Introduction

Hanqi Su¹ and Jay Lee¹

1 Center for Industrial Artificial Intelligence, Department of Mechanical Engineering, University of Maryland, College Park, 20742, United States of America

E-mail: hanqisu@umd.edu

The evolution toward smart manufacturing can be traced through several stages over the past six decades. Summarized in Figure 1, it highlights five phases: the early foundational phase (1960s–2000s), the new foundational phase (2000s–2010s), the rise of AI/ML-enabled smart manufacturing (2014–2025), and the progression toward next-generation AI for future manufacturing (2025–2035). In the mid-1960s, the concept of flexible manufacturing systems (FMS) was introduced to enable automated machining that could adapt to different products, with the first implementations appearing in the late 1960s. By the 1970s and 1980s, advances in digital technologies gave rise to computer-integrated manufacturing systems (CIMS), emphasizing the integration of CAD, CAM, robotics, and enterprise systems for end-to-end production management. In the early 1990s, the notion of agile manufacturing systems (AMS) emerged, focusing on responsiveness and adaptability in the face of globalization and rapidly changing customer needs. Around the same time, intelligent manufacturing systems (IMS) became a formal international research program. The IMS Program, launched in 1990–1991, was a collaborative initiative involving Japan, European Union, United States, and later other countries. It was coordinated by organizations such as the International IMS Steering Committee and supported by governments and industries. The goal was to develop intelligent, distributed, and adaptive manufacturing systems through global cooperation. These milestones collectively formed the early foundational phase, establishing automation and integration as the baseline for modern manufacturing. With the rise of digital infrastructure in the 2000s and 2010s, the new foundational phase was shaped by advances in data-driven connectivity, sensing, and operation. The introduction of the Internet of Things (IoT), cyber-physical systems (CPS), cloud computing, industrial big data analytics, prognostics and health management (PHM), and digital twins accelerated the digital and automated integration of design, production, inspection, and supply chain systems [1-4]. These advances laid the foundation for Industry 4.0, transforming traditional production into a mode characterized by digitalization, automation, and intelligence [5,6]. From 2014 onward, AI/ML-enabled smart manufacturing emerged as a central theme. Smart manufacturing, the cornerstone of modern manufacturing, is defined by the National Institute of Standards and Technology (NIST) as “fully integrated, collaborative manufacturing systems that respond in real time to meet changing demands and conditions in the factory, in the supply network, and in customer needs” [7]. This phase began with the application of deep learning (deep neural networks), transfer learning, explainable ML, and early multimodal fusion to solve specific manufacturing tasks and pilot deployments [8-10]. Building on the foundations of Industry 4.0, Industry 5.0 was proposed in 2021, introducing a vision that emphasizes human-centricity, sustainability, and resilience [11]. While Industry 4.0 focused on digitalization and automation for efficiency [6], Industry 5.0 stresses collaboration between humans and smart machines, circular economy principles to support sustainable production, and robust systems to ensure adaptability under disruptions [12]. At this stage, AI and ML methods are increasingly applied to reinforce these

principles, placing human needs at the center of manufacturing processes. More recently, the focus has expanded to accelerating and scaling advanced methods such as federated learning, multi-modal learning, physics-informed learning, and large language models (LLMs) into complex workflows [8-10]. Looking forward, the next-generation AI for manufacturing phase (2025–2035) is expected to be enabled by generative AI, agentic AI, industrial LLMs, and large-scale foundation models, driving new productive ecosystems for scalable and sustainable manufacturing [13, 14].

Figure 1. An evolution process of AI and ML in smart manufacturing.

In recent years, AI and ML have emerged as foundational technologies applied to many aspects of smart manufacturing [10, 15, 16]. On the one hand, with the growing availability of data and the increasing demand for analytics in the era of industrial big data, AI and ML techniques enable the efficient processing of large-scale data streams from equipment, sensor networks, and supply chains. These methods support the extraction of actionable insights from complex datasets and help manufacturers to perform intelligent decision-making in a timely manner. On the other hand, the integration of AI and ML has significantly advanced the automation of manufacturing processes. Previously, tasks such as quality control, product inspection, equipment maintenance, and production scheduling relied heavily on manual intervention. Right now, these tasks can be handled by intelligent algorithms. This shift reduces human workload and improves accuracy, consistency, and operational efficiency. Furthermore, with growing demand for customized and personalized products, AI and ML models can facilitate the analysis of consumer preferences and enable dynamic adjustments to production lines. As a result, manufacturers can reduce time-to-market (TTM) and improve overall market responsiveness. In addition, another major advancement in smart manufacturing is the shift from reactive to predictive maintenance. Traditional manufacturing depends on periodic maintenance and manual inspections to prevent equipment failures. In contrast, AI-and ML-based techniques for predictive maintenance and fault diagnosis can detect early signs of degradation or abnormal behaviour, and provide timely alerts before failures occur. By leveraging data-driven prediction and classification methods, manufacturing enterprises can reduce sudden shutdowns and production disruptions, achieving lower maintenance costs. Although AI and ML have brought significant potential to smart manufacturing, notable gaps remain between current AI and ML capabilities and the practical requirements of modern manufacturing systems. One of the most critical challenges is data quality. Industrial data are often noisy, incomplete, or presented in inconsistent formats, which can make them difficult to use in practice. These issues prevent effective training of AI and ML models and compromise the reliability of their predictions [17]. Another limitation is

the lack of interpretability in many AI models. While black-box models such as deep neural networks can achieve strong performance, their internal decision-making processes are often unclear. This lack of transparency makes it difficult for engineers and practitioners to fully understand or trust the outputs of AI models, particularly in high-stakes industrial applications [18]. Moreover, many manufacturing processes are governed by complex physical principles that are difficult to represent using purely data-driven models. Traditional approaches often struggle to effectively integrate these complex physical laws into AI and ML models [19]. In addition, AI and ML models trained on data from one machine, production line, or factory often perform poorly when applied to different domains, due to distribution shifts and limited availability of labeled data in the target domain. This challenge is commonly referred to as domain adaptation [20]. Finally, although AI models often demonstrate strong performance in laboratory settings, their deployment in real-world production faces challenges. These include system integration complexity, real-time performance requirements, and scalability across multiple factories and distributed supply chains.

Figure 2. This figure provides a summary of key topics in AI-enabled smart manufacturing and

various non-traditional ML techniques for smart manufacturing.

In response to the growing importance of AI and ML in transforming modern manufacturing, the aim of this roadmap on artificial intelligence and machine learning for smart manufacturing is to provide an overview of different research areas and technological developments driving progress in smart manufacturing, as shown in Figure 2. It outlines opportunities, challenges, and technological advancements for the next-generation manufacturing industry. The roadmap is organized into three main sections:

• Foundations and trends in AI for smart manufacturing: This section reviews the evolution of AI and ML in manufacturing, emphasizing their growing role in enhancing efficiency, adaptability, and automation for future manufacturing. It also discusses the outlook of AI technologies and their potential to transform manufacturing industry and value chains.

• Key topics in AI-enabled smart manufacturing: This section focuses on major application areas of AI and ML in smart manufacturing. It covers topics such as industrial big data analytics, autonomous manufacturing, additive manufacturing, process monitoring and optimization, digital twins, engineering design, smart supply chain and logistics, sustainable and green manufacturing, and AI-enhanced robotics.

• Non-traditional machine learning techniques for smart manufacturing: This section discusses emerging non-traditional machine learning paradigms and their relevance to smart manufacturing. Topics covered include deep learning, generative AI, physics-informed ML, semantic frameworks, and trustworthy and explainable AI. It also highlights developments in data-centric metrology, reliability, availability, maintainability, and safety (RAMS) in AI-enabled systems, as well as advances in large language models, industrial large knowledge models, domain adaptation, transfer learning, stream-of-quality analysis, and federated learning. We hope that this roadmap offers a comprehensive perspective and a long-term strategic foundation for advancing AI and ML in smart manufacturing. Each contribution, authored by leading researchers in their domains, presents the current state of the field, identifies key challenges, outlines the advances in science and technology needed to address them, and proposes future perspectives. In the end, we encourage broader collaboration among academic researchers, industry practitioners, funding agencies, and policymakers to work together to shape the future of manufacturing.

Acknowledgements

This work was supported by the U.S. Department of Education through the Fund for the Improvement of Postsecondary Education (FIPSE) under Grant No. P116S230014.

References

[1] Hu Y, Jia Q, Yao Y, Lee Y, Lee M, Wang C *et al.*2024 Industrial internet of things intelligence empowering smart manufacturing: a literature review *IEEE Internet Things J.*1119143–67 (doi: 10.1109/JIOT.2024.3367692) [2] Lee J, Bagheri B and Kao H-A 2015 A cyber-physical systems architecture for Industry 4.0-based manufacturing systems Manuf. *Lett.*318–23 (doi: 10.1016/j.mfglet.2014.12.001) [3] Su H and Lee J 2024 Machine learning approaches for diagnostics and prognostics of industrial systems using open source data from PHM data challenges: a review *Int. J. Progn. Health Manag.*15 (2) (doi: 10.36001/ijphm.2024.v15i2.3993) [4] Tao F, Zhang H, Liu A and Nee A Y C 2019 Digital twin in industry: state-of-the-art *IEEE Trans. Ind. Inform.*152405–15 (doi:

10.1109/TII.2018.2873186) [5] Lasi H, Fettke P, Kemper H-G, Feld T and Hoffmann M 2014 Industry 4.0 *Bus. Inf. Syst. Eng.*6239–42 (doi: 10.1007/s12599-014- 0334-4) [6] Zheng P, Wang H, Sang Z, Zhong R Y, Liu Y, Liu C *et al.*2018 Smart manufacturing systems for Industry 4.0: conceptual framework, scenarios, and future perspectives *Front. Mech. Eng.*13137–50 (doi: 10.1007/s11465-018-0499-5) [7] NIST 2014 Smart Manufacturing Operations Planning and Control Program [Internet] Natl. Inst. Stand. Technol. Available from: https://www.nist.gov/programs-projects/smart-manufacturing-operations-planning-and-control-program [8] Sahoo S and Lo C-Y 2022 Smart manufacturing powered by recent technological advancements: a review *J. Manuf. Syst.*64236– 50 (doi: 10.1016/j.jmsy.2022.06.008) [9] Phuyal S, Bista D and Bista R 2020 Challenges, opportunities and future directions of smart manufacturing: a state-of-the-art review Sustain. Futures 2100023 (doi: 10.1016/j.sftr.2020.100023) [10] Lee J and Su H 2025 Rethinking industrial artificial intelligence: a unified foundation framework *Int. J. AI Mater. Des.*2

(2):56-68. (doi: 10.36922/ijamd025080006) [11] Zhang C, Wang Z, Zhou G, Chang F, Ma D, Jing Y *et al.*2023 Towards new-generation human-centric smart manufacturing in Industry 5.0: a systematic review *Adv. Eng. Inform.*57102121 (doi: 10.1016/j.aei.2023.102121) [12] Xu X, Lu Y, Vogel-Heuser B and Wang L 2021 Industry 4.0 and Industry 5.0—inception, conception and perception J. *Manuf. Syst.*61530–5 (doi: 10.1016/j.jmsy.2021.10.006)

[13] Ren L, Wang H, Dong J, Jia Z, Li S, Wang Y et al. 2025 Industrial foundation model *IEEE Trans. Cybern.*55 (5) 2286–2301 (doi: 10.1109/TCYB.2025.3527632) [14] Lee J and Su H 2025 Agentic AI for smart manufacturing *Manuf. Lett.*4692–96 (doi: 10.1016/j.mfglet.2025.10.013) [15] Nti I K, Adekoya A F, Weyori B A and Nyarko-Boateng O 2022 Applications of artificial intelligence in engineering and manufacturing: a systematic review *J. Intell. Manuf.*331581–601 (doi: 10.1007/s10845-021-01771-6) [16] Haricha K, Khiat A, Issaoui Y, Bahnasse A and Ouajji H 2023 Recent technological progress to empower smart manufacturing: review and potential guidelines IEEE Access 1177929–51 (doi: 10.1109/ACCESS.2023.3246029) [17] Peixoto T, Oliveira B, Oliveira Ó and Ribeiro F 2025 Data quality assessment in smart manufacturing: a review Systems 13 243 (doi: 10.3390/systems13040243) [18] Puthanveettil Madathil A, Luo X, Liu Q, Walker C, Madarkar R and Qin Y 2025 A review of explainable artificial intelligence in smart manufacturing *Int. J. Prod. Res.*1–44 (doi: 10.1080/00207543.2025.2513574) [19] Wang J, Li Y, Gao R X and Zhang F 2022 Hybrid physics-based and data-driven models for smart manufacturing: modelling, simulation, and explainability J. Manuf. Syst.63381–91 (doi: 10.1016/j.jmsy.2022.04.004) [20] Lee J, Su H, Ji D-Y, Minami T.**2025. Engineering artificial intelligence: framework, challenges, and future direction. Mach. *Learn.: Eng.*1(1):013001. (doi: 10.1088/3049-4761/adce0d)