We present a robust pipeline for reading sidewall text on vehicle tyres in industrial settings. Our system addresses the challenge of accurately interpreting low contrast, distorted and curved text. The architecture combines classical computer vision (Hough transforms, Larson–Sekanina filtering) with state-of-the-art deep learning models (YOLO11m-OBB for detection and the Decoder-only Transformers for Optical Character Recognition (DTrOCR)). On a diverse dataset of 1,200 tyre images, our modular approach achieves a text detection mean average precision of mAP0.5= 96.2% and mAP[0.5:0.95]= 85.6%, along with a character error rate (CER) as low as 0.05. These findings demonstrate a feasibility for industrial applications ranging from quality control at assembly lines to automotive service and maintenance.

In the era of digital transformation and increasing data-centric operations, efficient and secure management of the supply chain remains a critical challenge. This article identifies the research gap in leveraging emerging technologies to enhance data-sovereign collaboration in the supply chain for manufacturers. To address this, we introduce TwinFlow, a novel architecture designed to facilitate the sharing of material flow data and information among manufacturers in the supply chain, following the principles of IDS (International Data Spaces) and ecosystems like Gaia-X while applying the digital twin methodology. TwinFlow enables knowledge representation of in-plant logistics through ontology modeling and fosters collaboration among manufacturers and relevant stakeholders through a shared data ecosystem. The proof-of-concept implementation of the proposed TwinFlow architecture further validates its efficacy in managing in-plant logistics operations. This study paves the way for a data-sovereign, interoperable, and real-time monitoring-enabled approach to optimizing industrial material flow, contributing significantly to the discourse on digital transformation in supply chain management.

In light of the growth in data availability, the manufacturing industry is experiencing a growing change in its needs and shape, which necessitates the use of more efficient data-driven methodologies in real-time production scheduling. Conventional dynamic scheduling approaches, designed for non-dynamic problem sizes, struggle in dealing with the inherent volatility and complexity of contemporary production scheduling scenarios. As an attempt to meet the demands of near real-time decision-making mechanisms on the shop floor, this study explores variations of QMIX, including local QMIX (LQMIX), where separate instances are employed for specific tasks, and gated QMIX (GQMIX), which utilizes specific agents for tasks while employing a central mixing network. Scheduling systems may not satisfy the recent requirements, emphasizing the need for more adaptive systems. Utilizing QMIX, manufacturing operations can be streamlined by integrating the collaborative synergy of multiple agents. Reinforcement learning models are trained using QMIX and benchmarked against heuristic dispatch strategies such as Shortest Path First as one of the most popular method used in the community. The experimental findings highlight the effectiveness of QMIX, in particular the original version, in tackling the challenges of dynamic scheduling within the manufacturing domain. QMIX exhibits superior performance compared to alternative algorithms and heuristic dispatching methods in specific contexts. Nonetheless, the study underscores the imperative of striking a balance between adaptability and specialization.

For many years, universal robots have been exten-sively popular across a wide range of industries to accommodate a broad range of manufacturing requirements. The complexity of manufacturing environments has resulted in a growing demand for automated programming approaches for robotic systems, where artificial intelligence approaches have recently proven effective. Despite recent advancements, the development of AI-driven controllers for universal robots performing complex, multi-step tasks continues to face several challenges, including stability issues. This paper aims to address these challenges by leveraging reinforcement learning to automate robot programming, with a particular focus on ensuring the stability of robot movements during multi-step robot positioning for inspection purposes. Our research formulates the multi-step robot positioning task as a reinforcement learning problem and develops a reward function to account for the robots' stability at the checkpoints. We conducted experiments using four different RL methods, namely PPO, TRPO, SAC, and TD3. Our findings indicate that TRPO outperforms the other methods, converging to an optimal controller. This study contributes to the field of robotics by providing a robust approach to enhancing the stability and efficiency of robot programming in complex manufacturing environments.

The development of scalable industrial computer vision (ICV) services is still in an early stage. While ICV ecosystems are being developed on an application-specific basis, there is no standardised technical infrastructure for cross-company or cross-organizational applications. This paper presents a novel approach for federated ICV services, exemplified through the need for visual assembly inspection in the quality assurance process of passenger car production. The use cases include the automated training of computer vision models from synthetic data and the execution of the inspection service itself. A prototypical pipeline is designed to demonstrate the federated processing. The novelty of the approach is the application of the AAS (Asset Administration Shell) digital twin standard for the provisioning of service relevant data combined with the EDC (Eclipse Dataspace Connector) to ensure secure and self-sovereign data exchange.

OPC UA and the Asset Administration Shell cover neighboring domains in modern industrial information architectures. Since their meta-models differ fundamentally, correct integration is vital for sustained interoperability. This paper shows how enterprise information systems benefit from using transformed OPC UA information models as a proxy for shopfloor transparency. All transformation rules facilitating this meta-model mapping are motivated and specified. When implemented, a new Submodel mirroring an OPC UA Nodeset can be built up automatically carrying all necessary context from its origins into the realm of cross-company data exchange.

Object detection is a task in computer vision that involves detecting instances of visual objects of a particular class in digital images. Numerous computer vision tasks highly depend on object detection such as instance segmentation, image captioning and object tracking. A major purpose of object detection is to develop computational models that provide inputs crucial to computer vision applications. Convolutional Neural Networks (CNNs) have recently become popular due to their key roles in enabling object detection. However, the performance of CNNs is largely dependent upon the quality and quantity of training datasets, which are often difficult to obtain in real-world applications. In order to ensure the robustness of such models, it is vital that training instances are provided under various randomized conditions. These conditions are typically a combination of a variety of factors, including lighting conditions, object location, the presence of multiple objects in the scene, varieties of backgrounds, and the angle of the camera. In particular, companies, depending on their applications (such as fault detection, anomaly detection, condition monitoring, predictive quality and so on), require specialized models for their custom products and so always face difficulties in providing a large number of randomized conditioned instances of their objects. The primary reason is that the process of capturing randomized conditioned images of real objects is usually costly, time-consuming, and challenging in practice. Due to the efficiency gained so far via the use of synthetic data for training such systems, synthetic data has recently attracted considerable attention. This paper presents an end-to-end synthetic data generation method for building a robust object detection model for customized products using NVIDIA Omniverse and CNNs. In this paper, we demonstrate and evaluate our contribution to the modeling of chess pieces, where a total accuracy of 98.8 % was obtained.

Companies are facing increasing pressure to optimize their energy supply. On the one hand, rising procurement costs are a factor - driven by a shortage of raw materials and legal regulations aimed at minimizing greenhouse gas emissions (e.g., CO2pricing), with the current geopolitical situation also increasingly fuelling the pressure to act. On the other hand, the optimization of energy supply also serves the purpose of meeting the requirements of customers and society for sustainable production. One field of action that is still in its infancy in practical use is the exploitation of energy flexibility potential in discrete manufacturing.This is not least because tools for exploiting energy flexibility potentials (e.g., taking advantage of fluctuating prices on short-term electricity markets, providing system services on the balancing energy markets, or increasing the share of self-consumption from on-site generated renewable energy have so far tended to represent stand-alone solutions that have not yet been implemented in conventional ERP or MES systems.)This paper aims to provide an insight into potential business models for the use of energy flexibility in discrete production, to assess the market potential for this and to show why integration into existing systems across all levels of automation is essential.

The use of sensor technology in companies is steadily increasing and until now there has been a lack of a fundamental architecture for making heterogeneous actuators and sensors available to the consumer in a uniform manner. In this paper, we will propose an architecture which supports protocol independent automatic integration and configuration of such devices into a system of systems. Measured values will be stored as SenML documents in a MQTT broker. The used IoT automation framework Eclipse Arrowhead assists consumers in finding required devices by use of metadata description.

In order to guarantee the quality and the productivity of a production system in a competitive marketplace, it is important to be able anticipate the changes in specifications of products and systems. The time limits in running productions, the complexity of manufacturing systems, and the diversification of components, are the challenges that human experts cannot handle without cognitive systems. Capabilities of cognitive Systems in observing, learning, and predicting the behavior and the evolution of the manufacturing systems make them special candidates for solving these problems. This mapping study provides an insight into the application of cognitive systems in the domain of production. We categorize different approaches and estimate their progress. We also discuss the optimizations and persisting problems and barriers. These representations can help in recognizing the concrete problems of the field. According to the results of our mapping study, Human-Machine Interaction and Knowledge Gaining/Sharing represents the largest categories of the domain. A gain in efficiency and maximized effectiveness can be achieved as optimization. The most common problem is the missing or only difficult generalization of the presented concepts.

In the next few years, Blockchain will playa central role in IoT as a technology. It enables the traceability of processes between multiple parties independent of a central instance. Blockchain allows to make the processes more transparent, cheaper, and safer. This research paper was conducted as systematic literature search. Our aim is to understand current state of implementation in context of Blockchain Technology for digital protection of communication in industrial cyber-physical systems. We have extracted 28 primary papers from scientific databases and classified into different categories using visualizations. The results show that the focus in around 14% papers is on solution proposal and implementation of use cases Secure transfer of order data using Ethereum Blockchain, 7% papers applying Hyperledger Fabric and Multichain. The majority of research (around 43%) is focusing on solution development for supply chain and process traceability.

Being successful and competitive on the market means that companies have to adapt to the demands of their customers. Personalised products are increasingly becoming a matter of course for consumers, which leads to a reduction in the number of similar orders for manufacturing companies. To satisfy these requirements, new information and communication technologies are needed in industrial manufacturing. Industry 4.0 aims to address these challenges. However, many approaches are not yet implemented or mature, so there is a need for further research in this field. For this reason, a comprehensive and systematic mapping study was conducted to structure and categorize the current state of research in the field of self-describing and self-organizing manufacturing. The literature considered was published between January 2014 and May 2019. The research carried out is based on the guidelines for conducting systematic mapping studies. With regard to the technical implementation of this technology, a number of research questions are carefully defined. Based on these questions, data from different levels of information are extracted and analyzed for the considered papers. The study results show in which areas more work is required and where there are future research perspectives. Furthermore, this study can help to better understand the field of research and the research gaps identified.

For the reliable communication network design (RCND) problem unreliable links are available, each bearing several options which have different levels of reliability and varying costs. The goal is to find the most cost-effective communication network design that satisfies a predefined overall reliability constraint. This paper presents two new evolutionary algorithm (EA) approaches to solving the RCND problem: LaBORNet and BaBORNet. LaBORNet uses an encoding that represents the network topology as well as the used link options while repairing infeasible solutions using an additional repair heuristic (CURE). BaBORNet encodes only the network topology and determines the link options by using the repair heuristic CURE as a local search method. The experimental results show that the new EA approaches using repair heuristics outperform existing EA approaches from the literature using penalties for infeasible solutions. They also find better solutions for existing problems from the literature, as well as for new and larger test problems.