Offenburg University of Applied Sciences
Hochschulschriftenserver der Hochschule OffenburgNot a member yet
6641 research outputs found
Sort by
Asynchrones Proximal Policy Optimization
No full textDen Menschen in Spielen zu übertreffen war schon immer eines der zentralen Forschungsfelder von Künstlicher Intelligenz. Auch in der beliebtesten Sportart der Welt, Fußball, soll es 2050 so weit sein. Die RoboCup Federation verfolgt dieses Ziel bereits seit 1997 und veranstaltet zu diesem Zweck regelmäßig Turniere. Der amtierende Vizeweltmeister Team Magma von der Hochschule Offenburg leistet einen Beitrag auf diesem Weg. Mittels einer eigenen Reinforcement-Learning-Bibliothek mit dem Namen RL_X, erlernen simulierte Roboter das Fußballspielen. Um die Trainingszeit zu reduzieren, wird in dieser Arbeit eine Asynchrone Proximal Policy Optimization (PPO) entwickelt. Durch die Einführung eines Environment-Runners wurde die Lernzeit für das Erlernen des Kickens reduziert. Die Implementierung wird umfassend evaluiert, wobei Stärken sowie Schwächen aufgezeigt werden
Influence of Intra-Particle Concentration Gradients and Lithium Plating on the Thickness Change of a Lithium-Ion Pouch Cell
Get PDFLithium plating is a critical factor that limits the fast-charging capabilities of lithium-ion batteries. Here we investigate the contributions of lithium plating and intra-particle lithium concentration gradients to the rate-dependent thickness change observed experimentally in a commercial high-power lithium-ion pouch cell with graphite negative electrode and a blend of lithium nickel cobalt aluminum oxide and lithium cobalt oxide positive electrode. This cell exhibits a thickness “overshoot” during charging at high C-rates, which partially recovers during the constant-voltage (CV) phase. We utilize a previously-developed pseudo-three-dimensional (P3D) thermo-electro-mechanical model to simulate these effects and validate the model against experimental data. Our findings show that intra-particle lithium concentration gradients significantly contribute to the thickness change at moderate C-rates due to the nonlinear volume expansion of graphite. At higher C-rates, lithium plating becomes a major factor, contributing both reversibly and irreversibly to the thickness change. A model including these effects accurately captures experimentally-observed behavior, providing insights into the complex thermo-electro-mechanical interactions inside the electrodes. This understanding is crucial for optimizing in operando lithium plating detection methods relying on thickness or mechanical stress measurements
Dispositif et procédé de méthanisation biologique avec une colonne de gazéification et de dégazage reliée en circuit
No full textHochdynamische Regelung elektrischer Antriebe
No full textIm Buch wird zunächst die Erstellung von regelungstechnischen Modellen der am häufigsten eingesetzten elektrischen Maschinen wie der Asynchronkäfigläufermaschine, der permanentmagneterregten Synchronmaschine und der Synchronreluktanzmaschine behandelt. Außerdem befasst sich das Buch mit Modellen für die aktive Netzeinspeisung elektrischer Antriebe. Auf dieser Grundlage werden geeignete Strukturen für die Strom-, Drehzahl-, Lage- und Zwischenkreisspannungsregelung für diese Antriebe vorgestellt und die zugehörigen Parameter für ein hochdynamisches Regelverhalten berechnet
Data-driven machine learning model estimates efficiency gains from passive filters under variable loads
Get PDFAccurately estimating power loss reduction from passive filters before installation is challenging due to variable loads and power quality conditions across grid points. Existing studies rely on simulation or analytical models. These approaches often fail to capture real-world variability through data-driven methods. This gap limits effective, site-specific filter deployment decisions. We present a two-step machine learning approach to estimate energy efficiency gains from passive filters under variable conditions using high-resolution power analyzer data. Ridge Regression identifies key predictive variables, achieving baseline R² = 0.591. XGBoost then captures nonlinear interactions between load variability, power quality disturbances, and filter performance, improving accuracy to R² = 0.755. The methodology was validated through deployment at three industrial facilities in collaboration with Livarsa GmbH. Results demonstrate 9.9% average relative error across measured efficiency gains, confirming reliability under real-world conditions. Comprehensive validation through k-fold cross-validation, ensemble methods, and external testing quantified prediction uncertainty inherent in small industrial datasets (25 training samples). The approach offers a scalable, data-driven decision-support tool overcoming simulation-based limitations. Computational efficiency enables real-time assessment during client consultations without specialized software. Economic value derives from reduced performance guarantee margins, accelerated assessment timelines, and minimized warranty exposure. Limitations include statistical constraints from limited training data, reflected in cross-validation overfitting and wide confidence intervals. External validity requires site-specific validation for facilities with substantially different electrical characteristics. Despite these constraints, the findings provide practical value for energy professionals seeking efficient power quality solutions, enabling confident passive filter deployment decisions based on quantified performance predictions
Marker-less motion capture - Differences to conventional 3D gait analysis in a clinical setting
No full textMarker-less motion capture has become more and more important in recent times. AI driven processes opened up new possibilities and increased precision. Literature currently shows differences between the gold standard of marker-based (MB) and marker-less (ML) systems [1]. However, unimpaired subjects with typical gait patterns have been monitored in these studies. Hence it is not yet fully known, how this methodology work in populations with impaired gait. So how effective can a ML system clinically relevant data on gait deviations in people with gait disorders? To what extent is a ML system an alternative or complement to the gold standard of MB motion analysis in a clinical setting
Förderung von KI-Kompetenz – Lernen mit und über Chatbots in einem Making-Szenario
No full textKünstliche Intelligenz (KI) nimmt eine immer größere Rolle im Berufs- und Alltagsleben ein. Ein bedeutendes Anwendungsgebiet von KI sind Chatbots, die Menschen als digitale Assistenten bei diversen Aufgaben unterstützen können. Der vorliegende Beitrag stellt die Umsetzung eines Moduls zum Themenbereich Chatbots vor, das konzipiert wurde, um die anwendungsbezogene KI-Kompetenz in der Hochschullehre zu fördern. Das Didaktische Design des Moduls kombiniert passgenau die Wissensvermittlung zur Funktionsweise von Chatbots mit einer praktischen Umsetzung basierend auf dem Making-Konzept.Artificial intelligence (AI) plays an increasingly important role in professional and everyday life. Chatbots, which can support people as digital assistants in various tasks, are an important area of AI application. This paper presents the implementa-tion of a module on the topic of chatbots, which was designed to promote applica-tion-related AI skills in university teaching. The didactic design of the module per-fectly combines the transfer of knowledge about how chatbots work with a practical implementation based on the making concept
Additive Manufactured Capacitive Sensor for Slip Detection with Seamless Integration of Object Recognition and Differentiation in Robot Grippers
Get PDFIn this contribution, we present a novel additive manufactured capacitive sensor for slip detection with seamless integration of object recognition and differentiation in robot grippers. The sensor consists of a measuring electrode in combination with guarding and shielding electrodes to reduce edge effects and disruptive influences. The results show precise slip detection through linear sensor behavior. In addition, object detection and differentiation are possible
Additively Manufactured Multimaterial Cantilever for Electromagnetic Vibration Harvesting
No full textThis study presents a novel, additively manufactured, multimaterial vibration harvester that employs fused filament fabrication (FFF) and wire embedding. The device features a clamped-free cantilever design. We investigate the effects of the embedding height of the coil layer and the coil’s displacement within the layer on the cantilever’s eigenfrequency. These parameters enable tuning of the cantilever’s first resonance frequency within a range of 27 Hz without altering its dimensions. In addition, the embedded wire serves as a coil for energy harvesting. We determined the electrical power output under mechanical excitations of up to 3 g. Piezo jetting (PJ) is also explored as an alternative printing technique to fabricate a double-sided coil structure on the cantilever. Both printing techniques are discussed in detail
Additive Manufactured Sensitive Gripper Jaw for Extended Gripping Force Ranges in Robotics
No full textIn this contribution, we present a novel additive manufactured sensitive gripper jaw based on strain gauge technology for extended gripping force ranges in robotics up to 40 N using fused layer manufacturing (FLM). The present sensor design is based on an arbitrary number of bending beams arranged one above the other, based on a single-point load cell. With this multiple-bending-beam structure, it is possible to individually reduce the mechanical stresses while maintaining the same deflection of the gripper jaw and enabling parallel gripping of parts. By varying the number of bending beams, it is also possible to adapt the measurement range accordingly. The investigations focus on the general sensor design, the fabrication process, and the sensor behavior in terms of zero-point deviation, characteristic value deviation, linearity behavior, repeatability, and viscoelastic behavior. The results show that higher gripping force measurement ranges are possible with good sensor linearity of 0.14% of full scale (FS). The arising viscoelastic behavior can be attributed to the properties of plastics. Nevertheless, this sensor is suitable for gripping force measurement in most industrial robot applications