocols, 149
Battery management systems (BMS), 152 wireless communication, 149
Bayesian approach, 34 Charging station integration
Big data, 237, 240 cloud service, 156
associative rules, 5 DC vs. EV standard, 156
cluster analysis, 5 guidelines and technical
collection phase, 4 measures, 155
data cleaning, 4 navigation systems, 156
data integration, 4 online payment service, 156
decision-making, 4 safety and security, 155
frequent path data mining, 5 CLONALG, 105
Industry 4.0, 3 Cloud computing model, 21, 32
knowledge extraction, 4 CNC method, 81
mining techniques, 4 Code division multiple access (CDMA)
powerful and useful, 3 technology, 42
real-time data record processes, 3 Collaborative management model
re-configuration, 4 (CMM), 169
Binder jetting process (BJG), 82 Collaborative sensor-cloud (CSC), 26
Bio-inspired algorithm, 103, 104 Computer-aided design (CAD), 28
ABC (see artificial bee colony (ABC)) Computer-aided manufacturing
ACO (see Ant colony optimization (ACO)) (CAM), 28
BA, 131–132 Computer-based support systems, 116
BFO (see Bacteria foraging optimization Computer vision (CV), 145
(BFO)) Constrained Application Protocol (CoAP),
CSO (see Cuckoo search optimization 31, 172
(CSO)) Convolution neural network (CNN), 94, 95
Index 251

Cuckoo search optimization (CSO) Decision support systems (DSS), 116

algorithm, 129, 130 artificial life
biological behaviour, 130 hard, 60
brood parasitism, 129 Mother Nature, 60
drawbacks, 130 soft, 60
Levy flights, 129, 130 wet, 60
Current transformer (CT), 182 biodegradability prediction, 67
Curse of Dimensionality, 245 cellular automata, 60
CustomPartNet estimation tool, 86 expert system/survey models, 67
Cyber-physical production systems (CPPS), 21 fault tolerance, 64
Cyber-physical systems (CPSs), 1, 10, hardware constraints, 66–67
18–20, 219 LCS (see Learning classifier systems (LCS))
Cybersecurity, 18, 19, 21, 169 machine-based learning
application layer, 25 (see Machine-­based learning)
categories, 22 membrane computing
communication layer, 24 cell-like, 63
features, 22 neural-like, 64
IIoT system, 22 P systems, 63
measures deployed, 26 tissue-like, 64
perception layer, 23–24 neural networks
processing layer, 25 approaches, 59
software defenses, 23 data processing algorithms, 58
perception network, 59
real-time application, 59
D operating environment, 66
DarkNet ImageNet database, 144 production costs, 65
Darknet neural system structure, 140 regression models, 67
Data Distribution Service (DDS), 45, 174 scalability, 65
Data fusion, 19, 27, 28, 34–36 sensor networks, 64, 66
DC fast charging swarm intelligence, 58
fault diagnostics study, 150 time-conserving methods, 57
features and advancements, 150 valuable methods, 57
investigation encounters, 150 Decision tree (DT), 5
requirements, 150 Deep Belief Network (DBN), 243, 245
Decision making, 70 Deep learning (DL) method, 221, 223, 228
AI techniques, 115, 117 AI, 93
ANN (see Artificial neural networks (ANN)) applications, 242
bio-inspire (see Bio-inspired algorithms) artificial intelligence, 240, 243
computer-based support systems, 116 big data, 238, 240
DSS, 116 concept, 238
FL, 118–119 description, 93
human process, 115 gas sensors, 244
Industry 4.0 HMT algorithm, 243, 245
ISSDM (see Intelligent Support ignition technology, 241
Systems for Decision Making in industries, 240
(ISSDM)) IoT, 239, 245
intelligent algorithms, 116 learning frameworks, 93
perform tasks, 116 lifesaving technologies, 241
real-world problems, 134 machine learning algorithm, 241, 243
reasoning-based decision-making medical field, 240
problems, 115 neural networks, 239
semi-structured decision problems, 115 1SVM model, 245
structured decision problems, 115 state-of-the-art machine learning
unstructured decision problems, 115 method, 240
252 Index

Deep learning (DL) method (cont.) cost-efficient ways, 180

technique, 239 feedback, 179
time series analysis, 244 load control unit, 187
traditional approach, 239 measurement, 182–184
types of, 239, 242–243 point-to-multipoint and mesh networks, 180
waste management system, 246 power factor, 185
Deep neural network systems (DNNs), 4, 137 reactive power, 186
Demand response (DR), 195 real-time energy consumption, 180
Digital Advanced Mobile Phone System RTCs, 187
(DAMPS), 51 single-phase AC electricity meter, 178
Digital era, 161 smart grid, 177
Digital Out of Home (DOOH), 225 smart meters (see Smart meters)
Digital technology, 161 solid-state electric meter type, 178
Digital transformation, 170 sources and energy storage devices, 180
Digital twin, 219 traditional meter, 178
Direct-to-home (DTH) services, 193 types, 177
Distributed control system (DCS), 48 unauthorized consumption and electricity
Distributed generation systems (DGs), 195 theft, 180
Domain name system (DNS), 17 uses, 179
voltage measurement, 184–185
Enterprise architecture (EA), 19
E Enterprise integration (EI), 19
Economical manufacturing method, 86 Equivalent circuit model (ECM), 152
Electric vehicle (EV) Evolution of technology, 162
battery charging Evolution, IIOT
AEKF, 152 applications (see Applications, IIoT)
ageing characteristics, 152 DCS, 162
BMS, 152 industrial 4.0, 163
communication technology, 154 innovation, 163
ECM, 152 revolution, 164
fused sensor network, 154 sensors, 163
optimisation algorithm, 154 versions (see Versions, IIoT)
optimisation models, 154 warehouse sensors, 163
pack and motor controller, 152
parameters, 153
physical parameters, 152 F
Q-estimation algorithm, 152 Fault tolerance, 64
sensor networks, 154 Feedback network, 117
SOC, 152 Fieldbus control system (FCS), 48
SOH, 152 Filtering method, 230
battery management Firefly algorithm (FA)
cloud computing systems, 154 application, 127
communication protocol, 154 biological behaviour, 126
GUI, 155 flashing behaviour, 126
hardware integration, 155 pseudo-code, 127
internet gateway, 155 Fish swarm algorithm (FSA)
Kalman filter, 155 bio-inspired technique, 125
TCP/IP, 154 food satisfaction factor, 125
Electronic Product Code (EPC), 6, 7 optimization solution, 126
Electronic vehicle charging (EVC), 195 pseudo-code, 125
Emergency traffic, 216 Food satisfaction factor, 125
Energy management Foraging strategy, 128
active power, 186 Fourth Industrial Revolution, see Industry 4.0
conventional meter, 179 Fragmentation redundancy scattering (FRS), 26
Index 253

Frequency division duplex (FDD) systems, 41 Industrial automation, 165

Fruit fly optimization algorithm (FFA), 130, 131 Industrial control systems (ICS), 28
Future Public Land Mobile Industrial Ethernet, 169
Telecommunication System Industrial Internet Architecture Framework
(FPLMTS), 51 (IIAF), 7
Fuzzy inference, 119 Industrial Internet Consortium (IIC), 7
Fuzzy logic (FL), 118, 119 Industrial Internet of Things Systems (IIoTS)
application, 39, 55
architectural framework, 17
G collaboration and connection, 18
Gas sensors, 244 communication
Genetic algorithms (GA) architecture of cloud, 32
challenges, 120 cloud-based applications, 32
convincing results, 121 Ethernet-based approaches, 32
optimization tool, 120 fieldbus systems, 32
pseudo-code, 120 protocols, 31
single/multi-objective problems, 121 vehicular networks, 31
Global search technique, 113 connectivity, 162
Graphical user interface (GUI), 199, 209, 210 cybersecurity (see Cybersecurity)
Grid, 149, 150 data fusion, 19, 34–36
industrial tools, 162
Industry 4.0, 18–22
H innovation, 39, 55
Hierarchical Temporal Memory (HTM), 244 IP, 17
Honey bee algorithm sensors (see Sensors)
Anode, 110 standard
cluster maintenance and communication, 112 AMPS (see Advanced Mobile Phone
collecting food information, 108 System (AMPS))
greedy mechanism, 112 fieldbus, 47–50
HBAC, 109 IEEE 802.15.4, 54
Hnode, 110 and stack, 47
intrusion detection, 108 WLAN, 52–53
nectar positions, 111 WPAN, 54–55
Nnode, 110 technological developments, 162
nourishment sources, 109 technology and protocol
novel CH, 111 AMQP, 44
operating system, 109–110 CoAP, 46
Pnode, 110 DDS, 45
pursuit system, 111 5G, 40
Q(Node), 111 ISA100.11a, 46, 47
types, 108, 109 Modbus, 44
waggle dance, 109 MQTT, 44
Honey bees calculations (HBAC), 109 NB-IoT, 43
Hub degree (Anode), 110 1G/2G/3G/4G, 41–43
Hub portability (Pnode), 110 OPC UA, 45
Human machine interface (HMI), 19, 224, 225 wireless sensor network, 40
Hydraulic pressure sensor, 207 WirelessHART, 47
Hypertext Transfer Protocol (HTTP), 31 WLAN, 40
WPAN, 40
XMPP, 43
I Zigbee, 46
IEEE 802.11 Access Point Interoperability warehouse sensors, 163
Protocol (IAPP), 53 Industrial Internet Reference Architecture
Independent sensor-cloud (ISC), 26 (IIRA), 7
254 Index

Industrial Internet system (IIS), 7 Internet of Services (IoS), 1

Industry 4.0, 1–3, 73, 76–78 Internet of Things (IoT)
advantages, 75 applications, 7
AM process architectures, 6
electron beam melting, 79 description, 6
laser beam melting, 79 EPC, 7
liquid bed process, 79 evolution, 6
material deposition process, 78 FOG computing layer, 7
SM, 76 IIoT, 7
3D printing, 78 Industry 4.0 (see Industry 4.0)
vs. traditional manufacturing, 75 localization, 8
barcode, 11 Logistics 4.0, 8
cutting-edge technology, 165 RAMI4.0, 7
description, 163 RFID technology, 8
development, 165 wire/wireless networks, 8
digital twin, 219 Internet Protocol (IP), 17
disadvantages, 75 (see also Energy Intrusion detection, 106–108
management) Intuitive approach, 237
Fourth Industry Revolution, 8 IoT-based additive manufacturing (AM),
HMI, 224–226 83–84, 87
injection molding, 75
intelligent actions, 165
interrelated smart devices, 162 L
machine learning (see Machine learning) Laser beam method, 79
mobile technology, 11 Learning classifier systems (LCS)
NFC, 11 ACS, 63
producing and manufacturing processes, 11 classifier, 61
RFID method, 11 discovery component, 62
Roland 3D CNC Milling Machine, 77 factors, 63
smart industries, 161 four components, 61
supply chain management, 9 performance component, 62
Information and communications technology reinforcement component, 62
(ICT), 19 rule-based systems, 61
Intelligent production systems Learning rate, 117
CNC machining, 74 Levy flight mechanism, 129
Industry 4.0 (see Industry 4.0) LibWebSockets (LWS), 201, 202
injection molding, 73 Light-emitting diode (LED), 178
plastic forming, 74 Liquid crystal display (LCD), 178, 187–189, 191
plastic joining, 74 Logistics 4.0, 8
traditional manufacturing process, 73 Low-rate wireless personal area network
Intelligent robots, 166 (LR-WPAN), 54
Intelligent Support Systems for Decision
Making (ISSDM)
AI techniques, 116 M
application, 133 Machine-based learning
bio-inspired algorithms, 133 AI and Industry 4.0, 222
challenge, 135 AI and predictive maintenance, 223
decision making, 133, 134 algorithm, 228
hybridization, 134 applications, 229
Internal combustion engine (ICE), 147 automobile industry, 221
International Electrotechnical Commission bees, 70
(IEC), 46 bio-inspired computing, 68
International Telecommunication Union bio-robotics, 68
(ITU), 51 classification, 230
Index 255

control systems-based industries, 232 Open Network Operating System (ONOS),

decision-making, 70 201, 208
digital monitoring and control, 222 Open vSwitch (OVS), 199, 203
DL, 223, 228 OpenWRT, 199
electric cars, 71 Outage detection (OD), 195
filtering method, 230 Over the Air (OTA), 197, 201
fish robots, 69
Industry 4.0, 221, 222
intelligence algorithms, 68 P
Internet, 227 Packet data unit (PDU), 204
IoT, 220 Pan-European Digital Mobile
locomotive robots, 68 Telecommunications Network
multivariate approach, 231 (GSM) system, 50
neural networks, 68 Particle swarm optimization (PSO), 154
newer and dynamic spaces, 69 algorithm, 121
real-time processing methods, 228 multidimensional multi-objective
regression, 230 problem, 122
swarm intelligence, 68 pseudo-code, 122
types, 229 stochastic optimization technique, 121
univariate approach, 231 thrust areas, 122
wrapper method, 230 P colony, 64
Machine intelligence Plug-in hybrid electric vehicles (xEVs), 147,
big data, 237 149, 150, 152, 155–157
deep learning (see Deep learning) Potential transformer (PT), 182
intuitive approach, 237 Predictive maintenance, 166
Machine-to-machine communication Prepaid smart meter, 193
(M2M), 6 Product life cycle management (PLM), 30
Marketing monsters, 167 Programmable logic controllers (PLCs), 17,
Message Queuing Telemetry Transport 29, 31, 44
(MQTT), 44, 172 Proposed system architecture
Mining techniques, 4 data optimization, 112–113
Multivariate approach, 231 design process, 112
Mutual sensor-cloud (MSC), 26 detector technique, 112
naïve detector, 112
Proteus 8 simulator, 189
N Protocols, IIoT
Naïve Bayes, 245 CoAP, 172
Narrowband Internet of Things DDS, 174
(NB-IoT), 43 HTTP protocol, 172
Nature-inspired algorithms, 120 MQTT system, 172
Negative selection algorithms (NSA), 103, OPC UA, 172
104, 107, 114 REST, 172
Neighbor quality (Nnode), 110
Network Time Protocol (NTP), 212
Neural network (NN) algorithm, 5, 35, 114 Q
Non-Dispersive Infrared Method, 244 Q-estimation algorithm, 153
Non-SDN device critical traffic, 212 Quality of service (QOS), 53, 197
Numenta Anomaly Benchmark (NAB), 244

R
O Radio-frequency identification (RFID), 19
OLE Process Control (OPC), 30 applications, 1
OPC Unified Architecture (OPC UA), 45, 172 big data (see Big data)
Open Images Dataset, 139 history, 2–3
256 Index

Radio-frequency identification (RFID) (cont.) protocols and interfaces, 30

Industry 4.0, 1 types, 27
IoT (see Internet of Things (IoT)) warehouse, 163
logistics, 12–13 Sensory data, 225
manufacturing, 13 Service quality (QoS), 42
method, 11 Simulation-based communication system, 242
retail, 12 Single board computer (SBC)
technology, 2–3 CNN, 94, 95
Radio transmission technology (RTT), 51 deep learning method (see deep learning
Rapid injection molding method, 84 method)
Real-time clock (RTC), 187 image processing, 95, 96
Real-time location system (RTLS), 2, 11 imagining processing (see imagining
Real-time processing methods, 228 processing)
Reference Architecture Model for Industry 4.0 math model, 96, 97
(RAMI4.0), 7 minimum cost, 100
Remote terminal units (RTUs), 31 outcomes, 98, 99
Revolution, IoT, 164 python program code, 95
Reynolds’ boids model, 60 safety equipment, 96
Round trip time (RTT), 210 secure working environment, 100
Routing method, 198 TensorFlow, 94, 95
Rule-based systems, 61 working principle, 95
WSN, 94
Small- and medium-sized enterprises
S (SMEs), 219
Scalability, 65 Smart grid, 177
Secured optimization technique, 106, 107 Smart meters
Security, 106, 112 applications, 195
Security Assertion Markup Language conventional meters, 181–182
(SAML), 227 developing countries, 181
Selective laser melting (SLM), 79 flowchart, 194
Selective laser sintering (SLS), 79 hardware results, 191–193
Self-nonself discrimination approach, 107 parameters and suggestions, 189
Semi-supervised learning, 242 prepaid facility, 193–195
Sensor networks, 64 prototype, 182, 187–189
Sensors, 34 Smart robotics, 166
artificial intelligence, 28 Smart technology, 167, 169
automation tools, 166 Smart wearables, 167
built-in sensors, 164 Software-defined networking (SDN)
and business machines, 163 application development, 198
cyber-physical production-control architecture
systems, 28 GUI topology, 210
data fusion, 27 Non-SDN device critical traffic, 212
data transmission, 29 ONOS, 209
DDS, 174 OpenFlow v1.4, 211
design software, 28 OpenWRT-based Open vSwitch, 208
digital counterparts, 30 parameters, 212
IIoT systems, 28, 30 real-time critical traffic, 211
industrial 4.0, 163 virtual switch, 209
industrial network, 29 complications, 198
modern industrial machines, 166 controller, 213, 215
packages, 168 emergency traffic, 216
physiological factors, 27 ESP-32 LWS, 200–201
processing method, 28 HTTP/2, 214
Index 257

implementation, 198 Time-Sensitive Networking (TSN), 32

IoT, 198 Total Access Communication System
components, 203 (TACS), 50
determination and data visualization, 204 Traditional Communication system, 243
ecosystem, 206 Traditional manufacturing process, 73
Ethernet switch, 204 Transmission Control Protocol (TCP), 26
industrial management systems, 205
network services, 203
open-hardware approaches, 204 U
open-source platforms, 203 Ultra-wideband (UWB), 2
PDU, 204 Unified Configuration Interface
programming development, 203 (UCI), 199
WSN, 205 Univariate approach, 231
next generation technology, 206 Unsupervised learning, 242
Non-SDN Critical Traffic, 215
ONOS, 201
OpenWRT OVS, 199–200 V
PRIORITY, 215 Versions, IIoT
Raspberry Pi v3, 199 industry 1.0, 164
real-time environment, 208 industry 2.0, 164
robotized machines, 206 industry 3.0, 165
Route Critical Traffic, 214 industry 4.0, 165
sensors, 202, 213 Virtual reality (VR), 28
server and client approach, 206 Voltage measurement, 184
traditional frameworks, 198
Solid-state relay (SSR), 187
Standard Tessellation Language (STL), 83 W
State of health (SOH), 152 Web Services Trust (WS-Trust), 227
State-of-the-art machine learning method, 240 Wi-Fi Protected Access (WPA), 53
Stochastic feature selection method, 58 Wireless local area network (WLAN), 40
Strong Security Network (SSN), 53 Wireless personal area network (WPAN),
Subtractive manufacturing (SM), 76 40, 54
Supervised learning, 242 Wireless sensor network (WSN), 6, 22, 94,
Supervisory Control and Data Acquisition 198, 205
(SCADA) Systems, 31 Wrapper method, 230
Support vector machine (SVM), 4, 5
Swarm intelligence, 58, 68
Y
You Only Look Once (YOLO), 137, 138,
T 141, 144
TensorFlow, 94 You Only Look Once–v3 (YOLOv3)
Terrible hub, 110 anchor boxes, 144
3D printing applications, 145
AM vs. traditional manufacturing Batch Normalization, 144
CNC machine, 80, 81 bounding box prediction, 144
EBM, 82 class location frameworks, 138
IoT-based, 80 class prediction, 144
CustomPartNet, 86 comparison v1 & v2, 138, 139
designing, 87 DarkNet, 144
gas-assisted injection molding, 84 decisions, 145
injection molding methods, 84, 86 design model
product paradigms, 87 bounding box prediction, 139, 140
Sculpteo batch control, 84 class prediction, 139
 258 Index

You Only Look Once–v3 (YOLOv3) (cont.) outcomes

Darknet-53, 140 art work image testing, 142, 143
network architecture, 141 real time image testing, 142
training and testing, 140 sample video testing, 142–144
end-to-end network, 144 scales prediction, 145
fine-grain features, 144 speed trade off, 145
high resolution classifier, 144
learning object location, 138
multi-scale training, 144 Z
object discovery, 138 Zero crossing detector (ZCD), 191

View publication stats