Received 8 August 2024, accepted 8 September 2024, date of publication 11 September 2024, date of current version 24 September 2024.

Digital Object Identifier 10.1109/ACCESS.2024.3458830

Advancing Manufacturing Through Artificial

Intelligence: Current Landscape, Perspectives,
Best Practices, Challenges, and Future Direction
RAJNISH RAKHOLIA , ANDRÉS L. SUÁREZ-CETRULO , MANOKAMNA SINGH ,
AND RICARDO SIMÓN CARBAJO
Ireland’s Centre for Artificial Intelligence (CeADAR), School of Computer Science, University College Dublin (UCD), Dublin 4, D04 V1W8 Ireland
Corresponding author: Rajnish Rakholia ([email protected])
This work was supported in part by the PeRCEPTION Project, which focuses on the predictive development of functional foods using data
analytics and machine learning, was funded by Enterprise Ireland under the Disruptive Technologies Innovation Fund (DTIF), Grant
Agreement number DT2020214.

ABSTRACT The industrial sector is currently undergoing a transformative era of intelligent automation
driven by Artificial Intelligence (AI) capabilities. This synergy greatly enhances efficiency and seamlessly
enables data-driven decision-making processes. These advantages enable more efficient resource allocation
and enhance production planning precision. This paper aims to provide state-of-the-art and ongoing develop-
ments in the AI landscape within the manufacturing industry. In addition, the review explores the key areas
where AI is being applied in manufacturing, such as predictive maintenance, quality control, process opti-
mization, supply chain management, robotics and automation, and intelligent decision support systems. The
review also encompasses an exploration of the challenges encountered by the manufacturing sector, alongside
an investigation into the potential of AI to mitigate these challenges. Furthermore, this work thoroughly
reviews recent AI advancements, including explainable AI, human-robot collaboration, edge computing, and
the Internet of Things (IoT) integration. The review concludes by providing recommendations, highlighting
best practices, and providing insights into potential collaborative opportunities.

INDEX TERMS Smart manufacturing, Industry 4.0, artificial intelligence, automation, machine learning,
quality control, Internet of Things, robotics.

I. INTRODUCTION on expenses, and boost their output using the power of

A. BACKGROUND AI [4], [5].
AI is significantly transforming conventional manufacturing AI is facilitating the creation of smart factories, or Industry
practices [1] by initiating a paradigm shift in the entire 4.0 [6], where networked systems, IoT devices, and AI algo-
lifecycle of production processes and management [2]. This rithms combine to provide a highly automated and intelligent
transformation is driven by AI’s ability to process and analyze manufacturing environment [7], [8]. Incorporating AI into
extensive datasets, make intelligent decisions, and adapt in these smart factories paves the way for improved productivity,
real-time, resulting in significant and influential changes. real-time monitoring of manufacturing processes, and data-
AI integration in manufacturing involves assimilat- driven decision-making [9], [10].
ing advanced technologies and algorithms that empower The concept of Industry 4.0 and its impact on production
machines and systems to execute tasks historically under- processes, emphasizes the need for capacity optimization
taken by humans, achieving heightened efficiency, precision, over maximization to enhance operational efficiency and
and automation [3]. Factories may streamline their opera- organizational value [11]. The application of Artificial Intel-
tions, raise the bar on product quality while cutting down ligence (AI) methods for Predictive Maintenance (PdM) in
the steel industry, highlights its importance in Industry 4.0.
The associate editor coordinating the review of this manuscript and The survey focuses on AI techniques like deep learning and
approving it for publication was Anandakumar Haldorai . machine learning used for PdM tasks such as fault detection
2024 The Authors. This work is licensed under a Creative Commons Attribution 4.0 License.
VOLUME 12, 2024 For more information, see https://creativecommons.org/licenses/by/4.0/ 131621
 R. Rakholia et al.: Advancing Manufacturing Through Artificial Intelligence: Current Landscape

and remaining useful life estimation [12]. Also, it emphasizes for planning production [26], managing inventories, and dis-
the challenges of implementing these methods in real-world tributing goods. Streamlining these operations could enhance
production environments and the need for more accessible customer satisfaction, reduce stockouts, and shorten pro-
and reproducible research. It also discusses the significant duction times for manufacturers. However, the use of AI
environmental, economic, and safety benefits of integrating in manufacturing is fraught with numerous challenges [27],
PdM solutions in the steel industry. [28]. AI system implementation requires large expenditures
Industry 5.0, promoted by the European Commission, in infrastructure, data collection, and analytics capabilities.
focuses on a sustainable, human-centric, and resilient indus- Some critical challenges that need to be addressed include
try, while Society 5.0, proposed by Japan, aims to balance integrating AI systems with existing infrastructure, data pri-
economic advancement with social problem resolution [13]. vacy and security, as well as the impact of AI on employment.
The field of predictive maintenance is one of the most Manufacturers must solve these challenges by investing in
notable areas where AI is revolutionizing the industrial sec- the appropriate resources, training employees, and ensuring
tor [14]. The traditional maintenance procedures often adhere ethical and responsible AI implementation.
to a fixed schedule, resulting in unnecessary downtime and
inefficient resource allocation. On the other hand, AI-based
predictive maintenance systems evaluate real-time data from B. OBJECTIVES OF THE REVIEW
sensors and equipment using machine learning algorithms to • To the best of our knowledge, the literature needs stud-
predict impending failures or maintenance requirements [15]. ies on smart manufacturing using artificial intelligence.
AI systems may detect trends and abnormalities and provide In fact, our current research shows that there needs to be
alerts to operators, allowing for proactive maintenance and a comprehensive review of this. With this in mind, our
reducing unscheduled downtime [16]. aim in this paper is:
Another significant application of AI in manufacturing is • To provide a thorough overview of the applications of
quality control. AI algorithms can analyze large volumes of AI in diverse aspects of manufacturing, encompassing
data, such as images, sensor readings, and historical data, production planning, quality control, predictive main-
to identify defects or deviations from the desired specifica- tenance, supply chain management, and other relevant
tions [17]. This enables manufacturers to identify and rectify areas.
quality concerns in real-time, ensuring that only items with • To provide an extensive analysis of AI’s recent advance-
desired standards are released on the market quality control ments, challenges, opportunities, and future directions
solutions driven by AI may reduce errors and waste [18], in manufacturing. This valuable resource caters to
resulting in significant savings and increased customer sat- researchers, practitioners, and decision-makers who aim
isfaction. to utilize AI techniques to improve manufacturing pro-
Furthermore, AI is leading to developments in robotics cesses and achieve better outcomes.
and the automation of production processes [19]. AI-enabled • To investigate the seamless integration of AI techniques
collaborative robots may operate alongside human opera- with existing manufacturing systems and technologies
tors to perform routine tasks with remarkable accuracy and while providing valuable insights into the practical
speed [20]. These robots can adapt to new environments, aspects of deploying AI solutions in manufacturing
gain knowledge from their experiences, and improve over environments.
time. AI-powered automation [21] also makes integrating • To address essential aspects regarding the implementa-
various industrial components and systems easier, enhanc- tion of AI, including regulatory frameworks, workforce
ing workflow efficiency and overall productivity. Intelligent reskilling, data governance, and the importance of col-
robotics has the potential to revolutionize various industries laboration among academia, industry, and government.
by amplifying output, streamlining operations, and enriching • Therefore, this study aims to help readers understand the
customer interactions [22]. current state of the art, bridge the gap among the field’s
The biotechnology industry is rapidly evolving, driven subtopics, and address promising future lines of research
by advancements in manufacturing technology [23]. The in this domain.
integration of artificial intelligence (AI) in bioprocess-
ing, automation, and data integration within biotechnologi-
cal engineering. AI enhances industrial operations through C. SCOPE AND ORGANIZATION OF THE PAPER
predictive maintenance, process optimization, and quality This paper discusses the different AI technologies used in
control, significantly improving supply chain management manufacturing and their benefits to the industry. The paper
[23], [24]. also focuses on the most recent research and the applica-
Moreover, AI’s influence in manufacturing extends beyond tion of AI in manufacturing. It identifies the challenges and
the factory environment, encompassing various dimensions opportunities associated with AI integration in manufactur-
of supply chain management [25]. AI algorithms can analyze ing and discusses the implications of AI for the future of
vast amounts of data related to inventory levels, demand the manufacturing landscape. The intended audience for this
forecasts, and logistical constraints to find the best strategies paper includes researchers, AI developers and engineers,

131622 VOLUME 12, 2024

R. Rakholia et al.: Advancing Manufacturing Through Artificial Intelligence: Current Landscape

technology enthusiasts, policymakers, consultants, and pro- to advancing intelligent manufacturing practices. Artificial
fessionals within the manufacturing industry. Intelligence (AI) and automation have garnered significant
• The study presents several contributions to the field of attention due to their transformative potential across diverse
AI in manufacturing. industries [29]. (Fig. 1) discusses various technologies in the
• It provides a comprehensive overview of the cur- context of AI and manufacturing.
rent state-of-the-art AI applications in manufacturing,
A. TRADITIONAL MACHINE LEARNING
including predictive maintenance, quality control, and
process optimization. Traditional machine learning algorithms are versatile tools
• The study identifies the industry’s critical challenges, for data analysis, pattern recognition, and modeling tasks.
such as integration with existing infrastructure and data Each algorithm has its strengths and limitations, making
privacy concerns, and discusses AI’s potential to address them suitable for specific applications. For instance, Support
these issues. Vector Machines offer robust binary classification capabil-
• It offers insights into recent advancements in AI, like ities [30], while k-means clustering efficiently groups data
explainable AI and human-robot collaboration, and sug- points [31]. Decision trees provide transparency and inter-
gests future directions for research and development. pretability [32]. In contrast, in a random forest, each tree is
• The review highlights best practices and recom- trained on a random subset of the data and a random subset
mends potential collaborative opportunities among of the features. This helps to prevent the trees from overfitting
academia, industry, and government to advance AI in the training data and makes the forest more accurate [33].
manufacturing. Naive Bayes is simple and efficient for text classifica-
• These contributions aim to enhance manufacturing effi- tion [34], and Linear Regression offers quick interpretations
ciency, precision, and automation while addressing the of linear relationships. While these algorithms have their
challenges and setting a path for future innovation. strengths, they also face challenges such as overfitting
and scalability. As the field of machine learning evolves,
researchers continue to explore innovative techniques and
1) HERE ARE SEVERAL POTENTIAL RESEARCH QUESTIONS
hybrid approaches to overcome these limitations and advance
THAT WILL BE EXPLORED
the capabilities of traditional machine learning on diverse
1) What AI technologies are currently being utilized within
datasets.
the field of manufacturing? 2) What are the potential advan-
In the machine learning domain, scikit-learn [35], and
tages of integrating AI into the manufacturing process?
XGBoost [36] libraries offer a comprehensive suite of
3) What are the key challenges associated with the implemen-
user-friendly tools, empowering seamless implementation.
tation of AI in the manufacturing sector? 4) What are AI’s
Concurrently, cloud platforms like Google Cloud, Azure,
prospective impacts on the future manufacturing landscape?
Amazon SageMaker, OpenAI, and IBM Watson have opened
up new horizons for customers, enabling them to harness the
2) THE STRUCTURE OF THE PAPER IS AS FOLLOWS: power of machine learning algorithms through scalable and
First, we introduce the paper’s background, review objectives, flexible cloud solutions. Within this dynamic landscape, the
and scope in Section I. Section II delves into AI techniques AI4EU platform [37] emerged as a collaborative hub, housing
and algorithms in the manufacturing context. Section III advanced algorithms such as IoT data analysis models and
examines data collection and storage. Section IV focuses on the Markov Decision Process, poised to drive transformative
AI applications within manufacturing. Section V presents a advancements in machine learning applications.
perspective on AI adoption in the manufacturing ecosystem. Chen et al. [38] explores the critical role of Machine Learn-
In Section VI, we emphasize recommendations and best prac- ing (ML) in the digitalization of manufacturing operations
tices. Section VII provides a summary of the key findings. within the context of Industry 4.0. It introduces a framework
Section VIII discusses the challenges and future directions, called ‘Four-Know’ (Know-what, Know-why, Know-when,
and lastly, Section IX wraps up with a conclusion. Know-how) and ‘Four-Level’ (Product, Process, Machine,
System) to categorize ML tasks and analyze development
II. AI TECHNIQUES AND ALGORITHMS IN trends. The study also presents an implementation pipeline
MANUFACTURING for ML solutions, detailing steps from data collection to
This section delves into the diverse array of Artificial Intel- model deployment. Additionally, it discusses various ML
ligence (AI) techniques and algorithms revolutionizing the methods (supervised, semi-supervised, unsupervised, and
manufacturing industry. From traditional machine learn- reinforcement learning) and their applications in manufactur-
ing methods to cutting-edge deep learning networks, these ing, highlighting current challenges and future directions for
technologies reshape manufacturers’ operations, enabling ML in this field.
unprecedented efficiency, precision, and automation. Also,
it explores the strengths and applications of various AI B. DEEP LEARNING AND NEURAL NETWORKS
approaches, including computer vision, reinforcement learn- Deep learning provides a diverse range of techniques that
ing, and active learning, and discusses how they contribute leverage the capabilities of neural networks to address

VOLUME 12, 2024 131623

 R. Rakholia et al.: Advancing Manufacturing Through Artificial Intelligence: Current Landscape

complex tasks effectively. Convolutional Neural Networks for developing this model, emphasizing the need for a
excel in analyzing images [39], while Recurrent Neural Net- comprehensive understanding of both technical and organiza-
works are skilled at processing sequential data [40], and tional aspects to improve cost-efficiency, time-efficiency, and
Long Short-Term Memory Networks overcome challenges product quality. The study also addresses the challenges faced
with long-term dependencies. These techniques are driving by small and medium-sized companies in adapting Industry
revolutions across various domains. 4.0 technologies [51].
Generative Adversarial Networks facilitate the genera-
tion of realistic synthetic data [41], while Autoencoders D. REINFORCEMENT LEARNING
and Variational Autoencoders are powerful tools for unsu-
Reinforcement Learning (RL) holds immense potential for
pervised learning and data reconstruction [42]. Initially
revolutionizing manufacturing processes by enabling agents
designed for natural language processing, transform model
to learn and make decisions based on interactions with the
architectures demonstrate broad adaptability [43], and Deep environment and feedback in the form of rewards. From
Reinforcement Learning enables intelligent decision-making foundational algorithms like Q-Learning to advanced deep
in dynamic environments [44]. learning-based methods such as DQNs, DDPG, and PPO,
Moreover, other techniques such as Siamese Networks, RL offers solutions for optimizing complex processes, con-
Zero or Few-Shot Learning, Meta-Learning, and Capsule trolling robotic systems, and enhancing overall efficiency in
Networks [45], [46] are just a few examples of innovative the manufacturing industry [52]. As RL continues to evolve
techniques driving the frontiers of artificial intelligence and and adapt, it presents a promising avenue for furthering
machine learning. automation and optimization, empowering manufacturers to
Deep learning excels at automatically learning hierarchi- achieve higher productivity and superior performance in their
cal features from raw data, reducing the need for extensive operations. RL solutions can be built using various libraries
manual feature engineering. Additionally, its adeptness in such as Gym, Keras-rl, Pyqlearning, and Tensorforce. These
handling large and complex datasets empowers it to achieve libraries provide essential tools and functionalities to develop
state-of-the-art performance in various domains. The exten- RL agents effectively. Additionally, platforms like Sage-
sive adoption of libraries such as TensorFlow, Keras, and maker RL and Deepmind Lab have been specifically designed
PyTorch [47] has expedited the deployment of deep learning to streamline the implementation and utilization of RL algo-
models, exerting a lasting influence across a wide array of rithms, further simplifying the adoption of RL in practical
domains. Moreover, these libraries seamlessly integrate with applications.
platforms such as Google Cloud, Azure, Amazon SageMaker, Figure 1 outlines various AI techniques and algorithms
and IBM Watson [48], enabling the development of cloud- revolutionizing the manufacturing industry. It highlights the
based solutions. use of traditional machine learning algorithms like Support
Vector Machines and k-means clustering, as well as deep
C. COMPUTER VISION AND IMAGE PROCESSING
learning techniques such as Convolutional Neural Networks
and Recurrent Neural Networks. Figure 1 also emphasizes
The manufacturing industry relies on diverse Computer
the importance of computer vision for automation and quality
Vision and Image Processing techniques and algorithms to
control, reinforcement learning for optimizing processes, and
enable automation, quality control, defect detection, and pro-
active learning to enhance model performance with limited
cess optimization [49]. From classic methods like histogram
labeled data. Additionally, it mentions transfer learning for
equalization, edge detection, and Hough Transform to state-
leveraging knowledge from one task to improve another.
of-the-art deep learning approaches such as CNNs, GANs,
These AI techniques collectively contribute to advancing
and YOLO [50], these tools form the backbone of various
intelligent manufacturing practices.
applications. Whether it’s object recognition, defect detec-
tion, motion analysis, or 3D reconstruction, these techniques
are crucial in improving efficiency and accuracy in manu- E. ACTIVE LEARNING
facturing processes. Integrating these methods continues to Active learning is a powerful approach that addresses the
drive innovation and progress in the manufacturing sector as challenges of limited labeled data in the manufacturing indus-
technology advances. try [53]. Active learning optimizes the learning process and
Albers et al. [51] discusses the initial phase of an provides substantial benefits by carefully choosing the most
Industry 4.0 project focused on developing an intelligent informative samples for annotation. Various active learning
quality control system. It highlights the importance of con- techniques [54], including uncertainty sampling, query by
tinuous communication between humans, machines, and committee, density-based sampling, expected model change,
products, enabled by cyber-physical production systems query by bagging, information density sampling, margin
(CPPS). The research analyzes a production process involv- sampling, and diversity sampling, can enhance model perfor-
ing three companies to identify quality-related issues and mance and generalization.
proposes a descriptive model to represent relevant systems The adaptability to real-time changes in the manufac-
and their interdependencies. The study outlines a procedure turing process enables the model to continuously improve

131624 VOLUME 12, 2024

R. Rakholia et al.: Advancing Manufacturing Through Artificial Intelligence: Current Landscape

FIGURE 1. AI techniques and algorithms in manufacturing.

its performance, ensuring its relevance in dynamic produc- deploy an increasing number of sensors and actuators in their
tion environments. Furthermore, by actively selecting diverse manufacturing processes, reliance on cloud services has risen,
samples, active learning addresses data bias issues [55], leading to higher network traffic and potentially affecting
ensuring that the model is trained on a representative dataset network transmission performance.
from various stages of the production process. With the sup- To address these challenges, various libraries such as
port of Python frameworks like modAL and open-source TensorFlow Lite, uTensor, the Embedded Learning Library
platforms like KNIME, implementing active learning mod- (EEL), CMIS NN, ARM Compute Library, and deepC have
els has become more accessible. Projects like CANDELA emerged as useful tools for implementing Edge AI solu-
under H2020 [56] also provide a comprehensive set of tions [58]. Notably, companies like NVIDIA and Advantech
algorithms, including active learning, as part of their data offer tailored Edge AI solutions to enable customers to inte-
mining functions. These advancements underscore the grow- grate and operate Edge AI effectively within their industries.
ing significance of active learning in overcoming data-related By leveraging Edge AI, industries can achieve faster and more
obstacles and driving innovation across various domains. efficient decision-making while ensuring data privacy and
reducing dependency on the network.
F. EDGE AI
Unlike traditional data warehouses that primarily process
data in a centralized manner, there is a growing need to G. EXPLAINABLE AI
develop models and deploy them on devices to make real- The complexity of features learned by various machine
time decisions. This paradigm, known as Edge AI, proves learning techniques often results in models acting as black-
essential in situations where devices may not always have box solutions, making it challenging to interpret their
access to a network or when processing and storing all data decisions intuitively. This has recently become a regula-
on a central server is slow and impractical. Kubiak et al. [57] tory must-have to achieve AI trustworthiness [59]. Con-
explore the potential applications of Edge computing in the sequently, the emergence of techniques such as Local
manufacturing industry, shedding light on its diverse range Interpretable Model-Agnostic Explanations (LIME), Shap-
of use cases and benefits. The key advantages of Edge AI are ley Additive explanations (SHAP), PDP, feature importance
speed and privacy, as computations are performed locally on methods, rule-based models, Layer-wise Relevance Propa-
the devices themselves, eliminating the need for data to be gation (LRP), counterfactual explanations, and anchors in
sent back and forth to a central entity. This not only reduces the domain of Explainable AI (XAI) aims to overcome this
latency but also enhances data privacy [57]. challenge by enabling users to understand and interpret the
However, the adoption of AI at the edge faces certain obsta- model’s features and decision-making process [32]. This
cles, particularly concerning power and computation effi- approach aims to empower stakeholders to justify actions
ciency. Traditional processors were not originally optimized taken based on algorithm recommendations, thereby increas-
for handling these specific calculations [58]. As industries ing trust in the model’s results [60].

VOLUME 12, 2024 131625

 R. Rakholia et al.: Advancing Manufacturing Through Artificial Intelligence: Current Landscape

The momentum in the XAI field has given rise to libraries A. DATA COLLECTION
like ELI5, Skater, SHAP, Tensorflow What-if Tool, and var- 1) INTERNET OF THINGS (IoT) SENSORS
ious tools offered by companies such as Google and Fiddler In manufacturing settings, Internet of Things (IoT) sensors
Labs. With these tools, customers can build machine learn- play a pivotal role in data collection, revolutionizing pro-
ing systems that prioritize high transparency, explaining the duction processes’ monitoring and optimization [66]. These
model’s learned features and their impact in predictions. smart sensors are strategically deployed throughout the man-
As a result, this fosters enhanced transparency and trust in ufacturing unit to gather real-time data from machines,
decision-making processes guided by AI. Additionally, plat- equipment, and various production parameters. They contin-
forms like AI4EU provide algorithms like LORE, LioNets, uously collect information on factors such as temperature,
and SHAP-Tree-MCDA for use in explainable AI projects. pressure, humidity, machine performance, energy consump-
tion, and more. This wealth of data gives manufacturers
H. TRANSFER LEARNING valuable insights into their operations, enabling them to make
Transfer learning has emerged as a valuable asset in machine informed decisions for process improvement, predictive
learning applications, enabling the transfer of knowledge maintenance, and quality control. With IoT sensors, manu-
gained from one task to improve the performance of another facturers can achieve enhanced efficiency, reduced downtime,
related task while saving computational costs [61]. By over- and improved product quality, ultimately leading to cost
coming data scarcity and reducing training time, transfer savings and a competitive edge in the industry [67].While
learning allows manufacturers to build more effective models FIWARE makes it easier to manage and connect IoT devices,
with limited labeled data [62]. The ability to extract mean- a direct connection is still required to retrieve the data they
ingful representations from pre-trained models, fine-tune generate [68]. This connection must adhere to the standard
them for specific manufacturing tasks, and leverage general for semantic interoperability (ISO 21823-3). Various libraries
patterns from other domains leads to improved model per- have been created to interact with devices using widely-used
formance in many tasks, such as defect recognition, process programming languages like Python. Some examples of
optimization, and outcome prediction. Furthermore, transfer such libraries include aws/aws-iot-device-sdk-python, ibm-
learning fosters generalization and robustness [63], making watson-iot/iot-python, Iotc, and Python Client for Cloud IoT
models better equipped to handle the variations and noise API.Lampropoulos et al. [69] reviews how augmented reality
inherent in manufacturing processes. (AR) can be enhanced by integrating deep learning, semantic
Traditional methods like Markov logic and Bayesian net- web, and knowledge graphs. It highlights the exponential
works have commonly been used to tackle these issues [64]. growth of data and the need for real-time, adaptive, and
However, in recent times, deep learning has gained popular- personalized information. AR provides a dynamic way to
ity, and researchers have been employing techniques where interact with both the physical and digital worlds. By combin-
the final few layers of deep models are trained on new data ing AR with deep learning, systems can gain intelligence and
to address these problems effectively. This approach has been improve computer vision. Semantic web technologies and
extensively reviewed by Weiss et al. [65]. The adaptability of knowledge graphs offer semantically interconnected infor-
pre-trained models to new data ensures that manufacturing mation, enhancing the retrieval and understanding of data.
systems can continuously learn and evolve, responding to The study discusses various case studies and applications
changes in the manufacturing environment. Additionally, the that utilize these technologies to improve user experience
incorporation of expert knowledge through transfer learning and interaction, ultimately suggesting that this integration can
further enhances the model’s performance and applicability lead to the development of intelligent, user-friendly applica-
to specific manufacturing challenges [65]. tions [69].

III. DATA COLLECTION AND STORAGE

In the era of Industry 4.0, the collection and storage of 2) SUPERVISORY CONTROL AND DATA ACQUISITION
data serve as the foundational pillars that support the vast (SCADA) SYSTEMS
landscape of AI-driven manufacturing. This section delves SCADA systems play a crucial role in manufacturing as
into the intricate web of technologies and methodologies that they effectively collect, monitor, and control data. They con-
enable the seamless acquisition and preservation of critical tinuously gather real-time data from diverse sensors and
data. From deploying sophisticated Internet of Things (IoT) devices [70], enabling swift issue identification and pro-
sensors to utilizing advanced Manufacturing Execution Sys- duction optimization. Additionally, SCADA systems store
tems (MES), this section explores the diverse mechanisms historical data, aiding in trend analysis. Configurable alarms
that gather invaluable insights from every corner of the man- and event notifications promptly address abnormal situations.
ufacturing floor. Furthermore, it examines the robust storage Moreover, SCADA systems ensure compliance in indus-
solutions, including both traditional databases and modern tries with strict regulations by facilitating data collection
data lakes, that ensure the integrity and accessibility of this and reporting for audits. Furthermore, SCADA systems are
data, paving the way for intelligent analysis and informed indispensable in modern manufacturing, offering comprehen-
decision-making in the innovative factory environment. sive data collection capabilities that enhance productivity,

131626 VOLUME 12, 2024

R. Rakholia et al.: Advancing Manufacturing Through Artificial Intelligence: Current Landscape

4) COMPUTER VISION SYSTEMS

Computer Vision Systems (CVS) are powerful tools for
data collection across diverse industries and applications,
capturing information primarily in images and videos [1].
In manufacturing, CVS is crucial in enhancing data col-
lection, streamlining processes, and boosting efficiency.
By identifying and classifying defects in real-time, CVS
empowers manufacturers to address root causes, enhance
production efficiency, and minimize waste [77]. Further-
more, through real-time monitoring, CVS ensures worker
safety [78] by detecting unsafe actions and fostering a safer
work environment. Additionally, CVS conducts automated
visual inspections of products, ensuring compliance with
quality standards and offering valuable data on product
quality.

B. DATA STORAGE
1) RELATIONAL AND NOSQL DATABASES
FIGURE 2. Illustration of MES functions and their interconnections with The adoption of both relational and NoSQL databases in man-
other systems. The extent of these linkages can vary depending on the
specific application. Source: [71]. ufacturing [79] brings notable advantages, as each excels in
managing distinct data types with their strengths [80]. Rela-
tional databases such as MySQL, PostgreSQL, Microsoft
quality, and efficiency while ensuring safety and compli- SQL Server, and Oracle excel at managing structured data,
ance. Libraries like Ignition enable Python to interface with ensuring data integrity, and supporting powerful querying
SCADA systems, typically storing data in a database for easy capabilities, making them suitable for critical manufacturing
extraction. information such as product specifications and inventory lev-
els. On the other hand, NoSQL databases like MongoDB,
3) MANUFACTURING EXECUTION SYSTEMS (MES) Cassandra, and Amazon DynamoDB handle unstructured
It has become a prevalent practice to implement a Manufac- and semi-structured data, enabling scalability, flexibility, and
turing Execution System (MES) as outlined by [71]. This high performance, making them ideal for managing data
system facilitates bi-directional communication throughout from sensors, computer vision systems, and IoT devices [19].
the enterprise, connecting data from all production facilities, Additionally, databases like InfluxDB are gaining traction
as illustrated in Fig. 2, depicting the functions of an MES. in storing and retrieving time series data efficiently, making
Some studies have explored optimal methods for extracting them a great choice for real-time monitoring and analytics.
data from this MES, such as the research conducted by By leveraging both database types, manufacturing companies
[72], [73]. may achieve optimal data storage, integrity, and scalability,
Implementing MES provides a centralized and compre- tailored to the diverse needs of their data sources and appli-
hensive platform for continuous improvement initiatives cations [81].
and operational excellence in modern manufacturing [74].
MES enables seamless information flow and better 2) DATA LAKES AND WAREHOUSES
decision-making across the organization by integrating with Data lakes are versatile data repositories capable of stor-
other enterprise systems, such as Enterprise Resource Plan- ing diverse data types from various sources and providing
ning (ERP) and Supply Chain Management (SCM). MES are a unified view of the manufacturing process through data
integral to the success of the manufacturing industry and play integration [82]. They enable manufacturers to retain raw
a critical role in multiple aspects of manufacturing, including data without transformation, facilitating analysis and insights
production planning and scheduling, real-time production at a later stage as needed. Additionally, data lakes support
monitoring, quality management, inventory control, trace- real-time monitoring and analysis of data from connected
ability, equipment maintenance, workforce optimization, and devices, offering a cost-effective solution for long-term data
data analytics [75]. storage. On the other hand, data warehouses are designed to
The diverse range of data accessible within smart man- store structured and processed data [83], ready for query-
ufacturing systems is illustrated in Fig. 3, as presented by ing and reporting. They offer an organized and curated
Cui et al. [76]. This figure encapsulates their comprehensive approach to data storage, allowing manufacturing managers
review of cutting-edge research in the domain of big data in and executives to generate reports and gain insights into the
manufacturing. company’s overall performance, including key performance

VOLUME 12, 2024 131627

 R. Rakholia et al.: Advancing Manufacturing Through Artificial Intelligence: Current Landscape

FIGURE 3. Depicts the diverse range of data formats encountered within smart manufacturing systems. Source: [71].

indicators (KPIs) related to production efficiency, quality, effectively leveraging specialized libraries and technologies
and downtime [84]. By leveraging both of them, manufac- to extract valuable insights from their data. With proper atten-
turing companies may fully exploit their data’s potential for tion to data security and compliance, file systems provide
informed decision-making and process optimization. a solid foundation for unlocking the full potential of data-
driven manufacturing.
3) CLOUD STORAGE
Cloud storage serves as a flexible solution for manufactur- 5) DATA SECURITY, BACKUP, AND RECOVERY
ing industries, facilitating effortless scaling of data storage
Data security is of utmost importance in manufacturing indus-
needs to accommodate the vast volumes of data generated
tries, as they handle sensitive information, including produc-
from sensors, machines, IoT devices, and other systems [85].
tion processes, customer data, and more. Robust measures
They eliminate on-premises hardware and maintenance,
such as encryption, access controls, and user authentica-
delivering significant cost savings for manufacturers while
tion safeguard against unauthorized access and cyber threats,
ensuring robust security, including encryption and access
ensuring data confidentiality and integrity [88]. Addition-
controls, for data privacy and protection [86]. With seam-
ally, data security protects valuable intellectual property,
less integration into AI platforms, cloud storage empowers
preventing data theft or unauthorized disclosure that could
manufacturers to harness AI and machine learning algo-
harm a company’s competitive advantage. Regular data
rithms for data analysis and modeling [87]. Moreover, cloud backups and off-site recovery enable swift restoration of
storage solutions’ automated data backup and redundancy critical information, minimizing disruptions caused by data
ensure continuous data availability and disaster recovery loss and facilitating seamless production resumption. Proac-
readiness. Various cloud-based data storage solutions, includ- tive testing of data recovery procedures further enhances
ing industry-leading platforms like Amazon S3, Microsoft resilience, reducing the potential impact of cybersecurity
Azure, Google Cloud, IBM Cloud, Snowflake’s Cloud, and threats, accidental data deletion, and hardware failures, pro-
Oracle Cloud Infrastructure, are available. They offer a wide viding manufacturers with peace of mind and a proactive
range of storage options and services, catering to diverse data approach to risk management.
storage requirements for manufacturing industries. The manufacturing industry can leverage a diverse range
of powerful tools to tackle data backup, recovery, and secu-
4) FILE SYSTEMS rity challenges effectively. Solutions like Veeam, Acronis,
In the manufacturing industry, file systems are crucial for and Commvault facilitate regular backups and efficient data
effectively storing and managing diverse data types, span- restoration, ensuring critical data remains protected. Cloud
ning from images and video to structured formats such as providers such as AWS, Microsoft Azure, and GCP offer
JSON and CSV. Manufacturers rely on this data from sen- secure cloud storage and backup options. At the same time,
sors, cameras, and machines to optimize operations, ensure companies like IBM Resiliency Services and Datto pro-
quality control, and make informed decisions. Manufacturers vide specialized disaster recovery services tailored to the
can improve their production processes and efficiency by unique needs of manufacturing. VPN software like OpenVPN

131628 VOLUME 12, 2024

R. Rakholia et al.: Advancing Manufacturing Through Artificial Intelligence: Current Landscape

and Cisco AnyConnect facilitates secure remote access to utilization of both distributed and centralized systems across
manufacturing networks and data. Mobile Device Manage- various manufacturing processes.
ment (MDM) solutions, such as VMware Workspace ONE
and Microsoft Intune, effectively manage and secure mobile B. WASTE REDUCTION AND SUSTAINABILITY
devices within manufacturing environments. By leveraging AI integration in manufacturing is crucial for achieving
these advanced tools and technologies, manufacturers can waste reduction and sustainability goals [92]. AI optimizes
confidently safeguard their data, optimize operations, and production processes, minimizing emissions and ensuring
maintain a resilient and secure manufacturing environment. compliance with environmental regulations. It assesses the
environmental impact of different practices and materials,
IV. AI APPLICATIONS IN MANUFACTURING guiding manufacturers toward more sustainable options [93].
Integrating AI into the manufacturing industry has initiated Additionally, AI facilitates the adoption of circular econ-
a revolutionary phase known as Industry 4.0. This section omy practices, focusing on recycling and reusing products
explores the applications of AI (Fig. 4) in the manufacturing and materials [94]. These AI-driven initiatives can make the
sector, where conventional procedures are being transformed manufacturing industry more environmentally friendly and
by automation, data analysis, and machine learning. From economically efficient.
predictive maintenance and quality control to supply chain
optimization and human-robot collaboration, AI enhances the C. PRODUCTION PLANNING AND RESOURCE
manufacturing industry’s efficiency, productivity, and com- OPTIMIZATION
petitiveness.
AI offers invaluable capabilities in revolutionizing produc-
tion planning and resource optimization within the man-
ufacturing industries [95]. Manufacturers can achieve cost
efficiencies and maintain seamless operations by leverag-
ing AI’s capabilities to forecast future demand accurately,
optimize production schedules, manage energy consump-
tion, and refine inventory levels. Furthermore, AI’s ability
to analyze resource utilization patterns facilitates optimal
allocation, making production processes more efficient and
effective [96]. Adopting AI technologies can empower man-
ufacturing industries to achieve higher productivity, reduce
costs, and ultimately achieve a competitive advantage in the
market. However, it is essential to implement AI technologies
FIGURE 4. Applications of AI in manufacturing. thoughtfully and consider the potential implications for the
workforce, data privacy, and security aspects [97].

A. PREDICTIVE MAINTENANCE D. HUMAN-ROBOT COLLABORATION

Predictive Maintenance (PdM) in manufacturing employs Human-Robot Collaboration (HRC) in manufacturing repre-
data analysis and advanced machine learning algorithms sents a revolutionary idea, uniting the strengths of humans
to prevent equipment failures proactively. By predicting and robots within shared workspaces to enhance produc-
potential issues, manufacturers can optimize maintenance tivity, efficiency, and safety [98]. Humans bring adaptabil-
schedules, increasing productivity and minimizing produc- ity, creativity, and critical thinking to collaboration, while
tion delays [89]. robots provide precision, speed, and endurance for repet-
PdM also yields cost savings, improved equipment relia- itive tasks [99]. HRC optimizes manufacturing workflows
bility, enhanced workplace safety, and energy efficiency [90]. by strategically allocating tasks, streamlining processes, and
IoT and interconnected devices facilitate remote monitoring, implementing user-friendly interfaces. As HRC develops and
while the data they gather fosters continuous improvement gains prominence, it is poised to reshape the manufacturing
in manufacturing processes. PdM empowers manufacturers landscape, driving efficiency, innovation, and safety to new
to operate efficiently, avoid disruptions, and make informed heights. However, the successful integration of HRC demands
decisions for a more reliable and productive future. Several careful planning, thoughtful task allocation, and comprehen-
leading manufacturing companies have adopted AI-driven sive training and support for human workers to adapt and
predictive maintenance strategies, as outlined in a case study thrive in this automated future.
presented by [58].
The SER-ENA (VerSatilE plug-and-play platform enabling E. RECOMMENDER SYSTEMS
remote pREdictive mainteNAnce) project, funded by the Recommender systems play a crucial role in manufacturing
EU, has successfully developed micro-services that facilitate by helping companies optimize their operations, improve
predictive analytics [91]. These micro-services allow for the product quality, and enhance decision-making processes.

VOLUME 12, 2024 131629

 R. Rakholia et al.: Advancing Manufacturing Through Artificial Intelligence: Current Landscape

These systems leverage data analytics, machine learning, and 2) AGEING WORKFORCE
artificial intelligence techniques to suggest the most suitable As the baby boom generation continues to age and birth
products, materials, processes, and actions to meet specific rates decline, we are witnessing a growing presence of older
manufacturing goals. Throughout the research and develop- workers in various industries. Stedmon et al. [105] discuss the
ment process, recommendation systems leverage customer valuable experience these workers bring. Still, as they natu-
feedback and market trends to facilitate successful product rally slow down with age, it is essential to ensure they are not
innovation. Reinforcement learning and collaborative filter- placed in situations where they can no longer perform their
ing have emerged as prominent techniques in contemporary tasks safely. Additionally, older workers possess a wealth
recommender systems. A study by Zhang et al. [100] and of institutional knowledge, though it can be challenging to
Zhang et al. [101] delved into the current state of the art articulate why their methods are the most effective [105].
concerning AI-based recommender systems. Their findings Identifying this tacit knowledge can be a complex
highlighted the wide variety of algorithms and applications endeavor, as it is not always straightforward to express why
in practice. The choice of algorithms depends on the specific certain approaches are superior in words. An AI system
manufacturing problem and the available data. Companies has the potential to observe and compare the interactions of
like Syte and Yupsify offer services that help businesses experienced workers with newer ones. Through this analysis,
create recommender systems. it can uncover behaviors exhibited by experienced work-
ers that may not have been previously recognized as best
F. DIGITAL TWINS practices.
The concept of a digital twin holds the promise of playing
a crucial role in realizing the vision of smart manufac- 3) CAREER PROGRESSION
turing [102]. Digital twins in manufacturing provide an Subjectivity becomes challenging in many organizations
advanced simulation platform that enables organizations to regarding career progression unless worker performance can
digitally test and enhance their processes, reducing the need be perfectly measured. Work products may be transferred
for physical prototypes and minimizing costly errors [103]. across shifts in certain roles, making it difficult to pinpoint
Recent technological advancements, including smart sensors, the origin of issues and identify the individuals responsi-
cloud computing, AI, and the emergence of Industry 4.0, have ble for outstanding work. Enhanced connectivity and the
paved the way for digitizing manufacturing processes and implementation of AI can aid in mitigating bias and tracking
the development of digital twins. In the manufacturing work- the relationships between product output and contributors.
flow, cyber-physical systems utilize sensors to collect data According to Buck and Morrow [106], using AI in perfor-
from the real-world environment. This data is subsequently mance management not only assists managers in responding
fed as input into the digital twin. The Digital Twin then to and motivating employee performance but also reduces
processes and analyzes this data, producing outputs through the time spent on evaluation. Furthermore, it enhances trans-
virtual simulations [103]. This approach enables the assess- parency and diminishes bias [106].
ment of possible outcomes and the early identification of
issues before initiating the manufacturing process, all without 4) WORKING CONDITIONS AND INCLUSION
incurring additional costs. Monitoring and analyzing work environments along with
human interaction with different components of a produc-
G. HUMAN FACTOR tion line can alleviate monotonous tasks, enhance worker
1) DIVERSITY efficiency, and mitigate inefficiencies within the processes.
A diverse team can create an environment that fosters new Incorporating AI in human resource management can also
ideas and innovation. However, many industries need help enhance organizational diversity and inclusivity while pro-
to improve diversity or meet diversity targets at all levels moting fairer career progression by minimizing bias in the
of an organization. AI can be employed to reduce bias in process.
hiring by automatically screening CVs and treating applicants
fairly. Furthermore, it can be utilized to target advertisements H. ENHANCING SAFETY FOR WORKERS AND THE PUBLIC
toward candidates who may have yet to consider applying to a Worker safety can be enhanced by implementing com-
manufacturing setting, potentially leading to increased diver- puter vision for monitoring the use of personal protective
sity among job applicants. In a study conducted by Harver in equipment (PPE) [107]. Additionally, predictive maintenance
2020 [104], several dimensions of diversity were identified, can play a crucial role in accident prevention and injury
including ethnicity, gender, marital status, socio-economic avoidance by generating maintenance requests before the
background, work style, and professional background. Sur- need for repairs becomes apparent. By establishing intercon-
prisingly, the research found that few companies incorporate nected systems, it becomes possible to continuously monitor
discussions about diversity that include the ‘majority’ group, the human-machine interface at every stage of the pro-
and inclusion policies often overlook aspects of diversity, duction line, thereby ensuring equipment’s safe and proper
such as age. utilization.

131630 VOLUME 12, 2024

R. Rakholia et al.: Advancing Manufacturing Through Artificial Intelligence: Current Landscape

I. QUALITY CONTROL to real-world manufacturing challenges. These collaborations

Quality control is crucial because defects can lead to penalties lead to process optimization, defect reduction, and improved
in later stages of production or pose a risk of injury to the product quality. Collaborating with industry also allows
end user of the final product. AI systems excel at reducing academia to understand commercial needs better and tailor
defects, often identifying optimal parameter configurations their AI research for practical applications, leading to more
within established control guidelines without requiring sig- market-ready solutions.
nificant alterations to the production process. Companies like Additionally, industries play a crucial role in the validation
DataProphet and Bosch have demonstrated this capability by of AI concepts and conducting large-scale pilot projects.
identifying internal casting defects and spotting missing parts They can provide the necessary infrastructure, data, and
on a production line, respectively. resources to test AI applications in authentic manufacturing
The EU-funded GO0D MAN (Agent-Oriented Zero environments. This enables researchers to understand practi-
Defect Multi-stage Manufacturing) project has developed cal challenges deeply and fine-tune their models accordingly.
Multi-Agent Systems (MAS) and smart inspection tools Governments and funding agencies often provide funding
to integrate into industrial systems seamlessly. These tools and support for AI research and development in manufac-
observe and optimize the key parameters to achieve consis- turing [110]. This financial backing helps academia and
tently high-quality results. industry to undertake ambitious projects, develop prototypes,
and overcome initial barriers to adoption. Governments also
have a crucial role in establishing a regulatory framework
J. ENVIRONMENTAL FOOTPRINT
for AI in manufacturing. This ensures ethical and responsible
The integration of AI into manufacturing represents a
AI use, data privacy, and safety standards, giving confidence
transformative approach to curbing the environmental foot-
to both academia and industry to invest in AI technologies.
print [108]. Through advanced predictive analytics and
Governments can act as facilitators and advocates, promoting
real-time monitoring, AI efficiently manages resources, cuts
collaboration between academia and industry and organizing
waste, and boosts energy efficiency [109]. Manufacturers can
workshops, conferences, and seminars to foster knowledge
make data-driven choices that foster sustainable practices,
exchange and partnership opportunities.
ultimately curbing carbon emissions and preserving vital
resources. Automation and process refinement facilitated by
AI significantly promote eco-friendly manufacturing, paving VI. RECOMMENDATIONS AND BEST PRACTICES
the way for a greener and more environmentally responsible In recent years, manufacturing processes have evolved sig-
industrial sphere. nificantly, driven by technological advancements and the
relentless pursuit of excellence in production. Manufacturers
across various sectors have recognized the immense potential
V. PERSPECTIVES ON AI ADOPTION IN THE
of AI to optimize their operations. As the adoption of AI in
MANUFACTURING ECOSYSTEM
manufacturing continues to accelerate, it becomes increas-
Collaboration among academia, industry, and government
ingly relevant for organizations to have a comprehensive
plays a crucial role in AI adoption in manufacturing. With
understanding of the best practices and recommendations that
distinct strengths and resources, each stakeholder contributes
can guide them on their AI journey. We will explore the key
to developing and implementing state-of-the-art AI solutions
recommendations and best practices manufacturers should
in the manufacturing sector.
consider when implementing AI solutions.
Through fundamental research, universities and research
institutions drive breakthroughs in AI technologies, mak- • Ensure the availability of high-quality, reliable, and rel-
ing manufacturing processes more advanced and efficient. evant data for AI models and concurrently implement
Additionally, academia’s commitment to training the next data collection and management protocols to maintain
generation of AI experts ensures a skilled workforce. At the data integrity.
same time, their emphasis on open innovation and collab- • Promote collaborative efforts and interdisciplinary
oration fosters knowledge sharing and benefits industries approaches involving AI experts, researchers, data sci-
at large. In the present era, with AI becoming increasingly entists, and domain-specific professionals to gain a
prevalent, it’s essential to recognize the significance of AI comprehensive understanding of specific needs and
companies, startups, and consulting firms in facilitating the challenges in the integration of AI into existing manu-
widespread adoption of AI across various industries. Beyond facturing processes.
research efforts, the successful implementation of AI into • Prioritize data security and privacy to protect sensitive
practical applications necessitates a strong foundation in soft- manufacturing information from potential breaches.
ware engineering skills, which can be effectively integrated • Begin with pilot projects and small-scale implementa-
with research and development (R&D) teams. tions of AI in manufacturing processes to understand its
Manufacturing companies have access to valuable impact and gain valuable insights; subsequently, scale
domain-specific knowledge and data, and they actively col- up gradually to minimize risks and ensure seamless
laborate with researchers and engineers to apply AI solutions integration.

VOLUME 12, 2024 131631

 R. Rakholia et al.: Advancing Manufacturing Through Artificial Intelligence: Current Landscape

• Provide training programs to upskill the existing work- • AI analytics and algorithms have been used to extract
force in AI-related technologies to cultivate a culture of insights from vast amounts of manufacturing data, help-
innovation and continuous improvement. ing identify trends, patterns, and potential areas of
• Incorporate ethical considerations throughout the devel- improvement.
opment and deployment of AI systems to foster trust • Successfully integrating AI into the manufacturing
among employees and stakeholders and ensure fairness, sector necessitates a proficient workforce and a cul-
responsibility, and transparency in decision-making tural transformation that embraces data-driven decision-
processes. making and fosters harmonious collaboration between
• Maintain awareness of industry regulations and priori- human expertise and machine capabilities.
tize compliance with safety and quality standards for AI • Security concerns related to AI implementation, such as
systems. data privacy and the potential for cyber-attacks, must be
• Consider a combination of cloud and edge computing carefully addressed to protect intellectual property and
for AI applications in manufacturing to balance compu- sensitive information.
tational power and real-time processing needs. • Despite the numerous benefits, some manufacturing
• Define key performance indicators (KPIs) and bench- companies (especially smaller ones with limited finan-
marks to measure the success and impact of AI imple- cial resources) face challenges related to high initial
mentations in manufacturing processes. investments, integration complexities, and resistance to
• Implement feedback loops, data-driven analytics, and change, which can create a cultural barrier to AI adop-
robust monitoring processes to identify areas for tion within the organization. Additionally, specialized
improvement and continuously refine AI models based AI talent and expertise may be needed to develop,
on the most recent data. This adaptive approach enables deploy, and maintain AI systems.
the AI models to optimize performance over time • AI systems heavily rely on large volumes of high-quality
effectively. data to operate effectively. Some manufacturing compa-
nies may lack the necessary data infrastructure, or their
VII. FINDINGS
data might be incomplete, outdated, or low-quality.
This section presents the primary discoveries, prevailing Without access to adequate data, AI systems may not
provide the desired level of accuracy and efficiency.
trends, and noteworthy advancements in AI applications
• The implementation of AI in manufacturing must also
within the manufacturing industries. It provides a concise
adhere to industry regulations and ethical guidelines.
summary of key insights. It emphasizes how these findings
currently influence future research and real-world imple- Companies must be aware of the legal implications of
mentation, serving as a valuable guide for readers and using AI and ensure that their AI systems comply with
decision-makers navigating this dynamic field. relevant laws and standards.
• Ongoing research and development efforts focus on
• AI adoption in manufacturing industries has signif- creating more robust AI solutions for manufacturing,
icantly impacted productivity, efficiency, and overall aiming to overcome existing limitations and unlock even
operational performance. greater potential for the industry.
• Smart manufacturing systems that leverage AI and IoT
technologies enable seamless communication between VIII. CHALLENGES AND FUTURE DIRECTIONS
machines and facilitate data-driven decision-making.
In recent years, AI has experienced remarkable growth and
• Machine learning algorithms have been extensively used
has brought about transformative changes across numerous
to optimize production processes, resulting in reduced
domains within the manufacturing industry. Nonetheless,
downtime, improved quality control, and increased
to fully harness the potential of AI in manufacturing, it is
throughput.
crucial to address several challenges and explore future direc-
• Predictive maintenance powered by AI has demon-
tions.
strated its ability to identify equipment failures before
The following are significant challenges and future direc-
they occur, leading to substantial cost savings and min-
tions to consider:
imizing unplanned downtime.
• AI-powered computer vision systems have revolution- • AI algorithms heavily depend on extensive data to
ized quality inspection, allowing for real-time defect make accurate predictions and decisions. Neverthe-
detection with higher accuracy than human inspection. less, obtaining relevant data from various sources and
• AI-driven supply chain optimization has improved ensuring its quality and consistency in manufactur-
inventory management, reduced lead times, and more ing can pose significant challenges. Future endeavors
effective demand forecasting. should prioritize the establishment of robust data col-
• Autonomous collaborative robots have become integral lection processes, the integration of data from disparate
to manufacturing processes, enhancing flexibility and systems, and the assurance of data accuracy. These
worker safety. efforts are essential to maximize the effectiveness of AI

131632 VOLUME 12, 2024

R. Rakholia et al.: Advancing Manufacturing Through Artificial Intelligence: Current Landscape

algorithms in manufacturing. In manufacturing environ- sustainable supply chain management, and facilitate the
ments, a complex ecosystem of interconnected systems, adoption of circular economy principles. The field of AI
such as enterprise resource planning (ERP), customer is progressing swiftly, and constant innovation and col-
relationship management (CRM), and manufacturing laboration among academia, industry, and policymakers
execution systems (MES), is prevalent. Seamless inte- are essential to propel advancements in AI-enabled man-
gration and interoperability between these systems ufacturing. Moving forward, promoting collaborative
and AI platforms are vital for optimizing manufac- research initiatives, knowledge-sharing platforms, and
turing processes. Future directions should focus on public-private partnerships is crucial. These efforts will
developing standardized interfaces and protocols that foster innovation, tackle challenges, and seize opportu-
facilitate smooth data exchange and interoperability nities in AI-driven manufacturing.
across diverse systems. This approach will contribute to • Manufacturing systems face cybersecurity vulnerabili-
streamlining operations and maximizing the benefits of ties, which are further amplified by the integration of AI.
AI in manufacturing. Safeguarding sensitive data’s intellectual property and
• The integration of AI in manufacturing can raise con- ensuring the security of AI-enabled manufacturing sys-
cerns regarding job displacement and workforce adap- tems is paramount. Future directions should prioritize
tation. While AI has the potential to automate certain the implementation of robust cybersecurity measures,
tasks, it can also create new opportunities and enhance encryption techniques, and ongoing monitoring to detect
productivity. To optimize benefits, it is essential to prior- and prevent potential attacks. With the growing inte-
itize investments in reskilling and upskilling the existing gration of AI in manufacturing processes, it is also
workforce, enabling them to collaborate effectively with essential to establish regulatory frameworks and eth-
AI systems. Future directions should emphasize the ical guidelines to address critical concerns like data
implementation of workforce development programs privacy, algorithmic bias, and liability. Looking ahead,
that concentrate on acquiring skills that complement it is important to define clear regulations and ethical
AI technologies. With the increasing sophistication frameworks that promote responsible and accountable
of AI systems, comprehending and explaining their use of AI within the manufacturing industry.
decision-making processes becomes crucial. In manu- By effectively addressing these challenges and embracing
facturing, where decisions have substantial implications AI’s future directions in manufacturing, the industry could
for production and quality, the absence of explainability unlock tremendous potential for enhanced productivity, effi-
can impede the adoption of AI solutions. Future research ciency, sustainability, and competitiveness.
should emphasize the development of AI models and
algorithms that offer transparent explanations for their IX. CONCLUSION
decisions, thereby fostering trust and acceptance among The impact of AI on shaping the future of manufacturing is
users and stakeholders. profound, leading to a revolution in the industry and affect-
• AI can play a crucial role in predictive maintenance, ing economic growth, sustainability, and competitiveness
enabling manufacturers to anticipate and prevent equip- in various dimensions. AI’s integration into manufacturing
ment failures, reduce downtime, and optimize mainte- processes has the power to substantially bolster economic
nance schedules. Future directions should involve the growth. By optimizing production and enhancing product
integration of AI with Internet of Things (IoT) sensors, quality using AI, manufacturers can achieve increased output
advanced analytics, and machine learning algorithms and reduce costs. This leads to heightened competitiveness
to enable real-time monitoring, predictive maintenance, and improved profitability for manufacturing companies.
and continuous process optimization. In manufactur- Furthermore, AI-driven innovations unlock new opportuni-
ing as a field, the future of AI resides in the progress ties for product development and market expansion, con-
of autonomous and collaborative systems. Autonomous tributing to economic progress on a broader scale. AI plays
systems can autonomously make decisions and execute a crucial role in advancing sustainability within the man-
tasks, while collaborative systems involve a synergistic ufacturing sector. As businesses face mounting pressure to
partnership between humans and AI. Future research adopt environmentally friendly practices, AI offers solutions
and development endeavors shall advance autonomous for more resource-efficient operations. Through data-driven
and collaborative technologies, including human-robot insights, AI can optimize energy usage, minimize waste, and
collaboration, to enhance productivity, flexibility, and promote circular economy principles, thereby reducing the
agility in manufacturing processes. industry’s ecological footprint. Embracing AI-driven sustain-
• AI can significantly contribute to sustainable manufac- ability practices aligns with global environmental goals and
turing practices by enhancing energy usage, minimizing enhances a company’s reputation and appeal to eco-conscious
waste generation, and optimizing resource efficiency. consumers.
To further advance in this area, it is crucial to pri- Furthermore, AI elevates competitiveness by transforming
oritize the development of AI algorithms and models traditional manufacturing models. Its ability to analyze vast
that actively mitigate environmental impacts, support amounts of data and predict trends empowers companies to

VOLUME 12, 2024 131633

 R. Rakholia et al.: Advancing Manufacturing Through Artificial Intelligence: Current Landscape

make informed decisions and stay ahead of the competition. [6] M. Soori, B. Arezoo, and R. Dastres, ‘‘Artificial intelligence, machine
AI ensures manufacturers remain relevant in an increasingly learning and deep learning in advanced robotics, a review,’’ in Cognitive
Robotics, vol. 3. Elsevier, 2023, doi: 10.1016/j.cogr.2023.04.001.
dynamic business landscape by enabling agile responses [7] R. Y. Zhong, X. Xu, E. Klotz, and S. T. Newman, ‘‘Intelligent manufac-
to changing market demands and consumer preferences. turing in the context of industry 4.0: A review,’’ Engineering, vol. 3, no. 5,
The insights generated by AI can also lead to the creation pp. 616–630, Oct. 2017, doi: 10.1016/j.eng.2017.05.015.
[8] M. Ryalat, H. ElMoaqet, and M. AlFaouri, ‘‘Design of a smart fac-
of innovative products and services, giving companies a tory based on cyber-physical systems and Internet of Things towards
unique selling point in their markets. As the manufactur- industry 4.0,’’ Appl. Sci., vol. 13, no. 4, p. 2156, Feb. 2023, doi:
ing industry continues to embrace AI, collaboration and 10.3390/app13042156.
[9] M. Suvarna, K. S. Yap, W. Yang, J. Li, Y. T. Ng, and X. Wang,
knowledge-sharing become vital elements for sustainable ‘‘Cyber–physical production systems for data-driven, decentralized, and
progress. Companies, researchers, and policymakers must secure manufacturing—A perspective,’’ Engineering, vol. 7, no. 9,
pp. 1212–1223, Sep. 2021, doi: 10.1016/j.eng.2021.04.021.
work together to develop best practices and address ethi- [10] R. Rai, M. K. Tiwari, D. Ivanov, and A. Dolgui, ‘‘Machine
cal considerations. Responsible AI adoption ensures that the learning in manufacturing and industry 4.0 applications,’’ Int.
benefits of this technology are equitably distributed and that J. Prod. Res., vol. 59, no. 16, pp. 4773–4778, Aug. 2021, doi:
10.1080/00207543.2021.1956675.
potential challenges, such as job displacement, are proac- [11] T. Pereira, L. Barreto, and A. Amaral, ‘‘Network and information secu-
tively addressed. rity challenges within industry 4.0 paradigm,’’ Proc. Manuf., vol. 13,
pp. 1253–1260, Jan. 2017, doi: 10.1016/j.promfg.2017.09.047.
CRediT AUTHORSHIP CONTRIBUTION STATEMENT [12] J. Jakubowski, N. Wojak-Strzelecka, R. P. Ribeiro, S. Pashami,
S. Bobek, J. Gama, and G. J. Nalepa, ‘‘Artificial intelligence approaches
All authors have made equal and significant contributions for predictive maintenance in the steel industry: A survey,’’ 2024,
to this work, actively participating in research, conceptual- arXiv:2405.12785.
ization, design, data collection, analysis, interpretation, and [13] M. Faheem and V. C. Gungor, ‘‘Energy efficient and QoS-aware routing
protocol for wireless sensor network-based smart grid applications in
manuscript drafting. the context of industry 4.0,’’ Appl. Soft Comput., vol. 68, pp. 910–922,
Jul. 2018, doi: 10.1016/j.asoc.2017.07.045.
DECLARATION OF COMPETING INTEREST [14] L. Silvestri, A. Forcina, V. Introna, A. Santolamazza, and V. Cesarotti,
The authors declare that they have no known competing ‘‘Maintenance transformation through industry 4.0 technologies: A
systematic literature review,’’ Comput. Ind., vol. 123, Dec. 2020,
financial interests or personal relationships that could have Art. no. 103335, doi: 10.1016/j.compind.2020.103335.
appeared to influence the work reported in this paper. [15] T. Zonta, C. A. da Costa, R. da Rosa Righi, M. J. de Lima,
E. S. da Trindade, and G. P. Li, ‘‘Predictive maintenance in the indus-
DATA AVAILABILITY STATEMENT try 4.0: A systematic literature review,’’ Comput. Ind. Eng., vol. 150,
Dec. 2020, Art. no. 106889, doi: 10.1016/j.cie.2020.106889.
We do not generate or use any datasets as our work is primar- [16] M. Achouch, M. Dimitrova, K. Ziane, S. S. Karganroudi, R. Dhouib,
ily theoretical in nature. H. Ibrahim, and M. Adda, ‘‘On predictive maintenance in industry 4.0:
Overview, models, and challenges,’’ Appl. Sci., vol. 12, no. 16, p. 8081,
ACKNOWLEDGMENT Aug. 2022, doi: 10.3390/app12168081.
[17] S. Sundaram and A. Zeid, ‘‘Artificial intelligence-based smart quality
The authors acknowledge the financial support from Enter- inspection for manufacturing,’’ Micromachines, vol. 14, no. 3, p. 570,
prise Ireland’s Disruptive Technologies Innovation Fund Feb. 2023, doi: 10.3390/mi14030570.
[18] H. D. Nguyen, P. H. Tran, T. H. Do, and K. P. Tran, ‘‘Quality control
(DTIF) (Grant No. DT2020214), which made this work pos- for smart manufacturing in industry 5.0,’’ in Artificial Intelligence for
sible. They also extend their sincere gratitude to their partners Smart Manufacturing (Springer Series in Reliability Engineering). Cham,
in the PeRCEPTION project–Dawn Farms Foods Ltd. and Switzerland: Springer, 2023, pp. 35–64, doi: 10.1007/978-3-031-30510-
8_3.
Creme Global–for their invaluable contributions in helping [19] C. U. Akpuokwe, A. O. Adeniyi, S. S. Bakare, and N. E. Eneh,
to initiate and shape this research. ‘‘Legal challenges of artificial intelligence and robotics: A comprehensive
review,’’ Comput. Sci. IT Res. J., vol. 5, no. 3, pp. 544–561, Mar. 2024,
doi: 10.51594/csitrj.v5i3.860.
REFERENCES [20] M. Soori, B. Arezoo, and R. Dastres, ‘‘Internet of Things for smart facto-
[1] S. K. Baduge, S. Thilakarathna, J. S. Perera, M. Arashpour, P. Sharafi, ries in industry 4.0, a review,’’ in Internet of Things and Cyber-Physical
B. Teodosio, A. Shringi, and P. Mendis, ‘‘Artificial intelligence and smart Systems, vol. 3. Elsevier, 2023, doi: 10.1016/j.iotcps.2023.04.006.
vision for building and construction 4.0: Machine and deep learning [21] S. Elhajjar, L. Yacoub, and H. Yaacoub, ‘‘Automation in business
methods and applications,’’ Autom. Construct., vol. 141, Sep. 2022, research: Systematic literature review,’’ Inf. Syst. e-Bus. Manage., vol. 21,
Art. no. 104440, doi: 10.1016/j.autcon.2022.104440. no. 3, pp. 675–698, Sep. 2023, doi: 10.1007/s10257-023-00645-z.
[2] L. Wang, Z. Liu, A. Liu, and F. Tao, ‘‘Artificial intelligence in product [22] J. T. Licardo, M. Domjan, and T. Orehovački, ‘‘Intelligent robotics—
lifecycle management,’’ Int. J. Adv. Manuf. Technol., vol. 114, nos. 3–4, A systematic review of emerging technologies and trends,’’ Electronics,
pp. 771–796, May 2021, doi: 10.1007/s00170-021-06882-1. vol. 13, no. 3, p. 542, Jan. 2024, doi: 10.3390/electronics13030542.
[3] S. S. M. Yusoh, D. A. Wahab, H. A. Habeeb, and A. H. Azman, ‘‘Intelli- [23] A. Mehta, M. Niaz, A. Adetoro, and U. Nwagwu, ‘‘Advancements in
gent systems for additive manufacturing-based repair in remanufacturing: manufacturing technology for the biotechnology industry: The role of
A systematic review of its potential,’’ PeerJ Comput. Sci., vol. 7, p. e808, artificial intelligence and emerging trends,’’ Int. J. Chem., Math. Phys.,
Dec. 2021, doi: 10.7717/peerj-cs.808. vol. 8, no. 2, pp. 12–18, 2024.
[4] M. Javaid, A. Haleem, R. P. Singh, and R. Suman, ‘‘Artificial [24] B. Ferreira and J. Reis, ‘‘Artificial intelligence in supply chain manage-
intelligence applications for industry 4.0: A literature-based study,’’ ment: A systematic literature review and guidelines for future research,’’
J. Ind. Integr. Manage., vol. 7, no. 1, pp. 83–111, Mar. 2022, doi: in Industrial Engineering and Operations Management (Springer Pro-
10.1142/s2424862221300040. ceedings in Mathematics and Statistics), vol. 431. Springer, 2023,
[5] A. T. Rizvi, A. Haleem, S. Bahl, and M. Javaid, ‘‘Artificial intelligence pp. 339–354, doi: 10.1007/978-3-031-47058-5_27.
(AI) and its applications in Indian manufacturing: A review,’’ in Cur- [25] R. Toorajipour, V. Sohrabpour, A. Nazarpour, P. Oghazi, and M. Fischl,
rent Advances in Mechanical Engineering (Lecture Notes in Mechanical ‘‘Artificial intelligence in supply chain management: A systematic lit-
Engineering), vol. 52. Springer, 2021, pp. 825–835, doi: 10.1007/978- erature review,’’ J. Bus. Res., vol. 122, pp. 502–517, Jan. 2021, doi:
981-33-4795-3_76. 10.1016/j.jbusres.2020.09.009.

131634 VOLUME 12, 2024

R. Rakholia et al.: Advancing Manufacturing Through Artificial Intelligence: Current Landscape

[26] A. Bueno, M. Godinho Filho, and A. G. Frank, ‘‘Smart production [49] L. Zhou, L. Zhang, and N. Konz, ‘‘Computer vision techniques in
planning and control in the industry 4.0 context: A systematic literature manufacturing,’’ IEEE Trans. Syst., Man, Cybern., Syst., vol. 53, no. 1,
review,’’ Comput. Ind. Eng., vol. 149, Nov. 2020, Art. no. 106774, doi: pp. 105–117, Jan. 2023.
10.1016/j.cie.2020.106774. [50] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, ‘‘You only look once:
[27] M. Sharma, S. Luthra, S. Joshi, and A. Kumar, ‘‘Implementing challenges Unified, real-time object detection,’’ in Proc. IEEE Conf. Comput. Vis.
of artificial intelligence: Evidence from public manufacturing sector of an Pattern Recognit., Jun. 2016, pp. 779–788.
emerging economy,’’ Government Inf. Quart., vol. 39, no. 4, Oct. 2022, [51] A. Albers, B. Gladysz, T. Pinner, V. Butenko, and T. Stürmlinger, ‘‘Proce-
Art. no. 101624, doi: 10.1016/j.giq.2021.101624. dure for defining the system of objectives in the initial phase of an industry
[28] S. Phuyal, D. Bista, and R. Bista, ‘‘Challenges, opportunities 4.0 project focusing on intelligent quality control systems,’’ Proc. CIRP,
and future directions of smart manufacturing: A state of art vol. 52, pp. 262–267, Jan. 2016, doi: 10.1016/j.procir.2016.07.067.
review,’’ Sustain. Futures, vol. 2, Jan. 2020, Art. no. 100023, doi: [52] M. Panzer and B. Bender, ‘‘Deep reinforcement learning in production
10.1016/j.sftr.2020.100023. systems: A systematic literature review,’’ Int. J. Prod. Res., vol. 60, no. 13,
[29] R. Anane-Simon and S. O. Atiku, ‘‘Artificial intelligence and automa- pp. 4316–4341, Jul. 2022.
tion for the future of startups,’’ in Ecosystem Dynamics and Strategies [53] P. Zajec, J. M. Rožanec, I. Novalija, B. Fortuna, D. Mladenić, and
for Startups Scalability. IGI Global, Apr. 2023, pp. 133–153, doi: K. Kenda, ‘‘Towards active learning based smart assistant for manufac-
10.4018/979-8-3693-0527-0.ch007. turing,’’ in Proc. IFIP Int. Conf. Adv. Prod. Manag. Syst. Nantes, France:
[30] A. V. Asimit, I. Kyriakou, S. Santoni, S. Scognamiglio, and R. Zhu, Springer, Sep. 2021, pp. 295–302.
‘‘Robust classification via support vector machines,’’ Risks, vol. 10, no. 8, [54] G. Chen, Y. Li, C. Mehdi-Souzani, and X. Liu, ‘‘Active transfer learning
p. 154, Aug. 2022, doi: 10.3390/risks10080154. for data-driven manufacturing process modelling,’’ Proc. CIRP, vol. 118,
[31] A. M. Ikotun, A. E. Ezugwu, L. Abualigah, B. Abuhaija, and J. Heming, pp. 98–103, Jan. 2023.
‘‘K-means clustering algorithms: A comprehensive review, variants anal- [55] P. Ren, Y. Xiao, X. Chang, P.-Y. Huang, Z. Li, B. B. Gupta, X. Chen,
ysis, and advances in the era of big data,’’ Inf. Sci., vol. 622, pp. 178–210, and X. Wang, ‘‘A survey of deep active learning,’’ ACM Comput. Surv.,
Jan. 2023, doi: 10.1016/j.ins.2022.11.139. vol. 54, no. 9, pp. 1–40, Oct. 2021.
[32] P. Linardatos, V. Papastefanopoulos, and S. Kotsiantis, ‘‘Explainable AI: [56] W. Yao, C. O. Dumitru, J. Lorenzo, and M. Datcu, ‘‘Data mining
A review of machine learning interpretability methods,’’ Entropy, vol. 23, on the CANDELA cloud platform,’’ in Proc. IGARSS, Sep. 2020,
no. 1, p. 18, Dec. 2020. pp. 6945–6948.
[33] L. Breiman, ‘‘Random forests,’’ Mach. Learn., vol. 45, no. 1, pp. 5–32, [57] K. Kubiak, G. Dec, and D. Stadnicka, ‘‘Possible applications of edge
2001, doi: 10.1023/a:1010933404324. computing in the manufacturing industry—Systematic literature review,’’
[34] S. Raschka, ‘‘Naive Bayes and text classification I—Introduction and Sensors, vol. 22, no. 7, p. 2445, Mar. 2022.
theory,’’ Oct. 2014, arXiv:1410.5329. Accessed: Aug. 3, 2024. [58] S. M. Lee, D. Lee, and Y. S. Kim, ‘‘The quality management ecosystem
[35] F. Pedregosa et al., ‘‘Scikit-learn: Machine learning in Python,’’ for predictive maintenance in the industry 4.0 era,’’ Int. J. Quality Innov.,
J. Mach. Learn. Res., vol. 12, pp. 2825–2830, Nov. 2011. [Online]. vol. 5, no. 1, pp. 1–11, Dec. 2019.
[59] J. Laux, S. Wachter, and B. Mittelstadt, ‘‘Trustworthy artificial intelli-
Available: http://scikit-learn.sourceforge.net
[36] T. Chen and C. Guestrin, ‘‘XGBoost: A scalable tree boosting system,’’ gence and the European Union AI act: On the conflation of trustworthi-
in Proc. 22nd ACM SIGKDD Int. Conf. Knowl. Discovery Data Mining, ness and acceptability of risk,’’ Regulation Governance, vol. 18, no. 1,
Aug. 2016, pp. 785–794, doi: 10.1145/2939672.2939785. pp. 3–32, Jan. 2024.
[60] R. Goebel, A. Chander, K. Holzinger, F. Lecue, Z. Akata, S. Stumpf,
[37] T. Veiga, H. A. Asad, F. A. Kraemer, and K. Bach, ‘‘Towards container-
P. Kieseberg, and A. Holzinger, ‘‘Explainable AI: The new 42?’’ in Proc.
ized, reuse-oriented AI deployment platforms for cognitive IoT applica-
Int. Cross-Domain Conf. Mach. Learn. Knowl. Extraction. Hamburg,
tions,’’ Future Gener. Comput. Syst., vol. 142, pp. 4–13, May 2023, doi:
Germany: Springer, Aug. 2018, pp. 295–303.
10.1016/j.future.2022.12.029.
[61] F. Zhuang, Z. Qi, K. Duan, D. Xi, Y. Zhu, H. Zhu, H. Xiong, and Q. He,
[38] T. Chen, V. Sampath, M. C. May, S. Shan, O. J. Jorg, J. J. A. Martín,
‘‘A comprehensive survey on transfer learning,’’ Proc. IEEE, vol. 109,
F. Stamer, G. Fantoni, G. Tosello, and M. Calaon, ‘‘Machine learning
no. 1, pp. 43–76, Jan. 2021.
in manufacturing towards industry 4.0: From ‘for now’ to ‘four-know,’’’
[62] M. Ramezankhani, B. Crawford, A. Narayan, H. Voggenreiter,
Appl. Sci., vol. 13, no. 3, p. 1903, Feb. 2023, doi: 10.3390/app13031903.
R. Seethaler, and A. S. Milani, ‘‘Making costly manufacturing smart
[39] W. Rawat and Z. Wang, ‘‘Deep convolutional neural networks for image
with transfer learning under limited data: A case study on composites
classification: A comprehensive review,’’ Neural Comput., vol. 29, no. 9,
autoclave processing,’’ J. Manuf. Syst., vol. 59, pp. 345–354, Apr. 2021.
pp. 2352–2449, Sep. 2017, doi: 10.1162/neco_a_00990. [63] A. Shafahi, P. Saadatpanah, C. Zhu, A. Ghiasi, C. Studer, D. Jacobs,
[40] X. Zhang, C. Zhong, J. Zhang, T. Wang, and W. W. Y. Ng, ‘‘Robust
and T. Goldstein, ‘‘Adversarially robust transfer learning,’’ 2019,
recurrent neural networks for time series forecasting,’’ Neurocomputing,
arXiv:1905.08232.
vol. 526, pp. 143–157, Mar. 2023. [64] P. Hofmann and Z. Tashman, ‘‘Hidden Markov models and their applica-
[41] C. Little, M. Elliot, R. Allmendinger, and S. S. Samani, ‘‘Generative tion for predicting failure events,’’ in Proc. 20th Int. Conf. Comput. Sci.
adversarial networks for synthetic data generation: A comparative study,’’ (ICCS). Amsterdam, The Netherlands: Springer, Jun. 2020, pp. 464–477.
2021, arXiv:2112.01925. [65] K. Weiss, T. M. Khoshgoftaar, and D. Wang, ‘‘A survey of transfer
[42] D. Bank, N. Koenigstein, and R. Giryes, ‘‘Autoencoders,’’ in Machine learning,’’ J. Big data, vol. 3, no. 1, pp. 1–40, May 2016.
Learning for Data Science Handbook: Data Mining and Knowledge [66] L. Haghnegahdar, S. S. Joshi, and N. B. Dahotre, ‘‘From IoT-based cloud
Discovery Handbook. Springer, 2023, pp. 353–374. manufacturing approach to intelligent additive manufacturing: Industrial
[43] K. Han, A. Xiao, E. Wu, J. Guo, C. Xu, and Y. Wang, ‘‘Transformer Internet of Things—An overview,’’ Int. J. Adv. Manuf. Technol., vol. 119,
in transformer,’’ in Proc. Adv. Neural Inf. Process. Syst., vol. 34, 2021, pp. 1461–1478, Jan. 2022.
pp. 15908–15919. [67] M. Javaid, A. Haleem, R. P. Singh, S. Rab, and R. Suman, ‘‘Upgrading
[44] C. Li, P. Zheng, Y. Yin, B. Wang, and L. Wang, ‘‘Deep reinforcement the manufacturing sector via applications of industrial Internet of Things
learning in smart manufacturing: A review and prospects,’’ CIRP J. (IIoT),’’ Sensors Int., vol. 2, Jan. 2021, Art. no. 100129.
Manuf. Sci. Technol., vol. 40, pp. 75–101, Feb. 2023. [68] V. Araujo, K. Mitra, S. Saguna, and C. Åhlund, ‘‘Performance evaluation
[45] S. Dutta, ‘‘TrajNet: Outlier detection in vehicle trajectory data using of FIWARE: A cloud-based IoT platform for smart cities,’’ J. Parallel
capsule network based one-shot learning,’’ in Proc. 29th Int. Conf. Adv. Distrib. Comput., vol. 132, pp. 250–261, Oct. 2019.
Geographic Inf. Syst., Nov. 2021, pp. 670–671. [69] G. Lampropoulos, E. Keramopoulos, and K. Diamantaras, ‘‘Enhancing
[46] Y. Chen, Z. Liu, H. Xu, T. Darrell, and X. Wang, ‘‘Meta-baseline: Explor- the functionality of augmented reality using deep learning, semantic web
ing simple meta-learning for few-shot learning,’’ in Proc. IEEE/CVF Int. and knowledge graphs: A review,’’ Vis. Informat., vol. 4, no. 1, pp. 32–42,
Conf. Comput. Vis. (ICCV), Oct. 2021, pp. 9042–9051. Mar. 2020, doi: 10.1016/j.visinf.2020.01.001.
[47] B. J. Erickson, P. Korfiatis, Z. Akkus, T. Kline, and K. Philbrick, ‘‘Toolk- [70] A. Nechibvute and H. D. Mafukidze, ‘‘Integration of SCADA and indus-
its and libraries for deep learning,’’ J. Digit. Imag., vol. 30, no. 4, trial IoT: Opportunities and challenges,’’ IETE Tech. Rev., vol. 41, no. 3,
pp. 400–405, Aug. 2017. pp. 312–325, May 2024.
[48] D. Chahal, M. Mishra, S. Palepu, and R. Singhal, ‘‘Performance and [71] I. Mesa. White Paper Number 6, MES Explained; A High Level Vision for
cost comparison of cloud services for deep learning workload,’’ in Executives, MESA 1997. Accessed: May 7, 2024. [Online]. Available:
Proc. Companion ACM/SPEC Int. Conf. Perform. Eng., Apr. 2021, http://alvarestech.com/temp/smar/www.delt.ufmg.br/seixas/PaginaII/
pp. 49–55. Download/DownloadFiles/pap6.pdf

VOLUME 12, 2024 131635

 R. Rakholia et al.: Advancing Manufacturing Through Artificial Intelligence: Current Landscape

[72] S. Lee, S. J. Nam, and J.-K. Lee, ‘‘Real-time data acquisition system and [95] D. Luo, S. Thevenin, and A. Dolgui, ‘‘A state-of-the-art on produc-
HMI for MES,’’ J. Mech. Sci. Technol., vol. 26, no. 8, pp. 2381–2388, tion planning in industry 4.0,’’ Int. J. Prod. Res., vol. 61, no. 19,
Aug. 2012. pp. 6602–6632, Oct. 2023.
[73] Y. Wang, M. Wang, J. Wang, and Y. Zhou, ‘‘Design and implementation of [96] L. Waltersmann, S. Kiemel, J. Stuhlsatz, A. Sauer, and R. Miehe,
device integration framework based on agent technology in MES,’’ Proc. ‘‘Artificial intelligence applications for increasing resource efficiency
CIRP, vol. 83, pp. 485–489, Jan. 2019. in manufacturing companies—A comprehensive review,’’ Sustainability,
[74] A. Shojaeinasab, T. Charter, M. Jalayer, M. Khadivi, O. Ogunfowora, vol. 13, no. 12, p. 6689, Jun. 2021.
N. Raiyani, M. Yaghoubi, and H. Najjaran, ‘‘Intelligent manufactur- [97] A. S. George and A. H. George, ‘‘Revolutionizing manufacturing: Explor-
ing execution systems: A systematic review,’’ J. Manuf. Syst., vol. 62, ing the promises and challenges of industry 5.0,’’ Partners Universal Int.
pp. 503–522, Jan. 2022. Innov. J., vol. 1, no. 2, pp. 22–38, 2023.
[75] S. Jaskó, A. Skrop, T. Holczinger, T. Chován, and J. Abonyi, ‘‘Develop- [98] A. K. Inkulu, M. V. A. R. Bahubalendruni, A. Dara, and
ment of manufacturing execution systems in accordance with industry 4.0 K. J. I. R. SankaranarayanaSamy, ‘‘Challenges and opportunities in
requirements: A review of standard- and ontology-based methodologies human robot collaboration context of industry 4.0—A state of the art
and tools,’’ Comput. Ind., vol. 123, Dec. 2020, Art. no. 103300. review,’’ Ind. Robot, Int. J. Robot. Res. Appl., vol. 49, no. 2, pp. 226–239,
[76] Y. Cui, S. Kara, and K. C. Chan, ‘‘Manufacturing big data ecosystem: Feb. 2022.
A systematic literature review,’’ Robot. Comput.-Integr. Manuf., vol. 62, [99] S. Kumar, C. Savur, and F. Sahin, ‘‘Survey of human–robot collaboration
Apr. 2020, Art. no. 101861. in industrial settings: Awareness, intelligence, and compliance,’’ IEEE
[77] M. Shahin, F. F. Chen, A. Hosseinzadeh, and M. Maghanaki, ‘‘Waste Trans. Syst., Man, Cybern., Syst., vol. 51, no. 1, pp. 280–297, Jan. 2021.
reduction via computer vision-based inspection: Towards lean systems [100] Q. Zhang, J. Lu, and Y. Jin, ‘‘Artificial intelligence in recommender
in metal production,’’ Res. Square, Apr. 2023, doi: 10.21203/rs.3.rs- systems,’’ Complex Intell. Syst., vol. 7, no. 1, pp. 439–457, Feb. 2021.
2782987/v1. [101] S. Zhang, Y. Tay, L. Yao, B. Wu, and A. Sun, ‘‘DeepRec: An
[78] W. Fang, L. Ding, P. E. D. Love, H. Luo, H. Li, F. Peña-Mora, B. Zhong, open-source toolkit for deep learning based recommendation,’’ 2019,
and C. Zhou, ‘‘Computer vision applications in construction safety assur- arXiv:1905.10536.
ance,’’ Autom. Construct., vol. 110, Feb. 2020, Art. no. 103013. [102] I. Onaji, D. Tiwari, P. Soulatiantork, B. Song, and A. Tiwari, ‘‘Digital
[79] V. F. de Oliveira, M. A. D. O. Pessoa, F. Junqueira, and P. E. Miyagi, twin in manufacturing: Conceptual framework and case studies,’’ Int. J.
‘‘SQL and NoSQL databases in the context of industry 4.0,’’ Machines, Comput. Integr. Manuf., vol. 35, no. 8, pp. 831–858, Aug. 2022.
vol. 10, no. 1, p. 20, Dec. 2021. [103] E. Karaarslan, Ō. Aydin, Ū. Cali, and M. Challenger, Eds., Digital
[80] G. Ekren and A. Erkollar, ‘‘The potential and capabilities of NoSQL Twin Driven Intelligent Systems and Emerging Metaverse. Singapore:
databases for ERP systems,’’ in Advanced MIS and Digital Transfor- Springer, 2023, doi: 10.1007/978-981-99-0252-1.
mation for Increased Creativity and Innovation in Business. IGI Global, [104] White Paper: The State of Diversity Recruiting in 2020, Harver,
2020, pp. 147–168. New York, NY, USA, 2020.
[81] J. Wang, W. Zhang, Y. Shi, S. Duan, and J. Liu, ‘‘Industrial big [105] A. W. Stedmon, H. Howells, J. R. Wilson, and I. Dianat,
data analytics: Challenges, methodologies, and applications,’’ 2018, ‘‘Ergonomics/human factors needs of an ageing workforce in the
arXiv:1807.01016. manufacturing sector,’’ Health Promotion Perspectives, vol. 2, no. 2,
[82] R. Hai, C. Koutras, C. Quix, and M. Jarke, ‘‘Data lakes: A survey of p. 112, 2012.
functions and systems,’’ IEEE Trans. Knowl. Data Eng., vol. 35, no. 12, [106] B. Buck and J. Morrow, ‘‘AI, performance management and engage-
pp. 12571–12590, Dec. 2023. ment: Keeping your best their best,’’ Strategic HR Rev., vol. 17, no. 5,
[83] B. Shiyal, ‘‘Modern data warehouses and data lakehouses,’’ in Beginning pp. 261–262, Oct. 2018.
Azure Synapse Analytics: Transition From Data Warehouse to Data Lake- [107] M. I. B. Ahmed, L. Saraireh, A. Rahman, S. Al-Qarawi, A. Mhran,
house, B. A. Synapse, Ed., Berkeley, CA, USA: Apress, 2021, pp. 21–48. J. Al-Jalaoud, D. Al-Mudaifer, F. Al-Haidar, D. AlKhulaifi, M. Youldash,
[84] M. Kurth, M. Göggel, C. Schleyer, J. Gratt, and B. Martin, ‘‘Filling the and M. Gollapalli, ‘‘Personal protective equipment detection: A deep-
industrial data lake in manufacturing companies based on KPIs,’’ Athens learning-based sustainable approach,’’ Sustainability, vol. 15, no. 18,
J. Technol. Eng., vol. 7, pp. 1–22, Jul. 2020. p. 13990, Sep. 2023.
[85] Q. Qi and F. Tao, ‘‘A smart manufacturing service system based on edge [108] Z. Liu, G. Yang, and Y. Zhang, ‘‘Carbon footprint assessment in manu-
computing, fog computing, and cloud computing,’’ IEEE Access, vol. 7, facturing industry 4.0 using machine learning with intelligent Internet of
pp. 86769–86777, 2019. Things,’’ Int. J. Adv. Manuf. Technol., vol. 170, pp. 1–8, Aug. 2023.
[86] C. Li, Y. Chen, and Y. Shang, ‘‘A review of industrial big data for decision [109] T. Ahmad, D. Zhang, C. Huang, H. Zhang, N. Dai, Y. Song, and
making in intelligent manufacturing,’’ Eng. Sci. Technol., Int. J., vol. 29, H. Chen, ‘‘Artificial intelligence in sustainable energy industry: Status
May 2022, Art. no. 101021. quo, challenges and opportunities,’’ J. Cleaner Prod., vol. 289, Mar. 2021,
[87] D. Soni and N. Kumar, ‘‘Machine learning techniques in emerging cloud Art. no. 125834.
computing integrated paradigms: A survey and taxonomy,’’ J. Netw. [110] EU-Research and Innovation. (2023). Digital, Industry
Comput. Appl., vol. 205, Sep. 2022, Art. no. 103419. and Space. Accessed: Oct. 6, 2023. [Online]. Available:
[88] S. Duggineni, ‘‘Impact of controls on data integrity and information https://research-and-innovation.ec.europa.eu/funding/funding-
systems,’’ Sci. Technol., vol. 13, no. 2, pp. 29–35, 2023. opportunities/funding-programmes-and-open-calls/horizon-
[89] Z. M. Çınar, A. A. Nuhu, Q. Zeeshan, O. Korhan, M. Asmael, and europe/cluster-4-digital-industry-and-space_en
B. Safaei, ‘‘Machine learning in predictive maintenance towards sustain-
able smart manufacturing in industry 4.0,’’ Sustainability, vol. 12, no. 19,
p. 8211, Oct. 2020, doi: 10.3390/su12198211.
[90] J. Lee, J. Ni, J. Singh, B. Jiang, M. Azamfar, and J. Feng, ‘‘Intelligent
maintenance systems and predictive manufacturing,’’ J. Manuf. Sci. Eng.,
RAJNISH RAKHOLIA received the bachelor’s
vol. 142, no. 11, Nov. 2020, Art. no. 110805. degree in physics, the master’s degree in com-
[91] T. Kuehne, X. Song, G. Caire, K. Rasilainen, T. H. Le, M. Rossi, I. Ndip, puter science, and the Ph.D. degree in artificial
and C. Fager, ‘‘Performance simulation of a 5G hybrid beamforming intelligence. He is currently a Data Scientist with
millimeter-wave system,’’ in Proc. 24th Int. ITG Workshop Smart Anten- Ireland’s Centre for AI (CeADAR), University
nas. Frankfurt, Germany: VDE, Feb. 2020, pp. 1–6. College Dublin, Ireland, with over a decade of
[92] A. Jamwal, R. Agrawal, and M. Sharma, ‘‘Deep learning for man- experience in research and academia. He has
ufacturing sustainability: Models, applications in industry 4.0 and been involved in developing AI solutions for
implications,’’ Int. J. Inf. Manage. Data Insights, vol. 2, no. 2, Nov. 2022, optimizing production processes and predictive
Art. no. 100107. maintenance within the food manufacturing indus-
[93] J.-P. Schöggl, M. Rusch, L. Stumpf, and R. J. Baumgartner, ‘‘Implemen-
try. His research interests include creating innovative AI-based solutions to
tation of digital technologies for a circular economy and sustainability
management in the manufacturing sector,’’ Sustain. Prod. Consumption, address climate change, enhance environmental sustainability, and advance
vol. 35, pp. 401–420, Jan. 2023. agricultural practices. Previously, he conducted research on monitoring and
[94] M. H. Ronaghi, ‘‘The influence of artificial intelligence adoption on cir- forecasting air pollution, evaluating its impacts on human health and environ-
cular economy practices in manufacturing industries,’’ Environ., Develop. mental ecosystems, in collaboration with prestigious institutions. His work
Sustainability, vol. 25, no. 12, pp. 14355–14380, Dec. 2023. has been featured in leading environmental journals.

131636 VOLUME 12, 2024

R. Rakholia et al.: Advancing Manufacturing Through Artificial Intelligence: Current Landscape

ANDRÉS L. SUÁREZ-CETRULO received the RICARDO SIMÓN CARBAJO received the Ph.D.
B.Sc., M.Sc., and Ph.D. degrees in computer sci- degree in computer science from the Trinity Col-
ence from the University Carlos III of Madrid, lege Dublin, in 2012, where his research focused
Spain, in 2013, 2014, and 2022, respectively. He is on wireless networking, signal processing, and
currently a Data Science Architect with Ireland’s real-time data analytics. Subsequently, he was a
Centre for Applied AI, University College Dublin. Senior Researcher, the Manager, and a Lecturer
His current interests include online machine learn- in various industry-focused AI research centres,
ing for data streams, regime changes in financial and the Co-Founder of start-ups. He is currently
markets, deep learning, transformers, and genera- the Director of the Innovation and Develop-
tive models. ment, Ireland’s Centre for Artificial Intelligence
(CeADAR), where he leads a large team of data science and software engi-
neering professionals. He is also the Director of European Digital Innovation
Hub in AI Program, CeADAR, and leads European program of the centre.
He coordinates large national and EU R&I projects and together with his
MANOKAMNA SINGH received the B.E. degree team work closely with industry in Ireland and Europe to help them adopt AI.
in information technology and the M.Sc. degree His research interest includes advanced machine learning with application to
in data science and analytics from Maynooth Uni- a large range of verticals. He represents CeADAR in AI associations across
versity, Ireland, in 2010 and 2018, respectively, Europe and acts as the Advisor to Irish Government.
where he is currently pursuing the Ph.D. degree
in computer science. Previously, he was a Soft-
ware Engineer with Nucleus Software Exports
Ltd., and TSYS India. He is a Data Scientist
with Ireland’s Centre for AI (CeADAR), Univer-
sity College Dublin, Ireland. He has experience in
research projects across the food manufacturing, financial, education, and
telecom domains. He has published numerous research papers in journals
and conferences. His research interests include natural language processing
and computer vision.

VOLUME 12, 2024 131637