Digital twins (DTs) represent a transformative technology in manufacturing, facilitating significant advancements in monitoring, simulation, and optimization. This paper offers an extensive bibliographic review of AI-Based DT applications, categorized into three principal dimensions: operator, process, and product. The operator dimension focuses on enhancing safety and ergonomics through intelligent assistance, utilizing real-time monitoring and artificial intelligence, notably in human–robot collaboration contexts. The process application concerns itself with optimizing production flows, identifying bottlenecks, and dynamically reconfiguring systems through predictive models and real-time simulations. Lastly, the product dimension emphasizes the applications focused on the improvements in product design and quality, employing lifecycle and historical data to satisfy evolving market requirements. This categorization provides a structured framework for analyzing the specific capabilities and trends of DTs, while also identifying knowledge gaps in contemporary research. This review highlights the key challenges of technological interoperability, data integration, and high implementation costs while emphasizing how digital twins, supported by AI, can drive the transition toward sustainable, human-centered manufacturing systems in line with Industry 5.0. The findings provide valuable insights for advancing the state of the art and exploring future opportunities in digital twin applications.

The growing need for sustainable and flexible automation solutions has led to the exploration of transforming traditional industrial robots into collaborative robots (cobots). This paper presents a framework for the conversion of conventional industrial robots into safe, intelligent, and sustainable cobots, leveraging advancements in artificial intelligence and computer vision and the principles of the circular economy. The proposed modular framework contains key components such as visual perception, cognitive adaptability, safe human–robot interactions, and reinforcement learning-based decision-making. Our methodology includes a comprehensive analysis of safety standards (e.g., ISO/TS 15066), robot typologies suitable for retrofitting, and sustainability strategies, including remanufacturing and lifecycle extension. A multi-phase implementation approach is laid out for a theoretical design to contribute to the development of cost-effective and environmentally responsible robotic systems, offering a scalable solution for extending the usability and social acceptance of legacy robotic platforms in collaborative settings.

The performance of manufacturing operations relies heavily on the operators’ performance. When operators begin to exhibit signs of fatigue, both their individual performance and the overall performance of the manufacturing plant tend to decline. This research presents a methodology for analyzing fatigue in assembly operations, considering indicators such as the EAR (Eye Aspect Ratio) indicator, operator pose, and elapsed operating time. To facilitate the analysis, a dataset of assembly operations was generated and recorded from three different perspectives: frontal, lateral, and top views. The top view enables the analysis of the operator’s face and posture to identify hand positions. By labeling the actions in our dataset, we train a deep learning system to recognize the sequence of operator actions required to complete the operation. Additionally, we propose a model for determining the level of fatigue by processing multimodal information acquired from various sources, including eye blink rate, operator pose, and task duration during assembly operations.

In the context of Industry 5.0, Virtual Reality (VR) is transforming workforce training by enhancing skill acquisition and smart assistance to improve safety, among other things. This study presents a VR-based welding training simulator, developed in Unreal Engine 5 and deployed on Meta Quest 3, designed to teach and standardize soldering techniques through progressive training stations, interactive guidance, and real-time feedback. A mixed-methods evaluation, combining quantitative metrics (execution time, positioning errors, grip errors) and qualitative user feedback, demonstrated a significant reduction in execution time and elimination of errors in positioning and grip. Additionally, no errors were made in real-world soldering post-training, confirming a high transfer of learning and reduced variability in performance among users. These findings highlight VR as a viable and scalable alternative to traditional training, ensuring process consistency, reduced learning time, and increased training efficiency. Future work will explore enhanced haptic feedback and the integration of generative artificial intelligence to improve real-time analysis.