Open Access

Grading student answers and providing feedback are essential yet time-consuming tasks for educators. Recent advancements in Large Language Models (LLMs), including ChatGPT, Llama, and Mistral, have paved the way for automated support in this domain. This paper investigates the efficacy of instruction-following LLMs in adhering to predefined rubrics for evaluating student answers and delivering meaningful feedback. Leveraging the Mohler dataset and a custom German dataset, we evaluate various models, from commercial ones like ChatGPT to smaller open-source options like Llama, Mistral, and Command R. Additionally, we explore the impact of temperature parameters and techniques such as few-shot prompting. Surprisingly, while few-shot prompting enhances grading accuracy closer to ground truth, it introduces model inconsistency. Furthermore, some models exhibit non-deterministic behavior even at near-zero temperature settings. Our findings highlight the importance of rubrics in enhancing the interpretability of model outputs and fostering consistency in grading practices.

Von der ersten Hausarbeit bis zum Examen: Wissenschaftliches Arbeiten ist eine Kernkompetenz in jedem Studium. Zum Erlernen der wichtigsten Methoden und Regeln des wissenschaftlichen Arbeitens geben Ihnen Martin Wördenweber und Paul R. Melcher einen prägnanten Leitfaden mit vielen Praxisbeispielen an die Hand. (Verlagsangaben)

A novel multidimensional index modulation-based differential chaos shift keying (DCSK) technique, designated as Joint Subcarrier Time Reference Index Modulation-aided Differential Chaos Shift Keying (JSTRIM-DCSK), is proposed for efficient data transmission in chaotic communication systems. The JSTRIM-DCSK system integrates subcarrier, time slot, and reference signal indexing to transmit information and offers two variants: JSTRIM-DCSK-I and JSTRIM-DCSK-II. The data is organized into L subblocks, each containing Ns subcarrier index bits ps and Nt time slot index bits pt , reference index bits pr , and modulated bits pm . The subcarrier and time slot index bits jointly select an active or inactive subcarrier time slot combination from a total of Ns ⋅ Nt possibilities, categorizing the system as either JSTRIM-DCSK-I (active) or JSTRIM-DCSK-II (inactive). The reference indexed bits select a single chaotic reference signal from Nr orthogonal chaotic vectors generated using the Gram-Schmidt orthogonalization process. The modulated bits are transmitted using a DCSK modulation scheme. Analytical expressions for the bit error rate (BER) performance of the JSTRIM-DCSK system are derived under both additive white Gaussian noise (AWGN) and multipath Rayleigh fading channel (MRFC) conditions. Furthermore, the potential for energy savings, bandwidth efficiency, and system complexity of the JSTRIM-DCSK system are thoroughly analyzed and compared with those of the existing techniques. The simulation results validate the analytical expressions and demonstrate the potential of JSTRIM-DCSK to achieve high data rates, efficient energy savings, and a competitive BER performance.

Achieving perfect flatness, tension-free surfaces, and exceptional resistance to spring back are important characteristics of packaging steel, setting the standard for high-quality material performance. To guarantee these crucial parameters, the implementation of the stretch-bending process at the final stage of the production route is indispensable. Besides improving the flatness properties, the induced plastic deformation results in an accompanying change in the mechanical properties. This investigation focuses on understanding this change in mechanical properties due to different stretch-bending straightening process parameters. A multivariate predictive model is created to calculate a process window for achieving the desired flatness and also mechanical properties in the production of packaging steel. This model is validated by experiments with a laboratory facility.

This study examines the impact of integrating solar photovoltaic (PV) systems on power factor (PF) within low-voltage radial distribution networks, using empirical data from the Energy Self-Sufficiency for Health Facilities in Ghana (EnerSHelF) project sites in Ghana. The research included simulations focusing on optimal PV integration, with and without PF considerations, and the strategic placement of PV and shunt capacitors (SC). Three scenarios evaluated PV injection at high-load demand nodes, achieving penetration levels of 85.00 percent, 82.88 percent with high voltage drop, and 100.00 percent with high loss nodes. Additionally, three scenarios assessed SC allocation methods: proportional to the node's reactive power demand (Scenario I), even distribution (Scenario II), and proportional to installed PV capacity at PV nodes (Scenario III). The analysis used a twin-objective index (TOI), combining voltage deviations and power factor degradation. Results showed significant PV curtailment was necessary to achieve standard PF. Optimal penetration levels, considering TOI, reduced PV penetration from 85.00 percent to 63.75 percent, 82.88 percent to 57.38 percent, and 100.00 percent to 72.50 percent for high load, high voltage drops, and high loss nodes, respectively. Notably, all scenarios showed a concerning PF of 0.00 at dead-end nodes (P20, P21, P22). Scenario I achieved PF ranges of -0.26 to 1.00 with PV at high load, -0.69 to 1.00 with PV at high voltage drop, and 0.95 to 1.00 with PV at high loss nodes. Scenario II produced similar ranges, -0.48 to 1.00, -1.00 to 0.99, and 0.30 to 0.96, with PV placement at high load, voltage drops, and loss nodes, respectively. Scenario III yielded ranges of -0.19 to 0.97 (high load), -0.23 to 1.00 (high voltage drop), and 0.86 to 0.96 (high losses). The study concluded that the most effective strategy involves installing PVs at high-loss nodes and distributing SCs proportionally to the node's reactive power demand (Scenario I). This approach achieved a more uniform PF pattern throughout the network, highlighting the practical implications of strategic PV placement and targeted reactive power compensation for maintaining a healthy and efficient distribution system with solar PV integration.

The recent transformation of the energy sector brings new challenges in areas such as supply security, efficiency, and reliability. Especially the increase of decentralized power plants leads to a more complex energy system and an increasing complexity. This requires expansion and digitization of the power grid as well as an initiative-taking operation of the grid operator. To investigate such complex systems and its phenomena, modern development methods such as real-time simulation or digital twins (DT) can be used. In this approach a digital replica of the real-world system, a grid section, is developed, which can represent or predict the behavior of the real distribution grid. For this, a model of the real-world system is derived and implemented in a co-simulation environment, in which it receives data via an analyzer or measurement system from the grid model. This paper focuses on the development of the digital twin of a testing grid and a grid analyzer for the measurement. With the digital twin of the testing grid, a first approach is achieved in a real-time capable environment showing the functionalities and interactions of a digital twin. Subsequently the development of the digital twin model is explained, and the results are discussed.

The goal of this study was to explore a route for introducing functionalities into agarose-based hydrogels to tune the physical, chemical, and biological properties. Several agarose derivatives were prepared by homogeneous synthesis, including anionic agarose sulfates (ASs), reactive azido agaroses (AZAs), and cationic agarose carbamates (ACs), as well as agarose tosylates (ATOSs) and agarose phenyl carbonates (APhCs). The products were characterized in terms of their molecular structure and solubility behavior. The results suggest that the native gel-forming ability of agarose is retained if the introduced functionalities are hydrophilic, and the overall degree of substitution is low (DS < 0.5). Thus, functional hydrogels from several agarose derivatives could be obtained. The mechanical stability of the functional hydrogels was decreased compared to native agarose gels but was still in a range that enables safe handling. An increase in mechanical strength could be achieved by blending functional agarose derivatives and agarose into composite hydrogels. Finally, it was demonstrated that the novel functional agarose hydrogels are biocompatible and can potentially stimulate interactions with cells and tissue.

OBJECTIVES: The present study aimed to evaluate the trueness and precision of monolithic zirconia crowns (MZCs) fabricated by 3D printing and milling techniques. METHODS: A premolar crown was designed after scanning a prepared typodont. Twenty MZCs were fabricated using milling and 3D-printing techniques (n=10). All the specimens were scanned with an industrial scanner, and the scanned data were analyzed using 3D measurement software to evaluate the trueness and precision of each group. Root mean square (RMS) deviations were measured and statistically analyzed (One-way ANOVA, Tukey's, p≤0.05). RESULTS: The trueness of the printed MZC group (140 ± 14 μm) showed a significantly higher RMS value compared to the milled MZCs (96 ± 27 μm,p<0.001). At the same time, the precision of the milled MZCs (61±17 μm) showed a significantly higher RMS value compared to that of the printed MZCs (31±5 μm,p<0.001). CONCLUSIONS: The Fabrication techniques had a significant impact on the accuracy of the MZCs. Milled MZCs showed the highest trueness, while printed MZCs showed the highest precision. All the results were within the clinically acceptable error values. CLINICAL SIGNIFICANCE: Although the trueness of the milled MZCs is higher, the manufacturing accuracy of the 3D-printed MZCs showed clinically acceptable results in terms of trueness and precision. However, additional clinical studies are recommended. Furthermore, the volumetric changes of the material should be considered.

A building’s energy storage demand depends on a variety of factors related to the specific local conditions such as building type, self-sufficiency-rate, and grid connection. Here, a newly developed bottom-up procedure is presented for classifying buildings in an urban building portfolio according to specific criteria. The algorithm uses publicly available building data such as building use, ground floor area, roof ridge height, solar roof potential, and population statistics. In addition, it considers the local gas grid (GG) as well as the district heating (DH) network. The building classification is developed for identifying typical building situations that can be used to estimate the demand for residential energy storage capacity. The developed algorithm is used to identify potential implementation of private photovoltaic(PV)-metal-hydride-storage (MHS) systems, for three scenarios, into the urban infrastructure for the city of Cologne. As result the statistical confidence interval of all analyzed buildings regarding their classification as well as corresponding maps is shown. Since similar data sets as used are available for many German or European metropolitan areas, the method developed with the assumptions presented in this work, can be used for classification of other urban and semi-urban areas including the assessment of their grid infrastructure.

To respond to the increasing demand for hyaluronic acid (HA) in dietary supplements (DSs) and nutricosmetics marketed for the treatment of osteoarthritis or moistening, it is essential to have an accurate and reliable method for its analysis in the final products. The study aimed to develop and validate alternative method for the quality control of HA in DSs using low-field (LF) and high-field (HF) nuclear magnetic resonance (NMR) spectroscopy at 80 MHz and 600 MHz, respectively. Moreover, chondroitin sulphate (CH), another active ingredient in DSs, can be simultaneously quantified. The 1H-NMR methods have been successfully validated in terms of limit of detection (LOD) and limit of quantitation (LOQ), which were found to be 0.1 mg/mL and 0.2 mg/mL (80 MHz) as well as 0.2 mg/mL and 0.6 mg/mL (600 MHz). Recovery rates were estimated to be between 92 and 120% on both spectrometers; precision including sample preparation was found to be 4.2% and 8.0% for 600 MHz and 80 MHz, respectively. Quantitative results obtained by HF and LF NMR were comparable for 16 DSs with varying matrix. HF NMR experiments at 70 ℃ serve as a simple and efficient quality control tool for HA and CH in multicomponent DSs. Benchtop NMR measurements, upon preceding acid hydrolysis, offer a cost-effective and cryogen-free alternative for analyzing DSs in the absence of CH and paramagnetic matrix components.

In the coming years, the European Union plans to establish Proton Exchange Membrane (PEM) electrolyzers, each with a 100 MW capacity. However, the selection of their locations has not been systematically optimized to leverage potential benefits, such as utilizing waste heat from large facilities for district heating. Presently, there are hardly any corresponding system models in the literature dynamically simulating a PEM electrolyzer of this size. This paper introduces a first model approach for such systems, drawing on parameters from existing literature. It addresses the inconsistency found in the literature regarding the use of the exchange current density, which varies by a factor of . A novel optimization process is developed by using an auxiliary parameter to fit the exchange current density with a newfound condition between the anode and cathode side. The outcome is a comprehensive model of a PEM electrolyzer plant, exemplarily adapted to the Siemens Silyzer 300.