International Workshop of Advanced Manufacturing and Automation (IWAMA 2016)

Industry 4.0 – Potentials for Predictive Maintenance

Zhe Li and Kesheng Wang Yafei He

Department of Production and Quality Engineering College of Engineering
Norwegian University of Science and Technology Shanghai Polytechnic University
Trondheim, Norway Shanghai, P.R. China
{zhe.li & kesheng.wang}@ntnu.no [email protected]

Abstract— Industry 4,0 represents the coming fourth Therefore, this paper aims to provide empirical information
industrial revolution on the way to combine modern industries on the potentials of Industry 4.0 in the research field of
with Cyber-Physical Systems, Internet of Things and Internet of maintenance. It may help academics and practitioners to
Services. In an Industry 4.0 factory, machines are connected as a identify and prioritize their steps towards predictive
collaborative community to collect, exchange and analyse data maintenance and condition-based maintenance management
systematically. This paper investigates the potentials and trends under the environment of Industry 4.0.
of predictive maintenance and maintenance management in
industrial big data and Cyber-Physical Systems environment. The remaining part of the paper is organized as follows.
Furthermore, the development of predictive maintenance, its Section 2 briefly presents the classification of maintenance
technical challenges, and the potentials under Industry 4.0 era strategy and related challenges. Section 3 discusses the related
was researched to discover the linkage between Industry 4.0 and trends accelerated and encouraged under Industry 4.0 era.
predictive maintenance. Section 4 describes a case study, which illustrates how the
predictive maintenance and maintenance management can be
Keywords— Industry 4.0; industrial big data; predictive applied under Industry 4.0 era. Conclusions are summarized in
maintenance; maintenance management the last section of this paper.

I. INTRODUCTION II. REVIEW OF MAINTENANCE STRATEGIES

The goal of predictive maintenance is to reduce downtime A maintenance action may include a set of maintenance
and cost of maintenance under the premise of zero failure activities: monitoring, condition analysis, routine maintenance,
manufacturing through monitoring the working condition of overhaul, repair and rebuilding. According to the European
equipment and predicting when equipment failure might occur. Standard, maintenance is defined as the combination of all
Prediction for future potential fault allows maintenance to be technical, administrative and managerial actions during the life
planned before the fault happens. Ideally, maintenance cycle of an item intended to retain it in, or restore it to, a state
schedule can be optimized to minimize the cost of maintenance in which it can perform the required function [5]. Progress in
and achieve zero failure manufacturing. However, it is difficult maintenance strategy benefits from a long historical
to realize all the advantages of predictive maintenance without development
the foundation of correlation techniques like big data and
cloud-computing. Many manufacturing systems are not ready As shown in Fig. 1, the classification of maintenance
to manage big data due to the high demands on data access and strategies can be divided into three classes as corrective
data quality and use multiple data sources to extract relevant maintenance (CM), preventive maintenance (PM), and
information [1]. predictive maintenance (PdM). CM is similar to repair work,
which is performed when the equipment has an obvious failure
Industry 4.0 refers to a fourth generation of industrial or a breakdown. That is why CM is also called as run-to failure
activity as a result of the fourth industrial revolution maintenance, breakdown maintenance or maintenance-on-
characterized by smart systems and Internet-based solutions failure. PM is carried out at predetermined intervals or
[2]. It combines strengths of optimized industrial according to pre-described criteria with the intention to
manufacturing with internet technologies and changes decrease the probability of failure or degradation of certain
manufacturing process, maintenance strategies and functions. PM is scheduled without the occurrence of any
maintenance management significantly. Therefore, many monitoring activities. The scheduling may be based on the
companies face the challenge to assess the diversity of number of working hours or times. These two strategies are
developments and concepts summarized the term Industry 4.0 very well known in most industries [6].
and to develop their own strategies [3]. However, since lack of
research of the potential use of Industry 4.0 and prospect of
predictive maintenance, many companies and organizations are
exposed to a dilemma, neither to wait too long with their
Industry 4.0 implementation nor to start too early and commit
fatal errors [4].

© 2016. The authors - Published by Atlantis Press 42

 interact with humans through many new modalities. The key is
Maintenance Strategies
the ability to interact with, and expand the capabilities of, the
physical world through computation, communication, and
control [11]. IoT is defined as the ubiquitous access to entities
Corrective Maintenance Preventive Maintenance Predictive Maintenance on the internet for the extension of the physical world through
a variety of sensing, detection, identification, location tracking
and monitoring equipment [12]. IoS pursuits a similar approach
Reliability- Time- Statistical- Condition- with services instead of physical entities. The integration of
Deferred Immediate
based based based based
these developments promotes the cooperation between the
partners alone the entire system. From the perspective of
Opportunity Design-Out predictive maintenance, Industry 4.0 accelerates and
encourages several developments, which will be reviewed and
Fig. 1 Classification of maintenance strategies [6]. discussed in this section.
As one of the most popular and modern maintenance
policies, PdM measures parameters in the condition of A. Cloud computing environment
equipment carry out the appropriate tasks to optimize the Industry 4.0 includes the increasing impact of information
service life of machines and processes without increasing the and communication technologies on industrial production
risk of failure [7]. Based on the approaches of measuring the processes. As one of the driving forces behind Industry 4.0,
symptom of failures, there are two groups of PdM: statistical- cloud computing has rapidly emerged as an accepted
based PdM and condition-based PdM. computing paradigm in many enterprises worldwide due to its
flexibility and many other advantages [13]. It can manage
Comparing with other maintenance strategies, PdM has shared data from multiple sources efficiently and flexibly in a
some advantages: equipment that requires maintenance is shut self-service way and provide a unified service delivery
down only before imminent failure; reducing the total time platform for the IoT applications [14]. A cloud-based system
spent maintaining equipment; reducing maintenance costs by provides the technological basis for the provision of data and
avoiding catastrophic damage; increasing availability and allows not only creation of community-type services but also
reliability of machines; extending life of equipment and building of an open service platform environment which may
processes. However, PdM is facing several challenges as have features of interactive, collaborative and customizable on
following: demand. From the view point of predictive maintenance, cloud
(1) High demands on data access, data quality and data computing environment can efficiently support various smart
fusion from multiple sources for data sharing and data services and solve several issues such as the memory capacity
publication. Since these sources of data often operate in a of equipment, computing power of processor, data security and
heterogeneous environment, integration between the systems is data fusion from multiple sources.
problematic [8].
B. Industrial big data environment
(2) The capability to deal with industrial big data. To
leverage big data, industrial businesses need the ability to Under the Industry 4.0 era, another significant development
support different types of information, the infrastructure to is the combination of intelligent analytics and Cyber-Physical
store massive data sets, and the flexibility to leverage the systems for achieving a new type of manufacturing
information once it is collected and stored—enabling historical management and factory transformation. The trend behand this
analysis of critical trends to enable real-time predictive analysis combination is the environment of industrial big data. The
[9]. application of industrial big data has been demonstrated in
different areas of manufacturing, including production, supply
(3) The prediction accuracy for PdM. The inaccurate chain, maintenance and diagnosis, quality management, and
predictive information may result in either unnecessary energy [15].
maintenance, such as early replacement of components, or
production downtime because of unexpected machine failures. In the research fields of maintenance, fault diagnosis and
Therefore, the accuracy of remaining useful life prediction, prognosis, appropriate sensor installations and various signals
particularly the long-term prediction, which gives sufficient can be applied to monitor the working condition of the
time to prepare for a maintenance operation, plays an essential equipment. In addition, through comparing the current and
role in the full realization of the potentials of predictive historical data with data mining technologies, the degree of
maintenance [10]. potential faults, the reliability of certain components and other
useful information may be harvested.
III. FROM THE PERSPECTIVE OF MAINTENANCE
C. Smart Factory
TOWARDS INDUSTRY 4.0
Industry 4.0 also facilitate the vision and execution of the
Industry 4.0 is the superposition of several technological idea "Smart Factory", in which cyber-physical systems monitor
developments related to Cyber-Physical Systems (CPS), physical processes, create the virtual copy to represent the
Internet of Things (IoT), Internet of Services (IoS) and Data physical world and make decentralized decisions. Products
Mining (DM). CPS refers to a new generation of systems with may not only provide their identity but also record their
integrated computational and physical capabilities that can properties, history and status via Radio Frequency

43
Identification (RFID) technology. Over cloud computing and faults. Fig.2 shows the overall structure of the system, the data
distributed control, cyber-physical systems communicate and is acquired from remote customer site and transported to data
cooperate with each and via the Internet of Services, both mining center for data analysis. Data includes two types: (1)
internal and cross-organizational services are offered and the signals collected from the sensors, which can be used to
utilized to make machines self-aware and actively prevents evaluate the working condition of the equipment or quality of
potential performance issues [16]. the products. (2) the logistics information in the RFID tags,
which can be collected by local RFID readers to identify the
From the perspective of maintenance, a self-aware and self- components in the equipment.
maintained machine system can be considered as a system
which can self-assess its own health and degradation, and All the data will be uploaded on the information network,
further use similar information from other peers for smart which is controlled by ICB, a software company, for data
maintenance decisions to avoid potential faults [1]. To achieve sharing and publication. The system flowchart is shown in Fig.
such intelligence, smart analytics may be used at the individual 3 [17]. After a series of data mining process, including Signal
machine or fleet levels. For a mechanical system, self-aware Pre-process and Feature Extraction, the result of fault diagnosis
means the capability to assess the current, past or future and prognosis can form the Key Performance Indicator (KPI)
working condition of a machine, and output the evaluation to evaluate the current working condition and predict the
result. Such health assessment can be performed through data degree of potential faults.
mining technologies to analyse the information collected from
the given machine and its ambient environment. The result of data mining will be recorded and published
back to the network for remote condition monitoring and
information sharing. The recommended maintenance strategy
IV. CASE STUDY: REMOTE PREDICTIVE MAINTENANCE SYSTEMS will be formed in the decision support system according to the
In this section, a case study about the application of prediction or evaluation result. Normally, it requires some
predictive maintenance will be introduced. This case study relevant optimization and association theories to capture the
comes from a project called Green Monitoring, in which the trade-off between several factors such as maintenance cost,
remote prognostics and monitoring system focused on machining accuracy, or defective percentage, and then acquire
evaluating and predicting the health of equipment to reduce the the optimal decision. In addition, local RFID readers can
maintenance cost on faults, defects and maintenance during the collect and identify all the RFID tags shown in Fig. 4, which
manufacturing process. In addition, RFID technology is represent the components in the machine.
applied to identify the components, which may have potential

Fig. 2 The Structure of remote predictive maintenance system

44
 Key Performance Indicator (KPI) Fault Prognosis
CPS Maintenance IOS Auto-regressive Moving Averaging
Management (ARMA)
KPI
Leading
Fuzzy Logic Prediction

Maintenance Maintenance
Actions and Decision-
ANN Prediction
Instructions Logistics making and Degradation Prediction
Information sheduling
Match Matrix Prediction

KPI
Logging

Fault Diagnosis

Support Support Machine (SVM)

IOT
Signal Pre-process Feature Extraction
Wireless Data Collection Data Mining (Decision Tree &
Equipments Association rules)
RFID Reader 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. 3 System flowchart.

Fig. 5 shows a sample of Information in a RFID Tag. Here,

Tag ID “0xE20093747411026111005007” can be used for
components identity. Normally, the 512 bits user data are
reserved for the customized information, which can be
converted to 64 ASCII characters. In this case, the condition
evaluation related information is written in the user memory of
the RFID tag. The information comprises the current working
condition, whether has potential faults in the prediction period,
when the potential faults may happen, evaluation time and
date, in the form of “COND=0 POT=1 FAULT=28.05.16
TIME=14:35 DATE=21.05.16”. Since the tag memory is
written in the form of hexadecimal, the texts have to be
converted to HEX before writing. In this case, the current
working condition of the component, which is a spindle in a
machine center, is normal (COND=0). Potential faults are
detected (POT=1), which may happen in 28.05.2016
(FAULT=28.05.16). In addition, the detection is taken at 14:35
on 21.05.2016 (TIME=14:35 DATE=21.05.16).
Fig. 5 Information in a RFID Tag.

Combining the result of diagnosis and prognosis with

logistics information collected from RFID tags, customers can
figure out when, where, which equipment, and which
components may have faults.

V. CONCLUSIONS
Industry 4.0 promotes the predictive and smart
manufacturing in the future industry. In an industry 4,0 factory,
the machines are connected as a collaborative community,
which generates a great potential for predictive maintenance.
This article has investigated the development of predictive
maintenance, its technical challenges, and the potential under
the environment of Industry 4.0. On one hand, the environment
of cloud computing, industrial big data and smart factory is
accelerated and encouraged in Industry 4.0 era, which establish
the foundation to achieve predictive maintenance. On the other
hand, predictive maintenance will play a very important role in
future maintenance activities and meet the requirement of
Fig. 4 Record of Tags in RFID Readers

45
smart manufacturing, self-aware machine under the terms of [7] M. C. Garcia, M. A. Sanz-Bobi, and J. del Pico, "SIMAP: Intelligent
Industry 4.0. System for Predictive Maintenance: Application to the health condition
monitoring of a windturbine gearbox," Computers in Industry, vol. 57,
The case study introduced in this paper illustrates how the pp. 552-568, 2006.
predictive maintenance and maintenance management can be [8] M. Aljumaili, K. Wandt, R. Karim, and P. Tretten, "eMaintenance
applied under the environment of cloud computing and ontologies for data quality support," Journal of Quality in Maintenance
Engineering, vol. 21, pp. 358-374, 2015.
industrial big data based on data mining and RFID techniques.
[9] The Rise of Industrial Big Data. Available:
http://leadwise.mediadroit.com/files/19174the_rise_of_industrial_big_d
ACKNOWLEDGMENT ata_wp_gft834.pdf
[10] J. Liu, D. Djurdjanovic, J. Ni, N. Casoetto, and J. Lee, "Similarity based
This study has been supported by a grant from Norway method for manufacturing process performance prediction and
through the Norwegian Financial Mechanism 2009-2014, in diagnosis," Computers in Industry, vol. 58, pp. 558-566, 8// 2007.
the frame of the Green Industry Innovation Programme [11] R. Baheti and H. Gill, "Cyber-physical systems," The impact of control
Bulgaria. Moreover, the authors gratefully thank InterConsult technology, vol. 12, pp. 161-166, 2011.
Bulgaria (ICB) and Kongsberg Terotech (KTT) for their [12] L. W. Chaves and Z. Nochta, "Breakthrough towards the internet of
assistance in the software and data sources used in this study. things," in Unique Radio Innovation for the 21st Century, ed: Springer,
2011, pp. 25-38.
[13] J. Bughin, M. Chui, and J. Manyika, "Clouds, big data, and smart assets:
REFERENCES Ten tech-enabled business trends to watch," McKinsey Quarterly, vol.
[1] J. Lee, H.-A. Kao, and S. Yang, "Service innovation and smart analytics 56, pp. 75-86, 2010.
for industry 4.0 and big data environment," Procedia CIRP, vol. 16, pp. [14] L. Li, L. Xinrui, and L. Xinyu, "Cloud-Based Service Composition
3-8, 2014. Architecture for Internet of Things," Communications in Computer and
[2] H. Lasi, P. Fettke, H.-G. Kemper, T. Feld, and M. Hoffmann, "Industry Information Science, Springer, vol. 312, pp. 559-564, 2012.
4.0," Business & Information Systems Engineering, vol. 6, p. 239, 2014. [15] P. O’Donovan, K. Leahy, K. Bruton, and D. T. O’Sullivan, "Big data in
[3] A. Hochschild, "EMPIRE OF COTTON A Global History," ed: NEW manufacturing: a systematic mapping study," Journal of Big Data, vol.
YORK TIMES 620 8TH AVE, NEW YORK, NY 10018 USA, 2015. 2, pp. 1-22, 2015.
[4] R. Schmidt, M. Möhring, R.-C. Härting, C. Reichstein, P. Neumaier, and [16] M. Hermann, T. Pentek, and B. Otto, "Design principles for Industrie
P. Jozinović, "Industry 4.0-Potentials for Creating Smart Products: 4.0 scenarios: a literature review," Technische Universität Dortmund,
Empirical Research Results," in Business Information Systems, 2015, Dortmund, 2015.
pp. 16-27. [17] K. Wang, "Intelligent Predictive maintenance (IPdM) system – Industry
[5] E. Standard, "Maintenance terminology," in European Committee for 4.0 scenario", editors: K. Wang, Y. Wang, J. O. Strandhagen and T. Yu,
Standardization 13306, ed. Brussels, 2001. Proceedings of Advanced Manufacturing and Automation V, WIT
[6] K.-S. Wang, Z. Li, J. Braaten, and Q. Yu, "Interpretation and Transaction on Engineering Science, Vol 113, pp. 259-268, 2016
compensation of backlash error data in machine centers for intelligent
predictive maintenance using ANNs," Advances in Manufacturing, vol.
3, pp. 97-104, 2015.

46