Abstract Battery management systems (BMSs) are essential for accessing and managing battery performance information, with state of health (SOH) estimation providing insights into the battery’s life expectancy. Electrochemical impedance spectroscopy (EIS) is a non-destructive method for SOH assessment. However, collecting EIS data across diverse operating conditions and battery types is both time-intensive and costly, presenting challenges related to data distribution and heterogeneity. This work investigates a lightweight gradient-based fusion strategy to enable collective learning across independently trained models without sharing raw data. Specifically, the collective inference via gradient aggregation (CIGAR) algorithm is applied to multiple convolutional neural network–bidirectional long short-term memory models trained on disjoint EIS datasets. The approach is evaluated on a real-world SOH prediction task, demonstrating that gradient-based collective learning can facilitate knowledge exchange among models under compatible conditions. The results highlight both the potential and the limitations of CIGAR in heterogeneous battery scenarios, indicating that further optimisation and validation on larger and more diverse datasets are required to improve robustness and generalisation.

Collaborative Learning (CL) grows in relevance within Industry 4.0, as it facilitates knowledge exchange among networked participants and integrates learnings from models into local manufacturing processes. This allows organisations to efficiently harness information from distributed data, reduce costs, and accelerate innovation. Existing research focuses on conventional CL strategies, wherein model aggregation is performed on a central server. This centralised architecture, however, introduces inherent vulnerabilities, including a single point of failure and privacy risks. Recent advancements indicate a paradigm shift toward decentralised collaborative learning (DeCL) systems, yet no unified set of requirements has been proposed for supporting their development and implementation. This study addresses the aforementioned research gap by conducting a systematic literature review. Drawing on a synthesis and analysis of existing studies, this review identifies a comprehensive set of 23 system requirements, organised into six key categories, specifically focused on designing DeCL systems in manufacturing. Future research should align with Industry 5.0 principles by incorporating the human-centric design for ethical artificial intelligence and worker inclusion, as well as incorporating sustainable, resource-efficient technologies to develop environmentally responsible DeCL systems.

3835

Asset Administration Shell (AAS) serves as the digital representation of industrial assets, facilitating the creation and management of Industry 4.0 digital twins. In the context of integrating 5G into the industrial domain, AAS addresses the requirements of industrial applications and operational technology (OT) processes by providing an abstraction layer that enables factory operators to manage networks according to their needs. This seamless horizontal integration of 5G networks into OT processes lays the foundation for a holistic, end-to-end automation framework. Moreover, AAS offers valuable insights from diverse systems that enhance both network and asset management performance. In this paper, we detail the design, development, and implementation of 5G AAS, which encompasses both the 5G network and 5G user equipment components. We also demonstrate the practical application of our approach through the integration of the 5G-Industry Campus Europe and a corresponding use case.

With the growing reliance on deep learning (DL) models for tasks in computer vision, high-quality training data is essential for the development of accurate models. However, generating large, labeled datasets from real-world sources is time-consuming, laborious, and expensive. Synthetic datasets offer an alternative, allowing for the automated generation of large amounts of training data. Despite their advantages, synthetic datasets often fail to capture the complexity of real-world images, leading to a performance gap when DL models trained on synthetic data are applied to real-world tasks. To improve the performance of synthetically trained DL models, it is crucial to identify factors that contribute to the domain gap between synthetic and real-world datasets first. Yet, the high dimensionality of image data yields a challenge that impedes a direct comparison between image datasets. This work explores the utilization of the UMAP algorithm to assess the similarity between synthetic and real-world image datasets through their embedding and analysis in a low-dimensional space.

In the automotive context, the quality of products or assemblies is inspected regularly at quality gates, constituting a laborious and error-prone process if performed manually. Machine Vision (MV) systems can partially automate the inspection processes, relieving workers from such tasks and thus increasing the overall productivity. Although MV systems offer high potential in theory, their performance depends on large volumes of annotated data specific to the considered use cases. Yet, acquiring and annotating such data in industrial settings poses significant challenges due to accessibility constraints, confidentiality issues, high manual effort, and costs. As an alternative to the acquisition of real-world data, this work investigates the utilization of synthetic data for the training of deep learning (DL) models for defect detection, providing high-quality annotations by design and reducing manual effort significantly. The authors present a methodology for image generation and demonstrate that a DL model trained on synthetic data achieves comparable accuracy in object detection to one trained on real-world data. Moreover, the presented approach offers the capability to generate explicit error cases, providing valuable insights for error mitigation strategies in the context of automotive production and assembly processes.

The Job Shop Problem (JSP) is a well-known NP-hard problem with numerous applications in manufacturing and other fields.Efficient scheduling is critical for producing customized products in the manufacturing industry in time.Typically, the quality metrics of a schedule, such as the makespan, can only be assessed after all tasks have been assigned, leading to sparse reward signals when framing JSP as a reinforcement learning (RL) problem.Sparse rewards pose significant challenges for many RL algorithms, often resulting in slow learning behavior.Curiosity algorithms, which introduce intrinsic reward signals, have been shown to accelerate learning in environments with sparse rewards.In this study, we explored the effectiveness of the Intrinsic Curiosity Module (ICM) and Episodic Curiosity (EC) by benchmarking them against state-of-the-art methods.Our experiments demonstrate that the use of curiosity significantly increases the amount of states encountered by the RL agent.When the intrinsic and extrinsic reward signals are of comparable magnitude, the agent is with ICM module are able to escape local optima and discover better solutions.

Exploiting data and information is known to be essential for tapping into unrealized (business) potential. In the context of the Industrial Internet of Things (IIoT), concerns related to the sensitivity of data frequently hinder its sharing (across organizations). Despite this situation, universal approaches that account for and appropriately model the confidentiality needs of stakeholders are still missing. In this paper, we address this research gap by proposing ConfMod, a middleware that simplifies the fine-granular modeling of confidentiality requirements while striving for interoperability with other tools and standardization in the area. We evaluate ConfMod in a diverse set of twelve real-world use cases from industry and show its general feasibility. Hence, we are confident that the functionality and simplicity of ConfMod facilitate an important building block for the IIoT, which will fuel inter-organizational data sharing in the future.

A key problem in the adoption of exoskeletons in industry is that workers are incorrectly fitting the device, leading to discomfort and suboptimal functioning of the exoskeleton. Although biomechanical modeling and design optimization strategies have tried to resolve this issue, we propose a user-centric, real-time fitting aid to guide and control the correct fitting process, which has not been achieved in practice. Inspired by previous work that uses exergame-like feedback to instruct a user, we compared augmented reality (AR)-based visual cues to guide the accurate fitting of a passive upper limb exoskeleton. We selected visual cues through a systematic literature review and evaluated their efficacy and usability for different aspects of exoskeleton fitting in a study with sixteen participants. The study outcome suggests a statistically significant preference for a semi-transparent overlay instead of a more abstract arrow-based method. Moreover, the results indicate high usability and satisfaction with our approach, improved user acceptance, and potentially enhanced fitting accuracy. These findings advance understanding of the viability of exergame-like real-time guidance as a means to increase exoskeleton acceptance and adoption in industrial settings.

The adoption of Millimeter-Wave (mmWave) technology in industrial environments presents significant challenges in maintaining consistent Quality of Service (QoS) under dynamic and complex conditions. This study investigates the performance of a multi-cell mmWave network deployed in a large-scale industrial hall, emphasizing the mobility of end devices and their interaction with environmental factors. Measurements were conducted using a Non-Standalone (NSA) 5G network configuration with one sub-6 GHz anchor and two mmWave Radio Units (RUs) deployed for comprehensive coverage. The evaluation highlights key aspects such as Secondary Node (SN) changes, end device orientation, and network load conditions. Results from mobile measurements reveal the influence of device alignment and environmental changes on performance, transmission power and connectivity, particularly in SN change areas. Additionally, the impact of static load generation on multi-cell network performance is examined, demonstrating the interplay between mobility and network capacity. These findings underscore the importance of precise device orientation and environmental awareness in optimizing mmWave deployments. Index Terms-mmWave communications, indoor measurements, multi-cell, multi-user, mobility.

The fifth generation of mobile communications, 5G, has been introduced to various application domains, enabling significant progress towards networked and adaptive systems. As the development of 5G was specifically tailored towards the requirements of industry, a broad knowledge of the technology has been built up in industrial production while also contributing to increasing the spread of data-driven technologies like machine learning. Other application domains are still lacking the widespread use and adoption of digital and data-driven technologies. The objective of the research project 5G.NAMICO is to utilize the gained expertise and knowledge from industrial production and contribute to the adoption of 5G in the application domains of construction and underground mining. We set up trial sites on a construction site and in an underground mine to determine how a 5G network must be designed to meet the domain-specific requirements. Use cases were designed and implemented to verify the functionality as well as the benefits of the employed 5G networks. First network tests were conducted showing the potential of 5G to enable an end-to-end coverage, which provides the basis for the use of digital and data-driven technologies in the application domains of construction and underground mining.

Manufacturing process optimization is an open question, where Bayesian decision theoretic methods have shown considerable promise. One such is Bayesian optimization, with Gaussian Process (GP) surrogate model. This paper explores Gaussian Processes networks to jointly use parameter and observed state to predict the output(s) of a manufacturing process. The Gaussian process network that represents the paths from parameters to state-space to tasks, provides a methodology to ‘look inside’ the black-box of complex manufacturing processes. We present a comparative analysis of this method against the multi-task Gaussian processes and single-task counterparts, highlighting the benefits and drawbacks of each in modelling the behavior of such processes. We show the benefits of the proposed approach using numerical experiments. We show that we are able to improve the output prediction by additional sensor observations from inside the process at training time without needing those sensor observations for predicting product quality given the process parameters.

Abstract The advent of AlphaGo and its successors marked the beginning of a new paradigm in playing games using artificial intelligence. This was achieved by combining Monte Carlo tree search, a planning procedure, and deep learning. While the impact on the domain of games has been undeniable, it is less clear how useful similar approaches are in applications beyond games and how they need to be adapted from the original methodology. We perform a systematic literature review of peer-reviewed articles detailing the application of neural Monte Carlo tree search methods in domains other than games. Our goal is to systematically assess how such methods are structured in practice and if their success can be extended to other domains. We find applications in a variety of domains, many distinct ways of guiding the tree search using learned policy and value functions, and various training methods. Our review maps the current landscape of algorithms in the family of neural monte carlo tree search as they are applied to practical problems, which is a first step towards a more principled way of designing such algorithms for specific problems and their requirements.

The long-term utility and reusability of measurement data from production processes depend on the appropriate contextualization of the measured values. These requirements further mandate that modifications to the context need to be recorded. To be (re-)used at all, the data must be easily findable in the first place, which requires arbitrary filtering and searching routines. Following the FAIR guiding principles, fostering findable, accessible, interoperable and reusable (FAIR) data, in this paper, the FAIR Sensor Ecosystem is proposed, which provides a contextualization middleware based on a unified data metamodel. All information and relations which might change over time are versioned and associated with temporal validity intervals to enable full reconstruction of a system’s state at any point in time. A technical validation demonstrates the correctness of the FAIR Sensor Ecosystem, including its contextualization model and filtering techniques. State-of-the-art FAIRness assessment frameworks rate the proposed FAIR Sensor Ecosystem with an average FAIRness of 71%. The obtained rating can be considered remarkable, as deductions mainly result from the lack of fully appropriate FAIRness metrics and the absence of relevant community standards for the domain of the manufacturing industry.

5G New Radio (NR) non-public network deployments for industrial and enterprise applications are becoming highly popular in locally licensed and/or operator spectrum. The interference from coexisting networks on adjacent channels (in same or adjacent spectrum bands) could potentially deteriorate the performance characteristics in certain deployment scenarios. Appropriate interference mitigation is thus required to achieve the desired performance levels. In this paper, we present our detailed empirical results on the performance impact of coexisting 5G NR networks operating on adjacent channels. Our experimental study conducted on an industrial shopfloor reports the impact on the downlink and uplink latency and throughput when using the same and different Time Division Duplexing patterns for coexisting networks. Our empirical evaluation includes realistic user equipment deployment locations and traffic load conditions. We also present our results on different mitigation techniques to counter the adjacent channel interference effects.

Dynamic and flexible business relationships are expected to become more important in the future to accommodate specialized change requests or small-batch production. Today, buyers and sellers must disclose sensitive information on products upfront before the actual manufacturing. However, without a trust relation, this situation is precarious for the involved companies as they fear for their competitiveness. Related work overlooks this issue so far: existing approaches protect the information of a single party only, hindering dynamic and on-demand business relationships. To account for the corresponding research gap of inadequately privacy-protected information and to deal with companies without an established trust relation, we pursue the direction of innovative privacy-preserving purchase inquiries that seamlessly integrate into today’s established supplier management and procurement processes. Utilizing well-established building blocks from private computing, such as private set intersection and homomorphic encryption, we propose two designs with slightly different privacy and performance implications to securely realize purchase inquiries over the Internet. In particular, we allow buyers to consider more potential sellers without sharing sensitive information and relieve sellers of the burden of repeatedly preparing elaborate yet discarded offers. We demonstrate our approaches’ scalability using two real-world use cases from the domain of production technology. Overall, we present deployable designs that offer two-way privacy for purchase inquiries and, in turn, fill a gap that currently hinders establishing dynamic and flexible business relationships. In the future, we expect significantly increasing research activity in this overlooked area to address the needs of an evolving production landscape.

Autonomous Mobile Robots (AMRs) and mobile manipulators are becoming increasingly popular in industrial use cases, especially for flexible lineless assembly systems. These use cases require wireless communication with high reliability and jitter-minimized bounded latency. This paper describes an industrial use case, where edge-controlled AMRs collaboratively perform various screw installation tasks on a truck chassis. We integrate Time Sensitive Networking (TSN) features such as time synchronization and IEEE 802.1CB based Frame Replication and Elimination for Reliability (FRER) scheme with 5G wireless communication to realize this use case. Moreover, we apply hold and forward buffer (HFB) mechanism to minimize communication jitters. This paper highlights the performance benefits of the aforementioned TSN features for 5G communication in the edge-controlled AMR use case. Our empirical evaluation with over-the-air performance results obtained on an industrial shopfloor brings significant insights on using 5G for edge-controlled robotic use cases.

In this paper, we present a solution for simultaneous handling of large components with industrial robots performing synchronized motions with an AGV in flexible flow assembly. For this purpose, we implement an Extended Kalman Filter with a global localization system to track an AGV and multiple manipulators. We propose a model-predictive controller for force compliance and trajectory tracking in multi-robot cooperative, decentralized, and fast manipulation tasks. In order to show the effectiveness of our system, we assemble a truck windshield using two industrial robots and an AGV in motion. In our experiments, we reliably achieve assembly tolerances of 1.5mm at AGV velocities up to $400\frac{{{\text{mm}}}}{{\text{s}}}$. The presented system makes flexible assembly systems with AGVs and freely reconfigurable manipulators possible. It enables the automation of high variant, low volume, large size assembly tasks such as aircraft, truck or steel beam assembly, which are mostly manual processes at present.

Concepts such as Industry 4.0 and Industrial Internet of Things aim for a digital transformation of manufacturing companies. One aspect that is to be transformed is the shop floor, where the interconnection of machines and sensor devices plays a major role. 5G positions itself to be a key technology to enable this transformation by providing reliable, low latency, and high bandwidth communication which is required by industrial use cases. However, so far it is still unclear which role 5G will play in industrial networks and how it can be integrated into existing factory ecosystems. Therefore, this paper presents three different data flow architectures that illustrate how 5G can be integrated into the production IT and how it can coexist with a factory cloud system. Furthermore, we describe a use case that reflects typical industrial communication requirements and will be used in the future to evaluate the different architectures. We conclude by presenting a validation plan and an outlook on future work.

The estimated economic impact of 5G Technology in production industry is immense. Until 2030, worldwide production industry gross domestic product is expected to by up to $740 billion [4]. Providing low latencies, high data transmission rates and the possibility of operating many devices simultaneously in narrowly restricted radio cells, 5G is expected to meet the demands of networked production systems and has great potential to accelerate the ongoing digital transformation. Despite these prospected advantages, benefits of integrating 5G Technology in a production process can barely be quantified yet. 40 % of enterprises cite poor measurability of economic benefits of 5G for their specific processes as key concern [6]. Thus, improvement potential for production processes and their monetary benefit need to be quantified in order to provide decision-makers with a sound base for their investment decisions. This paper describes the requirements and a first approach of a model to quantify economic potential of 5G Technology in production. Therefore, existing approaches and models to quantify economic benefits of 5G Technology and of digitalization in production in general are analyzed. Then, the model is derived. In the end, future research needs are given.

To face the requirements of current and future industry and to push the digitalization of factories, an international consortium in the project 5G-SMART develops a versatile multi-sensor platform communicating via 5G. To achieve an adaptable and flexible system, the embedded device is designed in a modular approach, consisting of a local processing core, integrable field sensors and a 5G modem. This paper covers the concept and design of the elaborated versatile multi-sensor platform and therewith presents a system that overcomes the limitations of current sensor systems and enables an interconnected real-time monitoring for production industry.

Managing project risks is challenging for many enterprises, especially smaller ones, because they generally only have very limited method or tool support, i.e. basic, qualitative and rather short term reaction to the occurrence of risks. This results in higher vulnerability and reduced competitiveness. This paper proposes a risk management framework fitting SMEs needs by providing a way to adequately quantify risks and address them at two levels. First, an Analytical Hierarchy Process (AHP) is used to perform cost-benefit analyses of the possible mitigation actions assessed through Monte-Carlo simulations. Second, an on-line optimisation tool is used to make sure the planning is following the minimal risk path and reschedule mitigation action as soon as a risk has materialised. To address the limited SMEs resources, the core components are provided as Open Source with a clean application programming interface for easing integration with existing tools. A reference integration with the Open Source Redmine project management tool is also provided.

Since energy efficiency has obtained much attention from researchers and this situation will last for several decades, various simulation tools were developed to describe and predict the energy-consumption behavior of both production processes and building facilities in order to optimize energy efficiency. Physically based modelling (PBM) to describe thermodynamic interaction between production activities and building facilities is one of most utilized solutions, which involves a huge amount of analytical efforts and causes some difficulties in practice. This research work aims to develop a hybrid simulation tool to quantify the heating/cooling requirement considering different production and weather scenarios. Combined with a data-driven model (DDM) through the application of machine learning (ML) and a PBM based on thermodynamic interactions, this tool achieves a good accuracy as well as a manageable modeling effort. A minimal heating/cooling effort can be derived through using thermal impacts from production and atmosphere. Additionally, a comparison of various production plans under the same condition shows up to 40 % difference of energy consumption. This finding indicates that an adjustment of production planning can lead to a further energy saving from heating/cooling.

S.327-332

This article presents a proposal for a multi-agent system for controlling the production of printed circuit boards in small series based on a new multi-agent approach. The multi-agent system was designed following the methodology Prometheus AEOlus and implemented using the framework JaCaMo. Unlike other multi-agent system used to control the production of printed circuit boards that are based on the agent-oriented paradigm, this system was completely designed and implemented using an integrated platform that follows the multi-agent oriented paradigm: the framework JaCaMo. So far this is the only implementation of this approach in the small series production.

The dynamic conditions of global markets force manufacturers to invest in flexible production strategies to cope with demanding clients and still survive in a competitive economic scenario. In this sense, small series production appears as a trend for many manufacturing niches and brings many challenges regarding manufacturing and quality assurance aspects. Investing in production flexibility implies increasing production control complexity and planning. This flexibility usually does not correlate with higher degrees of manufacturing automation or with quality assurance strategies. The concept of Cognitive Metrology strives for handling the challenging automation and quality inspection requirements of small series production with a new approach based on self-optimizing systems. This paper introduces the concepts of self-optimization and Cognitive Metrology and focuses especially on a multiagent-based approach for supporting flexible automation and quality assurance in small series production, as a basis for the development of the Cognitive Metrology technology. Initial results of the application of this approach into industrial prototypes are introduced and discussed as well as the migration of this system to different industrial scenarios.

The performance of Large-Volume Metrology has substantially increased during the last number of years. Systems such as indoor global positioning system (iGPS) provide precise measurements in complete production environments and will eventually lead to the development of new manufacturing principles. The understanding of measurement uncertainty has always been a critical step in the integration of measurement systems in production lines. The laboratory of machine tools and production engineering WZL is addressing this important step by working with a focus on cooperating robot movements to provide a suitable application case. An initial performance evaluation of the iGPS system at WZL is undertaken. Theoretical simulation of the triangulation allows the development of a virtual iGPS. However, at this moment in time parameters are not fully known. Experimental results from a laser tracker for reference measurement are used to validate simulations. An overview of the primary use of the iGPS to control and calibrate robots will show the potential that iGPS possesses to be used in industrial applications, not being limited to robots only.