Materials & Design 244 (2024) 113086

Contents lists available at ScienceDirect

Materials & Design

journal homepage: www.elsevier.com/locate/matdes

Big data, machine learning, and digital twin assisted additive

manufacturing: A review
Liuchao Jin a,b,c , Xiaoya Zhai d,a , Kang Wang e,f , Kang Zhang a,g , Dazhong Wu h , Aamer Nazir i,j ,
Jingchao Jiang k,∗ , Wei-Hsin Liao a,l,∗
a
Department of Mechanical and Automation Engineering, The Chinese University of Hong Kong, Hong Kong, China
b
Department of Mechanical and Energy Engineering, Southern University of Science and Technology, Shenzhen, 518055, China
c
Shenzhen Key Laboratory of Soft Mechanics & Smart Manufacturing, Southern University of Science and Technology, Shenzhen, 518055, China
d
School of Mathematical Sciences, University of Science and Technology of China, Hefei, 230026, China
e
School of Nursing, The Hong Kong Polytechnical University, Hong Kong, China
f
School of Mechanical Engineering, Zhejiang University, Hangzhou, 310027, China
g
Nano and Advanced Materials Institute Limited, Hong Kong Science Park, Hong Kong, China
h
Department of Mechanical and Aerospace Engineering, University of Central Florida, Orlando, 32816, FL, USA
i
Department of Mechanical Engineering, King Fahd University of Petroleum & Minerals, Dhahran, 31261, Saudi Arabia
j
Interdisciplinary Research Center for Advanced Materials, King Fahd University of Petroleum & Minerals, Dhahran, 31261, Saudi Arabia
k
Department of Engineering, University of Exeter, Exeter, United Kingdom
l
Institute of Intelligent Design and Manufacturing, The Chinese University of Hong Kong, Hong Kong, China

A R T I C L E I N F O A B S T R A C T

Keywords: Additive manufacturing (AM) has undergone significant development over the past decades, resulting in vast
Additive manufacturing amounts of data that carry valuable information. Numerous research studies have been conducted to extract
Big data insights from AM data and utilize it for optimizing various aspects such as the manufacturing process, supply
Machine learning
chain, and real-time monitoring. Data integration into proposed digital twin frameworks and the application of
Digital twin
machine learning techniques is expected to play pivotal roles in advancing AM in the future. In this paper, we
Data-driven
provide an overview of machine learning and digital twin-assisted AM. On one hand, we discuss the research
domain and highlight the machine-learning methods utilized in this field, including material analysis, design
optimization, process parameter optimization, defect detection and monitoring, and sustainability. On the other
hand, we examine the status of digital twin-assisted AM from the current research status to the technical approach
and offer insights into future developments and perspectives in this area. This review paper aims to examine
present research and development in the convergence of big data, machine learning, and digital twin-assisted
AM. Although there are numerous review papers on machine learning for additive manufacturing and others
on digital twins for AM, no existing paper has considered how these concepts are intrinsically connected and
interrelated. Our paper is the first to integrate the three concepts big data, machine learning, and digital twins
and propose a cohesive framework for how they can work together to improve the efficiency, accuracy, and
sustainability of AM processes. By exploring latest advancements and applications within these domains, our
objective is to emphasize the potential advantages and future possibilities associated with integration of these
technologies in AM.

1. Introduction production of complex geometries that were once impractical or impos-

sible with traditional manufacturing methods [12–21]. This transfor-
Additive manufacturing (AM), also known as 3D printing, has rev- mative technology has found applications in various sectors, including
olutionized the manufacturing industry by offering unprecedented de- aerospace [22–31], automotive [32–40], healthcare [41–54], food in-
sign freedom [1–5], reducing material waste [6–11], and enabling the dustry [55–69], and even construction industry [70–83]. As AM contin-

* Corresponding authors.
E-mail addresses: [email protected] (J. Jiang), [email protected] (W.-H. Liao).

https://doi.org/10.1016/j.matdes.2024.113086
Received 15 March 2024; Received in revised form 26 May 2024; Accepted 11 June 2024
Available online 25 June 2024
0264-1275/© 2024 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

ues to evolve and gain wider adoption in various industries, the focus keyword co-occurrence and cluster analysis, to analyze literature on
has shifted towards optimizing the manufacturing process, enhancing various aspects of AM, such as design for AM, material analytics, moni-
product quality, and accurately predicting material properties [84,85]. toring, defect detection, property prediction, and sustainability. Guo et
The complex nature of AM, with its intricate geometries, diverse ma- al. [131] addressed the challenge of interpreting machine learning out-
terials, and intricate interplay of process parameters, poses challenges comes within the context of complex thermodynamics governing AM.
in achieving consistent and reliable outcomes. To address these chal- To address this, the concept of physics-informed machine learning was
lenges, the integration of advanced data-driven technologies such as introduced, integrating data-driven methods with underlying physical
big data analytics, machine learning (ML), and digital twin (DT) sim- principles. Chinchanikar et al. [132] provided a comprehensive review
ulations becomes crucial. The vast amount of data generated during of recent advancements in the application of machine learning to AM,
the AM process provides valuable insights into process performance, with a focus on the aerospace industry. (3). For digital twins, Phua
material behavior, and product characteristics. By harnessing big data et al. [133] critically examined the application of digital twin tech-
analytics, these data can be analyzed and interpreted to identify pat- nology in the context of metal AM. They explored how digital twins
terns, correlations, and hidden trends that may impact the final product offer a viable solution to the challenges inherent in metal AM, partic-
quality. ularly related to part qualification, certification, and optimization. The
Big data plays a pivotal role in AM by enabling the collection, evaluation encompassed diverse aspects such as modeling, sensing, con-
storage, and analysis of vast amounts of data generated during the man- trol, and machine intelligence. They identified a four-level hierarchy for
ufacturing process [86,87]. By leveraging big data analytics techniques, the development of a digital twin for metal AM, highlighting essential
valuable insights and patterns can be extracted so that they can be used components like surrogate modeling, in-situ sensing, hardware control
to optimize process parameters [88–90], monitor product quality [91], systems, and intelligent control policies. Zhang et al. [134] delved into
and improve overall performance [92]. The utilization of big data in the potential of digital twin applications in AM. They recognized the
AM has the potential to enhance efficiency, reduce costs, and facilitate challenges posed by the trial-and-error approach to achieve optimal
innovation. structural integrity and mechanical properties in AM-produced compo-
In recent years, machine learning, a subset of artificial intelligence nents. They proposed that digital twins could address these challenges
(AI), has also emerged as a game-changer in the AM domain. ML al- by offering a comprehensive digital representation of production sys-
gorithms and techniques enable computers to learn from data patterns tems or products. Digital twins were identified as a potential solution
and make data-driven decisions [93], optimizing manufacturing pro- to enhance part quality, shorten product qualification time, and im-
cesses [94], enhancing product quality [95], and reducing costs [96]. prove overall production efficiency. Chen et al. [135] proposed a novel,
The integration of machine learning methods into AM has opened up service-oriented framework for AM digital twin development, address-
new possibilities for material analysis, design optimization, process op- ing challenges stemming from the uniqueness of each AM process. They
timization, defect detection, real-time monitoring, and sustainable man- outlined the discrepancies in defining and developing AM digital twins,
ufacturing [97–99]. which often lead to high costs and low adaptability. To mitigate these
Additionally, the concept of digital twins has emerged as a transfor- issues, They summarized existing AM digital twin approaches and in-
mative technology in the manufacturing industry [100–102]. A digital troduces a four-layered framework: service, model, data, and interface.
twin is a virtual representation of a physical object or system, capturing This framework aimed to enhance reusability across different levels
its behavior, performance, and lifecycle in real-time. This technology of AM digital twin development. Ladani [136] explored the applica-
has found diverse applications in healthcare [103–108], smart cities tions of artificial intelligence and machine learning in metal AM. They
[109–112], and, more importantly, AM [113–115]. Digital twins en- highlighted the potential synergies between AI and AM due to the vast
able real-time monitoring, predictive maintenance, and optimization of amount of digital data generated in AM processes. Specifically focusing
AM processes, paving the way for data-driven decision-making and in- on powder bed AM technology, they discussed the types and sources
creased operational efficiency. of data, potential variabilities in experimental and simulation data, and
There have been a lot of review papers exploring the application how these data can be utilized in ML algorithms. They also presented
of these three techniques in AM [116–127]. (1). For big data, Bi et al. several innovative ideas that demonstrate how combining AI and AM
[90] conducted a comprehensive review of the integration of big data in could significantly impact various fields.
AM, highlighting its potential impact on design, production, and supply Although there are numerous review papers on machine learning
chain processes. They explored how big data analytics can effectively for additive manufacturing and others on digital twins for AM, no
handle the vast data generated by AM, benefiting both research and existing paper has considered how these concepts are intrinsically con-
production. Additionally, big data methods were identified as aiding nected and interrelated. Our paper is the first to integrate the three
designers, engineers, and customers through valuable information col- concepts—big data, machine learning, and digital twins—and propose
lection. Wang & Alexander [128] reviewed the utilization of big data a cohesive framework for how they can work together to improve the
in AM, emphasizing its influence on supply chains and production effi- efficiency, accuracy, and sustainability of additive manufacturing pro-
ciency. They explored the advantages, applications, and technological cesses as demonstrated in Fig. 1. This distinct approach is lacking in
advancements of AM, including its limitations and cybersecurity consid- current literature, where the interplay between these three elements is
erations. Additionally, the integration of big data and the application of often overlooked. By amalgamating the capabilities of big data, ma-
big data analytics in AM were discussed to underscore their significance chine learning, and digital twin, our review provides an integrated
in enhancing the overall AM process. (2). For machine learning, Wang framework that demonstrates their symbiotic relationship. Our com-
et al. [98] examined the use of machine learning in AM. They explored prehensive analysis not only highlights the individual strengths of each
how machine learning enhances various aspects of AM, from design op- technology but also elucidates their collective potential. This inclusive
timization to production quality assessment and data security. Meng review not only offers insights into their individual significance but also
et al. [129] extensively assessed the utilization of machine learning in unveils how they synergistically shape the future of AM. In doing so, our
AM, categorizing applications like parameter optimization and anomaly paper not only aims to contribute to the academic discourse but also
detection into distinct machine learning tasks such as regression, classi- presents a forward-looking perspective that is aligned with the evolv-
fication, and clustering. Qin et al. [130] conducted an in-depth review ing landscape of advanced manufacturing technologies. The objective
of the role of machine learning in addressing challenges within AM. of this review paper is to provide a comprehensive overview of the cur-
They recognized the potential of AM for revolutionizing manufactur- rent state of research and development in the intersection of big data
ing and emphasized the obstacles related to complex systems, product analytics, machine learning, and digital twin-assisted AM. By examin-
quality, and adoption. They employed systematic methods, including ing the latest advancements and applications in these areas, we aim to

2
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Fig. 1. Interconnected landscape: big data, machine learning, and digital twin.

highlight the potential benefits, challenges, and future prospects of in- with the application of machine learning and digital twins, this review
tegrating these technologies in the AM process. does not list the application of big data analysis in AM as a separate sec-
The subsequent sections of this review paper are structured as fol- tion. By shedding light on the latest advancements and applications, we
lows: In Section 2, we will provide a detailed discussion of the concepts seek to underscore the potential benefits of these integrated technolo-
of big data analytics, machine learning, and digital twin technology, gies and inspire further research and innovation in this exciting and
emphasizing their relevance and significance in the AM domain. We rapidly evolving field.
will delve into the specific application areas of machine learning in
AM in Section 3. This includes material analysis, where machine learn- 2. Basic concepts
ing techniques can be employed to optimize material properties and
enhance material selection processes. We will also explore design opti- In this section, we will explore the basic concepts of AM, big data,
mization, where geometric and topology optimization methods driven machine learning, and digital twins.
by machine learning can revolutionize how products are designed and
manufactured. This section will also delve into the realm of process 2.1. Additive manufacturing
optimization, discussing how machine learning algorithms can be uti-
lized to achieve better control over AM processes, leading to improved ISO/ASTM categorizes additive manufacturing into seven distinct
efficiency and reduced defects. Additionally, we will discuss the sig- groups, namely vat photopolymerization, material jetting, binder jet-
nificance of defect detection and real-time monitoring in ensuring the ting, powder bed fusion, material extrusion, directed energy deposition,
quality of the AM process. Moreover, we will explore the concept of and sheet lamination [137]. In this subsection, we will introduce the
sustainability in AM, emphasizing how machine learning and digital concept of AM through reviewing the developments of seven different
twin technologies can contribute to environmentally friendly practices categories of AM demonstrated in Fig. 2.
through energy consumption modeling, waste reduction, and optimiza-
tion of support structures. Section 4 will assess the current research 2.1.1. Vat photopolymerization
status of digital twin-assisted AM and discuss the technical approaches Vat photopolymerization is an AM process that uses a vat of liquid
being used to integrate digital twins into the AM workflow. Addition- photopolymer resin that is selectively cured (hardened) by exposing it
ally, we will explore future development directions and perspectives to a UV light source, usually in the range of 350-405 nm [138]. This
to highlight the potential of this transformative technology in revo- process is also known as stereolithography (SLA) or digital light pro-
lutionizing the manufacturing industry. Finally, Section 5 presents a cessing (DLP) 3D printing. In SLA, a laser is used to selectively cure the
comprehensive discussion on the implications and potential areas of fu- resin. In DLP, a digital projector is used to project a pattern of light onto
ture research, followed by a conclusion. Through this comprehensive the resin.
review, we aim to highlight the challenges, opportunities, and future The process involves a build platform that is lowered into the vat of
prospects of leveraging big data, machine learning, and digital twin liquid photopolymer resin, and a UV light source that selectively cures
technologies in AM. Because big data analysis is usually interspersed the resin layer by layer, creating a solid object. Once a layer has been

3
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Fig. 2. Additive manufacturing process categories.

cured, the build platform is lowered into the vat by a small amount, 2.1.3. Binder jetting
typically in the range of 0.05-0.15 mm, and the next layer of resin is Binder jetting is an AM process that uses a binder material to fuse
spread on top of the cured layer [139]. The process is repeated until together a powdered material, such as metal, plastic, or ceramic, to cre-
the entire object has been printed. After printing is complete, the object ate a solid object. It works by depositing a liquid binder onto a thin
is removed from the vat and washed to remove any uncured resin. The layer of powder, which then solidifies and binds the particles together.
object may also be post-processed to smooth or finish the surface, or to The process is repeated layer by layer until the object is complete. Once
add additional features or colors. the object is printed, it is typically sintered or cured to strengthen the fi-
Vat photopolymerization produces high-resolution, detailed parts nal product. Binder jetting is often used in industries such as aerospace,
with smooth surfaces and high accuracy. It is commonly used in indus- automotive, and medical device manufacturing. It has been used to pro-
tries such as jewelry, dentistry, and prototyping. However, the process duce parts such as engine components, dental implants, and surgical
can be slow compared to other 3D printing techniques, and the cost of instruments [146].
One of the main advantages of binder jetting is its ability to print
the photopolymer resin can be relatively high [140]. In addition, the
with a wide range of materials, including metals, plastics, and ceramics.
cured resin may be brittle and have limited mechanical properties, de-
This makes it a versatile manufacturing method for creating a variety
pending on the specific material used [141].
of parts and components. While binder jetting is a relatively fast pro-
cess, it does have some limitations. The resolution and accuracy of the
final product can be lower than other 3D printing methods, which may
2.1.2. Material jetting
limit its use for certain applications [147]. Additionally, post-processing
Material jetting is a 3D printing technology that uses a print head
steps such as sintering or curing may be required to strengthen the final
to deposit small droplets of material onto a build platform. This process
product [148]. Ongoing research is focused on improving the resolution
is similar to inkjet printing, where tiny droplets of ink are deposited
and accuracy of the final product, as well as developing new materials
onto paper. In material jetting, the print head can deposit droplets of
and applications for binder jetting [149].
multiple materials, such as plastics, metals, or ceramics, onto the build
platform. The droplets are then cured using heat, light, or other methods
2.1.4. Powder bed fusion
to solidify the material and create a 3D object [142].
Powder bed fusion is a 3D printing technology that uses a laser or
One of the key advantages of material jetting is its ability to pro-
electron beam to selectively melt or fuse together layers of powdered
duce parts with very high resolution and accuracy [143]. This makes
material, such as metal, plastic, or ceramic, to create a solid object.
it well-suited for applications that require fine details, such as jewelry,
There are two main types of powder bed fusion: selective laser melting
dental and medical devices, and small mechanical parts. Material jet-
(SLM) and electron beam melting (EBM). SLM uses a high-powered laser
ting also allows for the creation of parts with multiple materials, which to selectively melt the powder, while EBM uses an electron beam to
can be useful for creating parts with varying properties, such as parts melt the powder. The process begins with a thin layer of powder spread
that need to be both rigid and flexible [144]. However, material jet- evenly over a build platform. A laser or electron beam is then used to se-
ting can be relatively slow compared to other 3D printing technologies, lectively melt or fuse together the powder in the desired areas, creating
particularly for larger parts [145]. It can also be more expensive, as a solid layer. The build platform then moves down by a layer thickness,
it typically requires specialized equipment and materials. Despite these and the process is repeated layer by layer until the object is complete.
limitations, material jetting remains an important 3D printing technol- Powder bed fusion is typically used to create small to medium-sized
ogy, particularly in industries such as product design, prototyping, and parts, although larger parts can be created by using multiple machines
the creation of high-precision parts for aerospace and automotive appli- or by printing and joining individual components together. The pro-
cations. Ongoing research is also exploring ways to improve the speed, cess can be relatively slow compared to other 3D printing technologies,
cost, and versatility of material jetting. particularly for parts with complex geometries [150].

4
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

One of the main advantages of powder bed fusion is its ability to metal parts and components. Ongoing research is focused on improv-
create parts with very high accuracy and resolution [151]. This makes ing the speed and efficiency of the process, as well as expanding the
it an excellent choice for creating parts with intricate geometries and range of materials and geometries that can be used.
fine details. Additionally, powder bed fusion can be used with a wide
range of materials, including metals [152], plastics [153], and ceramics 2.1.7. Sheet lamination
[154]. Powder bed fusion has some limitations, including the need for Sheet lamination is a 3D printing technology that involves bonding
post-processing steps such as heat treatment to improve the strength of together thin layers of material, usually paper or plastic, to create a
the final product [155–159]. Additionally, the process can be relatively solid object. The process begins with a 3D model that is sliced into thin
expensive, as it requires specialized equipment and materials [160]. layers using 3D printing software. The printer then applies adhesive to
Ongoing research is focused on multi-material printing [161], reducing each layer and bonds them together, layer by layer, until the object
residual stresses and distortions [162], and new materials development is complete. Sheet lamination is a relatively low-cost and low-tech 3D
[163]. printing technology, making it accessible to a wide range of users. It is
often used in educational settings, such as schools and libraries, as well
2.1.5. Material extrusion as for rapid prototyping and small-scale production runs [185].
Material extrusion is a 3D printing technology that uses a heated One of the main advantages of sheet lamination is its low cost and
nozzle to melt and extrude a thermoplastic material, such as ABS or accessibility [186]. The materials used are relatively inexpensive, and
PLA, layer by layer, to create a solid object. The process begins with the process does not require sophisticated equipment or technical exper-
a 3D model that is sliced into thin layers using 3D printing software. tise. This makes it a popular choice for users who are new to 3D printing
The printer then extrudes the melted material through a nozzle, build- or who have limited resources. Another advantage of sheet lamination
ing up the object layer by layer. Material extrusion is typically used for is its ability to create objects with full-color graphics or images [187].
creating prototypes, low-volume production runs, and one-off parts. It This is achieved by printing the color or image onto the surface of
is also commonly used in education and hobbyist applications, as desk- each layer before bonding them together. This makes it well-suited for
top 3D printers using material extrusion technology can be relatively creating models or prototypes of products that require detailed visual
affordable and easy to use [164]. representations. However, sheet lamination also has some limitations.
One of the main advantages of material extrusion is its ability to The resolution of the final object is typically lower than other 3D print-
print with a wide range of thermoplastic materials, including PLA, ABS, ing technologies, as the layers are relatively thick compared to other
PETG, and nylon [165]. This makes it a versatile manufacturing method methods [188]. Additionally, the materials used may not be as strong
for creating a variety of parts and components. However, material ex- or durable as those used in other methods [189], which limits the range
trusion does have some limitations. The resolution and accuracy of the
of applications for which it is suitable.
final product can be lower than other 3D printing methods, particularly
Despite these limitations, sheet lamination remains an important 3D
for parts with complex geometries [166]. Additionally, the layer-by-
printing technology, particularly for applications that require full-color
layer construction of parts can result in a visible texture on the final
graphics or images. Ongoing research is focused on improving the reso-
product [167]. Despite these limitations, material extrusion remains
lution and strength of the final product, as well as expanding the range
an important 3D printing technology, particularly for creating simple,
of materials that can be used. Newer techniques, such as hybrid lami-
functional parts quickly and affordably. Ongoing research is focused
nation, which combines sheet lamination with other AM or subtractive
on bioprinting and tissue engineering [168–173], multi-axis printing
manufacturing techniques, are also being explored to further improve
[174–178], as well as advanced filament materials that can be used
the capabilities of the technology.
[179–181].
The comparison of seven different kinds of AM technologies, includ-
ing advantages, disadvantages, materials, and applications, is listed in
2.1.6. Directed energy deposition
Table 1.
Directed energy deposition is a 3D printing technology that uses a
Most AM systems typically align with one of the seven established
high-energy heat source, such as a laser or electron beam, to melt and
categories outlined in current industry standards. However, there are
fuse together metal powders or wire feedstock to create a solid object.
latest exceptions to this categorization [255], such as cold spray tech-
There are two main types of directed energy deposition: laser-based
nology. Cold spray, unlike traditional AM methods, involves the depo-
and electron beam-based. Laser-based directed energy deposition uses
sition of metal or composite particles onto a substrate at supersonic
a high-powered laser to melt and fuse the metal material, while elec-
velocities, forming a solid build-up without melting the material. This
tron beam-based directed energy deposition uses an electron beam to
could require an update to the ISO/ASTM 52900 standard.
achieve the same result. The process begins with a 3D model that is
sliced into thin layers using 3D printing software. The printer then di-
rects the heat source to melt and fuse the metal material onto the build 2.2. Big data
platform, layer by layer, until the object is complete. Directed energy
deposition can be used with a wide range of metal materials, including Big data refers to the presence of extremely large and complex
titanium, stainless steel, and aluminum [182]. It is often used in indus- data sets that are challenging to manage, process, and analyze using
tries such as aerospace, automotive, and medical device manufacturing traditional data processing tools and methods [256]. Big data is char-
to create complex, high-performance parts and components. acterized by its volume (enormous amounts of data), velocity (high
One of the advantages of directed energy deposition is its ability to speed of data generation and processing), and variety (diversity of data
create large, complex parts with a high degree of accuracy and precision types and sources) [257]. The term “big data” encompasses various
[183]. It is also well-suited for repairing or adding material to existing entities, including social phenomena, information assets, data sets, ana-
parts, making it a popular choice for maintenance and repair applica- lytical techniques, storage technologies, processes, and infrastructures.
tions. However, directed energy deposition does have some limitations. Microsoft describes big data as the application of “serious computing
The process can be relatively slow compared to other 3D printing tech- power” to massive amounts of information, while the National Insti-
nologies, particularly for parts with complex geometries. Additionally, tute of Standards and Technology (NIST) emphasizes the need for a
post-processing steps such as machining or polishing may be required “scalable architecture for efficient storage, manipulation, and analysis”
to achieve the desired surface finish and tolerances [184]. Despite these when defining big data [258]. The development histories of big data
limitations, directed energy deposition remains an important 3D print- are listed in Table 2 and the core concept of big data can be expressed
ing technology, particularly for creating complex, high-performance through the following key aspects:

5
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Table 1
Comparison of additive manufacturing technologies.

Process Advantages Disadvantages Printable Materials Applications Ref

Vat photopoly- ∙ high-resolution ∙ long process time pure photopolymer or pure Industries such as jewelry, [190–199]
merization ∙ smooth surfaces ∙ expensive waxes, or they can contain dentistry, and prototyping
∙ high accuracy ∙ weak parts ceramic or metal particles

Material ∙ high accurate parts ∙ weak parts pure photopolymer or pure Industries such as product design, [200–204]
jetting ∙ smooth surface ∙ expensive waxes, or they can contain prototyping, high-precision parts
∙ low wastage ∙ material limitation ceramic or metal particles for aerospace and automotive
applications

Binder jetting ∙ create large, complex parts ∙ low resolution metals, ceramics, and polymers Industries such as aerospace, [205–214]
quickly ∙ brittle parts automotive, and medical device
∙ low cost manufacturing

Powder bed ∙ high-resolution ∙ need for post processing metals including titanium Rapid prototyping of parts; [215–233]
fusion ∙ wide range of materials used steps alloys, stainless steel, aluminum aerospace components; high
∙ expensive alloys; polymers including complex and delicate objects
∙ long process time nylon, PA

Material ∙ wide range of materials used ∙ low resolution thermoplastic polymer, Creating prototypes, low-volume [234–242]
extrusion ∙ easy to use ∙ visible texture surface including PLA, TPU, ABS; production runs, and one-off parts
∙ low cost elastomers; graphite; metals

Directed ∙ create large, complex parts ∙ long process time metals including titanium Industries such as aerospace, [243–251]
energy ∙ high-resolution ∙ post processing steps alloys, stainless steel, aluminum automotive, and medical device
deposition ∙ easy for repairing or adding ∙ material limitation alloys manufacturing to create complex,
material to existing parts high-performance parts and
components

Sheet ∙ easy to use ∙ low resolution paper, plastic film, metal foil, Rapid prototyping and small-scale [252–254]
lamination ∙ low cost ∙ weak parts composites including carbon production runs; educational
∙ material limitation fiber composites, glass fiber settings
composites

• “Volume,” “Velocity,” and “Variety”: These terms describe the ployed for a given task. The evolution histories of ML are demonstrated
characteristics of the involved information, highlighting the vast in Table 2.
amounts of data, the high speed at which it is generated and pro- Machine learning models can be classified into various types based
cessed, and the diverse types and sources of data. on their learning algorithms, objectives, and underlying mathematical
• Specific “Technology” and “Analytical Methods”: Big data necessi- techniques. Some commonly used machine learning models include:
tates unique requirements to effectively utilize such vast amounts
of information, including specialized technologies and analytical
• Supervised Learning Models: Linear Regression, Logistic Regression,
methods designed to handle the complexities and challenges associ-
ated with big data. Decision Trees, Random Forest, Support Vector Machines, Gradient
• Transformation into insights and creation of economic value: The Boosting models (e.g., XGBoost, LightGBM), Neural Networks (e.g.,
primary impact of big data on companies and society lies in its Multi-layer Perceptron).
ability to extract valuable insights from the data, leading to the gen- • Unsupervised Learning Models: K-means Clustering, Hierarchical
eration of economic value. By leveraging the power of big data an- Clustering, Principal Component Analysis (PCA), Gaussian Mixture
alytics, organizations can uncover meaningful patterns, trends, and Models (GMM), Self-Organizing Maps (SOM), Autoencoders.
correlations within the data, enabling informed decision-making • Reinforcement Learning Models: Q-Learning, Deep Q-Networks
and driving innovation. (DQN), Policy Gradient Methods, Actor-Critic Methods, Monte Carlo
Tree Search (MCTS).
Overall, big data represents a paradigm shift in the way data is
• Deep Learning Models: Convolutional Neural Networks (CNN), Re-
managed and analyzed, emphasizing the importance of scalable tech-
current Neural Networks (RNN), Long Short-Term Memory (LSTM),
nologies, advanced analytics, and the potential for generating valuable
insights and economic benefits. Generative Adversarial Networks (GAN), Transformers.
• Bayesian Models: Naive Bayes, Bayesian Networks, Hidden Markov
2.3. Machine learning Models (HMM), Gaussian Processes.
• Ensemble Learning Models: Bagging (Bootstrap Aggregating), Boost-
Machine learning is a specialized field focused on developing and ing (e.g., AdaBoost, Gradient Boosting), Stacking, Voting.
understanding methods that enable machines to “learn” and improve • Dimensionality Reduction Models: Principal Component Analysis
their performance on specific tasks by leveraging data [269,270]. It (PCA), t-Distributed Stochastic Neighbor Embedding (t-SNE), Lin-
is a subfield of artificial intelligence that emphasizes the creation of ear Discriminant Analysis (LDA).
algorithms and models capable of autonomously learning, making pre-
• Recommender Systems Models: Collaborative Filtering, Content-
dictions, and making decisions without explicit programming [271]. ML
Based Filtering, Hybrid Approaches.
involves the development of computer systems that can analyze and in-
terpret large volumes of data, identify patterns, and extract meaningful
insights to enhance their performance over time [272]. ML tasks are These models represent a range of techniques used in machine learn-
typically categorized into different types based on learning approaches ing, each with its own strengths and areas of application. Researchers
(supervised/unsupervised), learning models (classification, regression, and practitioners select appropriate models based on the specific prob-
clustering, dimensionality reduction), or the specific algorithms em- lem they are addressing and the characteristics of their data.

6
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Table 2
Evolution of big data, machine learning, and digital twins.

Time Big Data Machine Learning Digital Twin

1950- ∙ Development of relational database systems like ∙ Development of the perceptron algorithm by −−−−−−−
1980 IBM’s IMS and Oracle. Frank Rosenblatt [259] in 1957, one of the
∙ Creation of the World Wide Web by Tim earliest machine learning algorithms inspired by
Berners-Lee in 1989. biological neurons.
∙ Introduction of decision tree learning algorithms
like ID3 by Ross Quinlan [260] in the 1980s.

1980- ∙ Introduction of support vector machines (SVM)

2000 by Vladimir Vapnik [261] in the 1990s, providing
a powerful method for classification and
regression tasks.

2000- ∙ Google’s publication of the MapReduce and GFS ∙ Emergence of ensemble learning methods like ∙ Introduction of the “Digital Twin” concept by
2010 papers in 2004, laying the groundwork for random forests [262] and gradient boosting Michael Grieves [264] in 2002, highlighting the
distributed computing and the Hadoop ecosystem. machines [263], improving predictive potential for virtual replicas of physical assets.
∙ Apache Hadoop project launch in 2006, performance by combining multiple models. ∙ Adoption of simulation and modeling techniques
providing an open-source framework for in manufacturing and aerospace industries for
distributed storage and processing of big data. virtual prototyping and testing.

2010- ∙ Introduction of Apache Spark in 2014, offering ∙ Breakthroughs in deep learning, including the ∙ Implementation of digital twins in industries
2020 faster and more versatile big data processing development of convolutional neural networks like energy, healthcare, and transportation for
compared to MapReduce. (CNNs) [265] for image recognition and recurrent predictive maintenance and performance
∙ Growth of cloud computing platforms like neural networks (RNNs) [266] for sequential data. optimization.
Amazon Web Services (AWS), Google Cloud ∙ Introduction of the Transformer architecture by ∙ Development of IoT platforms and sensors for
Platform (GCP), and Microsoft Azure. Vaswani et al. [267] in 2017, revolutionizing collecting real-time data from physical assets and
natural language processing (NLP) by enabling environments.
efficient training on large datasets through
attention mechanisms.

2020- ∙ Advancements in real-time data processing ∙ Continued research in areas like reinforcement ∙ Expansion of digital twins into smart cities and
technologies like Apache Kafka and Apache Flink. learning, generative adversarial networks (GANs) urban planning, enabling holistic management of
[268], and self-supervised learning. infrastructure and resources. ∙ Integration of AI
∙ Advanced multimodal models like GPT, DALL-E, and machine learning into digital twins for
Sora demonstrated integration across text, advanced analytics, anomaly detection, and
images, and other data types, enabling more scenario simulation.
sophisticated and contextually aware outputs.

2.4. Digital twin by the utilization of a Digital Thread. The Digital Thread incorporates
the best available models (physical, behavioral, etc.) and updated in-
The initial definition of Digital Twin was formulated by NASA (Ta- formation to emulate the life cycle, actions, and operations of its real
ble 2), describing it as “an integrated multi-physics, multi-scale, proba- counterpart [276]. This concept is widely embraced because it offers a
bilistic simulation of a vehicle or system that utilizes the best available broad and inclusive definition, encompassing all the key characteristics
physical models, sensor updates, fleet history, etc., to replicate the life and elements of a DT, while also governing its performance. Impor-
of its flying counterpart” [273]. According to this definition, the DT con- tantly, unlike the initial NASA definition, this concept can be applied
sists of three principal components: (i) physical objects in the Physical to various development areas, as its primary objective is not limited to
World, (ii) digital objects in the Digital World, and (iii) the connections replicating the life of an aircraft but rather capturing the life cycle of
that link the digital and physical elements together. any element, product, or system that serves as its physical twin.
While the research on DT primarily focused on the aerospace field
initially, in 2013, there were initial studies exploring its application in 3. Machine learning assisted additive manufacturing
the manufacturing sector. Lee and colleagues proposed the DT to be the
virtual representation of production resources, extending beyond just 3.1. Overview
the product itself. This sparked a discussion about the role of DT in
advanced manufacturing environments, particularly in relation to In- Machine learning has revolutionized AM by enabling advancements
dustry 4.0, which encompasses core technologies like big data analytics in defect detection, process parameter optimization, design optimiza-
and cloud platforms [274]. This ongoing debate forms the context for tion, material analysis, and sustainability. Through the application of
the present work. various machine learning algorithms such as k-nearest neighbors [277],
It became evident that the significance of DT in the manufactur- support vector machines [278,279], deep learning [280,281], decision
ing industry lies in its definition as the virtual counterpart of physical trees [282], and genetic algorithms [283], researchers and industry
devices. These virtual representations are based on semantic data mod- professionals have been able to improve the quality, efficiency, and
els and enable simulations across various disciplines. They facilitate overall performance of AM processes. By harnessing the power of ma-
not only static prognostic assessments during the design stage but also chine learning, they have successfully detected defects [277–279,284],
the continuous updating of the virtual representation through real-time optimized process parameters [283,285–293], generated optimized de-
synchronization with sensed data. This enables the representation to signs [294–301], analyzed materials [302–304], and made significant
reflect the current state of the system, facilitating real-time optimiza- strides towards sustainable manufacturing practices [305,306]. These
tion, decision-making, and predictive maintenance based on the sensed developments showcase the tremendous potential of machine learning
conditions. in transforming AM into a more reliable, cost-effective, and environ-
A widely acknowledged and utilized concept of DT is presented in mentally friendly production method.
[275]. According to this concept, a DT is a comprehensive, physical, Furthermore, machine learning algorithms play a pivotal role in re-
unified, and stochastic simulation of an as-built system, made possible fining AM processes by adjusting parameters based on insights learned

7
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

from historical data, thus ensuring higher quality outputs and improved
operational efficiency [277–279,281,282,284,307–311]. These algo-
rithms significantly augment defect detection mechanisms and quality
assurance practices by enabling automated inspection of manufactured
parts, drastically reducing human errors and time spent on manual as-
sessments [277–279,281,284,307–310]. By leveraging operational data
patterns, machine learning can foresee potential equipment failures, fa-
cilitating a proactive approach to maintenance and consequently reduc-
ing unexpected downtime [277–279,281,284,307–310]. Machine learn-
ing also accelerates the materials research segment of AM by predicting
the properties of potential materials, substantially speeding up the dis-
covery and testing processes [282,311]. In addition, the amalgamation
of machine learning with design tools introduces an avenue for design
optimization in AM, taking into account factors like manufacturability,
material usage, and performance characteristics [294–301]. Machine
learning also extends its capabilities to real-time monitoring of AM pro-
cesses, ensuring consistency and quality of production by identifying
and rectifying anomalies instantaneously [277–279,281,284,307–310].
Finally, by optimizing manufacturing processes and improving defect
detection, machine learning has a substantial impact on waste reduc-
tion, leading to considerable cost savings and environmental benefits
[283,285–293,312–327]. Thus, the confluence of machine learning and
AM unlocks a realm of opportunities, leading to more efficient pro-
cesses, improved quality, and innovative possibilities in design and
materials. In this section, we endeavor to present a comprehensive
overview of the applications of machine learning techniques within the Fig. 3. Research domain of machine learning in additive manufacturing.
diverse domains of AM processes. This synthesis draws on the most
recent scholarly literature, offering a critical examination of current and identify optimal designs for AM. Machine learning algorithms can
trends. Additionally, this discussion includes our insightful projections consider various factors such as manufacturability, material usage, and
on promising and impactful research directions that are either currently performance characteristics to generate designs that are both functional
under active exploration or possess significant potential for the future and efficient in terms of AM processes. This approach enables the cre-
development of AM. ation of complex geometries and customized designs that maximize the
capabilities of AM.
3.1.1. Research domain Material analysis is another domain where machine learning plays a
In this section, we will explore the wide-ranging applications of ma- vital role in AM [302–304]. Machine learning algorithms have been em-
chine learning in AM. Machine learning techniques have revolutionized ployed to analyze the properties of different materials used in AM pro-
the AM landscape, offering significant advancements in material analy- cesses. By considering various material characteristics, such as strength,
sis and selection, design optimization, process parameter optimization, thermal properties, and chemical composition, machine learning mod-
defect detection and real-time monitoring, and sustainability as shown els can assist in identifying the most suitable materials for specific
in Fig. 3. By harnessing the power of machine learning algorithms, applications, optimizing material selection, and even predicting mate-
researchers and industry professionals have been able to enhance the rial behavior under different conditions.
quality, efficiency, and overall performance of AM processes. Furthermore, sustainability is a growing concern in the field of AM,
One of the primary applications of machine learning in AM is in the and machine learning can contribute to addressing these challenges
domain of defect detection and real-time monitoring. With the ability to [305,306,341–347]. Machine learning is being employed to enhance
analyze large volumes of data and detect subtle patterns, machine learn- sustainability in AM by reducing material waste and energy consump-
ing algorithms enable automated inspection of manufactured parts, en- tion. By predicting potential sources of waste and inefficiencies in the
suring high-quality outputs and minimizing human errors [277]. These process, machine learning models can guide process improvements that
algorithms can quickly identify defects or anomalies in the manufac- lead to a more environmentally friendly and sustainable AM practice.
turing process, allowing for timely corrective actions [278,279,284]. In conclusion, machine learning has significantly advanced AM by
By reducing the reliance on manual inspection and streamlining the enhancing defect detection, process parameter optimization, design op-
quality assurance process, machine learning significantly improves the timization, material analysis, and sustainability. It has greatly improved
efficiency and reliability of AM [281,307–310]. the quality, efficiency, and sustainability of AM processes, offering new
Another crucial research domain where machine learning has made opportunities for complex designs. Continued research in this field holds
significant contributions is process parameter optimization. AM pro- great potential for further advancements in AM with the help of ma-
cesses involve a multitude of parameters that can affect the final pro- chine learning techniques.
duct’s quality and performance. Machine learning algorithms can an-
alyze historical data on process parameters and their corresponding 3.1.2. Machine learning methods
outcomes to identify the optimal parameter settings [283,285–293]. Machine learning methods commonly used in AM encompass var-
By leveraging insights learned from the data, machine learning algo- ious approaches. These include supervised learning methods such as
rithms can adjust the parameters in real-time, leading to improved k-Nearest Neighbors (k-NN) [322], Support Vector Machines (SVM)
process efficiency, reduced material waste, and enhanced product qual- [322,330], Random Forests (RF) [322], Decision Trees (DTree) [295,
ity [312–340]. 297,304,346], Linear Regression (LR) [322], and Gaussian Process
In addition to defect detection and process parameter optimization, Regression (GPR) [322,327]. Deep Learning (DL) methods, such as
machine learning techniques have found applications in design opti- neural networks (NN) [296,299–301] and Convolutional Neural Net-
mization [294–301]. By integrating machine learning algorithms with works (CNN) [337,339], are also prevalent. Gaussian Process (GP)
design tools, engineers and designers can explore vast design space methods utilize Gaussian distributions for regression and optimization

8
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

tasks [324,325,331,333–335]. Additionally, Particle Swarm Optimiza- further complicating data collection efforts. Consequently, AM primar-
tion (PSO) [342] and Bayesian Networks (BN) [338] are employed. ily deals with what can be termed as “small data.”
These methods are utilized for defect detection, real-time monitoring, The limitations posed by small data in AM necessitate the devel-
quality assessment, process parameter optimization, design optimiza- opment of techniques to overcome these challenges. Machine learning
tion, material analysis, and sustainability aspects in AM. models typically perform more robustly when trained on large volumes
The selection of a specific method depends on the goals and re- of data. Therefore, strategies must be devised to augment datasets and
quirements of the application at hand. In supervised learning, machine address data imbalances, enabling the effective utilization of ML algo-
learning models are used to predict the properties of interest based on rithms for process optimization and quality control. Several approaches
a labeled dataset. The model is trained on a collection of data-label have been proposed to mitigate data limitations in AM [357], including
pairs. Data can take various forms, such as fields, vectors, images, and experiment modularization, physics-informed machine learning, data
graphs. Supervised learning is typically divided into two main cate- augmentation, synthetic data generation, transfer learning, data-centric
gories: regression and classification, depending on the type of labels. machine learning models.
In the context of data-driven additive manufacturing, supervised learn-
ing is often employed to establish relationships between parameters and 3.2. Material analysis
performance. The data set usually describes a shape, represented by its
parameterization, and describes the quantity of interest, such as elastic AM has revolutionized the manufacturing industry by enabling the
components or response spectra. A key motivation for supervised learn- production of complex parts with unprecedented design freedom. Vari-
ing is to replace resource-intensive cell evaluation processes with faster ous materials, including polymers [358–369], metal [370–376], ceram-
alternative models. The types of models are primarily based on neural ics [377–381], concrete [382–384], ice [385–387], wood [388–390],
networks, sometimes also based on Gaussian Processes. Once trained on and even sugar [391,392], as well as their combinations, are utilized
a large dataset, data-driven models can instantly predict the properties in AM applications, and ongoing efforts are focused on developing new
of unseen cells. materials to expand the capabilities of the technology. The choice of
In contrast to supervised learning, unsupervised learning extracts materials in AM significantly impacts the performance and properties
information from unlabeled data and is mainly applied to learning of the final parts [393]. To optimize the manufacturing process and
the complex processes of additive manufacturing. Common unsuper- ensure high-quality results, it is crucial to analyze and understand the
vised learning models include autoencoders, variational autoencoders, relationships between material chemistry, material characteristics, and
and generative adversarial networks. An autoencoder is a type of neu- the performance of the printed parts using the available material data.
ral network that uses an encoder-decoder architecture to extract low- Most of the current material analysis for AM using machine learn-
dimensional latent variables from input data. Variational autoencoders ing focuses on metal powder material analysis in metal 3D printing
are a type of deep generative model that generates new data by sam- [394]. This is because metals AM has brought about a paradigm shift
pling low-dimensional latent variables that follow well-defined distribu- in the manufacturing of metal components, offering unique advantages
tions. Therefore, compared to traditional additive manufacturing opti- such as the simultaneous fabrication of materials and parts using a sin-
mization, the latent representations learned by variational autoencoders gle machine and the ability to produce highly complex geometries.
are often more efficient and interpretable, especially considering high However, ensuring the consistency and quality of parts and materials
design complexity and degrees of freedom. Semi-supervised learning is remains an ongoing challenge in many applications of metal AM. To
a machine learning method trained on partially labeled data so that the address this challenge, Integrated Computational Materials Engineering
model can predict the labels of unlabeled data. (ICME) approaches have been instrumental in accelerating the develop-
Compared to pure supervised and unsupervised learning, the advan- ment and adoption of materials technologies [395–398]. Traditionally,
tage of semi-supervised learning lies in its ability to leverage additional ICME approaches have relied on physics-based experimental data and
information provided by unlabeled data to improve model performance. simulations to understand the relationship between material properties
Semi-supervised learning bridges the gap between these two approaches and processing conditions [399]. However, in the context of metals AM,
by incorporating both labeled and unlabeled data into the training pro- where the physics are still being discovered, the development of com-
cess. The presence of unlabeled data allows the model to learn from putationally feasible physics-first ICME approaches remains an open
a larger dataset, capturing more comprehensive and nuanced patterns challenge. Additionally, the wealth of data generated in the field of
within the data. This can be particularly beneficial in scenarios where AM presents its own challenges, as the ability to store and analyze
obtaining labeled data is challenging or costly, but unlabeled data is this data is being pushed to its limits. Nevertheless, this abundance
abundant. of data has sparked a paradigm shift towards incorporating machine
The core idea of reinforcement learning is to simulate an agent in- learning into ICME approaches, enabling the extraction of valuable
teracting with an environment, learning how to maximize cumulative insights and the development of comprehensive and computationally
rewards by trying different actions. Typically, reinforcement learning feasible models. Motaman et al. [400] demonstrated a multi-scale and
is used to solve sequential decision-making problems, where the agent multi-physics ICME approach, highlighting the linkages between the
needs to select the optimal action among a series of decisions to achieve different stages of the process–structure–properties–performance chain.
specific goals. They also proposed a hybrid physics-based and data-driven strategy for
The machine learning methods that are commonly used in AM are optimal component design. Wang & Xiong [401] developed a method to
listed in Fig. 4 and the machine learning methods for different research optimize the composition of pre-alloyed powders in AM by employing
domains and different goals are shown in Table 3. a CALPHAD-based ICME framework. The high-strength low-alloy steel
was used as a case study, and process–structure–property relationships
3.1.3. Data limitation in additive manufacturing were analyzed for various compositions. Critical properties, including
Collecting extensive datasets for AM processes presents a significant yield strength, impact transition temperature, and weldability, were
challenge due to various factors, including process diversity and the evaluated to optimize the composition. The optimized composition in-
resource-intensive nature of experimentation. Unlike sectors like retail creased the probability of successful AM builds by 44.7%. While current
or streaming services, where vast amounts of data are readily available machine learning applications in material analysis for AM largely focus
from millions of customers, manufacturing, including AM, is character- on metal powder characterization, further exploration, and research are
ized by its segmented nature. Moreover, within the realm of AM, there needed to unlock the full potential of leveraging machine learning for
exist numerous categories based on material, technology, and hardware, material analysis, including analyzing material composition for alloy

9
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Fig. 4. Machine learning methods for additive manufacturing.

10
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Table 3
Application of machine learning in additive manufacturing.

Direction AM processes Materials Data type ML methods Goal Refs

Defect detection & real-time Powder bed fusion SS316L acoustic emission k-NN, DL, GPR quality [277]
monitoring
images CA quality [279]
images SVM, ANN quality [278]
printing parameters SVM quality [284]
SS304L acoustic emission SVM quality [307]
Ti-6Al-4V images SVM quality [308]
images SVM, k-NN, ANN quality [281]
images k-NN, SVM quality [309]
acoustic emission ANN quality [310]
Inconel 625 printing parameters ANN quality [311]
Al-5083 image DTree quality [282]
Material extrusion Carbon fiber reinforced real-time images DL quality [280]
polymer
ABS images k-NN, SVM, RF quality [348]
process parameter SVM quality [349]
Binder jetting SS316L images CNN quality [350]
Directed energy deposition Ti-6Al-4V images, optical data SVM quality [351]
optical emission DL, CA quality [352]
Vat photo-polymerization Liquid resin videos DL quality [353]
image GPR quality [354]
Material jetting liquefied metal video DL quality [355]
Dimatix images ANN quality [356]

Process parameter Powder bed fusion SS316L process parameters GPR quality [283]
optimization
process parameters GPR quality [285]
process parameters DTree, k-NN, LR quality [286]
process parameters BN quality [287]
process parameters SVM, DL, DTree, GPR quality [288]
images & data DL quality [289]
process parameters GPR property [290]
process parameters GPR quality [291]
process parameters GPR quality [283]
process parameters GPR quality [292]
process parameters ANN, RF, SVM property [293]
Inconel 625 & 718 process parameters SVM, DTree, LR quality [312]
Inconel 625 process parameters GA quality [313]
Inconel 718 process parameters DL property [314]
process parameters GA quality [315]
NiTi SMA process parameters DL quality [316]
AlSi10Mg process parameters GPR property [317]
Ti-6Al-4V process parameters ANN property [319]
process parameters GA property [318]
process parameters ANN property [319]
process parameters GPR property [320]
CoCr process parameters SVM quality [321]
Material extrusion PLA sensor data SVM, LR quality [322]
process parameter DL property [323]
process parameter GPR quality [324]
shape GPR quality [325]
ABS process parameter GA quality [326]
Silicone process parameter GPR property [327]
Binder jetting Co-Cr-Mo process parameters ANN quality [328]
Directed energy deposition Copper alloy process parameters SVM quality [329]
Al 4043 process parameters SVM model [330]
Al 5356 process parameters GPR quality [331]
SS304L process parameters GA property [332]
SS316L process parameters GPR quality [333]
Ti–Mn alloy process parameters GPR property [334]
Material jetting Silver process parameters GPR quality [335]
Glycerol process parameters DL quality [336]
Vat photo-polymerization Liquid resin images DL property [337]
CAD model BN quality [338]
CAD model DL quality [339]
process parameter GA quality [340]

Design optimization Powder bed fusion AlSi10Mg CAD model SVM design [294]
Inconel 625 images DTree design [295]
Material extrusion ABS 3D coordinates DL design [296]
build orientation ANN, BN design [297]
− − CAD model DL property [298]
CAD model GPR design [299]
CAD model DL design [300]
images DL design [301]
(continued on next page)

11
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Table 3 (continued)

Direction AM processes Materials Data type ML methods Goal Refs

Material analysis Powder bed fusion − radiation SVM classify [302]

images SVM classify [303]
features DTree classify [304]

Sustainability Material extrusion PLA process parameter GA cost [305]

ABS process parameter GA cost [306]
Directed energy deposition Inconel 718, Ti–6Al–4V process parameter GA efficiency [341]
Vat photo-polymerization Liquid resin process parameter PSO efficiency [342]
process parameter DL efficiency [343]
process parameter XGBoost, CA efficiency [344]
− − CAD model GA cost [345]
CAD model DTree efficiency [346]
process parameter DL efficiency [347]

development and modeling the intricate relationships between material 3.3. Design optimization
chemistry, properties, and the performance of the final printed parts.
Some other researches focus on the material properties prediction of Design optimization for AM [299] takes advantage of the unique ca-
AM. Jiang et al. [402] utilized machine learning models to predict the pabilities of AM technologies to optimize the design of parts or products
amorphization of crystalline drug formulations and the chemical stabil- that are stronger, lighter, and more efficient than traditional manufac-
ity of subsequent amorphous solid dispersions (ASDs) prepared through turing methods. Design optimization for AM is a critical aspect of AM,
hot-melt extrusion (HME). They trained the ML models using a dataset as it can significantly impact the performance and functionality of the
of 760 formulations containing 49 active pharmaceutical ingredients final product [405].
(APIs) and various excipients. The best ML model achieved an accuracy The entire AM pipeline is shown in Fig. 6. In our paper, design
of 92.8% in predicting amorphization and 96.0% in estimating chem- optimization for AM mainly focuses on shape computational optimiza-
ical stability. Feature importance analyses revealed critical processing tion [406] adopting machine learning methods. This can include op-
parameters and material attributes for accurate predictions. The study timizing the shape [407], topology [408] of the part to improve its
highlights the potential of ML in facilitating the development of chem- strength, reduce its weight or improve its functionality. According to
ically stable ASDs through HME, reducing the laborious trial-and-error the pipeline of AM, the process of computational optimization is di-
vided into the following steps:
approach traditionally used in ASD development. Chernyavsky et al.
[403] developed a heteroscedastic Gaussian process (HGP) model to
• Input optimization including macrostructure/material (microstruc-
predict microstructural properties, specifically amorphicity, of a glass-
ture) optimization [409–413]. AM processes, such as 3D printing,
forming alloy fabricated through laser powder bed fusion. The HGP
can produce intricate geometries that are difficult or impossible to
model accurately predicted mean amorphicity and quantified uncer-
achieve using traditional manufacturing methods. Input optimiza-
tainty, facilitating the assessment of dataset quality and identification
tion can help take advantage of these capabilities by designing
of underlying physical phenomena. The results of HGP prediction are structures that are optimized for the specific AM process being used
demonstrated in Fig. 5. while obtaining higher stiffness.
Additionally, machine learning also has a broad application in the • Segmentation/package optimization [414,415]. These techniques
material classification of AM. Schmidt et al. [404] utilized acoustic involve breaking down a large part into smaller segments or opti-
emissions for material classification and quality monitoring in laser mizing the layout of multiple parts within a build volume to achieve
AM. Acoustic emissions were recorded and processed using fast Fourier better use of the available space.
transformation. Machine learning models were trained and achieved • Optimization of the print direction [416–418]. It is to optimize the
high accuracy in classifying materials (up to 0.99) and process quality orientation of a part during printing, with the goal of improving its
(up to 0.81). Vrábel et al. [302] explored the use of laser-induced break- mechanical properties and reducing the need for support structures.
down spectroscopy (LIBS) for material analysis in SLM. The elemental • Slice optimization [419–421]. It involves breaking down a 3D
composition of raw materials and constructed parts is obtained from model into individual layers for printing. The goal of slice optimiza-
LIBS measurements. Multivariate data analysis algorithms, including tion is to achieve better print quality and reduce printing time and
principal component analysis and support vector machines, are em- material usage.
ployed for data processing and classification. • Path planning optimization [422–425]. It involves analyzing the ge-
In conclusion, material analysis plays a crucial role in AM to op- ometry of the part being printed and selecting the optimal path for
the print head to follow during printing. The goal of path planning
timize the manufacturing process and ensure high-quality results. Ma-
optimization is to reduce printing time, improve print quality, and
chine learning techniques have been applied to analyze and understand
reduce the wear and tear on the print head.
the relationships between material chemistry, characteristics, and per-
formance in AM. Current research in material analysis has primarily
3.3.1. Geometric processing based on machine learning
focused on metal powder characterization, while further exploration
This section delves into the remarkable synergy between geomet-
is needed to leverage machine learning for material composition anal- ric processing and ML, focusing on key optimization domains. From
ysis and modeling the relationships between material properties and package segmentation to path planning, the integration of ML not only
final part performance. Additionally, machine learning has been applied enhances these processes but also marks a paradigm shift in how AM
in predicting material properties and classifying materials in AM pro- leverages computational intelligence to propel manufacturing excel-
cesses, offering potential improvements in efficiency and quality con- lence. Through a comprehensive exploration of package/segmentation
trol. The integration of machine learning with integrated computational optimization, printing orientation optimization, slicing optimization,
materials engineering approaches has shown promise in advancing the and path planning optimization, this section unveils the transformative
understanding and optimization of AM processes. power of ML-driven geometric processing, underscoring its potential to

12
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Fig. 5. Predicting materials characteristics and their uncertainty by HGP [403]. (a). HGP model predictions for mean values of amorphicity and its total uncertainty.
(b). Two-dimensional contour maps of HGP model predictions for mean values of amorphicity and its total uncertainty. (c). Position-resolved aleatoric and epistemic
uncertainties predicted by the HGP model.

13
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Fig. 6. Design process of additive manufacturing technology.

revolutionize the landscape of AM. Fig. 7 illustrates a selection of rep- printed is very important for reducing energy consumption. Based on
resentative research examples in the field. this, the paper [434] explored the use of twelve machine learning algo-
rithms to optimize energy consumption in AM processes, specifically the
3.3.1.1. Package/segmentation optimization For larger objects, it may FDM process. The Gaussian process regressor model was found to esti-
be possible to print the object in smaller sections and then assemble mate energy consumption with high accuracy. Shi et al. [428] proposed
them together. This requires breaking down the object into smaller a reinforcement learning (RL) framework for optimal build orientation
pieces and then printing each section separately. Once all sections are in 3D printing to address local optimization and trial and error issues.
printed, they can be assembled to create the final object. However, this The proposed method outperforms existing approaches and can quickly
requires careful planning and assembly to ensure that the pieces fit to- discover optimal global solutions and generalize beyond the training
gether correctly. Object segmentation from AM is an extensive task in environment.
deep learning. Guo et al. [426] proposed a new method to find the opti-
mal build orientation for each submodel, based on a modified curvature 3.3.1.3. Slicing optimization Slicing optimization is used to convert the
shift strategy and weight factors for every flat facet of the regions. The CAD model into a series of layers depending on the type of slicing strat-
aim is to improve surface quality, reduce printing costs, and eliminate egy used. The most commonly used slicing method is planner slicing
the need for support structures. Li et al. [415] proposed a 3D model which slices the model into parallel layers. The thickness of each layer
segmentation method based on deep learning to improve the surface can be adjusted to control the objects’ details. By optimizing the slic-
quality and reduce the support structure of 3D printed models. The ing parameters, it is possible to achieve higher quality and accuracy
proposed method uses pre-segmentation and an Affinity Propagation in the final 3D print, while minimizing material usage and print time.
clustering method to segment a product model into several parts, which Checa et al. [435] proposed a strategy to select the best cutting tool de-
is validated by printing samples. Ng et al. [432] explored the integra- sign and cutting parameters using experimental tests, machine learning
tion of deep learning into 3D bioprinting, from image segmentation to modeling, and virtual reality visualization to optimize power consump-
tissue maturation. The paper discusses practical guidelines and poten- tion. Tamir et al. [436] presented open-loop and closed-loop machine
tial applications, highlighting synergistic interactions between biology, learning models to monitor the effects of processing parameters on the
materials, and computational design. quality of 3D printed parts, allowing a closed-loop control system in
AM. Jin et al. [437] proposed machine learning algorithms to precisely
3.3.1.2. Printing orientation optimization Optimizing the orientation of detect and localize in-plane printing conditions such as over-extrusion
a 3D model for printing involves considering various factors such as sta- and under-extrusion in fused filament fabrication. The system provides
bility, support requirements, surface details, and print time. To ensure real-time defect information and has the potential for automated con-
maximum stability during printing, the orientation should be chosen trol and correction.
carefully. Similarly, minimizing the need for support structures is im-
portant, as is ensuring that the surface details of the final product are as 3.3.1.4. Path planning optimization 3D printing path planning based on
smooth as possible. Finally, minimizing print time is essential to keep machine learning involves developing a model that learns from previ-
costs low and avoid quality problems. ous 3D printing jobs to suggest the most efficient printing path for a
Based on these factors, the 3D model can be rotated and adjusted new object. To do this, the machine learning model analyzes the char-
in the 3D printing software to find the optimal orientation for printing. acteristics of the object being printed, such as its shape and size, and
The software can also simulate the printing process to check for stability takes into account factors such as the type of material being used, the
and identify areas that may require additional support structures. Zhang printer’s capabilities, and any design constraints. It then uses this in-
et al. [427] proposed a perceptual model to find optimal 3D printing ori- formation to generate a printing path that minimizes printing time,
entations that avoid placing supports in perceptually significant regions, reduces material waste, and improves the overall quality of the final
resulting in fewer surface artifacts. The model is formulated as a com- product. Patrick et al. [438] proposed the use of reinforcement learn-
bination of metrics that include support area, visual saliency, preferred ing, a type of machine learning, to generate toolpaths for 3D printing.
viewpoint, and smoothness preservation. The study demonstrates the RL involves two agents, the actor and the critic, learning to maximize
performance of this model on natural and man-made objects. Malviya a score based on the actions of the actor in a defined space, and in
et al. [297] proposed a machine learning-based framework to optimize this context, the actor will learn to find the optimal toolpath to re-
the build orientation in FDM components by maximizing the minimum duce printhead lifts and print time. Fok et al. [439] introduced using
Factor of Safety. An ANN coupled with a Bayesian algorithm is used for artificial neural networks to improve the efficiency of the optimiza-
acceleration, and experiments show effective optimization with limita- tion process. The optimization process is computationally intensive, and
tions. Rezaei et al. [433] proposed an autonomous system that utilizes heuristics and metaheuristics are typically used to generate suboptimal
artificial intelligence to decide the build orientation of a shape based results. Ge et al. [430] described an intelligent path planning method
on similar parts previously encountered. Orientation of the part to be called Q-Path, which uses reinforcement learning to find optimal so-

14
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Fig. 7. Geometric optimization based on machine learning. (a). Segmentation methods using k-means, modified Cs and the Reeb graph method [426]. (b). Seg-
mentation methods using clustering [415]. (c). Optimal 3D printing orientations generated by MeshMixer (top) and Perceptual Models (bottom) [427]. (d) Support
optimization with different build orientations by random search, RotNet, genetic algorithm, and facet clustering [428]. (e). Zigzag toolpath along the vertical and
horizontal direction and selected temperature distribution (green point) and adaptive toolpath calculated by objective functions and selected temperature distri-
bution (green point) [429]. (f). The path plans using three algorithms in order: ZigZag-Path, Fleury-Path, and Q-Path [430]. (g). Reverse engineering of additive
manufactured composite part by toolpath reconstruction using imaging and machine learning [431].

15
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

lutions while adhering to 3D printing constraints, such as minimizing effectively conduct TO. Furthermore, the resolution of optimized struc-
printhead lifts and turns. Xie et al. [440] presented a tool path pat- tures remains constrained [458], thereby impeding the implementation
tern recommendation method based on deep learning neural networks of topology optimization in more intricate designs.
that utilize 3D point cloud segmentation and classification and transfer In this section, we will explore the application of machine learning
learning to identify object features and improve performance. Yana- in topology optimization from two aspects: macro-structure and micro-
mandra et al. [431] focused on using imaging and machine learning to structure topology optimization.
reverse engineer composite material parts made through AM. The aim is
to reconstruct not only the geometry, but also the tool path of 3D print- 3.3.2.1. Macro-structures topology optimization Macro-scale topology
ing, to prevent unauthorized production. The method achieves a high optimization is a design approach that seeks to optimize the overall
level of dimensional accuracy. Crockett et al. [441] proposed a varia- shape and layout of large structures, such as buildings or bridges, in
tion of k-means clustering for object division to reduce print time in 3D order to achieve a desired performance objective [459,460]. Machine
printing. The algorithm considers the balance of workload and discrete learning-based macro-scale topology optimization methods use algo-
build areas for parallel construction. The method is tested on several rithms that can learn from data to generate optimized designs based on
models to compare workload across the number of build points. Qin those data [461,462]. Machine learning has been developed to automat-
et al. [429] proposed adaptive toolpath generation algorithms for laser ically detect patterns in the data to accurately predict the distribution of
powder bed fusion to reduce part distortion. The algorithms are based material under different forces and boundary conditions. Recent years
on minimizing thermal gradients and include collision-free and smooth- have witnessed a number of studies devoted to applying ML techniques
ing constraints. Experimental results show that the proposed algorithms to solve topology optimization problems as illustrated in Fig. 8.
reduce distortion compared to traditional zigzag- and chessboard-based Some research work [97,98,469–472] reviewed the progress in us-
patterns and can be extended to fabricate complex 3D geometries. ing deep learning for photonic design and highlights key applications,
challenges, and perspectives. Sosnovik and Oseledets [463] adopted
3.3.2. Topology optimization for properties based on machine learning the convolution neural network to determine the distribution of the
Topology optimization is a layout optimization method to find the material. The deep learning method is used to speed up the whole
shape and material distribution of a structure to achieve the best perfor- optimization process. Lei et al. [464] adopted machine learning to
mance targeting minimizes weight, maximizes stiffness or strength, and achieve real-time structural topology optimization using an explicit
minimizes stress concentrations under certain constraints. It is a promis- framework and regression models, resulting in reduced training data
ing field in engineering and computational mechanics, which emerged and the potential for intuitive understanding. Examples demonstrate
in the 1980s [442] and 1990s [443]. However, the idea of optimizing effectiveness. Yu et al. [465] proposed a deep learning-based method
structures for maximum performance has been around for centuries. for predicting optimized structures without iterative schemes. Using an
When combined with AM and machine learning, topology optimization open source optimization code, a CNN-based encoder and cGAN net-
can become even more powerful, allowing for the creation of parts that work are trained to produce near-optimal structures with negligible
are optimized specifically for AM processes quickly and efficiently. computational time. Wang et al. [454] proposed a lightweight and high-
Machine learning has the potential to significantly enhance topology efficiency CNN, the convolutional block attention (Cba-U-Net) model,
optimization by providing more accurate and efficient optimization al- for predicting topology-optimized configurations in real time, achieving
gorithms, as well as enabling new types of design and material optimiza- accuracy rates of 91.42%, and being suitable for various optimization
tions. One benefit of using machine learning in topology optimization algorithms. It demonstrates the potential of combining deep learning
is the ability to incorporate large amounts of data into the design pro- with topology optimization for large-scale projects. Xing et al. [473]
cess. Machine learning algorithms can be trained on data from previous adopted an autonomous learning and prediction (ALP) scheme for ma-
designs and simulations to learn patterns and correlations that can be chine learning-assisted structural optimization (MLaSO) to reduce total
used to inform new designs. This can lead to faster and more accurate computational time while maintaining prediction accuracy. Four nu-
optimization results, as well as the discovery of new design possibili- merical examples demonstrate the benefits of integrating ALP with
ties that might not have been apparent through traditional optimization gradient-based MLaSO. Yoo et al. [474] presented a topology control
techniques. Another benefit of machine learning in topology optimiza- system for drone networks to optimize connectivity between drones
tion is the ability to optimize material distributions and designs for from the perspective of interference and energy consumption. The sys-
multiple objectives simultaneously. This is particularly useful in multi- tem uses reinforcement learning (DDPG) to adaptively conduct connec-
physics problems where there are multiple competing objectives, such tivity optimization, minimizing learning time by changing the number
as minimizing weight while maximizing stiffness or reducing cost while of learning steps. The proposed system was verified through simulation
maintaining performance. Machine learning algorithms can optimize experiments and theoretical analysis on various topologies consisting of
designs for these multiple objectives simultaneously, resulting in more multiple UAVs.
efficient and effective designs. Overall, the benefits of machine learning Overall, machine learning-based macro-scale topology optimization
in topology optimization include faster and more accurate optimiza- has the potential to significantly improve the design process for large
tion results, the discovery of new design possibilities, and the ability structures, but it requires careful consideration of computational re-
to optimize for multiple objectives simultaneously. Machine learning sources and validation to ensure the accuracy and feasibility of the
can also optimize the manufacturing process for topology-optimized de- resulting designs. Additionally, the incorporation of expert knowledge
signs, leading to more efficient and cost-effective manufacturing. and engineering principles is still necessary to ensure that the designs
Despite the promising potential of machine learning in topology are practical and can be implemented.
optimization, there are still several limitations that require attention
in current research. A notable limitation pertains to the scarcity of 3.3.2.2. Micro-structure topology optimization Meta-materials have
ML applications specifically tailored for AM in topology optimiza- unique properties that make them useful in a variety of applications,
tion. While deep learning techniques have made progress in generating including aerospace, biomedical, and marine engineering. However,
topologically-optimized designs, the majority of these efforts have con- designing meta-materials with desired properties can be challenging,
centrated on 2D structures [444–457]. Given that AM encompasses because of the complex and intricate structure of these materials.
three-dimensional fabrication, the extension of topology optimization Machine learning has been explored as a potential solution to this chal-
to 3D structures becomes imperative for fully capitalizing on the advan- lenge. Researchers have developed various methods to use machine
tages of AM. However, the introduction of an additional dimension in learning algorithms to generate meta-materials with specific prop-
the design space presents challenges for machine learning algorithms to erties, such as high refractive index or negative permittivity [475],

16
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Fig. 8. Topology optimization based on machine learning. (a). Results of the application of neural networks for topology optimization [463]. (b). Using ML-predicted
results as initial design for direct optimization [464]. (c). Deep learning for determining a near-optimal topological design without any iteration [465]. (d). Topology
optimization results of MMC and three deep learning models [454]. (e). Post-processed results of L-shaped beam design with machine learning combined topology
optimization for the functionally graded composite structure with cross-shaped fiber layout and the fiber-reinforced composite structure with fixed fiber volume
fraction [466]. (f). Cantilever model optimization and MBB beam model using machine learning combined topology optimization with 1, 30, 80, 200 steps [410]. (g).
The final bridge designs obtained from the standard and machine learning-based approaches for cases [444]. (h). Results for the MBB beam domain obtained through
the ML-driven topology optimization. [449] (i). Optimized results for topology optimization using neural networks [467]. (j). Machine learning-based parameter
tuning strategy for MMC-based topology optimization [468].

17
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

auxetic properties [476], maximum bulk/shear modulus [446], which costs. By identifying the optimal settings for process parameters, man-
are demonstrated in Fig. 9. These methods include the use of adversar- ufacturers can achieve better process efficiency, reduce material waste,
ial generative networks [477–479], deep neural networks [480–482], and minimize the need for post-processing operations. This can result in
and reinforcement learning [483–486]. significant cost savings, particularly for high-volume production runs.
While these approaches have shown promise, there are still some However, the optimization of process parameters in AM is a com-
limitations to using machine learning for meta-materials generation. plex problem. One of the challenges is that AM processes involve a
One limitation is the need for large amounts of training data to train multitude of variables that can impact the quality of the final product
the machine learning algorithms effectively. Another limitation is the [491]. Another challenge in optimizing process parameters is the need
computational resources required to run the algorithms, which can be to balance conflicting objectives [492]. There are often trade-offs be-
time-consuming and costly. Overall, the use of machine learning for tween different aspects of the print. For example, optimizing for print
meta-materials generation is an exciting area of research with promising strength may require sacrificing print speed or resolution. Similarly,
potential. Further advancements in this area could lead to the develop- optimizing for print quality may require sacrificing production speed
ment of new and improved meta-materials with tailored properties for or material usage. These trade-offs can make it difficult to find the op-
specific applications. timal set of parameters that balance all of the competing objectives.
Also, the range of materials and techniques that are used in AM is one
3.3.3. Section conclusion of the reasons for this complexity. Each material and technique has its
Design optimization is a pivotal aspect of AM, harnessing the unique own unique set of process parameters that need to be optimized for the
capabilities of AM technologies to produce stronger, lighter, and more best results. Another challenge in optimizing process parameters is the
efficient parts compared to traditional manufacturing methods. By lack of standardization in the industry [493]. While there are some gen-
adopting machine learning methods, researchers and industry profes- eral guidelines for process parameters, each manufacturer and printer
sionals have made significant advancements in shaping computational model may have different optimal settings. This means that optimiza-
optimization for AM. This optimization encompasses various steps, tion often requires a great deal of trial and error, as well as careful
including macrostructure/material (microstructure) optimization, seg- observation and analysis of the results. Besides, the optimization of pro-
mentation/package optimization, print direction optimization, slice op- cess parameters is not a one-time event. As materials, printer models,
timization, and path planning optimization. and techniques continue to evolve, the optimal process parameters will
Geometric processing based on machine learning has shown promis- also change. This means that manufacturers and researchers must con-
ing results in package/segmentation optimization, enabling the design tinually monitor and update their process parameter settings to ensure
of complex structures while reducing support structures and improv- that they are using the most up-to-date and effective methods. As a re-
sult, researchers are exploring various optimization methods to find the
ing surface quality. Moreover, machine learning has proven valuable
best set of parameters that achieve the desired trade-off between these
in printing orientation optimization, predicting optimal orientations
objectives.
to achieve stability, minimize support structures, and enhance sur-
In this section, we will thoroughly explore the concept of process
face details. Slicing optimization has benefited from machine learning
parameter optimization for AM, addressing it from three distinct but
techniques, leading to improved 3D print quality, reduced material us-
interconnected perspectives. Firstly, optimization objectives will be dis-
age, and shorter print times. Additionally, machine learning-based path
cussed, that is the specific goals and performance benchmarks we aim to
planning optimization has enabled the selection of efficient printing
achieve with optimization, as these factors directly influence the qual-
paths, further enhancing print quality and minimizing print time and
ity and applicability of the printed part. Secondly, optimization process
material consumption.
parameters in the AM process that can be adjusted to drive better out-
At the macrostructures level, machine learning-driven topology op-
comes will be examined in detail. The careful control and adjustment
timization has demonstrated the ability to optimize the overall shape
of these parameters is crucial to ensure optimal results. The overview
and layout of large structures with multiple objectives and constraints,
of optimized process parameters and optimization objectives in process
offering a powerful tool for engineering design. However, addressing
parameter optimization of AM is illustrated in Fig. 10. Lastly, optimiza-
computational complexity and ensuring interpretability remains a fo-
tion methods will be explored into the computational and mathematical
cus of ongoing research. On the microstructure front, machine learning
models or strategies employed to systematically determine the best
methods have been applied to design metamaterials with tailored prop-
combination of parameters that yield optimal results. This segment
erties, showcasing potential for various applications. Challenges, such
focuses on the role of advanced techniques like machine learning in
as the need for substantial training data and computational resources,
modeling the complex relationship between process parameters and the
persist and call for further investigation.
properties of the final part. These three components together form a
As AM and machine learning technologies continue to evolve, design
holistic approach to process parameter optimization in AM, providing
optimization is set to play an increasingly critical role in revolutioniz-
a nuanced understanding of how to achieve the best possible product
ing various industries. By combining the strengths of AM and machine
performance and quality.
learning, researchers aim to create innovative products with improved
performance and sustainability, advancing the realm of AM and its real-
3.4.1. Optimization objectives
world applications.
The main objectives of optimization in AM can be divided into three
principal categories: cost-efficiency, qualities, and properties. These cat-
3.4. Process optimization egories represent key targets that need to be achieved to ensure an
effective and high-quality production process. Each category covers a
Process parameter optimization (PPO) is a crucial aspect of AM that range of parameters that are subject to optimization in order to en-
can lead to significant benefits for manufacturers. First, the optimiza- hance production efficiency, product quality, and suitability for specific
tion of process parameters in AM can help improve the quality of the applications.
finished product [130]. This is because the right combination of pro- The first category, cost-efficiency, is concerned with making the
cess parameters can result in reduced porosity, improved surface finish, manufacturing process more economical and sustainable. Key fac-
and better dimensional accuracy. By optimizing process parameters, tors that need optimization in this category are the building time
manufacturers can ensure that their products meet the required spec- [305,306,313,324,326,331,333,494–508], energy consumption [341,
ifications, and are free from defects that may arise due to improper 342,501,509–516], and material usage [517]. By optimizing these fac-
process parameter settings. PPO can also help to reduce manufacturing tors, we can significantly increase production efficiency and reduce the

18
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Fig. 9. Materials design based on machine learning. (a). Metamaterial design with deep learning [476]. (b). Optimized results for the inverse machine learning
framework to optimize lightweight metamaterials [479]. (c). Inverse-designed spinodoid metamaterials using machine learning [487]. (d). New planar auxetic
metamaterial perforated with orthogonally aligned oval-shaped holes using machine learning [447]. (e). Inverse design of mechanical metamaterial based on shell
structures with customized loading curves based on machine learning and genetic algorithm [488]. (f). Pragmatic generative optimization of novel structural lattice
metamaterials with machine learning [489]. (g). Inverse design based on machine learning of auxetic metamaterial with zero Poisson’s ratio [490]. (h). Inverse
machine learning discovered metamaterials with record high recovery stress [448]. (i). Topology optimization of 2D metamaterials for the case of maximizing
the bulk modulus, minimizing the Poisson’s ratio, and maximizing the shear modulus [446]. (j). Machine learning-based prediction and inverse design of 2D
metamaterial structures with tunable deformation-dependent Poisson’s ratio [445].

19
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Fig. 10. Optimized process parameters and optimization objectives in process parameter optimization of additive manufacturing.

operational costs involved in manufacturing, which in turn makes the Lastly, the third category revolves around the properties of the man-
process more sustainable. ufactured products, with a particular focus on mechanical and thermal
The second category focuses on the quality of the products man- properties. These properties include tensile strength [290,317–320,327,
ufactured. This category encompasses parameters such as accuracy 332,334,501,518,528,541,542,557–580], shear strength [581], yield
[291,326,327,494,498,518–526], surface roughness [291,306,329,340, strength [518,558,567,581,582], flexural strength [558–561,565,566,
583–585], impact strength [558–561,565,566,586], elastic modulus
495,497,520,527–539], density [328,520,527,530,540–542], porosity
[518,528,558,562,563,567,568,581], hardness [509,528,552,573,587–
[320,321,331,520,543–551], microstructure [313,315,552,553], and
589], and fatigue resistance [293,564,590] among others. These at-
melt-pool appearance [283,292,554–556]. The optimization of these tributes determine the product’s durability, lifespan, and suitability for
qualities plays a crucial role in determining both the aesthetic and various applications. By optimizing these properties, we can ensure that
functional aspects of the final product. By enhancing precision in these the manufactured products meet the required performance standards in
areas, we ensure that products are manufactured to meet the highest their intended applications, providing long-lasting reliability and over-
standards and suit specific application requirements. all value.

20
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

3.4.2. Optimization process parameters aluminum alloy fabricated using LPBF. The framework involved the fab-
In this section, we will delve into the crucial aspect of optimizing rication of specimens, surface topography measurement, data extraction
process parameters in AM. The optimization of process parameters is and engineering, and the development of a deep neural network model.
essential for achieving high-quality, efficient, and cost-effective produc- Different scanning strategies and process parameters were considered,
tion. However, due to the diverse nature of various printing methods, and the inputs to the model were the AM process parameters. Lapointe
the parameters that require optimization vary for each AM category. et al. [634] focused on applying machine learning to optimize laser pro-
Hence, we will explore the optimization process parameters individu- cess parameters in AM. The irregular quality of parts produced through
ally for seven distinct AM categories. laser powder bed fusion was addressed, with a focus on dimensional in-
The optimization of the material extrusion process requires careful accuracies and defects in complex geometries. A data-driven approach
consideration of various parameters such as build orientation [499,500, was employed, utilizing experimental diagnostics to create a training
518,528,557,559,560,565,591–596] and layer thickness [498–500,518, dataset for supervised learning models. These models, a forward and an
525,528,559,560,562,565–567,581,591,592,595–598]. Also, parame- inverse model, were developed to predict photodiode signals and laser
ters like printing temperature [324,501,518], bed temperature [327], parameters based on track-wise geometry features. The inverse model
and air gap [499,500,559,560,591,592,595–597] play crucial roles in enabled the determination of required laser parameters for desired pho-
determining the characteristics of the part being manufactured. Param- todiode signals. The whole process and results are demonstrated in
eters like printing speed [498,518,525,566,597] have also been investi- Fig. 11.
gated for their impact on the extrusion process. Infill pattern and infill In directed energy deposition, laser power [635,636], feed rate
ratio play a significant role in both the cost and mechanical properties of [333,635,636], travel speed [333,635–637], interpass time [333], and
the final part [499–501,518,525,528,557–562,564,565,567,581,591– energy input [521,637] are among the parameters to optimize. Effec-
593,595–597,599–602]. Furthermore, support material and model inte- tive management of these parameters can control the quality of the
rior are also optimized for hardness, flexural modulus, tensile strength, deposited material and the subsequent cooling process, thus influenc-
and surface roughness [528]. Gao et al. [603] compared two optimiza- ing the mechanical properties and dimensional accuracy of the final
tion approaches–the Taguchi method and RSM. They identified four product. By tuning these parameters, manufacturers can enhance pro-
operating parameters and their interaction terms as control variables, cess efficiency and improve the performance characteristics of the final
and tensile strength and compressive strength as responses. The exper- products. Heat treatment also assumes a key role in the process op-
imental results showed that the conclusions regarding the significant timization of directed energy deposition. While directed energy depo-
ranking of parameters differed between the Taguchi method and RSM. sition distinguishes itself from powder bed fusion through its distinct
Mohamed et al. [604] examined the effects of FDM fabrication condi- method of material deposition and melting, the fundamental impor-
tions on the dimensional accuracy of cylindrical parts. A combination of tance of heat treatment remains. It contributes significantly to the
a second-order definitive screening design (DSD) and an artificial neural modulation of the microstructure, and by extension, the mechanical
network (ANN) was proposed for experimental design and prediction. properties of the produced parts. In the directed energy deposition pro-
The study successfully determined the optimum fabrication conditions cess, the feedstock material, typically in the form of powder or wire, is
for improved dimensional accuracies of cylindrical parts. melted using a focused thermal energy source such as a laser or elec-
For powder bed fusion, parameters like laser power [580,605–615], tron beam. The manipulation of heat treatment parameters during this
energy density [616,617], scanning speed [580,605,607,609,612,613, process can significantly alter the thermal gradients and cooling rates
615,618,619], hatch space [610,615,619], and powder layer thickness experienced by the material, thus influencing the microstructural evo-
[615,620,621] are pivotal to the success of the process. When consid- lution within the part. As a consequence, mechanical attributes such as
ering the process parameter optimization of powder bed fusion, heat tensile strength, hardness, ductility, and fatigue resistance can be finely
treatment emerges as a significant parameter that sets this method apart tuned. Much like in powder bed fusion, optimizing heat treatment in
from other kinds of AM. The role of heat treatment in the powder bed directed energy deposition is a complex and delicate task that requires
fusion process is a multifaceted one and demands careful consideration, a deep understanding of material science and the relationship between
as it has far-reaching implications on the properties and performance heat treatment conditions and resulting material properties. Strategies
of the final product. The heat treatment process in powder bed fu- such as preheating the substrate, controlling the interpass temperature,
sion plays an instrumental role in defining the microstructure of the and post-process heat treatment can all be deployed to manipulate the
manufactured parts, which is a determining factor of their mechanical material microstructure and achieve the desired mechanical properties
properties. The application of heat, both in terms of magnitude and du- [638–645]. Era et al. [646] applied machine learning algorithms (XG-
ration, impacts the atomic arrangement within the product, which can Boost and Random Forest) to predict the tensile behaviors of stainless
lead to variations in properties like tensile strength, hardness, ductility, steel 316L parts fabricated using laser-based directed energy deposi-
and resistance to wear and tear. Specifically, the appropriate control tion (L-DED). The process parameters, including laser power, scanning
and regulation of heat treatment can enhance the tensile strength of the speed, and layer thickness, were optimized to control the properties
printed parts, making them more durable and capable of withstanding of the fabricated parts. The models successfully predicted the tensile
higher mechanical stresses. This underscores the significance of heat properties, with XGBoost demonstrating superior accuracy compared to
treatment as a parameter in the powder bed fusion process and em- Ridge Regression and Random Forest. Jamnikar et al. [647] developed
phasizes its influence on the functional characteristics of the products a deep learning-based multi-modality convolutional neural network for
[622–630]. Liu et al. [631] developed a machine learning approach in-process geometry monitoring in wire-feed laser AM. The network
based on Gaussian process regression to optimize the LPBF process in was trained to estimate the geometric properties of the build bead us-
AM. The optimized processing window for manufacturing fully dense ing real-time molten pool sensing data. The effect of using temperature
AlSi10Mg samples was determined, resulting in previously unattainable data from different positions of the molten pool on the prediction per-
combinations of high strength and ductility. Gu et al. [632] utilized formance of the CNN model was analyzed. Pham et al. [648] proposed a
small dataset of 33 experiments to predict relative density and surface feedforward neural network surrogate model was developed for fast and
roughness of LPBF process. The ML models were improved by curat- accurate prediction of temperature evolutions and melting pool sizes in
ing the input data and using volumetric energy density as a universal metal bulk samples manufactured by the DED process. The surrogate
predictor. This approach reduced the mean absolute percentage error model was trained using high-fidelity data from finite element models
and increased the translation capability of the models across different and validated with experiments. The FFNN model achieved high ac-
machines and materials. Muhammad et al. [633] developed a DL frame- curacy, with 99% and 98% accuracy for temperature evolutions and
work to predict the process-induced surface roughness of AlSi10Mg melting pool sizes, respectively.

21
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Fig. 11. Photodiode-based machine learning for optimization of laser powder bed fusion parameters in complex geometries [634]. (a). Mean and signal-to-noise ratio
of the photodiode signal for single tracks printed at different laser powers and velocities. (b). Joint probability density function of the predicted and true photodiode
signal intensities. (c). Comparison of true and predicted photodiode signal for four tracks from the test set. (d). Comparison of the true and predicted laser powers
and velocities. (e). Laser power predicted by the inverse ML model for three layers of the test geometry. (f). Comparison of the photodiode signal on the first layer
above the overhang (layer 200) for a part printed with nominal laser parameters and a part printed with laser parameters predicted by the inverse model.

22
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

In the binder jetting process, various process parameters play pivotal Indeed, process parameter optimization in AM is often approached
roles in influencing the quality and characteristics of the final product. as a multi-objective optimization problem [660]. The multi-objective
Layer thickness and binder saturation are two of the most frequently op- optimization problem is a class of optimization problems that involve
timized parameters, which significantly impact the structural integrity optimizing multiple, often conflicting, objectives simultaneously. In the
and dimensional accuracy of printed parts [328,523,527,583,649–654]. context of process parameter optimization in AM, a multi-objective op-
Other parameters, such as roll speed and the feed-to-powder ratio, also timization problem arises when multiple quality measures need to be
require careful calibration for optimal performance [583]. In addition, optimized simultaneously. For example, in metal AM, the quality mea-
heater power ratio affects the drying time, which in turn impacts the sures include porosity, surface roughness, tensile strength, and fatigue
part strength and the rate of production [328,523]. The properties of strength. Each quality measure is typically optimized individually in a
the granules used, such as particle size and morphology, and the pow- single-objective optimization problem. However, in practice, these qual-
der bed’s characteristics, including roughness and packing density, also ity measures are often interdependent, and optimizing one measure may
significantly affect the printing process and the final part characteris- lead to a degradation in another measure. Thus, optimizing these mea-
tics [527,655]. Lastly, printing speed and spread speed are parameters sures simultaneously requires a multi-objective optimization problem.
that influence the printing efficiency and surface finish of the prod- The typical multi-objective optimization problem is given in Equation
uct [328,649]. By optimizing these process parameters, manufacturers (1).
can significantly enhance the quality and efficiency of binder jetting
∑
processes. Onler et al. [328] developed a robust approach to identify min 𝐹 (𝑥) = 𝑇𝑖=1 𝜔𝑖 𝑓𝑖 (𝑥) 𝑥 ∈ Ω ∈ ℝ𝑛
optimal process parameters for binder jetting of Co-Cr-Mo alloy. Exper- s.t.𝑔𝑖 (𝑥) ≤ 0 𝑖 = 1, 2, ..., 𝑘 (1)
imental investigations and machine learning techniques were used to ℎ𝑖 (𝑥) = 0 𝑖 = 1, 2, ..., 𝑘
establish the relationship between process parameters and output qual-
where 𝑓𝑖 (𝑥) is the objective function that needs to be optimized subject
ity, including density, dimensions, and surface quality of green parts.
to the constraints, and 𝑔𝑖 (𝑥) ≤ 0 and ℎ𝑖 (𝑥) = 0 are constraints that are
Artificial neural networks were employed to model this relationship,
required to be satisfied (these are called hard constraints).
and the weighted k-NN algorithm was used for qualitative classification.
Multi-objective optimization methods can be broadly categorized
A genetic algorithm-based multi-objective optimization was applied to
into three groups: priori methods, posteriori methods, and interactive
determine the optimum process parameters, achieving over 90% ac-
methods [661].
curacy in predicting part quality. Jimenez et al. [656] conducted a
Priori methods are also known as pre-decision methods or offline
parametric study to investigate the influence of seven process inputs on
methods [662]. These methods involve solving the multi-objective op-
the relative densities of printed alumina parts. Multivariable linear and
timization problem before the decision-maker is faced with making
Gaussian process regression models were developed to predict green
a choice. The solutions obtained from the pre-decision methods are
densities based on process inputs. The study identified recoat speed and
presented to the decision-maker, who then chooses the best solution
oscillator speed as key parameters affecting green densities.
based on their preferences. Some examples of priori methods include
Process parameter optimization research in vat photopolymeriza-
weighting methods [663], goal programming [664], and lexicographic
tion, material jetting, and sheet lamination is relatively scarce com-
methods [665].
pared to other techniques, possibly due to perceived limited opportu-
Posteriori methods are also known as post-decision methods or on-
nities for substantial quality improvement [657–659]. These methods
line methods [666]. These methods involve solving the multi-objective
already yield high-resolution products with smooth finishes, especially
in the realm of polymers. Consequently, the quality benefits achievable optimization problem in real-time as the decision-maker makes choices.
through process optimization might not be seen as profound enough The solutions obtained from the post-decision methods are presented
to justify the extensive time and resources required for in-depth study. to the decision-maker, who then chooses the best solution based on
The inherent limitations of these technologies, such as the restricted their preferences. Some examples of post-decision methods include non-
range of materials applicable in vat photopolymerization and mate- dominated sorting genetic algorithm-II (NSGA-II) [306,318,515,667],
rial jetting, or the weak interlayer bonding in Sheet Lamination, might and Pareto archive evolution strategy (PAES) [668]. Peng et al. [515]
also inhibit broad-scale enhancements. Thus, research efforts may be developed a predictive model for specific energy consumption and
directed more towards other aspects like new material development metallic powder usage rate in laser cladding using three different meth-
or integration with other techniques rather than process parameter ods. The comparison revealed that the method that demonstrated the
optimization, which might be seen as offering minimal potential for sig- highest fitting performance was the integrated Tabu search and GEP.
nificant quality advancements in these specific techniques. Segura et al. The predictive models were then used to determine the Pareto front us-
[336] developed a machine-learning approach to predict droplet behav- ing NSGA-II, which provided the optimal set of processing parameters.
iors in MJ AM. The Network of Tensor Time Series was used to capture Results showed that NSGA-II was effective in maximizing energy and
the underlying relationships shared across diverse material and pro- metallic powder efficiency. This study emphasizes the importance and
cess parameters. The Tensor Graph Convolutional Network and Tensor effectiveness of NSGA-II in the process optimization of AM. Asadollahi-
Recurrent Neural Network were employed to capture cross-linked and Yazdi et al. [306] proposed a method to simultaneously analyze critical
temporal relationships, respectively. The features extracted from these attributes and drawbacks of AM and optimize the manufacturing pa-
networks were fed into a Multilayer Perceptron for predicting future rameters for AM products. The study focuses on FDM and formulates
droplet behaviors. The proposed methodology demonstrated accurate a multi-objective optimization problem with layer thickness and part
and efficient predictions for both known and unknown material/process orientation as decision variables. NSGA-II is utilized to find optimal
parameters using simulated and experimental droplet evolution data. solutions. The approach is demonstrated through a case study and em-
phasizes the effectiveness of NSGA-II in AM process optimization. Meng
3.4.3. Optimization algorithms et al. [318] focused on optimizing multilayer bio-inspired sandwich
Optimization algorithms play a critical role in process parameter structures of Ti-6Al-4V alloy fabricated by SLM. The results showed that
optimization for AM [285]. After the relationships between process NSGA-II can effectively optimize the bio-inspired sandwich structures
parameters and output qualities are determined, the optimization algo- with a relative error rate of less than 10%. The study further demon-
rithms need to be applied to figure out the optimal solutions to obtain strated the comprehensive mechanical performances of the bio-inspired
the desired qualities. The quality and efficiency of an AM process can be sandwich structures and identified the optimal configuration as the two-
improved by optimizing the process parameters that control the manu- layered structure under cross-arranged configurations, emphasizing the
facturing process. effectiveness of NSGA-II in optimizing AM structures.

23
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Interactive methods involve the decision-maker actively participat- to find the optimal parameter settings. Additionally, the lack of stan-
ing in the optimization process [669]. These methods involve present- dardization in the industry adds complexity to the optimization process,
ing the decision-maker with a set of solutions and asking them to as optimal settings may vary among manufacturers and printer models.
choose the best solution based on their preferences. The optimization Furthermore, the range of materials and techniques used in AM fur-
process is then updated based on the decision-maker’s choices. Some ther contributes to the complexity of process parameter optimization.
examples of interactive decision-making methods include interactive Each material and technique have its own unique set of process param-
particle swarm optimization (PSO) [342,533], interactive genetic algo- eters that must be optimized for optimal results. Additionally, process
rithms (GA) [313,332,341,670], and interactive evolutionary strategies parameter optimization is an ongoing process, as materials, printer
[671–673]. Selvam et al. [533] used PSO to optimize FFF printing pa- models, and techniques evolve over time, necessitating continual mon-
rameters for improving surface quality and reducing the printing time itoring and updates to achieve the most effective methods.
of ABS polymer. The study employed RSM and PSO techniques to pre-
dict the optimum process parameter values. Through the PSO method, 3.5. Defect detection & real-time monitoring
high surface quality with a minimum printing time of 24 minutes was
achieved, showing finer optimal values compared to RSM. Kumar & Defect detection and real-time monitoring in AM hold significant
Maji [332] discussed the optimization of WAAM through the use of importance across various aspects of the manufacturing process [674].
multi-objective GA to minimize void and excess material. The focus It ensures quality assurance by identifying and addressing defects that
is on developing a single-bead geometry model using RSM and Box- can compromise the structural integrity and functional performance of
Behenken DoE for the deposition of SS304L. The study found that the AM parts [675]. Moreover, defect detection enhances reliability and
optimal processing conditions and deposition planning determined by safety, particularly in industries where structural integrity is critical
multi-objective GA led to the minimum void and maximum material [676]. By minimizing defects, manufacturers can achieve cost savings
yield. Additionally, it was discovered that optimal bead sizes and de- through reduced material waste and improved efficiency. Defect anal-
grees of overlapping are different for different geometries. The study ysis also provides valuable insights for process optimization [677], en-
also highlights the benefits of double wire feed deposition, which can abling manufacturers to fine-tune parameters, materials, and designs.
deliver a higher deposition rate and superior mechanical properties of Compliance with regulatory standards is ensured through defect detec-
the deposited material compared to single wire feed deposition. tion, and continuous improvement is fostered by identifying areas for
Each of these categories of multi-objective optimization methods has enhancement. Ultimately, defect detection plays a crucial role in en-
its advantages and disadvantages. Priori methods can be more compu- hancing the overall quality, reliability, cost-efficiency, and compliance
tationally efficient as they solve the optimization problem only once. of additively manufactured components.
However, they may not provide the decision-maker with enough flex-
ibility to make the best choice. Posteriori methods provide more flex- 3.5.1. Defect types
ibility to the decision-maker, but they can be more computationally AM processes are susceptible to various types of defects that can im-
expensive as the optimization problem is solved in real-time. Interac- pact the quality and performance of manufactured components. These
tive methods provide the most flexibility to the decision-maker but can defects can arise at different stages, including material preparation, part
be time-consuming as the decision-maker must actively participate in fabrication, and post-processing. Common defect types in AM include
the optimization process. geometric defects, surface defects, porosity and voids, inadequate fu-
sion or bonding, inclusions and contaminants, residual stresses, and mi-
3.4.4. Section conclusion crostructural defects [674,678]. Fig. 12 provides an overview of defect
In this section, the concept of process parameter optimization in detection within the context of AM. The explanations for each defect
AM is explored comprehensively from three interconnected perspec- are demonstrated below:
tives. Firstly, the optimization objectives, encompassing cost-efficiency,
qualities, and properties, are discussed. These objectives cover a range • Geometric defects: Warping [679], distortion [680], overhangs
of parameters that need optimization to enhance production efficiency, [681,682], undercuts [682], and shape inaccuracies affecting the
product quality, and suitability for specific applications. Secondly, the geometric accuracy and dimensional deviations of printed parts.
optimization of process parameters in the AM process is examined in • Surface defects: Rough surfaces [683,684], pits [685,686], cracks
detail. Various parameters, such as build orientation, layer thickness, [687,688], and surface porosity [689] impacting the surface finish
printing temperature, and infill pattern, are explored for their impact and smoothness of printed parts.
on the characteristics of the manufactured part. Lastly, optimization • Porosity and voids: Small voids or air pockets within the printed
methods, including computational and mathematical models, are ex- material, which can negatively impact mechanical properties and
plored, with a focus on advanced techniques like machine learning for structural integrity [547,690–695].
modeling the complex relationship between process parameters and the • Inadequate fusion or bonding: Insufficient fusion or inadequate
properties of the final part. bonding between successive layers or material interfaces, leading
Additionally, we delve into the crucial aspect of optimizing process to weak or delaminated regions within the part [696–699].
parameters in AM to achieve significant benefits for manufacturers. The • Inclusions and contaminants: Foreign particles or impurities incor-
optimization of process parameters plays a vital role in improving the porated into the printed material during manufacturing, causing
quality of the finished product, including reducing porosity, enhancing defects like weak spots or reduced properties [700–702].
surface finish, and improving dimensional accuracy. This optimization • Residual stresses: Stresses remaining within printed parts due to
helps ensure that products meet required specifications and are free rapid heating and cooling cycles, potentially resulting in distortion,
from defects arising from improper process parameter settings. More- warping, or cracking [493,680,703–706].
over, process parameter optimization contributes to reducing manufac- • Microstructural defects: Variations in grain size, phase transforma-
turing costs by achieving better process efficiency, minimizing material tions, or uneven distribution of alloying elements, affecting material
waste, and reducing the need for post-processing operations. properties and performance [547,707–710].
However, optimizing process parameters in AM presents several
challenges. AM processes involve numerous variables that impact the 3.5.2. Data types
final product’s quality, and balancing conflicting objectives poses a sig- Defect detection in AM relies on the analysis of various signals,
nificant challenge. Trade-offs between different aspects of the print, which can be categorized into two main types: image-based signals
such as strength, speed, resolution, and quality, can make it challenging and sensor-based signals [711]. Image-based signals involve visual data,

24
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Fig. 12. Machine learning assisted defect detection & real-time monitoring in additive manufacturing.

such as images or videos, that capture surface irregularities, cracks, or ing any anomalies present. Advanced image processing algorithms are
other visual defects. These images can be acquired using high-resolution then applied to these images, enabling the detection and characteriza-
cameras or microscopy techniques [712,713]. Advanced image process- tion of defects with high accuracy and precision. By leveraging image-
ing algorithms can then be applied to analyze the images, detect de- based signals, AM processes can be monitored in real-time, enabling
fects, and assess their severity. Image-based defect detection techniques early detection of defects and facilitating quality control throughout
provide valuable information about surface quality and visual anoma- the manufacturing process. The utilization of image-based defect detec-
lies [678]. Sensor-based signals, on the other hand, involve the use of tion techniques provides valuable insights into the visual aspects of the
various sensors to capture different aspects of the manufactured part. printed parts, ensuring the overall quality and integrity of the manufac-
Thermal imaging sensors can detect variations in temperature distribu- tured components. Scime & Beuth [314] utilized a high-speed camera
tion, which may indicate inadequate fusion or cooling issues [714–716]. to monitor the morphology of LPBF melt pools in the Inconel 718 mate-
Acoustic sensors can capture acoustic emissions or ultrasonic waves to rial system. Computer Vision techniques were employed to construct a
assess the structural integrity of the part, detecting anomalies such as scale-invariant description of melt pool morphology, and unsupervised
voids or delaminations [277,713,717–720]. X-ray or CT scanning em- machine learning was applied to differentiate between observed melt
ploys sensors to capture internal images, revealing defects like porosity pools. In-situ signatures indicating flaws were identified through the
or incomplete fusion [721–728]. Mechanical testing sensors, such as linkage of ex-situ and in-situ morphology. Supervised machine learn-
tensile testers or hardness testers, can measure the mechanical proper- ing was then used to classify melt pools observed during the fusion of
ties of the part, detecting deviations from the desired material behavior. non-bulk geometries, such as overhangs. Zhang et al. [734] applied a
deep-learning-based approach to detect porosity in laser AM. A high-
3.5.3. Current research speed digital camera captured melt-pool data during the process, and
As mentioned before, defect detection and real-time monitoring are CNN models were trained to analyze the features of the melt-pool and
one of the most popular research domains for the application of ma- predict porosity attributes in the deposited specimens. The developed
chine learning in AM. Therefore, in this section, we will explore the image processing tools automated the extraction of porosity information
current research for defect detection and real-time monitoring accord- from raw quality inspection data. The compact CNN models achieved a
ing to the data type used in the process. Some representative research high classification accuracy of 91.2% for porosity occurrence detection
is demonstrated in Fig. 13. and demonstrated predictive capability for micro pores. Furthermore,
the models accurately predicted local volume porosity with a low root
3.5.3.1. Image-based signals Image-based signals in defect detection of mean square error, showcasing their effectiveness for both high and
AM play a crucial role in capturing visual data that reveals surface ir- low-porosity specimens. Wright et al. [735] developed a novel opti-
regularities, cracks, and other visual defects. High-resolution cameras mization framework using computer vision and DL for the defect de-
or microscopy techniques are employed to acquire detailed images or tection and optimization of extrusion AM of thermoset composites. The
videos of the manufactured parts. These visual data serve as a rich framework dynamically adjusted printing parameters during extrusion
source of information for analyzing the surface quality and identify- to achieve optimal results. A DL-integrated extrusion AM system was

25
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Fig. 13. Current research for defect detection and real-time monitoring in additive manufacturing based on machine learning. (a). Clustering result showing surface
defects in 3D with in-situ point cloud processing and machine learning [729]. (b). Confusion matrix for deep learning-based quality identification method for metal
AM process [730]. (c). Confusion matrix for vision-based melt pool monitoring for wire-arc additive manufacturing using deep learning method [731]. (d). Confusion
matrix for A hybrid deep learning model of process-build interactions in additive manufacturing [732]. (e). 3D fitted surface of five ML algorithms for geometrical
defect detection on additive manufacturing parts with curvature feature and machine learning [733].

26
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

used, along with in-situ imaging and high-accuracy CNN, to determine monitoring and quality control of AM processes, ensuring the pro-
the ideal process parameters for composite AM. duction of high-quality parts with minimal defects. Kononenko et al.
[738] designed a situ crack detection in LPBF fabricated parts. Ma-
3.5.3.2. Sensor-based signals Sensor-based signals can provide parame- chine learning models were developed to differentiate crack AE events
ters such as temperature, pressure, vibration, sound, or electromagnetic from background noise sound, achieving high classification accuracy
signals during the AM process. In the context of defect detection in AM, of up to 99%. Wang et al. [739] proposed an in situ quality moni-
sensor-based signals provide valuable information for identifying and toring method, combining machine learning with an improved vari-
characterizing defects and abnormalities. ational modal decomposition (VMD) technique. The AE signals were
adaptively decomposed using VMD with adjusted parameters based on
3.5.3.2.1. Thermal imaging By capturing thermal images of the printed the whale optimization algorithm and average energy entropy. Xu et
part or the surrounding environment, thermal imaging sensors provide al. [740] developed a real-time monitoring system, combining AE and
valuable information for defect detection and quality control. laser scanning technology, to detect warpage defects during the print-
One of the key applications of thermal imaging in defect detection is ing process. The AE signal captured variations in vibration, indicating
the identification of inadequate fusion or incomplete melting. Thermal changes in internal tension and reflecting real-time part characteristics.
anomalies, such as cold spots or hot spots, can indicate regions where The AE signal was processed, and frequency and time domain features
the material has not properly fused or melted, leading to weak or de- were extracted for training machine learning models. Laser scanning
fective bonds between layers. By analyzing thermal images, machine depth images were used to quantify dimensional changes and measure
learning algorithms can be employed to identify and classify these ar- warpage. Machine learning models, including support vector machine,
eas of inadequate fusion, allowing for timely intervention or adjustment naïve Bayes classifier, and decision tree, were employed to accurately
of process parameters to ensure the production of defect-free parts. Tian identify warpage in real-time. Shevchik et al. [718] utilized an AE sen-
et al. [736] described a deep learning-based data fusion approach for sor to capture acoustic signals during the selective laser melting process,
predicting porosity in LBAM. The melt pool thermal history, captured with intentionally varied process parameters to induce different levels
by high-speed thermal imaging, was used to monitor the process and of pore formation in the workpiece. The collected AE signals were di-
detect potential defects. Two deep-learning neural networks, PyroNet vided into training and testing datasets, and the relative energies of
and IRNet, were developed to correlate in-process pyrometry images narrow frequency bands were extracted as acoustic features. A spec-
and sequential thermal images from an infrared camera with layer-wise tral convolutional neural network classifier was trained to differentiate
porosity. The predictions from PyroNet and IRNet were fused at the de- between different quality levels based on these features. The classifier
cision level to improve the accuracy of porosity detection. Khanzadeh et achieved confidence levels between 83% and 89% in quality classifica-
al. [309] developed a real-time porosity prediction method using mor- tion.
phological characteristics of the melt pool boundary, obtained through 3.5.3.2.3. Optical emission Optical emission signals play a crucial role
functional principal component analysis (FPCA). Thermal imaging was in the defect detection of AM. These signals are obtained through opti-
used to capture the time-varying melt pool signal, which was labeled cal sensors that capture the emission of light or other electromagnetic
as either pores or normal melt pools using X-ray tomography. Super- radiation during the manufacturing process. By analyzing the optical
vised learning methods were employed to build a black-box model that emission characteristics, such as intensity, wavelength, and spectral
predicts the probability distribution of porosity based on melt pool char- patterns, it is possible to identify various defects and anomalies in the
acteristics. printed parts [219,741,742]. Optical emission signals can provide valu-
In addition to inadequate fusion, thermal imaging can also help able insights into phenomena such as overheating, improper melting,
identify cooling-related issues during the AM process. Rapid cooling or lack of fusion, and surface irregularities. Machine learning algorithms
uneven cooling can result in thermal gradients across the printed part, can be applied to process and analyze optical emission data, enabling
leading to residual stresses, warping, or distortion. By capturing thermal automated defect detection and classification. This enables real-time
images at different stages of the cooling process, thermal imaging sen- monitoring and quality control during AM, helping to ensure the pro-
sors can detect and visualize these temperature gradients, enabling the duction of high-quality parts with minimal defects. Atwya & Panoutsos
identification of potential defects or areas of concern. This information [691] developed a data-driven neural network framework to achieve
can guide the optimization of cooling strategies or the implementation in-situ micro-porosity localization in laser powder bed fusion using
of post-processing techniques to mitigate these defects and improve the exclusively within hatch stripe optical emission data. The proposed
overall quality of the printed parts. Estalaki et al. [737] developed ML method leveraged prior-guided neural networks to incorporate process
models to predict micropore defects in LPBF stainless steel materials us- physics and utilize nominal data. Ren et al. [352] developed a novel un-
ing in-situ thermographic data. Two key features, the time above the supervised recognition model for in-situ defect detection in DED. The
apparent melting threshold and the maximum radiance, were used as model, which combines a long short-term memory-based autoencoder
inputs for the ML models. The models were trained and tested for binary (LSTM-Autoencoder) and k-means clustering, analyzes the optical emis-
classification, considering the state of each voxel as either defective or sion collected during the DED process to extract features and classify
normal. The inclusion of thermal features from neighboring voxels im- the deposition quality. Petrich et al. [743] proposed a comprehensive
proved the prediction accuracy. concept and validation scheme for inter-layer flaw detection in powder
bed fusion using supervised machine learning. The research established
3.5.3.2.2. Acoustic emission Acoustic emission (AE) signals play a cru- a statistical correlation between the multi-modal sensor footprint col-
cial role in the defect detection of AM. By utilizing acoustic sensors, lected during the build process and the presence of flaws identified
the emitted acoustic waves or vibrations can be captured and analyzed through post-build X-ray Computed Tomography scans. Various sen-
to assess the structural integrity of the printed part. These sensors are sor modalities, including layerwise imagery, acoustic and multi-spectral
sensitive to the acoustic emissions produced during the manufacturing emissions, and scan vector trajectory information, were integrated us-
process and can detect anomalies such as voids, cracks, or delamina- ing data registration techniques. A neural network was employed to
tions. The AE signals provide valuable information about the internal fuse the modalities and distinguish flaws from nominal build conditions
structural quality of the part, allowing for early detection of defects using only in-situ data.
that may compromise its performance or durability. Advanced signal
processing techniques, such as pattern recognition and machine learn- 3.5.4. Control and real-time intervention
ing algorithms, can be applied to analyze the AE signals and distin- For AM, the ability to control and intervene in real-time during the
guish between normal and defective conditions. This enables real-time manufacturing process is crucial for ensuring high-quality outputs and

27
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

optimizing performance. Optimization, defect detection, and real-time small-to-medium batch production, offering superior time and cost effi-
monitoring are prerequisites for effective control and intervention. Op- ciency compared to conventional manufacturing processes.
timization and real-time monitoring lay the groundwork by ensuring However, it is essential to exercise caution when considering the
that the manufacturing process is well-understood and that any devia- seemingly evident benefits that AM may offer. There are instances doc-
tions or anomalies can be detected promptly. These steps are essential umented in the literature where AM does not align perfectly with the
to gather the necessary data and insights that enable informed control notion of being a wholly “green” technology. Here are some examples:
decisions. Gunasegaram et al. [357] discussed the application of ma- (1) Economies of scale typically favor conventional mass production,
chine learning in closed-loop control (CLC) strategies for metal AM. leading to lower incremental costs compared to one-off AM produc-
It highlights the limitations of traditional deterministic and rule-based tion [748]. This aspect raises concerns about the cost-effectiveness of
CLC solutions in addressing the stochastic nature of the AM process. AM for certain production scenarios. (2) A comprehensive evaluation
They proposed a framework for ML-assisted CLC in AM, focusing on of the entire operating procedure, which includes accounting for stages
defect and anomaly control through three scenarios: avoidance, miti- like equipment warm-up/cool-down, may reveal that AM does not nec-
gation, and repair. It identified reinforcement learning and inverse ML essarily outperform conventional manufacturing processes in terms of
models as promising approaches for rapid, situation-aware control de- energy consumption [749]. This aspect challenges the assumption of
cisions. They emphasized the need for seamless integration of various AM being inherently more energy-efficient. (3) AM often necessitates
technologies to advance autonomous in-situ control in industrial set- the use of additional support structures for overhanging parts. These
tings. support structures can lead to wastage of material, time, and energy [6],
further highlighting the need to consider the overall environmental im-
3.5.5. Section conclusion pact of AM processes.
In conclusion, the research in defect detection of AM using machine In conclusion, while AM has undeniably demonstrated several envi-
learning techniques has made significant progress in recent years. Vari- ronmental benefits, it is crucial to acknowledge and address these con-
ous types of defects, including geometric defects, surface defects, poros- cerns to ensure its sustainable and responsible implementation. Careful
ity and voids, inadequate fusion or bonding, inclusions and contami- consideration of the specific manufacturing requirements and the over-
nants, residual stresses, and microstructural defects, have been iden- all ecological impact of AM is essential for making informed decisions
tified and characterized. Image-based signals, such as high-resolution about its usage in different contexts. The main concerns for the sustain-
cameras and microscopy techniques, have been utilized to capture vi- ability of AM are listed in Fig. 14. By considering sustainability from
sual data and detect surface irregularities and cracks. Sensor-based the outset, designers can create products that not only meet functional
signals, including thermal imaging, acoustic emission, and optical emis- requirements but also contribute to environmental and resource con-
sion, have been employed to monitor temperature variations, assess servation goals. Herein, we shine a spotlight on two aspects of design
structural integrity, and identify anomalies in the manufacturing pro- for sustainability, including energy consumption modeling and material
cess. Machine learning algorithms have been applied to analyze the col- waste reduction.
lected data and classify defects with high accuracy. The integration of
multiple sensor modalities and the use of prior-guided neural networks 3.6.1. Energy consumption modeling
have improved the reliability and efficiency of defect detection. Real- In general, recent approaches for energy consumption modeling
time monitoring and quality control based on these techniques have in AM technologies can be broadly divided into two groups: physics
the potential to enhance the reliability and performance of AM pro- modeling and machine learning methods [750,751]. Physics-modeling
cesses. Further research and development in this field will continue to approaches aim to uncover the relationships between energy consump-
advance the application of machine learning in defect detection of AM, tion and relevant parameters by using rigorous formulas. For instance,
contributing to the production of high-quality parts and the widespread Gutierrez-Osorio et al. [752] introduced a mathematical model to pre-
adoption of AM technology. dict energy consumption directly during the system stage, covering
processes such as extrusion, deposition, selection, gluing, and curing.
3.6. Design for sustainability Garcia et al. [753] used life cycle assessment and unit process life cycle
inventory methodology to evaluate the potential environmental impacts
Undoubtedly, ensuring the sustainability of present-day technology of injection molding and fused deposition modeling. Yi et al. [754] de-
is vital for meeting the needs of future generations [744]. It is worth veloped an energy simulation of a desktop AM system based on the bond
noting that historically, sustainability received serious attention with graph modeling approach. Ramesh et al. [755] presented an interpre-
a considerable delay of 50 years from the introduction of numerically tive structural energy consumption model that considers 18 influencing
controlled machining and mass production [269]. As for AM technolo- factors in the material extrusion-based AM process. Despite their poten-
gies, while research on their sustainability is currently being expedited, tial, physics-modeling approaches face challenges in AM due to their
the available documented knowledge remains limited [745]. Therefore, complex nature and the involvement of multiple data sources, including
it is imperative to conduct careful and comprehensive studies on the model designers and manufacturers. This complexity makes it difficult
sustainability aspects of AM in parallel with the further development of for physics-based methods to comprehensively account for all relevant
the industry. factors in AM’s energy consumption.
AM holds significant promise for reducing energy and resource- As a result, machine learning methods have emerged as an alter-
intensive manufacturing processes [269]. It offers notable advantages native approach [756]. These methods rely on large datasets and ma-
over conventional subtractive manufacturing, particularly when fabri- chine learning techniques to identify patterns and correlations between
cating parts with complex geometries, structures, and compositions. The input parameters and energy consumption [269]. Machine learning ap-
ease of manufacturing such intricate components using AM contributes proaches can handle the intricacies and multivariate nature of AM pro-
to improved materials efficiency, as many AM processes employ powder cesses more effectively, providing valuable insights for energy consump-
materials that are highly reusable and recyclable. This efficiency leads tion minimization. Unlike physics-modeling approaches, ML specializes
to reduced material waste, thereby helping to lower the carbon foot- in learning potential rules solely from data, making it particularly well-
print [746]. Additionally, AM’s potential for decentralized production suited for tackling the complexities of AM processes. For example,
brings production closer to the point of consumption. Consequently, Wang et al. [756] proposed a novel approach that involves construct-
supply chains can be compressed and simplified [747]. This reduction ing a multisource dataset comprising pixel-source, motion-source, and
in complexity may lead to lower environmental impact and easier man- processing-source data generated from the AM design stage. They then
agement. Overall, AM is widely recognized as a green technology for employed a deep network called 3DPECP-Net to map the learned rep-

28
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Fig. 14. Main concerns for the sustainability of additive manufacturing.

29
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

resentation from this dataset to energy consumption values, achieving crostructures and their corresponding mechanical properties. The use
accurate predictions with moderate computation requirements. Wang of a multiresponse latent-variable Gaussian process in this approach
et al. [751] also focused on addressing energy consumption (EC) and allows for efficient optimization of the AM design, considering both
material waste concerns associated with AM processes. A new energy ef- performance and support structure requirements [762].
ficiency design method was proposed, utilizing a Multimodal Attention By leveraging advanced optimization techniques and integrating
Fusion Network (MAFN) to optimize the AM process. The study devel- machine learning with topology optimization, researchers and engi-
oped a mathematical EC model for AM and used the MAFN to predict EC neers can significantly reduce waste, material consumption, and print-
by integrating data from computer-aided design and the manufacturing ing time, thus contributing to the overall sustainability and efficiency
process. The MAFN included a Multimodal Fusion Framework and an of AM processes.
Attentional Feature Fusion Module to effectively combine and analyze
diverse data sources. The whole process is demonstrated in Fig. 15. Sim-
ilarly, Yang et al. [343] utilized ML to explore the relationships between 4. Digital twin assisted additive manufacturing
energy consumption and various hand-crafted features of printing lay-
ers, such as layer thickness, surface areas, and shapes. Barrionuevo et
al. [757] applied three ML methods, including gaussian process regres- In recent years, the concept of the Metaverse has gained immense
sor, extreme gradient boosting regressor, and multi-layer perceptron, popularity and captured the attention of various sectors, including
to assess energy consumption in the wire arc AM process using input academia, business, and technology enthusiasts [763–768]. The indus-
variables such as wire diameter, wire feed speed, travel speed, and net trial version of Metaverse–the digital twin–has also attracted significant
power. Majeed et al. [89] proposed a big data-driven framework for attention as a powerful tool for improving the efficiency and reliabil-
sustainable and smart AM, which assists AM industry leaders in making ity of manufacturing processes [769–781]. The adoption of digital twin
better decisions during the beginning of the product life cycle. technology has experienced a rapid surge, driven by its potential to
These examples illustrate the potential of ML in providing valuable revolutionize the manufacturing landscape. The digital twin technology
insights into energy consumption patterns in AM. By leveraging large involves the creation of a virtual replica of a physical object or system
datasets and powerful learning algorithms, ML approaches contribute
that can be used for monitoring, analysis, and optimization. This tech-
to advancing energy-efficient practices in AM. Combining both physics-
nology offers a novel approach to bridge the gap between the physical
modeling and data-driven methods allows researchers to develop more
and digital worlds, enabling manufacturers to create virtual replicas of
accurate and comprehensive tools for assessing and reducing energy
their physical assets. These virtual representations are continuously up-
consumption in AM processes, further promoting the sustainable devel-
dated with real-time data from their physical counterparts [276,782].
opment of AM technology.
By harnessing this wealth of information, manufacturers can gain a deep
3.6.2. Material waste reduction understanding of their assets’ performance, behavior, and condition.
AM technology is not entirely waste-free, and thus, material waste Digital twins serve as powerful tools for monitoring, analysis, and
reduction becomes a critical aspect of sustainable AM [758–760]. Re- simulation of manufacturing processes [783]. They provide real-time
ducing material waste not only has a direct impact on the environmen- insights into the operation of physical systems, allowing manufactur-
tal dimension of sustainable AM by minimizing resource depletion and ers to optimize process parameters, identify inefficiencies, and enhance
waste generation, but it also significantly influences the economic di- overall efficiency. By capturing and analyzing data from sensors and
mension of the process. In numerous AM processes, support structures IoT devices embedded within the physical assets, digital twins facilitate
are a significant source of waste in AM processes [6]. Due to the na- predictive maintenance strategies [784]. This enables proactive identi-
ture of AM, some complex geometries may require support structures fication of potential faults or malfunctions, thereby reducing unplanned
to ensure the integrity of the object during the manufacturing process. downtime and improving overall equipment effectiveness.
However, these additional support structures often result in material Furthermore, digital twins hold tremendous potential for decision-
waste, as they are removed and discarded after the printing is com- making [785–787] in AM. AM processes often involve intricate geome-
plete. Moreover, the presence of support structures leads to increased tries and complex material interactions, which can lead to challenges in
print time and energy consumption. Printing these extra structures not achieving desired part quality and performance. The use of digital twins
only extends the overall manufacturing time but also consumes addi- allows manufacturers to simulate and analyze the entire AM workflow,
tional energy, further contributing to the environmental impact. from design to post-processing [788,789]. By virtually testing different
Optimizing the geometry and topology of support structures is in- design iterations, material choices, and process parameters, manufac-
deed a potential solution to reduce waste and improve the efficiency of turers can identify optimal settings that result in improved part quality,
AM processes [415,418,502]. Recent studies have explored innovative reduced waste, and enhanced production efficiency.
approaches to achieve self-supporting designs or to minimize the need
By integrating digital twin technology into AM, manufacturers can
for excessive support structures. One method involves using genetic al-
achieve greater control and visibility over the entire production lifecy-
gorithms to prune supporting lattice structures without compromising
cle. Real-time data gathered from sensors during the AM process can be
their supporting functionality [345]. This approach identifies and re-
utilized to validate and calibrate the digital twin, ensuring a high degree
moves unnecessary elements from the support structure, optimizing it
of accuracy and reliability [790–794]. This integration facilitates the
for minimal material usage while still providing the necessary support
identification of potential process variations or deviations and enables
to the printed object. Another technique is designing self-supporting
parts that eliminate the need for external support during the AM pro- real-time adjustments to maintain consistency and quality throughout
cess. In a recent study, explicit topology optimization was employed production. The main components and principles of digital twins are
to generate self-supporting designs by optimizing explicit geometry pa- demonstrated in Fig. 16.
rameters [761]. This approach involves adjusting the geometry of the In this section, we will explore the key components and function-
design in a way that allows it to be printed without requiring addi- alities of digital twins and their integration into the AM workflow. We
tional support structures. Moreover, combining topology optimization will examine the diverse ways in which digital twins are utilized for pro-
with machine learning techniques has shown promising results in re- cess optimization, predictive maintenance, quality assurance, iterative
ducing the need for inner and outer support structures. For instance, design, and lifecycle management in the context of AM. Furthermore,
a recent study proposed a data-driven topology optimization method we will analyze the benefits, limitations, and future prospects of digital
for AM that leverages microstructure libraries containing various mi- twin technology in this rapidly evolving field.

30
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Fig. 15. Energy efficiency design for eco-friendly additive manufacturing based on multimodal attention fusion [751]. (a). The energy efficiency design and EC
prediction frameworks during the AM process. (b). Grad-CAM overlaps with the corresponding LCI to interpret the diverse methods during the test phase in EC
prediction. (c). Sliced LCIs with six infill patterns. (d). Comparison results of measured EC and predicted EC by the proposed MAFN on the wheel support models
with diverse infill patterns.

31
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Fig. 16. Components and principles of digital twins for additive manufacturing.

4.1. Research status of digital twins Although the development of digital twins is very popular now,
there are still some challenges that have to be faced. One of the chal-
The current research status of digital twin technology remains highly lenges is the lack of existing physical data for new products [85,798].
active and dynamic, with a growing number of studies exploring its var- Developing digital twins for new products that are still in the design
ious applications and potential in different industries. Researchers and phase can be challenging due to the lack of historical data. Unlike ex-
practitioners still do not agree on a standardized definition of digital isting products, where data from past interactions and performance are
twin, leading to variations in its interpretations and applications based available, new products do not have a track record to rely on. This
on specific use cases and industries. Some recent research studies have scarcity of physical data makes it difficult to accurately model the
focused on establishing common definitions and components of digi- behavior and characteristics of the digital twin. Without this data, de-
tal twins. For instance, Tao et al. [769] conducted a comprehensive
signers may face uncertainties in making informed decisions during the
analysis of digital twin models from various perspectives and classified
design and conceptualization process. Also, the reliance on accurate
studies based on six modeling aspects within the digital twin model-
data of digital twins also sets a threshold for its development. The accu-
ing theoretical system. Enabling technologies and tools for digital twin
racy of the data collected from sensors embedded on physical entities is
modeling were also investigated and summarized. Additionally, Liu et
paramount to the effectiveness of digital twins. In industries like manu-
al. [779] did a systematic research on the basic components of digital
facturing, aerospace, and healthcare, where precise measurements and
twins, focusing on common definitions and characteristics. It clarified
monitoring are critical, any inaccuracies in data collection can result
the relationship between digital twins and cyber-physical systems. The
research methodology for core components (physical entities, virtual in faulty simulations and unreliable predictions. To ensure the fidelity
models, and twin data) was examined, and application areas of digital of the digital twin, rigorous data validation processes and calibration
twins were explored. of sensors are necessary. Additionally, the integration of real-time data
As industries undergo significant digital transformations and em- updates is essential to maintain accurate synchronization between the
brace Industry 4.0 technologies, the adoption of digital twin is expected physical counterpart and its virtual replica [799,800].
to increase [795]. The advancements in data collection, storage, and ex- While the potential benefits of the digital twin in improving process
change, driven by technologies like the Internet of Things and big data, management, monitoring, and predictive maintenance are promising,
are enhancing the capabilities of digital twins [796]. Moreover, the its implementation can be costly [778,801–803]. Creating a digital
integration of augmented reality provides an optimal human-machine twin requires significant investments in technology, infrastructure, and
interface, making digital twin information easily accessible and under- skilled personnel. The costs associated with acquiring and integrating
standable for both technical and non-technical users [797]. sensors, setting up data storage systems, and developing the digital twin

32
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

software can be substantial. Consequently, potential users need to care- ital twin ecosystem for AM using two data-driven approaches. One
fully evaluate the cost and benefits of adopting DTs, often leading to approach used an open-source 3D printer web controller application
their utilization in high-value projects or those that heavily rely on dig- to capture status and parameters, while the other relied on externally
ital twin-specific functionalities. mounted sensors for accurate synchronization between the physical and
Besides, the lack of standardized guidelines and specifications for virtual 3D printers. Near-real-time synchronization was achieved, and
digital twins still poses a challenge to their widespread adoption the digital twin was validated for position, temperature, and run dura-
[804–807]. Establishing a standardized framework for DTs is crucial to tion. The cost-efficient and reliable digital twin ecosystem can provide
building confidence in the technology’s validity, accuracy, and interop- legacy equipment with digital twin capabilities, collect historical data,
erability. Furthermore, the ease of integrating digital twins into existing and generate analytics.
manufacturing setups is an important issue that requires addressing, as In conclusion, the research status of digital twin technology is still
seamless data exchange between different digital twin components is in its infancy, with a focus on its applications in various industries,
essential for realizing their full potential. including AM. Despite challenges such as the lack of standardized def-
As research in the field of digital twin technology continues to initions and guidelines, researchers are exploring common components
progress, there is a particular focus on its application in manufactur- and definitions for digital twins. The integration of digital twins in AM
ing environments. Many recent studies have explored how DTs can be shows promising potential in optimizing the entire product life cycle,
integrated into various stages of a product’s life cycle, from conceptual- from design to production and post-production phases. By combining
ization and design to production, post-production, and retirement stages physics-based models with in-situ sensor data, researchers have suc-
[770,773,808–810]. Researchers are investigating the potential of digi- cessfully developed digital twins for flaw detection in AM processes,
tal twins to interact and influence each phase of the product life cycle, improving the accuracy of predictions and reducing production defects.
with a special emphasis on AM settings. Researchers are exploring how
digital twins can revolutionize various stages of the product life cycle, 4.2. Technical approach
placing particular emphasis on their applications within AM settings.
The integration of digital twins in the AM allows for accurate pro-
Digital twin is a system integration, which contains a lot of tech-
totyping, rapid iterations, and virtual testing, enabling more efficient
nologies from data analytics to virtual modeling and simulation. In this
and cost-effective product development. During the production phase,
section, we will introduce the whole process of building digital twin
digital twins facilitate real-time monitoring, predictive maintenance,
from a technical point of view, providing insights into the diverse range
and quality control, optimizing manufacturing processes and minimiz-
of technologies involved and their interconnections to create an effec-
ing downtime. Furthermore, in the post-production phase, digital twins
tive and comprehensive digital twin system.
continue to play a vital role in monitoring product performance, collect-
ing valuable data for product improvements, and enabling sustainable
lifecycle management. By harnessing the power of digital twins in AM, 4.2.1. Virtual modeling and simulation
researchers aim to enhance productivity, reduce waste, and pave the Virtual modeling and simulation are pivotal in the construction of
way for a more agile and intelligent manufacturing ecosystem. Gaikwad the digital twin, serving as essential tools to bridge the gap between
et al. [793] researched defect-free production of AM parts by combin- theoretical research and practical implementation. At the core of the
ing predictions from a physical model with in-situ sensor data in a digital twin lies its model, which is fundamental for the successful ap-
machine learning framework. They demonstrated the effectiveness of plication of digital twin technology. Within the realm of virtual space,
a digital twin approach, combining physics-based predictions and sen- the digital twin model is composed of four critical dimensions: shape
sor data, for real-time flaw detection in AM processes. The integration and structure, physical properties, dynamic response, and historical
of these approaches resulted in an statistical fidelity-score approaching data and knowledge [769]. The shape and structure model precisely
90%, outperforming using either approach alone (statistical fidelity- captures the geometric shape and intricate assembly relationships of
score of nearly 80%). Liu et al. [789] proposed a digital twin-enabled the physical entity. On the other hand, the physical model meticu-
collaborative data management framework for metal AM systems. The lously reflects the comprehensive physical properties, characteristics,
framework was validated through practical implementation in a dis- and constraints of the actual entity. By simulating the dynamic behav-
tributed metal AM system, showcasing its potential to enhance process ior of the physical entity in response to internal and external influences,
understanding, develop simulation models, reduce development times the dynamic response model offers valuable insights into its perfor-
and costs, and improve product quality and production efficiency. The mance under varying conditions. Moreover, the historical data and
representative application scenario of cloud-based and deep learning- knowledge model enriches the digital twin’s intelligence by integrat-
enabled metal AM layer defect analysis demonstrated the effectiveness ing historical data and leveraging tacit knowledge, contributing to a
of the digital twin approach in supporting intelligent process monitor- smarter digital twin model. Through the amalgamation of multidisci-
ing and optimization. Mandolla et al. [811] proposed a digital twin for plinary knowledge, the multidimensional digital twin model becomes
AM in the aircraft industry, utilizing Blockchain solutions to secure and empowered to perform advanced functions such as prognostication, op-
track the data generated throughout the end-to-end AM process. This timization, and control within the digital realm. As a result, virtual
approach ensures compliance with stringent technical standards and fa- modeling and simulation play a critical role in shaping the capabili-
cilitates the rapid prototyping of components, leading to reduced time- ties of the digital twin and unlocking its immense potential in various
to-market, improved quality, and cost containment. The digital twin real-world applications. Knapp et al. [813] developed and validated a
application enhances the security and organization of data, allowing first-generation digital twin for laser-based directed energy deposition
companies to build secure and connected manufacturing infrastructure AM. The digital twin utilized a transient, three-dimensional model to
in the aircraft industry. Cai et al. [812] developed a methodology using calculate temperature and velocity fields, cooling rates, solidification
augmented reality (AR) to communicate layout information between parameters, and deposit geometry. The predictions of metallurgical pa-
a reconfigurable AM system and its digital twin. The AR technique rameters, such as cooling rates, temperature gradients, solidification
facilitated toolpath planning and simulation for concurrent material de- rates, secondary dendrite arm spacing, and micro-hardness values, were
position using multiple independent actuators. The study demonstrated found to be more accurate compared to conventional heat conduction
that this approach enabled rapid retrieval of position information from calculations. This digital twin provides a practical and efficient tool to
the physical system into the digital twin, supporting layout optimiza- optimize AM processes and improve the properties and serviceability
tion and convenient deployment of the optimized layout back to the of additively manufactured components. Gunasegaram et al. [814] em-
physical system. Pantelidakis et al. [791] implemented a novel dig- phasized the application of digital twin modeling in metal AM to embed

33
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

artificial intelligence capabilities. They highlighted the value of high- tion of digital twins with other systems and technologies, facilitating
fidelity multiscale-multiphysics models for AM processes, addressing data exchange and collaboration.
technical hurdles related to model complexity and scarcity of experi- Research efforts in model lightweighting focus on developing ad-
mental data. Machine learning-based surrogate models were proposed vanced algorithms, compression techniques, and data reduction meth-
for real-time problem-solving. Non-technical barriers, including stan- ods to efficiently represent complex physical systems without sacrificing
dardization and collaboration difficulties, were identified, and potential accuracy. These advancements contribute to the broader adoption and
solutions were offered. Phua et al. [115] developed a new framework implementation of digital twins across various industries, enabling them
for the application of digital twin modeling in metal AM, using Bayesian to leverage the full potential of this technology for enhanced decision-
optimization to build and train surrogate models for real-time control of making, predictive maintenance, and process optimization. As digital
the layering process. The Smart Recoating approach demonstrated the twin technology continues to evolve, model lightweighting will remain
potential of the digital twin control system to mitigate process variation a crucial aspect in building practical and scalable digital twin solutions
and achieve consistent print quality in each layer, revealing new strate- for real-world applications.
gies for controlling the recoater and print stage displacements. Gamdha
et al. [815] developed a general framework for creating a digital twin of 4.2.2. In-situ monitoring and data analytics
the dynamic printing process in extrusion-based AM. Their approach in- In-situ monitoring of the digital twin in AM refers to the real-time
volved performing geometric modeling physics-based simulations with data collection and analysis of the physical manufacturing process,
intermediate print geometries, using parallel adaptive octree meshes which is then used to update and synchronize the corresponding vir-
for real-time predictions. The method demonstrated scalability to high tual model or digital twin [793]. This process involves the continuous
voxel resolutions and accurate predictions of transient heat distribu- monitoring of various process parameters, such as temperature, pres-
tion during the printing process. This work establishes the foundations sure, material deposition rates, and other relevant variables during the
for real-time digital twins, enabling rapid virtual print sequence explo- AM process. The data collected from sensors and monitoring devices on
ration to enhance print quality and reduce material waste. the physical AM system are fed back to the digital twin, allowing it to
In the construction of digital twin models, model lightweight is in- dynamically adjust and simulate the evolving conditions of the physi-
deed a crucial and significant research direction [769,816–820]. As cal process. In-situ monitoring plays a crucial role in ensuring accurate
digital twins are virtual replicas of physical assets, systems, or pro- and up-to-date representations of the manufacturing process in the digi-
cesses, they often involve complex and detailed models to accurately tal twin, enabling more informed decision-making, quality control, and
represent their real-world counterpart. However, creating and work- optimization throughout the AM lifecycle. In-situ quality monitoring in
ing with highly detailed models can be computationally intensive and digital twin-assisted AM goes beyond just being an objective; it serves as
resource-consuming, especially for large-scale systems and simulations. a comprehensive method to enhance the overall system, encompassing
Model lightweighting addresses this challenge by developing techniques both the virtual and physical aspects. By integrating real-time data from
to create simplified yet efficient digital twin models. These lightweight various sensors into the digital twin, the virtual replica can continu-
models retain essential features and information necessary for accurate ously monitor and analyze the manufacturing process. This data-driven
simulations and analysis while reducing the memory and computational approach enables the digital twin to identify defects, anomalies, or devi-
requirements. By optimizing the digital twin’s model, it becomes easier ations as they occur during the physical manufacturing process. Phua et
to handle large datasets, process real-time data, and achieve higher per- al. [133] evaluated and reviewed the application of in-situ monitoring
formance in simulations. Some common methods for model lightweight in the development of a digital twin for metal AM. They organized the
are listed here: research into four levels of increasing complexity, focusing on surrogate
modeling, in-situ sensing, hardware control systems, and intelligent
• Level of Detail (LOD) models [821–824]: Creating different versions control policies. The proposed digital twin hierarchy offers a framework
of the model with varying levels of detail. For example, a high-level for engineering digital twins in AM and other intelligent manufacturing
LOD model may include only the main features, while a low-level systems. Chen et al. [840,841] developed a multisensor fusion-based
LOD model contains fewer details for quicker processing. digital twin for in-situ quality monitoring and defect correction in AM.
• Decimation [825–827]: Removing certain elements or vertices from They synchronized and registered data from multiple sensors, including
the 3D model, reducing its polygon count while retaining the overall an acoustic sensor, an infrared thermal camera, a coaxial vision cam-
shape. era, and a laser line scanner, to predict location-specific quality using
• Feature suppression [828–830]: Temporarily hiding or simplifying machine learning. This allowed for on-the-fly identification of regions
specific features, annotations, or metadata that are not critical for requiring material addition or removal and enabled auto-tuned process
the current analysis or simulation. parameters for defect correction.
• Subdivision surfaces [826,831,832]: Representing complex sur- Absolutely, data analytics is also a crucial aspect of the digital twin
faces using a lower number of control points, resulting in a more of AM. It involves several key components, including:
lightweight model.
• Parameterization [833–835]: Using mathematical parameterization • Data Collection: The data collection process during additive man-
techniques to represent complex shapes with simpler equations. ufacturing involves utilizing various technologies, including Indus-
• Simplified physics models [836]: Employing reduced or approxi- trial IoT [842–844], sensors, digital gauges [845–847], RFID (Radio
mated physics models for simulations, sacrificing some accuracy for Frequency Identification) [848,849], cameras [850–852], and 3D
faster computations. scanning devices [853–855]. These technologies capture essential
• Data aggregation [837–839]: Grouping similar data points together data during the additive manufacturing process.
or using statistical methods to reduce the overall data volume. • Data Mapping: Once the data is collected, it needs to be or-
ganized and mapped in a way that makes sense for the digi-
The benefits of model lightweighting in digital twin construction tal twin. Data mapping is essential for organizing and structur-
are multi-fold. Firstly, it enables real-time or near-real-time simula- ing the collected data. Commonly used data mapping technologies
tions, making it possible to monitor and control physical assets and include XML (Extensible Markup Language) [856–860], Automa-
processes in an agile and responsive manner. Secondly, lightweight tionML [861–863], OPUUA (Open Platform Communications Uni-
models are particularly valuable in fields like AM, where precise and fied Architecture) [864–866], and MT connect [867–869]. These
rapid simulations are essential for optimizing designs and production technologies facilitate data standardization and efficient data ex-
processes. Additionally, model lightweight allows for seamless integra- change between the physical system and the digital twin.

34
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Fig. 17. Digital twins for additive manufacturing workflow.

• Data Processing: Data processing techniques are applied to ana- expertise in modeling and simulation software, as well as a deep un-
lyze and interpret the collected data. This includes data fusion derstanding of the AM process.
[870–873], where data from multiple sources are integrated to pro-
Step 4: Perform Virtual Experiments and Analysis Utilizing the simulation
vide a comprehensive view. Blockchain technology may also be
model within the digital twin, virtual experiments can be performed to
used for data security and integrity [874–879]. Edge computing is
simulate and analyze the AM process under various conditions. Virtual
used for local processing, reducing the burden on central servers
experiments allow for the exploration of different scenarios, the evalua-
[880–884]. Signal processing is employed to extract meaningful in-
formation from raw data [885–888]. tion of alternative process settings, and the prediction of manufacturing
• Data Transmission: The collected and processed data need to be outcomes. The results of these virtual experiments are thoroughly ana-
transmitted between the physical system and the digital twin. For lyzed to gain insights into the process performance, quality, efficiency,
this purpose, various data transmission technologies are used, such and other relevant metrics.
as Bluetooth [889–891], wireless sensor networks [892–895], field- Step 5: Optimize Process Parameters and Variables Based on the analysis
bus networks [896–898], and 5G [899–901]. These technologies of the virtual experiments, the digital twin enables the optimization of
enable real-time or near-real-time data exchange, ensuring seamless process parameters and variables for AM. The insights gained from the
communication between the physical and virtual environments. simulation model help identify optimal settings that result in improved
product quality, reduced production time, enhanced efficiency, and
4.2.3. Whole workflow minimized waste. The optimization process involves adjusting and fine-
According to the existing research on digital twins, the main steps tuning the parameters and variables within the digital twin to achieve
to construct digital twins for AM are demonstrated in Fig. 17 and also the desired manufacturing objectives.
summarized as follows [114,133,134,790,793,814,902–907]:
Step 6: Validate and Refine the Digital Twin The optimized process pa-
Step 1: Create a Digital Twin The first step in utilizing digital twin tech- rameters and variables obtained from the digital twin simulation are
nology for AM is to create a digital twin of the entire manufacturing validated through physical experiments and actual manufacturing runs.
process. A digital twin is a virtual representation of the physical sys- The physical validation allows for the comparison of the digital twin’s
tem, encompassing the equipment, materials, and process parameters predictions with real-world results, validating the accuracy and reliabil-
involved in AM. Creating a digital twin requires specialized software ity of the digital twin model. Any discrepancies or deviations between
and expertise in modeling and simulation to accurately capture the com- the digital twin and the physical outcomes and physical results are care-
plexities of the manufacturing process.
fully analyzed to identify potential gaps or areas for improvement in the
Step 2: Identify Key Parameters and Variables Once the digital twin has digital twin. The insights gained from the validation process are used
been created, the next step is to identify the key parameters and vari- to refine and update the digital twin model, ensuring its accuracy and
ables that have a significant impact on the AM process. These can reliability for future AM processes.
include process parameters, material properties, equipment settings, en-
Step 7: Real-Time Monitoring and Control To enhance the effectiveness of
vironmental conditions, and other relevant factors. Each parameter and
the digital twin-assisted AM process, real-time monitoring and control
variable must be carefully analyzed and understood to ensure their ac-
systems can be implemented. By integrating sensors and data acquisi-
curate representation within the digital twin model.
tion systems into the manufacturing setup, real-time data on process
Step 3: Develop a Comprehensive Simulation Model Using the digital twin parameters, equipment performance, material properties, and product
and the identified parameters and variables, a comprehensive simula- quality can be collected. This real-time data is fed into the digital twin,
tion model is developed to simulate the AM process. The simulation enabling continuous monitoring and control of the AM process. Real-
model should incorporate the physics, mechanics, and dynamics of the time monitoring allows for proactive identification of deviations, timely
process, accurately reflecting the interactions and dependencies be- adjustments, and optimization to maintain optimal manufacturing con-
tween various components. Developing a simulation model requires ditions and product quality.

35
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Step 8: Lifecycle Management and Predictive Maintenance In addition to digital twins with physical twins, considering the decentralized nature
process optimization, the digital twin can be utilized for lifecycle man- of products, and integrate data from different stakeholders and lifecycle
agement and predictive maintenance. By incorporating data on equip- phases. Digital twins can also serve as “soft sensors” to extend the mea-
ment condition, maintenance history, and usage patterns into the digital surement range and facilitate information flow throughout the entire
twin model, manufacturers can monitor the health and performance of lifecycle.
the manufacturing equipment. Predictive maintenance algorithms can Finally, to accelerate the adoption of digital twins in metal AM, a
analyze the data from the digital twin to predict equipment failures and suggested high-level roadmap involves forming a global partnership in
recommend maintenance actions, minimizing downtime and optimizing research through international collaboration. Key topics for discussion
equipment utilization. and research include software and hardware challenges in multiscale-
multiphysics modeling, standardization, uncertainty quantification, ver-
Step 9: Data-Driven Decision-Making and Continuous Improvement Lever-
ification and validation, and the use of machine learning for developing
aging the wealth of data generated from the digital twin, real-time
surrogate models for real-time queries by digital twins.
monitoring, and other sources, manufacturers can employ data analytics
and machine learning techniques to extract insights and make data- 5. Conclusions
driven decisions. By analyzing historical data, identifying patterns, and
detecting correlations, manufacturers can gain a deeper understanding In this essay, we presented a comprehensive review of the inter-
of the AM process. This allows for continuous improvement through section between big data, machine learning, and digital twins in the
the identification of optimization opportunities, root cause analysis of context of AM and examined how they can work together to improve
issues, and the implementation of strategies to enhance overall opera- the efficiency, accuracy, and sustainability of additive manufacturing
tional efficiency and productivity. processes. For machine learning, we explored its application in AM from
Step 10: Integration with Supply Chain and Design Processes To fully har- five aspects, including material analysis, design optimization, process
ness the potential of digital twin-assisted AM, the digital twin can be optimization, defect detection and real-time monitoring, and design for
integrated with other aspects of the supply chain and design processes. sustainability. The ability to leverage big data from various sources
By connecting the digital twin with supply chain management systems, further reinforces the effectiveness of machine learning in providing
manufacturers can optimize material flow, inventory management, and data-driven insights and decision-making capabilities. For digital twin,
production scheduling. Furthermore, integrating the digital twin with we reviewed its current research status and the technical approach to
design software enables design optimization, rapid prototyping, and build the digital twins for additive manufacturing. We also outlined
design-for-manufacturability analysis, leading to improved product de- future development directions of application of digital twins to addi-
signs and reduced time-to-market. tive manufacturing. The combination of big data, machine learning,
and digital twins has the potential to reshape the AM landscape, fos-
4.3. Future development and perspective tering innovation, sustainability, and efficiency. As this field continues
to evolve, further research and collaboration among academia, indus-
The future development directions of digital twin-assisted AM en- try, and policymakers are essential to realize the full potential of these
compass a multi-faceted approach that leverages advancements in vari- technologies in AM.
ous domains to enhance its capabilities and impact on the manufactur-
CRediT authorship contribution statement
ing industry. To fully capitalize on this potential, several critical areas
require focused attention.
Liuchao Jin: Writing – review & editing, Writing – original draft,
Firstly, the research has shown that digital twins have tremendous
Visualization, Validation, Methodology, Investigation, Formal analysis,
potential for optimizing build speed and part performance in metal
Conceptualization. Xiaoya Zhai: Writing – review & editing, Writ-
AM systems. To unlock this potential, future work needs to focus on
ing – original draft, Validation, Investigation, Conceptualization. Kang
improving surrogate modeling techniques to enable deeper integration
Wang: Writing – original draft, Validation, Investigation, Formal anal-
with higher levels of the digital twin hierarchy. Multi-physics validation
ysis. Kang Zhang: Writing – review & editing, Writing – original draft,
and enhancement of surrogate model performance, facilitated by larger
Investigation, Formal analysis. Dazhong Wu: Writing – review & edit-
datasets and emerging machine learning methods, are key areas to ad-
ing, Validation, Investigation, Formal analysis. Aamer Nazir: Writing –
dress. Additionally, sensor development for AM-specific monitoring,
review & editing, Validation, Investigation, Formal analysis. Jingchao
such as computer vision for defect detection, requires further refine-
Jiang: Writing – review & editing, Writing – original draft, Validation,
ment and integration into digital twins to enable real-time monitoring
Supervision, Investigation. Wei-Hsin Liao: Writing – review & editing,
and control of the AM process.
Writing – original draft, Validation, Supervision, Resources, Investiga-
Secondly, there is a need for a unified ontology for digital twin de-
tion.
velopment in AM. Efforts should be made to bridge the gap between
low digital design levels and the high demand for digital twin technol- Declaration of competing interest
ogy. This includes developing mathematical and simulation models to
support digital twin construction and integration across various stake- The authors declare that they have no known competing financial
holders. Blockchain technology can be used to ensure seamless data interests or personal relationships that could have appeared to influence
access while protecting intellectual property. the work reported in this paper.
Thirdly, the construction of digital twins requires the integration of
a vast amount of manufacturing data from various sources, including Data availability
5M1E data (Manpower, Machine, Material, Method, Measurement, En-
vironment) that affects product quality. Complex manufacturing system The data supporting the findings of this study are available within
phenomena, including external factors like orders and supply chains the article.
and internal factors like machine degradation and workers’ skills, need
to be considered for a comprehensive understanding of manufacturing Acknowledgements
systems.
Fourthly, when digital twins come to implement, they are critical for This work is supported by Research Grants Council (C4074-22G),
understanding the lifecycle of products and enabling predictive mainte- Hong Kong Special Administrative Region, China, The Chinese Uni-
nance, fault detection, and diagnosis. Efforts should be made to connect versity of Hong Kong (Project ID: 3110174). For the purpose of open

36
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

Table 4
Acronym used in this paper.

Term Acronym Term Acronym

Accelerated Process Optimization APO k-Nearest Neighbors k-NN

Acrylonitrile Butadiene Styrene ABS Laser-Based Additive Manufacturing LBAM
Acoustic Emission AE Laser Powder-Bed Fusion LPBF
Additive Manufacturing AM Linear Regression LR
Analysis of Variance ANOVA Machine Learning ML
Artificial Intelligence AI Multi-Objective Genetic Algorithm MOGA
Artificial Neural Networks ANN Neural Networks NN
Augmented Reality AR Non-dominated Sorting Genetic Algorithm NSGA
Bayesian Networks BN Particle Swarm Optimization PSO
Cluster Analysis CA Process Parameter Optimization PPO
Central Composite Design CCD Random Forest RF
Computational Fluid Dynamics CFD Reinforcement Learning RL
Convolutional Neural Networks CNN Response Surface Modeling RSM
Directed Energy Deposition DED Stereolithography SLA
Digital Light Processing DLP Selective Laser Melting SLM
Design of Experiments DoE Selective Laser Sintering SLS
Digital Twin DT Simulated Annealing SA
Decision Tree DTree Stainless Steel 316L SS316L
Electron Beam Melting EBM Stereolithography SLA
Full Factorial Design FFD Support Vector Machines SVM
Fused Deposition Modeling FDM Support Vector Regression SVR
Gaussian Process Regression GPR Wire Arc Additive Manufacturing WAAM
Genetic Algorithms GA Extreme Gradient Boosting XGBoost

access, the author has applied a Creative Commons Attribution (CC BY) [15] W. Gao, Y. Zhang, D. Ramanujan, K. Ramani, Y. Chen, C.B. Williams, C.C. Wang,
licence to any Author Accepted Manuscript version arising from this Y.C. Shin, S. Zhang, P.D. Zavattieri, The status, challenges, and future of additive
manufacturing in engineering, Comput. Aided Des. 69 (2015) 65–89.
submission.
[16] C. Druzgalski, A. Ashby, G. Guss, W. King, T.T. Roehling, M.J. Matthews, Process
optimization of complex geometries using feed forward control for laser powder
Appendix A bed fusion additive manufacturing, Addit. Manuf. 34 (2020) 101169.
[17] S. Verma, C.-K. Yang, C.-H. Lin, J.Y. Jeng, Additive manufacturing of lattice struc-
tures for high strength mechanical interlocking of metal and resin during injection
The main terms and their acronyms used in this paper are listed in
molding, Addit. Manuf. 49 (2022) 102463.
Table 4. [18] M.J. Prajapati, A. Kumar, S.-C. Lin, J.-Y. Jeng, Multi-material additive manu-
facturing with lightweight closed-cell foam-filled lattice structures for enhanced
References mechanical and functional properties, Addit. Manuf. 54 (2022) 102766.
[19] X. Zhang, H. Tang, J. Wang, L. Jia, Y. Fan, M. Leary, M. Qian, Additive manu-
[1] J. Liu, V. Nguyen-Van, B. Panda, K. Fox, A. du Plessis, P. Tran, Additive manufac- facturing of intricate lattice materials: ensuring robust strut additive continuity to
turing of sustainable construction materials and form-finding structures: a review realize the design potential, Addit. Manuf. 58 (2022) 103022.
on recent progresses, 3D Print. Addit. Manuf. 9 (1) (2022) 12–34. [20] A. De Marzi, M. Vibrante, M. Bottin, G. Franchin, Development of robot assisted
[2] B.H. Jared, M.A. Aguilo, L.L. Beghini, B.L. Boyce, B.W. Clark, A. Cook, B.J. Kaehr, hybrid additive manufacturing technology for the freeform fabrication of lattice
J. Robbins, Additive manufacturing: toward holistic design, Scr. Mater. 135 (2017) structures, Addit. Manuf. 66 (2023) 103456.
141–147. [21] B. Jagadeesh, M. Duraiselvam, K. Prashanth, Deformation behavior of metallic
[3] C. Tan, R. Li, J. Su, D. Du, Y. Du, B. Attard, Y. Chew, H. Zhang, E.J. Lavernia, lattice structures with symmetrical gradients of porosity manufactured by metal
Y. Fautrelle, et al., Review on field assisted metal additive manufacturing, Int. J. additive manufacturing, Vacuum 211 (2023) 111955.
Mach. Tools Manuf. (2023) 104032. [22] U. Fasel, D. Keidel, L. Baumann, G. Cavolina, M. Eichenhofer, P. Ermanni, Com-
[4] R. Citarella, V. Giannella, Additive manufacturing in industry, Appl. Sci. 11 (2) posite additive manufacturing of morphing aerospace structures, Manuf. Lett. 23
(2021) 840. (2020) 85–88.
[5] X. Zhai, L. Jin, J. Jiang, A survey of additive manufacturing reviews, Mater. Sci. [23] V. Mohanavel, K.A. Ali, K. Ranganathan, J.A. Jeffrey, M. Ravikumar, S. Rajkumar,
Addit. Manuf. 1 (4) (2022) 21. The roles and applications of additive manufacturing in the aerospace and auto-
[6] J. Jiang, X. Xu, J. Stringer, Support structures for additive manufacturing: a review, mobile sector, Mater. Today Proc. 47 (2021) 405–409.
J. Manuf. Mater. Process. 2 (4) (2018) 64. [24] V. Madhavadas, D. Srivastava, U. Chadha, S.A. Raj, M.T.H. Sultan, F.S. Shahar,
[7] J. Jiang, X. Xu, J. Stringer, Optimization of process planning for reducing material A.U.M. Shah, A review on metal additive manufacturing for intricately shaped
waste in extrusion based additive manufacturing, Robot. Comput.-Integr. Manuf. aerospace components, CIRP J. Manuf. Sci. Technol. 39 (2022) 18–36.
59 (2019) 317–325. [25] S. Mohd Yusuf, S. Cutler, N. Gao, The impact of metal additive manufacturing on
[8] J. Jiang, X. Xu, J. Stringer, Optimisation of multi-part production in additive the aerospace industry, Metals 9 (12) (2019) 1286.
manufacturing for reducing support waste, Virtual Phys. Prototyp. 14 (3) (2019) [26] M. Khorasani, A. Ghasemi, B. Rolfe, I. Gibson, Additive manufacturing a powerful
219–228. tool for the aerospace industry, Rapid Prototyping J. 28 (1) (2022) 87–100.
[9] J. Jiang, J. Stringer, X. Xu, R.Y. Zhong, Investigation of printable threshold over- [27] F.H. Froes, R. Boyer, Additive Manufacturing for the Aerospace Industry, Elsevier,
hang angle in extrusion-based additive manufacturing for reducing support waste, Amsterdam, Netherlands; Cambridge, Ma, United States, 2019.
Int. J. Comput. Integr. Manuf. 31 (10) (2018) 961–969. [28] J.C. Najmon, S. Raeisi, A. Tovar, Review of additive manufacturing technologies
[10] J. Jiang, X. Xu, J. Stringer, A new support strategy for reducing waste in additive and applications in the aerospace industry, Addit. Manuf. Aerospace Indust. (2019)
manufacturing, in: The 48th International Conference on Computers and Industrial 7–31.
Engineering (CIE 48), 2018, pp. 1–7. [29] S.C. Altıparmak, B. Xiao, A market assessment of additive manufacturing potential
[11] A. Ghobadian, I. Talavera, A. Bhattacharya, V. Kumar, J.A. Garza-Reyes, N. O’re- for the aerospace industry, J. Manuf. Process. 68 (2021) 728–738.
gan, Examining legitimatisation of additive manufacturing in the interplay be- [30] B. Blakey-Milner, P. Gradl, G. Snedden, M. Brooks, J. Pitot, E. Lopez, M. Leary, F.
tween innovation, lean manufacturing and sustainability, Int. J. Prod. Econ. 219 Berto, A. du Plessis, Metal additive manufacturing in aerospace: a review, Mater.
(2020) 457–468. Des. 209 (2021) 110008.
[12] J. Jiang, X. Zhai, K. Zhang, L. Jin, Q. Lu, Z. Shen, W.-H. Liao, Low-melting-point [31] L. Jin, Y. Lou, L.-A. Chen, Q. Lu, The unified tracking controller for a tilt-rotor
alloys integrated extrusion additive manufacturing, Addit. Manuf. (2023) 103633. unmanned aerial vehicle based on the dual quaternion, in: 2022 IEEE International
[13] J. Gardan, Additive manufacturing technologies: state of the art and trends, in: Conference on Unmanned Systems (ICUS), IEEE, 2022, pp. 1356–1363.
Additive Manufacturing Handbook, 2017, pp. 149–168. [32] V. Juechter, M. Franke, T. Merenda, A. Stich, C. Körner, R. Singer, Additive man-
[14] M. Pérez, D. Carou, E.M. Rubio, R. Teti, Current advances in additive manufactur- ufacturing of Ti-45Al-4Nb-C by selective electron beam melting for automotive
ing, Proc. CIRP 88 (2020) 439–444. applications, Addit. Manuf. 22 (2018) 118–126.

37
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

[33] M.S. Muhammad, L. Kerbache, A. Elomri, Potential of additive manufacturing for [65] C. Guo, M. Zhang, B. Bhandari, Model building and slicing in food 3D printing
upstream automotive supply chains, Supply Chain Forum 23 (2022) 1–19, Taylor processes: a review, Compr. Rev. Food Sci. Food Saf. 18 (4) (2019) 1052–1069.
& Francis. [66] S. Zhu, M.A. Stieger, A.J. van der Goot, M.A. Schutyser, Extrusion-based 3D print-
[34] S. Salifu, D. Desai, O. Ogunbiyi, K. Mwale, Recent development in the additive ing of food pastes: correlating rheological properties with printing behaviour,
manufacturing of polymer-based composites for automotive structures—a review, Innov. Food Sci. Emerg. Technol. 58 (2019) 102214.
Int. J. Adv. Manuf. Technol. 119 (11–12) (2022) 6877–6891. [67] R. Kumar, R. Kumar, et al., 3D printing of food materials: a state of art review and
[35] M. Delic, D.R. Eyers, The effect of additive manufacturing adoption on supply chain future applications, Mater. Today Proc. 33 (2020) 1463–1467.
flexibility and performance: an empirical analysis from the automotive industry, [68] A.F. Ghazal, M. Zhang, Z. Liu, Spontaneous color change of 3D printed healthy
Int. J. Prod. Econ. 228 (2020) 107689. food product over time after printing as a novel application for 4D food printing,
[36] G. Dwivedi, S.K. Srivastava, R.K. Srivastava, Analysis of barriers to implement Food Bioprocess Technol. 12 (2019) 1627–1645.
additive manufacturing technology in the indian automotive sector, Int J. Phys. [69] P. Rando, M. Ramaioli, Food 3D printing: effect of heat transfer on print stability
Distrib. Logist. Manag. 47 (10) (2017) 972–991. of chocolate, J. Food Eng. 294 (2021) 110415.
[37] S.G. Sarvankar, S.N. Yewale, Additive manufacturing in automobile industry, Int. [70] A. Siddika, M.A.A. Mamun, W. Ferdous, A.K. Saha, R. Alyousef, 3D-printed con-
J. Res. Aeronaut. Mech. Eng. 7 (4) (2019) 1–10. crete: applications, performance, and challenges, J. Sustain. Cement-Based Mater.
[38] D. Böckin, A.-M. Tillman, Environmental assessment of additive manufacturing in 9 (3) (2020) 127–164.
the automotive industry, J. Clean. Prod. 226 (2019) 977–987. [71] S.J. Schuldt, J.A. Jagoda, A.J. Hoisington, J.D. Delorit, A systematic review and
[39] J.C. Vasco, Additive manufacturing for the automotive industry, in: Additive Man- analysis of the viability of 3D-printed construction in remote environments, Autom.
ufacturing, Elsevier, 2021, pp. 505–530. Constr. 125 (2021) 103642.
[40] R. Leal, F. Barreiros, L. Alves, F. Romeiro, J. Vasco, M. Santos, C. Marto, Additive [72] D. Lee, H. Kim, J. Sim, D. Lee, H. Cho, D. Hong, Trends in 3D printing technology
manufacturing tooling for the automotive industry, Int. J. Adv. Manuf. Technol. 92 for construction automation using text mining, Int. J. Prec. Eng. Manuf. 20 (2019)
(2017) 1671–1676. 871–882.
[41] E. Alabort, D. Barba, R.C. Reed, Design of metallic bone by additive manufacturing, [73] F. Tahmasebinia, S.M.E. Sepasgozar, S. Shirowzhan, M. Niemela, A. Tripp, S.
Scr. Mater. 164 (2019) 110–114. Nagabhyrava, K. Ko, Z. Mansuri, F. Alonso-Marroquin, Criteria development for
[42] X. Wang, S. Xu, S. Zhou, W. Xu, M. Leary, P. Choong, M. Qian, M. Brandt, Y.M. Xie, sustainable construction manufacturing in construction industry 4.0: theoretical
Topological design and additive manufacturing of porous metals for bone scaffolds and laboratory investigations, Constr. Inn. 20 (3) (2020) 379–400.
and orthopaedic implants: a review, Biomaterials 83 (2016) 127–141. [74] M.A. Hossain, A. Zhumabekova, S.C. Paul, J.R. Kim, A review of 3D printing in
[43] Y. Yang, G. Wang, H. Liang, C. Gao, S. Peng, L. Shen, C. Shuai, Additive manufac- construction and its impact on the labor market, Sustainability 12 (20) (2020)
turing of bone scaffolds, Int. J. Bioprint. 5 (1) (2019). 8492.
[44] A.P.M. Madrid, S.M. Vrech, M.A. Sanchez, A.P. Rodriguez, Advances in additive [75] R.G. Rivera, R.G. Alvarado, A. Martinez-Rocamora, F. Auat Cheein, A comprehen-
manufacturing for bone tissue engineering scaffolds, Mater. Sci. Eng. C 100 (2019) sive performance evaluation of different mobile manipulators used as displace-
631–644. able 3D printers of building elements for the construction industry, Sustainability
[45] D. Barba, E. Alabort, R. Reed, Synthetic bone: design by additive manufacturing,
12 (11) (2020) 4378.
Acta Biomater. 97 (2019) 637–656.
[76] P. Shakor, S. Nejadi, G. Paul, S. Malek, Review of emerging additive manufacturing
[46] Y. Huang, S.R. Schmid, Additive manufacturing for health: state of the art, gaps
technologies in 3D printing of cementitious materials in the construction industry,
and needs, and recommendations, J. Manuf. Sci. Eng. 140 (9) (2018).
Front. Built Env. 4 (2019) 85.
[47] H.K. Celik, O. Kose, M.-E. Ulmeanu, A.E. Rennie, T.N. Abram, I. Akinci, Design
[77] S. Pessoa, A.S. Guimarães, S.S. Lucas, N. Simões, 3D printing in the construction
and additive manufacturing of medical face shield for healthcare workers battling
industry-a systematic review of the thermal performance in buildings, Renew. Sus-
coronavirus (covid-19), Int. J. Bioprint. 6 (4) (2020).
tain. Energy Rev. 141 (2021) 110794.
[48] E. Özceylan, C. Çetinkaya, N. Demirel, O. Sabırlıoğlu, Impacts of additive man-
[78] R. Duballet, O. Baverel, J. Dirrenberger, Classification of building systems for con-
ufacturing on supply chain flow: a simulation approach in healthcare industry,
crete 3D printing, Autom. Constr. 83 (2017) 247–258.
Logistics 2 (1) (2017) 1.
[79] R. García-Alvarado, G. Moroni-Orellana, P. Banda-Pérez, Architectural evaluation
[49] P.K. Arora, R. Arora, A. Haleem, H. Kumar, Application of additive manufacturing
of 3D-printed buildings, Buildings 11 (6) (2021) 254.
in challenges posed by covid-19, Mater. Today Proc. 38 (2021) 466–468.
[80] M. Sakin, Y.C. Kiroglu, 3D printing of buildings: construction of the sustainable
[50] M. Ramola, V. Yadav, R. Jain, On the adoption of additive manufacturing in health-
houses of the future by BIM, Energy Proc. 134 (2017) 702–711.
care: a literature review, Int. J. Manuf. Technol. Manag. 30 (1) (2019) 48–69.
[81] I. Hager, A. Golonka, R. Putanowicz, 3D printing of buildings and building compo-
[51] E.R. Ghomi, F. Khosravi, R.E. Neisiany, S. Singh, S. Ramakrishna, Future of additive
nents as the future of sustainable construction?, Proc. Eng. 151 (2016) 292–299.
manufacturing in healthcare, Current Opin. Biomed. Eng. 17 (2021) 100255.
[82] Y.W.D. Tay, B. Panda, S.C. Paul, N.A. Noor Mohamed, M.J. Tan, K.F. Leong, 3D
[52] G. Liu, Y. He, P. Liu, Z. Chen, X. Chen, L. Wan, Y. Li, J. Lu, Development of bioim-
printing trends in building and construction industry: a review, Virtual Phys. Pro-
plants with 2D, 3D, and 4D additive manufacturing materials, Engineering 6 (11)
totyp. 12 (3) (2017) 261–276.
(2020) 1232–1243.
[53] Q. Yan, H. Dong, J. Su, J. Han, B. Song, Q. Wei, Y. Shi, A review of 3D printing [83] B. Lu, Y. Weng, M. Li, Y. Qian, K.F. Leong, M.J. Tan, S. Qian, A systematical review
technology for medical applications, Engineering 4 (5) (2018) 729–742. of 3D printable cementitious materials, Constr. Build. Mater. 207 (2019) 477–490.
[54] K. Wang, C.-C. Ho, C. Zhang, B. Wang, A review on the 3D printing of functional [84] R. Ashima, A. Haleem, S. Bahl, M. Javaid, S.K. Mahla, S. Singh, Automation and
structures for medical phantoms and regenerated tissue and organ applications, manufacturing of smart materials in additive manufacturing technologies using
Engineering 3 (5) (2017) 653–662. internet of things towards the adoption of industry 4.0, Mater. Today Proc. 45
[55] N. Nachal, J. Moses, P. Karthik, C. Anandharamakrishnan, Applications of 3D print- (2021) 5081–5088.
ing in food processing, Food Eng. Rev. 11 (3) (2019) 123–141. [85] F. Tao, J. Cheng, Q. Qi, M. Zhang, H. Zhang, F. Sui, Digital twin-driven product
[56] I. Tomašević, P. Putnik, F. Valjak, B. Pavlić, B. Šojić, A.B. Markovinović, D.B. design, manufacturing and service with big data, Int. J. Adv. Manuf. Technol. 94
Kovačević, 3D printing as novel tool for fruit-based functional food production, (2018) 3563–3576.
Current Opin. Food Sci. 41 (2021) 138–145. [86] Y. Cui, S. Kara, K.C. Chan, Manufacturing big data ecosystem: a systematic litera-
[57] T. Pereira, S. Barroso, M.M. Gil, Food texture design by 3D printing: a review, ture review, Robot. Comput.-Integr. Manuf. 62 (2020) 101861.
Foods 10 (2) (2021) 320. [87] F. Tao, Q. Qi, A. Liu, A. Kusiak, Data-driven smart manufacturing, J. Manuf. Syst.
[58] S. Caulier, E. Doets, M. Noort, An exploratory consumer study of 3D printed food 48 (2018) 157–169.
perception in a real-life military setting, Food Qual. Prefer. 86 (2020) 104001. [88] A. Majeed, J. Lv, T. Peng, A framework for big data driven process analysis
[59] L. Liu, O.N. Ciftci, Effects of high oil compositions and printing parameters on food and optimization for additive manufacturing, Rapid Prototyping J. 25 (2) (2018)
paste properties and printability in a 3D printing food processing model, J. Food 308–321.
Eng. 288 (2021) 110135. [89] A. Majeed, Y. Zhang, S. Ren, J. Lv, T. Peng, S. Waqar, E. Yin, A big data-driven
[60] A. Baiano, 3D printed foods: a comprehensive review on technologies, nutritional framework for sustainable and smart additive manufacturing, Robot. Comput.-
value, safety, consumer attitude, regulatory framework, and economic and sustain- Integr. Manuf. 67 (2021) 102026.
ability issues, Food Rev. Int. 38 (5) (2022) 986–1016. [90] K. Bi, D. Lin, Y. Liao, C.-H. Wu, P. Parandoush, Additive manufacturing embraces
[61] B. Pérez, H. Nykvist, A.F. Brøgger, M.B. Larsen, M.F. Falkeborg, Impact of macronu- big data, Progr. Addit. Manuf. 6 (2021) 181–197.
trients printability and 3D-printer parameters on 3D-food printing: a review, Food [91] D. Gu, X. Shi, R. Poprawe, D.L. Bourell, R. Setchi, J. Zhu, Material-structure-
Chem. 287 (2019) 249–257. performance integrated laser-metal additive manufacturing, Science 372 (6545)
[62] T. Manstan, M.B. McSweeney, Consumers’ attitudes towards and acceptance of 3D (2021) eabg1487.
printed foods in comparison with conventional food products, Int. J. Food Sci. [92] J.F. Arinez, Q. Chang, R.X. Gao, C. Xu, J. Zhang, Artificial intelligence in advanced
Technol. 55 (1) (2020) 323–331. manufacturing: current status and future outlook, J. Manuf. Sci. Eng. 142 (11)
[63] C. He, M. Zhang, Z. Fang, 3D printing of food: pretreatment and post-treatment of (2020) 110804.
materials, Crit. Rev. Food Sci. Nutr. 60 (14) (2020) 2379–2392. [93] S.S. Razvi, S. Feng, A. Narayanan, Y.-T.T. Lee, P. Witherell, A Review of Machine
[64] A. Le-Bail, B.C. Maniglia, P. Le-Bail, Recent advances and future perspective in Learning Applications in Additive Manufacturing, International Design Engineer-
additive manufacturing of foods based on 3D printing, Current Opin. Food Sci. 35 ing Technical Conferences and Computers and Information in Engineering Confer-
(2020) 54–64. ence, vol. 59179, American Society of Mechanical Engineers, 2019, V001T02A040.

38
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

[94] S. Kumar, T. Gopi, N. Harikeerthana, M.K. Gupta, V. Gaur, G.M. Krolczyk, C. Wu, [122] X. Qi, G. Chen, Y. Li, X. Cheng, C. Li, Applying neural-network-based machine
Machine learning techniques in additive manufacturing: a state of the art review learning to additive manufacturing: current applications, challenges, and future
on design, processes and production control, J. Intell. Manuf. 34 (1) (2023) 21–55. perspectives, Engineering 5 (4) (2019) 721–729.
[95] X. Li, M. Zhang, M. Zhou, J. Wang, W. Zhu, C. Wu, X. Zhang, Qualify assessment [123] F.W. Baumann, A. Sekulla, M. Hassler, B. Himpel, M. Pfeil, Trends of machine
for extrusion-based additive manufacturing with 3D scan and machine learning, J. learning in additive manufacturing, Int. J. Rapid Manuf. 7 (4) (2018) 310–336.
Manuf. Process. 90 (2023) 274–285. [124] A. Hamrani, A. Agarwal, A. Allouhi, D. McDaniel, Applying machine learning to
[96] J. Jiang, Y. Xiong, Z. Zhang, D.W. Rosen, Machine learning integrated design for wire arc additive manufacturing: a systematic data-driven literature review, J. In-
additive manufacturing, J. Intell. Manuf. 33 (4) (2022) 1073–1086. tell. Manuf. (2023) 1–33.
[97] Z. Jin, Z. Zhang, K. Demir, G.X. Gu, Machine learning for advanced additive man- [125] Y. Zhang, M. Safdar, J. Xie, J. Li, M. Sage, Y.F. Zhao, A systematic review on data of
ufacturing, Matter 3 (5) (2020) 1541–1556. additive manufacturing for machine learning applications: the data quality, type,
[98] C. Wang, X. Tan, S.B. Tor, C. Lim, Machine learning in additive manufacturing: preprocessing, and management, J. Intell. Manuf. (2022) 1–36.
state-of-the-art and perspectives, Addit. Manuf. 36 (2020) 101538. [126] J. Akhavan, S. Manoochehri, Sensory data fusion using machine learning methods
[99] J. Jiang, A survey of machine learning in additive manufacturing technologies, Int. for in-situ defect registration in additive manufacturing: a review, in: 2022 IEEE
J. Comput. Integr. Manuf. 36 (9) (2023) 1258–1280. International IOT, Electronics and Mechatronics Conference (IEMTRONICS), IEEE,
[100] D. Guo, S. Ling, H. Li, D. Ao, T. Zhang, Y. Rong, G.Q. Huang, A framework for 2022, pp. 1–10.
personalized production based on digital twin, blockchain and additive manufac- [127] J. Breitenbach, F. Seidenspinner, F. Vural, P. Beisswanger, R. Buettner, A system-
turing in the context of industry 4.0, in: 2020 IEEE 16th International Conference atic literature review of machine learning approaches for optimization in additive
on Automation Science and Engineering (CASE), IEEE, 2020, pp. 1181–1186. manufacturing, in: 2022 IEEE 46th Annual Computers, Software, and Applications
[101] R.B. Roy, D. Mishra, S.K. Pal, T. Chakravarty, S. Panda, M.G. Chandra, A. Pal, P. Conference (COMPSAC), IEEE, 2022, pp. 1147–1152.
Misra, D. Chakravarty, S. Misra, Digital twin: current scenario and a case study on [128] L. Wang, C.A. Alexander, Additive manufacturing and big data, Int. J. Math. Eng.
a manufacturing process, Int. J. Adv. Manuf. Technol. 107 (2020) 3691–3714. Manag. Sci. 1 (3) (2016) 107.
[102] D.P. Möller, H. Vakilzadian, W. Hou, Intelligent manufacturing with digital twin, [129] L. Meng, B. McWilliams, W. Jarosinski, H.-Y. Park, Y.-G. Jung, J. Lee, J. Zhang,
in: 2021 IEEE International Conference on Electro Information Technology (EIT), Machine learning in additive manufacturing: a review, JOM 72 (2020) 2363–2377.
IEEE, 2021, pp. 413–418. [130] J. Qin, F. Hu, Y. Liu, P. Witherell, C.C. Wang, D.W. Rosen, T.W. Simpson, Y. Lu, Q.
[103] A. Croatti, M. Gabellini, S. Montagna, A. Ricci, On the integration of agents and Tang, Research and application of machine learning for additive manufacturing,
digital twins in healthcare, J. Med. Syst. 44 (2020) 1–8. Addit. Manuf. 52 (2022) 102691.
[104] G. Ahmadi-Assalemi, H. Al-Khateeb, C. Maple, G. Epiphaniou, Z.A. Alhaboby, S. [131] S. Guo, M. Agarwal, C. Cooper, Q. Tian, R.X. Gao, W.G. Grace, Y. Guo, Machine
Alkaabi, D. Alhaboby, Digital twins for precision healthcare, in: Cyber Defence in learning for metal additive manufacturing: towards a physics-informed data-driven
the Age of AI, Smart Societies and Augmented Humanity, 2020, pp. 133–158. paradigm, J. Manuf. Syst. 62 (2022) 145–163.
[105] M. Alazab, L.U. Khan, S. Koppu, S.P. Ramu, M. Iyapparaja, P. Boobalan, T. Baker,
[132] S. Chinchanikar, A.A. Shaikh, A review on machine learning, big data analytics,
P.K.R. Maddikunta, T.R. Gadekallu, A. Aljuhani, Digital twins for healthcare 4.0-
and design for additive manufacturing for aerospace applications, J. Mater. Eng.
recent advances, architecture, and open challenges, IEEE Cons. Electron. Mag.
Perform. 31 (8) (2022) 6112–6130.
(2022).
[133] A. Phua, C. Davies, G. Delaney, A digital twin hierarchy for metal additive manu-
[106] R. Kaul, C. Ossai, A.R.M. Forkan, P.P. Jayaraman, J. Zelcer, S. Vaughan, N. Wick-
facturing, Comput. Ind. 140 (2022) 103667.
ramasinghe, The role of AI for developing digital twins in healthcare: the case of
[134] L. Zhang, X. Chen, W. Zhou, T. Cheng, L. Chen, Z. Guo, B. Han, L. Lu, Digital twins
cancer care, Wiley Interdiscip. Rev. Data Min. Knowl. Discov. 13 (1) (2023) e1480.
for additive manufacturing: a state-of-the-art review, Appl. Sci. 10 (23) (2020)
[107] Y. Wang, Q. Tan, F. Pu, D. Boone, M. Zhang, A review of the application of additive
8350.
manufacturing in prosthetic and orthotic clinics from a biomechanical perspective,
[135] Z. Chen, K. Surendraarcharyagie, C. Granland, Keenan Chen, X. Xu, Y. Xiong, C.
Engineering 6 (11) (2020) 1258–1266.
Davies, Y. Tang, Service oriented digital twin for additive manufacturing process,
[108] C. Li, D. Pisignano, Y. Zhao, J. Xue, Advances in medical applications of additive
J. Manuf. Syst. 74 (2024) 762–776.
manufacturing, Engineering 6 (11) (2020) 1222–1231.
[136] L.J. Ladani, Applications of artificial intelligence and machine learning in metal
[109] S.P. Ramu, P. Boopalan, Q.-V. Pham, P.K.R. Maddikunta, T. Huynh-The, M. Alazab,
additive manufacturing, J. Phys. Mater. 4 (4) (2021) 042009.
T.T. Nguyen, T.R. Gadekallu, Federated learning enabled digital twins for smart
[137] International Organization for Standardization, ASTM International, Additive man-
cities: concepts, recent advances, and future directions, Sustain. Cities Soc. 79
ufacturing – general principles – terminology, ISO/ASTM 52900, 2021.
(2022) 103663.
[138] F. Zhang, L. Zhu, Z. Li, S. Wang, J. Shi, W. Tang, N. Li, J. Yang, The recent devel-
[110] G. Mylonas, A. Kalogeras, G. Kalogeras, C. Anagnostopoulos, C. Alexakos, L.
opment of vat photopolymerization: a review, Addit. Manuf. 48 (2021) 102423.
Muñoz, Digital twins from smart manufacturing to smart cities: a survey, IEEE
[139] A. Al Rashid, W. Ahmed, M.Y. Khalid, M. Koc, Vat photopolymerization of poly-
Access 9 (2021) 143222–143249.
mers and polymer composites: processes and applications, Addit. Manuf. 47 (2021)
[111] L. Deren, Y. Wenbo, S. Zhenfeng, Smart city based on digital twins, Comput. Urban
102279.
Sci. 1 (2021) 1–11.
[112] F. Dembski, U. Wössner, M. Letzgus, M. Ruddat, C. Yamu, Urban digital twins for [140] R.V. Pazhamannil, P. Govindan, Current state and future scope of additive manu-
smart cities and citizens: the case study of Herrenberg, Germany, Sustainability facturing technologies via vat photopolymerization, Mater. Today Proc. 43 (2021)
12 (6) (2020) 2307. 130–136.
[113] F. Oettl, S. Hörbrand, T. Wittmeir, J. Schilp, Method for evaluating the monetary [141] I. Cazin, M.O. Gleirscher, M. Fleisch, M. Berer, M. Sangermano, S. Schlögl, Spatially
added value of the usage of a digital twin for additive manufacturing, Proc. CIRP controlling the mechanical properties of 3D printed objects by dual-wavelength vat
118 (2023) 717–722. photopolymerization, Addit. Manuf. 57 (2022) 102977.
[114] H. Mu, F. He, L. Yuan, P. Commins, H. Wang, Z. Pan, Toward a smart wire arc ad- [142] S. Tyagi, A. Yadav, S. Deshmukh, Review on mechanical characterization of 3D
ditive manufacturing system: a review on current developments and a framework printed parts created using material jetting process, Mater. Today Proc. 51 (2022)
of digital twin, J. Manuf. Syst. 67 (2023) 174–189. 1012–1016.
[115] A. Phua, P.S. Cook, C.H. Davies, G.W. Delaney, Smart recoating: a digital twin [143] A. Elkaseer, K.J. Chen, J.C. Janhsen, O. Refle, V. Hagenmeyer, S.G. Scholz, Ma-
framework for optimisation and control of powder spreading in metal additive terial jetting for advanced applications: a state-of-the-art review, gaps and future
manufacturing, J. Manuf. Process. 99 (2023) 382–391. directions, Addit. Manuf. (2022) 103270.
[116] K. Zhu, J.Y.H. Fuh, X. Lin, Metal-based additive manufacturing condition monitor- [144] A. Pugalendhi, R. Ranganathan, S. Ganesan, Impact of process parameters on
ing: a review on machine learning based approaches, IEEE/ASME Trans. Mecha- mechanical behaviour in multi-material jetting, Mater. Today Proc. 46 (2021)
tron. (2021). 9139–9144.
[117] D. Mahmoud, M. Magolon, J. Boer, M. Elbestawi, M.G. Mohammadi, Applications [145] H. Yang, J.C. Lim, Y. Liu, X. Qi, Y.L. Yap, V. Dikshit, W.Y. Yeong, J. Wei, Perfor-
of machine learning in process monitoring and controls of L-PBF additive manu- mance evaluation of projet multi-material jetting 3D printer, Virtual Phys. Proto-
facturing: a review, Appl. Sci. 11 (24) (2021) 11910. typ. 12 (1) (2017) 95–103.
[118] M. Parsazadeh, S. Sharma, N. Dahotre, Towards the next generation of machine [146] M. Yakout, M. Elbestawi, S.C. Veldhuis, A review of metal additive manufacturing
learning models in additive manufacturing: a review of process dependent material technologies, Solid State Phenom. 278 (2018) 1–14.
evolution, Prog. Mater. Sci. (2023) 101102. [147] X. Lv, F. Ye, L. Cheng, S. Fan, Y. Liu, Binder jetting of ceramics: powders, binders,
[119] M.D. Xames, F.K. Torsha, F. Sarwar, A systematic literature review on recent trends printing parameters, equipment, and post-treatment, Ceram. Int. 45 (10) (2019)
of machine learning applications in additive manufacturing, J. Intell. Manuf. 34 (6) 12609–12624.
(2023) 2529–2555. [148] Y. Bai, C.B. Williams, Binder jetting additive manufacturing with a particle-free
[120] A. Raza, K.M. Deen, R. Jaafreh, K. Hamad, A. Haider, W. Haider, Incorporation of metal ink as a binder precursor, Mater. Des. 147 (2018) 146–156.
machine learning in additive manufacturing: a review, Int. J. Adv. Manuf. Technol. [149] X. Shen, H.E. Naguib, A robust ink deposition system for binder jetting and material
122 (3–4) (2022) 1143–1166. jetting, Addit. Manuf. 29 (2019) 100820.
[121] G.K. Sarkon, B. Safaei, M.S. Kenevisi, S. Arman, Q. Zeeshan, State-of-the-art re- [150] N.D. Dejene, H.G. Lemu, Current status and challenges of powder bed fusion-based
view of machine learning applications in additive manufacturing; from design to metal additive manufacturing: literature review, Metals 13 (2) (2023) 424.
manufacturing and property control, Arch. Comput. Methods Eng. 29 (7) (2022) [151] S. Qu, J. Ding, J. Fu, M. Fu, B. Zhang, X. Song, High-precision laser powder bed
5663–5721. fusion processing of pure copper, Addit. Manuf. 48 (2021) 102417.

39
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

[152] N. Sanaei, A. Fatemi, N. Phan, Defect characteristics and analysis of their variabil- [179] M. Alhijjaj, J. Nasereddin, P. Belton, S. Qi, Impact of processing parameters on
ity in metal L-PBF additive manufacturing, Mater. Des. 182 (2019) 108091. the quality of pharmaceutical solid dosage forms produced by fused deposition
[153] S. Yu, S. Park, D.Y. Kang, G.S. Shin, M.W. Lee, S.Y. Moon, J.Y. Hwang, Strategic modeling (FDM), Pharmaceutics 11 (12) (2019) 633.
dual laser 3D printing of structural metal-plastic hybrid materials, Composites, Part [180] A. Sharma, A. Rai, Fused deposition modelling (FDM) based 3D & 4D Printing: a
B, Eng. 261 (2023) 110794. state of art review, Mater. Today Proc. 62 (2022) 367–372.
[154] S. Mu, Y. Hong, H. Huang, A. Ishii, J. Lei, Y. Song, Y. Li, K.S. Brinkman, F. Peng, [181] M.S. Alsoufi, M.W. Alhazmi, D.K. Suker, T.A. Alghamdi, R.A. Sabbagh, M.A. Felem-
H. Xiao, et al., A novel laser 3D printing method for the advanced manufacturing ban, F.K. Bazuhair, Experimental characterization of the influence of nozzle tem-
of protonic ceramics, Membranes 10 (5) (2020) 98. perature in FDM 3D printed pure PLA and advanced PLA+, Am. J. Mech. Eng. 7 (2)
[155] J. Rindler, C. Slone, E. Herderick, M. Mills, A. Ramirez, Investigation on the poten- (2019) 45–60.
tial effects of laser stitching and subsequent heat treatment on the microstructure [182] J. Zhou, X. Zhou, H. Li, J. Hu, X. Han, S. Liu, In-situ laser shock peening for
and mechanical properties of Nickel Alloy 718 produced via Laser Powder Bed improved surface quality and mechanical properties of laser-directed energy-
Fusion (L-PBF), Addit. Manuf. 57 (2022) 102906. deposited AlSi10Mg alloy, Addit. Manuf. 60 (2022) 103177.
[156] W. Yang, Y.-G. Jung, T. Kwak, S.K. Kim, H. Lim, D.-H. Kim, Microstructure and [183] T. Feldhausen, L. Heinrich, K. Saleeby, A. Burl, B. Post, E. MacDonald, C. Saldana,
mechanical properties of an Al–Mg–Si–Zr alloy processed by L-PBF and subsequent L. Love, Review of computer-aided manufacturing (cam) strategies for hybrid di-
heat treatments, Materials 15 (15) (2022) 5089. rected energy deposition, Addit. Manuf. (2022) 102900.
[157] H. Javidrad, S. Salemi, Effect of the volume energy density and heat treatment [184] D. De Oliveira, M.C. Gomes, A.G. Dos Santos, K.S.B. Ribeiro, I.J. Vasques, R.T.
on the defect, microstructure, and hardness of L-PBF Inconel 625, Metall. Mater. Coelho, M.B. Da Silva, N.W. Hung, Abrasive and non-conventional post-processing
Trans. A 51 (2020) 5880–5891. techniques to improve surface finish of additively manufactured metals: a review,
[158] W. Qin, C. Man, K. Pang, H. Zhang, Z. Cui, L. Wang, D. Kong, C. Dong, H. Cui, Progr. Addit. Manuf. (2022) 1–18.
Corrosion behavior of L-PBF Ti6Al4V with heat treatments in the F–containing [185] A. Pilipović, Sheet lamination, in: Polymers for 3D Printing, Elsevier, 2022,
environments, Corros. Sci. 210 (2023) 110811. pp. 127–136.
[159] M. Avateffazeli, S.I. Shakil, M.F. Khan, H. Pirgazi, N. Shamsaei, M. Haghshenas, [186] M.A.M. Asri, W.C. Mak, S.A. Norazman, A.N. Nordin, Low-cost and rapid proto-
The effect of heat treatment on fatigue response of laser powder bed fused Al-Cu- typing of integrated electrochemical microfluidic platforms using consumer-grade
Mg-Ag-TiB2 (A20X) alloy, Mater. Today Commun. 35 (2023) 106009. off-the-shelf tools and materials, Lab Chip 22 (9) (2022) 1779–1792.
[160] L. Wang, C. Shen, P. Zhang, X. Hua, Y. Zhang, F. Li, J. Xin, K. Wu, Fabrication [187] J. Yuan, G. Chen, H. Li, H. Prautzsch, K. Xiao, Accurate and computational: a
of Fe–30Al alloy using plasma arc welding powered twin-wire directed energy review of color reproduction in full-color 3D printing, Mater. Des. 209 (2021)
deposition-arc process: droplet transfer, microstructure, and mechanical property 109943.
investigation, Intermetallics 161 (2023) 107961. [188] S.M. El Awad Azrak, C.M. Clarkson, R.J. Moon, G.T. Schueneman, J.P. Youngblood,
[161] A. Nazir, O. Gokcekaya, K.M.M. Billah, O. Ertugrul, J. Jiang, J. Sun, S. Hussain, Wet-stacking lamination of multilayer mechanically fibrillated cellulose nanofibril
Multi-material additive manufacturing: a systematic review of design, properties, (CNF) sheets with increased mechanical performance for use in high-strength and
applications, challenges, and 3D Printing of materials and cellular metamaterials, lightweight structural and packaging applications, ACS Appl. Polym. Mater. 1 (9)
Mater. Des. (2023) 111661. (2019) 2525–2534.
[162] R.J. Williams, C.M. Davies, P.A. Hooper, A pragmatic part scale model for residual [189] P. Kiss, W. Stadlbauer, C. Burgstaller, H. Stadler, S. Fehringer, F. Haeuserer, V.-M.
stress and distortion prediction in powder bed fusion, Addit. Manuf. 22 (2018) Archodoulaki, In-house recycling of carbon-and glass fibre-reinforced thermoplas-
416–425. tic composite laminate waste into high-performance sheet materials, Composites,
[163] M.A. Dechet, J.S.G. Bonilla, M. Grünewald, K. Popp, J. Rudloff, M. Lang, J. Part A, Appl. Sci. Manuf. 139 (2020) 106110.
Schmidt, A novel, precipitated polybutylene terephthalate feedstock material for [190] X. Wu, J. Teng, X. Ji, C. Xu, D. Ma, S. Sui, Z. Zhang, Research progress of the
powder bed fusion of polymers (PBF): material development and initial PBF pro- defects and innovations of ceramic vat photopolymerization, Addit. Manuf. (2023)
cessability, Mater. Des. 197 (2021) 109265. 103441.
[164] A.M. Gohn, D. Brown, G. Mendis, S. Forster, N. Rudd, M. Giles, Mold inserts for [191] M. Pagac, J. Hajnys, Q.-P. Ma, L. Jancar, J. Jansa, P. Stefek, J. Mesicek, A review
injection molding prototype applications fabricated via material extrusion additive of vat photopolymerization technology: materials, applications, challenges, and
manufacturing, Addit. Manuf. 51 (2022) 102595. future trends of 3D printing, Polymers 13 (4) (2021) 598.
[165] D.E. Samoylenko, K.S. Rodygin, V.P. Ananikov, Sustainable application of calcium [192] A. Andreu, P.-C. Su, J.-H. Kim, C.S. Ng, S. Kim, I. Kim, J. Lee, J. Noh, A.S. Sub-
carbide residue as a filler for 3D printing materials, Sci. Rep. 13 (1) (2023) 4465. ramanian, Y.-J. Yoon, 4D printing materials for vat photopolymerization, Addit.
[166] K.M. Agarwal, P. Shubham, D. Bhatia, P. Sharma, H. Vaid, R. Vajpeyi, Analyzing Manuf. 44 (2021) 102024.
the impact of print parameters on dimensional variation of ABS specimens printed [193] U. Shaukat, E. Rossegger, S. Schlögl, A review of multi-material 3D printing of
using fused deposition modelling (FDM), Sensors Int. 3 (2022) 100149. functional materials via vat photopolymerization, Polymers 14 (12) (2022) 2449.
[167] B.S. Shim, J.H. Choe, J.-U. Hou, Source identification of 3D printer based on lay- [194] X. Xu, A. Awad, P. Robles-Martinez, S. Gaisford, A. Goyanes, A.W. Basit, Vat
ered texture encoders, IEEE Trans. Multimed. (2023). photopolymerization 3D printing for advanced drug delivery and medical device
[168] S.Y. Hann, H. Cui, N.C. Zalud, T. Esworthy, K. Bulusu, Y.-L. Shen, M.W. Plesniak, applications, J. Control. Release 329 (2021) 743–757.
L.G. Zhang, An in vitro analysis of the effect of geometry-induced flows on en- [195] I. Sachdeva, S. Ramesh, U. Chadha, H. Punugoti, S.K. Selvaraj, Computational AI
dothelial cell behavior in 3D printed small-diameter blood vessels, Biomater. Adv. models in VAT photopolymerization: a review, current trends, open issues, and
137 (2022) 212832. future opportunities, Neural Comput. Appl. 34 (20) (2022) 17207–17229.
[169] S.Y. Hann, H. Cui, T. Esworthy, L.G. Zhang, 4D thermo-responsive smart hiPSC-CM [196] A. Medellin, W. Du, G. Miao, J. Zou, Z. Pei, C. Ma, Vat photopolymerization 3D
cardiac construct for myocardial cell therapy, Int. J. Nanomed. (2023) 1809–1821. printing of nanocomposites: a literature review, J. Micro Nano-Manuf. 7 (3) (2019).
[170] Y. Wang, H. Cui, T. Esworthy, D. Mei, Y. Wang, L.G. Zhang, Emerging 4D printing [197] N.A. Chartrain, C.B. Williams, A.R. Whittington, A review on fabricating tissue
strategies for next-generation tissue regeneration and medical devices, Adv. Mater. scaffolds using vat photopolymerization, Acta Biomater. 74 (2018) 90–111.
34 (20) (2022) 2109198. [198] H. Liu, D. Mei, S. Qian, Y. Wang, Vat photopolymerization of polymer-derived SiC
[171] W. Zhu, N.J. Castro, Y.-L. Shen, L.G. Zhang, Nanotechnology and 3D/4D bioprint- ceramic with triply periodic minimal surface structure for hydrogen production,
ing for neural tissue regeneration, in: 3D Bioprinting and Nanotechnology in Tissue Addit. Manuf. (2023) 103694.
Engineering and Regenerative Medicine, Elsevier, 2022, pp. 427–458. [199] J. Cheng, R. Wang, Z. Sun, Q. Liu, X. He, H. Li, H. Ye, X. Yang, X. Wei, Z. Li, et al.,
[172] S.Y. Hann, H. Cui, G. Chen, M. Boehm, T. Esworthy, L.G. Zhang, 3D printed Centrifugal multimaterial 3D printing of multifunctional heterogeneous objects,
biomimetic flexible blood vessels with iPS cell-laden hierarchical multilayers, Nat. Commun. 13 (1) (2022) 7931.
Biomed. Eng. Adv. 4 (2022) 100065. [200] S. Mora, N.M. Pugno, D. Misseroni, 3D printed architected lattice structures by
[173] L.G. Zhang, K. Leong, J.P. Fisher, 3D Bioprinting and Nanotechnology in Tissue material jetting, Mater. Today (2022).
Engineering and Regenerative Medicine, Academic Press, 2022. [201] I. Gibson, D. Rosen, B. Stucker, M. Khorasani, I. Gibson, D. Rosen, B. Stucker, M.
[174] C. Wu, C. Dai, G. Fang, Y.-J. Liu, C.C. Wang, General support-effective decomposi- Khorasani, Material jetting, Addit. Manuf. Technol. (2021) 203–235.
tion for multi-directional 3-D printing, IEEE Trans. Autom. Sci. Eng. 17 (2) (2019) [202] Y.L. Yap, C. Wang, S.L. Sing, V. Dikshit, W.Y. Yeong, J. Wei, Material jetting
599–610. additive manufacturing: an experimental study using designed metrological bench-
[175] M. Qin, S. Gao, C.C. Wang, W.-H. Liao, Multi-axis direct metal deposition process marks, Precis. Eng. 50 (2017) 275–285.
with effective regrouping strategy, J. Manuf. Process. 81 (2022) 707–716. [203] O. Gülcan, K. Günaydın, A. Tamer, The state of the art of material jetting—a critical
[176] Z. Zhang, C. Wu, C. Dai, Q. Shi, G. Fang, D. Xie, X. Zhao, Y.-J. Liu, C.C. Wang, X.-J. review, Polymers 13 (16) (2021) 2829.
Wang, A multi-axis robot-based bioprinting system supporting natural cell function [204] J. Dilag, T. Chen, S. Li, S.A. Bateman, Design and direct additive manufacturing
preservation and cardiac tissue fabrication, Bioactive Mater. 18 (2022) 138–150. of three-dimensional surface micro-structures using material jetting technologies,
[177] G. Fang, T. Zhang, S. Zhong, X. Chen, Z. Zhong, C.C. Wang, Reinforced FDM: multi- Addit. Manuf. 27 (2019) 167–174.
axis filament alignment with controlled anisotropic strength, ACM Trans. Graph. [205] K. Sen, T. Mehta, S. Sansare, L. Sharifi, A.W. Ma, B. Chaudhuri, Pharmaceutical
39 (6) (2020) 1–15. applications of powder-based binder jet 3D printing process–a review, Adv. Drug
[178] T. Zhang, G. Fang, Y. Huang, N. Dutta, S. Lefebvre, Z.M. Kilic, C.C. Wang, S3-slicer: Deliv. Rev. 177 (2021) 113943.
a general slicing framework for multi-axis 3D printing, ACM Trans. Graph. 41 (6) [206] P. Shakor, S. Chu, A. Puzatova, E. Dini, Review of binder jetting 3D printing in the
(2022) 1–15. construction industry, Progr. Addit. Manuf. 7 (4) (2022) 643–669.

40
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

[207] S. Mirzababaei, S. Pasebani, A review on binder jet additive manufacturing of 316L [237] A. Oleff, B. Küster, M. Stonis, L. Overmeyer, Process monitoring for material extru-
stainless steel, J. Manuf. Mater. Process. 3 (3) (2019) 82. sion additive manufacturing: a state-of-the-art review, Progr. Addit. Manuf. 6 (4)
[208] W. Du, X. Ren, Z. Pei, C. Ma, Ceramic binder jetting additive manufacturing: a (2021) 705–730.
literature review on density, J. Manuf. Sci. Eng. 142 (4) (2020) 040801. [238] C. Suwanpreecha, A. Manonukul, A review on material extrusion additive man-
[209] W. Du, X. Ren, C. Ma, Z. Pei, Binder jetting additive manufacturing of ceram- ufacturing of metal and how it compares with metal injection moulding, Metals
ics: a literature review, in: ASME International Mechanical Engineering Congress 12 (3) (2022) 429.
and Exposition, vol. 58493, American Society of Mechanical Engineers, 2017, [239] J. Huang, Q. Chen, H. Jiang, B. Zou, L. Li, J. Liu, H. Yu, A survey of design methods
V014T07A006. for material extrusion polymer 3D printing, Virtual Phys. Prototyp. 15 (2) (2020)
[210] F. Dini, S.A. Ghaffari, J. Jafar, R. Hamidreza, S. Marjan, A review of binder jet pro- 148–162.
cess parameters: powder, binder, printing and sintering condition, Metal Powder [240] J. Gonzalez-Gutierrez, S. Cano, S. Schuschnigg, C. Kukla, J. Sapkota, C. Holzer, Ad-
Rep. 75 (2) (2020) 95–100. ditive manufacturing of metallic and ceramic components by the material extrusion
[211] A. Lores, N. Azurmendi, I. Agote, E. Zuza, A review on recent developments in of highly-filled polymers: a review and future perspectives, Materials 11 (5) (2018)
binder jetting metal additive manufacturing: materials and process characteristics, 840.
Powder Metall. 62 (5) (2019) 267–296. [241] G. Hsiang Loh, E. Pei, J. Gonzalez-Gutierrez, M. Monzón, An overview of material
[212] A. Mostafaei, A.M. Elliott, J.E. Barnes, F. Li, W. Tan, C.L. Cramer, P. Nandwana, extrusion troubleshooting, Appl. Sci. 10 (14) (2020) 4776.
M. Chmielus, Binder jet 3D printing: process parameters, materials, properties, [242] G.D. Goh, Y.L. Yap, H. Tan, S.L. Sing, G.L. Goh, W.Y. Yeong, Process–
modeling, and challenges, Prog. Mater. Sci. 119 (2021) 100707. structure–properties in polymer additive manufacturing via material extrusion: a
[213] M. Li, W. Du, A. Elwany, Z. Pei, C. Ma, Metal binder jetting additive manufacturing: review, Crit. Rev. Solid State Mater. Sci. 45 (2) (2020) 113–133.
a literature review, J. Manuf. Sci. Eng. 142 (9) (2020). [243] A. Jinoop, C. Paul, K. Bindra, Laser-assisted directed energy deposition of nickel
[214] M. Ziaee, N.B. Crane, Binder jetting: a review of process, materials, and methods, super alloys: a review, Proc. Inst. Mech. Eng., Part L: J. Mater. Des. Appl. 233 (11)
Addit. Manuf. 28 (2019) 781–801. (2019) 2376–2400.
[215] Z. Snow, A.R. Nassar, E.W. Reutzel, Invited review article: review of the formation [244] A. Saboori, D. Gallo, S. Biamino, P. Fino, M. Lombardi, An overview of additive
and impact of flaws in powder bed fusion additive manufacturing, Addit. Manuf. manufacturing of titanium components by directed energy deposition: microstruc-
36 (2020) 101457. ture and mechanical properties, Appl. Sci. 7 (9) (2017) 883.
[216] A. Awad, F. Fina, A. Goyanes, S. Gaisford, A.W. Basit, Advances in powder bed [245] A. Saboori, A. Aversa, G. Marchese, S. Biamino, M. Lombardi, P. Fino, Application
fusion 3D printing in drug delivery and healthcare, Adv. Drug Deliv. Rev. 174 of directed energy deposition-based additive manufacturing in repair, Appl. Sci.
(2021) 406–424. 9 (16) (2019) 3316.
[217] W.E. King, A.T. Anderson, R.M. Ferencz, N.E. Hodge, C. Kamath, S.A. Khairal- [246] D. Feenstra, R. Banerjee, H. Fraser, A. Huang, A. Molotnikov, N. Birbilis, Critical
lah, A.M. Rubenchik, Laser powder bed fusion additive manufacturing of metals; review of the state of the art in multi-material fabrication via directed energy
physics, computational, and materials challenges, Appl. Phys. Rev. 2 (4) (2015). deposition, Curr. Opin. Solid State Mater. Sci. 25 (4) (2021) 100924.
[218] M. Khorasani, I. Gibson, A.H. Ghasemi, E. Hadavi, B. Rolfe, Laser subtractive and [247] D. Svetlizky, M. Das, B. Zheng, A.L. Vyatskikh, S. Bose, A. Bandyopadhyay, J.M.
laser powder bed fusion of metals: review of process and production features, Rapid Schoenung, E.J. Lavernia, N. Eliaz, Directed energy deposition (DED) additive man-
Prototyping J. 29 (5) (2023) 935–958. ufacturing: physical characteristics, defects, challenges and applications, Mater.
[219] M. Grasso, A. Remani, A. Dickins, B. Colosimo, R.K. Leach, In-situ measurement Today 49 (2021) 271–295.
and monitoring methods for metal powder bed fusion: an updated review, Meas. [248] A. Singh, S. Kapil, M. Das, A comprehensive review of the methods and mecha-
Sci. Technol. 32 (11) (2021) 112001. nisms for powder feedstock handling in directed energy deposition, Addit. Manuf.
[220] C. Zitelli, P. Folgarait, A. Di Schino, Laser powder bed fusion of stainless steel 35 (2020) 101388.
grades: a review, Metals 9 (7) (2019) 731. [249] Z.-j. Tang, W.-w. Liu, Y.-w. Wang, K.M. Saleheen, Z.-c. Liu, S.-t. Peng, Z. Zhang, H.-
[221] A. Aversa, G. Marchese, A. Saboori, E. Bassini, D. Manfredi, S. Biamino, D. Ugues, c. Zhang, A review on in situ monitoring technology for directed energy deposition
P. Fino, M. Lombardi, New aluminum alloys specifically designed for laser powder of metals, Int. J. Adv. Manuf. Technol. 108 (2020) 3437–3463.
bed fusion: a review, Materials 12 (7) (2019) 1007. [250] D.-G. Ahn, Directed energy deposition (DED) process: state of the art, Int. J. Precis.
[222] S. Sanchez, P. Smith, Z. Xu, G. Gaspard, C.J. Hyde, W.W. Wits, I.A. Ashcroft, H. Eng. Manuf.-Green Technol. 8 (2021) 703–742.
Chen, A.T. Clare, Powder bed fusion of nickel-based superalloys: a review, Int. J. [251] A. Dass, A. Moridi, State of the art in directed energy deposition: from additive
Mach. Tools Manuf. 165 (2021) 103729. manufacturing to materials design, Coatings 9 (7) (2019) 418.
[223] E. Malekipour, H. El-Mounayri, Common defects and contributing parameters in [252] P.M. Bhatt, A.M. Kabir, M. Peralta, H.A. Bruck, S.K. Gupta, A robotic cell for per-
powder bed fusion am process and their classification for online monitoring and forming sheet lamination-based additive manufacturing, Addit. Manuf. 27 (2019)
control: a review, Int. J. Adv. Manuf. Technol. 95 (2018) 527–550. 278–289.
[224] A. Khorasani, I. Gibson, J.K. Veetil, A.H. Ghasemi, A review of technological im- [253] Y. Zhang, L. Wu, X. Guo, S. Kane, Y. Deng, Y.-G. Jung, J.-H. Lee, J. Zhang, Additive
provements in laser-based powder bed fusion of metal printers, Int. J. Adv. Manuf. manufacturing of metallic materials: a review, J. Mater. Eng. Perform. 27 (2018)
Technol. 108 (2020) 191–209. 1–13.
[225] R. Singh, A. Gupta, O. Tripathi, S. Srivastava, B. Singh, A. Awasthi, S. Rajput, P. So- [254] C. Thakar, S.P. Deshmukh, T. Mulla, A review on selective deposition lamination
nia, P. Singhal, K.K. Saxena, Powder bed fusion process in additive manufacturing: 3D printing technique, Int. J. Adv. Sci. Res. Eng. Trends 4 (2020) 7–11.
an overview, Mater. Today Proc. 26 (2020) 3058–3070. [255] L. Ladani, Additive Manufacturing of Metals: Materials, Processes, Tests, and Stan-
[226] M. Mehrpouya, D. Tuma, T. Vaneker, M. Afrasiabi, M. Bambach, I. Gibson, Multi- dards, DEStech Publications, 2021.
material powder bed fusion techniques, Rapid Prototyping J. (2022). [256] J. Fan, F. Han, H. Liu, Challenges of big data analysis, Nat. Sci. Rev. 1 (2) (2014)
[227] J.L. Bartlett, X. Li, An overview of residual stresses in metal powder bed fusion, 293–314.
Addit. Manuf. 27 (2019) 131–149. [257] S. Sagiroglu, D. Sinanc, Big data: a review, in: 2013 International Conference on
[228] L. Dowling, J. Kennedy, S. O’Shaughnessy, D. Trimble, A review of critical repeata- Collaboration Technologies and Systems (CTS), IEEE, 2013, pp. 42–47.
bility and reproducibility issues in powder bed fusion, Mater. Des. 186 (2020) [258] W.L. Chang, N. Grady, Nist Big Data Interoperability Framework: Volume 1, Defi-
108346. nitions, 2019.
[229] A.T. Sutton, C.S. Kriewall, M.C. Leu, J.W. Newkirk, Powder characterisation tech- [259] F. Rosenblatt, et al., Principles of Neurodynamics: Perceptrons and the Theory of
niques and effects of powder characteristics on part properties in powder-bed Brain Mechanisms, vol. 55, Spartan Books, Washington, DC, 1962.
fusion processes, Virtual Phys. Prototyp. 12 (1) (2017) 3–29. [260] J.R. Quinlan, Induction of decision trees, Mach. Learn. 1 (1986) 81–106.
[230] M. Grasso, B.M. Colosimo, Process defects and in situ monitoring methods in metal [261] O. Chapelle, P. Haffner, V.N. Vapnik, Support vector machines for histogram-based
powder bed fusion: a review, Meas. Sci. Technol. 28 (4) (2017) 044005. image classification, IEEE Trans. Neural Netw. 10 (5) (1999) 1055–1064.
[231] V. Bhavar, P. Kattire, V. Patil, S. Khot, K. Gujar, R. Singh, A review on powder bed [262] L. Breiman, Random forests, Mach. Learn. 45 (2001) 5–32.
fusion technology of metal additive manufacturing, Addit. Manuf. Handb. (2017) [263] J.H. Friedman, Greedy function approximation: a gradient boosting machine, Ann.
251–253. Stat. (2001) 1189–1232.
[232] D.D. Singh, T. Mahender, A.R. Reddy, Powder bed fusion process: a brief review, [264] M. Grieves, J. Vickers, Digital twin: mitigating unpredictable, undesirable emer-
Mater. Today Proc. 46 (2021) 350–355. gent behavior in complex systems, in: Transdisciplinary Perspectives on Complex
[233] S. Vock, B. Klöden, A. Kirchner, T. Weißgärber, B. Kieback, Powders for powder Systems: New Findings and Approaches, 2017, pp. 85–113.
bed fusion: a review, Progr. Addit. Manuf. 4 (2019) 383–397. [265] A. Krizhevsky, I. Sutskever, G.E. Hinton, Imagenet classification with deep convo-
[234] I.J. Solomon, P. Sevvel, J. Gunasekaran, A review on the various processing pa- lutional neural networks, Adv. Neural Inf. Process. Syst. 25 (2012).
rameters in FDM, Mater. Today Proc. 37 (2021) 509–514. [266] M. Schuster, K.K. Paliwal, Bidirectional recurrent neural networks, IEEE Trans.
[235] M.E. Lamm, L. Wang, V. Kishore, H. Tekinalp, V. Kunc, J. Wang, D.J. Gardner, Signal Process. 45 (11) (1997) 2673–2681.
S. Ozcan, Material extrusion additive manufacturing of wood and lignocellulosic [267] A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A.N. Gomez, Ł. Kaiser,
filled composites, Polymers 12 (9) (2020) 2115. I. Polosukhin, Attention is all you need, Adv. Neural Inf. Process. Syst. 30 (2017).
[236] P. Zhuo, S. Li, I.A. Ashcroft, A.I. Jones, Material extrusion additive manufacturing [268] I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A.
of continuous fibre reinforced polymer matrix composites: a review and outlook, Courville, Y. Bengio, Generative adversarial nets, Adv. Neural Inf. Process. Syst. 27
Composites, Part B, Eng. 224 (2021) 109143. (2014).

41
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

[269] K. Wang, J. Xu, S. Zhang, J. Tan, J. Qin, Towards low-budget energy efficiency [298] N. Després, E. Cyr, P. Setoodeh, M. Mohammadi, Deep learning and design for
design in additive manufacturing based on variational scale-aware transformer, J. additive manufacturing: a framework for microlattice architecture, JOM 72 (2020)
Clean. Prod. 393 (2023) 136168. 2408–2418.
[270] T.M. Mitchell, et al., Machine Learning, vol. 1, McGraw-Hill, New York, 2007. [299] G. Liu, Y. Xiong, D.W. Rosen, Multidisciplinary design optimization in design for
[271] K. Wang, Contrastive learning-based semantic segmentation for in-situ stratified additive manufacturing, J. Comput. Des. Eng. 9 (1) (2022) 128–143.
defect detection in additive manufacturing, J. Manuf. Syst. 68 (2023) 465–476. [300] J. Huang, T.-H. Kwok, C. Zhou, W. Xu, Surfel convolutional neural network for sup-
[272] M.I. Jordan, T.M. Mitchell, Machine learning: trends, perspectives, and prospects, port detection in additive manufacturing, Int. J. Adv. Manuf. Technol. 105 (2019)
Science 349 (6245) (2015) 255–260. 3593–3604.
[273] M. Shafto, M. Conroy, R. Doyle, E. Glaessgen, C. Kemp, J. LeMoigne, L. Wang, [301] N. Hertlein, P.R. Buskohl, A. Gillman, K. Vemaganti, S. Anand, Generative ad-
Modeling, simulation, information technology & processing roadmap, Nat. Aero- versarial network for early-stage design flexibility in topology optimization for
naut. Space Adm. 32 (2012) (2012) 1–38. additive manufacturing, J. Manuf. Syst. 59 (2021) 675–685.
[274] J. Lee, E. Lapira, B. Bagheri, H.-a. Kao, Recent advances and trends in predictive [302] J. Vrábel, P. Pořízka, J. Klus, D. Prochazka, J. Novotnỳ, D. Koutnỳ, D. Paloušek,
manufacturing systems in big data environment, Manuf. Lett. 1 (1) (2013) 38–41. J. Kaiser, Classification of materials for selective laser melting by laser-induced
[275] E.M. Kraft, The air force digital thread/digital twin-life cycle integration and use breakdown spectroscopy, Chem. Pap. 73 (2019) 2897–2905.
of computational and experimental knowledge, in: 54th AIAA Aerospace Sciences [303] B.L. DeCost, H. Jain, A.D. Rollett, E.A. Holm, Computer vision and machine learn-
Meeting, 2016, p. 0897. ing for autonomous characterization of am powder feedstocks, JOM 69 (3) (2017)
[276] E. Negri, L. Fumagalli, M. Macchi, A review of the roles of digital twin in CPS-based 456–465.
production systems, Proc. Manuf. 11 (2017) 939–948. [304] R. Valente, A. Ostapenko, B.C. Sousa, J. Grubbs, C.J. Massar, D.L. Cote, R. Neamtu,
[277] M.G. Mohammadi, D. Mahmoud, M. Elbestawi, On the application of machine Classifying powder flowability for cold spray additive manufacturing using ma-
learning for defect detection in L-PBF additive manufacturing, Opt. Laser Tech- chine learning, in: 2020 IEEE International Conference on Big Data, IEEE, 2020,
nol. 143 (2021) 107338. pp. 2919–2928.
[278] Y. Zhang, G.S. Hong, D. Ye, K. Zhu, J.Y. Fuh, Extraction and evaluation of melt [305] X. Wang, J. Cao, A novel multi-objective optimization of 3D printing adaptive
pool, plume and spatter information for powder-bed fusion am process monitoring, layering algorithm based on improved NSGA-II and fuzzy set theory, Int. J. Adv.
Mater. Des. 156 (2018) 458–469. Manuf. Technol. 123 (3–4) (2022) 957–972.
[279] X. Zhang, J. Saniie, A. Heifetz, Detection of defects in additively manufactured [306] E. Asadollahi-Yazdi, J. Gardan, P. Lafon, Multi-objective optimization of additive
stainless steel 316L with compact infrared camera and machine learning algo- manufacturing process, IFAC-PapersOnLine 51 (11) (2018) 152–157.
rithms, JOM 72 (12) (2020) 4244–4253. [307] D.S. Ye, Y. Fuh, Y. Zhang, G. Hong, K.P. Zhu, Defects Recognition in Selective
[280] L. Lu, J. Hou, S. Yuan, X. Yao, Y. Li, J. Zhu, Deep learning-assisted real-time
Laser Melting with Acoustic Signals by svm Based on Feature Reduction, IOP Con-
defect detection and closed-loop adjustment for additive manufacturing of con-
ference Series: Materials Science and Engineering, vol. 436, IOP Publishing, 2018,
tinuous fiber-reinforced polymer composites, Robot. Comput.-Integr. Manuf. 79
p. 012020.
(2023) 102431.
[308] S.H. Seifi, W. Tian, H. Doude, M.A. Tschopp, L. Bian, Layer-wise modeling and
[281] F. Imani, A. Gaikwad, M. Montazeri, P. Rao, H. Yang, E. Reutzel, Process mapping
anomaly detection for laser-based additive manufacturing, J. Manuf. Sci. Eng.
and in-process monitoring of porosity in laser powder bed fusion using layerwise
141 (8) (2019) 081013.
optical imaging, J. Manuf. Sci. Eng. 140 (10) (2018).
[309] M. Khanzadeh, S. Chowdhury, M. Marufuzzaman, M.A. Tschopp, L. Bian, Porosity
[282] A.-I. García-Moreno, J.-M. Alvarado-Orozco, J. Ibarra-Medina, E. Martínez-Franco,
prediction: supervised-learning of thermal history for direct laser deposition, J.
Image-based porosity classification in Al-alloys by laser metal deposition using
Manuf. Syst. 47 (2018) 69–82.
random forests, Int. J. Adv. Manuf. Technol. 110 (2020) 2827–2845.
[310] H. Gaja, F. Liou, Defect classification of laser metal deposition using logistic re-
[283] G. Tapia, S. Khairallah, M. Matthews, W.E. King, A. Elwany, Gaussian process-
gression and artificial neural networks for pattern recognition, Int. J. Adv. Manuf.
based surrogate modeling framework for process planning in laser powder-bed
Technol. 94 (2018) 315–326.
fusion additive manufacturing of 316L stainless steel, Int. J. Adv. Manuf. Technol.
[311] T. Özel, A. Altay, B. Kaftanoğlu, R. Leach, N. Senin, A. Donmez, Focus variation
94 (2018) 3591–3603.
measurement and prediction of surface texture parameters using machine learning
[284] Z. Lu, D. Li, B. Lu, A. Zhang, G. Zhu, G. Pi, The prediction of the building precision
in laser powder bed fusion, J. Manuf. Sci. Eng. 142 (1) (2020) 011008.
in the laser engineered net shaping process using advanced networks, Opt. Lasers
[312] S. Lee, J. Peng, D. Shin, Y.S. Choi, Data analytics approach for melt-pool geometries
Eng. 48 (5) (2010) 519–525.
in metal additive manufacturing, Sci. Technol. Adv. Mater. 20 (1) (2019) 972–978.
[285] L. Meng, J. Zhang, Process design of laser powder bed fusion of stainless steel using
[313] Y.M. Arısoy, L.E. Criales, T. Özel, B. Lane, S. Moylan, A. Donmez, Influence of
a gaussian process-based machine learning model, JOM 72 (1) (2020) 420–428.
scan strategy and process parameters on microstructure and its optimization in
[286] C. Kamath, Y.J. Fan, Regression with small data sets: a case study using code sur-
additively manufactured nickel alloy 625 via laser powder bed fusion, Int. J. Adv.
rogates in additive manufacturing, Knowl. Inf. Syst. 57 (2018) 475–493.
Manuf. Technol. 90 (2017) 1393–1417.
[287] N. Hertlein, S. Deshpande, V. Venugopal, M. Kumar, S. Anand, Prediction of se-
lective laser melting part quality using hybrid bayesian network, Addit. Manuf. 32 [314] L. Scime, J. Beuth, Using machine learning to identify in-situ melt pool signatures
(2020) 101089. indicative of flaw formation in a laser powder bed fusion additive manufacturing
[288] G.O. Barrionuevo, J.A. Ramos-Grez, M. Walczak, C.A. Betancourt, Comparative process, Addit. Manuf. 25 (2019) 151–165.
evaluation of supervised machine learning algorithms in the prediction of the rela- [315] W. Halsey, J. Ferguson, A. Plotkowski, R. Dehoff, V. Paquit, Geometry-independent
tive density of 316L stainless steel fabricated by selective laser melting, Int. J. Adv. microstructure optimization for electron beam powder bed fusion additive manu-
Manuf. Technol. 113 (2021) 419–433. facturing, Addit. Manuf. 35 (2020) 101354.
[289] A. Gaikwad, B. Giera, G.M. Guss, J.-B. Forien, M.J. Matthews, P. Rao, Heteroge- [316] M. Mehrpouya, A. Gisario, A. Rahimzadeh, M. Nematollahi, K.S. Baghbaderani,
neous sensing and scientific machine learning for quality assurance in laser powder M. Elahinia, A prediction model for finding the optimal laser parameters in addi-
bed fusion–a single-track study, Addit. Manuf. 36 (2020) 101659. tive manufacturing of NiTi shape memory alloy, Int. J. Adv. Manuf. Technol. 105
[290] K. Huang, C. Kain, N. Diaz-Vallejo, Y. Sohn, L. Zhou, High throughput mechani- (2019) 4691–4699.
cal testing platform and application in metal additive manufacturing and process [317] P. He, Q. Liu, J.J. Kruzic, X. Li, Machine-learning assisted additive manufacturing
optimization, J. Manuf. Process. 66 (2021) 494–505. of a TiCN reinforced AlSi10Mg composite with tailorable mechanical properties,
[291] L. Cao, J. Li, J. Hu, H. Liu, Y. Wu, Q. Zhou, Optimization of surface roughness and Mater. Lett. 307 (2022) 131018.
dimensional accuracy in LPBF additive manufacturing, Opt. Laser Technol. 142 [318] L. Meng, J. Zhao, X. Lan, H. Yang, Z. Wang, Multi-objective optimisation of bio-
(2021) 107246. inspired lightweight sandwich structures based on selective laser melting, Virtual
[292] S. Mondal, D. Gwynn, A. Ray, A. Basak, Investigation of melt pool geometry control Phys. Prototyp. 15 (1) (2020) 106–119.
in additive manufacturing using hybrid modeling, Metals 10 (5) (2020) 683. [319] E. Maleki, S. Bagherifard, M. Guagliano, Application of artificial intelligence to op-
[293] L. Song, W. Huang, X. Han, J. Mazumder, Real-time composition monitoring using timize the process parameters effects on tensile properties of Ti-6Al-4V fabricated
support vector regression of laser-induced plasma for laser additive manufacturing, by laser powder-bed fusion, Int. J. Mech. Mater. Des. (2021) 1–24.
IEEE Trans. Ind. Electron. 64 (1) (2016) 633–642. [320] Z. Wang, P. Liu, Y. Xiao, X. Cui, Z. Hu, L. Chen, A data-driven approach for process
[294] X. Yao, S.K. Moon, G. Bi, A hybrid machine learning approach for additive manu- optimization of metallic additive manufacturing under uncertainty, J. Manuf. Sci.
facturing design feature recommendation, Rapid Prototyping J. (2017). Eng. 141 (8) (2019).
[295] H. Ko, P. Witherell, Y. Lu, S. Kim, D.W. Rosen, Machine learning and knowledge [321] K. Aoyagi, H. Wang, H. Sudo, A. Chiba, Simple method to construct process maps
graph based design rule construction for additive manufacturing, Addit. Manuf. 37 for additive manufacturing using a support vector machine, Addit. Manuf. 27
(2021) 101620. (2019) 353–362.
[296] J.D. Alejandrino, R.S. Concepcion II, S.C. Lauguico, R.R. Tobias, L. Venancio, D. [322] D. Wu, Y. Wei, J. Terpenny, Predictive modelling of surface roughness in fused de-
Macasaet, A.A. Bandala, E.P. Dadios, A machine learning approach of lattice infill position modelling using data fusion, Int. J. Prod. Res. 57 (12) (2019) 3992–4006.
pattern for increasing material efficiency in additive manufacturing processes, Int. [323] J. Zhang, P. Wang, R.X. Gao, Deep learning-based tensile strength prediction in
J. Mech. Eng. Robot. Res. 9 (9) (2020) 1253–1263. fused deposition modeling, Comput. Ind. 107 (2019) 11–21.
[297] M. Malviya, K. Desai, Build orientation optimization for strength enhancement of [324] F. Wang, S. Fathizadan, F. Ju, K. Rowe, N. Hofmann, Print surface thermal mod-
FDM parts using machine learning based algorithm, Comput-Aided Des. Appl. 17 eling and layer time control for large-scale additive manufacturing, IEEE Trans.
(2019) 783–796. Autom. Sci. Eng. 18 (1) (2020) 244–254.

42
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

[325] Z. Zhu, N. Anwer, Q. Huang, L. Mathieu, Machine learning in tolerancing for addi- [353] X.Y. Lee, S.K. Saha, S. Sarkar, B. Giera, Automated detection of part quality during
tive manufacturing, CIRP Ann. 67 (1) (2018) 157–160. two-photon lithography via deep learning, Addit. Manuf. 36 (2020) 101444.
[326] N.A. Fountas, N.M. Vaxevanidis, Optimization of fused deposition modeling pro- [354] Y. Shan, A. Krishnakumar, Z. Qin, H. Mao, Smart Resin vat: Real-Time Detecting
cess using a virus-evolutionary genetic algorithm, Comput. Ind. 125 (2021) Failures, Defects, and Curing Area in vat Photopolymerization 3D Printing, Interna-
103371. tional Manufacturing Science and Engineering Conference, vol. 85802, American
[327] E. Luis, H.M. Pan, S.L. Sing, A.K. Bastola, G.D. Goh, G.L. Goh, H.K.J. Tan, R. Bajpai, Society of Mechanical Engineers, 2022, V001T01A030.
J. Song, W.Y. Yeong, Silicone 3D printing: process optimization, product biocom- [355] T. Wang, T.-H. Kwok, C. Zhou, S. Vader, In-situ droplet inspection and closed-loop
patibility, and reliability of silicone meniscus implants, 3D Print. Addit. Manuf. control system using machine learning for liquid metal jet printing, J. Manuf. Syst.
6 (6) (2019) 319–332. 47 (2018) 83–92.
[328] R. Onler, A.S. Koca, B. Kirim, E. Soylemez, Multi-objective optimization of binder [356] M. Ogunsanya, J. Isichei, S.K. Parupelli, S. Desai, Y. Cai, In-situ droplet monitoring
jet additive manufacturing of Co-Cr-Mo using machine learning, Int. J. Adv. Manuf. of inkjet 3D printing process using image analysis and machine learning models,
Technol. 119 (1–2) (2022) 1091–1108. Proc. Manuf. 53 (2021) 427–434.
[329] Y. Li, Y. Sun, Q. Han, G. Zhang, I. Horváth, Enhanced beads overlapping model [357] D. Gunasegaram, A. Barnard, M. Matthews, B. Jared, A. Andreaco, K. Bartsch,
for wire and arc additive manufacturing of multi-layer multi-bead metallic parts, A. Murphy, Machine learning-assisted in-situ adaptive strategies for the control
J. Mater. Process. Technol. 252 (2018) 838–848. of defects and anomalies in metal additive manufacturing, Addit. Manuf. (2024)
[330] D. Ding, F. He, L. Yuan, Z. Pan, L. Wang, M. Ros, The first step towards intelli- 104013.
gent wire arc additive manufacturing: an automatic bead modelling system using [358] Z.U. Arif, M.Y. Khalid, A. Zolfagharian, M. Bodaghi, 4D bioprinting of smart poly-
machine learning through industrial information integration, J. Ind. Inf. Integr. 23 mers for biomedical applications: recent progress, challenges, and future perspec-
(2021) 100218. tives, React. Funct. Polym. (2022) 105374.
[331] M. Naveen Srinivas, K. Vimal, N. Manikandan, G. Sritharanandh, Parametric opti-
[359] S. Park, W. Shou, L. Makatura, W. Matusik, K.K. Fu, 3D printing of polymer com-
mization and multiple regression modelling for fabrication of aluminium alloy thin
posites: materials, processes, and applications, Matter 5 (1) (2022) 43–76.
plate using wire arc additive manufacturing, Int. J. Interact. Des. Manuf. (2022)
[360] S. Singh, S. Ramakrishna, F. Berto, 3D Printing of polymer composites: a short
1–11.
review, Mater. Des. Process. Commun. 2 (2) (2020) e97.
[332] A. Kumar, K. Maji, Selection of process parameters for near-net shape deposition
[361] S.C. Ligon, R. Liska, J. Stampfl, M. Gurr, R. Mülhaupt, Polymers for 3D printing and
in wire arc additive manufacturing by genetic algorithm, J. Mater. Eng. Perform.
customized additive manufacturing, Chem. Rev. 117 (15) (2017) 10212–10290.
29 (2020) 3334–3352.
[333] S.H. Lee, Optimization of cold metal transfer-based wire arc additive manufactur- [362] J.W. Stansbury, M.J. Idacavage, 3D printing with polymers: challenges among ex-
ing processes using gaussian process regression, Metals 10 (4) (2020) 461. panding options and opportunities, Dent. Mater. 32 (1) (2016) 54–64.
[334] X. Gong, Y.C. Yabansu, P.C. Collins, S.R. Kalidindi, Evaluation of Ti-Mn alloys for [363] M. Askari, M.A. Naniz, M. Kouhi, A. Saberi, A. Zolfagharian, M. Bodaghi, Recent
additive manufacturing using high-throughput experimental assays and Gaussian progress in extrusion 3D bioprinting of hydrogel biomaterials for tissue regener-
process regression, Materials 13 (20) (2020) 4641. ation: a comprehensive review with focus on advanced fabrication techniques,
[335] H. Zhang, S.K. Moon, T.H. Ngo, Hybrid machine learning method to determine Biomater. Sci. 9 (3) (2021) 535–573.
the optimal operating process window in aerosol jet 3D printing, ACS Appl. Mater. [364] Q. Ge, A.H. Sakhaei, H. Lee, C.K. Dunn, N.X. Fang, M.L. Dunn, Multimaterial 4D
Interfaces 11 (19) (2019) 17994–18003. printing with tailorable shape memory polymers, Sci. Rep. 6 (1) (2016) 31110.
[336] L.J. Segura, Z. Li, C. Zhou, H. Sun, Droplet evolution prediction in material jetting [365] Q. Ge, Z. Chen, J. Cheng, B. Zhang, Y.-F. Zhang, H. Li, X. He, C. Yuan, J. Liu, S.
via tensor time series analysis, Addit. Manuf. 66 (2023) 103461. Magdassi, et al., 3D printing of highly stretchable hydrogel with diverse UV curable
[337] A. Khadilkar, J. Wang, R. Rai, Deep learning–based stress prediction for bottom-up polymers, Sci. Adv. 7 (2) (2021) eaba4261.
SLA 3D printing process, Int. J. Adv. Manuf. Technol. 102 (2019) 2555–2569. [366] H. Ye, Q. Liu, J. Cheng, H. Li, B. Jian, R. Wang, Z. Sun, Y. Lu, Q. Ge, Multimaterial
[338] R. d, S.B. Ferreira, A. Sabbaghi, Q. Huang, Automated geometric shape deviation 3D printed self-locking thick-panel origami metamaterials, Nat. Commun. 14 (1)
modeling for additive manufacturing systems via bayesian neural networks, IEEE (2023) 1607.
Trans. Autom. Sci. Eng. 17 (2) (2019) 584–598. [367] M. Lalegani Dezaki, M. Bodaghi, Sustainable 4D printing of magneto-electroactive
[339] Z. Shen, X. Shang, M. Zhao, X. Dong, G. Xiong, F.-Y. Wang, A learning-based frame- shape memory polymer composites, Int. J. Adv. Manuf. Technol. 126 (1–2) (2023)
work for error compensation in 3D printing, IEEE Trans. Cybern. 49 (11) (2019) 35–48.
4042–4050. [368] M.Y. Khalid, Z.U. Arif, R. Noroozi, A. Zolfagharian, M. Bodaghi, 4D printing of
[340] G. Strano, L. Hao, R. Everson, K. Evans, Multi-objective optimization of selective shape memory polymer composites: a review on fabrication techniques, applica-
laser sintering processes for surface quality and energy saving, Proc. Inst. Mech. tions, and future perspectives, J. Manuf. Process. 81 (2022) 759–797.
Eng., B J. Eng. Manuf. 225 (9) (2011) 1673–1682. [369] L. Jin, X. Zhai, J. Jiang, K. Zhang, W.-H. Liao, Optimizing stimuli-based 4D printed
[341] J. Yan, I. Battiato, G.M. Fadel, A mathematical model-based optimization method structures: a paradigm shift in programmable material response, in: Sensors and
for direct metal deposition of multimaterials, J. Manuf. Sci. Eng. 139 (8) (2017). Smart Structures Technologies for Civil, Mechanical, and Aerospace Systems 2024,
[342] J. Qin, Y. Liu, R. Grosvenor, F. Lacan, Z. Jiang, Deep learning-driven particle vol. 12949, SPIE, 2024, pp. 321–332.
swarm optimisation for additive manufacturing energy optimisation, J. Clean. [370] Q. Wei, H. Li, G. Liu, Y. He, Y. Wang, Y.E. Tan, D. Wang, X. Peng, G. Yang, N.
Prod. 245 (2020) 118702. Tsubaki, Metal 3D printing technology for functional integration of catalytic sys-
[343] Y. Yang, M. He, L. Li, Power consumption estimation for mask image projection tem, Nat. Commun. 11 (1) (2020) 1–8.
stereolithography additive manufacturing using machine learning based approach, [371] C. Buchanan, L. Gardner, Metal 3D printing in construction: a review of meth-
J. Clean. Prod. 251 (2020) 119710. ods, research, applications, opportunities and challenges, Eng. Struct. 180 (2019)
[344] Y. Li, F. Hu, J. Qin, M. Ryan, R. Wang, Y. Liu, A hybrid machine learning approach 332–348.
for energy consumption prediction in additive manufacturing, in: Pattern Recogni-
[372] S. Das, D.L. Bourell, S. Babu, Metallic materials for 3D printing, Mater. Res. Soc.
tion. ICPR International Workshops and Challenges: Virtual Event, January 10–15,
Bull. 41 (10) (2016) 729–741.
2021, Proceedings, Part IV, Springer, 2021, pp. 622–636.
[373] T. Duda, L.V. Raghavan, 3D metal printing technology, IFAC-PapersOnLine 49 (29)
[345] B. Vaissier, J.-P. Pernot, L. Chougrani, P. Véron, Genetic-algorithm based frame-
(2016) 103–110.
work for lattice support structure optimization in additive manufacturing, Comput.
[374] J. Ding, S. Qu, L. Zhang, M.Y. Wang, X. Song, Geometric deviation and compen-
Aided Des. 110 (2019) 11–23.
sation for thin-walled shell lattice structures fabricated by high precision laser
[346] F. Hu, Y. Liu, J. Qin, X. Sun, P. Witherell, Feature-level data fusion for energy
powder bed fusion, Addit. Manuf. 58 (2022) 103061.
consumption analytics in additive manufacturing, in: 2020 IEEE 16th Interna-
[375] J. Xu, Y. Ding, Y. Gao, B. Liu, S. Xue, Y. Hu, D. Zhang, X. Song, Improving high-
tional Conference on Automation Science and Engineering (CASE), IEEE, 2020,
temperature mechanical properties of laser powder bed-fused Inconel 738 alloy by
pp. 612–617.
[347] F. Hu, J. Qin, Y. Li, Y. Liu, X. Sun, Deep fusion for energy consumption prediction hot isostatic pressing: tailoring precipitates and healing defects, Mater. Sci. Eng. A
in additive manufacturing, Proc. CIRP 104 (2021) 1878–1883. 862 (2023) 144285.
[348] R. Li, M. Jin, V.C. Paquit, Geometrical defect detection for additive manufacturing [376] J. Jiang, X. Zhai, L. Jin, K. Zhang, J. Chen, Q. Lu, W.-H. Liao, Design for reversed
with machine learning models, Mater. Des. 206 (2021) 109726. additive manufacturing low-melting-point alloys, J. Eng. Des. (2023) 1–14.
[349] H. Hu, K. He, T. Zhong, Y. Hong, Fault diagnosis of FDM process based on support [377] Z. Chen, X. Sun, Y. Shang, K. Xiong, Z. Xu, R. Guo, S. Cai, C. Zheng, Dense ceramics
vector machine (SVM), Rapid Prototyping J. 26 (2) (2020) 330–348. with complex shape fabricated by 3D printing: a review, J. Adv. Ceram. 10 (2)
[350] N. Satterlee, E. Torresani, E. Olevsky, J.S. Kang, Automatic detection and charac- (2021) 195–218.
terization of porosities in cross-section images of metal parts produced by binder [378] S.M. Sajadi, L. Vásárhelyi, R. Mousavi, A.H. Rahmati, Z. Kónya, Á. Kukovecz, T.
jetting using machine learning and image augmentation, J. Intell. Manuf. (2023) Arif, T. Filleter, R. Vajtai, P. Boul, et al., Damage-tolerant 3D-printed ceramics via
1–23. conformal coating, Sci. Adv. 7 (28) (2021) eabc5028.
[351] M. Montazeri, A.R. Nassar, C.B. Stutzman, P. Rao, Heterogeneous sensor-based con- [379] Z. Chen, Z. Li, J. Li, C. Liu, C. Lao, Y. Fu, C. Liu, Y. Li, P. Wang, Y. He, 3D printing
dition monitoring in directed energy deposition, Addit. Manuf. 30 (2019) 100916. of ceramics: a review, J. Eur. Ceram. Soc. 39 (4) (2019) 661–687.
[352] W. Ren, G. Wen, Z. Zhang, J. Mazumder, Quality monitoring in additive manufac- [380] L.C. Hwa, S. Rajoo, A.M. Noor, N. Ahmad, M. Uday, Recent advances in 3D printing
turing using emission spectroscopy and unsupervised deep learning, Mater. Manuf. of porous ceramics: a review, Curr. Opin. Solid State Mater. Sci. 21 (6) (2017)
Process. 37 (11) (2022) 1339–1346. 323–347.

43
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

[381] Z.C. Eckel, C. Zhou, J.H. Martin, A.J. Jacobsen, W.B. Carter, T.A. Schaedler, Ad- [411] N.A. Kallioras, N.D. Lagaros, Mlgen: generative design framework based on ma-
ditive manufacturing of polymer-derived ceramics, Science 351 (6268) (2016) chine learning and topology optimization, Appl. Sci. 11 (24) (2021) 12044.
58–62. [412] D.W. Abueidda, S. Koric, N.A. Sobh, Topology optimization of 2D structures with
[382] Y. Weng, M. Li, T.N. Wong, M.J. Tan, Synchronized concrete and bonding agent de- nonlinearities using deep learning, Comput. Struct. 237 (2020) 106283.
position system for interlayer bond strength enhancement in 3D concrete printing, [413] N.S. Iyer, A.M. Mirzendehdel, S. Raghavan, Y. Jiao, E. Ulu, M. Behandish, S. Nela-
Autom. Constr. 123 (2021) 103546. turi, D.M. Robinson, Pato: producibility-aware topology optimization using deep
[383] J.H. Lim, Y. Weng, Q.-C. Pham, 3D printing of curved concrete surfaces using learning for metal additive manufacturing, preprint, arXiv:2112.04552, 2021.
Adaptable Membrane Formwork, Constr. Build. Mater. 232 (2020) 117075. [414] X. Chen, H. Zhang, J. Lin, R. Hu, L. Lu, Q.-X. Huang, B. Benes, D. Cohen-Or, B.
[384] Y. Weng, M. Li, S. Ruan, T.N. Wong, M.J. Tan, K.L.O. Yeong, S. Qian, Comparative Chen, Dapper: decompose-and-pack for 3D printing, ACM Trans. Graph. 34 (6)
economic, environmental and productivity assessment of a concrete bathroom unit (2015) 213–1.
fabricated through 3D printing and a precast approach, J. Clean. Prod. 261 (2020) [415] R. Li, Q. Peng, Deep learning-based optimal segmentation of 3D printed product
121245. for surface quality improvement and support structure reduction, J. Manuf. Syst.
[385] P.P. Kamble, S. Chavan, R. Hodgir, G. Gote, K. Karunakaran, Multi-jet ice 3D print- 60 (2021) 252–264.
ing, Rapid Prototyping J. (2021). [416] B. Ezair, F. Massarwi, G. Elber, Orientation analysis of 3D objects toward minimal
[386] F. Zheng, Z. Wang, J. Huang, Z. Li, Inkjet printing-based fabrication of microscale support volume in 3D-printing, Comput. Graph. 51 (2015) 117–124.
3D ice structures, Microsyst. Nanoeng. 6 (1) (2020) 1–10. [417] W.M. Wang, C. Zanni, L. Kobbelt, Improved Surface Quality in 3D Printing by Op-
[387] A. Pronk, Y. Moonen, C. Ao, P. Luo, Y. Wu, 3D printing of ice, in: Proceedings of timizing the Printing Direction, Computer Graphics Forum, vol. 35, Wiley Online
IASS Annual Symposia, vol. 2017, International Association for Shell and Spatial Library, 2016, pp. 59–70.
Structures (IASS), 2017, pp. 1–8. [418] H. Shen, X. Ye, G. Xu, L. Zhang, J. Qian, J. Fu, 3D printing build orientation opti-
[388] A.K. Das, D.A. Agar, M. Rudolfsson, S.H. Larsson, A review on wood powders in 3D mization for flexible support platform, Rapid Prototyping J. 26 (1) (2020) 59–72.
printing: processes, properties and potential applications, J. Mater. Res. Technol. [419] W. Wang, H. Chao, J. Tong, Z. Yang, X. Tong, H. Li, X. Liu, L. Liu, Saliency-
15 (2021) 241–255. Preserving Slicing Optimization for Effective 3D Printing, Computer Graphics Fo-
[389] A. Le Duigou, M. Castro, R. Bevan, N. Martin, 3D printing of wood fibre biocompos- rum, vol. 34, Wiley Online Library, 2015, pp. 148–160.
ites: from mechanical to actuation functionality, Mater. Des. 96 (2016) 106–114. [420] W. Wang, H. Shao, X. Liu, B. Yin, Printing direction optimization through slice
[390] K. Henke, S. Treml, Wood based bulk material in 3D printing processes for appli- number and support minimization, IEEE Access 8 (2020) 75646–75655.
cations in construction, Eur. J. Wood Prod. 71 (1) (2013) 139–141. [421] N. Umetani, R.M. Schmidt, Cross-sectional structural analysis for 3D printing opti-
[391] B.N. Johnson, A sweet solution to complex microprinting, Science 378 (6622) mization, in: SIGGRAPH Asia Technical Briefs, Citeseer, 2013, pp. 5–1.
(2022) 826–827. [422] T. Wang, N. Li, G. Link, J. Jelonnek, J. Fleischer, J. Dittus, D. Kupzik, Load-
[392] P.Y.V. Leung, Sugar 3D printing: additive manufacturing with molten sugar for
dependent path planning method for 3D printing of continuous fiber reinforced
investigating molten material fed printing, 3D Print. Addit. Manuf. 4 (1) (2017)
plastics, Composites, Part A, Appl. Sci. Manuf. 140 (2021) 106181.
13–18.
[423] R.D. Weeks, R.L. Truby, S.G. Uzel, J.A. Lewis, Embedded 3D printing of multi-
[393] A. Dey, I.N. Roan Eagle, N. Yodo, A review on filament materials for fused filament
material polymer lattices via graph-based print path planning, Adv. Mater. 35 (5)
fabrication, J. Manuf. Mater. Process. 5 (3) (2021) 69.
(2023) 2206958.
[394] N. Johnson, P. Vulimiri, A. To, X. Zhang, C. Brice, B. Kappes, A. Stebner, Invited
[424] S. Elagandula, L. Poudel, Z. Sha, W. Zhou, Multi-robot path planning for coopera-
review: machine learning for materials developments in metals additive manufac-
tive 3D printing, in: International Manufacturing Science and Engineering Confer-
turing, Addit. Manuf. 36 (2020) 101641.
ence, vol. 84256, American Society of Mechanical Engineers, 2020, V001T01A034.
[395] J.H. Panchal, S.R. Kalidindi, D.L. McDowell, Key computational modeling issues in
[425] A.V. Shembekar, Y.J. Yoon, A. Kanyuck, S.K. Gupta, Trajectory planning for con-
integrated computational materials engineering, Comput. Aided Des. 45 (1) (2013)
formal 3D printing using non-planar layers, in: International Design Engineering
4–25.
Technical Conferences and Computers and Information in Engineering Conference,
[396] P.-A. Pires, O. Desmaison, M. Megahed, Icme manufacturability assessment in pow-
vol. 51722, American Society of Mechanical Engineers, 2018, V01AT02A026.
der bed fusion additive manufacturing, JOM 70 (2018) 1677–1685.
[426] H. Guo, J. Xu, S. Zhang, Y. Zhang, J. Tan, Multi-orientation optimization of com-
[397] R. Kannan, G.L. Knapp, P. Nandwana, R. Dehoff, A. Plotkowski, B. Stump, Y. Yang,
plex parts based on model segmentation in additive manufacturing, J. Mech. Sci.
V. Paquit, Data mining and visualization of high-dimensional icme data for additive
Technol. (2023) 1–15.
manufacturing, Integr. Mater. Manuf. Innov. 11 (1) (2022) 57–70.
[427] X. Zhang, X. Le, A. Panotopoulou, E. Whiting, C.C. Wang, Perceptual models of
[398] M. Seifi, A. Salem, J. Beuth, O. Harrysson, J.J. Lewandowski, Overview of materials
preference in 3D printing direction, ACM Trans. Graph. 34 (6) (2015) 1–12.
qualification needs for metal additive manufacturing, JOM 68 (2016) 747–764.
[428] P. Shi, Q. Qi, Y. Qin, F. Meng, S. Lou, P.J. Scott, X. Jiang, Learn to rotate: part
[399] W.Y. Wang, J. Li, W. Liu, Z.-K. Liu, Integrated computational materials engineering
orientation for reducing support volume via generalizable reinforcement learning,
for advanced materials: a brief review, Comput. Mater. Sci. 158 (2019) 42–48.
[400] S.A.H. Motaman, F. Kies, P. Köhnen, M. Létang, M. Lin, A. Molotnikov, C. Haase, IEEE Trans. Ind. Inform. (2023).
Optimal design for metal additive manufacturing: an integrated computational ma- [429] M. Qin, S. Qu, J. Ding, X. Song, S. Gao, C.C. Wang, W.-H. Liao, Adaptive tool-
terials engineering (icme) approach, JOM 72 (2020) 1092–1104. path generation for distortion reduction in laser powder bed fusion process, Addit.
[401] X. Wang, W. Xiong, Uncertainty quantification and composition optimization for Manuf. 64 (2023) 103432.
alloy additive manufacturing through a calphad-based icme framework, npj Com- [430] J. Ge, Y. Wang, J. Li, H. Bai, L. Liu, S. Wang, X. Xue, F. Li, A reinforcement learning-
put. Mater. 6 (1) (2020) 188. based path planning method for complex thin-walled structures in 3D printing,
[402] J. Jiang, A. Lu, X. Ma, D. Ouyang, R.O. Williams III, The applications of machine in: 2021 the 5th International Conference on Innovation in Artificial Intelligence,
learning to predict the forming of chemically stable amorphous solid dispersions 2021, pp. 234–240.
prepared by hot-melt extrusion, Int. J. Pharm. X 5 (2023) 100164. [431] K. Yanamandra, G.L. Chen, X. Xu, G. Mac, N. Gupta, Reverse engineering of ad-
[403] D. Chernyavsky, D.Y. Kononenko, J.H. Han, H.J. Kim, J. van den Brink, K. Kosiba, ditive manufactured composite part by toolpath reconstruction using imaging and
Machine learning for additive manufacturing: predicting materials characteristics machine learning, Compos. Sci. Technol. 198 (2020) 108318.
and their uncertainty, Mater. Des. 227 (2023) 111699. [432] W.L. Ng, A. Chan, Y.S. Ong, C.K. Chua, Deep learning for fabrication and matu-
[404] R.R. Schmidt, J. Hildebrand, I. Kraljevski, F. Duckhorn, C. Tschöpe, A study for ration of 3D bioprinted tissues and organs, Virtual Phys. Prototyp. 15 (3) (2020)
laser additive manufacturing quality and material classification using machine 340–358.
learning, in: 2022 IEEE Sensors, IEEE, 2022, pp. 1–4. [433] M.R. Rezaei, M. Houshmand, O. Fatahi Valilai, An autonomous intelligent frame-
[405] S. Yang, Y.F. Zhao, Additive manufacturing-enabled design theory and methodol- work for optimal orientation detection in 3D printing, Int. J. Comput. Integr.
ogy: a critical review, Int. J. Adv. Manuf. Technol. 80 (2015) 327–342. Manuf. (2023) 1–39.
[406] K.V. Bhat, G. Capasso, S. Coniglio, J. Morlier, C. Gogu, On some applications of [434] M.A. El youbi El idrissi, L. Laaouina, A. Jeghal, H. Tairi, M. Zaki, Energy consump-
Generalized Geometric Projection to optimal 3D printing, Comput. Graph. 102 tion prediction for fused deposition modelling 3D printing using machine learning,
(2022) 199–212. Appl. Syst. Innov. 5 (4) (2022) 86.
[407] V. Hassani, Z. Khabazi, F. Raspall, C. Banon, D. Rosen, Form-finding and struc- [435] D. Checa, G. Urbikain, A. Beranoagirre, A. Bustillo, L.N. Lopez de Lacalle, Using
tural shape optimization of the metal 3D-printed node, in: Proceedings of CAD’19, machine-learning techniques and virtual reality to design cutting tools for energy
Singapore, 2019, pp. 24–28. optimization in milling operations, Int. J. Comput. Integr. Manuf. 35 (9) (2022)
[408] G. Kazakis, I. Kanellopoulos, S. Sotiropoulos, N.D. Lagaros, Topology optimization 951–971.
aided structural design: interpretation, computational aspects and 3D printing, He- [436] T.S. Tamir, G. Xiong, Q. Fang, Y. Yang, Z. Shen, M. Zhou, J. Jiang, Machine-
liyon 3 (10) (2017) e00431. learning-based monitoring and optimization of processing parameters in 3D print-
[409] Y. Wang, W. Du, H. Wang, Y. Zhao, Intelligent generation method of innovative ing, Int. J. Comput. Integr. Manuf. (2022) 1–17.
structures based on topology optimization and deep learning, Materials 14 (24) [437] Z. Jin, Z. Zhang, J. Ott, G.X. Gu, Precise localization and semantic segmentation
(2021) 7680. detection of printing conditions in fused filament fabrication technologies using
[410] Y. Maksum, A. Amirli, A. Amangeldi, M. Inkarbekov, Y. Ding, A. Romagnoli, S. machine learning, Addit. Manuf. 37 (2021) 101696.
Rustamov, B. Akhmetov, Computational acceleration of topology optimization us- [438] S. Patrick, A. Nycz, M. Noakes, Reinforcement learning for generating toolpaths in
ing parallel computing and machine learning methods–analysis of research trends, additive manufacturing, in: 2018 International Solid Freeform Fabrication Sympo-
J. Ind. Inf. Integr. 28 (2022) 100352. sium, University of Texas at Austin, 2018.

44
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

[439] K.-Y. Fok, N. Ganganath, C.-T. Cheng, H.H.-C. Iu, K.T. Chi, Tool-path optimization [469] W. Ma, Z. Liu, Z.A. Kudyshev, A. Boltasseva, W. Cai, Y. Liu, Deep learning for the
using neural networks, in: 2019 IEEE International Symposium on Circuits and design of photonic structures, Nat. Photonics 15 (2) (2021) 77–90.
Systems (ISCAS), IEEE, 2019, pp. 1–5. [470] H. Liu, Z. Fu, K. Yang, X. Xu, M. Bauchy, Machine learning for glass science and
[440] Z. Xie, N. Somani, Y.J.S. Tan, J.C.Y. Seng, Automatic toolpath pattern recom- engineering: a review, J. Non-Cryst. Solids 557 (2021) 119419.
mendation for various industrial applications based on deep learning, in: 2021 [471] J. Seo, R.K. Kapania, Development of deep convolutional neural network for struc-
IEEE/SICE International Symposium on System Integration (SII), IEEE, 2021, tural topology optimization, AIAA J. 61 (3) (2023) 1366–1379.
pp. 60–65. [472] A. Prathyusha, G.R. Babu, A review on additive manufacturing and topology op-
[441] B. Crockett, M. Borish, Toolpath planning for multiple build points using k-means timization process for weight reduction studies in various industrial applications,
clustering, in: 2022 International Solid Freeform Fabrication Symposium, 2022. Mater. Today Proc. 62 (2022) 109–117.
[442] M.P. Bendsøe, N. Kikuchi, Generating optimal topologies in structural design us- [473] Y. Xing, L. Tong, An online autonomous learning and prediction scheme for ma-
ing a homogenization method, Comput. Methods Appl. Mech. Eng. 71 (2) (1988) chine learning assisted structural optimization, Thin-Walled Struct. 184 (2023)
197–224. 110500.
[443] O. Sigmund, On the design of compliant mechanisms using topology optimization, [474] T. Yoo, S. Lee, K. Yoo, H. Kim, Reinforcement learning based topology control for
J. Struct. Mech. 25 (4) (1997) 493–524. UAV networks, Sensors 23 (2) (2023) 921.
[444] H. Chi, Y. Zhang, T.L.E. Tang, L. Mirabella, L. Dalloro, L. Song, G.H. Paulino, Uni- [475] S.K. Patel, J. Surve, J. Parmar, A. Natesan, V. Katkar, Graphene-based metasurface
versal machine learning for topology optimization, Comput. Methods Appl. Mech. refractive index biosensor for hemoglobin detection: machine learning assisted op-
Eng. 375 (2021) 112739. timization, IEEE Trans. Nanobiosci. (2022).
[445] J. Tian, K. Tang, X. Chen, X. Wang, Machine learning-based prediction and in- [476] J.K. Wilt, C. Yang, G.X. Gu, Accelerating auxetic metamaterial design with deep
verse design of 2D metamaterial structures with tunable deformation-dependent learning, Adv. Eng. Mater. 22 (5) (2020) 1901266.
Poisson’s ratio, Nanoscale 14 (35) (2022) 12677–12691. [477] S. Jafar-Zanjani, M.M. Salary, D. Huynh, E. Elhamifar, H. Mosallaei, Tco-based ac-
[446] H.T. Kollmann, D.W. Abueidda, S. Koric, E. Guleryuz, N.A. Sobh, Deep learning for tive dielectric metasurfaces design by conditional generative adversarial networks,
topology optimization of 2D metamaterials, Mater. Des. 196 (2020) 109098. Adv. Theory Simul. 4 (2) (2021) 2000196.
[447] H. Wang, S.-H. Xiao, C. Zhang, Novel planar auxetic metamaterial perforated with [478] C. Gurbuz, F. Kronowetter, C. Dietz, M. Eser, J. Schmid, S. Marburg, Generative
orthogonally aligned oval-shaped holes and machine learning solutions, Adv. Eng. adversarial networks for the design of acoustic metamaterials, J. Acoust. Soc. Am.
Mater. 23 (7) (2021) 2100102. 149 (2) (2021) 1162–1174.
[448] A. Challapalli, J. Konlan, G. Li, Inverse machine learning discovered metamaterials [479] A. Challapalli, D. Patel, G. Li, Inverse machine learning framework for optimizing
with record high recovery stress, Int. J. Mech. Sci. 244 (2023) 108029. lightweight metamaterials, Mater. Des. 208 (2021) 109937.
[449] F.V. Senhora, H. Chi, Y. Zhang, L. Mirabella, T.L.E. Tang, G.H. Paulino, Machine [480] T. Gahlmann, P. Tassin, Deep neural networks for the prediction of the optical
learning for topology optimization: physics-based learning through an independent properties and the free-form inverse design of metamaterials, Phys. Rev. B 106 (8)
training strategy, Comput. Methods Appl. Mech. Eng. 398 (2022) 115116. (2022) 085408.
[450] Z. Xia, H. Zhang, Z. Zhuang, C. Yu, J. Yu, L. Gao, A machine-learning framework [481] O. Khatib, S. Ren, J. Malof, W.J. Padilla, Learning the physics of all-dielectric
for isogeometric topology optimization, Struct. Multidiscip. Optim. 66 (4) (2023) metamaterials with deep Lorentz neural networks, Adv. Opt. Mater. 10 (13) (2022)
83. 2200097.
[451] H. Jeong, J. Bai, C.P. Batuwatta-Gamage, C. Rathnayaka, Y. Zhou, Y. Gu, A [482] J. Chen, W. Ding, X.-M. Li, X. Xi, K.-P. Ye, H.-B. Wu, R.-X. Wu, Absorption
physics-informed neural network-based topology optimization (pinnto) framework and diffusion enabled ultrathin broadband metamaterial absorber designed by
for structural optimization, Eng. Struct. 278 (2023) 115484. deep neural network and pso, IEEE Antennas Wirel. Propag. Lett. 20 (10) (2021)
[452] J. Seo, R.K. Kapania, Topology optimization with advanced cnn using mapped 1993–1997.
physics-based data, Struct. Multidiscip. Optim. 66 (1) (2023) 21. [483] T. Shah, L. Zhuo, P. Lai, A. De La Rosa-Moreno, F. Amirkulova, P. Gerstoft, Re-
[453] C.M. Parrott, D.W. Abueidda, K.A. James, Multidisciplinary topology optimization inforcement learning applied to metamaterial design, J. Acoust. Soc. Am. 150 (1)
using generative adversarial networks for physics-based design enhancement, J. (2021) 321–338.
Mech. Des. 145 (6) (2023) 061704. [484] L. Rosafalco, J.M. De Ponti, L. Iorio, R. Ardito, A. Corigliano, Optimised graded
[454] L. Wang, D. Shi, B. Zhang, G. Li, W.M. Helal, M. Qi, Deep learning driven real time metamaterials for mechanical energy confinement and amplification via reinforce-
topology optimization based on improved convolutional block attention (Cba-U- ment learning, Eur. J. Mech. A, Solids 99 (2023) 104947.
Net) model, Eng. Anal. Bound. Elem. 147 (2023) 112–124. [485] L. Rosafalco, J.M. De Ponti, L. Iorio, R. Ardito, A. Corigliano, Optimised graded
[455] A. Chandrasekhar, A. Mirzendehdel, M. Behandish, K. Suresh, Frc-tounn: topol- metamaterials for energy harvesting via reinforcement learning, preprint, arXiv:
ogy optimization of continuous fiber reinforced composites using neural network, 2211.09528, 2022.
Comput. Aided Des. 156 (2023) 103449. [486] S.B. Oliva, F.T. Bölle, A. Las, X. Xia, I. Castelli, Reinforcement Learning-Based De-
[456] A. Chandrasekhar, S. Sridhara, K. Suresh, Graded multiscale topology optimization sign of Shape-Changing Metamaterials, 2023.
using neural networks, Adv. Eng. Softw. 175 (2023) 103359. [487] S. Kumar, S. Tan, L. Zheng, D.M. Kochmann, Inverse-designed spinodoid metama-
[457] S. Liu, Q. Li, J. Liu, W. Chen, Y. Zhang, A realization method for transforming a terials, npj Comput. Mater. 6 (1) (2020) 73.
topology optimization design into additive manufacturing structures, Engineering [488] Y. Wang, Q. Zeng, J. Wang, Y. Li, D. Fang, Inverse design of shell-based mechan-
4 (2) (2018) 277–285. ical metamaterial with customized loading curves based on machine learning and
[458] W. Wang, D. Feng, L. Yang, S. Li, C.C. Wang, Topology optimization of self- genetic algorithm, Comput. Methods Appl. Mech. Eng. 401 (2022) 115571.
supporting lattice structure, Addit. Manuf. 67 (2023) 103507. [489] A.P. Garland, B.C. White, S.C. Jensen, B.L. Boyce, Pragmatic generative optimiza-
[459] H. Li, Z. Luo, L. Gao, Q. Qin, Topology optimization for concurrent design of struc- tion of novel structural lattice metamaterials with machine learning, Mater. Des.
tures with multi-patch microstructures by level sets, Comput. Methods Appl. Mech. 203 (2021) 109632.
Eng. 331 (2018) 536–561. [490] Y. Chang, H. Wang, Q. Dong, Machine learning-based inverse design of auxetic
[460] C. Lu, M. Hsieh, Z. Huang, C. Zhang, Y. Lin, Q. Shen, F. Chen, L. Zhang, Architec- metamaterial with zero poisson’s ratio, Mater. Today Commun. 30 (2022) 103186.
tural design and additive manufacturing of mechanical metamaterials: a review, [491] Y. Singh, J. Singh, S. Sharma, A. Sharma, J.S. Chohan, Process parameter opti-
Engineering 17 (2022) 44–63. mization in laser cutting of coir fiber reinforced epoxy composite-a review, Mater.
[461] S. Qian, H. Liu, Y. Wang, D. Mei, Structural optimization of 3D printed SiC scaf- Today Proc. 48 (2022) 1021–1027.
fold with gradient pore size distribution as catalyst support for methanol steam [492] R. Chaudhari, H. Parmar, J. Vora, V.K. Patel, Parametric study and investigations
reforming, Fuel 341 (2023) 127612. of bead geometries of GMAW-based wire–arc additive manufacturing of 316L stain-
[462] Z. Gan, H. Li, S.J. Wolff, J.L. Bennett, G. Hyatt, G.J. Wagner, J. Cao, W.K. Liu, less steels, Metals 12 (7) (2022) 1232.
Data-driven microstructure and microhardness design in additive manufacturing [493] Z. Chen, C. Han, M. Gao, S.Y. Kandukuri, K. Zhou, A review on qualification
using a self-organizing map, Engineering 5 (4) (2019) 730–735. and certification for metal additive manufacturing, Virtual Phys. Prototyp. 17 (2)
[463] I. Sosnovik, I. Oseledets, Neural networks for topology optimization, Russ. J. Nu- (2022) 382–405.
mer. Anal. Math. Model. 34 (4) (2019) 215–223. [494] N. Fountas, J. Kechagias, D. Manolakos, N. Vaxevanidis, Single and multi-objective
[464] X. Lei, C. Liu, Z. Du, W. Zhang, X. Guo, Machine learning-driven real-time topol- optimization of FDM-based additive manufacturing using metaheuristic algo-
ogy optimization under moving morphable component-based framework, J. Appl. rithms, Proc. Manuf. 51 (2020) 740–747.
Mech. 86 (1) (2019) 011004. [495] N. Padhye, K. Deb, Multi-objective optimisation and multi-criteria decision making
[465] Y. Yu, T. Hur, J. Jung, I.G. Jang, Deep learning for determining a near-optimal in SLS using evolutionary approaches, Rapid Prototyping J. (2011).
topological design without any iteration, Struct. Multidiscip. Optim. 59 (3) (2019) [496] O.A. Mohamed, S.H. Masood, J.L. Bhowmik, Mathematical modeling and FDM
787–799. process parameters optimization using response surface methodology based on Q-
[466] C. Kim, J. Lee, J. Yoo, Machine learning-combined topology optimization for func- optimal design, Appl. Math. Model. 40 (23–24) (2016) 10052–10073.
tionary graded composite structure design, Comput. Methods Appl. Mech. Eng. 387 [497] V. Wankhede, D. Jagetiya, A. Joshi, R. Chaudhari, Experimental investigation of
(2021) 114158. FDM process parameters using Taguchi analysis, Mater. Today Proc. 27 (2020)
[467] A. Chandrasekhar, K. Suresh, Tounn: topology optimization using neural networks, 2117–2120.
Struct. Multidiscip. Optim. 63 (2021) 1135–1149. [498] A. Peng, X. Xiao, R. Yue, Process parameter optimization for fused deposition mod-
[468] X. Jiang, H. Wang, Y. Li, K. Mo, Machine Learning based parameter tuning strategy eling using response surface methodology combined with fuzzy inference system,
for MMC based topology optimization, Adv. Eng. Softw. 149 (2020) 102841. Int. J. Adv. Manuf. Technol. 73 (1–4) (2014) 87–100.

45
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

[499] P.K. Gurrala, S.P. Regalla, Optimization of Support Material and Build Time in [525] N. Garg, V. Rastogi, P. Kumar, Process parameter optimization on the dimensional
Fused Deposition Modeling (FDM), Applied Mechanics and Materials, vol. 110, accuracy of additive manufacture Thermoplastic Polyurethane (TPU) using RSM,
Trans Tech Publ, 2012, pp. 2245–2251. Mater. Today Proc. 62 (2022) 94–99.
[500] M. Srivastava, S. Rathee, S. Maheshwari, T. Kundra, Multi-objective optimisation [526] W. Mycroft, M. Katzman, S. Tammas-Williams, E. Hernandez-Nava, G. Panoutsos,
of fused deposition modelling process parameters using RSM and fuzzy logic for I. Todd, V. Kadirkamanathan, A data-driven approach for predicting printability in
build time and support material, Int. J. Rapid Manuf. 7 (1) (2018) 25–42. metal additive manufacturing processes, J. Intell. Manuf. 31 (2020) 1769–1781.
[501] A. Rinanto, A. Nugroho, H. Prasetyo, E. Pujiyanto, Simultaneous optimization of [527] Q. Chen, E. Juste, M. Lasgorceix, F. Petit, A. Leriche, Binder jetting process with
tensile strength, energy consumption, and processing time on FDM process using ceramic powders: influence of powder properties and printing parameters, Open
Taguchi and PCR-TOPSIS, in: 2018 4th International Conference on Science and Ceram. 9 (2022) 100218.
Technology (ICST), IEEE, 2018, pp. 1–5. [528] M. Raju, M.K. Gupta, N. Bhanot, V.S. Sharma, A hybrid pso–bfo evolutionary algo-
[502] D. Coupek, J. Friedrich, D. Battran, O. Riedel, Reduction of support structures rithm for optimization of fused deposition modelling process parameters, J. Intell.
and building time by optimized path planning algorithms in multi-axis additive Manuf. 30 (2019) 2743–2758.
manufacturing, Proc. CIRP 67 (2018) 221–226. [529] B. Yang, Y. Lai, X. Yue, D. Wang, Y. Zhao, Parametric optimization of laser additive
[503] A. Pajonk, A. Prieto, U. Blum, U. Knaack, Multi-material additive manufacturing manufacturing of Inconel 625 using Taguchi method and grey relational analysis,
in architecture and construction: a review, J. Build. Eng. 45 (2022) 103603. Scanning 2020 (2020).
[504] E. Ulu, R. Huang, L.B. Kara, K.S. Whitefoot, Concurrent structure and process op- [530] C. Lu, J. Shi, Relative density and surface roughness prediction for Inconel 718 by
timization for minimum cost metal additive manufacturing, J. Mech. Des. 141 (6) selective laser melting: central composite design and multi-objective optimization,
(2019). Int. J. Adv. Manuf. Technol. (2022) 1–19.
[505] R. Huang, E. Ulu, L.B. Kara, K.S. Whitefoot, Cost minimization in metal additive [531] C. Xia, Z. Pan, J. Polden, H. Li, Y. Xu, S. Chen, Modelling and prediction of surface
manufacturing using concurrent structure and process optimization, in: Interna- roughness in wire arc additive manufacturing using machine learning, J. Intell.
tional Design Engineering Technical Conferences and Computers and Information Manuf. (2022) 1–16.
in Engineering Conference, vol. 58127, American Society of Mechanical Engineers, [532] A. Yaseer, H. Chen, Machine learning based layer roughness modeling in robotic
2017, V02AT03A030. additive manufacturing, J. Manuf. Process. 70 (2021) 543–552.
[506] S.H. Nikam, N.K. Jain, M.S. Sawant, Optimization of parameters of micro-plasma [533] A. Selvam, S. Mayilswamy, R. Whenish, K. Naresh, V. Shanmugam, O. Das, Multi-
transferred arc additive manufacturing process using real coded genetic algorithm, objective optimization and prediction of surface roughness and printing time in
Int. J. Adv. Manuf. Technol. 106 (2020) 1239–1252. FFF printed ABS polymer, Sci. Rep. 12 (1) (2022) 16887.
[507] W. Wu, J. Xue, W. Xu, H. Lin, H. Tang, P. Yao, Parameters optimization of auxil- [534] D. Dong, C. Chang, H. Wang, X. Yan, W. Ma, M. Liu, S. Deng, J. Gardan, R. Bolot,
iary gas process for double-wire SS316L stainless steel arc additive manufacturing, H. Liao, Selective laser melting (SLM) of CX stainless steel: theoretical calcula-
Metals 11 (2) (2021) 190. tion, process optimization and strengthening mechanism, J. Mater. Sci. Technol.
[508] A.M. Aboutaleb, L. Bian, A. Elwany, N. Shamsaei, S.M. Thompson, G. Tapia, Accel- 73 (2021) 151–164.
erated process optimization for laser-based additive manufacturing by leveraging
[535] M. Oyesola, K. Mpofu, N. Mathe, S. Fatoba, S. Hoosain, I. Daniyan, Optimization of
similar prior studies, IISE Trans. 49 (1) (2017) 31–44.
selective laser melting process parameters for surface quality performance of the
[509] T. Peng, J. Lv, A. Majeed, X. Liang, An experimental investigation on energy-
fabricated Ti6Al4V, Int. J. Adv. Manuf. Technol. 114 (2021) 1585–1599.
effective additive manufacturing of aluminum parts via process parameter selec-
[536] J. Gockel, L. Sheridan, B. Koerper, B. Whip, The influence of additive manufactur-
tion, J. Clean. Prod. 279 (2021) 123609.
ing processing parameters on surface roughness and fatigue life, Int. J. Fatigue 124
[510] M. Khalid, Q. Peng, Investigation of printing parameters of additive manufactur-
(2019) 380–388.
ing process for sustainability using design of experiments, J. Mech. Des. 143 (3)
[537] B. Whip, L. Sheridan, J. Gockel, The effect of primary processing parameters on
(2021).
surface roughness in laser powder bed additive manufacturing, Int. J. Adv. Manuf.
[511] J. Watson, K. Taminger, A decision-support model for selecting additive manufac-
Technol. 103 (2019) 4411–4422.
turing versus subtractive manufacturing based on energy consumption, J. Clean.
[538] J.C. Fox, S.P. Moylan, B.M. Lane, Effect of process parameters on the surface rough-
Prod. 176 (2018) 1316–1322.
ness of overhanging structures in laser powder bed fusion additive manufacturing,
[512] Z. Liu, C. Li, X. Fang, Y. Guo, Energy consumption in additive manufacturing of
Proc. CIRP 45 (2016) 131–134.
metal parts, Proc. Manuf. 26 (2018) 834–845.
[539] Z. Li, Z. Zhang, J. Shi, D. Wu, Prediction of surface roughness in extrusion-based
[513] Z. Ma, M. Gao, Q. Wang, N. Wang, L. Li, C. Liu, Z. Liu, Energy consumption dis-
additive manufacturing with machine learning, Robot. Comput.-Integr. Manuf. 57
tribution and optimization of additive manufacturing, Int. J. Adv. Manuf. Technol.
(2019) 488–495.
116 (2021) 3377–3390.
[540] J. Sun, Y. Yang, D. Wang, Parametric optimization of selective laser melting for
[514] A. Verma, R. Rai, Energy efficient modeling and optimization of additive manu-
forming Ti6Al4V samples by Taguchi method, Opt. Laser Technol. 49 (2013)
facturing process, in: 2013 International Solid Freeform Fabrication Symposium,
118–124.
University of Texas at Austin, 2013.
[515] S. Peng, T. Li, J. Zhao, S. Lv, G.Z. Tan, M. Dong, H. Zhang, Towards energy and [541] M. Yakout, M. Elbestawi, S.C. Veldhuis, Density and mechanical properties in selec-
material efficient laser cladding process: modeling and optimization using a hybrid tive laser melting of Invar 36 and stainless steel 316L, J. Mater. Process. Technol.
TS-GEP algorithm and the NSGA-II, J. Clean. Prod. 227 (2019) 58–69. 266 (2019) 397–420.
[516] A.K. Singh, A. Sadhu, A.K. Das, D.K. Pratihar, A.R. Choudhury, An approach to- [542] A.M. Aboutaleb, M.J. Mahtabi, M.A. Tschopp, L. Bian, Multi-objective accelerated
wards energy and material efficient additive manufacturing: multi-objective opti- process optimization of mechanical properties in laser-based additive manufactur-
mization of stellite-6 deposition on SS304, Opt. Laser Technol. 148 (2022) 107799. ing: case study on Selective Laser Melting (SLM) Ti-6Al-4V, J. Manuf. Process. 38
[517] A.G. Dharmawan, Y. Xiong, S. Foong, G.S. Soh, A model-based reinforcement learn- (2019) 432–444.
ing and correction framework for process control of robotic wire arc additive [543] D.-s. Shim, Effects of process parameters on additive manufacturing of aluminum
manufacturing, in: 2020 IEEE International Conference on Robotics and Automa- porous materials and their optimization using response surface method, J. Mater.
tion (ICRA), IEEE, 2020, pp. 4030–4036. Res. Technol. 15 (2021) 119–134.
[518] A. Alafaghani, A. Qattawi, B. Alrawi, A. Guzman, Experimental optimization of [544] G. Tapia, A.H. Elwany, H. Sang, Prediction of porosity in metal-based additive
fused deposition modelling processing parameters: a design-for-manufacturing ap- manufacturing using spatial gaussian process models, Addit. Manuf. 12 (2016)
proach, Proc. Manuf. 10 (2017) 791–803. 282–290.
[519] S. Mahmood, A. Qureshi, D. Talamona, Taguchi based process optimization for [545] J.A. Lee, M.J. Sagong, J. Jung, E.S. Kim, H.S. Kim, Explainable machine learning
dimension and tolerance control for fused deposition modelling, Addit. Manuf. 21 for understanding and predicting geometry and defect types in Fe-Ni alloys fabri-
(2018) 183–190. cated by laser metal deposition additive manufacturing, J. Mater. Res. Technol. 22
[520] A.H. Maamoun, Y.F. Xue, M.A. Elbestawi, S.C. Veldhuis, Effect of selective laser (2023) 413–423.
melting process parameters on the quality of Al alloy parts: powder characteri- [546] S. Bland, N.T. Aboulkhair, Reducing porosity in additive manufacturing, Metal
zation, density, surface roughness, and dimensional accuracy, Materials 11 (12) Powder Rep. 70 (2) (2015) 79–81.
(2018) 2343. [547] A. Sola, A. Nouri, Microstructural porosity in additive manufacturing: the forma-
[521] C. Wacker, M. Köhler, M. David, F. Aschersleben, F. Gabriel, J. Hensel, K. Dilger, K. tion and detection of pores in metal parts fabricated by powder bed fusion, J. Adv.
Dröder, Geometry and distortion prediction of multiple layers for wire arc additive Manuf. Process. 1 (3) (2019) e10021.
manufacturing with artificial neural networks, Appl. Sci. 11 (10) (2021) 4694. [548] G. Kasperovich, J. Haubrich, J. Gussone, G. Requena, Correlation between porosity
[522] M.A. Matos, A.M.A. Rocha, A.I. Pereira, Improving additive manufacturing perfor- and processing parameters in TiAl6V4 produced by selective laser melting, Mater.
mance by build orientation optimization, Int. J. Adv. Manuf. Technol. 107 (2020) Des. 105 (2016) 160–170.
1993–2005. [549] H. Gong, K. Rafi, H. Gu, T. Starr, B. Stucker, Analysis of defect generation in Ti–
[523] H. Chen, Y.F. Zhao, Process parameters optimization for improving surface qual- 6Al–4V parts made using powder bed fusion additive manufacturing processes,
ity and manufacturing accuracy of binder jetting additive manufacturing process, Addit. Manuf. 1 (2014) 87–98.
Rapid Prototyping J. 22 (3) (2016) 527–538. [550] A. Singh, D. Cooper, N. Blundell, G. Gibbons, D. Pratihar, Modelling of direct metal
[524] A.M. Aboutaleb, M.A. Tschopp, P.K. Rao, L. Bian, Multi-objective accelerated pro- laser sintering of eos dm20 bronze using neural networks and genetic algorithms,
cess optimization of part geometric accuracy in additive manufacturing, J. Manuf. in: Proceedings of the 37th International MATADOR Conference, Springer Science
Sci. Eng. 139 (10) (2017). & Business Media, 2012, p. 395.

46
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

[551] B. Kappes, S. Moorthy, D. Drake, H. Geerlings, A. Stebner, Machine learning to [576] C. Wang, X. Tan, E. Liu, S.B. Tor, Process parameter optimization and mechani-
optimize additive manufacturing parameters for laser powder bed fusion of In- cal properties for additively manufactured stainless steel 316L parts by selective
conel 718, in: Proceedings of the 9th International Symposium on Superalloy 718 electron beam melting, Mater. Des. 147 (2018) 157–166.
& Derivatives: Energy, Aerospace, and Industrial Applications, Springer, 2018, [577] M. Elsayed, M. Ghazy, Y. Youssef, K. Essa, Optimization of SLM process parameters
pp. 595–610. for Ti6Al4V medical implants, Rapid Prototyping J. 25 (3) (2018) 433–447.
[552] D. Svetlizky, B. Zheng, D.M. Steinberg, J.M. Schoenung, E.J. Lavernia, N. Eliaz, [578] N. Diaz Vallejo, C. Lucas, N. Ayers, K. Graydon, H. Hyer, Y. Sohn, Process op-
The influence of laser directed energy deposition (DED) processing parameters for timization and microstructure analysis to understand laser powder bed fusion of
Al5083 studied by central composite design, J. Mater. Res. Technol. 17 (2022) 316L stainless steel, Metals 11 (5) (2021) 832.
3157–3171. [579] M. Zhang, C.-N. Sun, X. Zhang, P.C. Goh, J. Wei, D. Hardacre, H. Li, High cycle
[553] M. Ackermann, C. Haase, Machine learning-based identification of interpretable fatigue life prediction of laser additive manufactured stainless steel: a machine
process-structure linkages in metal additive manufacturing, Addit. Manuf. 71 learning approach, Int. J. Fatigue 128 (2019) 105194.
(2023) 103585. [580] G. Shi, L. Li, Z. Yu, R. Liu, P. Sha, Z. Xu, Y. Guo, R. Xi, J. Liu, R. Xin, L. Chen, X.
[554] Z. Yang, D. Eddy, S. Krishnamurty, I. Grosse, P. Denno, P.W. Witherell, F. Lopez, Wang, Z. Zhang, The interaction effect of process parameters on the phase transfor-
Dynamic metamodeling for predictive analytics in advanced manufacturing, Smart mation behavior and tensile properties in additive manufacturing of Ni-rich NiTi
Sustain. Manuf. Syst. 2 (2018) 18–39. alloy, J. Manuf. Process. 77 (2022) 539–550.
[555] P. Akbari, F. Ogoke, N.-Y. Kao, K. Meidani, C.-Y. Yeh, W. Lee, A.B. Farimani, Melt- [581] J. Torres, J. Cotelo, J. Karl, A.P. Gordon, Mechanical property optimization of FDM
poolnet: melt pool characteristic prediction in metal additive manufacturing using PLA in shear with multiple objectives, JOM 67 (2015) 1183–1193.
machine learning, Addit. Manuf. 55 (2022) 102817. [582] X. Yang, R.A. Barrett, N.M. Harrison, S.B. Leen, A physically-based structure-
property model for additively manufactured Ti-6Al-4V, Mater. Des. 205 (2021)
[556] T. Moges, Z. Yang, K. Jones, S. Feng, P. Witherell, Y. Lu, Hybrid modeling approach
109709.
for melt-pool prediction in laser powder bed fusion additive manufacturing, J.
[583] S. Shrestha, G. Manogharan, Optimization of binder jetting using Taguchi method,
Comput. Inf. Sci. Eng. 21 (5) (2021).
JOM 69 (2017) 491–497.
[557] J.F. Rodrı´guez, J.P. Thomas, J.E. Renaud, Design of fused-deposition ABS compo-
[584] H. Xiao, W. Han, Y. Ming, Z. Ding, Y. Duan, A sensitivity analysis-based parameter
nents for stiffness and strength, J. Mech. Des. 125 (3) (2003) 545–551.
optimization framework for 3D printing of continuous carbon fiber/epoxy compos-
[558] O.S. Es-Said, J. Foyos, R. Noorani, M. Mendelson, R. Marloth, B.A. Pregger, Effect
ites, Materials 12 (23) (2019) 3961.
of layer orientation on mechanical properties of rapid prototyped samples, Mater.
[585] A.D. Tura, H.B. Mamo, Characterization and parametric optimization of additive
Manuf. Process. 15 (1) (2000) 107–122.
manufacturing process for enhancing mechanical properties, Heliyon 8 (7) (2022)
[559] S.K. Panda, S. Padhee, S. Anoop Kumar, S.S. Mahapatra, Optimization of fused e09832.
deposition modelling (FDM) process parameters using bacterial foraging technique, [586] F. Górski, W. Kuczko, R. Wichniarek, Impact strength of ABS parts manufac-
Intell. Inf. Manag. 1 (02) (2009) 89. tured using Fused Deposition Modeling technology, Arch. Mech. Technol. Automat.
[560] A.K. Sood, R.K. Ohdar, S.S. Mahapatra, Parametric appraisal of mechanical prop- 34 (1) (2014) 3–12.
erty of fused deposition modelling processed parts, Mater. Des. 31 (1) (2010) [587] M. Moradi, A. Hasani, Z. Pourmand, J. Lawrence, Direct laser metal deposition
287–295. additive manufacturing of Inconel 718 superalloy: statistical modelling and opti-
[561] A. Fatimatuzahraa, B. Farahaina, W. Yusoff, The effect of employing different raster mization by design of experiments, Opt. Laser Technol. 144 (2021) 107380.
orientations on the mechanical properties and microstructure of fused deposition [588] M. Dada, P. Popoola, N. Mathe, S. Pityana, S. Adeosun, Parametric optimization
modeling parts, in: 2011 IEEE Symposium on Business, Engineering and Industrial of laser deposited high entropy alloys using response surface methodology (RSM),
Applications (ISBEIA), IEEE, 2011, pp. 22–27. Int. J. Adv. Manuf. Technol. 109 (2020) 2719–2732.
[562] B. Tymrak, M. Kreiger, J.M. Pearce, Mechanical properties of components fab- [589] B.B. Ravichander, A. Rahimzadeh, B. Farhang, N. Shayesteh Moghaddam, A.
ricated with open-source 3-D printers under realistic environmental conditions, Amerinatanzi, M. Mehrpouya, A prediction model for additive manufacturing of
Mater. Des. 58 (2014) 242–246. Inconel 718 superalloy, Appl. Sci. 11 (17) (2021) 8010.
[563] T. Letcher, B. Rankouhi, S. Javadpour, Experimental study of mechanical proper- [590] Z. Zhan, H. Li, Machine learning based fatigue life prediction with effects of ad-
ties of additively manufactured ABS plastic as a function of layer parameters, in: ditive manufacturing process parameters for printed SS 316L, Int. J. Fatigue 142
ASME International Mechanical Engineering Congress and Exposition, vol. 57359, (2021) 105941.
American Society of Mechanical Engineers, 2015, V02AT02A018. [591] B.H. Lee, J. Abdullah, Z.A. Khan, Optimization of rapid prototyping parameters
[564] S. Ziemian, M. Okwara, C.W. Ziemian, Tensile and fatigue behavior of layered for production of flexible ABS object, J. Mater. Process. Technol. 169 (1) (2005)
acrylonitrile butadiene styrene, Rapid Prototyping J. (2015). 54–61.
[565] X. Liu, M. Zhang, S. Li, L. Si, J. Peng, Y. Hu, Mechanical property parametric [592] J. Laeng, Z.A. Khan, S. Khu, Optimizing flexible behaviour of bow prototype using
appraisal of fused deposition modeling parts based on the gray Taguchi method, Taguchi approach, J. Appl. Sci. 6 (3) (2006) 622–630.
Int. J. Adv. Manuf. Technol. 89 (2017) 2387–2397. [593] A. Arivazhagan, S. Masood, Dynamic mechanical properties of ABS material
[566] X. Deng, Z. Zeng, B. Peng, S. Yan, W. Ke, Mechanical properties optimization of processed by fused deposition modelling, Int. J. Eng. Res. Appl. 2 (3) (2012)
poly-ether-ether-ketone via fused deposition modeling, Materials 11 (2) (2018) 2009–2014.
216. [594] H. Jami, S.H. Masood, W. Song, Dynamic Response of FDM Made ABS Parts in
[567] J. Fernandes, A.M. Deus, L. Reis, M.F. Vaz, M. Leite, Study of the influence of Different Part Orientations, Advanced Materials Research, vol. 748, Trans Tech
3D printing parameters on the mechanical properties of PLA, in: Proceedings of Publ, 2013, pp. 291–294.
the 3rd International Conference on Progress in Additive Manufacturing (Pro-AM [595] O.A. Mohamed, S.H. Masood, J.L. Bhowmik, M. Nikzad, J. Azadmanjiri, Effect of
2018), Singapore, 2018, pp. 14–17. process parameters on dynamic mechanical performance of FDM PC/ABS printed
parts through design of experiment, J. Mater. Eng. Perform. 25 (2016) 2922–2935.
[568] G. Dong, G. Wijaya, Y. Tang, Y.F. Zhao, Optimizing process parameters of fused
[596] M. Srivastava, S. Rathee, Optimisation of FDM process parameters by Taguchi
deposition modeling by Taguchi method for the fabrication of lattice structures,
method for imparting customised properties to components, Virtual Phys. Proto-
Addit. Manuf. 19 (2018) 62–72.
typ. 13 (3) (2018) 203–210.
[569] K. Swarna Lakshmi, G. Arumaikkannu, Influence of process parameters on tensile
[597] Y. Zhang, K. Chou, A parametric study of part distortions in fused deposition mod-
strength of additive manufactured polymer parts using Taguchi method, Adv. 3D
elling using three-dimensional finite element analysis, Proc. Inst. Mech. Eng., B J.
Print. Addit. Manuf. Technol. (2017) 1–7.
Eng. Manuf. 222 (8) (2008) 959–968.
[570] N. Tagscherer, A.M. Bär, S. Zaremba, K. Drechsler, Mechanical analysis of param-
[598] A. Elkholy, R. Kempers, Investigation into the Influence of Fused Deposition Mod-
eter variations in large-scale extrusion additive manufacturing of thermoplastic
eling (FDM) Process Parameters on the Thermal Properties of 3D-Printed Parts,
composites, J. Manuf. Mater. Process. 6 (2) (2022) 36.
2018.
[571] K. Rashed, A. Kafi, R. Simons, S. Bateman, Fused filament fabrication of nylon [599] L. Baich, G. Manogharan, H. Marie, Study of infill print design on production cost-
6/66 copolymer: parametric study comparing full factorial and Taguchi design of time of 3D printed ABS parts, Int. J. Rapid Manuf. 5 (3–4) (2015) 308–319.
experiments, Rapid Prototyping J. (2022). [600] Y.Y. Aw, C.K. Yeoh, M.A. Idris, P.L. Teh, K.A. Hamzah, S.A. Sazali, Effect of printing
[572] M. Spoerk, F. Arbeiter, H. Cajner, J. Sapkota, C. Holzer, Parametric optimization parameters on tensile, dynamic mechanical, and thermoelectric properties of FDM
of intra-and inter-layer strengths in parts produced by extrusion-based additive 3D printed CABS/ZnO composites, Materials 11 (4) (2018) 466.
manufacturing of poly (lactic acid), J. Appl. Polym. Sci. 134 (41) (2017) 45401. [601] K. Chin Ang, K. Fai Leong, C. Kai Chua, M. Chandrasekaran, Investigation of the
[573] N.A. Fountas, J.D. Kechagias, N.M. Vaxevanidis, Optimization of selective laser mechanical properties and porosity relationships in fused deposition modelling-
sintering/melting operations by using a virus-evolutionary genetic algorithm, Ma- fabricated porous structures, Rapid Prototyping J. 12 (2) (2006) 100–105.
chines 11 (1) (2023) 95. [602] J.T. Cantrell, S. Rohde, D. Damiani, R. Gurnani, L. DiSandro, J. Anton, A. Young, A.
[574] A. Du Plessis, I. Yadroitsava, I. Yadroitsev, Effects of defects on mechanical prop- Jerez, D. Steinbach, C. Kroese, P.G. Ifju, Experimental characterization of the me-
erties in metal additive manufacturing: a review focusing on X-ray tomography chanical properties of 3D-printed ABS and polycarbonate parts, Rapid Prototyping
insights, Mater. Des. 187 (2020) 108385. J. 23 (4) (2017) 811–824.
[575] T.A. Rodrigues, V. Duarte, J.A. Avila, T.G. Santos, R. Miranda, J. Oliveira, Wire and [603] G. Gao, F. Xu, J. Xu, Parametric optimization of FDM process for improving me-
arc additive manufacturing of HSLA steel: effect of thermal cycles on microstruc- chanical strengths using Taguchi method and response surface method: a compar-
ture and mechanical properties, Addit. Manuf. 27 (2019) 440–450. ative investigation, Machines 10 (9) (2022) 750.

47
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

[604] O.A. Mohamed, S.H. Masood, J.L. Bhowmik, Modeling, analysis, and optimization [630] K. Li, R. Ma, Y. Qin, N. Gong, J. Wu, P. Wen, S. Tan, D.Z. Zhang, L.E. Murr, J. Luo,
of dimensional accuracy of FDM-fabricated parts using definitive screening design A review of the multi-dimensional application of machine learning to improve
and deep learning feedforward artificial neural network, Adv. Manuf. 9 (2021) the integrated intelligence of laser powder bed fusion, J. Mater. Process. Technol.
115–129. (2023) 118032.
[605] D. Gu, Y.-C. Hagedorn, W. Meiners, G. Meng, R.J.S. Batista, K. Wissenbach, R. [631] Q. Liu, H. Wu, M.J. Paul, P. He, Z. Peng, B. Gludovatz, J.J. Kruzic, C.H. Wang,
Poprawe, Densification behavior, microstructure evolution, and wear performance X. Li, Machine-learning assisted laser powder bed fusion process optimization for
of selective laser melting processed commercially pure titanium, Acta Mater. 60 (9) AlSi10Mg: new microstructure description indices and fracture mechanisms, Acta
(2012) 3849–3860. Mater. 201 (2020) 316–328.
[606] T.-W. Na, W.R. Kim, S.-M. Yang, O. Kwon, J.M. Park, G.-H. Kim, K.-H. Jung, C.-W. [632] Z. Gu, S. Sharma, D.A. Riley, M.V. Pantawane, S.S. Joshi, S. Fu, N.B. Dahotre,
Lee, H.-K. Park, H.G. Kim, Effect of laser power on oxygen and nitrogen concentra- A universal predictor-based machine learning model for optimal process maps in
tion of commercially pure titanium manufactured by selective laser melting, Mater. laser powder bed fusion process, J. Intell. Manuf. (2022) 1–23.
Charact. 143 (2018) 110–117. [633] W. Muhammad, J. Kang, O. Ibragimova, K. Inal, Experimental investigation and
[607] B. Zhang, H. Liao, C. Coddet, Microstructure evolution and density behavior of CP development of a deep learning framework to predict process-induced surface
Ti parts elaborated by self-developed vacuum selective laser melting system, Appl. roughness in additively manufactured aluminum alloys, Weld. World 67 (4) (2023)
Surf. Sci. 279 (2013) 310–316. 897–921.
[608] H. Attar, S. Ehtemam-Haghighi, D. Kent, X. Wu, M.S. Dargusch, Comparative study [634] S. Lapointe, G. Guss, Z. Reese, M. Strantza, M. Matthews, C. Druzgalski,
of commercially pure titanium produced by laser engineered net shaping, selective Photodiode-based machine learning for optimization of laser powder bed fusion
laser melting and casting processes, Mater. Sci. Eng. A 705 (2017) 385–393. parameters in complex geometries, Addit. Manuf. 53 (2022) 102687.
[609] C. Qiu, N.J. Adkins, M.M. Attallah, Microstructure and tensile properties of selec- [635] C. Guo, S. He, H. Yue, Q. Li, G. Hao, Prediction modelling and process optimization
tively laser-melted and of HIPed laser-melted Ti–6Al–4V, Mater. Sci. Eng. A 578 for forming multi-layer cladding structures with laser directed energy deposition,
(2013) 230–239. Opt. Laser Technol. 134 (2021) 106607.
[610] T.C. Dzogbewu, Laser powder bed fusion of Ti6Al4V-xCu: process parameters, J. [636] B. Manjunath, A. Vinod, K. Abhinav, S. Verma, M.R. Sankar, Optimisation of
Metals Mater. Miner. 31 (2) (2021) 62–70. process parameters for deposition of colmonoy using directed energy deposition
[611] B. AlMangour, D. Grzesiak, J.-M. Yang, In-situ formation of novel TiC-particle- process, Mater. Today Proc. 26 (2020) 1108–1112.
reinforced 316L stainless steel bulk-form composites by selective laser melting, J. [637] P.-Y. Lin, F.-C. Shen, K.-T. Wu, S.-J. Hwang, H.-H. Lee, Process optimization for
Alloys Compd. 706 (2017) 409–418. directed energy deposition of SS316L components, Int. J. Adv. Manuf. Technol.
[612] S. Dadbakhsh, M. Speirs, J.-P. Kruth, J. Schrooten, J. Luyten, J. Van Humbeeck, 111 (2020) 1387–1400.
Effect of SLM parameters on transformation temperatures of shape memory nickel [638] C. Félix-Martínez, J. Ibarra-Medina, D.A. Fernández-Benavides, L.A. Cáceres-Díaz,
titanium parts, Adv. Eng. Mater. 16 (9) (2014) 1140–1146. J.M. Alvarado-Orozco, Effect of the parametric optimization and heat-treatment on
[613] S. Dadbakhsh, M. Speirs, J.-P. Kruth, J. Van Humbeeck, Influence of SLM on shape the 18Ni-300 maraging steel microstructural properties manufactured by directed
memory and compression behaviour of NiTi scaffolds, CIRP Ann. 64 (1) (2015) energy deposition, Int. J. Adv. Manuf. Technol. 115 (11–12) (2021) 3999–4020.
209–212. [639] C. Wei, Z. Zhao, H. Ye, Y. Yang, J. Tang, X. Shen, G. Le, Two optimized post-heat
[614] T. Bormann, B. Müller, M. Schinhammer, A. Kessler, P. Thalmann, M. De Wild, treatments to achieve high-performance 90W–7Ni–3Fe alloys fabricated by laser-
Microstructure of selective laser melted nickel–titanium, Mater. Charact. 94 (2014) directed energy deposition, Mater. Sci. Eng. A 833 (2022) 142561.
189–202. [640] C. Li, P. Hodgson, M. Preuss, Y. Chen, X. Wu, Y. Zhu, Y. Tian, A. Huang, Rolling-
[615] W.M. Tucho, V.H. Lysne, H. Austbø, A. Sjolyst-Kverneland, V. Hansen, Investi- assisted direct energy deposited Inconel 718: microstructural evolution and me-
gation of effects of process parameters on microstructure and hardness of SLM chanical properties after optimized heat treatment, J. Mater. Sci. Technol. 144
manufactured SS316L, J. Alloys Compd. 740 (2018) 910–925. (2023) 118–127.
[616] R. Li, J. Liu, Y. Shi, L. Wang, W. Jiang, Balling behavior of stainless steel and [641] Y. Yao, Y. Huang, B. Chen, C. Tan, Y. Su, J. Feng, Influence of processing parame-
nickel powder during selective laser melting process, Int. J. Adv. Manuf. Technol. ters and heat treatment on the mechanical properties of 18Ni300 manufactured by
59 (2012) 1025–1035. laser based directed energy deposition, Opt. Laser Technol. 105 (2018) 171–179.
[617] J.M. Walker, C. Haberland, M. Taheri Andani, H.E. Karaca, D. Dean, M. Elahinia, [642] D. Kats, Z. Wang, Z. Gan, W.K. Liu, G.J. Wagner, Y. Lian, A physics-informed ma-
Process development and characterization of additively manufactured nickel– chine learning method for predicting grain structure characteristics in directed
titanium shape memory parts, J. Intell. Mater. Syst. Struct. 27 (19) (2016) energy deposition, Comput. Mater. Sci. 202 (2022) 110958.
2653–2660. [643] T. Pham, T. Hoang, X. Tran, S. Fetni, L. Duchêne, H.S. Tran, A. Habraken, Char-
[618] K. Antony, T.R. Rakeshnath, Study on selective laser melting of commercially acterization, propagation, and sensitivity analysis of uncertainties in the directed
pure titanium powder, Proc. Inst. Mech. Eng., B J. Eng. Manuf. 233 (7) (2019) energy deposition process using a deep learning-based surrogate model, Probab.
1794–1807. Eng. Mech. 69 (2022) 103297.
[619] C.Y. Yap, H.K. Tan, Z. Du, C.K. Chua, Z. Dong, Selective laser melting of nickel [644] J.-H. Kim, W.-J. Oh, C.-M. Lee, D.-H. Kim, Achieving optimal process design
powder, Rapid Prototyping J. 23 (4) (2017) 750–757. for minimizing porosity in additive manufacturing of Inconel 718 using a deep
[620] M. Letenneur, A. Kreitcberg, V. Brailovski, Optimization of laser powder bed fusion learning-based pore detection approach, Int. J. Adv. Manuf. Technol. 121 (3–4)
processing using a combination of melt pool modeling and design of experiment (2022) 2115–2134.
approaches: density control, J. Manuf. Mater. Process. 3 (1) (2019) 21. [645] W.-J. Oh, C.-M. Lee, D.-H. Kim, Prediction of deposition bead geometry in wire arc
[621] M. Hiren, M. Gajera, K.G. Dave, V.P. Jani, Experimental investigation and analysis additive manufacturing using machine learning, J. Mater. Res. Technol. 20 (2022)
of dimensional accuracy of laser-based powder bed fusion made specimen by appli- 4283–4296.
cation of response surface methodology, Progr. Addit. Manuf. 4 (2019) 371–382. [646] I.Z. Era, M. Grandhi, Z. Liu, Prediction of mechanical behaviors of L-DED fabricated
[622] D. Ivanov, A. Travyanov, P. Petrovskiy, V. Cheverikin, E. Alekseeva, A. Khvan, I. SS 316L parts via machine learning, Int. J. Adv. Manuf. Technol. 121 (3–4) (2022)
Logachev, Evolution of structure and properties of the nickel-based alloy EP718 2445–2459.
after the SLM growth and after different types of heat and mechanical treatment, [647] N.D. Jamnikar, S. Liu, C. Brice, X. Zhang, In-process comprehensive prediction of
Addit. Manuf. 18 (2017) 269–275. bead geometry for laser wire-feed DED system using molten pool sensing data and
[623] B. Vrancken, L. Thijs, J.-P. Kruth, J. Van Humbeeck, Heat treatment of Ti6Al4V multi-modality CNN, Int. J. Adv. Manuf. Technol. 121 (1–2) (2022) 903–917.
produced by Selective Laser Melting: microstructure and mechanical properties, J. [648] T.Q.D. Pham, T.V. Hoang, X. Van Tran, Q.T. Pham, S. Fetni, L. Duchêne, H.S.
Alloys Compd. 541 (2012) 177–185. Tran, A.-M. Habraken, Fast and accurate prediction of temperature evolutions in
[624] J. Wang, Y. Liu, P. Qin, S. Liang, T. Sercombe, L. Zhang, Selective laser melting of additive manufacturing process using deep learning, J. Intell. Manuf. 34 (4) (2023)
Ti–35Nb composite from elemental powder mixture: microstructure, mechanical 1701–1719.
behavior and corrosion behavior, Mater. Sci. Eng. A 760 (2019) 214–224. [649] S. Atre, J. Porter, T. Batchelor, K.K.M. Bulger, P. Gangopadhya, Process parameter
[625] T. Riipinen, S. Metsä-Kortelainen, T. Lindroos, J.S. Keränen, A. Manninen, J. optimization for binder jetting using 420 stainless steel, in: European Congress
Pippuri-Mäkeläinen, Properties of soft magnetic Fe-Co-V alloy produced by laser and Exhibition on Powder Metallurgy. European PM Conference Proceedings, The
powder bed fusion, Rapid Prototyping J. 25 (4) (2019) 699–707. European Powder Metallurgy Association, 2016, pp. 1–6.
[626] F.C. Ewald, F. Brenne, T. Gustmann, M. Vollmer, P. Krooß, T. Niendorf, Laser [650] S.-J. Huang, C.-S. Ye, H.-P. Zhao, Z.-a. Fan, Parameters optimization of binder
powder bed fusion processing of Fe-Mn-Al-Ni shape memory alloy—on the effect jetting process using modified silicate as a binder, Mater. Manuf. Process. 35 (2)
of elevated platform temperatures, Metals 11 (2) (2021) 185. (2020) 214–220.
[627] N.T. Aboulkhair, M. Simonelli, L. Parry, I. Ashcroft, C. Tuck, R. Hague, 3D printing [651] K. Kreft, Z. Lavrič, j. Stanić, P. Perhavec, R. Dreu, Influence of the binder jetting
of Aluminium alloys: additive Manufacturing of Aluminium alloys using selective process parameters and binder liquid composition on the relevant attributes of
laser melting, Prog. Mater. Sci. 106 (2019) 100578. 3D-printed tablets, Pharmaceutics 14 (8) (2022) 1568.
[628] J. Zhang, B. Song, Q. Wei, D. Bourell, Y. Shi, A review of selective laser melting [652] Y. Mao, J. Li, W. Li, D. Cai, Q. Wei, Binder jetting additive manufacturing of 316L
of aluminum alloys: processing, microstructure, property and developing trends, J. stainless-steel green parts with high strength and low binder content: binder prepa-
Mater. Sci. Technol. 35 (2) (2019) 270–284. ration and process optimization, J. Mater. Process. Technol. 291 (2021) 117020.
[629] T.B. Sercombe, X. Li, Selective laser melting of aluminium and aluminium metal [653] N. Lecis, M. Mariani, R. Beltrami, L. Emanuelli, R. Casati, M. Vedani, A. Molinari,
matrix composites, Mater. Technol. 31 (2) (2016) 77–85. Effects of process parameters, debinding and sintering on the microstructure of

48
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

316L stainless steel produced by binder jetting, Mater. Sci. Eng. A 828 (2021) [682] X. Qian, Undercut and overhang angle control in topology optimization: a density
142108. gradient based integral approach, Int. J. Numer. Methods Eng. 111 (3) (2017)
[654] Z.-f. Zhang, L. Wang, L.-t. Zhang, P.-f. Ma, B.-h. Lu, C.-w. Du, Binder jetting 3D 247–272.
printing process optimization for rapid casting of green parts with high tensile [683] D. Obilanade, C. Dordlofva, P. Törlind, Surface roughness considerations in de-
strength, China Foundry 18 (2021) 335–343. sign for additive manufacturing-a literature review, Proc. Des. Soc. 1 (2021)
[655] H. Miyanaji, K.M. Rahman, M. Da, C.B. Williams, Effect of fine powder particles 2841–2850.
on quality of binder jetting parts, Addit. Manuf. 36 (2020) 101587. [684] U.S. Kim, J.W. Park, High-quality surface finishing of industrial three-dimensional
[656] E.M. Jimenez, D. Ding, L. Su, A.R. Joshi, A. Singh, B. Reeja-Jayan, J. Beuth, Para- metal additive manufacturing using electrochemical polishing, Int. J. Precis. Eng.
metric analysis to quantify process input influence on the printed densities of Manuf.-Green Technol. 6 (2019) 11–21.
binder jetted alumina ceramics, Addit. Manuf. 30 (2019) 100864. [685] T. Voisin, R. Shi, Y. Zhu, Z. Qi, M. Wu, S. Sen-Britain, Y. Zhang, S. Qiu, Y. Wang, S.
[657] E. Aznarte Garcia, A.J. Qureshi, C. Ayranci, A study on material-process interac- Thomas, et al., Pitting corrosion in 316L stainless steel fabricated by laser powder
tion and optimization for vat-photopolymerization processes, Rapid Prototyping J. bed fusion additive manufacturing: a review and perspective, JOM 74 (4) (2022)
24 (9) (2018) 1479–1485. 1668–1689.
[658] E.A. Garcia, C. Ayranci, A.J. Qureshi, Material property-manufacturing process [686] K. Tantratian, H. Yan, L. Chen, Predicting pitting corrosion behavior in additive
optimization for form 2 vat-photo polymerization 3D Printers, J. Manuf. Mater. manufacturing: electro-chemo-mechanical phase-field model, Comput. Mater. Sci.
Process. 4 (1) (2020) 12. 213 (2022) 111640.
[687] B. Guo, Y. Zhang, Z. Yang, D. Cui, F. He, J. Li, Z. Wang, X. Lin, J. Wang, Cracking
[659] Y. Wang, Y. Wang, C. Mao, D. Mei, Printing depth modeling, printing process quan-
mechanism of hastelloy x superalloy during directed energy deposition additive
tification and quick-decision of printing parameters in micro-vat polymerization,
manufacturing, Addit. Manuf. 55 (2022) 102792.
Mater. Des. 227 (2023) 111698.
[688] S. Zhou, Y. Su, H. Wang, J. Enz, T. Ebel, M. Yan, Selective laser melting ad-
[660] K. Ransikarbum, R. Pitakaso, N. Kim, A decision-support model for additive man-
ditive manufacturing of 7xxx series Al-Zn-Mg-Cu alloy: cracking elimination by
ufacturing scheduling using an integrative analytic hierarchy process and multi-
co-incorporation of Si and TiB2, Addit. Manuf. 36 (2020) 101458.
objective optimization, Appl. Sci. 10 (15) (2020) 5159.
[689] P. Jiang, M. Rifat, S. Basu, Impact of surface roughness and porosity on lat-
[661] S. Yacoubi, G. Manita, A. Chhabra, O. Korbaa, S. Mirjalili, A multi-objective chaos
tice structures fabricated by additive manufacturing-a computational study, Proc.
game optimization algorithm based on decomposition and random learning mech-
Manuf. 48 (2020) 781–789.
anisms for numerical optimization, Appl. Soft Comput. (2023) 110525.
[690] S. Mojumder, Z. Gan, Y. Li, A. Al Amin, W.K. Liu, Linking process parameters with
[662] J.L.J. Pereira, G.A. Oliver, M.B. Francisco, S.S. Cunha, G.F. Gomes, A review of lack-of-fusion porosity for laser powder bed fusion metal additive manufacturing,
multi-objective optimization: methods and algorithms in mechanical engineering Addit. Manuf. 68 (2023) 103500.
problems, Arch. Comput. Methods Eng. (2021) 1–24. [691] M. Atwya, G. Panoutsos, In-situ porosity prediction in metal powder bed fusion
[663] A. Guo, D. Kong, X. Zhou, P. Qu, S. Wang, J. Li, F. Li, L. Wang, Y. Hu, Evaluation additive manufacturing using spectral emissions: a prior-guided machine learning
of material reuse degree in additive manufacturing by the improved resolution co- approach, J. Intell. Manuf. (2023) 1–24.
efficient grey correlation method, Process Saf. Environ. Prot. 166 (2022) 451–460. [692] Y. Wei, F. Liu, F. Liu, D. Yu, Q. You, C. Huang, Z. Wang, W. Jiang, X. Lin, X. Hu,
[664] M. Tamiz, D.F. Jones, E. El-Darzi, A review of goal programming and its applica- Effect of arc oscillation on porosity and mechanical properties of 2319 aluminum
tions, Ann. Oper. Res. 58 (1995) 39–53. alloy fabricated by cmt-wire arc additive manufacturing, J. Mater. Res. Technol.
[665] B. Pérez-Cañedo, J.L. Verdegay, E.R. Concepción-Morales, A. Rosete, Lexicographic 24 (2023) 3477–3490.
methods for fuzzy linear programming, Mathematics 8 (9) (2020) 1540. [693] Y. Cui, J. Cai, Z. Li, Z. Jiao, L. Hu, J. Hu, Effect of porosity on dynamic response of
[666] T.H.B. Huy, P. Nallagownden, K.H. Truong, R. Kannan, D.N. Vo, N. Ho, Multi- additive manufacturing Ti-6Al-4V alloys, Micromachines 13 (3) (2022) 408.
objective search group algorithm for engineering design problems, Appl. Soft Com- [694] P. Mondal, A. Das, A. Wazeer, A. Karmakar, Biomedical porous scaffold fabrication
put. 126 (2022) 109287. using additive manufacturing technique: porosity, surface roughness and process
[667] K. Deb, A. Pratap, S. Agarwal, T. Meyarivan, A fast and elitist multiobjective ge- parameters optimization, Int. J. Lightweight Mater. Manuf. 5 (3) (2022) 384–396.
netic algorithm: NSGA-II, IEEE Trans. Evol. Comput. 6 (2) (2002) 182–197. [695] N. Béraud, A. Chergui, M. Limousin, F. Villeneuve, F. Vignat, An indicator of
[668] D. Lei, Pareto archive particle swarm optimization for multi-objective fuzzy job porosity through simulation of melt pool volume in aluminum wire arc additive
shop scheduling problems, Int. J. Adv. Manuf. Technol. 37 (2008) 157–165. manufacturing, Mech. Ind. 23 (2022) 1.
[669] Y.-J. Li, H.-N. Li, Interactive evolutionary multi-objective optimization and [696] J. Yang, D. Gu, K. Lin, Y. Zhang, M. Guo, L. Yuan, H. Zhang, H. Zhang, Laser
decision-making on life-cycle seismic design of bridge, Adv. Struct. Eng. 21 (15) additive manufacturing of bio-inspired metallic structures, Chin. J. Mech. Eng.,
(2018) 2227–2240. Addit. Manuf. Front. 1 (1) (2022) 100013.
[670] T. Murata, H. Ishibuchi, Moga: Multi-Objective Genetic Algorithms, IEEE Interna- [697] H. Liu, M. Ye, Z. Ye, L. Wang, G. Wang, X. Shen, P. Xu, C. Wang, High-quality sur-
tional Conference on Evolutionary Computation, vol. 1, IEEE, Piscataway, NJ, USA, face smoothening of laser powder bed fusion additive manufacturing AlSi10Mg via
1995, pp. 289–294. intermittent electrochemical polishing, Surf. Coat. Technol. 443 (2022) 128608.
[671] T.S. Leirmo, K. Martinsen, Evolutionary algorithms in additive manufacturing sys- [698] L. Ladani, J. Razmi, M. Sadeghilaridjani, Fabrication of cu-cnt composite and
tems: discussion of future prospects, Proc. CIRP 81 (2019) 671–676. cu using laser powder bed fusion additive manufacturing, Powders 1 (4) (2022)
[672] J. Zhang, X. Yao, Y. Li, Improved evolutionary algorithm for parallel batch process- 207–220.
ing machine scheduling in additive manufacturing, Int. J. Prod. Res. 58 (8) (2020) [699] E. Maleki, S. Bagherifard, M. Bandini, M. Guagliano, Surface post-treatments
2263–2282. for metal additive manufacturing: progress, challenges, and opportunities, Addit.
Manuf. 37 (2021) 101619.
[673] A. Du Plessis, C. Broeckhoven, I. Yadroitsava, I. Yadroitsev, C.H. Hands, R. Kunju,
[700] A.D. Brandão, R. Gerard, J. Gumpinger, S. Beretta, A. Makaya, L. Pambaguian, T.
D. Bhate, Beautiful and functional: a review of biomimetic design in additive man-
Ghidini, Challenges in additive manufacturing of space parts: powder feedstock
ufacturing, Addit. Manuf. 27 (2019) 408–427.
cross-contamination and its impact on end products, Materials 10 (5) (2017) 522.
[674] Y. Chen, X. Peng, L. Kong, G. Dong, A. Remani, R. Leach, Defect inspection tech-
[701] E. Santecchia, P. Mengucci, A. Gatto, E. Bassoli, S. Defanti, G. Barucca, Cross-
nologies for additive manufacturing, Int. J. Extr. Manuf. 3 (2) (2021) 022002.
contamination quantification in powders for additive manufacturing: a study on
[675] S.K. Everton, M. Hirsch, P. Stravroulakis, R.K. Leach, A.T. Clare, Review of in-situ
Ti-6Al-4V and maraging steel, Materials 12 (15) (2019) 2342.
process monitoring and in-situ metrology for metal additive manufacturing, Mater.
[702] A. du Plessis, S.G. le Roux, Standardized X-ray tomography testing of additively
Des. 95 (2016) 431–445.
manufactured parts: a round robin test, Addit. Manuf. 24 (2018) 125–136.
[676] I.S. Ramírez, F.P.G. Márquez, M. Papaelias, Review on additive manufacturing and [703] S. Xu, J. Liu, Y. Ma, Residual stress constrained self-support topology optimization
non-destructive testing, J. Manuf. Syst. 66 (2023) 260–286. for metal additive manufacturing, Comput. Methods Appl. Mech. Eng. 389 (2022)
[677] P. Sperling, G. Maeurer, N. Achilles, X-ray and Computed Tomography as a tool 114380.
for quality assurance and process optimization in the field of additive manufac- [704] W. Huang, Q. Wang, N. Ma, H. Kitano, Distribution characteristics of residual
turing, in: International Symposium on Digital Industrial Radiology and Computed stresses in typical wall and pipe components built by wire arc additive manufac-
Tomography, 2019. turing, J. Manuf. Process. 82 (2022) 434–447.
[678] Y. Fu, A.R. Downey, L. Yuan, T. Zhang, A. Pratt, Y. Balogun, Machine learning algo- [705] K.S. Ramani, C. He, Y.-L. Tsai, C.E. Okwudire, SmartScan: an intelligent scanning
rithms for defect detection in metal laser-based additive manufacturing: a review, approach for uniform thermal distribution, reduced residual stresses and deforma-
J. Manuf. Process. 75 (2022) 693–710. tions in PBF additive manufacturing, Addit. Manuf. 52 (2022) 102643.
[679] P. Stavropoulos, P. Foteinopoulos, A. Papacharalampopoulos, G. Tsoukantas, [706] Q. Wu, T. Mukherjee, A. De, T. DebRoy, Residual stresses in wire-arc additive
Warping in SLM additive manufacturing processes: estimation through thermo- manufacturing–hierarchy of influential variables, Addit. Manuf. 35 (2020) 101355.
mechanical analysis, Int. J. Adv. Manuf. Technol. 104 (2019) 1571–1580. [707] S. Islam, G.-J. Seo, M.R. Ahsan, H. Villarraga-Gómez, H.-J. Lee, D.B. Kim, In-
[680] D. Xie, F. Lv, Y. Yang, L. Shen, Z. Tian, C. Shuai, B. Chen, J. Zhao, A review on vestigation of microstructures, defects, and mechanical properties of titanium-
distortion and residual stress in additive manufacturing, Chin. J. Mech. Eng., Addit. zirconium-molybdenum alloy manufactured by wire arc additive manufacturing,
Manuf. Front. (2022) 100039. Int. J. Refract. Met. Hard Mater. 110 (2023) 106042.
[681] B. Cheng, K. Chou, Geometric consideration of support structures in part over- [708] M. Caputo, C.V. Solomon, P.-K. Nguyen, A.E. Berkowitz, Electron microscopy in-
hang fabrications by electron beam additive manufacturing, Comput. Aided Des. vestigation of binder saturation and microstructural defects in functional parts
69 (2015) 102–111. made by additive manufacturing, Microsc. Microanal. 22 (S3) (2016) 1770–1771.

49
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

[709] J. Bustillos, J. Kim, A. Moridi, Exploiting lack of fusion defects for microstructural [737] S.M. Estalaki, C.S. Lough, R.G. Landers, E.C. Kinzel, T. Luo, Predicting defects in
engineering in additive manufacturing, Addit. Manuf. 48 (2021) 102399. laser powder bed fusion using in-situ thermal imaging data and machine learning,
[710] J.L. Bartlett, A. Jarama, J. Jones, X. Li, Prediction of microstructural defects in Addit. Manuf. 58 (2022) 103008.
additive manufacturing from powder bed quality using digital image correlation, [738] D.Y. Kononenko, V. Nikonova, M. Seleznev, J. van den Brink, D. Chernyavsky,
Mater. Sci. Eng. A 794 (2020) 140002. An in situ crack detection approach in additive manufacturing based on acoustic
[711] X. Chen, F. Kong, Y. Fu, X. Zhao, R. Li, G. Wang, H. Zhang, A review on wire- emission and machine learning, Addit. Manuf. Lett. 5 (2023) 100130.
arc additive manufacturing: typical defects, detection approaches, and multisensor [739] H. Wang, B. Li, F.-Z. Xuan, Acoustic emission for in situ process monitoring of
data fusion-based model, Int. J. Adv. Manuf. Technol. 117 (2021) 707–727. selective laser melting additive manufacturing based on machine learning and im-
[712] C. Gobert, E.W. Reutzel, J. Petrich, A.R. Nassar, S. Phoha, Application of supervised proved variational modal decomposition, Int. J. Adv. Manuf. Technol. 122 (5–6)
machine learning for defect detection during metallic powder bed fusion additive (2022) 2277–2292.
manufacturing using high resolution imaging, Addit. Manuf. 21 (2018) 517–528. [740] K. Xu, J. Lyu, S. Manoochehri, In situ process monitoring using acoustic emission
[713] M.S. Hossain, H. Taheri, In situ process monitoring for additive manufacturing and laser scanning techniques based on machine learning models, J. Manuf. Pro-
through acoustic techniques, J. Mater. Eng. Perform. 29 (10) (2020) 6249–6262. cess. 84 (2022) 357–374.
[714] M. Borish, B.K. Post, A. Roschli, P.C. Chesser, L.J. Love, Real-time defect correc- [741] W. Ren, J. Mazumder, In-situ porosity recognition for laser additive manufactur-
tion in large-scale polymer additive manufacturing via thermal imaging and laser ing of 7075-Al alloy using plasma emission spectroscopy, Sci. Rep. 10 (1) (2020)
profilometer, Proc. Manuf. 48 (2020) 625–633. 19493.
[715] J. Lewis, A.L. Moore, et al., In situ infrared temperature sensing for real-time defect [742] Y. Tan, F. Lin, M. Ali, Z. Su, H. Wong, Development of a novel beam profiling
detection in additive manufacturing, Addit. Manuf. 47 (2021) 102328. prototype with laser self-mixing via the knife-edge approach, Opt. Lasers Eng. 169
[716] P. Charalampous, I. Kostavelis, D. Tzovaras, Non-destructive quality control meth- (2023) 107696.
ods in additive manufacturing: a survey, Rapid Prototyping J. 26 (4) (2020) [743] J. Petrich, Z. Snow, D. Corbin, E.W. Reutzel, Multi-modal sensor fusion with ma-
777–790. chine learning for data-driven process monitoring for additive manufacturing,
[717] N.A. Surovi, S. Hussain, G.S. Soh, A study of machine learning framework for en- Addit. Manuf. 48 (2021) 102364.
abling early defect detection in wire arc additive manufacturing processes, in: [744] K. Wang, J. Xu, S. Zhang, J. Tan, Economically evaluating energy efficiency perfor-
International Design Engineering Technical Conferences and Computers and In- mance in fused filament fabrication using a multi-scale hierarchical transformer,
formation in Engineering Conference, vol. 86229, American Society of Mechanical Int. J. Adv. Manuf. Technol. (2023) 1–15.
Engineers, 2022, V03AT03A002. [745] M.R. Khosravani, T. Reinicke, On the environmental impacts of 3D printing tech-
[718] S.A. Shevchik, C. Kenel, C. Leinenbach, K. Wasmer, Acoustic emission for in situ nology, Appl. Mater. Today 20 (2020) 100689.
quality monitoring in additive manufacturing using spectral convolutional neural [746] W.E. Frazier, Metal additive manufacturing: a review, J. Mater. Eng. Perform. 23
networks, Addit. Manuf. 21 (2018) 598–604. (2014) 1917–1928.
[719] N.A. Surovi, G.S. Soh, Acoustic feature based geometric defect identification in [747] S. Ford, M. Despeisse, Additive manufacturing and sustainability: an exploratory
wire arc additive manufacturing, Virtual Phys. Prototyp. 18 (1) (2023) e2210553. study of the advantages and challenges, J. Clean. Prod. 137 (2016) 1573–1587.
[748] M. Khorram Niaki, F. Nonino, G. Palombi, S.A. Torabi, Economic sustainability of
[720] O.K. Bowoto, B.I. Oladapo, S. Zahedi, F.T. Omigbodun, O.P. Emenuvwe, Analyt-
additive manufacturing: contextual factors driving its performance in rapid proto-
ical modelling of in situ layer-wise defect detection in 3D-printed parts: additive
typing, Int. J. Manuf. Technol. Manag. 30 (2) (2019) 353–365.
manufacturing, Int. J. Adv. Manuf. Technol. 111 (2020) 2311–2321.
[749] S.H. Huang, P. Liu, A. Mokasdar, L. Hou, Additive manufacturing and its societal
[721] J. Lesseur, B. Tranchand, T. Mancier, A. Montauzier, C. Larignon, S. Perusin, On
impact: a literature review, Int. J. Adv. Manuf. Technol. 67 (2013) 1191–1203.
the use of X-ray microtomography to control artificial defect geometries produced
[750] K. Wang, Y. Song, Z. Huang, Y. Sun, J. Xu, S. Zhang, Additive manufacturing energy
by metal additive manufacturing, Nondestruct. Test. Eval. 37 (5) (2022) 611–630.
consumption measurement and prediction in fabricating lattice structure based on
[722] F.H. Kim, A.L. Pintar, S.P. Moylan, E.J. Garboczi, The influence of X-ray computed
recallable multimodal fusion network, Measurement 196 (2022) 111215.
tomography acquisition parameters on image quality and probability of detection
[751] K. Wang, Y. Song, H. Sheng, J. Xu, S. Zhang, J. Qin, Energy efficiency design
of additive manufacturing defects, J. Manuf. Sci. Eng. 141 (11) (2019) 111002.
for eco-friendly additive manufacturing based on multimodal attention fusion, J.
[723] J.L. McNeil, K. Sisco, C. Frederick, M. Massey, K. Carver, F. List, C. Qiu, M. Mader,
Manuf. Process. 79 (2022) 720–730.
S. Sundarraj, S. Babu, In-situ monitoring for defect identification in nickel alloy
[752] A.H. Gutierrez-Osorio, L. Ruiz-Huerta, A. Caballero-Ruiz, H.R. Siller, V. Borja,
complex geometries fabricated by L-PBF additive manufacturing, Metall. Mater.
Energy consumption analysis for additive manufacturing processes, Int. J. Adv.
Trans. A 51 (2020) 6528–6545.
Manuf. Technol. 105 (1) (2019) 1735–1743.
[724] A. Du Plessis, I. Yadroitsev, I. Yadroitsava, S.G. Le Roux, X-ray microcomputed
[753] F.L. Garcia, A.O. Nunes, M.G. Martins, M.C. Belli, Y.M.B. Saavedra, D.A.L. Silva,
tomography in additive manufacturing: a review of the current technology and
V.A.d.S. Moris, Comparative LCA of conventional manufacturing vs. additive man-
applications, 3D Print. Addit. Manuf. 5 (3) (2018) 227–247.
ufacturing: the case of injection moulding for recycled polymers, Int. J. Sustain.
[725] S. Bellens, P. Vandewalle, W. Dewulf, Deep learning based porosity segmentation
Eng. 14 (6) (2021) 1604–1622.
in X-ray CT measurements of polymer additive manufacturing parts, Proc. CIRP 96
[754] L. Yi, B. Ravani, J.C. Aurich, Development and validation of an energy simulation
(2021) 336–341.
for a desktop additive manufacturing system, Addit. Manuf. 32 (2020) 101021.
[726] V.R. Duarte, T.A. Rodrigues, M.A. Machado, J.P. Pragana, P. Pombinha, L. [755] P. Ramesh, S. Vinodh, Analysis of factors influencing energy consumption of ma-
Coutinho, C.M. Silva, R.M. Miranda, C. Goodwin, D.E. Huber, et al., Benchmark- terial extrusion-based additive manufacturing using interpretive structural mod-
ing of nondestructive testing for additive manufacturing, 3D Print. Addit. Manuf. elling, Rapid Prototyping J. 27 (7) (2021) 1363–1377.
8 (4) (2021) 263–270. [756] K. Wang, L. Yu, J. Xu, S. Zhang, J. Qin, Energy consumption intelligent modeling
[727] A. Thompson, I. Maskery, R.K. Leach, X-ray computed tomography for additive and prediction for additive manufacturing via multisource fusion and inter-layer
manufacturing: a review, Meas. Sci. Technol. 27 (7) (2016) 072001. consistency, Comput. Ind. Eng. (2022) 108720.
[728] Z. Guo, P. Ni, Y. Dai, W. Zhang, K. Huang, Measurement of small defect testing ac- [757] G.O. Barrionuevo, P.M. Sequeira-Almeida, S. Ríos, J.A. Ramos-Grez, S.W. Williams,
curacy in additive manufacturing alloy using industrial CT method, J. Phys. Conf. A machine learning approach for the prediction of melting efficiency in wire arc
Ser. 1827 (2021) 012039, IOP Publishing. additive manufacturing, Int. J. Adv. Manuf. Technol. 120 (5) (2022) 3123–3133.
[729] L. Chen, X. Yao, P. Xu, S.K. Moon, G. Bi, Rapid surface defect identification for [758] K. Wang, J. Xu, S. Zhang, J. Tan, Antivibration and energy efficiency design for
additive manufacturing with in-situ point cloud processing and machine learning, large stroke additive manufacturing based on dynamic trajectory adaption, Int. J.
Virtual Phys. Prototyp. 16 (1) (2021) 50–67. Adv. Manuf. Technol. 118 (9) (2022) 3015–3034.
[730] X. Li, X. Jia, Q. Yang, J. Lee, Quality analysis in metal additive manufacturing with [759] T. Li, J. Yeo, Strengthening the sustainability of additive manufacturing through
deep learning, J. Intell. Manuf. 31 (2020) 2003–2017. data-driven approaches and workforce development, Adv. Intell. Syst. 3 (12)
[731] C. Xia, Z. Pan, Y. Li, J. Chen, H. Li, Vision-based melt pool monitoring for wire-arc (2021) 2100069.
additive manufacturing using deep learning method, Int. J. Adv. Manuf. Technol. [760] J. Xu, K. Wang, H. Sheng, M. Gao, S. Zhang, J. Tan, Energy efficiency optimization
120 (1–2) (2022) 551–562. for ecological 3D printing based on adaptive multi-layer customization, J. Clean.
[732] R.M. Yazdi, F. Imani, H. Yang, A hybrid deep learning model of process-build Prod. 245 (2020) 118826.
interactions in additive manufacturing, J. Manuf. Syst. 57 (2020) 460–468. [761] X. Guo, J. Zhou, W. Zhang, Z. Du, C. Liu, Y. Liu, Self-supporting structure design in
[733] R. Li, M. Jin, Z. Pei, D. Wang, Geometrical defect detection on additive manu- additive manufacturing through explicit topology optimization, Comput. Methods
facturing parts with curvature feature and machine learning, Int. J. Adv. Manuf. Appl. Mech. Eng. 323 (2017) 27–63.
Technol. 120 (5–6) (2022) 3719–3729. [762] L. Wang, S. Tao, P. Zhu, W. Chen, Data-driven topology optimization with multi-
[734] B. Zhang, S. Liu, Y.C. Shin, In-process monitoring of porosity during laser additive class microstructures using latent variable gaussian process, J. Mech. Des. 143 (3)
manufacturing process, Addit. Manuf. 28 (2019) 497–505. (2021) 031708.
[735] W.J. Wright, J. Darville, N. Celik, H. Koerner, E. Celik, In-situ optimization of ther- [763] H. Ning, H. Wang, Y. Lin, W. Wang, S. Dhelim, F. Farha, J. Ding, M. Danesh-
moset composite additive manufacturing via deep learning and computer vision, mand, A survey on the metaverse: the state-of-the-art, technologies, applications,
Addit. Manuf. 58 (2022) 102985. and challenges, IEEE Int. Things J. (2023).
[736] Q. Tian, S. Guo, E. Melder, L. Bian, W. Guo, Deep learning-based data fusion [764] M.M. Queiroz, S. Fosso Wamba, S.C.F. Pereira, C.J. Chiappetta Jabbour, The meta-
method for in situ porosity detection in laser-based additive manufacturing, J. verse as a breakthrough for operations and supply chain management: implications
Manuf. Sci. Eng. 143 (4) (2021) 041011. and call for action, Int. J. Oper. Prod. Manag. (2023).

50
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

[765] H. Xu, A. Berres, Y. Shao, C.R. Wang, J.R. New, O.A. Omitaomu, 13 toward a smart [798] Y. Zheng, S. Yang, H. Cheng, An application framework of digital twin and its case
metaverse city, Adv. Scalable Intell. Geospat. Anal., Chall. Appl. (2023) 245. study, J. Ambient Intell. Humaniz. Comput. 10 (2019) 1141–1153.
[766] Y.K. Dwivedi, L. Hughes, Y. Wang, A.A. Alalwan, S.J. Ahn, J. Balakrishnan, S. [799] C. Boje, A. Guerriero, S. Kubicki, Y. Rezgui, Towards a semantic construction digi-
Barta, R. Belk, D. Buhalis, V. Dutot, et al., Metaverse marketing: how the metaverse tal twin: directions for future research, Autom. Constr. 114 (2020) 103179.
will shape the future of consumer research and practice, Psychol. Mark. 40 (4) [800] R. Zhao, Z. Chen, F. Xue, A blockchain 3.0 paradigm for digital twins in construc-
(2023) 750–776. tion project management, Autom. Constr. 145 (2023) 104645.
[767] T. Huynh-The, T.R. Gadekallu, W. Wang, G. Yenduri, P. Ranaweera, Q.-V. Pham, [801] M. Panarotto, O. Isaksson, V. Vial, Cost-efficient digital twins for design space
D.B. da Costa, M. Liyanage, Blockchain for the metaverse: a review, Future Gener. exploration: a modular platform approach, Comput. Ind. 145 (2023) 103813.
Comput. Syst. (2023). [802] J. Argota Sánchez-Vaquerizo, Getting real: the challenge of building and validating
[768] X. Chen, D. Zou, H. Xie, F.L. Wang, Metaverse in education: contributors, cooper- a large-scale digital twin of barcelona’s traffic with empirical data, ISPRS Int. J.
ations, and research themes, IEEE Trans. Learn. Technol. (2023). Geo-Inf. 11 (1) (2021) 24.
[769] F. Tao, B. Xiao, Q. Qi, J. Cheng, P. Ji, Digital twin modeling, J. Manuf. Syst. 64 [803] R. van Dinter, B. Tekinerdogan, C. Catal, Predictive maintenance using digital
(2022) 372–389. twins: a systematic literature review, Inf. Softw. Technol. (2022) 107008.
[770] M. Singh, E. Fuenmayor, E.P. Hinchy, Y. Qiao, N. Murray, D. Devine, Digital twin: [804] M. Atalay, P. Angin, A digital twins approach to smart grid security testing and
origin to future, Appl. Syst. Innov. 4 (2) (2021) 36. standardization, in: 2020 IEEE International Workshop on Metrology for Industry
[771] D. Jones, C. Snider, A. Nassehi, J. Yon, B. Hicks, Characterising the digital twin: a 4.0 & IoT, IEEE, 2020, pp. 435–440.
systematic literature review, CIRP J. Manuf. Sci. Technol. 29 (2020) 36–52. [805] M. Vuković, D. Mazzei, S. Chessa, G. Fantoni, Digital Twins in Industrial IoT: a
[772] S. Haag, R. Anderl, Digital twin–proof of concept, Manuf. Lett. 15 (2018) 64–66. survey of the state of the art and of relevant standards, in: 2021 IEEE International
[773] M. Liu, S. Fang, H. Dong, C. Xu, Review of digital twin about concepts, technolo- Conference on Communications Workshops (ICC Workshops), IEEE, 2021, pp. 1–6.
gies, and industrial applications, J. Manuf. Syst. 58 (2021) 346–361. [806] O.J. Pinon Fischer, J.F. Matlik, W.D. Schindel, M.O. French, M.H. Kabir, J.S. Gan-
[774] S. Boschert, R. Rosen, Digital twin—the simulation aspect, in: Mechatronic Fu- guli, M. Hardwick, S.M. Arnold, A.D. Byar, J.-H. Lewe, et al., Digital twin: reference
tures: Challenges and Solutions for Mechatronic Systems and Their Designers, model, realizations, and recommendations, Insight 25 (1) (2022) 50–55.
2016, pp. 59–74. [807] K. Wang, Y. Wang, Y. Li, X. Fan, S. Xiao, L. Hu, A review of the technology stan-
[775] Q. Qi, F. Tao, T. Hu, N. Anwer, A. Liu, Y. Wei, L. Wang, A. Nee, Enabling technolo- dards for enabling digital twin, Digital Twin 2 (2022) 4.
gies and tools for digital twin, J. Manuf. Syst. 58 (2021) 3–21. [808] W. Yu, P. Patros, B. Young, E. Klinac, T.G. Walmsley, Energy digital twin tech-
[776] B. Schleich, N. Anwer, L. Mathieu, S. Wartzack, Shaping the digital twin for design nology for industrial energy management: classification, challenges and future,
and production engineering, CIRP Ann. 66 (1) (2017) 141–144. Renew. Sustain. Energy Rev. 161 (2022) 112407.
[777] L. Wright, S. Davidson, How to tell the difference between a model and a digital
[809] X. Fang, H. Wang, G. Liu, X. Tian, G. Ding, H. Zhang, Industry application of digital
twin, Adv. Model. Simul. Eng. Sci. 7 (1) (2020) 1–13.
twin: from concept to implementation, Int. J. Adv. Manuf. Technol. 121 (7–8)
[778] M. Attaran, B.G. Celik, Digital Twin: benefits, use cases, challenges, and opportu-
(2022) 4289–4312.
nities, Dec. Anal. J. (2023) 100165.
[810] D.-G.J. Opoku, S. Perera, R. Osei-Kyei, M. Rashidi, Digital twin application in the
[779] X. Liu, D. Jiang, B. Tao, F. Xiang, G. Jiang, Y. Sun, J. Kong, G. Li, A systematic
construction industry: a literature review, J. Build. Eng. 40 (2021) 102726.
review of digital twin about physical entities, virtual models, twin data, and appli-
[811] C. Mandolla, A.M. Petruzzelli, G. Percoco, A. Urbinati, Building a digital twin for
cations, Adv. Eng. Inform. 55 (2023) 101876.
additive manufacturing through the exploitation of blockchain: a case analysis of
[780] S. Liu, J. Bao, P. Zheng, A review of digital twin-driven machining: from digitiza-
the aircraft industry, Comput. Ind. 109 (2019) 134–152.
tion to intellectualization, J. Manuf. Syst. 67 (2023) 361–378.
[812] Y. Cai, Y. Wang, M. Burnett, Using augmented reality to build digital twin for
[781] C. Semeraro, A. Olabi, H. Aljaghoub, A.H. Alami, M. Al Radi, M. Dassisti, M.A.
reconfigurable additive manufacturing system, J. Manuf. Syst. 56 (2020) 598–604.
Abdelkareem, Digital twin application in energy storage: trends and challenges, J.
[813] G. Knapp, T. Mukherjee, J. Zuback, H. Wei, T. Palmer, A. De, T. DebRoy, Build-
Energy Storage 58 (2023) 106347.
ing blocks for a digital twin of additive manufacturing, Acta Mater. 135 (2017)
[782] Y.-H. Kuo, F. Pilati, T. Qu, G.Q. Huang, Digital twin-enabled smart industrial sys-
390–399.
tems: recent developments and future perspectives, Int. J. Comput. Integr. Manuf.
[814] D.R. Gunasegaram, A. Murphy, A. Barnard, T. DebRoy, M. Matthews, L. Ladani, D.
34 (7–8) (2021) 685–689.
Gu, Towards developing multiscale-multiphysics models and their surrogates for
[783] L. Xu, P. de Vrieze, X. Lu, W. Wang, Digital twins approach for sustainable indus-
digital twins of metal additive manufacturing, Addit. Manuf. 46 (2021) 102089.
try, in: International Conference on Advanced Information Systems Engineering,
[815] D. Gamdha, K. Saurabh, B. Ganapathysubramanian, A. Krishnamurthy, Geomet-
Springer, 2022, pp. 126–134.
ric modeling and physics simulation framework for building a digital twin of
[784] K. Classens, W.M. Heemels, T. Oomen, Digital twins in mechatronics: from model-
extrusion-based additive manufacturing, preprint, arXiv:2305.07120, 2023.
based control to predictive maintenance, in: 2021 IEEE 1st International Confer-
[816] L. Fang, Q. Liu, D. Zhang, A digital twin-oriented lightweight approach for 3D
ence on Digital Twins and Parallel Intelligence (DTPI), IEEE, 2021, pp. 336–339.
[785] M. Braun, J. Krutzinna, Digital twins and the ethics of health decision-making assemblies, Machines 9 (10) (2021) 231.
concerning children, Patterns 3 (4) (2022). [817] X. Lai, X. He, S. Wang, X. Wang, W. Sun, X. Song, Building a lightweight digital
[786] S. West, O. Stoll, J. Meierhofer, S. Züst, Digital twin providing new opportunities twin of a crane boom for structural safety monitoring based on a multifidelity
for value co-creation through supporting decision-making, Appl. Sci. 11 (9) (2021) surrogate model, J. Mech. Des. 144 (6) (2022) 064502.
3750. [818] W. Sun, S. Lian, H. Zhang, Y. Zhang, Lightweight digital twin and federated learn-
[787] A. Suvorova, Towards digital twins for the development of territories, in: Digi- ing with distributed incentive in air-ground 6g networks, IEEE Trans. Netw. Sci.
tal Transformation in Industry: Digital Twins and New Business Models, Springer, Eng. 10 (3) (2022) 1214–1227.
2022, pp. 121–131. [819] X. Zhang, B. Hu, G. Xiong, X. Liu, X. Dong, D. Li, Research and practice of
[788] L. Scime, A. Singh, V. Paquit, A scalable digital platform for the use of digital twins lightweight digital twin speeding up the implementation of flexible manufacturing
in additive manufacturing, Manuf. Lett. 31 (2022) 28–32. systems, in: 2021 IEEE 1st International Conference on Digital Twins and Parallel
[789] C. Liu, L. Le Roux, C. Körner, O. Tabaste, F. Lacan, S. Bigot, Digital twin- Intelligence (DTPI), IEEE, 2021, pp. 456–460.
enabled collaborative data management for metal additive manufacturing systems, [820] Z. Wang, Digital twin technology, in: Industry 4.0-Impact on Intelligent Logistics
J. Manuf. Syst. 62 (2022) 857–874. and Manufacturing, IntechOpen, 2020.
[790] P. Nath, S. Mahadevan, Probabilistic digital twin for additive manufacturing pro- [821] G. Piao, J.G. Breslin, Factorization machines leveraging lightweight linked open
cess design and control, J. Mech. Des. 144 (9) (2022) 091704. data-enabled features for top-n recommendations, in: Web Information Systems
[791] M. Pantelidakis, K. Mykoniatis, J. Liu, G. Harris, A digital twin ecosystem for ad- Engineering–WISE 2017: 18th International Conference, Puschino, Russia, October
ditive manufacturing using a real-time development platform, Int. J. Adv. Manuf. 7-11, 2017, Proceedings, Part II 18, Springer, pp. 420–434, 2017.
Technol. 120 (9–10) (2022) 6547–6563. [822] L. Tang, S. Ying, L. Li, F. Biljecki, H. Zhu, Y. Zhu, F. Yang, F. Su, An application-
[792] F. Corradini, M. Silvestri, A digital twin-based self-calibration tool for fault predic- driven LOD modeling paradigm for 3D building models, ISPRS J. Photogramm.
tion of FDM additive manufacturing systems, Ann. DAAAM Proc. 10 (2) (2021). Remote Sens. 161 (2020) 194–207.
[793] A. Gaikwad, R. Yavari, M. Montazeri, K. Cole, L. Bian, P. Rao, Toward the digi- [823] J. Huo, J. Liu, G. Pei, T. Wang, Research on LOD lightweight method of railway
tal twin of additive manufacturing: integrating thermal simulations, sensing, and four electric BIM model, in: Proceedings of the 2022 6th International Confer-
analytics to detect process faults, IISE Trans. 52 (11) (2020) 1204–1217. ence on Electronic Information Technology and Computer Engineering, 2022,
[794] R.S. Kenett, J. Bortman, The digital twin in Industry 4.0: a wide-angle perspective, pp. 492–497.
Qual. Reliab. Eng. Int. 38 (3) (2022) 1357–1366. [824] J. Wang, X. Xia, Z. Zhang, Y. Zhang, Y. Shen, B. Ren, Class continuous lod algorithm
[795] F. Pires, A. Cachada, J. Barbosa, A.P. Moreira, P. Leitão, Digital twin in industry for lightweight webgl rendering optimization, in: 2022 International Conference on
4.0: technologies, applications and challenges, in: 2019 IEEE 17th International Networking and Network Applications (NaNA), IEEE, 2022, pp. 489–494.
Conference on Industrial Informatics (INDIN), vol. 1, IEEE, 2019, pp. 721–726. [825] V. Kamra, P. Kudeshia, S. ArabiNaree, D. Chen, Y. Akiyama, J. Peethambaran,
[796] M. Jacoby, T. Usländer, Digital twin and internet of things—current standards Lightweight reconstruction of urban buildings: data structures, algorithms, and fu-
landscape, Appl. Sci. 10 (18) (2020) 6519. ture directions, IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 16 (2022) 902–917.
[797] Z. Zhang, F. Wen, Z. Sun, X. Guo, T. He, C. Lee, Artificial intelligence-enabled sens- [826] C. Zhang, B. He, R. Guo, D. Ma, When a tree model meets texture baking: an
ing technologies in the 5G/internet of things era: from virtual reality/augmented approach for quality-preserving lightweight visualization in virtual 3D scene con-
reality to the digital twin, Adv. Intell. Syst. 4 (7) (2022) 2100228. struction, Int. J. Digit. Earth 16 (1) (2023) 645–670.

51
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

[827] A. Jagannath, J. Jagannath, Embedding-assisted attentional deep learning for real- [855] H.H. Hosamo, M.H. Hosamo, Digital twin technology for bridge maintenance using
world rf fingerprinting of bluetooth, IEEE Trans. Cogn. Commun. Netw. (2023). 3D laser scanning: a review, Adv. Civ. Eng. (2022) 2022.
[828] Y. Feng, J. Chen, Z. Huang, H. Wan, R. Xia, B. Wu, L. Sun, M. Xing, A lightweight [856] H. Wang, L. Lv, X. Li, H. Li, J. Leng, Y. Zhang, V. Thomson, G. Liu, X. Wen, C. Sun,
position-enhanced anchor-free algorithm for sar ship detection, Remote Sens. et al., A safety management approach for industry 5.0’s human-centered manufac-
14 (8) (2022) 1908. turing based on digital twin, J. Manuf. Syst. 66 (2023) 1–12.
[829] X. Zhang, H. Wang, C. Xu, Y. Lv, C. Fu, H. Xiao, Y. He, A lightweight fea- [857] L. Arnemann, S. Winter, N. Quernheim, B. Schleich, Product life cycle management
ture optimizing network for ship detection in sar image, IEEE Access 7 (2019) with digital twins for product generation development, Proc. Des. Soc. 3 (2023)
141662–141678. 2955–2964.
[830] W. Bao, X. Yang, D. Liang, G. Hu, X. Yang, Lightweight convolutional neural net- [858] D. Kibira, G. Shao, R. Venketesh, Building a digital twin of an automated robot
work model for field wheat ear disease identification, Comput. Electron. Agric. 189 workcell, in: 2023 Annual Modeling and Simulation Conference (ANNSIM), IEEE,
(2021) 106367. 2023, pp. 196–207.
[831] T. Oberbichler, K. Bletzinger, Cad-integrated form-finding of structural membranes [859] F. Hu, X. Qiu, G. Jing, J. Tang, Y. Zhu, Digital twin-based decision making
using extended catmull–clark subdivision surfaces, Comput. Aided Des. 151 (2022) paradigm of raise boring method, J. Intell. Manuf. 34 (5) (2023) 2387–2405.
103360. [860] M. Rico, M.L. Taverna, M.R. Galli, M.L. Caliusco, Context-aware representation of
[832] S. Rosso, A. Curtarello, F. Basana, L. Grigolato, R. Meneghello, G. Concheri, G. digital twins’ data: the ontology network role, Comput. Ind. 146 (2023) 103856.
Savio, Modeling symmetric minimal surfaces by mesh subdivision, in: Advances [861] J. Vyskočil, P. Douda, P. Novák, B. Wally, A digital twin-based distributed manu-
on Mechanics, Design Engineering and Manufacturing III: Proceedings of the Inter- facturing execution system for Industry 4.0 with AI-powered on-the-fly replanning
national Joint Conference on Mechanics, Design Engineering & Advanced Manu- capabilities, Sustainability 15 (7) (2023) 6251.
facturing, JCM 2020, June 2-4, 2020, Springer, 2021, pp. 249–254. [862] G. Ke, Z. Dashun, M. Liqin, W. Jing, D. Juan, L. Shubo, Q. Huan, S. Lizhong, Intel-
[833] H. Shi, W. Ma, Z. Xu, P. Lin, A novel integrated strategy of easy pruning, parameter ligent machine plus production line digital twin model construction technology, J.
searching, and re-parameterization for lightweight intelligent lithology identifica- Phys. Conf. Ser. 2478 (2023) 102011, IOP Publishing.
tion, Expert Syst. Appl. (2023) 120657. [863] G.N. Schroeder, C. Steinmetz, R.N. Rodrigues, R.V.B. Henriques, A. Rettberg, C.E.
[834] L. Duan, N.-c. Xiao, Z. Hu, G. Li, A. Cheng, An efficient lightweight design strategy Pereira, A methodology for digital twin modeling and deployment for industry 4.0,
for body-in-white based on implicit parameterization technique, Struct. Multidis- Proc. IEEE 109 (4) (2020) 556–567.
cip. Optim. 55 (2017) 1927–1943. [864] Q. Charrier, N. Hakam, K. Benfriha, V. Meyrueis, C. Liotard, A. Aoussat, A roadmap
[835] J. Xu, Y. Zhao, F. Xu, Rdpnet: a single-path lightweight cnn with re- for monitoring cyber-physical systems through the digital twin using an IoT sensing
parameterization for cpu-type edge devices, J. Cloud Comput. 11 (1) (2022) 54. architecture, Available at SSRN 4406361.
[836] A.M. Madni, C.C. Madni, S.D. Lucero, Leveraging digital twin technology in model- [865] A. Haghshenas, A. Hasan, O. Osen, E.T. Mikalsen, Predictive digital twin for off-
based systems engineering, Systems 7 (1) (2019) 7. shore wind farms, Energy Inf. 6 (1) (2023) 1–26.
[837] P. Gope, P.K. Sharma, B. Sikdar, An ultra-lightweight data-aggregation scheme
[866] S. Cavalieri, S. Gambadoro, Proposal of mapping digital twins definition language
with deep learning security for smart grid, IEEE Wirel. Commun. 29 (2) (2022)
to open platform communications unified architecture, Sensors 23 (4) (2023) 2349.
30–36.
[867] A.I. Khalyasmaa, A.I. Stepanova, S.A. Eroshenko, P.V. Matrenin, Review of the dig-
[838] J. Qian, Z. Cao, X. Dong, J. Shen, Z. Liu, Y. Ye, Two secure and efficient lightweight
ital twin technology applications for electrical equipment lifecycle management,
data aggregation schemes for smart grid, IEEE Trans. Smart Grid 12 (3) (2020)
Mathematics 11 (6) (2023) 1315.
2625–2637.
[868] J.V.A. Cabral, E.A.R. Gasca, A.J. Alvares, Digital twin implementation for machin-
[839] B.O. Soufiene, A.A. Bahattab, A. Trad, H. Youssef, LSDA: lightweight secure data
ing center based on ISO 23247 standard, IEEE Latin Am. Trans. 21 (5) (2023)
aggregation scheme in healthcare using IoT, in: Proceedings of the 10th Interna-
628–635.
tional Conference on Information Systems and Technologies, 2020, pp. 1–4.
[869] Z. Huang, M. Fey, C. Liu, E. Beysel, X. Xu, C. Brecher, Hybrid learning-based digital
[840] L. Chen, X. Yao, K. Liu, C. Tan, S.K. Moon, Multisensor fusion-based digital twin in
twin for manufacturing process: modeling framework and implementation, Robot.
additive manufacturing for in-situ quality monitoring and defect correction, Proc.
Comput.-Integr. Manuf. 82 (2023) 102545.
Des. Soc. 3 (2023) 2755–2764.
[870] K. Abutalip, A. Al-Lahham, A. El Saddik, Digital twin of atmospheric environment:
[841] L. Chen, G. Bi, X. Yao, C. Tan, J. Su, N.P.H. Ng, Y. Chew, K. Liu, S.K. Moon, Mul-
sensory data fusion for high-resolution PM 2.5 estimation and action policies rec-
tisensor fusion-based digital twin for localized quality prediction in robotic laser-
ommendation, IEEE Access 11 (2023) 14448–14457.
directed energy deposition, Robot. Comput.-Integr. Manuf. 84 (2023) 102581.
[871] M. La Guardia, M. Koeva, Towards digital twinning on the web: heterogeneous 3D
[842] E.E.-D. Hemdan, W. El-Shafai, A. Sayed, Integrating digital twins with IoT-based
data fusion based on open-source structure, Remote Sens. 15 (3) (2023) 721.
blockchain: concept, architecture, challenges, and future scope, Wirel. Pers. Com-
[872] W. Zhao, C. Zhang, J. Wang, S. Wang, D. Lv, F. Qin, Research on digital twin driven
mun. (2023) 1–24.
rolling bearing model-data fusion life prediction method, IEEE Access (2023).
[843] L. Chalal, A. Saadane, A. Rachid, Unified environment for real-time control of
[873] K. de Koning, J. Broekhuijsen, I. Kühn, O. Ovaskainen, F. Taubert, D. Endresen, D.
hybrid energy system using digital twin and IoT approach, Sensors 23 (12) (2023)
Schigel, V. Grimm, Digital twins: dynamic model-data fusion for ecology, Trends
5646.
[844] N.H. Motlagh, M.A. Zaidan, L. Lovén, P.L. Fung, T. Hänninen, R. Morabito, P. Ecol. Evol. (2023).
Nurmi, S. Tarkoma, Digital twins for smart spaces: beyond IoT analytics, IEEE Int. [874] M. Li, Y. Fu, Q. Chen, T. Qu, Blockchain-enabled digital twin collaboration plat-
Things J. (2023). form for heterogeneous socialized manufacturing resource management, Int. J.
[845] R. Revetria, F. Tonelli, L. Damiani, M. Demartini, F. Bisio, N. Peruzzo, A real-time Prod. Res. 61 (12) (2023) 3963–3983.
mechanical structures monitoring system based on digital twin, IoT and augmented [875] A. Sasikumar, S. Vairavasundaram, K. Kotecha, V. Indragandhi, L. Ravi, G. Sel-
reality, in: 2019 Spring Simulation Conference (SpringSim), IEEE, 2019, pp. 1–10. vachandran, A. Abraham, Blockchain-based trust mechanism for digital twin em-
[846] T. Moi, A. Cibicik, T. Rølvåg, Digital twin based condition monitoring of a knuckle powered industrial internet of things, Future Gener. Comput. Syst. 141 (2023)
boom crane: an experimental study, Eng. Fail. Anal. 112 (2020) 104517. 16–27.
[847] M. Vlaeyen, H. Haitjema, W. Dewulf, Digital twin of an optical measurement sys- [876] S. Liu, Y. Lu, J. Li, X. Shen, X. Sun, J. Bao, A blockchain-based interactive approach
tem, Sensors 21 (19) (2021) 6638. between digital twin-based manufacturing systems, Comput. Ind. Eng. 175 (2023)
[848] M. Braglia, R. Gabbrielli, M. Frosolini, L. Marrazzini, L. Padellini, Using rfid tech- 108827.
nology and discrete-events, agent-based simulation tools to build digital-twins of [877] S. Huang, G. Wang, Y. Yan, X. Fang, Blockchain-based data management for digital
large warehouses, in: 2019 IEEE International Conference on RFID Technology and twin of product, J. Manuf. Syst. 54 (2020) 361–371.
Applications (RFID-TA), IEEE, 2019, pp. 464–469. [878] I. Yaqoob, K. Salah, M. Uddin, R. Jayaraman, M. Omar, M. Imran, Blockchain for
[849] M. Cai, S. Wang, X. Shen, Y. Jin, A scheme for anomalous rfid trajectory detection digital twins: recent advances and future research challenges, IEEE Netw. 34 (5)
based on improved clustering algorithm under digital-twin-driven, in: Proceedings (2020) 290–298.
of the 16th EAI International Conference on Mobile and Ubiquitous Systems: Com- [879] B. Teisserenc, S. Sepasgozar, Adoption of blockchain technology through digital
puting, Networking and Services, 2019, pp. 126–134. twins in the construction industry 4.0: a pestels approach, Buildings 11 (12) (2021)
[850] Y. Liu, Z. Wang, K. Han, Z. Shou, P. Tiwari, J.H. Hansen, Sensor fusion of camera 670.
and cloud digital twin information for intelligent vehicles, in: 2020 IEEE Intelligent [880] J. Protner, M. Pipan, H. Zupan, M. Resman, M. Simic, N. Herakovic, Edge comput-
Vehicles Symposium (IV), IEEE, 2020, pp. 182–187. ing and digital twin based smart manufacturing, IFAC-PapersOnLine 54 (1) (2021)
[851] X. Zhou, K. Sun, J. Wang, J. Zhao, C. Feng, Y. Yang, W. Zhou, Computer vision 831–836.
enabled building digital twin using building information model, IEEE Trans. Ind. [881] F.P. Knebel, R. Trevisan, G.S. do Nascimento, M. Abel, J.A. Wickboldt, A study on
Inform. 19 (3) (2022) 2684–2692. cloud and edge computing for the implementation of digital twins in the oil & gas
[852] D.G. Broo, M. Bravo-Haro, J. Schooling, Design and implementation of a smart industries, Comput. Ind. Eng. 182 (2023) 109363.
infrastructure digital twin, Autom. Constr. 136 (2022) 104171. [882] P. Durana, V. Krastev, K. Buckner, Digital twin modeling, multi-sensor fusion tech-
[853] C. Rausch, R. Lu, S. Talebi, C. Haas, Deploying 3D scanning based geometric digital nology, and data mining algorithms in cloud and edge computing-based smart city
twins during fabrication and assembly in offsite manufacturing, Int. J. Construct. environments, Geopolit. History Int. Relat. 14 (1) (2022) 91–106.
Manag. 23 (3) (2023) 565–578. [883] J. Wang, Y. Liu, S. Ren, C. Wang, S. Ma, Edge computing-based real-time schedul-
[854] F. Jiang, L. Ma, T. Broyd, K. Chen, Digital twin and its implementations in the civil ing for digital twin flexible job shop with variable time window, Robot. Comput.-
engineering sector, Autom. Constr. 130 (2021) 103838. Integr. Manuf. 79 (2023) 102435.

52
L. Jin, X. Zhai, K. Wang et al. Materials & Design 244 (2024) 113086

[884] Z. Zhou, Z. Jia, H. Liao, W. Lu, S. Mumtaz, M. Guizani, M. Tariq, Secure and [896] F. Pires, V. Melo, J. Almeida, P. Leitão, Digital twin experiments focusing vir-
latency-aware digital twin assisted resource scheduling for 5G edge computing- tualisation, connectivity and real-time monitoring, in: 2020 IEEE Conference on
empowered distribution grids, IEEE Trans. Ind. Inform. 18 (7) (2021) 4933–4943. Industrial Cyberphysical Systems (ICPS), vol. 1, IEEE, 2020, pp. 309–314.
[885] Y. Xie, K. Lian, Q. Liu, C. Zhang, H. Liu, Digital twin for cutting tool: modeling, [897] Y. Yi, Y. Yan, X. Liu, Z. Ni, J. Feng, J. Liu, Digital twin-based smart assembly
application and service strategy, J. Manuf. Syst. 58 (2021) 305–312. process design and application framework for complex products and its case study,
[886] S.K. Pal, D. Mishra, A. Pal, S. Dutta, D. Chakravarty, S. Pal, S.K. Pal, D. J. Manuf. Syst. 58 (2021) 94–107.
Mishra, A. Pal, S. Dutta, et al., Signal processing for digital twin, in: Digital [898] Q. Zhang, S. Shen, H. Li, W. Cao, W. Tang, J. Jiang, M. Deng, Y. Zhang, B. Gu,
Twin–Fundamental Concepts to Applications in Advanced Manufacturing, 2022, K. Wu, et al., Digital twin-driven intelligent production line for automotive mems
pp. 117–187. pressure sensors, Adv. Eng. Inform. 54 (2022) 101779.
[887] P. Aivaliotis, K. Georgoulias, G. Chryssolouris, The use of digital twin for predic- [899] J. Mertes, M. Glatt, C. Schellenberger, M. Klar, H.D. Schotten, J.C. Aurich, Devel-
tive maintenance in manufacturing, Int. J. Comput. Integr. Manuf. 32 (11) (2019) opment of a 5G-enabled digital twin of a machine tool, Proc. CIRP 107 (2022)
1067–1080. 173–178.
[888] Y. He, J. Guo, X. Zheng, From surveillance to digital twin: challenges and recent [900] Y. Tai, L. Zhang, Q. Li, C. Zhu, V. Chang, J.J. Rodrigues, M. Guizani, Digital-twin-
advances of signal processing for industrial internet of things, IEEE Signal Process. enabled IoMT system for surgical simulation using rAC-GAN, IEEE Int. Things J.
Mag. 35 (5) (2018) 120–129. 9 (21) (2022) 20918–20931.
[889] S.H. Khajavi, N.H. Motlagh, A. Jaribion, L.C. Werner, J. Holmström, Digital twin: [901] Z. Lv, D. Chen, H. Feng, R. Lou, H. Wang, Beyond 5G for digital twins of UAVs,
vision, benefits, boundaries, and creation for buildings, IEEE Access 7 (2019) Comput. Netw. 197 (2021) 108366.
147406–147419. [902] K. Bartsch, A. Pettke, A. Hübert, J. Lakämper, F. Lange, On the digital twin appli-
[890] D. Burnett, J. Thorp, D. Richards, K. Gorkovenko, D. Murray-Rust, Digital twins cation and the role of artificial intelligence in additive manufacturing: a systematic
as a resource for design research, in: Proceedings of the 8th ACM International review, J. Phys. Mater. 4 (3) (2021) 032005.
Symposium on Pervasive Displays, 2019, pp. 1–2. [903] D.R. Gunasegaram, A.B. Murphy, M. Matthews, T. DebRoy, The case for digital
[891] R. Martinez-Velazquez, R. Gamez, A. El Saddik, Cardio twin: a digital twin of the twins in metal additive manufacturing, J. Phys. Mater. 4 (4) (2021) 040401.
human heart running on the edge, in: 2019 IEEE International Symposium on Med- [904] J. Haw, S.L. Sing, Z.H. Liu, Digital twins in design for additive manufacturing,
ical Measurements and Applications (MeMeA), IEEE, 2019, pp. 1–6. Mater. Today Proc. 70 (2022) 352–357.
[892] Z. Lv, D. Chen, H. Feng, W. Wei, H. Lv, Artificial intelligence in underwater digital [905] N. Jyeniskhan, A. Keutayeva, G. Kazbek, M.H. Ali, E. Shehab, Integrating machine
twins sensor networks, ACM Trans. Sens. Netw. 18 (3) (2022) 1–27. learning model and digital twin system for additive manufacturing, IEEE Access
[893] P. Angin, M.H. Anisi, F. Göksel, C. Gürsoy, A. Büyükgülcü, Agrilora: a digital twin (2023).
framework for smart agriculture, J. Wirel. Mob. Networks Ubiquitous Comput. [906] D.B. Kim, G. Shao, G. Jo, A digital twin implementation architecture for wire+ arc
Dependable Appl. 11 (4) (2020) 77–96. additive manufacturing based on ISO 23247, Manuf. Lett. 34 (2022) 1–5.
[894] W. Zhou, Research on wireless sensor network access control and load balancing [907] D. Guerra-Zubiaga, V. Kuts, K. Mahmood, A. Bondar, N. Nasajpour-Esfahani, T.
in the industrial digital twin scenario, J. Sens. 2022 (2022) 1–12. Otto, An approach to develop a digital twin for industry 4.0 systems: manufactur-
[895] R.-S. Balica, et al., Machine and deep learning technologies, wireless sensor net- ing automation case studies, Int. J. Comput. Integr. Manuf. 34 (9) (2021) 933–949.
works, and virtual simulation algorithms in digital twin cities, Geopolit. History
Int. Relat. 14 (1) (2022) 59–74.

53