2019 8th International Conference on Industrial Technology and Management

Smart Manufacturing with Prescriptive Analytics

A review of the current status and future work

Johannes Vater, Lars Harscheidt Alois Knoll

Planning and Production of Electrified Powertrains Robotics, Artificial Intelligence and Real-time Systems
BMW Group Technical University of Munich
Munich, Germany Munich, Germany
e-mail: Johannes.JV.Vater@bmw.de, e-mail: knoll@in.tum.de
Lars.Harscheidt@bmw.de

Abstract—Automotive industry faces challenges in of the state of the art about smart manufacturing, Internet of
manufacturing like increasingly individualized products with a Things and data analytics, which are key concepts for an
short lead-time to market and higher quality. Additionally to autonomous production system based on prescriptive
that, new technologies, such as Internet of Things (IoT), big analytics. Section V is a review of key elements and
data, data analytics and cloud computing, are changing the concepts for prescriptive analytics, on the shop floor. Based
production into the next generation of industry. To address on this review Section VI provides a recommendation for
these challenges intelligent manufacturing in combination with action for a framework to control production proactively and
data analytics plays an important role. In this sense, the autonomously based on prescriptive analytics. Finally,
integration of prescriptive analytics in manufacturing may
Section VII concludes the paper.
help industry to increase productiveness. This paper provides
first a comprehensive review of key elements for prescriptive II. SMART MANUFACTURING
analytics in manufacturing. Furthermore, this paper highlights
requirements for a prescriptive analytics based production Smart manufacturing uses advanced sensing-, control-,
control, so called prescriptive automation, and finally points modelling- and platform technologies with the aim of
out field of activities in this topic. optimizing production transactions through the full use of
advanced information and manufacturing technologies [3].
Keywords-industry 4.0; smart manufacturing; data analytics; The visualization of the process performance through the
prescriptive analytics; prescriptive automation; internet of things; acquisition of huge volumes of real time data is one
review application [4]. Smart control, as part of smart manufacturing,
not only considers the visualization, but also the intelligent
I. INTRODUCTION control of production facilities that can interact in real time
[6]. The goal is to use methods to control and optimize the
Whereas production systems in earlier times consisted of
production process. Depending on which input is available
purely mechanical and electrical components, they are
and which output should be controlled, different models for
nowadays complex systems that combine hardware and
control strategies can be used [7]. Based on the
software in different ways. Digital processes have an
understanding of data analytics, a feedback can be provided
exponentially increased range of functions compared to
to employees or machines, and thus the utilization and
conventional processes and redefine the traditional process
production processes can be optimized [5].
boundaries [1]. Digitized production systems are based on
A pioneer in smart manufacturing is the IoT. Its devices
the acquisition, processing and provision of information for
include sensors, actuators and computers with wireless
machines and objects. The distribution, analysis and target-
networks among other things that contribute to automation
orientated use of the information offers manifold potentials
and monitoring [8]. IoT realizes its potential through the
for an autonomous control of production processes. For that
holistic integration of its three components: intelligent
it is important to know what is happening, what will happen
devices, intelligent systems and intelligent decisions [9].
and how to react proactively and autonomously.
Nevertheless, this proactive control leads to new III. INTERNET OF THINGS
requirements and can only be achieved on the basis of
comprehensive data, so called big data [2] and a change of IoT is defined as “a dynamic global network
the architecture for the control and data network in infrastructure with self-configuring capabilities based on
production [3]. standard and interoperable communication protocols where
Therefore, this paper provides a review about key physical and virtual ‘things’ have identities, physical
elements for data analytics, especially prescriptive analytics, attributes, and virtual personalities and use intelligent
on the shop floor. Based on this literature review, this paper interfaces, and are seamlessly integrated into the information
identifies requirements for efficient data analytics in network” [10].
manufacturing and a smart production control and highlights While IoT applications in the consumer goods market
fields of action. From this point forward, the paper is receive a great deal of public attention, the Industrial Internet
structured as follows: Section II, III, IV provide an overview of Things (IIoT) represents an enormous potential for the

978-1-7281-3268-6/19/$31.00 ©2019 IEEE 224

industry and can cover almost any industrial sector [11]. The decision-making follows the analysis by Stein, Meller and
IIoT, also known as the Industrial Internet, characterizes the Flath [18]:
integration and cooperation between machines, analytics and x Decision making at the enterprise level
people. The aim is to connect the information processes to x Decision making at departmental level
make production more intelligent [12]. Therefore, e.g. x Decision making at the individual level
machines, robots, transport systems, workpieces and control Within the classification of decision making on an
systems must communicate with each other to achieve a individual level, one topic of interest is the intelligent shop
higher degree of integration. The enormous amount of data floor control with the help of prescriptive analytics [18].
generated by various elements of the intelligent factory The production control is responsible for the decision
allows optimizing manufacturing processes to be more process on the shop floor. As mentioned in the previous
efficient and flexible. Additionally, production costs can be section, this should optimize the operative work of individual
reduced and product quality can be improved [13]. However, processes. The relevant work in this research field is
to achieve useful information from the enormous amount of summarized in the following chapter and we categorized it
data, data analytics can offer a solution. on five essential categories that are required for prescriptive
analytics on shop floor. These include data acquisition,
IV. DATA ANALYTICS connectivity, data storage, data processing and control. The
Although the development of the IoT has enhanced data contributions are listed chronologically within the categories.
collection, the question remains how this data can be On the basis of this research work, this article recommends
properly processed to provide the right information for the further research for a proactive autonomous control of
right purpose at the right time [9]. Big data analytics become production processes based on prescriptive analytics.
a key base for competitiveness, productivity growth and
innovation, because in a big data environment, data sets are V. PRESCRIPTIVE ANALYTICS ON SHOP FLOOR
much larger and can be too complex for traditional data
analytics software. Therefore, it is a challenge to analyze the A. Data Acquisition
content of these huge, continuous data streams, and Data acquisition serves as a key factor for intelligent
additionally to build more robust and intelligent learning manufacturing, because real time status information is
systems [8]. essential for data analytics and production control. The
Analytics consists of two main areas: business concept of IoT apply the authors to the production area and
intelligence (BI) and advanced analytics (AA). With regard consider the use of an industrial wireless sensor network
to the analysis horizon, a distinction can be made between (IWSN) for condition monitoring and fault diagnosis of
four levels [14]: machines [19]. Therefore the article describes the usage of a
x Descriptive: What happened? (BI) hybrid framework of fog computing for data pre-processing
x Diagnostic: Why did it happen? (BI) within the sensors in combination with a cloud computing
x Predictive: What will happen? (AA) infrastructure for further data processing [19]. Zhong, Xu,.
x Prescriptive: What to do? (AA) Klotz and Newman provide a comprehensive overview of
Business intelligence focuses on reporting and queries. IoT-enabled manufacturing [9]. Also an IoT-based
Advanced analytics, also called business analytics, goes architecture for production measurement systems is
beyond business intelligence by using sophisticated developed by Hwang, Lee, Park and Chang based on ISA-95
modelling techniques to predict future events or discover and ISO-22400, which can use data and planning
patterns that otherwise cannot be identified. Advanced information from MES and ERP systems through interfaces
analytics is about optimizing, correlating and predicting the [20]. Giusti, Bevilacqua, Tedeschi and Emmanouilidis
next best action. Advanced analytics is divided in two parts: further show concrete examples of how IIoT can be used to
predictive and prescriptive analytics [14]. improve process monitoring and demonstrate a cost-effective
retrofit of old machines with IoT devices in an existing
A. Prescriptive Analytics production environment [5]. It offers sensor functions and
Prescriptive analytics serves to determine a sequence of internet connection. The system enables users to capture key
decisions to obtain a desired result and generally answers the process performance metrics and can be alerted of anomalies,
question "What do I have to do to achieve a desired goal?” as well as remote monitoring via a cloud infrastructure [5].
[15]. This ensures an adaptive, autonomous, time-based and
B. Connectivity
optimal decision and recommends the best approach to
achieve specific key performance indicators [16]. Connectivity enables extensive connections and
Prescriptive systems have two important characteristics. exchanges of knowledge. An architecture with smart
First, they deliver realizable results in the form of gateway and fog-computing is developed by Aazam and Huh,
recommendations for action. Secondly, the quality of the which allows a reduction of the communication load for the
recommended activity is reviewed with regard to its core network and more efficient cloud services as well as
correctness [17]. real time communication for delay-sensitive applications
The current literature suggests that prescriptive analytics [21]. Wang, Wan, Li and Zhang propose an architecture for a
is divided into three planning levels, at which companies use self-organizing IoT network for production control in
data-driven decision-making systems. This classification of conjunction with data analytics in the cloud and mobile

225
devices [22]. The approach of Nino, Saenz, Blanco and IoT framework with cloud computing and feedback of
Illarramendi considers an architecture scheme for data recommendations for action to process participants. The
acquisition at field level and data processing with cloud framework helps reducing the energy consumption of AM
computing in a heterogeneous IT infrastructure, which serves processes [32]. Bassat shows that a digital assistant enables
as a case study for an applied research project on error manufacturers to meet stringent requirements in the future
detection using data analytics [23]. The authors extend the and to remain sustainable and competitive. The developed
possibilities for production control through an architecture digital assistant enables the analysis of big data collected by
based on cloud computing, CPS and IoT [24]. Production IIoT sensors in combination with cloud computing and
control is the focus of Sunny, Liu and Shahriar. The research prescriptive analytics. This supports employees in production
shows an approach of an agent-adapter architecture for in the management of the raw material by specifying the
remote control over the internet and the cloud [25]. action and thus optimizing the work process and quality [33].
The paper proposes an architectural design and a software
C. Data Storage framework for the rapid development of prescriptive
An efficient data storage is essential to analyze big data analytics solutions for dynamic production processes in the
on shop floor. Therefore, the following approach considers automotive industry. The architecture supports the storage of
storage of the acquired data and describe a scalable service modular, expandable and reusable knowledge bases of
architecture for an analytics system that enables query process performance models. The decision support system is
processing and data analytics of data streams [26]. The demonstrated by a prototype to illustrate the principles of the
service provides uncompressed and correlated data in a proposed architectural framework [15]. Stein, Meller and
warehouse for further analysis. On this occasion, descriptive, Flath consider the development of a sensor-based decision
predictive and prescriptive analytics tools can be used [26]. support tool for a manual leak detection process and
From a business point of view the article recognizes that two discusses the development of a prescriptive framework for
central prerequisites are necessary for efficient and effective localizing a leak based on sensor data. The article mentions
manufacturing processes: process transparency and process an integrated framework for prescriptive analytics of manual
responsiveness [27]. The authors address process processes in production environments in form of an
transparency through a concept for a holistic, production- analytically supported production system. For this purpose,
specific process data warehouse. This integrates operating data is acquired during the production process and machine
and process data into a standardized multidimensional learning algorithms are applied to train predictive models
warehouse and is based on a generalized meta model of the based on the sensor data. The algorithm creates an individual
manufacturing process [27]. action prediction [18].
D. Data Processing E. Control
Analysis of the provided data plays an elementary role in New technologies such as IoT, big data and cloud-based
decision making on shop floor. Gröger, Schwarz and services are currently changing the field of control
Niedermann develop an analytics platform for the concept of technology. Babiceanu and Seker expect advanced
real time prediction and process optimization. This platform production environments to become reality. Therefore the
combines relevant data and provides data analytics paper proposes modeling guidelines that include IoT
functionalities. Based on local processing, the approach connectivity, complex event processing, and big data
concentrates on a data warehouse for prescriptive analytics analytics for operational prediction [2]. The article of Gupta
for production control [28]. and Chow identifies some of the key research topics related
The research project iPRODICT has the goal to realize to networked control systems (NCS). These include, for
predictive and prescriptive analytics and thus to optimize example, network delay compensation and resource
processes. The article analyses the integration of different allocation. With increasing applications for NCS, real time
technologies in order to enable sensor-based decision making control is an important issue [34]. The following paper also
in real time for process improvement in the process industry. takes up this topic and separates the physical location of the
Within the iPRODICT project, the authors address production control from the production itself. The author
prescriptive control of processes through event-based process recognizes that classical computing is gradually moving into
predictions based on big data, with a focus on production the cloud and offering completely new possibilities in the use
planning and control in the context of the process industry of information. Therefore, the approach proposes the
[29]. The paper concentrates on the adaptation and implementation of a cloud-based control architecture, the so-
optimization of production processes through the proactive called Machine Control as a Service [35]. Coupek, Lechler
analysis of part quality based on sensor data. For this and Verl connect sequential production and assembly
purpose the authors demonstrate a suitable IT architecture processes via a cloud-based architecture that allows
which enables the provision of various real time analyses information from a previous production step to be used in
[30]. Additive manufacturing (AM) is one of the most one of the subsequent steps for deviation compensation. In
popular applications of data analytics in production [31]. A this way, the authors recognize that cloud computing offers
generic prescriptive analytic method to understand the new possibilities in the use of information and develops a
geometric deformation of products in AM is the topic of Qin, cloud-based control architecture in the production line of
Liu and Grosvenor. The authors develop a service-oriented rotors. The aim is to generate deviation compensations for

226
subsequent production steps on the basis of collected data execute the prescriptive analytics model. This model controls
[36]. A Control System as a Service (CSaaS) is developed by the operational process autonomously. Based on the
the authors [37]. It enables the control of a production line in continuously grown database, the framework has to offer the
Auckland, New Zealand via the cloud from Stuttgart, possibility to update and train the prescriptive analytics
Germany by using an NCS. Within the framework, the model. Finally, this updated model has to be deployed by the
controller, sensors, actuators and other system components framework to the controlling system.
exchange information via a common network. However, it
turned out that CSaaS between New Zealand and Germany is TABLE I. OVERVIEW OF RELEVANT SCIENTIFIC CONTRIBUTIONS
FOR PRESCRIPTIVE ANALYTICS ON SHOP FLOOR
not possible due to network challenges, so the controller
should be closer to the machine [37]. Steiner and Poledna

shop floor with

shop floor with

Hybrid cloud –

manual closed
fog computing

based control

based control
conclude that the IIoT softens the rigid layers of the

analytics on

analytics on
Prescriptive

Prescriptive

closed loop
computing

computing

automated
automation pyramid by introducing fog computing as an

Cloud

loop
Fog
architectural measure for linking IIoT and process
automation [38]. According to Patel, Ali and Sheth, fog
computing offers production control advantages over cloud
computing, but it cannot replace cloud computing because [2] ○ ۚ ○ ۚ ۚ
many applications require both fog localization and cloud [6] ○ ۗ ○ ○ ۚ
globalization, particularly for analytics and big data. For this [15] ○ ○ ○ ● ○
reason the approach proposes a parallel use of fog and cloud
computing [39]. Barton, Maturana and Tilbury recognize that
[18] ○ ○ ○ ۛ ○
manufacturing systems should close the loop and transform [19] ۛ ○ ۛ ○ ○
IoT data into production knowledge. The authors develop a [28] ○ ○ ○ ۛ ۚ
bidirectional framework for a closed loop from sensor data to
machine control to compare simulation and operating data to
[29] ○ ○ ○ ۛ ۚ

optimize control approaches [40]. [30] ○ ○ ○ ۛ ۚ

[31] ○ ○ ○ ۗ ۚ
VI. RECOMMENDATION FOR FURTHER RESEARCH
[32] ○ ○ ○ ۗ ۚ
The following research topics are the result of the five
relevant categories for prescriptive analytics on shop floor as
[33] ○ ○ ○ ● ○
defined in Section V. Production control usually follows a [35] ○ ۗ ○ ○ ○
fixed logic, with conditional consideration of dependencies [36] ○ ۛ ○ ○ ○
and correlations based on a usually outdated data basis [41]. [37] ○ ۛ ○ ○ ○
Digitalization opens new ways to improve decision making
through up-to-date and comprehensive data. Table I shows
[39] ۗ ○ ۗ ۚ ۚ

an overview of the authors regarding relevant scientific [40] ۗ ۚ ○ ۗ ۚ

contributions for prescriptive analytics on the shop floor. Table ○ not ۚ idea ۗ concept ۛ concept ● concept
legend: considered considered considered developed implemented
However, it only lists papers dealing with connectivity,
control and data processing (see Section V). As the review VII. CONCLUSION
and Table I show, process data extraction for data analytics
is often considered, but analysis and optimization takes place This paper deals with the change from a purely
ex post the production process in the form of a decision mechanical and electrical to a data-driven and data analytics-
support system. However, the support system still requires a supported production. The paper focuses on the special case
human being as the top decision maker and manual executive of prescriptive analytics on shop floor. Based on the
element. Therefore, some authors mentioned the idea for scientific literature, five categories for prescriptive analytics
prescriptive analytics on the shop floor with an automated on shop floor with automated closed loop are determined. A
closed-loop, but they do not conceptually develop and comprehensive review of the scientific contributions in these
implement it. categories is given. As the review shows the idea for a
As a result, a research gap is an autonomous, proactive prescriptive automation is suggested, but not conceptually
control of the production process by optimizing process developed and implemented. To address this, a framework
parameters by prescriptive analytics during execution, so which has the in Section VI considered features have to be
called prescriptive automation. A key part of this gap is to developed.
develop an integrated framework for prescriptive automation.
REFERENCES
This framework allows to control the process ex ante based
on a prescriptive analytics model and autonomous without a [1] M. Porter, J. Heppelmann, “How smart, connected products are
transforming competition,” Harvard Business Report, Nov. 2014.
human as a manual executive. This framework has to be
[2] R. F. Babiceanu, R. Seker, “Big Data and virtualization for
interoperable with the heterogeneity of communication manufacturing cyber-physical systems: A survey of the current status
networks and changes in IIoT for a wide data acquisition on and future outlook,” Computers in Industry, vol. 81, Sep. 2016, pp.
the shop floor. Due to the fact that a huge amount of data is 128-137.
generated, scalable computational power is needed to

227
[3] J. Davis, T. Edgar, R. Graybill, P. Korambath, B. Schott, D. Swink, et [23] M. Nino, F. Saenz, J. M. Blanco, A. Illarramendi, “Requirements for
al., “Smart manufacturing,” Annual Review of Chemichal a big data capturing and integration architecture in a distributed
Biomolecular Engineering, vol. 6, 2015, pp. 141–160. manufacturing scenario,” 2016, pp. 1326–1329.
[4] P. Evants, M. Annunziata, “Industrial internet: pushing the [24] V. D. Majstorovic, D. Mourtzis, E. Vlachou, “Cloud-based cyber-
boundaries of minds and machines”. physical systems and quality of services,” The TQM Journal, vol. 28,
[5] F. Giusti, M. Bevilacqua, S. Tedeschi, C. Emmanouilidis, “Data 2016, pp. 704–733.
analytics and production efficiency evaluation on a flexible [25] S. M. N. A. Sunny, X. F. Liu, M. R. Shahriar, “Communication
manufacturing cell,” 2018, pp. 1–6. method for manufacturing services in a cyber–physical manufacturing
[6] P. Zheng, H. Wang, Z. Sang, R. Y. Zhong, Y. Liu, C. Liu, et al., cloud,” International Journal of Computer Integrated Manufacturing,
“Smart manufacturing systems for Industry 4.0: Conceptual vol. 31, 2018, pp. 636–652.
framework, scenarios, and future perspectives,” Front. Mech. Eng., [26] S. Ayhan, J. Pesce, P. Comitz, G. Gerberick, S. Bliesner, "Predictive
vol. 13, 2018, pp. 137–150. analytics with surveillance big data,” Proceedings of the 1st ACM
[7] H. Kagermann, W. Wahlster, J. Helbig, “Recommendations for SIGSPATIAL International Workshop on Analytics for Big
implementing the strategic initiative Industrie 4.0: Final report of the Geospatial Data2012, pp. 81–90.
Industrie 4.0 Working Group,” 2013. [27] C. Gröger, J. Schlaudraff, F. Niedermann, B. Mitschang,
[8] Q. P. He, J. Wang, D. Shah, N. Vahdat, “Statistical Process “Warehousing manufacturing data. A holistic process warehouse for
Monitoring for IoT-Enabled Cybermanufacturing: Opportunities and advanced manufacturing analytics,” 14th International Conference on
Challenges,” IFAC-PapersOnLine, vol. 50, 2017, pp. 14946–14951. Data Warehousing and Knowledge Discovery (DaWaK), 2012, pp.
142–155.
[9] R. Y. Zhong, X. Xu, E. Klotz, S. T. Newman, “Intelligent
manufacturing in the Context of Industry 4.0: A Review,” [28] C. Gröger, H. Schwarz, B. Mitschang, “Prescriptive analytics for
Engineering, vol. 3, 2017, pp. 616–630. recommendation-based business process optimization,” 17th
International Conference on Business Information Systems (BIS), pp.
[10] R. van Kranenburg, “The Internet of Things: A Critique of Ambient 25–37.
Technology and the All-Seeing Network of RFID,” Amsterdam, The
Netherlands: Institute of Network Cultures, 2007. [29] J. Krumeich, D. Werth, P. Loos, “Prescriptive control of business
processes,” Bus Inf Syst Eng, vol. 58, 2016, pp. 261–280.
[11] A. Gilchrist, "Industry 4.0: The Industrial Internet of Things, 2016.
[30] N. Mehdiyev, A. Emrich, B. Stahmer, P. Fettke, P. Loos, “iPRODICT
[12] B. Heinrich, P. Linke, M. Glöckler, “Basics of automation: sensor – Intelligent Process Prediction based on Big Data Analytics,” CEUR
technology, regulation, control (english); Grundlagen Workshop Proceedings, 2017, pp. 13–24.
Automatisierung: Sensorik, Regelung, Steuerung (german),” 2017.
[31] Y. Jin, S. J. Qin, Q. Huang, “Prescriptive analytics for understanding
[13] P. Gerbert, M. Lorenz, M. Rüßmann, M. Walder, J. Justus, P. Engel,
of out-of-plane deformation in additive manufacturing,” IEEE
et al., “Industry 4.0: The future of productivity and growth in
International Conference on Automation Science and Engineering,
manufacturing industries,” 2015.
2016, pp. 786–791.
[14] C.-H. Wang, H.-Y. Cheng, Y.-T. Deng, “Using Bayesian belief
[32] J. Qin, Y. Liu, R. Grosvenor, “A framework of energy consumption
network and time-series model to conduct prescriptive and predictive
modelling for additive manufacturing using Internet of Things,"
analytics for computer industries,” Computers & Industrial Procedia CIRP, vol. 63, 2017, pp. 307–312.
Engineering, vol. 115, 2018, pp. 486–494.
[33] A. Ben Bassat, “Your digital assistant: Ai and industrial iot in the
[15] A. Brodsky, G. Shao, M. Krishnamoorthy, A. Narayanan, D. Menascé, service of advanced manufacturing optimization,” International
R. Ak, “Analysis and optimization based on reusable knowledge base SAMPE Technical Converence, 2017, pp. 245–257.
of process performance models,” Int J Adv Manuf Technol, vol. 88,
2017, pp. 337–357. [34] R. A. Gupta, M.-Y. Chow, “Networked control system: Overview and
research trends,” IEEE Transactions on Industrial Electronics, vol. 57,
[16] A. Basu, “Executive Edge: Five pillars of prescriptive analytics 2010, pp. 2527–2535.
success,” 2013.
[35] A. Verl, A. Lechler, S. Wesner, A. Kirstädter, J. Schlechtendahl, L.
[17] R. Soltanpoor, T. Sellis, “Prescriptive analytics for big data,” 2016, Schubert, et al., “An Approach for a Cloud-based Machine Tool
pp. 245–256. Control,” Procedia CIRP, vol. 7, 2013, pp. 682–687.
[18] N. Stein, J. Meller, C. M. Flath, “Big data on the shop-floor: Sensor- [36] D. Coupek, A. Lechler, A. Verl, "Cloud-Based Control Strategy:
based decision-support for manual processes,” J Bus Econ, vol. 88, Downstream Defect Reduction in the Production of Electric Motors,"
2018, pp. 593–616. IEEE Trans. on Ind. Applicat., vol. 53, 2017, pp. 5348–5353.
[19] D. Wu, S. Liu, L. Zhang, J. Terpenny, R. X. Gao, T. Kurfess, et al.,
[37] J. Schlechtendahl, F. Kretschmer, Z. Sang, A. Lechler, X. Xu,
“A fog computing-based framework for process monitoring and
“Extended study of network capability for cloud based control
prognosis in cyber-manufacturing,” Journal of Manufacturing
systems,” Robotics and Computer-Integrated Manufacturing, vol. 43,
Systems, vol. 43, 2017, pp. 25–34.
2017, pp. 89–95.
[20] G. Hwang, J. Lee, J. Park, T.-W. Chang, “Developing performance
[38] W. Steiner, S. Poledna, “Fog computing as enabler for the Industrial
measurement system for Internet of Things and smart factory
Internet of Things,” Elektrotech. Inftech., vol. 133, 2016, pp. 310–
environment,” International Journal of Production Research, vol. 55, 314.
2017, pp. 2590–2602.
[39] P. Patel, M. Intizar Ali, A. Sheth, “On Using the Intelligent Edge for
[21] M. Aazam, E.-N. Huh, “Fog Computing and Smart Gateway Based IoT Analytics,” IEEE Intell. Syst., vol. 32, 2017, pp. 64–69.
Communication for Cloud of Things,” 2014, pp. 464–470.
[40] K. Barton, F. Maturana, D. Tilbury, “Closing the Loop in IoT-enabled
[22] S. Wang, J. Wan, Di Li, C. Zhang, “Implementing Smart Factory of Manufacturing Systems: Challenges and Opportunities,” Annual
Industrie 4.0: An Outlook,” International Journal of Distributed American Control Conference, 2018, pp. 5503–5509.
Sensor Networks, vol. 12, 2016.
[41] G. Wellenreuther, D. Zastrow, “Automation with PLC - theory and
practice (english); Automatisieren mit SPS - Theorie und Praxis
(german),” 2015. Wiesbaden: Springer Vieweg.

228