International Workshop of Advanced Manufacturing and Automation (IWAMA 2016)

Intelligent Predictive Maintenance (IPdM) for

Elevator Service
Through CPS, IOT&S and Data Mining

Kesheng Wang Guohong Dai and Lanzhong Guo

Knowledge Discovery Laboratory, Department of School of Mechanical Engineering
Production and Quality Engineering Changshu Institute of Technology,
NTNU, Trondheim, Norway Changshu, Jiangsu, China
[email protected] [email protected] and [email protected]

Abstract— With the rapid economic growth and urbanization safety of boarding passengers will be the central questions to
development, most of elevator production and service companies elevator industries.
have completely changed their service policy opting to eliminate
the standard preventive service policy. They now provide some The elevator service business has, and continues, to
form of predictive service policy and emphasize that they utilize undergo many changes. Traditional elevator maintenance
remote monitoring of elevators to detect faults and estimate when service is not able to meet the requirement of customers. For
components may need to be maintained due to actual usage. example, remote monitoring of elevators appeared in the late
However, most of them are not true predictive maintenance 1980s. While remote monitoring would alert the elevator
policy. They merely a preventive, slightly enhanced, usage-based company when an elevator had a breakdown, it did not in and
program. What is the future of elevator production and service of itself reduce the number of breakdowns. A decade later in
companies? This paper will challenge the word “predictive the 1990s, the model evolved to usage-based maintenance – a
maintenance” and present the framework of intelligent predictive concept based around adjusting the level and frequency of
maintenance system for smart elevator service based on industry repairs dependent on each elevator’s usage. This method had
4.0 concepts. been adapted from the automobile industry, where motor oil
was often changed based upon distance travelled. In the
Keywords—Industry 4.0; CPS; IOT&S; Big data/ Data Mining; elevator industry, this paved the way for gradual progression
Elevator service; Smart elevator
towards more condition-based maintenance policy.

I. INTRODUCTION Each of these developments demonstrated the ongoing

industry intent to have completely restructured their service
According to the World Health Organization (WHO) policies aiming to eliminate the standard “preventive
estimates that by 2050 there will be 70 per cent of people living maintenance” service. Market competition coupled with cost
in large cities. With the rapid economic growth and increases has been a driver for searching new innovative
urbanization development, China’s elevator industry and approaches of elevator maintenance service.
university are facing an unprecedented opportunity for research
and development. Elevator ownership, annual production and With the development of Industry 4.0 concept. It gives
annual growth are all ranked at top one of the world. In the more possibilities for elevator industry. A new methodology
recent decade, the Chinese elevator has kept an average annual with the power to take the learnings of these earlier
growth rate of around 20%, with the volume of elevators, maintenance systems and incorporate the benefits of the new
which are being used and registered in Quality Departments systems and technologies we now have access to, was needed
throughout the country growing from 346,067 in 2002 to to create a new efficient and effective practice for elevator
3,009,268 at the end of 2013. [1] repairs to enhance overall building efficiency. ThyssenKrupp
has cooperated with Microsoft to develop MAX system for
It is expected that in 2016 the total amount of elevators will predictive maintenance and manage to detect and predict
be over nearly 4.5 million units in China and the activities of failures of elevators. [3]
maintenance, repair, update, change of elevators will give
service companies tremendous opportunity and economic Nowadays, many elevator companies will say they now
value. The past five years, the number of various elevator provide some form of “predictive maintenance” service and
accident has increased doubling. Factors leading to the elevator emphasize that they utilize remote monitoring of elevators to
safety hazards are 16% for the manufacturing quality, 24% gain visibility of faults and estimate when components may
related to installation, while up to 60% caused by maintenance need to be replaced due to actual usage. However, when asked,
and usage of elevators. [2] Therefore, a reliable and safe usage the typical company representative will explain that their
and maintenance of elevator systems is most essential. How to policy monitors the elevator and therefore they know when it
protect the elevator routine maintenance and How to ensure the “faults” or that they will replace a consumable part before its

© 2016. The authors - Published by Atlantis Press 1

useful life has expired. Is this a “predictive maintenance” and similar problems as something that can be anticipated and
policy? Or, is it merely a preventive, slightly enhanced, usage- dealt with before they occur.
based policy? [4]
Proactive maintenance consists of Preventive maintenance
This paper will challenge the word “predictive and Predictive maintenance.
maintenance” and review what is required to be truly
“predictive maintenance”. Based on Industry 4.0 concept, the C. Preventive vs Predictive
framework of Intelligent Faults Diagnosis and Prognosis
1) Preventive manintenance
System (IFDaPS) is proposed for elevator industry to develop
their own predictive maintenance system in the future.  Definition: Applies a planned maintenance service
before a failure occurs.
II. CALSSIFICATION OF MAINTENANCE  Characteristics: Simple, easy to implement, reliable, but
low efficient.
There are many ways to define maintenance policy. [5-11], 2) Predictive maintenance
In general, the maintenance policy can be classified as two  Definition: Carries on the philosophy “execute at the
types: (1) Reactive maintenance and (2) Proactive right time”. Service actions take place only when
maintenance. We focus on the techniques that are used to do necessary.
maintenance tasks. The classification of Maintenance is shown
in Fig. 1.  Characteristics: High efficient, reliable, but requires
mass data and complicated analytical models
Maintenance
3) Implementation of predictive maintenance
The implementation of the predictive maintenance
addresses several concerns, customer needs, organizational
Reactive Proactive requirements and feasible technologies. Quickly the predictive
maintenance became the desired industry standard in most
markets. Some questions are related to implementation of
Corrective Preventive Predictive predictive maintenance, for example, Did the standard
preventive maintenance simply become something marketed as
“predictive” when it really didn’t have the necessary elements
Deferred Immediate
Reliability
based
Time
based
Statistical
based
Condition
based
to be “predictive”? or Was it merely a usage-based policy that
fell short of actually “predicting” failures based upon a
component’s, or a system’s, behavioral patterns?
Opportunity Design Out

4) Predictive maintenance for elevators

Fig. 1. Classification of Maintenance policy Predictive maintenance can be divided two types:
Statistical-based predictive maintenance and Condition-based
predictive maintenance. Statistical-based predictive
A. Reactive maintenance maintenance (SBM) depends on statistical data from the
Reactive maintenance (also known as "breakdown meticulous recording of the stoppages of the in-plant items and
maintenance") are repairs that are done when equipment has components in order to develop models for predicting failures,
already broken down. while Condition-based predictive maintenance (CBM, also
called Condition-based maintenance) depends on continuous or
Reactive maintenance focuses on restoring the equipment periodic condition monitoring equipment to detect the signs of
to its normal operating condition. The broken-down equipment failure. There are a slight difference between predictive
is returned to working within service specifications by maintenance and condition based maintenance. CBM is focus
replacing or repairing faulty parts and components. Emergency on condition monitoring and humans make the decision of
repairs cost to times more than planned repairs, so maintenance maintenance. While predictive maintenance more focuses on
plans that rely on reactive maintenance are generally the most data analysis and data mining for support maintenance decision
expensive. Reactive maintenance is so expensive because making automatically. Recently, Industry 4.0 will more
shutdowns happen during production runs (instead of pre- concern data analysis rather than only condition monitoring.
scheduled maintenance shutdowns during downtimes); because
expedited shipping for spare parts costs much more than Condition monitoring is a pre-condition to implement
regular shipping; and because maintenance staff is often forced predictive maintenance. Modern elevators are highly intelligent
to work overtime to repair machinery. systems running sophisticated, proprietary remote monitoring
system producing a wealth of data within every movement, or
B. Proactive Maintenance desired movement. In order to be beneficial in creating the
predictive maintenance, the remote monitoring, or M2M
Proactive maintenance is a preventive maintenance strategy
(Machine to Machine) data, must allow for immediate access to
for maintaining the reliability of machines or equipment. The
data, analysis of the elevator’s intelligence, component and
purpose of proactive maintenance is to view machine failure
system trend analysis, analytical models to predict component
and system failures based upon established patterns,

2
dissemination of proper knowledge to service operation teams, Industry 4.0, which is buzzwords discussed among
and accurate inventory predictions. Clearly, any service practitioners as well as theorists, will facilitates the version of
provider can perform “dumb” remote monitoring. The concepts smart factory. [15, 16] It was introduced at Hanover fair in
of Industry 4.0 have help us to develop a predictive 2011 in Germany to present a new trend towards the
maintenance system systematically. I session III, we will networking of traditionally industries such as manufacturing.
introduce the concepts of industry 4.0 and its technical Similarly, in the United States, an initiative known as the Smart
components. In session VI, a detail framework of intelligent Manufacturing Leadership Coalition (SMLC) is also working
predictive maintenance for elevators is presented. on the future of manufacturing [17]. UK has been working on a
strategy on “bring manufacturing back to UK” [18]. China
III. INDUSTRY 4.0 adopts “smart Manufacturing” strategy to seek innovation-
driven development, which is called “China Manufacturing
A. Indystry revolutions 2025” or “Made in China 2025” [19]. There many similar
programs and projects in the world, such as “Intelligent
A Manufacturing is and continues to be an essential part of manufacturing system” from Japan, Canada, European Union,
world’s economy. Smart manufacturing will be capable of Switzerland and Norway [20]. “Future of Manufacturing” from
rapidly adapting their physical and intellectual infrastructures Norway [21] and “Ubiquitous manufacturing” from South
to exploit changes in technology as manufacturing becomes Korea [22].Technical components and definition of Industry
faster, more responsive to changing global markets and closer 4.0
to customer’s needs.
Industry 4.0 enables the manufacturing of individual and
Industrial production has been changed since it is in very customized products at the same cost of mass production,
beginning. Often, the changes have been so powerful that the which are manufactured by a smart factory with high
term of revolution has been used to describe it. The name automation and efficiency. For production companies, this
Industry 4.0 recognized the existence of three previous specifically means that they are able to overcome predominant
industrial revolutions. interface issues between product development, production and
1) Industry 1.0 product usage, and thus orient all main value-adding processes
The first industrial revolution is the term used to describe towards the customer’s requirements. New development
the change from purely manual work to machine production, processes, e.g. integrated product and production system
which initially affected the cotton-spinning and weaving mills development, intensify exchange between departments and
in England from 1770. The great breakthrough came in 1782 companies. Furthermore, industry 4.0 enables intelligent and
with the steam engine invented by James Watt. From this on it flexible production control using IT-based intercommunicating
was possible to have an energy supply at any location and the and interacting machines, products, services, equipment and
manual work was no longer focused tools. M2M communication and networking bridges
department boundaries and promotes the company’s orientation
2) Industry 2.0 towards the customer.
The second industrial revolution was characterized by the
principles of rationalization by Taylor. It is mainly based on the 5) Technical components of Industry 4.0
division of labor, standardization, precision manufacturing, and In general, Industry 4.0 may consists of 4 components: 1.
assembly line work. Henry Ford applied the first conveyor belt Cyber-Physical Systems (CPS); 2. Internet of Things (IOT); 3.
for the production of the T-Model and achieved pioneering Big data & Data Mining (DM); 4. Internet of Service (IOS),
success with it in the automobile manufacturing at the seeing Fig. 1.
beginning of the 20th century.
a) Cyber-Physical Systems (CPS)
3) Industry 3.0 An important component is Cyber-Physical Systems (CPS),
The third industrial revolution was based on the which transfer the physical world into the virtual one. They can
development of the computer and IT technology. This led to be understood as a basic unit in the system. The development
numerically controlled machines, such as NC machines and of CPS is characterized by three stages. The first generation of
industrial robots, which could be modified much faster than CPS includes identification technologies like RFID tags, which
conventional mechanical automated machines. Thus, the allow unique identification of objects. Storage and analytics
flexible automation came into being and systems were have to be provided as a centralized service. The second
characterized by high productivity and flexibility generation of CPS are equipped with sensors and actuators
with a limited range of functions. CPS of the third generation
4) Industry 4.0
can store and analyze data, are equipped with multiple sensors
Now we are standing on the brink of the fourth industrial and actuators, and are network compatible.
revolution, so called Industry 4.0. Information and
communication technologies (ICT) are growing together and b) Internet of thing (IoT)
affecting all areas of life. Devices and systems in our real The Internet of Thing (IOT) allows “Things or Objects”
environment that are controlled by embedded software are interact with each other and cooperate with their “smart”
integrated into the global communication network, where components to reach common aims. Based on CPS given in
“internet” is the key term. The real world and the virtual world section 2.1, CPS can be defined as “Things or Objects”.
are clearly growing together.

3
Therefore, the IOT can be thought as a network where CPS Nevertheless, for quality assurance and traceability it is
cooperate with each other through unique addressing schemas. necessary to keep track of the whole system at any time.
c) Data mining (DM) d) Real-time capability
Real-time big data isn’t just a process for storing a huge For organizational tasks, it is necessary that data is
amount of data in a data base or warehouse. Data Mining collected and analyzed in real time. The status of the plant is
enables you to analyze and discover patterns, rules and permanently tracked and analyzed. Thus, the plant can react to
knowledge from big data collected from multiple sources. So the failure of a machine and reroute products to another
you can make the right decision at the right time and right machine.
place.
e) Service Orientation
d) Internet of service (IOS) The services of companies, CPS, and humans are available
The Internet of Services (IoS) enables service vendors to over the IoS and can be utilized by other participants. They can
offer their services via the internet. The IoS consists of be offered both internally and across company borders. All
business models, an infrastructure for services, the services CPS offer their functionalities as an encapsulated web service.
themselves and participants. Services are offered and combined As a result, the product specific process operation can be
into value-added services by various suppliers. They are composed based on the customer specific requirements
communicated to users as well as consumers and accessed by provided by the RFID tag.
them via various channels.
f) Modularity
Following the 4 key components of Industry 4.0, a general Modular systems are able to adapt to changing
and explicit definition of Industry 4.0 can be given as the requirements by replacing or expanding individual modules
following. flexibly. Therefore, modular systems can be easily adjusted in
case of seasonal fluctuations or changed product
B. Definition of Industry 4.0 characteristics.
Based on the literature review, the general definition of
Industry 4.0 is that Industry 4.0 is a collective term for g) Security
technologies and concepts of value chain organization. Within Industry 4.0 will use ICT technologies for data transmission
the Smart Factories of Industry 4.0, CPS monitor physical and processing. The security and privacy of the information
processes, create a virtual copy of the physical world and make shall be emphasised in the data exchange processes using both
decentralized decisions. Over the IoT, CPS communicate and hardware and software.
cooperate with each other and humans in real time. Data
Mining discover knowledge to support decision-making IV. FRAMEWORK OF A PREDICTIVE PROGRAM
process. Via the IoS, both internal and cross-organizational FOR ELEVATOR SERVICE
services are offered and utilized by participants of the value
chain. A. Predictive maintenance
Predictive maintenance is a set of activities that detect
C. Implementation principle of Industry 4.0
changes in the physical condition of equipment (signs of
Many scenarios of Industry 4.0 have been implemented and failure) in order to carry out the appropriate maintenance work
tested by research institutes and industries. Their results will for maximizing the service life of equipment without
affect the design of Industry 4.0. Referring to [12], seven increasing the risk of failure. It is classified into two kinds
design principles are driven for helping the companies who are according to the methods of detecting the signs of failure: (1)
interested in development of Industry 4.0 strategy. Statistical-based predictive maintenance and (2) Condition-
based predictive maintenance. Statistical-based predictive
a) Interoperability Maintenance (SBM) depends on statistical data from the
In Industry 4.0 companies, CPS and humans are connected meticulous recording of the stoppages of the in-plant items and
over the IoT and the IoS. Standards will be a key success factor components in order to develop models for predicting failures,
for communication between CPS of various manufacturers. while Condition-based predictive Maintenance depends on
continuous or periodic monitoring conditions of equipment to
b) Virtualization detect the signs of failure and make a maintenance decisions
Virtualization means that CPS are able to monitor physical using data mining technology.
processes. These sensor data are linked to virtual plant models
and simulation models. Thus, a virtual copy of the physical B. A IPdM (Intelligent Predicitve Mantenance) for Elevator
world is created. service
c) Decentralization Our intelligent predictive maintenance solution (Fig. 2)
The rising demand for individual products makes it opens up innovative new possibilities for companies. [13][14]
increasingly difficult to control systems centrally. Embedded Data generated by Cyber-physical systems (CPS) and
computers enable CPS to make decisions on their own. Only in transmitted by Internet if Things (IOT) monitoring
cases of failure, tasks are delegated to a higher level. machine/process condition is automatically reviewed to pick up

4
any patterns that indicate a possible fault through Data Mining way. It also means unplanned downtimes can be avoided and
systems. This decision use Internet of service (IOS) to allow both staff and resources can be employed more effectively.
the onset of a stoppage to be recognized early and corrective This innovative solution is called Industry 4.0 for Intelligent
measures to be planned and introduced in the most effective Predictive Maintenance (IPdM).
Key Performance Indicator (KPI) Fault Prognosis
CPS IOS Auto-regressive Moving Averaging
(ARMA)
KPI
Leading
Fuzzy Logic Prediction

Maintenance Maintenance
Actions and Maintenance Decision-
Instructions Management making and ANN Prediction

Scheduling
Match Matrix Prediction

KPI
Logging

Fault Diagnosis

Support Support Machine (SVM)

IOT
Elevator
Signal Pre-process Feature Extraction
Data Mining (Decision Tree &
Wireless Data Collection Association rules)
Networks
Denosing Time Domain
Artificial Neural Network (SOM &
SBP)
Compression Time-Frequency Domain
Data Acquisition
Statistical Maching
Frequency Domain (FFT, DFT)
Infrared thermography Extract Weak Signal

Acoustic Emission(AE)Sensors Wavelet Domain

Filter (WT, WPT) DM
Ultrasonic Sensors + AE
Principal Component Analysis
Amplification (PCA)
Vibration

Fig. 2. The framework of Intelligent Predictive Maintenance (IPdM) systems In Industry 4.0.

and feature extraction methods are shown in Table 1 and which

C. Some key techniques of IPdM systems features could be selected depend on the real machines or
IPdM systems based on many key techniques, such as CPS, system. All these kinds of methods are selectable in IFDPS and
IOT, IOS, Computational Intelligence (CI), Data Mining which methods are applied can be decided by real machine or
(knowledge discovery), Swarm Intelligence (SI), need to be system analysis.
researched and developed for fitting industry requirements.
There are six main modules in IPdM: (1) Sensor and data 3) Maintenance decision-making module.
acquisition (2) Signal pre-processing and feature extraction, (3) Maintenance decision-making module offers sufficient and
Maintenance decision-making, (4) key performance indicators, efficient information to maintenance personnel’ decision on
(5) maintenance scheduling optimization and (6) feedback taking maintenance actions. The models for decision-support
control and compensation. could be divided into 4 categories: (1) physical model; (2)
statistic model; (3) data-driven Model, and (4) hybrid model.
These modules has been developed at KDL laboratory, Because PdM strategy mostly depends on the signals and data
NTNU and they are described as the following: reflecting the condition of equipment, data-driven model will
1) Sensor and data acquisition module be in a dominate place. IPdM focus on data-driven and hybrid
It is the first step to implementing an IPdM strategy for model.
machinery diagnostics and prognostics. The task of this module IPdM focus on the data-driven techniques and hybrid
is selecting a suitable sensors and optimal sensor strategy. The techniques in maintenance decision-making module. If the
data acquisition process transforms the senor signals into historical data can be obtained easily, the data-driven is very
domains that are of the most information of the condition of the good to identify the fault and evaluate the condition. When
equipment. Various sensors, such as micro sensors, ultrasonic only part of historical can be obtained, the hybrid techniques
sensors, vibration sensors, acoustic emission sensors, have that combine the data-driven techniques and model-based
been designed to collect different data. techniques can be used to evaluate the condition of machine
effectively. The semi-supervised learning method also can be
2) Signal pre-processing and feature extraction module
used to evaluate condition and identify fault when only part of
Generally, there are two steps to deal with the signals from
historical data is available and it is very effective. All these
sensors. The one is signal processing, which enhances the
techniques are selectable according to the real manufacturing
signal characteristics and quality. The techniques in signal
system analysis. Techniques for maintenance decision-making
processing include filtering, amplification, data compression,
module can be divided into two main classes: diagnostics and
data validation, and de-noising that will improve the signal-to-
prognostics. Fault diagnostics focuses on detect, isolation and
noise ratio. The other is feature extraction, which extracts
identification of faults when they occur. While prognostics
features from processed signals that are characteristic of an
attempts to predict faults or failures before they occur. CI and
incipient failure or fault. Generally, the features can be
DM techniques have been increasingly applied to equipment
extracted from three domains: time domain, frequency domain
diagnosis and shown improved performance over traditional
and time-frequency domain. All possible signal pre-processing

5
approaches. However, in practice, it is not easy to apply CI quicker workflow optimization, leaner organizational structure,
techniques due to the lack of efficient procedures to obtain visibility, collaboration, and improved service margins.
training data and specific knowledge, which are required to
train the models. So far, most of applications in the literature The future work will be focus on the research and develop
just used experimental data for model training. The intelligent an efficient data mining and analysis model, which could
techniques used include Artificial Neural Networks, Fuzzy integrated model-based approaches with data driven methods.
logic Systems, fuzzy-neural networks, neural-fuzzy systems, We believe that industry 4.0 will give elevator industry and
evolutionary algorithms and Swarm intelligence. Compared to service great opportunities in the near future. The research goal
diagnostics, the no. of papers of prognostics is much smaller. of us is to develop real Elevator 4.0 for make elevator more
The most used prognostics is to how much time is left before a safe and robust.
failure occurs. The time left is usually called Remaining Useful References
Life (RUL). IPdM evaluate the remaining useful life using
[1] http://www.chinaexhibition.com/trade_events/6179-
data-driven model and try to find the relations between the The_4th_China_International_Elevator_Expo_2015.html
remaining useful life and the condition of machine or [2] http://www.aiweibang.com/yuedu/124765370.html]
component. [3] https://max.thyssenkrupp-elevator.com.
4) Key Performance Indicators (KPI) module [4] www.wipro.com
A diagram of KPI also called spider chart or health radar [5] K. Holmberg, A. Adgar, A. Arnaiz, E. A. M. J. Jantunen, and S.
chart is used for indicating the degradation of components. Mekid, E-maintenance, London: Springer London, (2010).
Each radio line shows the component condition of from zero [6] T. Lenahan,. Turnaround, shutdown and outage management:
effective planning and step-by-step execution of planned maintenance
(perfect) to one (damage). The colors shows the levels of the operations, Amsterdam: Elsevier, 2006.
components, such as safe, warning, alarm, fault and defect. The
[7] C. Moya, “The control of the setting up of a predictive maintenance
diagram will help operators or mangers to evaluate the programme using a system of indicators,” Omega, Vol. 32, No. 1, pp.
performance of the equipment visually. 57-75, 2004.
[8] H. P. Bloch and F. K. Geitner, Machinery failure analysis and
5) Maintenance schedule optimization module troubleshooting, Amsterdam: Elsevier/Butterworth-Heinemann,
Maintenance planning and scheduling optimization is a (2012).
kind of NP problem and the SI algorithms could be a very good [9] N. Ahamed, Y. Pandya, and A. Parey, “Spur gear tooth root crack
technique to solve this kind of problem. IPdM apply Genetic detection using time synchronous averaging under fluctuating speed,”
Algorithm (GA), Particle Swarm Optimization (PSO), Ant Measurement, Vol. 52, No. 0, pp. 1-11, 2014.
Colony Optimization (ACO) and Bee Colony Algorithm [10] S. K. Pinjala, L. Pintelona, and A. Vereecke, “An empirical
(BCA) and try to find the optimal dynamic predictive investigation on the relationship between business and maintenance
strategies,” Int. J. Production Economics, pp. 214-229, 2006.
maintenance scheduling. All these kind methods are selectable
in IPdM to solve maintenance scheduling optimization [11] R. F. Orsagh, H. Lee, M. Watson, C. S. Byington, J. Power, Advance
Vibration Monitoring for Wind turbine Health Management; Impact
problems. Technologies, LLC: Rochester, NY, USA, 2006.
6) Error correction, compensation and feedback control [12] K. Wang, “Intelligent predictive maintenance (IPdM) system –
Industry 4.0 scenario”, Advanced Manufacturing and Automation V,
module WIT transaction on Engineering Science Vol. 113, pp. 259-268, 2016.
This module will make error correction, compensation and [13] K. Wang, “Towards zero-defect manufacturing (ZDM) – a data
feedback control based on the results from the maintenance mining approach”, Journal of Advanced in Manufacturing, Vol. 1, pp.
decision-support module. 62-74, Spering-Verlag Berlin Heidelberg, 2013.
[14] K. Wang, “Intelligent and integrated RFID (II-RFID) system for
improving traceability in manufacturing,” Advances in
V. CONCLUSIONS Manufacturing;Vol. 2, pp.106-120, 2014.
This paper has proposed a general framework of intelligent [15] C. B. Zamfirescu, B. C. Pirvu, J. Schlick, and D. Zühlke, “Preliminary
predictive maintenance system for elevator service based on insides for an anthropocentric cyber-physical reference architecture of
the concepts of industry 4.0. The main technical components in the smart factory,” Studies in Informatics and Control, Vol. 22, No. 3,
2011.
the system are CPS, IOT&S, and Data Mining.
[16] M. Hermann, T. Pentek, and B. Otto, “Design principles for industry
The most difficult task in predictive maintenance is data 4.0 scenarios: A literature Review”, Working Paper, No. 01-2015,
Technical University of Dortmund, 2015.
analysis. Once advanced analytical models are applied to real-
time remote monitoring data, the integrated maintenance [17] https://smartmanufacturingcoalition.org/sites/default/files/implementin
g_21st_century_smart_manufacturing_report_2011_0.pdf
management system provides holistic management providing
[18] http://www.civitas.org.uk/pdf/BringingManufacturingBack
the organization with a variety of actionable predictions,
[19] http://news.xinhuanet.com/english/2015-03/25/c_134097374.htm
including failure predictions. The key to success is real-time
predictions allowing for real-time decision-making capabilities. [20] http://www.ims.org
[21] http://www.sfinorman.no
Real-time data driven predictive maintenance is a paradigm [22] S.-J. Shin, J.-S. Yoon and J.-M. Um, A new paradigm for product
shift from traditional reactive, preventive, to usage-based design and manufacturing via ubiquitous computing technology,
maintenance platforms, which are driven by fault resolution International Journal of Computer Integrated Manufacturing, Vol. 21,
practices versus fault avoidance practices. An intelligent No. 5, pp 540-549, 2008.
predictive maintenance solution well developed allows for