A state of the art of predictive maintenance techniques

Abstract

Nowadays in industry there is a great and increasing demand in resource management, taking into consideration the ever-growing complexity of technical systems. The concept of maintenance is one the most important topics of product development today. As the factories and the industry evolves, the need of proper maintenance plays a major factor in cost and efficiency optimization. In this paper a state of the art of maintenance techniques is presented, predictive maintenance being one of the biggest topics going forward. Predictive maintenance techniques are discussed and presented in detail creating the necessary links with nowadays industry advances: Industry 4.0.

Full text

IOP Conference Series: Materials Science and Engineering
PAPER • OPEN ACCESS
A state of the art of predictive maintenance techniques
To cite this article: P Coand et al 2020 IOP Conf. Ser.: Mater. Sci. Eng. 997 012039
View the article online for updates and enhancements.
This content was downloaded from IP address 216.74.76.225 on 31/12/2020 at 06:51

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd
ACME 2020
IOP Conf. Series: Materials Science and Engineering 997 (2020) 012039
IOP Publishing
doi:10.1088/1757-899X/997/1/012039
1
A state of the art of predictive maintenance techniques
P Coandă1, M Avram1 and V Constantin1
1Mechanical Engineering, Mechatronics and Precision Mechanics Department, University
“Politehnica” of Bucharest, Bucharest, Romania
E-mail: philipcoanda@gmail.com
Abstract. Nowadays in industry there is a great and increasing demand in resource
management, taking into consideration the ever-growing complexity of technical systems. The
concept of maintenance is one the most important topics of product development today. As the
factories and the industry evolves, the need of proper maintenance plays a major factor in cost
and efficiency optimization. In this paper a state of the art of maintenance techniques is
presented, predictive maintenance being one of the biggest topics going forward. Predictive
maintenance techniques are discussed and presented in detail creating the necessary links with
nowadays industry advances: Industry 4.0.
1. Introduction
The present paper aims to deal with the main maintenance concepts, focusing on the principle of
predictive maintenance.
The industrial revolution represents one of the most important moments in human history and
continues to play an important role in the modern world today. The main areas targeted by the
industrial revolution were the technological, socio-economic and cultural ones. Through the changes,
it has shaped the modern world in the way it is today, changing the way people live, how they spend
their free time and not least how the current political class is organized globally.
Some of the most important technological changes targeted by the industrial revolution are:
• Use of new materials, mainly iron and steel;
• Use of new energy sources, such as coal, steam engine, electricity, oil and internal combustion
engines;
• Inventing new machines for the textile industry (eg torsion wheel);
• Organizing the production in the factory type system that we meet today and the division of
labour;
• Development of transport and means of communication such as: steam locomotives, steam
vessels, airplanes, telegraph, radio;
• Application of scientific principles in industrialization.
All the technological changes have greatly contributed to the use of natural resources and have
enabled the mass production of goods [1]. Today, maintenance is an issue of the utmost importance in
the industrial environment and beyond. Maintenance performed correctly, regardless of its type, does
nothing but improve the working environment and implicitly increase the performances, having a
general effect. In the following chapters, maintenance concepts will be dealt with extensively, as well
as topical solutions in this field.

ACME 2020
IOP Conf. Series: Materials Science and Engineering 997 (2020) 012039
IOP Publishing
doi:10.1088/1757-899X/997/1/012039
2
2. The maintenance concept
When we talk about products, they represent the centre of interest in companies and factories around
the world, therefore a product management plan is needed. It ensures good development and
organization of all stages, resources and means involved in the production process.
The idea of maintenance is present from the design stage of a product. It must ensure the proper
functioning of the product according to its life. As can be seen in figure 1, according to [2], there are
five stages in the life cycle of a product: idea, definition, realization, use, recycling. These five phases
can be divided into 3 broad categories:
1. The incipient phase: idea, definition, realization;
2. Middle phase: use;
3. Final phase: recycling.
Figure 1. The lifecycle of a product.
As time goes on, the medium phase lasts the longest, the product being in the use phase, so it is
necessary to maintain it in order to function in optimal parameters.
The maintenance concept is defined by the standard EN 13306, as follows [3]: the combination of
all technical, administrative and managerial actions performed during the life cycle of an element to
maintain or restore the conditions under which it can perform the required function. In order to
perform the maintenance operation, a maintenance management plan is required. Maintenance
management is defined as the sum of all the management activities that determine the maintenance
objectives, strategies and responsibilities and their implementation by means of maintenance planning,
maintenance control and improvement of maintenance activities.
There are therefore several types of maintenance, also defined in standard EN 13306. Figure 2 shows
the schematic representation of the maintenance types.
Figure 2. Types of maintenance according to EN 13306 standard.
Equipment maintenance has evolved over time, from maintenance that was performed only when
the elements suffered damage to modern methods, proactive and predictive maintenance, the latter
being the most popular today [4]. Maintenance performed only in case of damage has proven to be an
inefficient method in time because in many cases alternative failures are observed in the system. This
has led to the adoption of the principle of scheduled maintenance where the equipment is subject to a
total overhaul whether it has problems.

ACME 2020
IOP Conf. Series: Materials Science and Engineering 997 (2020) 012039
IOP Publishing
doi:10.1088/1757-899X/997/1/012039
3
The disadvantage of the scheduled maintenance method is that it analyses the service life of each
element in the system, so some elements are changed even if they have not reached the end of their life
cycle. This has led to the evolution towards predictive maintenance.
Figure 2 shows a generalized classification of maintenance types. According to [5], maintenance tasks
can be classified into two broad categories:
1. Reactive maintenance - the system is used until breakdowns occur. When a fault occurs, the
damaged elements are replaced in order to restore the operating capacity of the system;
2. Proactive maintenance - maintenance actions are planned or take place as a result of tracking
indicators. Planning can be done using sensors that announce the early stages of a fault. This
type of maintenance aims to maintain the functionality of the system;
Regardless of the type of maintenance applied, the purpose of these activities is to reduce the stops
caused by damage as much as possible. To be able to minimize the stops, it is necessary to understand
the failure modes and mechanisms. Table 1 presents a comparative analysis of advantages and
disadvantages for reactive and proactive maintenance.
Table 1. Reactive and proactive maintenance advantages and disadvantages.
Advantages
Disadvantages
Reactive
maintenance
Low initial costs
Easy to implement
Unscheduled stops
High associated costs (work over schedule,
delivery urgency, manufacturing urgency)
Poorly optimized resources
Proactive
maintenance
Increased system reliability
Minimizing logistical stops
Reduction of non-scheduled
stops
Lowers costs (optimizing
parts, optimizing work)
Maintenance planning
Optimization of logistic
support
High initial set-up costs
The cost reduction is not immediate
Not reliable for all equipment
3. Predictive maintenance nowadays
Predictive maintenance is a relatively new concept that can be framed as a proactive maintenance
approach. When speaking about predictive maintenance techniques, conditional predictive
maintenance is one of the most used techniques. Conditional predictive maintenance is a strategy
where maintenance is performed by observing certain parameters or certain components of the system.
The advantage of this method is that the state of the system is presented in real time, based on the
parameters followed. Although the systems generally have a well-defined operating curve by the
manufacturer, depending on the operating conditions and the working environment it may undergo
changes, as a result, damage may occur earlier. At the same time, this method increases the ease with
which the faults are detected and especially, their place.
Regarding conditional predictive maintenance, there is a wide range of parameters that can be
monitored for predictive maintenance, some of the most important being:
• Vibration analysis - represents the most efficient method for detecting problems in the
equipment that perform rotational movements;
• Acoustic analysis - this can detect or monitor cracks in pipes or pipes;
• Lubrication oils analysis - the particles found in the oils used to determine the degree of wear
of the components are analysed;

ACME 2020
IOP Conf. Series: Materials Science and Engineering 997 (2020) 012039
IOP Publishing
doi:10.1088/1757-899X/997/1/012039
4
• Particle analysis in the working environment - a method generally used in equipment working
in a fluid environment;
• Corrosive analysis - ultrasound measurements are performed to determine the corrosion in
different structures;
• Thermal analysis - used especially in the case of mechanical and electrical systems to detect
overheating in general.
• Performance analysis - an efficient technique to determine operational problems in the system;
Using one or more of the above methods, maintenance solutions for equipment and systems are
developed. Therefore, considering that more than 30% of maintenance costs are caused by faulty
maintenance planning [6] leading to added costs in the production process, predictive maintenance is a
method of increased interest especially when considering industrial environments, where stops can
cause large losses.
A new approach to in the predictive maintenance techniques is represented by the usage of 4.0
Industry standards in order to facilitate the processes contained by the predictive maintenance
technique. As nowadays the industry is facing difficulties due to the development of technology, the
diminution of natural resources, the incidence of wars and natural disasters and at the same time,
social issues such as globalization and increasing the retirement age for the labor force, represents
difficulties that produce substantial economic effects [7]. In addition to the points listed above, it
should be noted that consumers today want a variety and high quality of products, as well as quality
services during use.
The vision of Industry 4.0 comes with a massive impact on the current way of working in the
industry characterized by a new model of socio-technological interaction. These new smart factories
have small, decentralized networks that act autonomously being able to make decisions in different
situations and which are interconnected globally with respect to the company, resulting in intelligent
products that can be tracked in real time and of whose properties and characteristics are known
throughout the production process. The benefits of Industry 4.0 can therefore be used to develop a
predictive maintenance strategy using the aforementioned technologies. They provide the premises for
the implementation of predictive maintenance strategies, this being achievable with reduced material
and human resources.
A brief definition of Industry 4.0 can be considered the one provided by [8] which says that
Industry 4.0 represents basically a synergy between Internet of Things, Internet of Services and of
course, the industrial process. It should be also noted that Industry 4.0 is shaping not only the
manufacturing and maintenance processes, but also the way human-machine interaction take place.
The benefits of predictive maintenance in comparison with scheduled maintenance are srtrongly
highlighted in paper [9]. In comparison with classical maintenance approaches with are at least in the
industrial field the main maintenance techniques applied, predictive maintenance promises to lower
the increasing failures caused by the incapacity to provide a time estimation for components failure.
Regarding predictive maintenance, it can be split in three basic techniques:
• Maintenance based on existing process sensors, which are already equipped;
• Maintenance based on test sensors;
• Maintenance based on test signal technique;
All of the above maintenance techniques aim to provide an improvement in failure analysis, yet they
represent different approaches. The first two techniques can be considered as “passive” and the third
as “active”, the implementation considering loop responses and real time testing.
Considering that the vast majority of industrial processes are conducted using electrical and
mechanical systems, paper [10] proposes an overview over predictive maintenance techniques used for
performance enhancements. The main topics considered are electrical motors performance and bearing
failures, these being the most common problems. The study concentrates on two case studies carried
on a 75-hp blower motor and a 200-hp air compressor motor in an active plant. Results and
conclusions were gathered from both experiments, confirming a misalignment in blower motor shaft
and a performance loss and a degradation for the compressor motor due to small periods of overload.

ACME 2020
IOP Conf. Series: Materials Science and Engineering 997 (2020) 012039
IOP Publishing
doi:10.1088/1757-899X/997/1/012039
5
Grall, Antoine, et al. [11] propose a method of predictive maintenance used for systems that are
subject to continuous deterioration. The paper aims to treat two main points, as follows:
• Creating a structure for implementing conditional predictive maintenance and determining a
mathematical model of cost analysis from the way the faults appear;
• Demonstrate that long-term maintenance costs can be substantially reduced by using such an
approach.
A random model is used to simulate the continuous deterioration of the system. It can simulate an
aging in the case of mechanical models, the evolution of defective products in the case of a production
line, the level of corrosion / erosion in the case of structures. The chosen model offers a cost efficiency
considering the time required to perform maintenance for an indefinite period. It aims to make a
minimum estimation of the costs considering the level of wear and the appropriate moment for
stopping the equipment. As maintenance based on continuous parameter monitoring performs better
than other maintenance methods, the mathematical model presented must be able to prevent random
occurrences. The solution is outlined around two main ideas:
• Determining the maintenance method for the system concerned according to the evolution of
the damage (implementation of the strategies of preventive maintenance, corrective
maintenance or the use of the system regardless of the degree of damage);
• Determining and establishing the data for the next inspection;
The simulations performed offer a wide range of solutions for determining the maintenance mode
or the use of the equipment until an imminent failure to obtain the desired result: the efficiency of the
maintenance costs.
An innovative approach is also taken by Wang, Jinjiang, et al. [12] who propose a predictive
maintenance system that is based on a new cloud architecture that uses a mobile client instead of the
classic server-client architecture. This approach aims to increase the flexibility and adaptability of the
system, reduce the volume of raw data that is transmitted and improve the response time to the
dynamic changes that occur in the monitored systems. This approach is not without problems, some of
the points that remain open are: automatic maintenance planning, energy management, management of
data analysis and transmission, data storage, large-scale system implementation. To address some of
these issues, as mentioned from the outset, this solution proposes using mobile agents to distribute
tasks in the cloud.
The mobile agent is a method derived from the field of study of artificial intelligence that aims to
divide the tasks and their parallel execution. Agents are represented by software programs that migrate
within the network. The main features of a mobile agent are autonomy, intelligence, adaptability,
which makes this approach suitable for cloud architecture. Such an implementation makes the
distribution of computing power more efficient, reduces the amount of data transmitted and allows the
modification and adaptation of algorithms running on equipment, thus increasing their versatility.
For system testing, six AC motors were used, five of them having different defects, and one being
used as a standard. Using mobile agents, software programs are sent to be run and to purchase
different parameters, according to predictive maintenance methodology. The data is then processed
with specific algorithms, after which a characteristic of the monitored equipment is obtained. The
results confirmed the efficiency of the system and provided relevant data on the condition of the
analysed engines compared to the standard engine.
In the paper [13], the authors propose a method of predictive maintenance that is performed with a
monitoring system subject to errors. For data generation, it is considered a system subject to
continuous degradation which is modelled using Markov chains. This method provides the means for
performing an efficient simulation to track the evolution of the main topic of this paper: correlating the
efficiency of data analysis and acquisition equipment with reducing costs in predictive maintenance
compared to other maintenance methods.
For a robust evaluation of the proposed system, the analysis is performed on 2000 simulated
elements, and the costs of procurement equipment used in predictive maintenance are also considered.
Of course, depending on the quality of the sensors and the cost of repairing the damage, predictive

ACME 2020
IOP Conf. Series: Materials Science and Engineering 997 (2020) 012039
IOP Publishing
doi:10.1088/1757-899X/997/1/012039
6
maintenance may or may not be the best option. When the simulation parameters are changed and the
quality of the sensors used for the acquisition decreases, the efficiency of the predictive maintenance
decreases considerably, being below the preventive maintenance. This can be solved only by using
sensors and modules for the acquisition and interpretation of qualitative data and also by implementing
filters for the acquired data.
The results of this simulation reinforce the idea of adopting predictive maintenance systems, which
improve maintenance costs, even when the cost of analytical equipment is taken as long as the sensors
and the acquisition systems provide qualitative data.
Ullah, Irfan, et al. [14] propose a predictive maintenance solution through thermography using ML
(Machine Learning) techniques. This type of anomaly occurs mainly in power equipment, where
imperfect contacts, corrosion, imperfect insulation, etc. causes overheating of the equipment. The use
of thermography is a non-invasive method that can monitor the equipment in operation. The analysis is
done on a total of 150 photos, which are determined around 300 points of interest. Then using ML
techniques, the points of interest in the photos are classified as defects or functioning in parameters.
This method is possible using an infrared thermographic camera. It captures and filters the infrared
light component and thus provides images on the thermal profile of the equipment being tracked. As in
the deterioration of the electrical equipment the main effect of the aging of the cables is represented by
an increase of the resistance, the thermographic analysis with infrared can offer relevant information
in the study of the equipment for the implementation of the practices of predictive maintenance.
Regarding the maintenance applied to electrical equipment, it is generally neglected, using the strategy
of use until failure and applying the corrective maintenance when the equipment requires it.
Using the means of Industry 4.0, Cachada, Ana, et al. [15] propose an innovative predictive
maintenance structure based on OSA-CBM (Open System Architecture for Condition Based
Monitoring) architecture. The OSA-CBM structure consists of six steps in comparison to the structure
proposed in this paper which consists of only four steps, as follows:
1. Data acquisition - is based on automatic, semi-automatic and manual data acquisition methods.
This is done using IoT (Internet of Things) and HMI (Human Machine Interface)
technologies;
2. Offline data analysis - this step is based on the use of several technologies such as machine
learning, cloud technology and advanced data analysis technology to make new monitoring
rules and procedures. Depending on the type of method and algorithms chosen, results with
different degrees of efficiency are obtained;
3. Dynamic monitoring - it is composed of two subassemblies: the visualization subassembly and
the early damage subassembly;
4. The module for decisions and support - when it is necessary to carry out tasks or interventions,
the information is sent to the module for decisions and support through an organizing tool;
The solution comes with improvements in comparison with OSA-CBM technique which makes
usage of the Industry 4.0 and cloud techniques in order to create an enhanced predictive maintenance
system.
Implementing predictive maintenance on an industrial basis represented by Industry 4.0 [16],
changes the processes in comparison with classical maintenance. One of the main problems is
represented by the overwhelming amount of data which is acquired and must be processed accordingly
[17]. This leads to big data paradigm, which must be treated using data management methods.
Data management processes have been around for a long time, yet new techniques are developed in
order to adapt to data environment changes. Data management processes aim to provide a standard
procedure for data analysis in order to evaluate and propose a solution based on analyzed data [18].
There are of course many data processes available [19], each suited to different data management
necessity, some of the most important ones being:
• Cross Industry Standard Process for data Mining (CRISP-DM) process model;
• Sample, Explore, Modify, Model and Access (SEMMA) process model;
• Team Data Science Process (TDSP) methodology;

ACME 2020
IOP Conf. Series: Materials Science and Engineering 997 (2020) 012039
IOP Publishing
doi:10.1088/1757-899X/997/1/012039
7
The first two models are around for about two decades, but the changes in the industrial processes
data analysis shaped around the new industries asked the necessity for new data management
processed, thus newer methods like TDSP emerged.
Figure 3 presents an intelligent predictive maintenance system framework implemented with
Industry 4.0 means [20].
Figure 3. Predictive maintenance in Industry 4.0 framework.
All of the above methods represent a novel approach in terms of predictive maintenance
techniques. As expected, the predictive maintenance started to become the standard in maintenance
techniques due to the multitude of benefits. With the new technological advances and the 4.0 Industry
movement, the factories and machines started to become more advanced and connected. This opened
the way for real time parameter monitoring which is the founding pillar in the predictive maintenance.
The speed and ease wherewith the data is now acquired and sent is something that must be taken
advantage of. With new innovations, like Internet of Things and Internet of Services new challenges
appear, data management and processing being one of them [21].
4. Conclusions
This paper starts with the beginning of the industrial revolution and briefly presents the industrial
evolution, but also its problems and changes over time. Initially seen as a means of eliminating
humans from the field of work, technology has transformed the world into what it is today.
Since the beginning of technological upgrades, the efficiency of systems and means of production
has been a priority to increase production and reduce costs. The maintenance of the equipment was a
problem that raised a major interest, being a broad topic on which continuous research was carried out.
Today, evolution allows the creation of intelligent and autonomous maintenance systems, which have
analysis capabilities far beyond those of human personnel.
The main subject of the present paper is predictive maintenance - an intelligent way of monitoring
the parameters of the equipment in real time in order to determine the possibility of damage. Further,
the concept of predictive maintenance in Industry 4.0 is presented and how the synergy between these
two revolutionary methods can generate more efficiency. Moreover, data management processes are
presented in order to deal with data acquired in order to implement predictive maintenance system.
The study of the current solutions offers an overview of the research and discoveries in this field.
There are presented several predictive maintenance solutions implemented in different environments
that have different systems analysis methods, use different algorithms, but which have a common
purpose: to determine the possible breakthroughs in advance in order to be able to plan a maintenance
strategy so that the efficiency to be as high as possible.
Therefore, intelligent maintenance started to gain weight in applied maintenance methods due to
the long term benefits it provides. Even if the implementation of Industry 4.0 may be yet in an

ACME 2020
IOP Conf. Series: Materials Science and Engineering 997 (2020) 012039
IOP Publishing
doi:10.1088/1757-899X/997/1/012039
8
incipient phase, the value added by the means offered by it can provide a starting point for intelligent
maintenance systems and methods which are superior to the old maintenance practices.
5. References
[1] Industrial Revolution. (2019). In: Encyclopaedia Britannica. Encyclopaedia Britannica
[2] Stark, John. "Product lifecycle management." Product lifecycle management (Volume 1).
Springer, Cham, 2015. 1-29
[3] EN 13306:2010, (2010) Maintenance Terminology. European Standard. CEN (European
Committee for Standardization), Brussels
[4] Scheffer, Cornelius, and Paresh Girdhar. Practical machinery vibration analysis and predictive
maintenance. Elsevier, 2004
[5] Sillivant, D. (2015, January). Reliability centered maintenance cost modeling: Lost opportunity
cost. In 2015 Annual Reliability and Maintainability Symposium (RAMS) (pp. 1-5). IEEE
[6] Park, Chulsoon, et al. "A predictive maintenance approach based on real-time internal parameter
monitoring." The International Journal of Advanced Manufacturing Technology 85.1-4
(2016): 623-632
[7] Petrasch, Roland, and Roman Hentschke. "Process modeling for Industry 4.0 applications:
Towards an Industry 4.0 process modeling language and method." 2016 13th International
Joint Conference on Computer Science and Software Engineering (JCSSE). IEEE, 2016.
[8] Bartodziej, Christoph Jan. The Concept Industry 4.0. Wiesbaden, Germany: Springer, 2017.
Print.
[9] Hashemian, Hashem M. "State-of-the-art predictive maintenance techniques." IEEE
Transactions on Instrumentation and measurement 60.1 (2010): 226-236.
[10] Lu, B., Durocher, D. B., & Stemper, P. (2009). Predictive maintenance techniques. IEEE
Industry Applications Magazine, 15(6), 52-60.
[11] Grall, Antoine, et al. "Continuous-time predictive-maintenance scheduling for a deteriorating
system." IEEE transactions on reliability 51.2 (2002): 141-150
[12] Wang, Jinjiang, et al. "A new paradigm of cloud-based predictive maintenance for intelligent
manufacturing." Journal of Intelligent Manufacturing 28.5 (2017): 1125-1137
[13] Curcurù, Giuseppe, Giacomo Galante, and Alberto Lombardo. "A predictive maintenance
policy with imperfect monitoring." Reliability Engineering & System Safety 95.9 (2010):
989-997
[14] Ullah, Irfan, et al. "Predictive maintenance of power substation equipment by infrared
thermography using a machine-learning approach." Energies 10.12 (2017): 1987
[15] Cachada, Ana, et al. "Maintenance 4.0: Intelligent and predictive maintenance system
architecture." 2018 IEEE 23rd International Conference on Emerging Technologies and
Factory Automation (ETFA). Vol. 1. IEEE, 2018
[16] Li, Z., Wang, Y., & Wang, K. S. (2017). Intelligent predictive maintenance for fault diagnosis
and prognosis in machine centers: Industry 4.0 scenario. Advances in Manufacturing, 5(4),
377-387.
[17] Lee, J., Kao, H. A., & Yang, S. (2014). Service innovation and smart analytics for industry 4.0
and big data environment. Procedia Cirp, 16(1), 3-8.
[18] Ghavami, P. (2019). Big Data Analytics Methods: Analytics Techniques in Data Mining, Deep
Learning and Natural Language Processing. Walter de Gruyter GmbH & Co KG.
[19] Olson, D. L., & Delen, D. (2008). Advanced data mining techniques. Springer Science &
Business Media.
[20] Wang, K. (2016). Intelligent predictive maintenance (IPdM) system–Industry 4.0 scenario. WIT
Transactions on Engineering Sciences, 113, 259-268.
[21] Ji, Changqing, et al. "Big data processing: Big challenges and opportunities." Journal of
Interconnection Networks 13.03n04 (2012): 1250009.