Content uploaded by Gökan MAY
Author content
All content in this area was uploaded by Gökan MAY on Sep 20, 2018
Content may be subject to copyright.
Predictive Maintenance Platform Based
on Integrated Strategies for Increased
Operating Life of Factories
Gökan May
1(&)
, Nikos Kyriakoulis
2
, Konstantinos Apostolou
3
,
Sangje Cho
1
, Konstantinos Grevenitis
3
, Stefanos Kokkorikos
2
,
Jovana Milenkovic
3
, and Dimitris Kiritsis
1
1
EPFL, ICT for Sustainable Manufacturing, EPFL SCI-STI-DK,
Station 9, 1015 Lausanne, Switzerland
gokan.may@epfl.ch
2
Core Innovation and Technology O.E, Athens, Greece
3
ATLANTIS Engineering S.A., Thessaloniki, Greece
Abstract. Process output and profitability of the operations are mainly deter-
mined by how the equipment is being used. The production planning, operations
and machine maintenance influence the overall equipment effectiveness
(OEE) of the machinery, resulting in more ‘good parts’at the end of the day.
The target of the predictive maintenance approaches in this respect is to increase
efficiency and effectiveness by optimizing the way machines are being used and
to decrease the costs of unplanned interventions for the customer. To this end,
development of ad-hoc strategies and their seamless integration into predic-
tive maintenance systems is envisaged to bring substantial advantages in terms
of productivity and competitiveness enhancement for manufacturing systems,
representing a leap towards the real implementation of the Industry 4.0 vision.
Inspired by this challenge, the study provides an approach to develop a novel
predictive maintenance platform capable of preventing unexpected-
breakdowns based on integrated strategies for extending the operating life
span of production systems. The approach and result in this article are based on
the development and implementation in a large collaborative EU-funded H2020
research project entitled Z-Bre4k, i.e. Strategies and predictive maintenance
models wrapped around physical systems for zero-unexpected-breakdowns and
increased operating life of factories.
Keywords: Industry 4.0 Predictive maintenance Big data
Asset management Smart factories Sustainable manufacturing
Industrial production
1 Introduction and State-of-the-Art
The requirement of competitiveness is a constant objective of manufacturers [1]. For a
successful shift toward Industry 4.0, companies must constantly innovate and imple-
ment new methods and approaches enabling them to reduce operating costs and
increase the availability and reliability of their production equipment [2]. For this
©IFIP International Federation for Information Processing 2018
Published by Springer Nature Switzerland AG 2018. All Rights Reserved
I. Moon et al. (Eds.): APMS 2018, IFIP AICT 536, pp. 279–287, 2018.
https://doi.org/10.1007/978-3-319-99707-0_35
reason, any downtime due to technical issues with the equipment needs to be avoided
and, if occurring, decreased as much as possible [3]. On that vein, maintenance in
general and predictive maintenance strategies in particular, are now facing significant
challenges to deal with the evolution of the equipment, instrumentation and manu-
facturing processes they should support. Preventive maintenance strategies designed
for traditional highly repetitive and stable mass production processes based on pre-
defined components and machine behaviour models are no longer valid and more
predictive-prescriptive maintenance strategies are needed. The success of those adap-
tive and responsive maintenance strategies highly depends on real-time and operation-
synchronous information from the production system, the production process and the
individual product, which should enrich and extend more traditional techniques and
models.
To meet the requirements mentioned above and aligned with the Industry 4.0 key
objectives toward eco-factories of the future [4,5], this study provides a holistic
framework and a comprehensive set of integrated strategies encompassing the whole
manufacturing line for addressing the issue of asset management in smart factories of
industry 4.0 in order to extend the life of production systems. Doing so, the research
aims at providing an answer as to what could be the proper strategies and associated
technologies to effectively minimize downtimes of manufacturing systems. A large
collaborative EU-funded H2020 research project entitled Z-Bre4k [6] has been the
main driver of the described approach and is designed for its validation. The project
consortium is formed by 17 organisations across Europe including industrial pilot
plants, academic institutions and technology providing companies.
To this end, novel strategies are designed in this research to be deployed at the field
in order to prevent/predict/diagnose/remediate failures, estimate remaining useful life
(RUL) of assets, manage alarms and mitigation actions, and synchronise with shop-
floor operations and plant management systems while ensuring the safety of workers.
The ultimate aim is to introduce and apply a holistic approach via integrated strategies
to increase maintainability, accurately predict the condition and the RUL of networked
machines, and adapt the performance to increase the operating life span of production
systems.
2 Strategies for Increased Operating Life of Production
Systems
The innovative synergies between online data gathering systems, real-time simulation
models, data-based models and the knowledge management system form the main
strategies which contribute to achieve zero breakdowns in manufacturing. In this
context, the proposed solution comprises the introduction of eight (8) scalable strate-
gies at component, machine and system level, all of which can be applied in the
existing manufacturing plants with minimum interventions, targeting (1) the prediction
occurrence of failure (Z-PREDICT), (2) the early detection of current or emerging
failure (Z-DIAGNOSE), (3) the prevention of failure occurrence, building up, or even
propagation in the production system (Z-PREVENT), (4) the estimation of the RUL of
assets (Z-ESTIMATE), (5) the management of the aforementioned strategies through
280 G. May et al.
event modelling, KPI monitoring and real-time decision support (Z-MANAGE), (6) the
replacement, reconfiguration, re-use, retirement, and recycling of components/assets
(Z-REMEDIATE), (7) synchronizing remedy actions, production planning and logis-
tics (Z-SYNCHRONISE), (8) preserving the safety, health, and comfort of the workers
(Z-SAFETY). Each of the developed strategies are triggered based on predicting,
detecting and assessing the impact of system level events that cause low performances,
generate failures, and increase the costs. Figure 1highlights the synergies and inter-
actions between the eight Z-Strategies for building a novel predictive maintenance
platform and the role of each strategy is further explained below.
Z-PREDICT: The events detected from the physical layer of the system are engi-
neered into high value data that stipulates new and more accurate process models. Such
an unbiased systems behaviour monitoring and analysis provides the basis for
enriching the existing knowledge of the system (experience) learning new patterns,
raising attention towards behaviour that cause operational and functional discrepancies
(e.g. alarms for predicted failures) and the general trends in the shop-floor. The more
the data pool is being increased the more precise (repeatability) and accurate the
predictions will be. The estimations for the future states involve the whole production
line –network of machines and components. The system can thus predict with high
confidence the expected performance of components and their maintenance needs,
predicting current or emerging failures, allowing better production planning and
decision making on their RUL. Hence, the ability to optimise the manufacturing pro-
cesses according to the RUL, production needs, and the maintenance operations is the
key innovation to fulfil the industrial requirements.
Z-PREVENT: The prevention of failure occurrence strategy is based on the prediction
strategy (i.e. degraded performance of assets or failure) realised across the shop-floor
for condition monitoring of machinery and respective produced quality. The Z-
PREDICT is predecessor of Z- PREVENT. The initial estimation of the future states is
based on the simulation and modelling of the parameters. For each predicted failure or
low performance (e.g. due to fatigue, wear), the responsible factors are identified and
Fig. 1. Synergies and interactions between the eight Z-Strategies
Predictive Maintenance Platform Based on Integrated Strategies 281
flagged through the FMEA system. The system analyses these factors based on an
initial estimation, which after the simulation these are updated recursively. The result of
this process is to avoid the building up or even propagation of a failure that leads to
breakdown based on each recorded event both from previous and current states. The
strategy thus prevents multiple alarm activations on similar failures.
Z-DIAGNOSE: This strategy is invoked when a current or an emerging failure is
detected considering the condition at all three levels –machine, product, shop-floor. In
such a scenario, an alarm is being triggered to flag the events that resulted in a failure or
system performance degradation. By mapping the true reasons, the system is then able
to avoid generating the failure or its emergence by weighting the system model. The
strategy also involves more actions and processes to deal both with the generation of
the diagnosed failure, and its severity increase to the next iterations as well as its impact
to the production line. Depending on the criticality of the generated failure, the system
can either adapt its parameters to prolong the RUL until the next maintenance, or plan
to the production for maintenance. The final decision on the actions is based on the Z-
MANAGE strategy.
Z-ESTIMATE: This strategy combines the information from the Z-DIAGNOSE and
Z-PREDICT estimating the RUL of the assets. The estimated values are also combined
with the information from the maintenance operations (physical examination from
operators) as well as from the specifications provided from the manufacturer. The latter
is used as the starting point for the estimation process, which after each iteration the
deviation of the real-model from the physical model is reduced having an accurate
virtual-model wrapped around the actual state of each machine and its components.
The trends for the fatigue and wear rates provide a confident RUL estimation.
Z-MANAGE: This strategy is executing the overall supervision and optimisation of
the system. The failures are processed with the Decision Support System (DSS) tools
and are interfaced with Manufacturing Execution Systems (MES). False positives and
false negatives are clustered within the Z-PREDICT and Z-PREVENT Strategies. To
achieve so, the previous acquired knowledge and incidents are also processed to fine
tune the system’s performance. Additionally, the production is optimised by better
scheduling (Z-SYNCHRONISE), taking into account the impact of each failure. The
optimised scheduling and adaptability of the manufacturing improves the overall
flexibility, placing a premium on the production systems, extending their operating life,
while preserve increased machinery availability.
Z-REMEDIATE: This strategy involves the decision making in the event of a failure,
which classifies and categorises the input in terms of criticality, type, etc. Based on the
component/assets types (repairable-non repairable) and their RUL the strategy decides
for the following: (1) replace, (2) reconfigure and/or re-use, (3) retire, and (4) recycle.
This strategy triggers the Z-SYNCHRONISE and Z-SAFETY strategies from which
the maintenance actions can be planned and organized.
Z-SYNCHRONISE: The predecessor Z-REMEDIATE strategy identifies the type of
action required for diagnosed failures which are then fused with the Z-MANAGE
output. This strategy synchronises all the remedy actions with internal and external
282 G. May et al.
supply-chain tiers, as well as with production planning and logistics. It is therefore
responsible to shift the production from one machine to another due to failure or
deteriorated condition/performance, acting as the “end-effector”thus leading to opti-
mised scheduling and reduced costs by carrying out maintenance activities on time.
Z-SAFETY: This strategy is invoked to increased Health & Safety during Z-Bre4k
shop-floor operations. Since most of the accidents occur during maintenance actions,
the Z-SAFETY prevents any activation to the machine that is under investigation or
repair. The “Safety-Mode”lifts any unauthorised control from the personnel for the
whole duration of the maintenance. Apart from reducing the accidents Z-SAFETY also
takes into account the comfort of the human personnel on the shop-floor, e.g. extreme
heat or noise may be tolerable for the machines but not for humans. Therefore, the
health & safety procedures are also taken into account towards the operation feedback
of the whole production line.
3 Predictive Maintenance Platform Based on Z-Strategies
Manufacturing enterprises are pushed to take local actions: thinking globally however
staying economically compatible within the local context. In order to achieve high pre-
cision manufacturing of complex products, there has to be a fundamental rethink on how
to improve the operation of machines and improved controls. The improvement should
not only concern the individual machines as isolated islands but encompass the totality of
production process as a system of interrelated elements that seek to maximise efficiency,
Fig. 2. Tools’landscape and logical architecture of the predictive maintenance platform
Predictive Maintenance Platform Based on Integrated Strategies 283
productivity, customer satisfaction; whilst at the same time eliminating waste and excess
inventory. For that purpose, aligned with the Z-Strategies, a set of technologies and
overall system architecture have been identified as a part of the proposed approach,
following the method and procedures developed and proposed by May et al. [7].
The first high-level description to lead to the definition of the predictive mainte-
nance platform consisted in identifying and classifying all components that can be
called as the tools’landscape and logical architecture, i.e. conceptual view. Figure 2
presents this landscape by proposing a compact representation of the involved tools.
Based on the proposed approach and defined conceptual view of the system, in Z-
Bre4k a novel predictive maintenance platform will be developed and demonstrated in
three pilot plans proving its universal applicability for the achievement of zero
breakdowns in manufacturing. Therefore, the predictive maintenance platform will:
•Introduce a novel design for predictive maintenance based on three levels: machine
(network of components), product, and shop-floor (network of machines). It will
reconfigure the system to increase its performance (shorter cycles), increase its
quality, and its availability by the employment of eight strategies to maximise these
factors.
•Make accurate predictions for the future states of the components/machines/systems
by the employment of intelligent and adaptive simulators forecasting the generation
of failures, the fatigue and wear levels, estimating the RUL triggering respective
remedy actions. The condition monitoring will provide data about the actual status
which will update the simulation results, increasing its accuracy. The DSS will
synchronise the plans for maintenance, production and logistics.
•Estimate the RUL through its simulation capabilities, calling for maintenance and
suggesting the optimal times to place orders for spare parts, reducing the related
costs. The increased predictability of the system and the failure prevention actions
will reduce the number of failures, maximise the performance, decrease the
repair/recover times reducing further the costs.
•Optimise the performance of the machines, based on the current and predicted
fatigue/wear levels allowing actions to maintain and increase the operating life of
these assets, as well as to reduce the unexpected failures and breakdowns.
Following the development of the conceptual view, the required components have
been highlighted in a preliminary architectural view, identifying services and depen-
dencies within the Z-Bre4k platform. Later, new components were added in order to
cover all the required functionalities of the resulting predictive maintenance platform.
As a result, definitions, identifications and classification of the system principle and its
process is presented in Fig. 3. Besides the defined overall architecture, Table 1presents
initial links on how these strategies are integrated within the overall architecture and
how they are associated with each component of Z-Bre4k platform.
Z-Bre4k components, their functionality and their interactions are thus described in
the overall architecture (Fig. 3). Initial data, generated by shop-floor assets (i.e. sensors,
cameras, industrial/IoT devices, etc.) is collected by Condition Monitoring, Cognitive
Embedded Condition Monitoring and Machine Simulators components. All data is sent
to the Industrial Data Spaces (IDS) reaching the overall Z-Bre4k platform after which
they are homogenized by the Semantic Framework and used by the components in a
284 G. May et al.
unified manner. Within the platform, users/employees are using HMI in order to add
data and/or parameterize components. Specifically, FMECA (Failure mode, effects, and
criticality analysis) component needs these inputs to calculate risks, Risk Priority
Numbers (RPNs), criticality matrices and alerts in order to send data to DSS, that
further generates and delivers strategies, recommendations, notifications, reports and
updated schedules. Finally, M3 Gage, M3 Software and VRfx components are related
to XYZ cloud points, 3D representations and visualization data of physical objects.
The resulting Z-Bre4k system will be demonstrated in three key sectors with the
strongest SME presence (i.e. automotive, food and beverage, consumer electronics) for
Fig. 3. Overall architecture of the predictive maintenance platform
Table 1. Z-Strategy and component association
No. Z-Strategy Components
1 Z-PREDICT Condition monitoring, Machine simulators, VRfx, Predictive
Maintenance
2 Z-PREVENT Machine simulators, Predictive Maintenance, FMECA
3 Z-DIAGNOSE Predictive Maintenance, FMECA
4 Z-ESTIMATE Machine simulators, VRfx, Predictive Maintenance
5 Z-MANAGE Predictive Maintenance, FMECA, DSS
6 Z-REMEDIATE FMECA, DSS
7 Z-SYNCHRONIZE AUTOWARE Communication Middleware
8 Z-SAFETY DSS
Predictive Maintenance Platform Based on Integrated Strategies 285
a wide range of components and machines with different operational requirements and
behaviours, illustrating the potential and full value of Z-Bre4k as a holistic framework
to address predictive maintenance strategies for operation in high diversity of
machinery (e.g. robotic systems, inline quality control equipment, injection moulding,
stamping press, high performance smart tooling/dies and fixtures), including highly
challenging and sometimes critical manufacturing processes (e.g. automated packaging
industry, multi-stage zero-defect adaptive manufacturing of structural light-weight
component for automotive industry, short-batch mass customised production process
for consumer electronics and health sector).
4 Discussion and Concluding Remarks
The main goal of the predictive maintenance approach and implementation is to pro-
vide machine builders and designers (OEMs), industrial component suppliers and
engineering software developers with novel solutions which will: (1) improve the
performance of the manufacturing processes; (2) increase the machine maintainability;
(3) provide predictions on damages/failures; (4) understand and interpret the source of
the failures enhancing eventually the design process; (5) increase the availability of the
machine builders while making them cost effective; and thus (6) increase OEE.
In this context, the study will: (a) act as a comprehensive and practical guide for
optimizing production machineries and processes by implementing predictive mainte-
nance principles; (b) transform machine tool and process related data into useful infor-
mation that could support machinery prognosis and optimization strategies by enabling
model-based control of machine tools based on actual machine life-cycle parameters.
Accordingly, the expected impacts are highlighted as follows: (1) improved pre-
dictive maintenance and system adaptability for manufacturing systems and processes;
(2) new maintainability concepts based on predictive maintenance with improved
machine reliability (MTBF) and reduced maintenance costs; and (3) incorporating
intelligent systems and data analysis methods for achieving smart factories of
Industry 4.0.
Future work will focus on implementing and validating the proposed approach on
several use cases in different industries, demonstrating its ability to support major
actors of the manufacturing sector to take advantage of the digital transformation.
Acknowledgements. This work has been carried out in the framework of Z-Bre4k Project,
which has received funding from the European Union’s Horizon 2020 research and innovation
programme under grant agreement Nº768869.
References
1. Swanson, L.: Linking maintenance strategies to performance. Int. J. Prod. Econ. 70(3), 237–
244 (2001)
2. Wang, S., Wan, J., Li, D., Zhang, C.: Implementing smart factory of industrie 4.0: an outlook.
Int. J. Distrib. Sens. Netw. 12(1), 3159805 (2016)
286 G. May et al.
3. Lee, J., Kao, H.-A., Yang, S.: Service innovation and smart analytics for industry 4.0 and big
data environment. Procedia CIRP 16,3–8 (2014)
4. Lee, J., Bagheri, B., Kao, H.-A.: A cyber-physical systems architecture for Industry 4.0-based
manufacturing systems. Manuf. Lett. 3,18–23 (2015)
5. May, G., Stahl, B., Taisch, M.: Energy management in manufacturing: toward eco-factories of
the future –a focus group study. Appl. Energy 164, 628–638 (2016)
6. Z-Bre4 k Project. https://www.z-bre4k.eu. Accessed 21 Mar 2018
7. May, G., Ioannidis, D., Metaxa, I.N., Tzovaras, D., Kiritsis, D.: An approach to development
of system architecture in large collaborative projects. In: Lödding, H., Riedel, R., Thoben, K.-
D., von Cieminski, G., Kiritsis, D. (eds.) APMS 2017. IAICT, vol. 513, pp. 67–75. Springer,
Cham (2017). https://doi.org/10.1007/978-3-319-66923-6_8
Predictive Maintenance Platform Based on Integrated Strategies 287
