Repository der Technischen Hochschule Ingolstadt
Not a member yet
4170 research outputs found
Sort by
Optimizing AI-Driven Production in Industry 4.0: A Morphological Box and Taxonomy Approach
No full text
A Practical Approach to Multivariate Time Series Anomaly Detection in Automotive Bus Systems Testing
No full textThe increasing complexity of modern vehicles and their testing procedures generates vast amounts of multivariate time series data, making manual anomaly detection during automotive testing increasingly challenging. This article investigates the application of deep learning algorithms for automated anomaly detection in automotive bus data collected during dynamic driving scenarios. Three distinct architectures are implemented and compared: a CNN-based forecasting approach (DeepAnT), an LSTM-based model (LSTM-AD), and a Convolutional Autoencoder (CAE).
Real-world driving data collected across various scenarios, ranging from normal operation to extreme maneuvers, is employed. Through evaluation across seven distinct test scenarios, findings reveal that while each architecture demonstrates specific strengths, their effectiveness varies significantly based on anomaly type and driving context. DeepAnT shows the most consistent performance across different scenarios, while LSTM-AD achieves superior detection capability for complex temporal patterns, particularly in scenarios involving coordinated changes across multiple features. The CAE excels at identifying pronounced deviations but shows limitations in detecting subtle anomalies.
This study demonstrates that while deep learning models effectively detect anomalies in automotive time series data, their practical implementation requires careful consideration of specific use cases, emphasizing the critical role of data preprocessing and threshold calculation in ensuring reliable anomaly detection
Advancing Generative AI Collaboration in Design-to-Code Workflows: Insights from Two Empirical Studies
No full textFor expert users to accept Generative AI (GenAI) as a true collaborative partner, it must move beyond simple task-awareness to an understanding of their workflow’s underlying structural rules. This paper introduces a paradigm for AI collaborators that moves beyond simple task awareness to an understanding of the semantic and hierarchical relationships within a component-based system. We investigate this concept within the context of the design-to-code workflow, where inefficiencies arise from the modification of components within design systems. Through two empirical studies with designers and developers, we found that GenAI output was often rejected because it violated the component hierarchy. Designers required granular and visual control for refinements, whereas developers valued automated setup but required transparent validation of the generated code’s logic. Based on these findings, we contribute design guidelines for achieving Component-Structure Awareness (CSA), with two core principles: the Atomic Recommender, which provides assistance that respects the component hierarchy, and Communication Archetypes, which allow GenAI to adapt its interaction style to the user’s role and the atomic nature of their task. This work provides a new, higher-level concept for designing the next generation of truly collaborative GenAI agents
Mini-grid performance in Sub-Saharan Africa: case studies from Tsumkwe and Gam, Namibia
No full textThis paper investigates the performance analysis and operational challenges of mini-grids in Sub-Saharan Africa, focusing on the Tsumkwe and Gam mini-grids, the only officially recognized mini-grids in Namibia. The study addresses a critical gap in understanding mini-grid efficiency, technical difficulties, and future potential. The key research questions focus on assessing mini-grid performance, identifying technical and operational challenges, quantifying the relationship between electricity demand and supply, and providing recommendations for enhancing rural electrification through mini-grids. These insights are crucial for evaluating the effectiveness of mini-grids in the African context and understanding their role in advancing rural electrification. By using a mixed-method approach, the research combines quantitative and qualitative data to offer a comprehensive analysis of these two mini-grid systems of Namibia. The data collected from 2017 to 2022 include metrics on energy supply, consumption, photovoltaic generation, community load, and meteorological conditions. Qualitative insights were gathered through field visits, surveys, and interviews with mini-grid operators, allowing for a thorough examination of community perspectives, operational issues, and technical performance. The findings reveal that the Tsumkwe mini-grid has seen a decline in efficiency due to maintenance problems and battery overheating, while the Gam mini-grid, initially oversized, now struggles with increased demand and new connections. This study provides the first detailed technical insights, identifies consumer archetypes, and evaluates the energy dynamics of mini-grids in Namibia. Recommendations include improved maintenance protocols, advanced battery management, and enhanced data monitoring. By analyzing Tsumkwe and Gam, this paper offers valuable lessons for mini-grid development in Sub-Saharan Africa, stressing the need for continuous evaluation and strategic improvements to achieve sustainable rural electrification
From a Social POV: The Impact of Point of View on Player Behavior, Engagement, and Experience in a Serious Social Simulation Game
No full textMultiplayer games with social aspects vary widely regarding client design, e.g., point of view or camera perspective. While design paradigms usually arise from gold standards that are set by previously successful games in the industry, the impact of such paradigms is under-researched for games that serve as scientific instruments, e.g., to research social behavior. Intending to investigate how such games should be designed, we built two multiplayer clients with the same game logic, one using a first-person point of view, while the other includes a top-down camera perspective. Then, we conducted an online user study in which players tested these game clients in extensive multiplayer sessions. Analyzing speech time, in-game logs, questionnaires, and qualitative feedback, we look at the perspectives’ impact on player behavior, engagement, and game experience in a scientific or "serious games" context. In addition, we have made our designed game UNISON and both clients available as open source to facilitate future empirical social science research
Predictive Maintenance of Ball Screws: A Comparative Study Using Real-World Industrial Data
No full textBall screws are widely used in machine tool feed drives. With increasing degradation of the ball screws, machining accuracy and economic efficiency decrease. Past investigations have shown that condition monitoring models can predict this degradation. However, these models are typically trained and evaluated using datasets derived from test benches, questioning their applicability to real machine tools. In this article, a comparative evaluation of a selection of these condition monitoring models using an industrial dataset is described. This dataset consists of measurement data from a total of nine ball screws used in three machine tools until failure. It was shown that when using data of multiple ball screws or machines, artificial neural networks or automated machine learning methods achieve a higher accuracy than statistical methods. However, for smaller datasets, statistical methods perform almost as well. The results provide an insight into the industrial applicability of the evaluated condition monitoring models
Andreas Schadauer: Wissen in Zahlen? Zur Herstellung quantitativen Wissens in der Sozialwissenschaft. Bielefeld: transcript 2022, 254 S., ISBN 978-3-8376-6398-3, 45,00 €
No full textMit der auf seiner Dissertation aufbauenden Monografie Wissen in Zahlen? Zur Herstellung quantitativen Wissens in der Sozialwissenschaft, zeigt Andreas Schadauer anhand von zwei empirischen Fallstudien auf, wie Umfragedaten zunächst zu Zahlen und Statistiken werden, denen im weiteren Verlauf der Rezeption ein nahezu faktischer Status der Objektivität zugeschriebenwird. Dafür zeichnet er in der ersten Fallstudie den Weg der Daten nach, die im Rahmen der Household Finance and Consumption Survey (HFCS) der Österreichischen Nationalbank (OeNB) zwischen 2010–2011 generiert wurden. Als zweite Fallstudie wählt er die österreichische Immobilienvermögenserhebung von 2008, die zum Zeitpunkt der Feldforschung bereits abgeschlossen war. Beide Datensätze wurden mit dem Anspruch an Repräsentativität generiert und stellen in der österreichischen Debatte um die nationale Vermögensverteilung wichtige Referenzen dar.
Mit den „multi-sited“ (S. 19, 59 ff., Hervorh. i. Orig.) Fallstudien verfolgt Schadauer zwei Ziele: Zum einen hinterfragt er ein in vielen Gesellschaftsteilen vorherrschendes normativ-positivistisches Wissenschaftsverständnis, welches er als elementar für die unkritische Rezeption von Statistiken als die Abbildung von Realität im Singular sieht. Zum anderen stellt er sich „gegen die Vorstellung, Wissenschaft werde von der Gesellschaft determiniert und Erfolg hänge dann davon ab, was gesellschaftlich vorgegeben und akzeptiert wird (vgl. z. B. Bloor, 1991)“ (S. 16). Um sich diesen Zielen anzunähern, geht er der Frage nach, wie Zahlen und Statistiken so wichtig werden, dass sie mediale wie politische Diskurse formen, gar für Gesellschaftsgruppen beziehungsweise eine ganze Nation sprechen können
SalaciaML-2-Arctic - a deep learning quality control algorithm for Arctic Ocean temperature and salinity data
No full textWe have extended a classical quality control (QC) algorithm by integrating a deep learning neural network, resulting in SalaciaML-2-Arctic , a tool for automated QC of Arctic Ocean temperature and salinity profile data. The neural network component was trained on the Unified Database for Arctic and Subarctic Hydrography (UDASH), which has been quality-controlled and labeled by expert oceanographers. SalaciaML-2-Arctic successfully reproduces human expertise by correcting misclassifications made by the classical algorithm, reducing False Negatives (samples incorrectly classified as “bad”) by 96% for temperature and 99% for salinity. When used in combination with a visual post-QC by human experts, it achieves a workload reduction of approximately 60% for temperature and 85% for salinity. All code and data required to reproduce the analysis or apply the method to other datasets are openly available via PANGAEA and GitHub. Moreover, SalaciaML-2-Arctic is accessible as a browser-based application at https://mvre.autoqc.cloud.awi.de, enabling its use without software installation or programming knowledge