20005 research outputs found
Sort by
Data-driven fault localization in cyber-physical systems using dependency graphs and anomaly detection
No full textThe early and automatic detection of faulty behavior is essential for maintaining the reliability of a cyber-physical system. In this paper we describe a fault localization approach for such a highly complex distributed system, the optical synchronization system of the European X-ray free-electron laser. Using a dependency graph, we model the relationships between the components and the influences of environmental effects. After we first resolve linear long-term dependencies between dependent components with a correlation analysis, we then use an unsupervised fault detection pipeline consisting of statistical feature extraction and unsupervised anomaly detection to accurately identify anomalies and localize their origins in the system
dynRDF: Using deep contextual bandits to optimize position flooding in urban UAV networks
No full textAdvances in mechanical capabilities and mass manufacturing of Unmanned Aerial Vehicles (UAVs) are driving their application in various fields from precision agriculture to infrastructure monitoring and on-demand parcel delivery. Especially in urban areas it is projected that large amount of UAVs will inhabit the airspace. To facilitate the safe and reliable operation of large-scale urban UAV deployments, an Unmanned Aerial Traffic Management (UTM) system is required. Such a system needs to be aware of all movements within the airspace to control and monitor urban UAV operations. One way to realize this is the establishment of an ad-hoc network, which UAVs use for network-wide dissemination of their positions. Recently, Rate Decay Flooding (RDF) has been proposed as a tailor-made protocol to realize such a system. Although RDF has been proven to be efficient in supporting UTM applications in larger networks than ordinarily possible, much of its success relies on the proper selection of protocol parameters. In this work, we propose a reinforcement-learning framework that automatically adapts the configuration of RDF to its perceived environment. We utilize deep contextual bandits as a light-weight, but effective method to capture the non-linear relationship between the perceived environment and the achieved performance. We name this extension Dynamic Rate Decay Flooding (dynRDF). In a simulation study, we show that this solution is effective in finding optimal configurations for RDF for varying network sizes. To achieve this, only 2.7 % of all possible configurations had to be explored. Allowing dynRDF to also take the local UAV density into account, a performance gain of more than 12 % is achieved in a relevant composite metric capturing both the timely dissemination of position updates to nearby UAVs and reliable network-wide dissemination
Statistical correlation of 3D scanned weld geometry distributions and fatigue life for different welding methods
No full textThe relationship between local weld geometry and fatigue life has been extensively studied over the past decades, driven by the need to enhance structural integrity and optimize costs throughout a structure’s service life. While numerous studies have explored the influence of weld geometry on fatigue strength, the comparative effect of different welding methods under comparable weld geometry quality remains largely unexplored. Furthermore, the influence of local geometric variations for each welding method has not been systematically evaluated. This study explores a large dataset of laser-scanned butt welds, analyzing key geometric parameters. The dataset is categorized by welding method (laser-hybrid welding, submerged arc welding, and flux core arc welding), and statistical distributions are examined to assess variations in weld geometry and compliance with ISO 5817 quality groups. The characteristic fatigue life for each quality group is estimated. The correlation between geometric factor and fatigue life is evaluated through the residual analysis of stress-life curve linear fitting. According to the findings, different geometry features dominate depending on the welding method. The fracture location is strongly influenced by angular misalignment, while fatigue strength is better explained by quantile-based analysis of local geometry. These results provide a basis for future predictive modeling and quality assessment in welded structures
Robust LFSR-based scrambling to mitigate stencil attack on main memory
No full textMain memory plays a pivotal role in the storage of computational data in a wide range of applications, including highly sensitive assets such as banking transactions, cryptographic keys, and user credentials. However, memory systems remain vulnerable to advanced physical and side-channel attacks, including cold boot attacks that exploit residual data after power-down. To mitigate such risks, Intel’s DDR3 memory scrambler uses a Linear Feedback Shift Register (LFSR)-based stream cipher to obscure memory contents. Nevertheless, this mechanism has been shown to be susceptible to stencil attack, a cold boot technique that reconstructs the scrambling key by leveraging the linear and periodic nature of the keystream. This article proposes a novel, lightweight, and secure scrambling architecture based on a generic LFSR designed to enhance the security of DDR3 memory against cold boot attacks. The proposed generic LFSR-based mechanism eliminates differential keystream periodicity by introducing an address- and seed-dependent LFSR structure, thereby rendering differential key recovery techniques computationally infeasible. Furthermore, unlike traditional AES-based memory encryption that incurs high latency and area overhead, the proposed approach achieves comparable security guarantees with low hardware complexity and zero access latency. The hardware implementation results on the Xilinx VCU118 FPGA show that the proposed scheme consumes only 252 LUTs, 256 registers and 104 slices, comparable to the Intel DDR3 scrambler, while offering superior resilience against the cold boot, warm boot, and probing attacks. These results demonstrate the practicality of the proposed scheme for secure memory systems in resource-constrained environments
Towards data-driven multi-stage OPF
No full textThe operation of large-scale power systems is usually scheduled ahead via numerical optimization. However, this requires models of grid topology, line parameters, and bus specifications. Classic approaches first identify the network topology, i.e., the graph of interconnections and the associated impedances. The power generation schedules are computed by solving a multi-stage optimal power flow (OPF) problem built around the model. In this paper, we explore the prospect of data-driven approaches to multi-stage optimal power flow. Specifically, we leverage recent findings from systems and control to bypass the identification step and to construct the optimization problem directly from data. We illustrate the performance of our method on a 118-bus system and compare it with the classical identification-based approach
Towards certifiable autonomous local public transport on waterways
No full textAutonomous and remotely operated vessels are poised to transform maritime mobility, logistics, and research. This paper presents the MS Wavelab, a comprehensive research platform for intelligent, certifiable, and resilient autonomous surface vessel (ASV) operation in coastal and inland waterways. We introduce a layered navigation framework that integrates sensor fusion, machine learning, and rule-aware path planning to achieve robust situational awareness (SA) and collision avoidance (COLAV) in dynamic maritime environments.
To address the lack of domain-specific training data, we curate a high-quality dataset of over 50,000 annotated images from 10,000 hours of video footage collected in real-world operations. This supports a camera-LiDAR fusion pipeline for visual object detection and distance estimation using YOLO-based models. For connectivity, we develop a hybrid communication architecture that combines 5G cellular and Starlink LEO satellite networks. By applying supervised learning models for bandwidth and handover prediction, we achieve stable, low-latency communication essential for teleoperation and high-resolution media streaming.
Our system architecture adopts a certification-oriented design, extending type-approved integrated navigation systems with autonomous modules while aligning with emerging international regulatory frameworks such as the IMO MASS Code. Together, these components enable safe, real-time decision-making and remote control in constrained and variable environments like the Kiel Fjord. The MS Wavelab serves as a scalable, modular platform to advance the state of the art in autonomous maritime systems and accelerate their real-world deployment
Lernprozessbegleitung neu gedacht - Rolle und Einsatzmöglichkeiten von KI-Lernagenten in der Berufsausbildung
No full textIm Kontext tiefgreifender gesellschaftlicher Transformationsprozesse, die insbesondere durch Digitalisierung, Künstliche Intelligenz (KI) und den kontinuierlichen Wandel der Arbeitswelt geprägt sind, gewinnt der bereits vor mehreren Jahrzehnten entwickelte berufspädagogische Ansatz der Lernprozessbegleitung (vgl. u. a. Buschmeyer 2015, Bauer et al. 2007) neue Relevanz. Die Digitalisierung beruflicher Arbeitsprozesse führt nicht nur zu veränderten fachlichen Anforderungen, sondern beeinflusst die Lernprozesse: Lernende sehen sich zunehmend mit der Notwendigkeit konfrontiert, selbstgesteuert, digital vernetzt sowie flexibel auf neue und komplexe Herausforderungen zu reagieren. Diese Entwicklungen erfordern eine Neuausrichtung didaktischer Konzepte und stellen die Lernprozessbegleitung vor die Aufgabe, adaptives Lernen zu fördern und Lernende bei dem Erwerb einer umfassenden beruflichen Handlungskompetenz systematisch zu unterstützen. In diesem Verständnis zielt der begleitete Prozess der Kompetenzentwicklung darauf ab, dass das handelnde Subjekt eigenständig und flexibel über die relevanten Fähigkeiten und Kenntnisse – als „Dispositionen für selbstorganisiertes Handeln“ (Brater 2011, S. 5) – verfügt.
Mit Blick auf die rasante Entwicklung digitaler Technologien und die wachsende Bedeutung von KI-Anwendungen in der Berufsbildung (vgl. Seufert et al. 2021) geht dieser Beitrag der Frage nach, inwiefern der Einsatz von KI-Lernagenten Einfluss auf die bisherigen Maßstäbe und Zielbilder der Lernprozessbegleitung nehmen könnte. Zugespitzt gefragt: Steht das Berufsbildungspersonal angesichts der Integration von KI in Lern- und Arbeitsprozesse vor der Herausforderung, seine Rolle und Haltung in der Begleitung beruflicher Lernprozesse neu definieren bzw. erweitern zu müssen
Enabling the circularity of printed flexible plastic packaging : delevopment of a decision tool supporting packaging designers
No full textReassessment of the energetic value of lignocellulosic biomass in closed carbon cycles
No full textRelated to its energetic value, lignocellulosic biomass is typically characterized by its gross and net calorific value. However, while aiming for closed carbon cycles, this characterization does not reflect that organic matter is a carbon carrier providing “green” carbon urgently needed within a defossilized world, where material utilization of carbon for synthesizing carbon-rich materials is pursued. To address this, the assessment of biomass is expanded to include the work required to separate CO2 from the atmosphere, which is vital for reducing carbon into biopolymers through photosynthesis. A biogenic and a technical pathway are compared to highlight the advantages of biomass utilization. The analysis reveals that relying solely on the calorific value underestimates the true energetic value of biomass in closed carbon cycles. Organic matter is predestined to be used as a carbon source to provide carbon-based materials to maximize the utilization of nature’s inherent separation and reduction capabilities. Consequently, a reassessment of the energetic value of biomass is necessary within the context of a non-fossil economy
Enabling non-planar load oriented deposition of carbon fiber reinforced polymers by varying layer height
No full textA common research goal for printing carbon fiber reinforced polymers (CFRP) using fused filament fabrication (FFF) has been the deposition along load paths to fully utilize the potential of the highly anisotropic material. Yet, the state-of-the-art solutions for load oriented non-planar slicing and path planning for neat polymers involve the dynamic variation of layer height. This variation is not possible in a single layer for the most commonly used process variant for printing CFRP, towpreg extrusion, because of the fixed ratio of matrix to fiber. This problem can be solved by printing interlayers which roughly double the layer count, introduce weak points, decrease the fiber volume fraction (FVF), and increase manufacturing time. Continuous fiber coextrusion (CFC) offers a possible solution to this problem, as the amount of polymer co-matrix can be controlled. This is possible because of the pre-impregnation of the fiber material, which allows active feed of both fiber and co-matrix. This study aims to investigate the possibility of using continuous fiber coextrusion to dynamically vary layer height during the printing process to enable the load oriented non-planar printing of CFRP. To this end, the process is described, a custom control scheme is mathematically derived, and an experimental plan is presented. The experiments include the printing of coupons to evaluate the minimum and maximum layer heights and the possibility to vary the layer height dynamically. A pipe and a bracket are printed to establish the applicability to manufacturing real-life parts. Micrographs are taken to assess the void content and fiber distribution. Surface roughness is evaluated with white light interferometry. To evaluate the impact of layer height variation on stiffness and strength, a mechanical investigation is performed involving tensile and compressive tests. In conclusion of this study, the possibility of dynamic layer height variation to continuous fiber coextrusion can be confirmed and its application for load oriented non-planar printing is enabled