Industrial deployments increasingly rely on Open Platform Communications Unified Architecture (OPC UA) as a secure and platform-independent communication protocol, while private Fifth Generation (5G) networks provide low-latency and high-reliability connectivity for modern automation systems. However, their combination introduces new attack surfaces and traffic characteristics that remain insufficiently understood, particularly with respect to machine learning-based intrusion detection systems (ML-based IDS). This paper presents an experimental study on detecting cyberattacks against OPC UA applications operating over an operational private 5G network. Multiple attack scenarios are executed, and OPC UA traffic is captured and enriched with statistical flow-, packet-, and protocol-aware features. Several supervised ML models are trained and evaluated to distinguish benign and malicious traffic. The results demonstrate that the proposed ML-based IDS achieves high detection performance for a representative set of OPC UA-specific attack scenarios over an operational private 5G network.
Machine learning (ML) based anomaly detection has been thoroughly studied in the literature where different architectures and models have been developed for different industrial cyber-physical systems (CPS). However, given that most CPS control systems are deployed on embedded platforms with limited storage and computational resources, deployment and real-time functionality remain challenging. This paper applies a combination of TinyML techniques, Pruning, and quantization to reduce the computation cost and memory footprint of a transformer-based anomaly detection for an industrial combustion system without degrading the accuracy and classification performance. To assess reliability, we compare the reduced model's performance and accuracy to the original model and the state-of-the-art Cumulative Sum Control Chart (CUSUM) analytical model. The experiment results demonstrate that, while reducing the memory footprint by 70% and computation cost by 8%, the resulting model does not degrade the classification accuracy. Moreover, the reduced transformer model outperforms the CUSUM classification accuracy by up to 99%.
Today's manufacturing environments are challenged by the adoption of upcoming data-driven technologies, such as artificial intelligen0ce or digital twins. The bottleneck for digital transformation are mostly traditionally grew industrial hardware and software systems. The current pace of changes in technologies, requires a strong adaptive architecture backbone to address rapidly changing tools for manufacturing. This paper proposes a cloud native architecture for manufacturing, using Kubernetes as orchestration platform for industrial applications. Since strict requirements need to be met for control and monitoring, the proposed approach includes a layer model for multi-cluster deployments able to fulfill divers requirements. A holistic multicluster controller is presented, which distributes applications on the respective layers, based on their needs.
The ADACOR holonic architecture introduced an adaptive reconfigurable manufacturing control approach that promotes a balance between centralized and decentralized control, reducing the reliance on a pure central node and enhancing the robustness, responsiveness and fault tolerance. In the digital transformation era, it is crucial to understand the alignment of ADACOR with Industry 4.0 and Industry 5.0 principles. In this context, the paper discusses the alignment of ADACOR with the RAMI4.0 model, as well as with the human-centric and sustainability dimensions defined by Industry 5.0. Finally, the research challenges to develop an ADACOR 4.0 architecture fully aligned with these principles were also discussed.
Industrial Cyber-Physical Systems increasingly rely on Machine Learning models-particularly generative approaches such as Variational Autoencoders - for critical tasks such as anomaly detection and predictive maintenance. However, these models are vulnerable to adversarial attacks that can covertly mislead classification, posing severe risks in safetycritical domains. This paper introduces a novel Adversarial Sample Generation Algorithm (ASGA), specifically designed to craft adversarial samples for Variational Autoencoder-based anomaly detection systems used in industrial time-series applications. ASGA uses an optimization process that iteratively adjusts abnormal samples by applying small, targeted perturbations to minimize reconstruction errors, ensuring that anomalies are misclassified as nominal system functions. The proposed algorithm incorporates techniques such as gradient-based optimization, perturbation projection, and adaptive hyperparameter tuning to maintain stealthiness and stability. We evaluate ASGA on two industrial datasets, demonstrating consistent reductions in key anomaly detection metrics (accuracy, F1-score, ROC-AUC) and high attack success rate of 70 %. These results highlight the efficacy of ASGA and emphasize the urgent need for resilient anomaly detection mechanisms to safeguard ICPS against sophisticated adversarial threats.
Energy flexibility is critical for integrating renewable energy into power grids, as it helps mitigate supply volatility and enhance grid stability. This study presents a machine learning-based approach to predict energy flexibility, focusing on capacity and duration metrics. An ensemble model using support vector regression (SVR) was developed and trained on real-world energy consumption data from commercial buildings, incorporating weather variables. The results demonstrate strong predictive performance across diverse building types, and even on unsmoothed data, equipping aggregators with the tools to optimize energy market operations and support renewable energy integration.
Ensuring maritime safety in dynamic and uncertain environments requires robust and adaptive ship trajectory prediction systems. This study presents a self-adaptive framework that enables continuous, uncertainty-aware prediction for surface vessels. Leveraging the MAPLE-K concept (Monitor, Analyze, Plan, Legitimate, Execute, and Knowledge), the proposed approach ensures safe and reliable adaptation. Subject to varying environments, the prediction models undergo degradation, which is detected through predictive uncertainty analysis. To maintain uninterrupted decision support during model adaptation, a multimodel prediction system is introduced. Experimental results validate the system's resilience and accuracy, underscoring its capability to deliver safe and reliable trajectory predictions across diverse scenarios and enhance maritime operational safety.
Alarm floods present challenges to human operators in industrial process plants, potentially impairing their situational awareness. Alarm flood classification addresses this issue by assigning observed alarm sequences to known alarm flood classes. However, the dynamic nature of evolving alarm sequences can lead to temporary class ambiguities, where insufficient information at early stages makes it challenging to distinguish between alarm flood classes. Recent advancements integrated conformal prediction and alarm flood classification to address this challenge, providing dynamically updated sets of plausible alarm flood classes. Despite their promise, such methods require extensive calibration datasets, which may be scarce in industrial contexts. We propose a novel method leveraging data-efficient crossconformal prediction to reduce calibration data requirements. Additionally, a novel postprocessing step is introduced to resolve erroneously empty prediction sets. Evaluated on a synthetic alarm dataset, our approach is compared to an existing inductive conformal prediction method across three alarm flood classification methods from the literature. The results demonstrate our method's effectiveness in maintaining a high coverage and a relatively small set of plausible alarm flood classes, while reducing calibration demands compared to the existing method.
The increasing sophistication of cybersecurity threats in Operational Technology (OT) environments highlights the necessity for a lightweight and efficient anomaly detection framework to identify and mitigate these threats. This paper proposes a hybrid multivariate anomaly detection framework for OT networks using Leaky Bucket Algorithm (LBA) and Principal Component Analysis (PCA). The framework employs a multi-level approach, where LBA performs the initial anomaly detection, and the identified anomalies are forwarded to PCA for in-depth analysis and validation. The proposed framework's performance is evaluated against standard PCA in detecting various cyberattacks, including Denial-of-Service (DoS), Man-in-the-Middle (MiTM) and Command Injection Attacks (CIA) in an OT testbed. The experimental results demonstrate that the proposed framework outperformed the standard PCA in detecting various attacks using performance metrics such as accuracy, precision, recall, F1-score and false negative rate.
Control software in the Cyber-Physical Production System domain continuously evolves to meet customer needs, resulting in multiple software variants that require effective management. Developers often adopt ad hoc techniques such as Clone-and-Own to generate new product variants. These offer short-term benefits but pose long-term maintenance challenges. Such cloned product variants contain numerous artifacts with diverse associations, exhibiting commonalities as well as variability across different variants and versions. Various approaches exist to identify these commonalities and variability and extract reusable artifacts from product variants to support systematic reuse. However, existing approaches are typically limited to particular programming languages and have not been applied to control software. Also, there is a lack of tool support to analyze, visualize, and store variable artifacts across multiple product variants. This paper introduces a prototype dashboard tool to enhance the analysis, visualization, and representation of control software artifacts across multiple product variants. The tool provides the foundation for adopting a Software Product Line engineering approach for control software development.
As cyber-physical systems (CPS) are becoming increasingly complex, the need for robust requirements engineering becomes apparent. Model-based requirements engineering (MBRE) is an essential approach that can help to establish traceability between requirements and system components. However, current approaches, such as the SysML requirements diagram only support modeling the formalized requirements and fail to depict the process leading up to those requirements. The process of stakeholder analysis, requirements elicitation, including the identification of stakeholder needs, is not supported in current requirements modeling approaches. To close this gap, this research presents a domain-specific language (DSL), implemented as a Unified Modeling Language (UML) profile for Enterprise Architect (EA). The proposed DSL allows for modeling the entire RE process as it is defined by the International Requirements Engineering Board (IREB), from stakeholder analysis to system validation, ensuring traceability across all phases.
This paper explores the influence of charging speed and battery capacity of electric vehicles (EVs) on the energy autarky of residential areas utilizing Vehicle-toQuarter (V2Q) settings. A simulation model based on the FINE framework has been developed to analyze a residential area equipped with photovoltaic systems and varying EV configurations. The findings reveal that increased charging speeds and larger battery capacities have only little effect on system autarky. This implies that moderate charging infrastructure and battery sizes are sufficient for maximizing V2Q benefits, emphasizing the need to prioritize renewable energy generation over costly infrastructure enhancements. These insights aim to guide stakeholders in optimizing local energy concepts and enhancing grid integration.
This paper demonstrates the critical role of the Asset Administration Shell (AAS) as a standardized framework for data exchange within the industrial automation domain, particularly considering the discrete manufacturing roots of automation projects. To address the lack of integration and collaboration in the supply chain of automation projects, we propose a digitally assisted and efficient solution using the Factory-X project as a case study. This approach connects various tools from manufacturers and integrates them into the toolchains used by system integrators, mechanical engineers, and operators. We demonstrate how the proposed method significantly enhances productivity and the solution-finding process in automation engineering by facilitating the exchange and integration of automation component information between customers and suppliers through AAS.
Uncertainty estimation is an emerging field in trustworthy artificial intelligence for industrial cyber-physical systems (CPSs), as it ensures reliable detection of unprecedented situations, also known as out-of-distribution (OOD) samples. In this paper, we introduce Feature Collapse Mitigation (FCM), a lightweight method that utilizes a self-supervised reconstruction loss and Radial Basis Function neurons to provide predictive uncertainty. Unlike conventional approaches such as Monte Carlo (MC) Dropout, which require multiple forward passes, significantly increasing the computational cost, FCM offers efficient single forward-pass uncertainty estimation. The proposed approach reduces the intensity of the phenomenon of feature collapse in a self-supervised manner, which commonly degrades uncertainty estimates, without relying on prior knowledge of the in-domain (ID) data or OOD samples. Experiments on synthetic, benchmark and real world datasets demonstrate the potential and reliability of FCM in delivering reliable uncertainty estimation in various scenarios.
In industrial manufacturing, production machines should be reusable for different products to accommodate long machine lifecvcles and uncertainties like market fluctuation. For reusability, the control software must be modularized and reconfigurable, for which this paper builds upon previous work regarding dynamic fault handling. The controller executes assigned tasks autonomously while interfacing with the controlled hardware in real time. The present work extends this approach by reconfiguring real-time work distribution and data exchange across multiple PLCs (i.e., Figure 1). Thus, a production system can be physically rearranged while the controllers are configured accordingly without software changes. Additionally, real-time communication channels are established, ensuring deterministic communication times of multiple connected PLC devices. The Industry 4.0 consortium has proposed Capabilities, Skills, and Services to decouple the configuration and implementation levels. In compliance with that standard, the main contribution of this work is a concept comprising reconfigurable Skill implementations as described above, their abstract Capability descriptions, and Services accessible via a global coordinator. The implementation level (Skills) complies with IEC 61131-3 and is thus compatible with state-of-the-practice control software. The concept has been successfully validated on a lab-scale manufacturing demonstrator using multiple independent tasks, enabling online process reconfiguration.
The increasing complexity of automated production systems (aPS), driven by the demands for flexibility and productivity, has introduced intricate interdependencies across diverse engineering disciplines. These interdependencies are critical for system coordination, adaptability, and fault diagnosis but remain challenging to capture due to heterogeneous data sources, dynamic requirements, and extended system lifecycles. This paper presents an integrated approach combining formalized knowledge representation with data-driven analysis to achieve comprehensive and flexible modeling of information dependencies within the Product-Process-Resource (PPR) structure. By employing weighted knowledge graphs, the proposed approach incorporates qualitative dependencies derived from expert knowledge and quantitative correlations identified through datadriven techniques. The presented concept is implemented on an intralogistics demonstrator and evaluated through a fault scenario, demonstrating its ability to effectively model and analyze interdependencies within aPS.
The advancement of the Industry 4.0 technologies is transcending traditional boundaries within manufacturing systems, extending its reach into the construction engineering sector. The implementation of intelligent monitoring systems has proven effective in the maintenance of manufacturing machinery, with analogous applications in the monitoring of constructed bridges. Such systems have the potential to increase the safety and the longevity of ageing structures and to supplement the scarce workforce available for inspection and maintenance. The digital twin, a fundamental concept of Industry 4.0, has been identified as a key component in digitalisation by enabling interoperability between cyber-physical systems. The utilisation of digital twin technology in the construction engineering sector is likely to have a significant role in facilitating intelligent maintenance of bridges and roads. This paper presents the design of a digital twin architecture, with a focus on the monitoring of various components of a constructed bridge. The implementation of the digital twin architecture utilises the standardised Asset Administration Shell, which facilitates the integration of both static and dynamic data sources. This integration enables a consolidated analysis of all the structural and environmental data relevant for the monitoring and condition assessment of bridges.
The AAS and OPC UA are pivotal technologies for Industry 4.0, enabling asset data exchange across the value chain based on different lifecycle stages or system requirements. Mapping their meta-models is essential but challenging due to ongoing divergences. Manual mapping efforts have proven unsustainable over time. This work introduces a model-driven approach with simple rules to automate and maintain mappings. We implement this in the aas-core-works of generators, demonstrating consistent asset representation in both AAS and OPC UA. This approach guarantees better long-term maintainability, accelerates development, enhances interoperability, and flexibility of digital twins.
Digitalization and technological innovations in maritime logistics bring significant improvements in the efficiency and safety of seaborne transport. However, integrating Information and Communication Technology (ICT) into ships' operational systems introduces cybersecurity risks, removing the myth that vessels are isolated from cyber threats. Like other industrial control system environments, ships often have long service lifespans, with some still relying on automation systems installed decades ago. The integration of modern ICT solutions—such as remote maintenance tools and internet connectivity for crew welfare—into these legacy systems can expose them to various threats, some of which may pose physically dangerous. Given recent global crises, the maritime logistics and transport sector has become an increasingly attractive target for cyber-attacks. This paper explores the tailoring of the MISP Threat Sharing (formerly Malware Information Sharing Platform), an open-source threat intelligence platform, for the maritime domain. By leveraging MISP's capabilities, including standardized formats for Indicators of Compromise (IOCs), visualization tools, and automated data sharing, the maritime sector can enhance its proactive cybersecurity measures and overall resilience.
In this work, a continual learning problem for time-varying Li-ion battery systems is investigated based on a Physics-Informed Bayesian Neural Network (PI-BNN) approach. To build an interpretable dynamic model, a second-order battery equivalent circuit model is incorporated into a BNN, and a pretrained multi-layer perception approximating the battery nonlinear function is connected between the hidden and output layers of the PI-BNN. The network parameters of the PI-BNN are designed as the unknown resistance and capacitance variables of the Li-ion battery to learn probability distributions, and then each uncertain parameter is updated online using new measurement data, where the learning rate is proportional to the variance of the parameter probability distribution. By considering the uncertainty in the evolution of the posterior distribution, the uncertain model parameter is learned with improved model adaptability, and the output voltage is predicted based on the PI-BNN model during the lifelong battery monitoring. The effectiveness of the proposed approach is validated through cyclic charge and discharge simulations, which achieves accurate dynamic Li-ion battery modeling and output voltage prediction compared with a basic physics-informed neural network.
