Fault Detection using machine Learning Techniques

OBJECTIVE
The objective of fault detection using machine learning techniques is to develop systems that can
automatically identify and diagnose faults in machines and systems. This can be done by training
machine learning models on historical data of machine performance, such as vibration data,
temperature data, and sensor readings. Once trained, these models can be used to monitor
machine performance in real time and identify any anomalies that may indicate a fault.

PROBLEM STATEMENT

Developing machine learning models that can accurately detect faults in systems and machines,
even in the presence of noise, concept drift, and multiple simultaneous faults.Collecting and
labeling large amounts of data to train machine learning models.Engineering the data to extract
the most relevant features for fault detection.Selecting and training the appropriate machine
learning algorithm and hyperparameters.Deploying and monitoring machine learning models in
production environments.Developing machine learning models that can detect faults in
manufacturing equipment, such as robots, CNC machines, and assembly lines, in real time, so
that we can prevent downtime and improve product quality.Developing machine learning models
that can detect faults in power generators, transformers, and transmission lines in real time, so
that we can prevent blackouts and ensure the reliability of the power grid.Developing machine
learning models that can detect faults in aircraft engines, avionics, and other critical systems, so
that we can ensure the safety of passengers and crew. Developing machine learning models that
can detect faults in medical devices, such as MRI machines, CT scanners, and pacemakers, so
that we can prevent patient harm.
ABSTRACT
Additive manufacturing (AM) is poised to bring a revolution due to its unique production
paradigm. It offers the prospect of mass customization, flexible production, on-demand and
decentralized manufacturing. However, a number of challenges stem from not only the
complexity of manufacturing systems but the demand for increasingly complex and high-quality
products, in terms of design principles, standardization and quality control. These challenges
build up barriers to the widespread adoption of AM in the industry and the in-depth research of
AM in academia. To tackle the challenges, machine learning (ML) technologies rise to play a
critical role as they are able to provide effective ways to quality control, process optimization,
modelling of complex systems, and energy management. Hence, this paper employs a systematic
literature review method as it is a defined and methodical way of identifying, assessing, and
analysing published literature. Then, a keyword co-occurrence and cluster analysis are employed
for analysing relevant literature. Several aspects of AM, including Design for AM (DfAM),
material analytics, in situ monitoring and defect detection, property prediction and sustainability,
have been clustered and summarized to present state-of-the-art research in the scope of ML for
AM. Finally, the challenges and opportunities of ML for AM are uncovered and discussed.
INTRODUCTION
Additive manufacturing (AM), also known as rapid prototyping, 3D printing, and freeform
fabrication, is capable of depositing, joining or solidifying materials to construct physical objects
from computer-aided design (CAD) models [1]. Compared with conventional manufacturing
methodologies, such as subtractive manufacturing and formative manufacturing, AM systems
show higher efficiency and flexibility within the high-yield production and offer a new
perspective for the design and processing of both parts and materials. However, the AM process
is well-known as a highly complex system including various technologies that combines material
science, mechanics, optics, and electronics with computer science. As a result, the quality of
produced parts is affected by numerous factors, such as material properties, processing
parameters, process stability, and working conditions. This leads to the challenges that are
summarized and highlighted as follows.

It is generally difficult to model the mathematical relations of the underlying AM process

because the correlated factors are from various heterogeneous perspectives and different process
stages.

High-fidelity physical-based models are generally too complicated considering the in-process
uncertainties of the AM process, which demand significant computational resources.

It is challenging to integrate AM digital models, pertinent to various phenomena, at multiple

scales into one unified framework [2].
The applications of machine learning (ML) technologies have been proved effective in a wide
range of fields, such as computer science, aviation, healthcare, and the manufacturing
industry [3]. With the advancement of data acquisition and storage technologies, data-driven
approaches based on ML technologies have been increasingly adopted to discover hidden
knowledge and build highly complex relationships in digital manufacturing systems [4]. By
using reliable datasets, ML models are capable of learning hidden patterns and uncovering latent
knowledge to support decision-making, in terms of process optimization, quality control, and
system improvement. As one of the most popular manufacturing systems in Industry 4.0, AM
has been incorporated with digital systems and sensor networks where high-volume data can be
obtained. Hence, a growing number of researchers have applied ML algorithms to tackle
challenges in AM, such as design optimization, in situ monitoring, process modelling,
and energy management. However, different researchers and organisations focus on various AM
issues by using diverse ML technologies. To clarify the significant research challenges and
future opportunities of ML for AM, a comprehensive review is necessarily crucial to summarise
and analyse current research topics. There have been some existing review articles focusing on
different perspectives of this topic [5], such as the ML for the material development of AM [6],
and ML applications in laser powder bed fusion processes [7]. Additionally, the highlighted
papers in these review articles were selected and reviewed based on authors’ research and
industry experience, which is valuable but subjective. In order to discover the exhaustive
challenges and opportunities in this increasingly growing research field, a systematic and data-
driven review method is needed.
This paper aims to review, summarise, analyse and present the latest research and applications of
ML for AM. Section 2 introduces the state-of-the-art of the AM process, including seven main
AM system categories and their characteristics. In Section 3, the methodology of systematic
review is introduced. Based on the results of the systematic review, this paper also applies a data-
driven method to analyse notable articles, a keyword co-occurrence and clustering method.
Several aspects of AM issues are clustered based on these notable articles which are reviewed
and summarised in Section 4. The challenges and opportunities are discovered and clarified in
concludes.
LITERATURE REVIEW
Additive manufacturing (AM) methods for rapid prototyping of 3D materials (3D printing) have
become increasingly popular with a particular recent emphasis on those methods used for
metallic materials. These processes typically involve an accumulation of cyclic phase changes.
The widespread interest in these methods is largely stimulated by their unique ability to create
components of considerable complexity. However, modeling such processes is exceedingly
difficult due to the highly localized and drastic material evolution that often occurs over the
course of the manufacture time of each component. Final product characterization and validation
are currently driven primarily by experimental means as a result of the lack of robust modeling
procedures. In the present work, the authors discuss primary detrimental hurdles that have
plagued effective modeling of AM methods for metallic materials while also providing logical
speculation into preferable research directions for overcoming these hurdles. The primary focus
of this work encompasses the specific areas of high-performance computing, multiscale
modeling, materials characterization, process modeling, experimentation, and validation for final
product performance of additively manufactured metallic components.
Additive manufacturing (AM) has yielded major innovations in the electronics, biomedical and
energy domains. One of the AM techniques which has witnessed widespread use is the inkjet 3D
printing (IJP). The IJP process fabricates parts by depositing colloidal liquid droplets on
substrates. Despite its advantages, variations in input process parameters and fluid properties can
have a profound impact on the print quality. This paper aims to address this issue by presenting a
novel vision-based approach for in-situ monitoring of droplet formation. Further, a machine
learning model was used to study the relationship between droplet attributes and droplet modes.
A drop watcher camera was used to capture a sequence of videos obtained from different
combinations of voltage and frequency. Custom source code was developed using python
libraries to capture variations in droplet attributes (droplet size, velocity, aspect ratio, and
presence of satellites) and their impact on the droplet modes (normal, satellite, and no-droplet)
using computer vision. A backpropagation neural network mode (BPNN) was applied, with the
droplet features as inputs, to classify output droplet modes. The BPNN classified droplet modes
with 90% (high) accuracy. This research forms the basis for future development of digital twin
model of inkjet 3D printing towards predictive analysis and process optimization.
In metals additive manufacturing (AM), materials and components are concurrently made in a
single process as layers of metal are fabricated on top of each other in the near-final topology
required for the end-use product. Consequently, tens to hundreds of materials and part design
degrees of freedom must be simultaneously controlled and understood; hence, metals AM is a
highly interdisciplinary technology that requires synchronized consideration of physics,
chemistry, materials science, physical metallurgy, computer science, electrical engineering,
and mechanical engineering. The use of modern machine learning approaches to model these
degrees of freedom can reduce the time and cost to elucidate the science of metals AM and to
optimize the engineering of these complex, multidisciplinary processes. New machine learning
techniques are not needed for most metals AM development; those used in other sects of
materials science will also work for AM. Most prolifically, the density functional theory (DFT)
community has used many of them since the early 2000s for evaluating numerous combinations
of elements and crystal structures to discover new materials. This materials technologies-focused
review introduces the basic mathematics and terminology of machine learning through the lens
of metals AM, and then examines potential uses of machine learning to advance metals AM,
highlighting the many parallels to previous efforts in materials science and manufacturing while
also discussing new challenges and adaptations specific to metals AM. Metals additive
manufacturing (AM) has created a paradigm shift in the way metal components are
manufactured; materials and parts are fabricated simultaneously using a single machine, highly
complex geometries are possible, and local variations of microstructure-property relationships
may be realized through local process variations. Although decades of scientific and engineering
work in industry, academia, and government have resulted in the commercialization of metals
AM technologies, the consistency and quality of parts and materials are still open challenges for
many applications. In recent decades, Integrated Computational Materials Engineering (ICME)
approaches have proven to accelerate the development and adoption of materials
technologies [1]. Traditionally, ICME approaches incorporate physics-based experimental data
with simulations that span different length and time scales. However, for metals AM, much of
the physics are still being discovered; hence, the development of comprehensive,
computationally feasible physics-first approaches to ICME are still an open challenge. The
diverse array of promises and problems in AM has resulted in a field of study that is rich with
data – so much so that our ability to store and analyze the data is challenged. At the same time,
this wealth of data is motivating a paradigm shift to incorporate machine learning into ICME
approaches.
The 20th century saw the maturation of materials science and engineering as a field of study,
enabling targeted materials discoveries and innovations for specific applications. Over the past
several decades, materials development cycles have greatly accelerated by formulating materials
problems through the process-structure-property-performance paradigm [2], [1].
The process-structure-property-performance (PSPP) paradigm is a core philosophy in materials
science and engineering that governs how the manufacturing of a material determines its ability
to be used in different engineering applications. The PSPP relationships break down materials
development into four key areas of scientific and engineering interest [2].
In AM, the processing of a material is dictated by the thermal, mechanical, and chemical changes
experienced during its manufacture. Controllable machine parameters like energy density of the
heat source, the path in which material is deposited or fused, the order in which part layers are
manufactured, or the location of parts on the build plate are determining factors of the material
process history. Table 1 shows many of the controllable parameters common to laser-based
additive manufacturing systems. The choice of these parameters largely impacts the processing
history. The true processing history, however, is better described by the thermal history of the
build volume, both during manufacture and post-processing, the mechanical forces it
experiences, and any chemical reactions that occur in or on the part. Processing routes are often
discussed in AM and typically refer to beneficial or detrimental processing histories that impact
the part’s structure.
Table 1. A possible design space for laser powder bed fusion additive manufacturing. There are
over 104 possible combinations of machine inputs, based on the listed ranges and step sizes. Any
possible combination of these parameters is a point in the design space.
The structure of a material is a wide-ranging concept that spans many length scales. Structure
can refer to the crystallographic structure at the atomic scale, to the morphology and orientation
of grains at the mesoscale, to the geometry being manufactured at the macroscale. Microstructure
is a term often used in materials science referring to a specific subset of the material structure.
Microstructure for metals most commonly refers to grain and sub-grain level information like
material phases, grain morphologies, texture, and any defects like pores or dislocations that
might be present. Microstructures are often considered in analysis of material structures because
they fundamentally dictate a material’s properties.
The properties of a material are characteristics that determine its qualities. Properties of metals
AM parts that have been of interest are wide-ranging and they vary depending on the desired
engineering application of the part. Mechanical properties are some of the most studied for AM
metals since the majority of metals applications are structural. Other properties of interest
include thermal conductivity, which determines the heat transfer through an AM part, chemical
properties, like corrosion resistance, and optical properties, like reflectivity.
The performance of a part is its ability to be successfully implemented in an engineering
application. Performance can be viewed through the lifetime of an AM part when subjected to
the mechanical, thermal, chemical, etc., forces it will experience. Early additively manufactured
alloys showed degraded-to-comparable static properties compared to traditionally manufactured
alloys [3]. Further research and development improved the static properties of AM materials, yet
high microstructure variability and defect density can still cause AM material to fail
unexpectedly in fatigue limited applications [4], [5]. Some recent AM developments have
resulted in material properties that exceed those of traditionally manufactured
materials [6], [7], [8], [9], [10], [11]. Ultimately AM processes are unique relative to other
metal fabrication techniques and it is difficult to make fair comparisons regarding performance
across various manufacturing methods. When properly designed, AM parts can meet the
intended performance needs in a wide variety of end-use applications. The large combinatorial
space of manufacturing options in AM often obfuscates how proper design choices can be made.
The materials scientist interacts with the process-structure-property paradigm in traditionally
manufactured materials. Traditional material manufacturing can be phrased in a cause-and-effect
relationship between process, structure, and property. Once the material has been developed and
characterized by the materials scientist or engineer, another engineer then considers the property-
performance linkage. Since material is made separately from an engineered part in traditional
manufacturing, the PSPP paradigm can be broken up into these two separate sets of relationships.
In AM, the material and the part are made simultaneously. Simultaneous material-part
manufacturing motivates consideration of linkages across the entire PSPP paradigm. The ICME
approach to materials science is focused on modeling, bridging, and predicting relationships
throughout the PSPP paradigm.
Computational materials science and engineering has enabled the prediction of microstructure
from processing and of properties from microstructure, reducing the need for costly and time
consuming experimentation in discovering or developing a new material and/or its
manufacturing. Today, ICME approaches tightly integrate physics-based computational models
into the industrial design process, allowing the desired performance requirements of a part to
guide the design of a material. Alloy specific examples include low-Rare Earth
Ni superalloys for better turbine performance [12] and lower cost and radioactive element free
Ferrium S53 alloy designed for corrosion-resistant landing gears [13]. Both cases reduced
materials innovation timelines from decades to years, demonstrating the practical capability of
designing and qualifying new materials within an industrial product development cycle.
Generalizing and accelerating this capability across different industries and materials is a
primary goal of the Materials Genome Initiative (MGI) [14].
Predicting PSPP linkages in metals AM is difficult with existing physics-based ICME
approaches. The physics of AM processes are more complex than traditional fabrication
methods, like casting, as they involve rapid solidification, vaporization and ingestion of volatile
elements, and complex thermal history that consists of dozens of heating and cooling cycles,
each one different. Furthermore, all of these additional complexities vary from one location to
another within a part, and from part to part within a build volume. For AM, physics-based ICME
tools have been mostly developed through attempts to adopt legacy manufacturing models to
AM data, with some success. However, today’s relatively low cost and time for performing AM
processing experiments has led to metals AM development being largely combinatorial, with a
chief strategy of adopting AM processing to legacy alloys that were developed for other types of
manufacturing using extensive design of experiments.
It is with awareness of the large amounts of data being generated in AM through these
combinatorial development cycles that machine learning (ML) has been targeted to accelerate
AM innovations and their commercialization. Machine learning as a technology development
accelerator has shown wide application in recent years across fields including finance [15],
molecule design for genomics, chemistry and pharmacology [16], social networking [17] and,
most relevant to this review, materials science and engineering [18], [19], [20]. Still, the use of
ML in materials science was relatively limited for a variety of reasons, especially the lack of
large curated datasets amenable to existing ML methodologies. Through the work done under the
MGI, this data limitation was identified as a primary impediment to future materials
innovations [14]. In response, there has been significant recent investment in materials database
developments to better enable materials data informatics innovations. It is now recognized and
accepted that ML frameworks can couple legacy physics-based ICME tools with experimental
data to produce more accurate process-structure-property models and to automate the iteration of
designed experiments for model improvement and optimized
materials [21], [22], [20], [23], [24].
We proceed to review how the paradigm shift from purely physics-based to coupled physics-
based/data-driven ICME approaches can be made through solving metals AM challenges. We
begin by phrasing terms and ideas from AM in ways that are compatible with machine learning.
We provide a basic review of machine learning algorithms and how they can be applied to
additive manufacturing. Following this introduction to using ML for AM problems, we review
other uses of machine learning in materials science and engineering and state the uses of such
approaches for solving AM challenges.
Machine learning (ML) has shown to be an effective alternative to physical models for quality
prediction and process optimization of metal additive manufacturing (AM). However, the
inherent “black box” nature of ML techniques such as those represented by artificial neural
networks has often presented a challenge to interpret ML outcomes in the framework of the
complex thermodynamics that govern AM. While the practical benefits of ML provide an
adequate justification, its utility as a reliable modeling tool is ultimately reliant on assured
consistency with physical principles and model transparency. To facilitate the fundamental
needs, physics-informed machine learning (PIML) has emerged as a hybrid machine learning
paradigm that imbues ML models with physical domain knowledge such as thermomechanical
laws and constraints. The distinguishing feature of PIML is the synergistic integration of data-
driven methods that reflect system dynamics in real-time with the governing physics underlying
AM. In this paper, the current state-of-the-art in metal AM is reviewed and opportunities for a
paradigm shift to PIML are discussed, thereby identifying relevant future research directions.
Additive manufacturing (AM) is well-recognized as a fast, flexible, and eco-friendly [1]
technology for manufacturing metal parts. Metal AM processes, e.g., direct energy deposition
(DED) and powder bed fusion (PBF), have been applied to aerospace, medicine, health care, and
other fields [2] to produce high-value, customized products. Accompanying these industrial
needs is a high standard for part quality, which requires effective monitoring of the metal AM
processes to optimize process parameters such that the parts built will exhibit minimal defects
and meet performance expectations. Toward this end, extensive efforts have been made in
optimizing AM processes through modeling and simulations in a digital environment before
experiments are conducted to minimize material costs associated with trial-and-error while
maximizing process performance. These efforts can be categorized into two families: physics-
based and data-driven methods. Physics-based methods present process models that capture the
physical principles underlying AM processes by using either analytical or numerical methods to
emulate physical behavior at varying scales, e.g., atomic, microscale, or macroscopic. They are
established either thermomechanically, based on process physics, or empirically through
experimental findings. As a result, they can explain process mechanics and provide explicit
insights into process phenomena, thereby providing guidance on minimizing AM process
variabilities, part deficiencies, and improving overall process optimization [3, 4]. However, due
to model simplifications that are necessary to reduce model derivation complexity, physics-based
models may be limited in comprehensively covering the entire spectrum of process variability.
Specifically for physics-based models that are numerical and iterative in nature, e.g., finite
element models, simulation has exponential time and memory complexity as a function of both
the resolution and the number of elements to be simulated. As a result, this category of physical
methods is infeasible for in-situ real-time usage and oftentimes impractical in the fast-paced
development environment of industrial AM. Data-driven methods such as machine learning
(ML) algorithms are capable of harnessing data of high dimensionality and heterogeneity and
leveraging the full spectrum of process parameters beyond those incorporated in physical
models. Complemented by sensing data measured during the AM process in real-time, such
methods can effectively complement the physical understanding of the process by automatically
and continually updating themselves through learning from the data and experience [5, 6]. These
characteristics imply several intrinsic advantages of ML as a means for metal AM process
modeling and monitoring. First, ML models can be computationally Page 3 of 40 efficient once
properly trained [7] thus providing a good fit for analyzing dynamic thermomechanical
phenomena [8] in metal AM processes. Second, ML methods do not pose restrictive assumptions
on the processes being analyzed [9-11], and are thus adaptable to a broad range of process
variability. Third, data-driven models are renewable and generalizable [12] as the basic structure
of pretrained models can be retained and updated using new data before being transferred to
other processes that are suited for analysis using the updated ML models [13, 14]. Finally, ML
can be leveraged to solve a wide variety of auxiliary problems in AM that are difficult to solve
using traditional approaches, including cost estimation [15], manufacturability assessment [16,
17], and closed-loop quality control [18]. The above advantages of data-driven models have been
well-recognized by the research community [19], and considerable efforts have been made to
leverage ML for metal AM studies. Examples include Khanzadeh et al. [20], Khanzadeh et al.
[21], and Guo et al. [22] that analyzed in-situ thermal images of melt pools from DED with ML
models and developed techniques for real-time prediction of porosity. Similarly, Shevchik et al.
[23] and Wasmer et al. [24] demonstrated the feasibility of training ML-based quality prediction
models with acoustic emissions from metal AM processes. Closed-loop control systems for
metal AM were enabled by ML in Jafari-Marandi et al. [25], Liu et al. [26], and Renken et al.
[27], which connected the design and printing/finishing stages in metal AM and achieved reverse
process parameter optimization. These studies have facilitated improvements in quality and
design for metal AM and expanded the scope of ML. While the effectiveness of ML-based metal
AM process modeling and control has been demonstrated, and more advances in ML-based
methods are foreseeable [19], a major hurdle in the widespread acceptance of ML-based methods
is their lack of physical interpretability. ML models’ understanding of the underlying physics in
metal AM is constrained by the physical meaningfulness of the input/output data as well as by
the model architecture itself. Since datasets collected from diverse applications may not fully
represent the complete AM process physics, how to intuitively enhance ML models’ awareness
of the underlying process physics toward full model interpretability has remained a challenge.
Further challenges due to the lack of physical intuition in model construction include impractical
model prediction results [28-30] and susceptibility to data pollution caused by noise, missing or
incorrect data labels, etc. An emerging topic in imbuing ML models with physical intuition is
Physics-Informed Machine Learning (PIML), which consists of hybrid methods that incorporate
physical knowledge and domain constraints into ML to achieve interpretable model design,
input, and output. Although still in its infancy at the current stage, the PIML paradigm has
already attracted an increasing level of attention due to its potential for future exploration. This
paper presents a systematic review of the state-of-the-art in ML-based metal AM process
modeling and control by summarizing common methods developed in this field and revealing
open questions and challenges. By highlighting of the potential of PIML, the review further aims
to provide a roadmap to researchers in both the AM and ML communities for future synergy that
enables PIML-enabled, “smart” metal AM applications. This paper distinguishes itself from
other AM and ML review articles in the published literature [8, 30-33] by providing the above-
described roadmap and constitutes the first work in PIML for AM to guide the reader through the
foundations of metal AM, identify outstanding challenges, and propose a comprehensive solution
based on PIML. Additionally, the content of the paper establishes a foundation to further develop
AM using physics, which is a research gap identified by previous works [8, 30, 31]. Page 4 of 40
The rest of this paper is organized as follows. Section 2 briefly explains the mechanisms of metal
AM and introduces major, representative AM processes. Section 3 reviews the state of ML-based
metal AM process modeling and monitoring techniques. Section 4 introduce PIML as a potential
enabler for higher level of effectiveness and efficiency in AM process modeling and control,
followed by Section 5 where a vision for PIML-based metal AM is presented. Section 6
summarizes findings from the paper with a conclusion. Freedom of design, mass customisation,
waste minimisation and the ability to manufacture complex structures, as well as fast
prototyping, are the main benefits of additive manufacturing (AM) or 3D printing. A
comprehensive review of the main 3D printing methods, materials and their development in
trending applications was carried out. In particular, the revolutionary applications of AM in
biomedical, aerospace, buildings and protective structures were discussed. The current state of
materials development, including metal alloys, polymer composites, ceramics and concrete, was
presented. In addition, this paper discussed the main processing challenges with void
formation, anisotropic behaviour, the limitation of computer design and layer-by-layer
appearance. Overall, this paper gives an overview of 3D printing, including a survey on its
benefits and drawbacks as a benchmark for future research and development. 3-D printing is an
additive manufacturing (AM) technique for fabricating a wide range of structures and complex
geometries from three-dimensional (3D) model data. The process consists of printing successive
layers of materials that are formed on top of each other. This technology has been developed by
Charles Hull in 1986 in a process known as stereolithography (SLA), which was followed by
subsequent developments such as powder bed fusion, fused deposition modelling (FDM), inkjet
printing and contour crafting (CC). 3D-printing, which involves various methods, materials and
equipment, has evolved over the years and has the ability to transform manufacturing and
logistics processes. Additive manufacturing has been widely applied in different industries,
including construction, prototyping and biomechanical. The uptake of 3D printing in the
construction industry, in particular, was very slow and limited despite the advantages e.g. less
waste, freedom of design and automation.

New applications are emerging as novel materials and AM methods are continuously being
developed. One of the main drivers for this technology to become more accessible is attributed to
the expiry of earlier patents, which has given manufacturers the ability to develop new 3D
printing devices. Recent developments have reduced the cost of 3D printers, thereby expanding
its applications in schools, homes, libraries and laboratories. Initially, 3D printing has been
extensively used by architects and designers to produce aesthetic and functional prototypes due
to its rapid and cost-effective prototyping capability. The use of 3D printing has minimised the
additional expenses that are incurred in the process of developing a product. However, it is only
in the past few years that 3D printing has been fully utilised in various industries from prototypes
to products. Product customisation has been a challenge for manufacturers due to the high costs
of producing custom-tailored products for end-users. On the other hand, AM is able to 3D print
small quantities of customised products with relatively low costs. This is specifically useful in
the biomedical field, whereby unique patient-customised products are typically required.
Customised functional products are currently becoming the trend in 3D printing as predicted by
Wohlers Associates, who envisioned that about 50% of 3D printing will revolve around the
manufacturing of commercial products in 2020 [1]. This technology has gained the attention of
those in the medical field, due to its ability to produce a wide variety of medical implants from
CT-imaged tissue replicas [2]. More recently, 3D printing is effectively being used in the
construction industry. A group of relatively cheap houses in China ($4800 USD per unit) were
successfully mass printed by WinSun in less than a day [3].

The growing consensus of adapting the 3D manufacturing system over traditional techniques is
attributed to several advantages including fabrication of complex geometry with high precision,
maximum material savings, flexibility in design, and personal customisation. A wide range of
materials that are currently used in 3D printing include metals, polymers, ceramics and concrete.
Polylactic acid (PLA) and acrylonitrile butadiene styrene (ABS) are the main polymers used in
the 3D printing of composites. Advanced metals and alloys are typically utilised in the aerospace
sector because traditional processes are more time-consuming, difficult and costly. Ceramics are
mainly used in 3D-printed scaffolds and concrete is the main material employed in the additive
manufacturing of buildings. However, the inferior mechanical properties and anisotropic
behaviour of 3D printed parts still limit the potential of large-scale printing. Therefore, an
optimised pattern of 3D priming is important to control flaw sensitivity and anisotropic
behaviour. Also, changes in the printing environment have an influence on the quality of finished
products [4]. AM is capable of fabricating parts of various sizes from the micro-to macro-scale.
However, the precision of the printed parts is dependent on the accuracy of the employed method
and the scale of printing. For instance, micro-scale 3D printing poses challenges with the
resolution, surface finish and layer bonding, which sometimes require post-processing techniques
such as sintering [5]. On the other hand, the limited materials available for 3D printing pose
challenges in utilising this technology in various industries. Hence, there is a need for developing
suitable materials that can be used for 3D printing. Further developments are also needed to
enhance the mechanical properties of 3D printed parts.

The advantages of 3D printing technology will continue to emerge through continuing research
efforts, which must be undertaken to understand and eliminate constraints that inhibit the use of
this technology. Design tools to assess life-cycle costs i.e., AM-oriented computer-aided design
(CAD) systems with more user-friendly and advanced simulation capabilities are some of the
key aspects that need to be realised. A distinguished advantage of 3D printing is mass
customisation i.e., production of a series of personalised goods such that each product can be
different while maintaining a low price due to mass production. 3D printing is devoid of the
added cost due to mould making and tooling for a customised product. Therefore, mass
production of a number of identical parts can be as cost-effective as the same number of different
personalised goods. The change between different designs is straightforward with negligible
added cost and no need for special preparation. AM also has the potential for mass production of
complex geometries such as lattice structures, where the application of traditional methods of
manufacturing such as casting is not straightforward and require further time-consuming tooling
and post-processing. However, improvements in the fabrication speed and cost reduction must be
resolved through the improvement of machine design. Also, the high costs and time-consumption
of the AM process remain to be major hurdles that inhibit mass production.
This paper aims to provide a comprehensive review of 3D printing techniques in terms of the
main methods employed, materials utilised, its current state and applications in various
industries. The paper will also present research gaps and challenges encountered in adopting this
technology.

Metal additive manufacturing has seen extensive research and rapidly growing applications for
its high precision, efficiency, flexibility, etc. However, the appealing advantages are still far from
being fully exploited, and the bottleneck problems essentially originate from the incomplete
understanding of the complex physical mechanisms spanning from the manufacturing
processes, microstructure evolutions, to mechanical properties. Specifically, for powder-fusion-
based additive manufacturing such as laser powder bed fusion, the manufacturing process
involves powder dynamics, heat transfer, phase transitions (melting, solidification, evaporation,
and condensation), fluid flow (gas, vapor, and molten metal liquid), and their interactions. These
interactions induce not only various defects but also complex thermal-mechanical-compositional
conditions. These transient conditions lead to highly non-equilibrium microstructure evolutions,
and the resultant microstructures, together with those defects, can significantly alter the
mechanical properties of the as-built parts, including strength, ductility and residual stress. We
believe that the most efficient approach to advance the fundamental understanding is
integrating in-situ experimentation and high-fidelity modeling. In this review, we summarize the
state of the art of these two powerful tools: in-situ synchrotron experimentation and high-fidelity
modeling, and provide an outlook for potential research directions. Metal additive
manufacturing has emerged as a promising alternative to conventional manufacturing in various
industries, offering advantages like complex geometries and reduced material waste. However,
there is a research gap in the form of comparative and industry-specific investigations and case
studies to comprehensively assess the environmental and economic impact of metal additive
manufacturing.
This study focuses on two significant sectors, industrial machinery and aeronautics, to
understand how metal additive manufacturing influences environmental and economic
sustainability compared to conventional manufacturing. Therefore, taking a life cycle approach,
the environmental and economic impacts of this technology are evaluated and compared to
conventional manufacturing through representative case studies in the two sectors using life
cycle assessment and life cycle cost methodologies.
The economic evaluation reveals that metal additive manufacturing reduces the total cost for
aeronautical components by 33.2 %, even considering several post-processing operations. While
for industrial machinery parts, this cost increases by 79.3 % due to high machine and material
cost. However, this technology reduces the potential environmental impact by more than 60 %
for both sectors, mainly due to reduced material consumption. The use phase for aeronautical
components further contributes to environmental and economic benefits, with significant fuel
and CO2 emissions savings. The results emphasize the sector-dependent nature of metal additive
manufacturing's environmental and economic impact and underscore the importance of a life
cycle perspective for informed decision-making.
Some critical points for sustainable adoption of metal additive manufacturing in each sector are
equipment investment, the post-processing applied, the material selected or, for the aeronautical
industry, the impact of the use phase. Nevertheless, future developments are needed to address
powder reuse, optimization, certification, and the social impact of metal additive manufacturing.
In recent years, there is a trend for replacing conventional manufacturing (CM) processes with
additive manufacturing (AM) ones. While traditional production technologies such as machining
and forging have long been the established methods for fabricating components in various
industries, there is a growing reliance on AM as a driver for innovation (Tavares et al.,
2023; Attaran, 2017). Likewise, the scientific community has followed this trend in parallel
(Jemghili et al., 2021) with an increasing investment worldwide, reflected in AM's global market
value, which is expected to grow from 11 billion dollars in 2019 to 35 billion dollars by 2024
(Qin et al., 2022), representing a compound annual growth rate of around 26 %.
This burgeoning attentiveness can be explained by the multiple advantages resulting from
producing components layer by layer, ranging from cost-effectiveness for low-volume
productions (Ford and Despeisse, 2016) and reduced material waste (Gao et al., 2015) to a
potentially improved supply chain performance (Manco et al., 2023) and the ability to create
more complex geometries that, among other aspects, enable lightweight components
(Kunovjanek et al., 2020). These capabilities have captured the attention of manufacturers and
companies in a wide range of sectors, such as aerospace (Blakey-Milner et al., 2021), medical
(Salmi, 2021), automotive (Böckin and Tillman, 2019a) or industrial machinery (Colosimo et al.,
2020). Notably, AM found applications in a variety of materials, including metals, ceramics,
plastics, and composites (Tofail et al., 2018). Consequently, when addressing the impact of this
technology, it is essential to consider these diverse applications and their sector-specific
implications.
The adoption of AM holds considerable potential to influence economies and societies in various
areas (Caviggioli and Ughetto, 2019). Notably, the economic factor has emerged as a prominent
driver behind AM implementation across industrial sectors, surpassing social and environmental
considerations (Niaki et al., 2019). Consequently, a comprehensive understanding and analysis
of the main economic impacts are essential, as this will help comprehend the primary economic
motives driving AM adoption. Additionally, a holistic analysis of the remaining sustainability
dimensions is crucial to uncover overlooked opportunities that manufacturers may not have
considered. Thus, this study will primarily focus on exploring the environmental and economic
impacts of AM while acknowledging the current limitations in understanding its social impact.
Prior research on AM has explored both environmental and economic perspectives by adopting
general approaches focused on the role of this technology in industrial sustainability (Ford and
Despeisse, 2016) and supply chains (Kunovjanek et al., 2020). Nevertheless, since each sector
has unique characteristics, the adoption of AM is context-dependent (Vafadar et al., 2021). Thus,
some studies had focused on evaluating the adoption of AM for case studies from specific
industries by adopting the life cycle assessment (LCA) methodology (Le and Paris, 2018; Lyons
et al., 2021; Peng et al., 2020), life cycle cost (LCC) (Laureijs et al., 2017; Ribeiro et al., 2020)
or even combining both (Stieberova et al., 2022; Gouveia et al., 2022; Kokare et al., 2023).
While others look at sectors like aerospace focusing on energy and material efficiency (Monteiro
et al., 2022), or automotive emphasizing the environmental pillar (Böckin and Tillman, 2019b).
Despite this, further research is needed focusing on specific industry-oriented investigations
(Sanguineti et al., 2023) and conducting in-depth case studies across various sectors (Niaki et al.,
2019).
Furthermore, few studies have integrated and compared comprehensively different sectors that
differ in multiple aspects, such as operational and regulatory requirements as well as relevant
product life cycle phases. Previous research (Ford and Despeisse, 2016; Agnusdei and Del Prete,
2022) had already stated the need for conducting comparative and industry-specific evaluations
for AM to effectively identify and pursue opportunities for enhancing sustainability. In addition,
these analyses should be performed by considering a life cycle perspective, which is often
overlooked in some studies, despite being crucial for understanding the full range of
environmental and economic implications of AM (Ribeiro et al., 2020).
To address these gaps, this paper presents a comprehensive analysis of the environmental and
economic sustainability impact of metal additive manufacturing (MAM) in two distinct sectors:
industrial machinery and aeronautics. These sectors have been selected due to their
high technological readiness level in MAM (Colosimo et al., 2020). The industrial machinery
sector encompasses various equipment types with different applications, such as machinery
plastic transformation or pulp and paper manufacturing. Unlike sectors like aerospace, industrial
machinery faces fewer regulatory restrictions, making MAM a viable alternative to replace CM
(Cardeal et al., 2023). Exceptions may exist, for example, to produce industrial machinery for the
food sector (Lipton et al., 2015). Conversely, the aerospace sector, of which the aeronautical
industry is a part, is an early adopter of AM (Khorasani et al., 2022) and is expected to
experience substantial growth in the coming years due to increased military spending and the
demand for new commercial aircraft (Blakey-Milner et al., 2021).
Therefore, this research aims to understand how MAM differently impacts environmental and
economic sustainability in both sectors, compared to CM.

To achieve this, the paper begins by providing an overview of the main applications of MAM in
both the industrial machinery and aeronautical sectors. By doing so, this study aims to identify
and examine the commonalities and differences in MAM adoption. This summary will serve as a
basis to identify the specific environmental and economic impacts of MAM within these sectors
across the life cycle of a product application.

Given that the impacts identified are primarily qualitative, a quantitative approach is imperative
to gain clearer insights into the critical factors influencing MAM adoption in each sector
regarding environmental and economic sustainability performance. To this end, representative
case studies have been selected for each sector, delving into real-world industrial applications of
MAM, namely on the prominent and high-precision AM technology for metals known as
laser powder bed fusion (L-PBF) (Chowdhury et al., 2022). These case studies involve the
evaluation of the economic sustainability implications using comprehensive LCC analysis,
employing process-based cost models and investment appraisal methods. Additionally,
environmental evaluations are performed using the LCA methodology to comprehensively assess
the environmental impacts associated with MAM throughout the product life cycle. Factors such
as energy and fuel consumption, raw material usage, emissions, and waste generation are
analysed to gain insights into the environmental implications of MAM in the industrial
machinery and aeronautical sectors. By quantifying some of the impacts previously identified in
a specific and tangible manner through representative case studies, this research enables
informed decision-making regarding the implementation of MAM and sustainable practices in
the industrial machinery and aeronautical sectors.
This work contributes to the current body of knowledge on the environmental and economic
sustainability of MAM by adopting an industry-dependent analysis which provides valuable
insights for industrial stakeholders, researchers, and policymakers, promoting an understanding
of the critical points for sustainable implementation in each sector while highlighting the
commonalities and differences concerning two pivotal sectors within MAM.

The sustainability of the fused-deposition modeling process and its optimization is studied in this
paper using a method in which consideration is simultaneously given to the design of the
structural component and its 3D-printing process. The proposed method is used to determine the
mechanical properties of the deposited materials, through an experimental study that follows an
experimental design with variable printing parameters. Having recorded the values of the
material properties, a topological optimization of each experimental point is performed. Then,
the printing processes are designed with the help of a slicing software package and, based on the
results, the corresponding key sustainability performance indicators are computed on the basis of
the printing parameters. Finally, an optimization process is conducted and the most convenient
solution is chosen. Validated using a real case study, the effectiveness of the proposed approach
is demonstrated for determining the most suitable choice when design and manufacturing are
simultaneously considered through the lens of sustainability. Sustainability-driven optimization
assessment and its integration criteria for manufacturing processes are gaining immense
academic, industrial, and social interest from technical, economic and environmental
perspectives, due to environmental concerns over rising carbon dioxide emissions, energy waste,
and high carbon footprints of several conventional manufacturing processes [1], [2], [3], [4], [5],
[6], [7]. Interest is centered on improving the effectiveness of machines, materials and tool life-
cycle indicators in such areas as: stochastic machining process variables (e.g. material removal
rate and tool wear), surface quality, environmental friendliness (e.g. specific cutting energy, and
oil consumption), machining costs, and carbon emissions.
Trends towards sustainability-oriented technology awareness and circular product life cycles
have been noted in studies reported in [8], [9]. The focus is on various manufacturing industry
methods, such as: Life Cycle Analysis (LCA) [10], [11], [12], eco-design [13], [14], and eco-
indicators [15] where the analysis of materials, energy sources, and emissions among other green
footprint indicators are monitored at every stage (design, manufacturing, useful life and post-use
or recycling) of the production value chain, in terms of 29 environmental impacts [16].

In particular, Additive Manufacturing (AM) is an emergent technology in Industry 4.0. leading

industrial development towards personalized intelligent production [17], [18]. Additive
manufacturing implies mass customization, flexible production, and on-demand and
decentralized manufacturing [19]. Among the different additive manufacturing technologies,
Fused-Deposition Modeling (FDM) stands out, because it provides low-cost solutions for
different applications from industrial prototyping to (do-it-yourself) home manufacturing of parts
[20].

An important set of works has been focused on analyzing the environmental impact of AM and
on how to decrease its impact [21]. A critical review of 3D-printing technologies and their
advantages was presented in [22], to highlight the advantages of AM technologies as sustainable
and environmentally friendly alternatives to conventional manufacturing technologies for
greener smart-manufacturing applications. Nevertheless, several improvements are yet to be
addressed in the manufacturing industry, which if developed will promote the use of fully natural
materials with design properties and complex geometries that can open further perspectives in
many fields, such as sensors, bio-medical applications, and soft robotics [23].

Besides the assessment methods and indicators, additive manufacturing sustainability can be
improved through two main approaches. In the first place, component design can be optimized to
increase the economic profit and to decrease the environmental footprint [24]. There are several
techniques that work towards this target, including Taguchi robust design [25], response surface
methodology [26] and bio-inspired optimization heuristics [27], [28]. Nevertheless, topology
optimization is the most popular approach [29], [30], [31], [32], for the reduction of material
mass, with the consequent decrease in the environmental impact [33], and economic benefits
[34], [35], [36]. Topological optimization is a very suitable technique for designing 3D-printed
parts, especial when combined with fused-deposition modeling that is capable of manufacturing
complex-shaped parts [37]. In second place, important sustainability improvements can be also
achieved by optimizing the printing parameters, such as layer height [38], [39], [40], [41], [25],
[42], [26], fill density [39], [38], [41], [25], [26], fill orientation [39], [40], [25], [26], printing
speed [39], [41], [42], printing temperature [39], [38], bed temperature [42], and nozzle diameter
[42].

The main concern over optimizing part design and printing parameters in AM is that both
functions are not independent of each other. It has been widely proven that printing parameters
have significant influence on the mechanical properties of the 3D-printed component [43], [39],
[40]. Therefore, both functions (i.e., component design and printing parameters) should be
simultaneously considered, in order to obtain an actual optimization of process sustainability. In
spite of certain antecedents [44], to the best of our knowledge there is no formalized technique
for solving this problem. Thus, our purpose in this study is to formalize an approach for
concurrently optimizing structural design and printing parameters in fused deposition modeling,
which will be illustrated and validated through a study case.
The paper is organized into five sections. After this introduction, the proposed approach is
presented and explained. Then, each step of the case study is detailed in the third section. The
fourth section is on the analysis of the study cases and their results and, finally, the concluding
remarks and some directions for future work are highlighted in the last section.
EXISTING SYSTEM
Existing fault detection systems using machine learning techniques have made significant
progress in improving the accuracy and reliability of fault detection in a wide range of
applications, including industrial manufacturing, power generation and transmission, aerospace,
and healthcare.Machine learning models can be trained to achieve high levels of accuracy in
detecting faults, even in the presence of noise and uncertainty.Machine learning-based fault
detection systems can help to reduce the costs associated with downtime, maintenance, and
repairs.Machine learning-based fault detection systems can help to improve the safety of systems
and machines by detecting faults early on, before they can lead to catastrophic failures.These
systems use machine learning to monitor the vibration data of machines to detect faults such as
bearing wear and misalignment.These systems use machine learning to monitor the sensor
readings of industrial processes to detect faults such as leaks and clogs. These systems use
machine learning to monitor the performance of power generators to detect faults such as
winding failures and bearing wear. These systems use machine learning to monitor the condition
of transmission lines to detect faults such as insulation breakdown and conductor
corrosion. These systems use machine learning to monitor the performance of aircraft engines to
detect faults such as compressor stall and turbine blade failure.These systems use machine
learning to monitor the performance of aircraft avionics systems to detect faults such as
navigation sensor failure and communication system failure.These systems use machine learning
to monitor the performance of medical devices, such as MRI machines and CT scanners, to
detect faults such as software bugs and hardware failures.These systems use machine learning to
monitor the vital signs of patients to detect anomalies that may indicate a medical
emergency.While existing fault detection systems using machine learning techniques have made
significant progress, there are still some challenges that need to be addressed. For example,
machine learning models can be computationally expensive to train and deploy, and they may
require large amounts of data to achieve high levels of accuracy. Additionally, machine learning
models can be vulnerable to adversarial attacks, where attackers intentionally craft inputs to the
model in order to cause it to make incorrect predictions.Researchers and practitioners are
actively working on developing new and innovative machine learning models and algorithms to
address these challenges and improve the accuracy, reliability, and security of fault detection
systems. As machine learning technology continues to develop, we can expect to see even more
 widespread adoption of fault detection systems using machine learning techniques in a wide
range of industries.

DISADVANTAGES

Data requirements: Machine learning models require large amounts of data to train, which can
be difficult and expensive to collect, especially for complex systems.

 Explainability: Machine learning models can be difficult to explain, which can make it
challenging to understand why the model makes certain predictions. This can be problematic in
safety-critical applications, where it is important to be able to explain why a fault was detected.

 Robustness: Machine learning models can be vulnerable to noise and adversarial attacks, which
can lead to false positives and false negatives.

 Computational cost: Training and deploying machine learning models can be computationally
expensive, especially for complex models.
 CONCEPTUALIZED SYSTEM

The system would use a variety of techniques to reduce the amount of data required for
training, such as transfer learning and data augmentation.The system would use a variety of
techniques to explain the predictions of the machine learning model, such as concept activation
vectors and Shapley values.The system would use a variety of techniques to make the machine
learning model more robust to noise and adversarial attacks, such as adversarial training and
robust optimization.The system would use a variety of techniques to reduce the computational
cost of training and deploying the machine learning model, such as model compression and
quantization.The system would be scalable to handle large numbers of machines and sensors.The
system would be able to detect faults in real time, so that corrective action can be taken
immediately.The system would be able to adapt to changes in the operating conditions of the
machines and sensors.

Here is a high-level overview of the conceptualized fault detection system:

1. The system would collect data from a variety of sensors, such as vibration sensors, temperature
sensors, and current sensors.

2. The system would pre-process the data to clean it and extract the most relevant features.

3. The system would train a machine learning model on the pre-processed data.

4. The system would deploy the trained machine learning model to production.

5. The system would collect data from the sensors in real time and feed it to the machine learning
model.

6. The machine learning model would predict the probability of a fault occurring.

7. If the predicted probability of a fault is above a certain threshold, the system would alert the
operator.
The conceptualized fault detection system would be a valuable tool for a wide range of
industries, including industrial manufacturing, power generation and transmission, aerospace,
and healthcare. By addressing the disadvantages of existing systems and incorporating the
features listed above, the conceptualized system would help to improve the accuracy, reliability,
and safety of fault detection.

Here are some specific examples of how the conceptualized fault detection system could be used
in different domains The system could be used to detect faults in manufacturing equipment, such
as robots, CNC machines, and assembly lines. This would help to reduce downtime and improve
product quality.The system could be used to detect faults in power generators, transformers, and
transmission lines. This would help to prevent blackouts and ensure the reliability of the power
grid.The system could be used to detect faults in aircraft engines, avionics, and other critical
systems. This would help to ensure the safety of passengers and crew.The system could be used
to detect faults in medical devices, such as MRI machines and CT scanners. This would help to
prevent patient harm.The conceptualized fault detection system is still in the early stages of
development, but it has the potential to revolutionize the way that faults are detected in a wide
range of industries.
ADVANTAGES

Increased accuracy and reliability: Machine learning models can be trained to achieve high
levels of accuracy and reliability in detecting faults, even in the presence of noise and
uncertainty. Conceptualized fault detection systems using machine learning techniques could
leverage this capability to improve the accuracy and reliability of fault detection over existing
systems.

Reduced downtime and maintenance costs: By detecting faults early on, before they can lead
to catastrophic failures, conceptualized fault detection systems using machine learning
techniques could help to reduce downtime and maintenance costs. This could lead to significant
financial savings for businesses.

Improved safety: Conceptualized fault detection systems using machine learning techniques
could help to improve safety by detecting faults in critical systems early on. For example, in the
aerospace industry, such systems could help to detect faults in aircraft engines before they can
lead to a disaster.

Scalability and real-time detection: Conceptualized fault detection systems using machine
learning techniques could be scaled to handle large numbers of machines and sensors.
Additionally, these systems could be designed to detect faults in real time, allowing for
corrective action to be taken immediately.

Adaptability: Conceptualized fault detection systems using machine learning techniques could
be designed to adapt to changes in the operating conditions of the machines and sensors being
monitored. This would help to ensure that the systems continue to be effective in detecting faults
even as conditions change.

Reduced data requirements: Conceptualized fault detection systems using machine learning
techniques could be designed to use a variety of techniques to reduce the amount of data required
for training, such as transfer learning and data augmentation. This would make it easier to deploy
and maintain these systems in real-world settings.

Explainability: Conceptualized fault detection systems using machine learning techniques could
be designed to use a variety of techniques to explain the predictions of the machine learning
model. This would help to improve the transparency and trustworthiness of these systems.

Robustness: Conceptualized fault detection systems using machine learning techniques could be
designed to be more robust to noise and adversarial attacks. This would help to ensure that the
systems continue to be effective in detecting faults even in challenging environments.
SYSTEM REQUIREMENTS

➢ H/W System Configuration:-

➢ Processor - Pentium –IV

➢ RAM - 4 GB (min)

➢ Hard Disk - 20 GB

➢ Key Board - Standard Windows Keyboard

➢ Mouse - Two or Three Button Mouse

➢ Monitor - SVGA

SOFTWARE REQUIREMENTS:

 Operating system : Windows 7 Ultimate.

 Coding Language : Python.

 Front-End : Python.

 Back-End : Django-ORM

 Designing : Html, css, javascript.

 Data Base : MySQL (WAMP Server

IMPLEMENTATION

MODULES

SERVICE PROVIDER

In this module, the Service Provider has to login by using valid user name and password. After
login successful he can do some operations such as

Login, Browse Water Data Sets and Train & Test, View Trained and Tested Water Data Sets
Accuracy in Bar Chart, View Trained and Tested Water Data Sets Accuracy Results, View
Predicted Water Quality Detection Type,

Find Water Quality Detection Type Ratio, Download Predicted Data Sets,

View Water Quality Detection Ratio Results, View All Remote Users.

VIEW AND AUTHORIZE USERS

In this module, the admin can view the list of users who all registered. In this, the admin can
view the user’s details such as, user name, email, address and admin authorizes the users.

REMOTE USER

In this module, there are n numbers of users are present. User should register before doing any
operations. Once user registers, their details will be stored to the database. After registration
successful, he has to login by using authorized user name and password. Once Login is
successful user will do some operations like REGISTER AND LOGIN, PREDICT WATER QUALITY
DETECTION TYPE, VIEW YOUR PROFILE.
SYSTEM DESIGN
ALGORITHMS
Decision tree classifiers

Decision tree classifiers are used successfully in many diverse areas. Their most important
feature is the capability of capturing descriptive decision making knowledge from the supplied
data. Decision tree can be generated from training sets. The procedure for such generation
based on the set of objects (S), each belonging to one of the classes C1, C2, …, Ck is as follows:

Step 1. If all the objects in S belong to the same class, for example Ci, the decision tree for S
consists of a leaf labeled with this class

Step 2. Otherwise, let T be some test with possible outcomes O1, O2,…, On. Each object in S has
one outcome for T so the test partitions S into subsets S1, S2,… Sn where each object in Si has
outcome Oi for T. T becomes the root of the decision tree and for each outcome Oi we build a
subsidiary decision tree by invoking the same procedure recursively on the set Si.

Gradient boosting

Gradient boosting is a machine learning technique used in regression and classification tasks,
among others. It gives a prediction model in the form of an ensemble of weak prediction
models, which are typically decision trees.[1][2] When a decision tree is the weak learner, the
resulting algorithm is called gradient-boosted trees; it usually outperforms random forest.A
gradient-boosted trees model is built in a stage-wise fashion as in other boosting methods, but
it generalizes the other methods by allowing optimization of an arbitrary differentiable loss
function.
K-Nearest Neighbors (KNN)

 Simple, but a very powerful classification algorithm

 Classifies based on a similarity measure

 Non-parametric
 Lazy learning
 Does not “learn” until the test example is given

 Whenever we have a new data to classify, we find its K-nearest neighbors from the
training data

Example

 Training dataset consists of k-closest examples in feature space

 Feature space means, space with categorization variables (non-metric variables)
 Learning based on instances, and thus also works lazily because instance close to the
input vector for test or prediction may take time to occur in the training dataset

Logistic regression Classifiers

Logistic regression analysis studies the association between a categorical dependent variable
and a set of independent (explanatory) variables. The name logistic regression is used when the
dependent variable has only two values, such as 0 and 1 or Yes and No. The name multinomial
logistic regression is usually reserved for the case when the dependent variable has three or
more unique values, such as Married, Single, Divorced, or Widowed. Although the type of data
used for the dependent variable is different from that of multiple regression, the practical use
of the procedure is similar.

Logistic regression competes with discriminant analysis as a method for analyzing categorical-
response variables. Many statisticians feel that logistic regression is more versatile and better
suited for modeling most situations than is discriminant analysis. This is because logistic
regression does not assume that the independent variables are normally distributed, as
discriminant analysis does.

This program computes binary logistic regression and multinomial logistic regression on both
numeric and categorical independent variables. It reports on the regression equation as well as
the goodness of fit, odds ratios, confidence limits, likelihood, and deviance. It performs a
comprehensive residual analysis including diagnostic residual reports and plots. It can perform
an independent variable subset selection search, looking for the best regression model with the
fewest independent variables. It provides confidence intervals on predicted values and provides
ROC curves to help determine the best cutoff point for classification. It allows you to validate
your results by automatically classifying rows that are not used during the analysis.

Naïve Bayes

The naive bayes approach is a supervised learning method which is based on a simplistic
hypothesis: it assumes that the presence (or absence) of a particular feature of a class is
unrelated to the presence (or absence) of any other feature .

Yet, despite this, it appears robust and efficient. Its performance is comparable to other
supervised learning techniques. Various reasons have been advanced in the literature. In this
tutorial, we highlight an explanation based on the representation bias. The naive bayes
classifier is a linear classifier, as well as linear discriminant analysis, logistic regression or linear
SVM (support vector machine). The difference lies on the method of estimating the parameters
of the classifier (the learning bias).
While the Naive Bayes classifier is widely used in the research world, it is not widespread
among practitioners which want to obtain usable results. On the one hand, the researchers
found especially it is very easy to program and implement it, its parameters are easy to
estimate, learning is very fast even on very large databases, its accuracy is reasonably good in
comparison to the other approaches. On the other hand, the final users do not obtain a model
easy to interpret and deploy, they does not understand the interest of such a technique.

Thus, we introduce in a new presentation of the results of the learning process. The classifier is
easier to understand, and its deployment is also made easier. In the first part of this tutorial, we
present some theoretical aspects of the naive bayes classifier. Then, we implement the
approach on a dataset with Tanagra. We compare the obtained results (the parameters of the
model) to those obtained with other linear approaches such as the logistic regression, the linear
discriminant analysis and the linear SVM. We note that the results are highly consistent. This
largely explains the good performance of the method in comparison to others. In the second
part, we use various tools on the same dataset (Weka 3.6.0, R 2.9.2, Knime 2.1.1, Orange 2.0b
and RapidMiner 4.6.0). We try above all to understand the obtained results.

Random Forest

Random forests or random decision forests are an ensemble learning method for classification,
regression and other tasks that operates by constructing a multitude of decision trees at
training time. For classification tasks, the output of the random forest is the class selected by
most trees. For regression tasks, the mean or average prediction of the individual trees is
returned. Random decision forests correct for decision trees' habit of overfitting to their
training set. Random forests generally outperform decision trees, but their accuracy is lower
than gradient boosted trees. However, data characteristics can affect their performance.
The first algorithm for random decision forests was created in 1995 by Tin Kam Ho[1] using the
random subspace method, which, in Ho's formulation, is a way to implement the "stochastic
discrimination" approach to classification proposed by Eugene Kleinberg.

An extension of the algorithm was developed by Leo Breiman and Adele Cutler, who registered
"Random Forests" as a trademark in 2006 (as of 2019, owned by Minitab, Inc.).The extension
combines Breiman's "bagging" idea and random selection of features, introduced first by Ho[1]
and later independently by Amit and Geman[13] in order to construct a collection of decision
trees with controlled variance.

Random forests are frequently used as "blackbox" models in businesses, as they generate
reasonable predictions across a wide range of data while requiring little configuration.

SVM

In classification tasks a discriminant machine learning technique aims at finding, based on an

independent and identically distributed (iid) training dataset, a discriminant function that can
correctly predict labels for newly acquired instances. Unlike generative machine learning
approaches, which require computations of conditional probability distributions, a discriminant
classification function takes a data point x and assigns it to one of the different classes that are
a part of the classification task. Less powerful than generative approaches, which are mostly
used when prediction involves outlier detection, discriminant approaches require fewer
computational resources and less training data, especially for a multidimensional feature space
and when only posterior probabilities are needed. From a geometric perspective, learning a
classifier is equivalent to finding the equation for a multidimensional surface that best
separates the different classes in the feature space.
SVM is a discriminant technique, and, because it solves the convex optimization problem
analytically, it always returns the same optimal hyperplane parameter—in contrast to genetic
algorithms (GAs) or perceptrons, both of which are widely used for classification in machine
learning. For perceptrons, solutions are highly dependent on the initialization and termination
criteria. For a specific kernel that transforms the data from the input space to the feature
space, training returns uniquely defined SVM model parameters for a given training set,
whereas the perceptron and GA classifier models are different each time training is initialized.
The aim of GAs and perceptrons is only to minimize error during training, which will translate
into several hyperplanes’ meeting this requirement.
CONCLUSION AND FUTURE WORK

In this review we have analyzed the Machine Learning based methods most widely used by the
scientific community for diagnosing machine failures. For each type of methodology, we first
provided the background necessary to understand the method and then we analyzed the most
representative works that have yielded these techniques to identify failures in the industrial
machine. The work carried out highlighted the strong use of these methods which confirms the
extreme usefulness of these techniques in identifying failures in scenarios heavily contaminated
by residual noise. The automatic extraction of knowledge today represents a valid tool for
identifying faults: Technicians who must manage the maintenance of an industrial process must
necessarily use these methods for a correct forecast of the mechanical parts to be replaced.
REFERENCES
[1] M. Singh, B. K. Panigrahi, and R. P. Maheshwari, "Transmission line fault detection and classification,"
Emerging _Trends in Electrical and Computer_Technology conference, 2011, pp. 15–22.

[2] Nikita V.Tomin et al. Machine-Learning Techniques for Power-System Security Assessment.
IFACPapersOnLine, 2016, 49(27), pp.445-450.

[3] S. Rinaldi, M. Pasetti, P. Ferrari, G. Massa, D. Della Giustina. "Experimental characterization of

communication-infrastructure for virtual power plant monitoring."IEEE International Workshop on
Applied Measurements for Power Systems (AMPS), pp. 1-6, 2016.

[4] S. Rinaldi, M. Pasetti, A. Flammini, F. De Simone, "Characterization of Energy Storage Sytems for
Renewable Generators: An Experimental Testbed," IEEE International Workshop on Applied
Measurements for Power Systems (AMPS), 2018.

[5] T. Hong et al., "Guest editorial big data analytics for grid modernization," IEEE Trans. Smart Grid, vol.
7, no. 5, pp. 2395-2396, Sep. 2016.

[6] B. Wang, B. Fang, Y. Wang, H. Liu, Y. Liu, "Power system-transient stability assessment based on big-
data and the core vector machine," IEEE Trans. Smart Grid, vol. 7, no. 5, pp. 2561-2570, Sep. 2016.

[7] S. M. Miraftabzadeh, F. Foiadelli, M. Longo and M. Pasetti, "A Survey of Machine Learning
Applications for Power System Analytics," 2019 IEEE International Conference on Environment and
Electrical Engineering and IEEE Industrial_ and _Commercial Power _Systems Europe (EEEIC / I&CPS
Europe), Genova, Italy, 2019, pp. 1-5.

[8] S. Rinaldi, A. Flammini, M. Pasetti, L. C. Tagliabue, A. C. Ciribini, S. Zanoni, "Metrological Issues in the
Integration of Heterogeneous IoT Devices for Energy Efficiency in Cognitive Buildings," IEEE International
Instrumentation and Measurement Technology Conference , 2018
[9] R. Diao, et al.Decision tree-based online voltage security assessment using PMU measurements.EEE
Trans. Power Syst., 24 (2) (2009), pp. 832-839

[10] K. Morison, L. Wang, P. Kundur, Power system security assessment, IEEE Power and Energy
Magazine, 2 (5) (2004), pp. 30-39

[11] Tomin N., Negnevitsky M., Rehtanz Ch. (2014). Preventing Large-Scale Emergencies in Modern
Power Systems: AI Approach, Journal of Advanced Computational Intelligence and Intelligent
Informatics, Vol. 18, No.5.

[12] A.Muthukrishnan, J.Charles Rajesh kumar, D.Vinod Kumar, M.Kanagaraj. "Internet of image things-
discrete wavelet transform and Gabor wavelet transform based image enhancement resolution
technique for IoT satellite applications".57, pp.46-53, 2019.

[13] Voumvoulakis, Emmanouil, Gavoyiannis, A.E. & Hatziargyriou, Nikos. (2007). Application of
MachineLearning on Power -System Dynamic Security Assessment. Intelligent Systems Applications to
Power Systems (ISAP). 1 - 6.

[14] Yuanjun Guo, Zhile-Yang, Shengzhong Feng, and Jinxing Hu, "Complex Power-System Status
Monitoring and Evaluation Using Big _Data Platform and MachineLearning Algorithms: A Review and a
Case Study," Complexity, vol. 2018, Article ID 8496187, 21 pages, 2018.

[15] XingLuo et.al. Overview of current development in electrical energy storage technologies and the
application potential in power system operation. Journal of applied energy, 2015, 137, pp.511-536.

[16] Alex Albert et.al. Safety risk-management for electricaltransmission and distribution line
construction, Journal of safety science, 2013, 51(1), 118-126.

[17] H.Kiani et.al. Coordinated transmission substations and sub-transmission networks expansion
planning incorporating distributed generation, Journal of Energy,2017,120(1),pp.996-1011.
[18] Andrédos Santos, M.T Correia de Barrosa. Stochastic modeling of power system faults. Journal of
Electric Power Systems Research, 2015,126, pp.29-37.

[19] Avagaddi Prasad, J.Belwin Edward, K.Ravi, A review on fault classification methodologies in power
transmission systems: Part—I. Journal of Electrical-Systems and Information Technology, 2018, 5(1),
pp.48-60.

[20] D. Vinod Kumar, A. Nagappan. "Performance Analysis of Security and Accuracy on Palmprint Based
Biometric Authentication System," International Journal of Innovative Research in Computer and
CommunicationEngineering, Vol. 3, Issue 7, July 2015, pp 6697-6704.

[21] D. Nguyen, R. Barella, S. A. Wallace, X. Zhao, and X. Liang, "Smart grid line event classification using
supervised learning over pmu data streams," in 2015 Sixth International Green and Sustainable
Computing Conference (IGSC), Dec 2015, pp. 1–8.

[22] F. Gao, J. S. Thorp, S. Gao, A. Pal, and K. A. Vance, "A voltage phasor based fault-classification
method for phasor measurement unit only state estimator output," Electric Power Components and
Systems, vol. 43, no. 1, pp. 22–31, 2015.

[23] D. Nguyen, S. Wallace, and X. Zhao, "Finding needles in a haystack: Line event detection on smart
grid pmu data streams," in 2016 Sixth International Conference on Smart Grids, Green Communications
and IT Energyaware Technologies, 2016, pp. 42–47.

[24] N.Stalin, P.Selvam." Power Transfer efficiency Analysis of Double Intermediate-Resonator for
Wireless Power Transfer". International -Journal of Advances in Engineering and Emerging Technology,
9(3), pp.130-141, 2018.

[25] D.Madeshwaran P Selvam." Back To Back Converter Based Real and Reactive Power Control with
Constant Speed Operation of Dfig in Wind Mill", 8(2), pp.1855- 1860, 2019.
[26] K. S. Swarup, H. S. Chandra sekharaiah. "Fault detection and diagnosis of power systems using
artificial neural networks." Proceedings of the First International Forum on Applications of Neural
Networks to Power Systems, pp. 102-106, Jul 1991.

[27] N. S. Coleman, C. Schegan, and K. N. Miu, "A study of power distribution system fault classification
with machine learning techniques," 2015 North American Power Symposium (NAPS), Charlotte, NC,
2015, pp. 1-6.

[28] Singh, N. Thakur, and A. Sharma, "A review of supervised machine learning algorithms," 2016 3rd
International-Conference on Computing for Sustainable Global-Development (INDIACom), New Delhi,
2016, pp. 1310-1315.

[29] Ravikumar, D. Thukaram, and H. P. Khincha, "Application of support vector machines for fault
diagnosis in power transmission system," in IET Generation, Transmission & Distribution, vol. 2, no. 1,
pp. 119-130, January 2008.

[30] Andersson, J. E. Solem, and B. Eliasson. "Classification of power system stability using support
vector machines," IEEE Power Engineering Society_General-Meeting, 2005, San_Francisco, CA, 2005, pp.
650-655 Vol. 1.