THE USA JOURNALS
THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN
–
2689-0984)
VOLUME 06 ISSUE07
94
https://www.theamericanjournals.com/index.php/tajet
PUBLISHED DATE: - 30-07-2024
https://doi.org/10.37547/tajet/Volume06Issue07-10
PAGE NO.: - 94-98
APPLICATION OF AI IN MAINTENANCE AND
RELIABILITY MANAGEMENT
Adilzhan Jailaganov
Graduate Student, MSC Global Supply Chain Management, University Of Southern California,
Los Angeles, USA
INTRODUCTION
The rapid evolution of AI technologies such as
machine learning, deep learning, the Internet of
Things, blockchains, and robotics is revolutionizing
the maintenance sector. This article will delve into
how the implementation of artificial intelligence is
fundamentally
transforming
traditional
maintenance management principles. It will
explore AI's potential benefits and the challenges it
faces today. This article will focus on AI's most
notable and effective applications in maintenance
management, such as predictive analytics, live
monitoring and data collection, operational
transparency, and inventory management.
PREDICTIVE MAINTENANCE
It is a matter of common knowledge that traditional
maintenance practices are based on two main
vectors
–
preventive and reactive maintenance.
Reactive maintenance includes situations when a
malfunction occurs unexpectedly, leading to
extensive downtimes of a particular equipment or,
in the worst scenario, to complete shutdown of
production operations. Meanwhile, preventive
maintenance involves regular asset inspections by
a schedule and specific service procedures, which
often can be time-costly and inefficient. To make
malfunctions more predictable, vast data should be
gathered and analyzed thoroughly. However,
processing and analysis of immense volumes of
such data are challenging and time-consuming for
human beings, and artificial intelligence is one of
the best tools to reach the best results in predicting
failures.
From a predictive maintenance perspective, AI
algorithms can forecast malfunctions by leveraging
historical data, sensor information, log records,
staff feedback, etc. Machine learning technologies
can identify deviations from standard patterns and
inform maintenance teams before failure occurs by
utilizing and analyzing vast amounts of data. Such
valuable forecasts can lead to sustainable
operations and significant cost savings. For
example, the US Air Force, by leveraging Amazon
RESEARCH ARTICLE
Open Access
Abstract
THE USA JOURNALS
THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN
–
2689-0984)
VOLUME 06 ISSUE07
95
https://www.theamericanjournals.com/index.php/tajet
Web Services (AWS), could save around $8.4M by
avoiding reactive maintenance activities within ten
months of regular operations. Furthermore, using
AI as a prediction tool helped the US Air Forces
avoid unnecessary parts replacement by 18%
during the same frame of time (1, Ann Claire
Carnahan, 2021).
Prescriptive maintenance
Besides prediction capabilities, ML technologies
can also use the input data for prescriptive
purposes. Deep learning is another type of machine
learning with higher analytical and self-learning
abilities. Moreover, it can process even more
complicated analyses in one piece of equipment
and in a group of assets connected to AI processing
machines (D. Shaidauf, E. Bowen, D. Williams,
E.Schoenbrun, H. Diaby, 2023). Deep learning
machines identify abnormalities in data that
outweigh normal values and subsequently inform
maintenance teams about it. By providing such
information, AI enables maintenance management
to identify equipment wear level and its remaining
useful life. It also helps to plan operations by
balancing equipment load and workforce
arrangement.
Another advantage of prescriptive maintenance is
the ability of ML, in collaboration with AI and the
Internet of Things (IoT), to provide estimates and
recommendations. By studying historical and real-
time data, the ML can recommend required actions
for specific equipment or adjustments even for the
whole system. In this case, ML generates different
models based on provided data and runs
simulations multiple times until it receives the
most precise outputs. Therefore, it can generate
recommendations for maintenance teams, the
latter can analyze the outcomes manually.
Moreover, ML, run by AI, can identify the asset's
remaining life by deep learning the operational
patterns and current condition from the sensors of
the equipment (Anwar et al., 2024). Besides, ML
has the capability to optimize service operations by
focusing on the most critical or vulnerable assets
that need extra attention from maintenance crews.
IoT and real-time monitoring
In the modern technological world, various
measuring equipment, whether mechanical or
digital, has been used for many years for
production operations control and analysis.
Nevertheless, digital sensors bring many benefits
today, such as precision, real-time monitoring and
control, remoteness, automation, and eligibility for
live analysis and data-driven model processing by
ML and AI. The modern name of sensors or
connectable measuring devices is the Internet of
Things (IoT). IoT, in collaboration with ML and AI,
provides opportunities to manage operations
smoothly, avoid risks, and have precise forecasts
on hand.
Having precise data about operational equipment
is crucial for any industry, and IoT can significantly
enhance information collection processes. Imagine
an oil refinery plant where thousands of gauges
and sensors measure and indicate different
operational parameters used for production
adjustment,
decision-making,
and
risk
management. So, a well-developed network of IoT
connected to an ML processing machine can make
that process times faster and provide deep
analytical reports and forecasted scenarios.
Operations Transparency and Blockchain
Another valuable feature of artificial intelligence is
blockchain technology, which was initially
developed for cryptocurrency registration only.
Nevertheless, today, blockchain is widely used in
various industries, such as supply chain
management, real estate, gaming, and healthcare.
The key feature of blockchain technology is that it
saves information on multiple servers and is
protected from edition and deletion (Danny Pehar,
2024). Thereby, the information entered will
remain on those servers forever. That feature can
be leveraged in maintenance management for
more transparent operations and to ensure
historical data backup storage.
Using blockchain in maintenance can bring many
benefits in terms of transparency. For instance,
historical data saved on multiple servers can be
used in different investigation or commissioning
processes. Imagine when a critical failure occurs on
a power plant, which leads to social or
environmental consequences
–
long-term power
and water outages or excessive pollution in
THE USA JOURNALS
THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN
–
2689-0984)
VOLUME 06 ISSUE07
96
https://www.theamericanjournals.com/index.php/tajet
surrounding areas. In such situations, truthful data
is key for investigators to identify root causes and
mitigate such compliance in operations and
maintenance practices. Therefore, blockchain
technology has great potential to improve
industries in terms of legal and environmental
compliance.
Inventory management
Spare parts and consumables are among the
significant issues in maintenance management,
which require quality, transparency, and
availability. Blockchain technologies can be utilized
to track the origin of spare parts or consumables to
ensure the quality of materials provided by
suppliers. Of course, it will depend on how the
blockchain system was designed and its options
and access levels, but it is possible to track the
origin up to the source of raw materials from which
details were made. Maintenance managers can
make informed decisions when choosing suppliers
and contract issues.
Inventory control and management
–
a widespread
problem of modern enterprises which not so easy
to solve. Maintaining inventory levels involves a
vast amount of data, and usually, it is challenging to
predict the exact quantity of needed parts or
consumables. AI and ML technologies can be more
than helpful in this situation. As in the prediction
and prescription sections, ML analyzes historical
data and generates patterns for creating a virtual
model on spare demand. Unlike traditional
planning, during analysis, AI can also consider
different occasions that occurred previously and
affected the inventory. Therefore, a collaboration
between AI and ML can significantly contribute to
supply chain management upgrades, perform
material planning, and provide quality confidence
and transparency.
Challenges
Costs
Implementation of AI into the maintenance process
varies and depends on the customer’s personal
requirements. Today, simple AI solutions would
cost at least $400-500K for small projects and up to
tens of millions of dollars for bigger ones (Jon
Reilly, 2024). Despite high prices, big organizations
and joint ventures do not hesitate to implement AI
technologies into their operations, and Repsol
petrochemical company is one of them. Repsol
from previous projects of implementing AI as a
predictive analytics tool has witnessed notable cost
savings of around $200M annually from reducing
reactive maintenance activities (Tom Mostyn,
2019). In this regard, Repsol is planning to expand
the AI implementation program and is planning
huge investments in the 2024-2027 Repsol
Strategic Update (Repsol S.A., 2024).
Technological complexity
Implementing AI in maintenance processes and
hardware systems would bring various difficulties
in addition to high prices (Jon Reilly, 2024). The
logic is the same as with prices: the higher the
requirements, the more complex the system will
be. The challenges can vary from hardware and
software to administrative or regulative
restrictions.
Moreover,
designing
and
implementing AI projects in maintenance
processes requires the application of knowledge
and expertise from both sides
—
the client and the
provider of AI services. Just imagine a situation
where a provider of AI services cannot completely
understand the operations philosophy of his client
due to a lack of expertise in that field. Such a
relationship is not effective and can barely bring
good results.
Another issue is the digitalization of the whole
process. As was described before, AI and ML
technologies require a vast amount of high-quality
data, and the best way to its collection is to
completely digitalize the system (Christian Langer
et al., 2024). That measure would give a raw of
benefits such as live monitoring, data recording,
and storage. However, not every system can be
wholly digitalized due to various environmental
obstacles. For instance, it is a challenging problem
to install some kind of sensors in zones with high
levels of radiation, temperature, pressure, or
vibration.
Personnel education
As with any innovation, AI and ML bring difficulties
for personnel using new systems and software that
require some expertise. In that regard, companies
THE USA JOURNALS
THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN
–
2689-0984)
VOLUME 06 ISSUE07
97
https://www.theamericanjournals.com/index.php/tajet
should consider that fact before implementing such
emerging
technologies.
Stakeholders
and
managers should consider arranging special
courses and practical exercises for laborers who
will be in touch with AI systems. Today, there are
plenty of ways to educate personnel on AI
principles, such as various online platforms,
workshops, simulators, or continuous education
programs.
Data volumes and quality
Despite ML's valuable benefits through its
analytical capabilities, it also needs a vast amount
of high-quality data to produce more precise
outcomes. The colossal information hub enables
ML to scrutinize the data to understand long-term
patterns and variability. The more information you
input, the more complex the system is (Hrvoje
Smolic, 2024). However, ML would struggle to
identify precise and accurate models in situations
without huge volumes of high-quality information.
Besides quantitative data requirements, ML also
needs systematic data entry, one of the most
challenging processes in applying ML technologies
to business operations. If the data collection
process is fully automated and digitalized, it will
allow the system to read and analyze the inputs and
generate accurate patterns steadily. However,
semi-automated or manual data collection systems
require strict rules and procedures in the data
collection process, particularly for personnel who
record and input the data into the ML processing
machine. Inaccurate or incorrect inputs will lead to
erroneous model generation and, consequently, to
inaccurate forecasts.
CONCLUSION
To conclude, AI and ML technologies have become
the new norm today, and many organizations tend
to apply those technologies to their business
operations. Applying AI in combination with ML
and IoT in maintenance management can
significantly enhance the reliability levels of any
enterprise.
Implementing
such
emerging
technologies can lead enterprises to sustainable,
transparent, and safe operations, which in turn can
strengthen the company's name and reputation on
the market. According to Deloitte's research,
predictive maintenance can substantially increase
the main KPIs
–
save costs in operations by up to
10%, increase equipment uptime by up to 20%,
reduce overall maintenance costs to 10%, and
significantly cut maintenance planning time by
almost twice (Olaf Peter Schleicher et al, 2017)
Another considerable benefit is cost savings
through failure and downtime reduction. British
Petroleum (BP) can be considered a perfect
example, having saved around $7B during the
2014- 17 period (Andrew Carr, 2017). So,
potentially costly investments can be returned
pretty fast, and the most value you gain is reliability
and customer satisfaction. Of course, any
investment in such technology should be
preliminary analyzed by stakeholders and
reviewed by financial and technical experts.
In general, AI has been conquering the industrial
sector for the last decade and should be considered
as one of the key vectors in maintenance
management development. Despite the variety of
challenges, AI with ML technology has a bright
future that can lead to sustainable supply chains
and environmental and operational safety.
Stakeholders, top managers, and professionals of
different industries should now focus on the
capabilities of emerging technologies such as AI
and ML and consider their implementation into
their operations because those technologies will
remarkably change the industrial world in the near
future.
REFERENCE
1.
Tom
Mostyn,
2019
https://www.hydrocarbonengineering.com
/refining/26112019/top-oil-and-gas-
companies-continue-to-expand-predictive-
maintenance-usage/
2.
Anwar Meddaoui, Adil Hachmoud ,
Mustapha Hain 2024 . Advanced ML for
predictive maintenance: a case study on
remaining useful life prediction and
reliability enhancement | The International
Journal
of
Advanced
Manufacturing
Technology (springer.com) Springer Link.
3.
How using AI for predictive maintenance can
help you become mission ready | AWS Public
THE USA JOURNALS
THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN
–
2689-0984)
VOLUME 06 ISSUE07
98
https://www.theamericanjournals.com/index.php/tajet
Sector Blog (amazon.com)
4.
D. Shaidauf, E. Bowen, D. Williams,
E.Schoenbrun, H. Diaby, (2023).
5.
Danny, Pehar (2024). How Blockchain
Revolutionizes
Data
Integrity
And
Cybersecurity,
How
Blockchain
Revolutionizes
Data
Integrity
And
Cybersecurity
(forbes.com),
Forbes
Technology Council.
6.
Repsol S.A., (2024). Repsol Strategic Update
2024-2027
–
Evolving from our strengths,
growing sustainable returns. Repsol S.A.
7.
Jon Reilly (2024). Cost of AI in 2024:
Estimating Development & Deployment
Expenses (akkio.com). Akkio.
8.
Hrvoje Smolic, 2024 How Much Data Do You
Need for Machine Learning (graphite-
note.com), Graphite Note.
9.
Andrew Carr, 2017). The Big Data
technologies
that
saved
BP
$7bn
(scottlogic.com)
10.
Christian Langer is a partner in McKinsey’s
Hamburg office; Daniel Leblanc is a partner
in the Dallas office; Dave Marcontell is an
associate partner in the Atlanta office, where
Joe Nutter is a consultant; Eric Porter is a
consultant in the Boston office; and Joel
Thibert, Aircraft maintenance companies
and gen AI | McKinsey
11.
Olaf Peter Schleichert, Dr. Björn Bringmann,
Dr. Sergey Zablotskiy, Dr. Hardy Kremer, Dr.
David Köpfer (2017) Deloitte_Predictive-
Maintenance_PositionPaper.pdf
