The evolution of Artificial Intelligence and Machine Learning for Smart Manufacturing
Marco Macchi¹ and Adalberto Polenghi¹
1 Department of Management, Economics and Industrial Engineering, Politecnico di Milano, Milan, Italy
E-mail: marco.macchi@polimi.it, adalberto.polenghi@polimi.it
Status
The advances in various cutting-edge technologies and the opportunities for transformation emerging in the industrial environment are today placing the power of Artificial Intelligence (AI) applied to industrial processes at the top of the research agenda. A new term, Industrial AI, has been recently coined [1,2]; this was introduced to emphasize that, based on AI as technological foundation, on data and algorithms, and on software and hardware components, it is now imperative to bring AI to work in industrial systems through scalable applications with sustainable performance [3]. As we know, AI is not the only driver. The push forward is leading to an evolution towards a new paradigm envisaged by smart manufacturing [4,5]. The transition is supported by Industrial Internet of Things (IIoT), virtual manufacturing, Industrial AI and other enablers. The cloud-to-edge continuum ultimately supports the proximity of physical and virtual spaces in the deployment of computational intelligence [6]. In a broader perspective, the physical-digital convergence is a long-lasting development, fostered by the adoption of Cyber-Physical Systems (CPS), but not only. It is an effect resulting from the development of multiple technologies due to computer science, information and communication technologies (ICT), manufacturing science and technology, finally leading to the convergence between the physical and virtual
worlds [7,8]. As an aggregate impact, the cyber-physical integration problem should be addressed, and Industrial AI is an integral part of this problem. Looking at the virtual space, one can even envision a synergistic development for the coming years where AI will be a key component of Digital Twins of physical entities and systems in manufacturing, thanks to its capability to provide insights aimed at identifying hidden patterns and establishing correlations, and making predictions and optimizations of the future behaviour [9,10]. Then, a foreseeable scenario is to proceed towards a convergence where the Industrial Metaverse (IM), employing advanced technologies such as IIoT, Blockchain and Augmented/Virtual Reality (A/VR), will enable the construction of an immersive virtual space able to seamlessly interact with the physical space, facilitating human interaction in an advanced collaborative manufacturing [11,12] where Human-To-X collaboration is central. In this evolutionary trend, AI will play an essential role, both in industrial processes and in the relationship with humans within the decision-making loop [13]. It will support knowledge representation, machine/deep learning, reasoning and optimal problem-solving, thus integrating the advanced modelling and simulation technologies at the core of Digital Twins and Industrial Metaverse.
Current and future challenges
Current challenges in AI/ML are long-lasting to be addressed so that R&D activities lead to the provision of industrially ready solutions in the context of large-scale implementation of AI and machine learning (ML) in smart manufacturing. The integration of AI in smart manufacturing is not straightforward: technological and conceptual developments are required to make those solutions effective and scalable so that companies can leverage them to improve and achieve sustainable performance. Identified challenges that are timely and relevant to face come from proper mix of research and industrial experiences and are hereafter synthetised:
- Data-driven approaches showed limited impacts in terms of adoption in manufacturing companies. Knowledge of industrial processes is essential and must complement data-driven approaches. Indeed, domain-specific knowledge is necessary in order to rapidly develop capabilities in new tasks for new technologies and products as well as manufacturing processes and equipment [14].
- AI and ML are mainly limited to reaction to local and confined drifts and anomalies [15], but AI and ML need to scale up to handle complex systems, predicting and optimizing their behaviours globally, and should be challenged by the increased responsiveness to adapt to changes arising from new products, equipment and technologies, processes.
- Most AI research and development focus on technical performance of the model/solution without tackling the way in which the solution is embedded in a complex socio-technical environment as, for example, manufacturing shopfloors are, where only a multi-disciplinary approach can work out. However, so far, AI state of art shows that there are very few examples of AI-powered solutions that embrace such new research paradigm [16].
- Cognitive adaptation of manufacturing systems, implying autonomous execution of actions based on certain inputs and triggers from the system, is currently a look-ahead in research. It requires the collection and elaboration of data related to the system and to the context, properly elaborated so that Digital Twins and, generally speaking, AI-powered solutions, may be agent rather than pure informative systems for human-based decisions and actions [17]. Solving these challenges will then open future ones connected to the way machines, humans and AI will interact. The physical-digital convergence will lead to shadowed shopfloors in which what is “real” is merged between physical and virtual inputs and the Industrial Metaverse will be new way in which manufacturing companies should work from the design to the management of manufacturing systems.
Advances in science and technology to meet challenges
AI and machine learning are expected to enhance the Digital Twins of manufacturing systems. This will lead to synergies that will enable decision intelligence to grow towards higher levels of adaptability, intelligence and cognitive traits [17]. To support this growth, the AI-powered Digital Twins of manufacturing systems will be enriched by capabilities built on the adoption of behavioural models of human operators, continuous and reciprocal learning between humans and AI/machine learning models, human experience between virtual and real worlds, and augmentation of decisions through an increasingly cognitive collaboration between physical systems and human decision-makers [17,18]. In this path, technological advances are drastically increasing the capabilities to improve manufacturing operations. Regarding AI and ML, the hype is today focused on no-code AI and vibe-coding, which are making easier the development of advanced solutions; it is anyway relevant to remind that industry and business-grade AI-powered applications require heavy computer engineering and ICT expertise. AI itself is seeing an empowerment in terms of explainability capabilities so as to better engage with humans, providing not only results but the reason why such results has been obtained. Finally, LLMs (Large Language Models) and co- pilots, and underneath foundation models, are completely reshaping the way in which humans and machines interact, leading to new forms of HMI (Human Machine Interface) and dashboarding, that is more natural for human decision-makers; this interaction mode will be even more fruitful if complemented by immersive XR (Extended Reality) technologies towards full Industrial Metaverse realisation. Besides technological growth, new advancements also concern frameworks and theories to first assess and then introduce and scale-up AI-powered Digital Twins within manufacturing organisations considering human decision-makers as core actor [19]. Approaches such as systems engineering and MBSE (Model Based System Engineering) are thus important to conceptualise the relation between AI-powered Digital Twins with other technologies and human as well non-human agents. Furthermore, theories like SUT (System Usability Theory) or UTAUT (Unified Theory of Acceptance and Use of Technology) must be considered and managed within the scope of AI engineering and deployment. To this end, it is advisable for researchers and practitioners to work synergistically so that the solution is first defined within a use case and then checked for fit against business/economics, operational performance and human behaviour [20]. The vision for the future is outlined in the following Figure 1. To achieve it, a manufacturing company follows a path toward smart manufacturing that starts with existing manufacturing plants and systems. AI is a key pillar for advances in decision intelligence and is integrated into an evolving platform resulting from the combination of different technology stacks, both due to legacy IT systems and manufacturing equipment, and new equipment and tools, also those designed to support humans in IT/OT systems. Therefore, physical-digital convergence is envisioned in a future Industrial Metaverse as a natural trend originating in the IIoT and evolving through the development of digital twins of machines, humans and manufacturing systems and their XR extension. In this framework, AI plays a key role for the intelligence in terms of perception, learning, prediction, interaction, adaptation, reasoning and creativity.
Figure 1- Look-ahead for smart manufacturing systems
Concluding remarks
The way in which AI is permeating manufacturing companies is continuously evolving especially pushed by technological advancements. AI-powered Digital Twins are now a foreseeable reality in the manufacturing realm and will support humans, from operators to engineers and managers, in design, production and maintenance of products and manufacturing systems. Most of the development as of now is one-way, from AI to human, but for the former to become a critical part in manufacturing, adaptability is necessary as key capability to look for autonomous systems able to manage and react to non-trivial, context-dependent events. Therefore, AI-powered solutions become crucial resources with which humans can interact, being mutually informed to empower and strengthen decision-making. What is envisioned is that current AI state of art must move forward by disrupting the current human-machine-AI communication means towards a seamless convergence of physical and virtual worlds into the Industrial Metaverse concept. A blend of technologies is necessary for this, from IIoT to XR. New HMI based on LLMs will be the new norm and AI agents with diversified capabilities will be actors in CPS-based smart factories where humans can equally interact between themselves with machines and intangible solutions with their own learning capabilities., leading to different forms of Human-to-X collaboration. This will bring collaboration to a new frontier in industrial engineering only if this will be optimally developed and orchestrated with human-focused methodologies, considering the human-in-the-loop, and combining traditional engineering performance evaluation with aspects such as learnedness, usability and ergonomics.
Acknowledgements
This research is part of the HumanTech Project, which is financed by the Italian Ministry of University and Research (MUR) for the 2023-2027 period as part of the ministerial initiative “Departments of Excellence”
(L. 232/2016). The initiative rewards departments that stand out for the quality of the research produced and funds specific development projects.
References
[1] Peres R S, Jia X, Lee J, Sun K, Colombo A W and Barata J 2020 Industrial Artificial Intelligence in Industry 4.0 - Systematic Review, Challenges and Outlook IEEE Access 8220121–39 [2] Lee J, Davari H, Singh J and Pandhare V 2018 Industrial Artificial Intelligence for industry 4.0-based manufacturing systems Manufacturing Letters 1820–3 [3] Lee J 2020 Industrial AI: Applications with Sustainable Performance (Singapore: Springer Singapore) [4] Kusiak A 2018 Smart manufacturing null 56508–17 [5] Tao F, Qi Q, Liu A and Kusiak A 2018 Data-driven smart manufacturing Journal of Manufacturing Systems 48157–69 [6] Anon Dynamic Edge and Cloud Service Integration for Industrial IoT and Production Monitoring Applications of Industrial Cyber-Physical Systems | IEEE Journals & Magazine | IEEE Xplore [7] Monostori L, Kádár B, Bauernhansl T, Kondoh S, Kumara S, Reinhart G, Sauer O, Schuh G, Sihn W and Ueda K 2016 Cyber-physical systems in manufacturing CIRP Annals 65621–41 [8] Weichhart G, Panetto H and Molina A 2021 Interoperability in the cyber-physical manufacturing enterprise Annual Reviews in Control 51346–56 [9] Semeraro C, Lezoche M, Panetto H and Dassisti M 2021 Digital twin paradigm: A systematic literature review Computers in Industry 130103469 [10] Negri E, Pandhare V, Cattaneo L, Singh J, Macchi M and Lee J 2021 Field-synchronized Digital Twin framework for production scheduling with uncertainty Journal of Intelligent Manufacturing 321207–28 [11] Yun H and Jun M B G 2022 Immersive and interactive cyber-physical system (I2CPS) and virtual reality interface for human involved robotic manufacturing Journal of Manufacturing Systems 62234–48 [12] Tea S, Panuwatwanich K, Ruthankoon R and Kaewmoracharoen M 2021 Multiuser immersive virtual reality application for real-time remote collaboration to enhance design review process in the social distancing era Journal of Engineering, Design and Technology 20281–98 [13] Jwo J-S, Lin C-S and Lee C-H 2021 Smart technology–driven aspects for human-in-the-loop smart manufacturing Int J Adv Manuf Technol 1141741–52 [14] Yang C, Yu H, Zheng Y, Feng L, Ala-Laurinaho R and Tammi K 2025 A digital twin-driven industrial context-aware system: A case study of overhead crane operation Journal of Manufacturing Systems 78394–409 [15] Mypati O, Mukherjee A, Mishra D, Pal S K, Chakrabarti P P and Pal A 2023 A critical review on applications of artificial intelligence in manufacturing Artif Intell Rev 56661–768 [16] 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 [17] ElMaraghy H and ElMaraghy W 2022 Adaptive Cognitive Manufacturing System (ACMS) – a new paradigm International Journal of Production Research 607436–49 [18] Huynh-The T, Pham Q-V, Pham X-Q, Nguyen T T, Han Z and Kim D-S 2023 Artificial intelligence for the metaverse: A survey Engineering Applications of Artificial Intelligence 117105581 [19] Reiman A, Kaivo-oja J, Parviainen E, Takala E-P and Lauraeus T 2021 Human factors and ergonomics in manufacturing in the industry 4.0 context – A scoping review Technology in Society 65101572 [20] Maghazei O, Lewis M A and Netland T H 2022 Emerging technologies and the use case: A multi-year study of drone adoption Journal of Operations Management 68560–91