Uploaded by gartzounis

SSRN-id4366923

advertisement
Predictive Maintenance of HVAC Systems using Machine Learning
Nitin Goyala*, Naman Guptab*, Palak Bhatnagarc, Poorva Jaind, Priya Raie, Mukul Malikf
nitin.goyal@imsec.ac.in, Computer Science and Engineering, IMS Engineering College, Ghaziabad, Uttar Pradesh, India
153namangupta@gmail.com, Computer Science and Engineering, IMS Engineering College, Ghaziabad, Uttar Pradesh, India
c
A2019CSE6548@student.imsec.ac.in, Computer Science and Engineering, IMS Engineering College, Ghaziabad, Uttar Pradesh, India
d
A2019CSE6671@student.imsec.ac.in, Computer Science and Engineering, IMS Engineering College, Ghaziabad, Uttar Pradesh, India
e
A2019CSE6675@student.imsec.ac.in, Computer Science and Engineering, IMS Engineering College, Ghaziabad, Uttar Pradesh, India
f
A2019CSE6785@student.imsec.ac.in, Computer Science and Engineering, IMS Engineering College, Ghaziabad, Uttar Pradesh, India
a
b
Abstract: Industry has decreased losses mostly due to motor failure and considerably improved system dependability by continuously monitoring the status
of electric motors and other equipment used in HVAC systems. Predictive maintenance is the process of predicting the actual faults that may occur and
preventing them to avoid system failure or component loss. Predictive maintenance, combined with intelligent sensors, artificial intelligence, and machine
learning algorithms, makes the automated industries work more efficiently. Industry 4.0 is the new trend throughout the internet, which can be seen in
newspapers, technical articles, or company websites. Predictive maintenance is the key concept for Industry 4.0, which helps the industry manage their
machines more efficiently and effectively with the help of intelligent sensors put together with the machines in the industry. Although many advancements
have been proposed till now in the process of predictive maintenance, it is also important for the process to be cost-efficient and efficient for small-scale
industries too, with which they can also evolve themselves with the concept of Industry 4.0. We studied the publications that were published in the recent
20 years using a systematic review of the literature (SLR). In the predictive maintenance process, the application is composed of many IoT devices, such as
Node-MCUs and RS-485 sensors, which are all connected to the HVAC systems and will collect data on a regular basis and push it over the databases with
which it is integrated. The data thus obtained is first filtered for further processing. With this filtered data, the analysis is performed, and finally, the user is
prompted if the system needs maintenance to avoid a complete failure. As of now, we are working on two fields, i.e., temperature and energy consumption,
but in the future, we can extend it to detect more failures in HVAC systems, such as gas leakage, etc.
1. Introduction
It is no secret that HVAC systems [1] can be expensive to operate and
maintain. Maintenance is a very crucial aspect in the industry to take care
of, with its effect on cost and reliability being significant. In fact, National
Institute of Standards and Technology (NIST) [2] claims, HVAC systems
are one of the single largest energy consumers in buildings, accounting for
approximately 35% of total energy use in large commercial buildings and
about 55% of total energy consumed in residential buildings. This is
especially problematic for large multi-facility organizations that have
expansive heating and cooling systems to manage across their portfolio. For
many of these organizations, improving the efficiency of their HVAC
systems has been a top priority for several years now, as anything they can
do to reduce their energy consumption and operational costs will have a
major impact on their bottom line [3]. Predictive maintenance solutions are
one method of doing this. An asset's system's key performance indicators
are monitored and analyzed using predictive maintenance [4], a technology
that employs artificial intelligence and machine learning to spot possible
issues before they arise and result in expensive operational interruptions.
The term "predictive" [5] refers to the capability to predict when a critical
fault might occur so that the problem can be addressed before it causes
serious damage to the asset or disrupts business operations. When
implemented properly, it can help reduce equipment downtime by as much
as 70% and increase operating efficiency by up to 10%, leading to
significant cost savings for your organization [6]. There are several benefits
to implementing predictive maintenance in your facility. These include:
Reduced Equipment Downtime [7] because predictive maintenance
systems are capable of detecting potential faults before they cause a
problem, they can help minimize the impact of an unplanned outage and
help the consumer get the most out of the product through proper
maintenance. The amount of downtime needed for maintenance is
decreased with the use of a well-defined predictive maintenance algorithm.
The predictive maintenance research in the current study [8] is concentrated
on internal components, such as gas leakage and motor rotation analysis.
Data is collected by sensors, and after additional processing, it becomes a
filtered dataset. With supervised learning algorithms [9] applied to this
dataset, an alert is generated if it is needed. Although advancements to the
process are continuously being made with more efficient algorithms and
more advanced sensors, Intelligent sensors [10] came into existence in the
process in the last 10 years, and with them, the process becomes more
effective as compared to traditional ones. Intelligent sensors aid in the
storage of everlasting data generated by sensors connected to devices.
Furthermore, the current studies [8] focus more on the working of internal
components such as gas leakage, motor rotation, capacitor faults, etc. The
majority of previous research has concentrated on the traditional method of
predicting failure in internal components. Some progress has been made in
recent years, with increased efficiency but less cost-effective models. The
models proposed [11] must be cost-effective to support small-scale
industries and allow solo entrepreneurs to scale their businesses with lower
maintenance costs. Predictive maintenance is the main pillar of Industry 4.0
[12]; it allows industries or factories to run more efficiently while limiting
the costs associated with faulty machines. Effective maintenance
techniques may cut down on building upkeep expenses and potentially
prolong the life of building parts. In order to avoid replacing equipment
before it has reached the end of its useful life, it is essential to create a
maintenance plan when a machine or piece of equipment begins to exhibit
warning signs of potential failure in the future [13]. This is significant
because, throughout the course of the system's lifespan, maintenance
expenses might account for up to 75% of all maintenance expenses
worldwide. Performance monitoring, which passively assesses the system
and notifies the user (technician, application engineer, facility manager)
about potential concerns, is a crucial step toward better HVAC control
(hardware defects, poor control, etc.). Some of problems include using the
wrong amount of outside air, which causes unnecessary mechanical
Electronic copy available at: https://ssrn.com/abstract=4366923
cooling, oscillating control signals [14] caused by improper PI constants, a
permanent setpoint offset, and more. There are currently a number of
HVAC monitoring dashboards on the market, but they are often just used
In [18] Trivedi et.al. presents a study on air-conditioning systems are
widely used appliances in industries and homes worldwide. Productivity is
also increased with the comfort provided by air conditioning systems in
industrial and commercial settings. Using distributed sensing and
supervised machine learning, a model for predictive AC maintenance is put
out in this research. In order to identify the location and defect in the ACs,
it uses a decision tree method and SVM.
In [19], Dalzochio et.al. presents a study on the advances in the predictive
failures of Industry 4.0. The major objective of this study is to analyze the
logic behind predictive maintenance in the context of Industry 4.0.
Predictive maintenance is a new, hot issue, however there are difficulties in
the field of machine learning. The review's main objective was to find
frameworks that implement ontology-based or ML-based reasoning.
for simple trends. Boilers are one piece of equipment that has an on-board
diagnostics and monitoring system. The remaining of the paper has been
organized as follows depicted in Figure 1: Section II of this paper illustrates
the literature survey on predictive maintenance on HVAC systems
proposed in the last 20 years of research papers. A literature review [15] is
being presented to have a clear understanding of the current literature on
predictive maintenance. In Section III, the different algorithms used so far
in predictive maintenance are described, along with their features,
advantages, and disadvantages. Section IV explains the conclusion and
future scope of the research work conducted.
Figure 1: Organization of the paper
2. Literature Survey
The initial approaches discuss fault detection in the motors or other internal
components of HVAC systems. The major problems they focus on are gas
leakage and motor rotation. The methodologies, which first launched the
residual generation and analysis approach, employ diagnosis algorithms for
HVAC systems. A comprehensive difference between all the paper is being
depicted in Table 1.
In [16] Sakali et.al. presents a study on the building sector, or corporate
sector, is found to be the largest energy consumers and greenhouse gas
emitters, ranging from 30% to 40% of the total energy consumption. The
majority of predictive models are used in HVAC systems with sensors for
preventive maintenance. Different types of use cases need to be handled
with different algorithms, such as rule-based models, fuzzy logic models,
case-based models, ANN, SVM, PCA, KNN, and some supervised learning
algorithms. Predictive maintenance has become more efficient as a result
of IoT and artificial intelligence. It is observed here that the DBN is less
considered in predictive maintenance, and SVM and ANN are the most
widely used techniques in the process.
In [17] Paolanti et.al. presents study on the cutting machine and tries to find
a new predictive maintenance technique that gives more accurate results.
Its contributions mainly focus on the area of cloud architecture in Industry
4.0 and the machine learning approach to the real datasets from the
machines working in the field. It enables the adoption of dynamic rules for
maintenance management, which may be accomplished by training the
Random Forest technique.
In [20] Marik et.al. presents a study on energy efficiency can be increased
with the help of intelligent control on HVAC equipment. MPC is a strong,
industry-proven technology for improved control of complex systems, but
for the time being, its use in buildings seems to be a long way off. Low-cost
sensing devices or equipment have already made significant progress in the
field of predictive maintenance with traditional push. The goal in this field
is to create non-intrusive, inexpensive sensors that are simple to instal and
can wirelessly connect with building management systems. With these
many advancements, further processing can be done that generally requires
real-time data that is not present.
In [21] Singh et.al. presents a study on induction motors consume the
majority of energy in industries. Induction motors consume a lot of energy,
hence many efforts have been made to decrease their energy consumption
by improving the motor's efficiency and lowering its fault ratio. The paper
proposed a methodology that can be found effective for motor maintenance
decision-making, which at first focuses on efficiency and then on the costeffective solution, if one can be derived.
In [22] Pech et.al. presents a study on new technologies become available
in modern smart factories, production automation can be linked to
automated predictive maintenance. Intelligent sensors enable the storage of
everlasting data processed by system-integrated sensors. The approach has
the benefit that it may be employed in real-time, lowering the financial
expenses associated with production failure or downtime. The outdated
notion of software as a service is being replaced by the new idea of machine
as a service.
In [23] Ran et.al. presents a study on examination of the architecture,
purposes, and approaches to predictive maintenance systems has been
completed. Cost minimization, availability and reliability maximization,
and multiple objectives are the main objectives for performing predictive
performance, with the result that these can be achieved. Deep learning
approaches are found to be more efficient and cost-saving in the case of
predictive maintenance machines in Industry 4.0.
In [24] Halm-Owoo et.al. presents a study on fault detection and diagnostic
techniques in HVAC systems are the main issue, and this paper proposes a
clear understanding of all the Fault Detection and Diagnostic (FDD)
concepts. In the process of locating system flaws, a single system is
inefficient. The model thus far proposed will not be accurate due to the less
diverse dataset of the sensors applied to the RAC system.
Electronic copy available at: https://ssrn.com/abstract=4366923
In [25] Osterrieder et.al. presents a study on industry 4.0 being the new
emerging concept in all the newspapers, scientific journals, company
websites, etc. The main constructs of it are smart factories that can be
operated without the intervention of human beings by collecting,
processing, transferring, and generating data. AI and ML are being used in
Industry 4.0 to make factories smarter in their functions by applying more
relevant models or algorithms.
Table 1: Literature Survey
Electronic copy available at: https://ssrn.com/abstract=4366923
NAME OF PAPER
OBJECTIVE
ALGORITHMS USED
FOCUS ON
CONCLUSIONS
Review of predictive
maintenance algorithms applied
to HVAC systems. [16]
To explain and draft
difference between existing
algorithms being used for
Predictive Maintenance.
ANN [27], SVM [28], KNN
[29], CNN [30], RNN [31],
fuzzy-logic based model [32],
supervised model [33]
Differentiation between the
algorithms and to predict
their accuracy.
SVM, ANN are most
widely used algorithms
whereas DBN [39] is
less considered.
Machine Learning approach for
Predictive Maintenance in
Industry 4.0. [17]
Testing conditioning
monitoring with predictive
maintenance on real-time
machines in Industry 4.0.
Decision Forest Machine
Learning Algorithm [34]
Motors used in different
devices in Industry 4.0.
Shows 95% accuracy in
the preliminary results,
observed thus far.
Predictive Maintenance of Air
Conditioning Systems Using
Supervised Machine Learning.
[18]
Machine learning and
reasoning for predictive
maintenance in Industry 4.0:
Current status and challenges.
[19]
Identifying gas leakage and
capacitor malfunction in Air
Conditioners.
Decision Tree Machine
Learning Algorithm [35] and
SVM algorithms
Gas Leakage and
capacitors malfunction
93.5 accuracy had been
identified with decision
tree algorithm.
Uses several models to boost
efficiency as it looks at
scholarly advancements in
failure prediction.
SLR methodologies [36]
Importance of Predictive
Maintenance in Industry
4.0 with AI and ML
algorithms.
Predictive maintenance's
primary challenges with
machine learning and
reasoning are discussed
and demonstrated.
Advanced HVAC Control:
Theory vs. Reality. [20]
Aims to list the causes of
poor HVAC and suggest
potential solutions for more
effective control methods. .
Model Predictive Control
(MPC) [37]
To minimize the cost
involved in predictive
maintenance techniques.
Smart Integration,
interoperability,
standardization has been
developed to achieve the
desired goals.
Efficiency monitoring as a
strategy for cost effective
maintenance of induction
motors for minimizing carbon
emission and energy
consumption. [21]
Predictive Maintenance and
Intelligent Sensors in Smart
Factory: Review. [22]
Industry 4.0 aims to lower
the power consumption of
motors.
Supervised learning algorithms
To minimize the energy
consumption by faulty
motors and C02 emission
by the same.
Combined the
methodologies for motor
maintenance with the
cost-effective methods.
Review the most recent
research on intelligent
sensors and predictive
maintenance in smart
industries.
Comprehensive analysis of
the literature on predictive
maintenance with a focus on
the goals, methods, and
designs of the systems.
Classification [38]
Internal components of
HVAC systems and
sensors being used in the
process.
Methods for Smart and
Intelligent Predictive
Maintenance.
Advanced Deep Learning
Techniques.
Focused on system
architecture and its process
to evolve a low-cost
approach for the process.
Concluded a
comprehensive survey of
system architecture used
in predictive
maintenance.
Understanding of
fundamental principles
correlated with RAC
systems.[26]
ANN
Applications used in RAC
systems to reduce the
faults.
A general understanding
of the state of currently
operational commercial
RAC systems.
A Survey of Predictive
Maintenance: Systems,
Purposes and Approaches. [23]
Applications of fault detection
and diagnostic techniques for
refrigeration and air
conditioning: a review of basic
principles. [24]
AUTHOR
4
Electronic copy available at: https://ssrn.com/abstract=4366923
The smart factory as a key
construct of industry 4.0: A
systematic literature review.
[25]
Grouping of the selected
publications into eight
different viewpoints on the
smart factory.
Supervised learning algorithms
Classification of literature
with respect to eight
different parameters.
Clear understanding of
previous literatures w.r.t
the new trends in the
Industry4.0.
Figure 2: Predictive Maintenance Algorithms
3. Related Work
A predictive maintenance software analyses data collected from the
operating system using condition monitoring and prognostics
algorithms. Data from a machine is used in condition monitoring to
evaluate the machine's present state and to find and fix machine defects.
Machine data is information gathered using specialised sensors, such as
measures of temperature, pressure, voltage, noise, or vibration. The
measurements, known as condition indicators, are derived from the data
by a condition monitoring algorithm. Any aspect of system data with
predictable behaviour changes as the system deteriorates is a condition
indicator. Any number obtained from the data that groups system
statuses with the same characteristics together and differentiates between
them is a condition indicator. So, by comparing fresh data, a conditionmonitoring algorithm may carry out defect identification or diagnosis. A
concise definition of the algorithms had been depicted in Table 2.
3.1.1.
3.1.2.
Types of Knowledge based models
These are mainly divided into three sub-categories as depicted in Figure
2, i.e.,
•
Knowledge-based models [40]
3.1.2.1.
Case-based model [43]
•
Models-based models [41]
•
Data-driven based models [42]
A knowledge base and an inference engine are the two components of a
knowledge-based system. Instead of being implicitly incorporated in
procedural code, the knowledge base explicitly specifies the facts about
the world. Other common approaches, rather than the subsumption
ology, have frames, conceptual graphs, and logical assertions. New
knowledge may be inferred thanks to the second component, also
referred to as the inference engine. It can take the most typical form of
IF-THEN rules using forward and backward chaining techniques. Other
strategies employ logic programming, blackboard systems, and
automated theorem provers.
Case-based reasoning is used in the knowledge-based reasoning system.
It involves the review of past knowledge of similar situations. And on
the basis of what it finds, the knowledge-based system provides solutions
that were effective in those given situations.
3.1.2.2.
3.1.
Knowledge Based Models
A computer software known as a knowledge-based system (KBS)
employs a knowledge base to resolve extremely complicated issues. This
phrase covers a wide range of system kinds and is quite general. The
most common kind that unifies all knowledge-based systems is the
endeavour to represent knowledge explicitly and a reasoning mechanism
that will allow it to derive new knowledge for the system.
Features
Rule-based model [44]
The rule-based systems depend on human-made, hard-coded rules. It
uses these hard-coded rules to analyse and manipulate the data to achieve
specific outputs. This may involve using IF-THEN rules that establish
that if a user makes a certain request, then the system can deliver a certain
outcome.
3.1.2.3.
Fuzzy Logic based model [45]
The Fuzzy logic-based model is a logical mathematical procedure which
is totally based on IF-THEN rules system. It allows the human thought
process to be produced in a mathematical form. It is a method of
processing variables that enables the processing of several potential truth
values through a single variable.
Advantages:
• Any choice can be accepted by the knowledge-based system as long
as it is solely based on the manager's prior knowledge or expertise.
• It makes it possible to quickly analyse a lot of data.
• It always takes current data into account.
• Previous experience is unimportant in this case.
• Different persons might make the decisions without the outcome
altering.
• A knowledge-based model is extremely dependable.
Electronic copy available at: https://ssrn.com/abstract=4366923
Disadvantages:
• There is no pattern-specific result.
• People are likely susceptible to the errors.
• Previous experiences were not always positive.
• The manager needs to devote a lot of time to data analysis.
• It does not accept judgements that depart from the law.
stakeholders understand the system.
• Information coherence: The complexity challenge has existed since
the beginning of Systems Engineering. It addresses the issue by enabling
the system engineer to create a much more comprehensive vision of the
system. This simple model allows for much greater consistency in
information.
3.2. Model based Models
Making unique models for each new application is possible with the
model-based approach to machine learning as depicted in Figure 3.
Although, it frequently does not, this model could occasionally match a
conventional machine learning method. In order to specify the model, a
model specification language is typically used, allowing for succinct
code definition. This code allows for the automated generation of the
program that will use that model.
Disadvantages
• Functionality: One of its major issues is usability; the approach
requires more effort to use, and simple characteristics are complicated.
If it is superior to existing systems engineering and project management
approaches, it will be easy to navigate and use.
• Static overtime: The model becomes static over time as it is built on
a static framework. This can result in a lack of agility and an inability to
respond to environmental changes. A model change will necessitate a
rethinking of everything that has come before, which can be a timeconsuming and difficult process.
• Effort to use: The approach necessitates more effort, and simple
characteristics may be overly complex. This can lead to frustration and
the impression that the strategy isn't working. The extra effort may also
be ineffective, especially when traditional methods are adequate.
3.3. Data Driven Models
Figure 3: Model Based Models Architecture
3.2.1.
Features:
•
The capacity to build an extremely wide range of models, together
with appropriate inference or learning algorithms, in which many
conventional machine learning approaches emerge as special
instances.
•
Separation between the model and the inference algorithm ensures
that, in the event that the model is amended, the correspondingly
adjusted inference method is automatically constructed.
Advantages
• Integration of time and effort: A model-based system is an
interconnected system. When you make changes to a model, all of its
dependents are automatically updated. In a DBMS [46], any change to a
document necessitates editing the document and updating all related
documents.
• Interaction: The use of models as the primary mode of
communication among team members improves understanding and
allows for more detailed communication. Diagrams and pictures help all
Data-driven models describe how a site's overall structure is described.
Relationships and regulations that restrict how data can be stored and
used are also incorporated into a database model. Models for data
management and storage were developed decades ago. Data has always
been essential to businesses and institutions. Data-driven models are
becoming the focus of much more investment due to their limitless
applicability. A data-driven model is built on an analysis of data
regarding a particular system. The fundamental idea underlying datadriven models is to identify correlations between system state variables
(input and output) without taking into account the system's physical
phenomena. This model primarily shows how the logical structure of a
database is modelled. It includes all the relationships and constraints that
determine the data. Data is a fundamental part of all companies, and due
to the infinite number of applications available. Data-driven models have
become important for structuring.
3.3.1 Features:
The generation, evaluation, and updating of data is continuous.
Additionally, it is crucial to software engineers' job since it offers
precise, useful feedback that enables them to determine where and how
to make changes to a product or procedure.
Data is an integral part of any digital transformation. The four pillars of
data-driven models are:
•
Company Vision— Although technically not one of the pillars, it
is a crucial component. Each pillar is influenced by the vision and
is defined by it.
•
KPIs are used to track continuous business performance, which
includes profitability and progress toward a goal.
•
OKRs: It measures what is happening now to achieve and improve
the company's results. Strong OKRs result in improved KPIs.
•
Engineering metrics typically reveal what constitutes a good
measure for software development or a great developer.
Electronic copy available at: https://ssrn.com/abstract=4366923
•
Good engineering metrics should lead to consensus on software
engineering standards, a high standard for work quality, and the
development of more and better features to support more value
work.
Table 2: Description of the Algorithms
S. No.
1.
Name of Algorithm
Knowledge Based Model
2.
Model Based Model
3.
Data-driven Based Model
Principle
It is a computer software that
employs established information
to unravel complicated issues.
It is a computer program that
divides the complex problems to
the small models.
Models are described on the
basis of previous old data stored.
3.3.2 Types of data driven models:
The different types of data-driven models are:
•
Hierarchical Model: data is organized hierarchically.
•
Network Model—improvement to the hierarchical model for
reducing data redundancy.
•
The Relational Model, which is the most common of all.
•
Object-Oriented Database Model—it defines a database as a
collection of reusable objects associated with its features and
methods. It also includes the multimedia and hypertext databases.
Advantages
•
Knowledge about the underlying process is not required.
•
Applicable to problems that cannot be physically analysed.
•
A random source for datasets.
Disadvantages
•
Ignore the physics of the underlying system.
•
Correlations are only valid within the range of the dataset.
•
Prediction accuracy is determined by the quality of the data.
Advantages
Accept any decision based on
previous knowledge.
Disadvantages
No pattern specific result,
Integration of time and effort,
Information coherence.
Static overtime, Complex to use.
Random dataset.
Accuracy is determined on the
basis of quality of data.
are used to implement deep learning, and the biological neuron—a.k.a.
brain cell—serves as the inspiration for these networks.
4. Conclusion and Future Work
As a result, using additional techniques to apply a hybrid multi-based
model will produce complete findings. It might thus be advised, based
on the results of the survey, to combine many ML or DL models in order
to get better predictions than when using a single model. It is thus also
suggested to mix classification and anomaly detection methods in order
to preserve the accuracy of classification models without sacrificing the
advantages of anomaly detection methodologies. A hybrid multi-based
model will therefore yield comprehensive results when used with
additional methodologies. In order to obtain better forecasts than when
using a single model, it may thus be suggested, depending on the survey's
findings. The accuracy of classification models may be maintained
without compromising the benefits of anomaly detection approaches by
combining classification and anomaly detection algorithms, it is claimed.
3.4. Examples of Data Driven Models
3.4.1.
Supervised Learning
A model is created by the supervised learning algorithm using a collection
of labelled data. Once the correctness of the model has been verified, you
may utilise it in production with actual data. The desired result serves as
the supervision, allowing you to modify the function based on the actual
output it generates.
3.4.2.
Unsupervised Learning
Unsupervised learning analyses and groups unlabelled datasets using
machine learning methods. The models themselves glean insights and
patterns from the data. It is comparable to the learning that occurs in the
brain when a person learns something new. Unsupervised learning aims
to discover the underlying structure of a dataset, categorise the data based
on similarities, and describe the dataset in a compact form.
3.4.3.
Deep Learning
In artificial intelligence, deep learning is a subset of machine learning.
By using different types of data and using different learning strategies,
deep learning differs from traditional machine learning. Neural networks
REFERENCES
[1]
MarijaTrčka, Jan L.M.Hensen, Overview of HVAC system
simulation, 2010.
[2]
Peter Mell, Timothy Grance, The Nist Definition of Cloud
Computing, 2011.
[3]
Cristina Gimenez, Vicenta Sierra, Juan Rodon, Sustainable
operations: Their impact on the triple bottom line, 2012.
[4]
Sule Selcuk, Predictive maintenance, its implementation and
latest trends, 2017.
[5]
Waljee AK, Higgins PD, Singal AG. A primer on predictive
models. Clin Transl Gastroenterol. 2014.
[6]
Kumar, S. and Harms, R. (2004), "Improving business processes
for increased operational efficiency: a case study", 2004.
[7]
Ridgway, Malcolm PhD, CCE; Atles, Leslie R. CCE, CBET;
Subhan, Arif MS, CCE. Reducing Equipment Downtime: A New Line of
Attack, 2009.
[8]
Helge Nordal, Idriss El-Thalji, Modeling a predictive
maintenance management architecture to meet industry 4.0 requirements: A
case study, 2020.
[9]
Vladimir Nasteski, An overview of the supervised machine
learning methods, 2017.
Electronic copy available at: https://ssrn.com/abstract=4366923
[10]
Pech M, Vrchota J, Bednář J. Predictive Maintenance and
Intelligent Sensors in Smart Factory: Review. Sensors. 2021
[11]
Serradilla, O., Zugasti, E., Rodriguez, J. et al. Deep learning
models for predictive maintenance: a survey, comparison, challenges and
prospects, 2022
[12]
Lasi, H., Fettke, P., Kemper, HG. et al. Industry 4.0, 2014
[13]
Low, Sui Pheng, and Hennie Faizathy Omar. "The effective
maintenance of quality management systems in the construction industry."
International Journal of Quality & Reliability Management 1997.
[14]
Lin, Jong-Yih, and Jauh-Hurng Ke. "Application of dampedoscillation control signals for wide-band feedback active noise control in
acoustic ducts”, 2003
[15]
Carvalho, Thyago P., et al. "A systematic literature review of
machine learning methods applied to predictive maintenance.", 2019
[16]
Niima Es-sakali, Moha Cherkaoui, Mohamed Oualid Mghazli,
Zakaria Naimi, Review of predictive maintenance algorithms applied to
HVAC systems, Energy Reports, Volume 8, Supplement 9, 2022, Pages
1003-1012, ISSN 2352-4847.
[17]
Marina Paolanti, Luca Romeo, Adriano Mancini, Emanuele
Frontoni, Machine Learning approach for Predictive Maintenance in Industry
4.0, 2018.
[18]
Shrishti Trivedi, Predictive Maintenance of Air Conditioning
Systems Using Supervised Machine Learning, 2019.
[19]
Jovani Dalzochio, Rafael Kunst, Edison Pignaton, Alecio Binotto,
Srijnan Sanyal, Jose Favilla, Jorge Barbosa, Machine learning and reasoning
for predictive maintenance in Industry 4.0: Current status and challenges,
2020.
[20]
Karel Mařík, Jiří Rojíček, Petr Stluka, Jiří Vass, Advanced
HVAC Control: Theory vs. Reality, 2011.
[21]
Gurmeet Singha, T.Ch.Anil Kumar, V.N.A.Naikana, Efficiency
monitoring as a strategy for cost effective maintenance of induction motors
for minimizing carbon emission and energy consumption, 2019.
[22]
Martin Pech , Jaroslav Vrchota, Jiˇrí Bedná, Predictive
Maintenance and Intelligent Sensors in Smart Factory: Review, 2021.
[23]
Yongyi Ran, Xin Zhou, Pengfeng Lin, Yonggang Wen, Ruilong
Deng, A Survey of Predictive Maintenance: Systems, Purposes and
Approaches, 2019.
[24]
A K Halm-Owoo and K O Suen, Applications of fault detection
and diagnostictechniques for refrigeration and air conditioning:a review of
basic principles, 2002.
[25]
Philipp Osterrieder, Lukas Budde, Thomas Friedli, The smart
factory as a key construct of industry 4.0: A systematic literature review, 2019.
[26]
Nasution, Dian Morfi, Muhammad Idris, and Nugroho Agung
Pambudi. "Room air conditioning performance using liquid-suction heat
exchanger retrofitted with R290.", 2019
[27]
Agatonovic-Kustrin, S., and Rosemary Beresford. "Basic
concepts of artificial neural network (ANN) modeling and its application in
pharmaceutical research.", 2000.
[28]
Vishwanathan, S. V. M., and M. Narasimha Murty. "SSVM: a
simple SVM algorithm.", 2000
[29]
Guo, Gongde, et al. "KNN model-based approach in
classification.", 2003
[30]
Kozek, Tibor, Tamás Roska, and Leon O. Chua. "Genetic
algorithm for CNN template learning.", 1993
[31]
Sherstinsky, Alex. "Fundamentals of recurrent neural network
(RNN) and long short-term memory (LSTM) network.", 2020
[32]
Sugeno, Michio, and Takahiro Yasukawa. "A fuzzy-logic-based
approach to qualitative modeling.", 1993
[33]
Jiang, Tammy, Jaimie L. Gradus, and Anthony J. Rosellini.
"Supervised machine learning: a brief primer.", 2020
[34]
Kontschieder, Peter, et al. "Deep neural decision forests.", 2015
[35]
Su, Jiang, and Harry Zhang. "A fast decision tree learning
algorithm.", 2006
[36]
van den Heuvel, Henk, Louis Boves, and Eric Sanders.
"Validation of Content and Quality of SLR: Overview and Methodology."
(2002).
[37]
Rawlings, James B. "Tutorial overview of model predictive
control.", 2000
[38]
Neelamegam, S., and E. Ramaraj. "Classification algorithm in
data mining: An overview.", 2013
[39]
Naskath, J., G. Sivakamasundari, and A. Begum. "A study on
different deep learning algorithms used in deep neural nets: MLP SOM and
DBN.", 2022
[40]
Peuget, Raphael, Stephane Courtine, and J-P. Rognon. "Fault
detection and isolation on a PWM inverter by knowledge-based model.",
1998
[41]
Kaptein, B. L., et al. "A new model-based RSA method validated
using CAD models and models from reversed engineering.", 2003
[42]
Solomatine, D., Linda M. See, and R. J. Abrahart. "Data-driven
modelling: concepts, approaches and experiences.", 2009
[43]
Corchado, Juan M., and Brian Lees. "A hybrid case-based model
for forecasting.", 2001
[44]
Setnes, Magne, Robert Babuska, and Henk B. Verbruggen.
"Rule-based modeling: Precision and transparency.", 1998
[45]
Aluclu, I., A. Dalgic, and Z. F. Toprak. "A fuzzy logic-based
model for noise control at industrial workplaces.", 2008
[46]
McCarthy, Dennis, and Umeshwar Dayal. "The architecture of an
active database management system.", 1989
Electronic copy available at: https://ssrn.com/abstract=4366923
Download