ABSTRACT Multi-variety and small-batch production is widely used in modern manufacturing. Monitoring the machining process in such systems is crucial for improving efficiency and reducing costs. However, most existing monitoring methods rely on intrusive sensors which often require equipment modifications and increase costs. Additionally, many methods can only determine the machining state after all operations on a machine are completed, making real-time monitoring difficult. The small sample size characteristic also limits the effectiveness of conventional machine learning approaches.Therefore, a real-time machining state monitoring method based on transfer learning and power signal is proposed. The input power of the machine tool is used as the original data, whose acquisition is non-intrusive. The power signal is segmented using the Bayesian online change point detection algorithm and converted into recurrence plots for state monitoring model. A transfer learning–based model is then developed to identify machining steps and processes using only a small amount of training data. Power signal features are extracted using frequency domain analysis and recurrence quantification analysis. Finally, machining anomalies are detected using a Z-score variant algorithm. Case studies show the proposed method achieves a monitoring accuracy of 97.1%, enabling effective real-time state monitoring in multi-variety and small-batch production systems.

ABSTRACT This study proposes a digital twin–enabled hybrid scheduling framework that integrates mixed integer linear programming (MILP) and genetic algorithms (GA) for efficient scheduling in multi-product cellular manufacturing systems. Setup time during product changeovers directly affects production efficiency, yet existing rule-based methods and single-algorithm approaches fall short of capturing the complexity of actual production environments. The proposed framework embeds a hybrid MILP–GA within a digital twin, where simulation serves a dual role: evaluating candidate schedules during GA evolution and validating the final optimized schedule. By combining the extensive solution space exploration capability of GA with the precise optimization ability of MILP, the framework addresses the single-machine scheduling problem with sequence-dependent setup times and achieves both makespan minimization and setup time reduction. Experimental results demonstrated that the proposed framework reduced makespan by up to 4.3% and setup time by up to 9.4% relative to the current system. This research contributes a practical and adaptable scheduling solution for small-batch multi-product cellular manufacturing systems, highlighting the value of integrating digital twin technology with hybrid optimization in smart manufacturing environments.

ABSTRACT Digital Servitization is the transformation process of replacing classical production with the digitally empowered and motorized services with the advent of Industry 4.0 (4IR) technologies. It is altering industrial ecosystems with digital technologies. Nevertheless, even though academic interest in the field has been increasing, no conceptual elaboration on this specific subject has been carried out and there is a lack of empirically encompassing depth. This paper fills these gaps by performing a systematic literature review based on use-cases. It evaluates 70 industrial cases, derived out of 35 peer-reviewed articles published between 2010 and 2025. Thematic analysis provides five prevailing clusters. The identified themes are reflective of the path of digital Servitization towards predictive maintenance, lifecycle intelligence, adaptive service models, and value creation amongst manufacturing industries. The article presents an innovative integrative model, Smart Sustainable Servitization Model (SSSM) that connects the facilitators of 4IRs with the multidimensional sustainability outcomes and PSS typologies (product, use, and result-oriented). The review is well structured and introduces relevant concepts in a road map that enhances theory as well as providing practical guidance. It has the potential to enlighten the scholars and practitioners planning to engage in the practice of sustainable digital Servitization to smart manufacturing.

ABSTRACT Metrology, the science of measurement, has significantly evolved to meet the demands of modern industries. This paper explores the historical significance, technological advancements, and contemporary applications of metrology, highlighting key technologies like laser trackers, photogrammetry, interferometers, optical fibres, coordinate measuring machines (CMM), coordinate measuring Arms (CMA), electromagnetic distance measurement (EDM), and digital metrology tools. Each technology’s unique capabilities, such as high precision, long-range measurement, and data integration, are compared, emphasizing their roles in fields like aerospace and manufacturing. The comparative analysis provides insights into the advantages and limitations of these technologies, offering guidance for selecting appropriate metrology solutions. Additionally, the potential future advancements in metrology, driven by innovations in sensor technology, artificial intelligence, and digital twins, are discussed. This paper accentuates metrology’s pivotal role in enhancing measurement accuracy, reliability, and efficiency in various engineering and manufacturing processes.

ABSTRACT Virtual-real fusion is a typical feature of digital twin technology, and the geometric scenes of digital twins are mainly achieved through geometric modelling techniques. To solve the problem of insufficient accuracy and low automation level in scene geometric modelling. A high-precision simulation scene generation framework based on calibration balls is proposed to guide geometric modelling. We propose a point cloud segmentation model, Small Scale Segmentation Network (SSSNet), to improve segmentation accuracy, especially for calibration ball recognition in cluttered scenes. More specifically, a hybrid attention module is proposed for fusing multi-source features to guide the network in learning the correlation between features. In addition, a novel balloon simulation algorithm was proposed to achieve adaptive extraction of regions of interest. Finally, geometric feature descriptors are introduced to enrich features and reduce interference from neighbouring point clouds. The results indicate that the segmentation accuracy has been significantly improved compared to existing methods. Meanwhile, the proposed framework has more advantages in simulating scene generation problems.

ABSTRACT Configure-to-Order (CTO) assembly systems are becoming pivotal in meeting the rising demand for mass customization. However, enabling rapid and cost-effective system reconfiguration remains a major challenge in manufacturing. Despite the flexibility of general-purpose machines, fixture design introduces significant bottlenecks, often requiring extensive manual effort to access, organize, and interpret fragmented or incomplete data. This leads to inefficient design pipelines that can compromise product functionality and result in costly redesign iterations, even when using advanced Computer-Aided Fixture Design (CAFD) tools. Although the Model-Based Definition (MBD) paradigm offers a promising response to these information-related challenges, its adoption within the CAFD domain remains limited. This work aims to support the broader adoption of MBD in this field through a multi-level contribution. First, a structured literature review identifies the essential semantic information required in an MBD dataset tailored to CAFD. Second, this content is structured in a machine-readable form using the ISO standard STEP AP242. Third, an MBD-driven CAFD framework is proposed. Finally, a proof-of-concept CAFD tool is developed by integrating established methodologies with commercial software to automate non-value-added tasks, such as configuring Computer-Aided Technologies (CAx). The results demonstrate the potential of this approach to streamline CAFD processes and establish a fully connected digital thread.

ABSTRACT Digital technologies can completely transform the manufacturing supply chain and satisfy the dynamic requirements of the market. Developing countries like India face huge challenges around Industry 4.0 technology adoption. While academicians and researchers have focused on either standalone firms or a specific Industry 4.0 technology, a broader technology adoption across manufacturing supply chains remains relatively unexplored. This paper emphasizes on identifying and assessing the barriers of Industry 4.0 technology adoption across manufacturing supply chain. This paper idntified 12 significant barriers for digital transformation in the manufacturing supply chain. The global weight of every barrier and sub-barrier is analysed and evaluated using the SWARA approach and interpretive structural modeling. Dependent, Linkage, Driving, and Autonomous barriers are further categories for barriers. Government initiatives and industry-friendly regulations that address data security concerns and the availability of skilled talent for the effective execution of integrated technologies can accelerate digital transformation. The focus should first be placed on the following issues: lack of maturity and readiness of the partner ecosystem (PE), financial challenges (FC), lack of management support (MS), and lack of regulatory support and legal legislation (RS), lack of standardization for technical collaboration (STC), and lack of data security (DS). Furthermore, the SWARA analysis reveals significant barriers to data sharing, including lack of standardized protocols, lack of manager willingness, a culture that discourages innovation, and insufficient workforce knowledge and skills.

ABSTRACT Inspection systems play a critical role in ensuring product quality and maintaining a company’s reputation. However, current industrial systems are heavily dependent on human labor, which can lead to errors and inefficiencies. Although deep learning architectures have the potential to effectively automate inspection systems, their complexity often introduces computational challenges that can impede system speed. To address this issue, we propose a hybrid deep learning approach called Defect-Lite, which adapts the YOLOv5 object detection algorithm. YOLOv5 is renowned for its robustness in real-time defect identification in industrial environments. Our innovative method integrates the inverted residual layers of MobileNetV2 with the PANet architecture of YOLOv5, improving the speed of the algorithm and reducing its computational complexity. This improvement makes it feasible for real-time applications on lower-power machines. We evaluated the proposed model using custom real-time datasets designed to simulate actual industrial settings, running the model on various systems to demonstrate its adaptability and robustness. Experimental results indicate a precision of 0.743, recall of 0.66, mean average precision (MAP) of 0.739, and an F1 score of approximately 0.69 at 240 frames per second (FPS), based on our custom dataset containing 783 images across five classes of defects. These results significantly surpass the performance of the standard YOLOv5 model and other state-of-the-art approaches. Additionally, we deployed our model on a low-computation Raspberry Pi 4B microprocessor to assess its performance in a real-time scenario, where it successfully provided accurate bounding boxes for defects in live footage. Hence, the proposed defect inspection model is proven to be effective in real-world industrial settings.

ABSTRACT Manufacturing industries are under increased pressure to implement more sustainable practices to safeguard the environment and advance the welfare of humanity due to evolving ecological conditions. Industry 4.0 technologies are essential for transforming conventional manufacturing into sustainable practices by integrating Cyber-Physical Systems. Nevertheless, the manufacturing sector in developing countries has been notably slow to implement Industry 4.0, primarily due to the numerous associated risks. Specifically in the context of developing nations, the existing literature demonstrates a substantial gap in research that addresses these Industry 4.0 risks. Therefore, it is imperative to assess the potential risks linked with the adoption of Industry 4.0 and to examine viable alternatives or solutions. This paper attempts to identify and prioritize the most effective solutions for mitigating the risks associated with adopting Industry 4.0 in manufacturing industries in developing countries. Thirty-seven risks and eighteen solutions related to Industry 4.0 adoption have been considered for proposing the framework based on the Spherical fuzzy full consistency method (SF-FUCOM) and Spherical fuzzy multi-attributive border approximation area comparison (SF-MABAC) methods. The results emphasized several critical risks, including economic, social, environmental, technological, operational, and strategic, that are associated with adopting Industry 4.0. The proposed framework is a valuable reference aid for researchers, policymakers, and decision-makers, as it eases the implementation of Industry 4.0 in manufacturing. The present study also provides a clear path for developing nations to implement Industry 4.0 and assists them in prioritizing critical factors that are pertinent to their manufacturing industries.

ABSTRACT In a highly digitalized world, Digital Twin (DT) technology is becoming vital in manufacturing, especially in additive manufacturing. This research work presents cost-effective digital twin development and implementation for Fused Deposition Modeling (FDM) 3D printers. This developed system enhances real-time process monitoring, anomaly detection, and autonomous control through a logical approach. Sensor data that measures and tracks filament flow, vibrations, and nozzle position are processed in real-time for the detection of anomalies such as layer shifting, under- and over-extrusion, and excessive vibrations. Magnetic encoders enable monitoring the filament flow with 0.73% error, position sensors identify nozzle displacement to monitor layer shifts. The sensor data are stored in the Firebase and visualized in the Unity interface with a 500- 700 ms data lag. In contrast to traditional systems, this developed system works independently without the need for external hosts, providing a very low-cost, modular solution (less than $50) appropriate for small-scale applications. This research also addresses the gap in FDM 3D printers and sensor-based DT by demonstrating practical, real-time interventions for quality assurance in FDM processes.