Promotionszentrum
Refine
Document Type
- conference proceeding (article) (29)
- Article (21)
- conference proceeding (presentation) (6)
- Preprint (5)
- Report (2)
- Other (1)
Publication reviewed
- peer review (40)
- nicht begutachtet (13)
- begutachtet (11)
Keywords
- SOTIF (5)
- Automated Driving System (3)
- Cybersecurity Management System (3)
- Functional Safety (3)
- HADF (3)
- Highly Automated Driving (3)
- Large Language Model (3)
- resource efficiency (3)
- Artificial Intelligence (2)
- Condition Monitoring (2)
Institute
- Promotionszentrum (64)
- Fakultät Maschinenbau (29)
- Fakultät Informatik (27)
- IFM - Institut für Fahrerassistenz und vernetzte Mobilität (23)
- EPT – Institut für Effiziente Produktionstechnik (16)
- IDF – Institut für Datenoptimierte Fertigung (9)
- IMS - Institut für Maschinelles Sehen (7)
- IEAT - Institut für Energie- und Antriebstechnik (4)
- Fakultät Elektrotechnik (2)
- IKR – Institut für angewandte Künstliche Intelligenz und Robotik (2)
This study investigates the influence of data distribution and image quality on the performance of fully supervised deep neural networks (DNNs) from the YOLO family for individual plant detection. Several sugar beet datasets (Sugar Beets 2016, Lincoln beet, CropAndWeed, and a newly collected dataset) were used for this analysis. A series of experiments was conducted to examine the effects of model size, random training seeds, dataset size and composition, image quality, and varying agricultural conditions on detection accuracy. The objective of this work is to identify the factors with the greatest impact on model performance in order to enable targeted dataset design and reduce annotation effort. The results show that missing or imprecise annotations have a significantly smaller impact on detection performance than noisy or blurred images. Furthermore, the findings indicate that using approximately 30–40% of the total 26,449 images is sufficient to achieve an optimal balance between detection accuracy and annotation effort.
In this paper, a quantization method for a FPGA platform is applied on three different deep neural networks (DNNs) for classification, detection and semantic segmentation tasks. The FPGA platform is based on a Xilinx Zynq UltraScale+ ZU5EV. The quantization process is done with the Vitis AI Quantizer tool of the AMD Vitis AI Environment. After a short description of the used DNNs, datasets and the technique behind the quantization process, the results are presented for different bit-widths of the quantized DNNs. The evaluation includes a post-training quantization of weights and activations in INT8 format with performance measurements in frames per second (FPS) on the target platform.
A coupled nonlinear simulation framework for a drone carrying a payload suspended via rope is presented. The tethered configuration introduces pendulum-like dynamics coupled to the drone’s motion. The framework is intended to support the development of strategies for safe trajectory-following with collision avoidance.
The payload is modeled as a rigid pendulum with a holonomic length constraint enforced via Lagrange multipliers and stabilized using the Baumgarte method, ensuring numerical consistency while preserving physically correct rope forces. The quadrotor model is calibrated up to the attitude-control level, enabling direct specification of thrust magnitude and direction consistent with state-of-the-art autopilot interfaces, while retaining actuator limitations and operating close to the physical origin of the pendulum dynamics.
Simulations demonstrate that the model reproduces physically plausible system behavior, validating its suitability as a basis for control design. System identification experiments confirm frequency-dependent coupling between drone and payload, including tension loss phenomena and nonlinear interaction effects.
The presented framework establishes a foundation for future research into trajectory tracking in the presence of pendulum dynamics, fast and safe payload placement during swing motion, and integrated collision avoidance under simultaneous consideration of pendulum motion, drone trajectory, and actuator constraints.
The utilization of uncrewed aerial vehicles (UAVs) in search and rescue (SAR) operations has become increasingly prevalent because the deployment of UAVs is expected to facilitate a higher degree of operational flexibility while simultaneously reducing costs. Currently, commercially available UAVs can be equipped with low-resolution thermal infrared (IR) cameras with typical resolutions of 640 × 512 pixels, which generally are evaluated manually by the SAR teams during an operation. Automatic person detection in IR images still remains a challenge. The objective of the proposed AIResQ dataset is to significantly enhance the performance of object detectors in the IR domain, employed in SAR operations for missing and potentially injured persons. AIResQ comprises 9,788 IR images with a resolution of up to 2048 × 1536 pixels captured from drone perspectives with a handheld camera under
varying weather conditions and in different terrains. Additionally, AIResQ displays persons in atypical poses. In order to test new object detectors in the context of SAR, we established a benchmark dataset stemming from exercises with real drone flights together with SAR organizations.
Swivel bending is an established forming process that minimizes material loss and enables efficient use of resources. However, the process requires complex optimizations that have traditionally relied heavily on the expertise of machine operators. This results in significant time and material costs, as optimization steps are performed iteratively. Given the shortage of skilled workers, a technological upgrade of the machines in line with Industry 4.0 is necessary. As part of a research project, intelligent sensor technology was used to record critical influencing factors that reveal correlations between product defects and machine deformations. Based on this, a methodology was developed that forms the foundation for inline compensation, enabling the equipment to autonomously adjust process parameters to correct product defects and, in the long term, enable defect-free production from the very first component.
Thread forming requires precise lubricant application because high contact pressures and process temperatures strongly influence tool loading, friction, and process stability. Although minimum quantity lubrication (MQL) systems are widely used, current spray-based approaches can still suffer from spray losses, insufficient wetting of the thread grooves, and unstable droplet transport. This article presents a concept for adaptive precision lubrication in thread forming based on computational fluid dynamics (CFD)-supported flow analysis, experimental validation, and artificial intelligence (AI)-assisted optimization. The focus is on droplet size, spray jet geometry, nozzle position, ambient flow conditions, and their influence on wetting intensity. Preliminary simulation-based investigations indicate that data-driven optimization can help identify wetting deficiencies and support the development of future control strategies for resource-efficient lubricant application.
Schwenkbiegen ist ein etabliertes Umformverfahren, bei dem Materialverlust vermieden und Ressourcen effizient genutzt werden. Der Prozess erfordert jedoch aufwändige Optimierungen, die bisher stark vom Fachwissen der Bediener abhängen. Dies führt zu hohem Zeit- und Materialaufwand, da Optimierungsschritte iterativ erfolgen. Angesichts des Fachkräftemangels ist eine technologische Aufrüstung der Anlagen im Sinne von Industrie 4.0 notwendig. Im Rahmen eines Projekts wurden mittels intelligenter Sensorik kritische Einflussfaktoren erfasst, die Korrelationen zwischen Produktfehlern und Anlagenverformungen aufzeigen. Darauf basierend wurde eine Methodik entwickelt, die die Grundlage für eine Inline-Kompensation schafft, bei der die Anlage eigenständig Prozessparameter anpasst, um Produktfehler zu korrigieren und perspektivisch eine fehlerfreie Fertigung ab dem ersten Bauteil zu ermöglichen.
Das Gewindeformen erfordert eine präzise Schmierstoffapplikation, da hohe Flächenpressungen und lokale Temperaturspitzen die Werkzeugbelastung erheblich beeinflussen. Aktuelle Sprüh- und Minimalmengenschmierungssysteme (MMS) weisen trotz etablierter Technik häufig Streuverluste, unzureichende Benetzung und instabile Tropfendynamik auf. Diese wissenschaftliche Betrachtung beinhaltet und untersucht einen integrativen Ansatz zur adaptiven Präzisionsbeölung beim Gewindeformen, der auf Computational Fluid Dynamics (CFD)-basierter Strömungsanalyse, experimenteller Validierung und Künstliche Intelligenz (KI)-gestützten Optimierungsverfahren basiert. Im Fokus stehen Tropfengröße, Strahlgeometrie, Düsenposition und Umgebungsströmung sowie deren Einfluss auf die Benetzungsintensität. Erste simulationsgestützteVoruntersuchungen zeigen das Potenzial einer datenbasierten Optimierung zur Reduktion von Benetzungsdefiziten und zur Auslegung künftiger Regelstrategien für eine ressourceneffiziente Schmierstoffapplikation.
